This is a reproduction of a library book that was digitized

by Google as part of an ongoing effort to preserve the

information in books and make it universally accessible.

https://books.google.com
 LIBRARY OF THE
UNIVERSITY OF ILLINOIS
AT URBANA-CHANMPAIGN

629. |3
L! ºn 325th
no. 65 71 - 6590

ENGINEERING
 The person charging this material is re
sponsible for its return on or before the
Latest Date stamped below.
Theft, mutilation and underlining of books
are reasons for disciplinary action and may
result in dismissal from the University.

| rº-

pºro REPRChuction
JUN 16 RECD

L161—O-1096
|CA
 º 2_z ENGIN

NASA T E C H N. I. C. A. L NOT E NASA TN D-657]

THE LIBRARY OF THE

JAN 3 1972
UNIVERSITY OF ILLINQS
Aſ URBANA-CHAMPAIGN

ſº OF MONTE CARLO TFCHNIQUES

AP

º OF HIGH-ENERGY BEAM
SPORT IN A STOCHASTIC ENVIRONMENT

º V. Parrish, James E. Dieudonme,

* Tassos A. Filippas
E. *earch Center
4
mpton, Va. 23365

|MI AFRON
AUTICS AND SPACE ADMINISTRATION . WASHINGTON, D. C. . DECEMBER 1971
*
*
 ſº /5 2
A -

º .. eT ENGINEERING LIBRARY ~~~~ 2" '* : *-

ºf v > - 6– }!
**

945 Z/-43 ſo
APPLICATION OF MONTE CARLOTECHNIQUES TO OPTIMIZATION
OF HIGH-ENERGY BEAM TRANSPORT IN

A STOCHASTIC ENVIRONMENT

By Russell V. Parrish, James E. Dieudonne,

and Tassos A. Filippas
Langley Research Center

SUMMARY

An algorithm employing a modified sequential random perturbation, or creeping

| Hºm search, has been applied to the problem of optimizing the parameters of a high
eng beam transport system. The stochastic solution of the mathematical model for
first-Order magnetic-field expansion allows the inclusion of state-variable constraints,
ºthe inclusion of parameter constraints allowed by the method of algorithm application
ºuts the possibility of infeasible solutions. The mathematical model and the algo
in were programed for a real-time simulation facility; thus, two important features
ºf provided to the beam designer: (1) a strong degree of man-machine communication
at to the extent of bypassing the algorithm and applying analog-matching techniques),
i:') extensive graphics for displaying information concerning both algorithm operation
ºf transport-system behavior. Chromatic aberration was also included in the mathe
ºtical model and in the optimization process.
Results presented show this method as yielding better solutions (in terms of resolu
is to the particular problem than those of a standard analog program and well as dem
*rating flexibility, in terms of elements, constraints, and chromatic aberration, allowed
H ser interaction with both the algorithm and the stochastic model.Also presented are
simples of slit usage and a limited comparison of predicted results
and actual results
*ained with the 600 MeV cyclotron at the NASA Space Radiation Effects Laboratory.

INTRODUCTION

With the advent of large-accelerator projects over the past 15 years, a demand has
º for optimization of not only the design of the accelerator but also the design of the
ºns Systems that must necessarily be coupled with the accelerator. As pointed
“in reference 1, the high cost of producing the strong magnetic fields over large vol
TeS **ssary for high-energy beam optics makes optimization a necessity.
 º:
Past efforts in the area of beam-transport optimization have generally fallen into
two classes of techniques: "analog matching" with a high-Speed repetitive analog com- --.
puter and iterative deterministic methods for parameter optimization implemented on a º
general-purpose digital computer. References 1 to 3 describe the general methods used º:
in the technique of analog matching: Generally in designing transport Systems the resultº
of parameter variations in the system are continuously displayed on an oscilloscope as *
the designer seeks by trial and error to determine the focusing strengths and separation
distances necessary to give the desired results. Aside from the problems of analog res
olution and repeatability, which are usually present even with advanced equipment, the
major disadvantage with this technique is that it relies heavily on the experience of the sº
analyst, especially as the number of parameters increases. It is extremely time consun
ing and is usually based on visual determination of improvement; thus, the problem of sº
visual resolution arises.

However, as references 1 to 3 strongly emphasize, the man-machine communicatiº

available with the analog computer is often felt to be the overriding consideration when
choosing between the two classes of techniques available to solve the problem of beam- º
transport design. This man-machine communication is generally sacrificed when the
analyst chooses to use a digital-computer optimization program. This sacrifice is not
unwarranted, however, according to references 4 and 5, which emphasize the advantages
optimization algorithms over analog methods. These advantages include a performance º
measure for evaluating a given set of parameters, an algorithm that automatically seeks
an extremum solution of the performance measure (which may or may not be feasible), *:
and digital resolution.

Thus a survey of the literature available in the field reveals strong advocates for >
each class of techniques. The present paper will describe a method that evolved in the
process of solving a particular problem in beam-transport design. This method combi º
some of the best features of both classes of techniques, that is, a digital algorithm with &
ensuing resolution, and man-machine communication (complete communication even to *

extent of bypassing the algorithm and applying analog-matching techniques). In additio. s

the method allows for the inclusion of state-variable constraints and parameter con–
Straints in the mathematical model. Other features of the program include stochastic º
representation of initial conditions and chromatic aberration and the introduction of the
aberration during the optimization process. These features make the simulation of the ``
beam transport system more realistic than in most previous work (e.g., ref. 1). *-
- .

The method will be presented in the context of solving a specific problem in bear
transport design, namely, that of transporting a beam of charged particles from one gi.
area to another through the use of four quadrupole magnets (other configurations of ele
ments are allowable). In this problem five parameters are available for the optimization
process. The maximum number of parameters that can be varied at one time is deter
mined by the choice of the optimization algorithm and the amount of computing time nec
essary to evaluate the performance measure. However, the method may be used to opti
mize Several elements at a time along the transport system, with the resulting beam used
as input to the next set of elements in large systems.

SYMBOLS

aibi constraint constants, meter-1

value of cost function with present parameter values, dimensionless

value of cost function at last success, dimensionless

cost function corresponding to horizontal axis, dimensionless

cost function corresponding to vertical axis, dimensionless

separation distance between second and third quadrupoles, meters

electron charge, coulombs

calculated test statistic, dimensionless

F value for 95-percent confidence level with (a,n) degrees of freedom,

dimensionless

ith magnetic-field gradient, webers/meter?

alternate hypothesis for a statistical test, dimensionless

H] null hypothesis for a statistical test, dimensionless

K. magnetic focusing strength of ith quadrupole, meter-1

h effective field width, meters

number of rays which passed through the system, dimensionless

number of rays transmitted with present parameter values,
dimensionless

number of rays transmitted at last success, dimensionless

II: '...It
percentage of rays which did not pass through the system,
In - In 1
*** ºr tº:
× 100, dimensionless -
'º' º:

horizontal focal point, meters

vertical focal point, meters

minimum value allowed for distance between horizontal and vertical focal

points of the first doublet, meters

minimum value allowed for horizontal focal point of the first doublet, meters

maximum value allowed for horizontal focal point of the first doublet, meters

minimum value allowed for vertical focal point of the first doublet, meters

maximum value allowed for vertical focal point of the first doublet, meters

average momentum, kilogram-meters/second

momentum perturbation, kilogram-meters/second

thickness of slit, meters

displacement along focal axis of transport system, meters

Sout - Sn, meters

position along focal axis of extremum in x, meters

position along focal axis of extremum in Z, meters

position along focal axis at which cost function is evaluated, meters

 *B beginning position of horizontal slit, meters

iE ending position of horizontal slit, meters

position of entrance to field (magnetic or drift) along focal axis, meters

position of exit from field (magnetic or drift) along focal axis, meters

beginning position of vertical slit, meters

ending position of vertical slit, meters

calculated test statistic, dimensionless

% t value associated with type I error, dimensionless

§
ill t value associated with type II error, dimensionless
'S
| º 2. \
- Student's "t" for 95-percent confidence level with (a + np - 2. degrees
of freedom, dimensionless

W weight corresponding to difference between cost function for vertical plane

º and cost function for horizontal plane, dimensionless
|
}
W weight corresponding to percentage of beam which passed through the system,
dimensionless

displacement along horizontal axis of transport system, meters

desired value of horizontal displacement at end point, meters

mean value of the horizontal slope at the end point of all rays of the beam
Which passed through the system, dimensionless

displacement of beam in horizontal plane parallel to focal axis, used to deter

mine slit opening, meters

ls)
displacement along vertical axis of transport system, meters
 desired value of vertical displacement at end point, meters

mean value of the vertical slope at the end point of all rays of the beam
which passed through the system, dimensionless

aperture radius, meters

detectable difference desired, dimensionless

variance of cost function for present parameter values, dimensionless

estimate of cost-function variance at last success, dimensionless

variance of cost function at last success, dimensionless

º:
: estimate of cost-function variance for present parameter values,
dimensionless

slit width along x-axis, meters sº

º,
-

(Dz slit width along z-axis, meters

Subscript:

desired value

Primes indicate derivatives with respect to the focal axis (s-axis).

MATHEMATICAL MODEL
s º,

The basic beam-transport problem discussed in this paper is the design of a high
energy beam transport system consisting of four quadrupoles, with variable magnetic
focusing strengths and separation distances, and slits with variable width, thickness, and
Separation distance. This system will transform a charged-particle beam of known initial
conditions into a beam meeting some desired phase-plane configuration defined by a math 2
ematical formula known as the cost function. In the design of this particular system,
illustrated in figure 1, five system parameters were to be determined. These were the *S*

four magnetic focusing strengths (K1, K2, K3, and K4)and the separation distance D2
Kºšitle doublets (pairs of quadrupoles). All other system parameters are considered
iºd. In general, the method used to determine the parameters imposes no restrictions
st number or types of the parameters.
For applications of interest here, the differential equations of motion for traversing
iſſuadrupole field (ref. 1, p. 9)

X" = -K,(1 + z'2 + .* + x2)x xz.

-

(1)

Z" - Kı(1 + z'2 + x2)

1/2
/ ſ + z:2)2 - wº
can be reduced to the uncoupled, linear equations (refs. 1 and 6)

x" + Kix = 0
(2)
Z" – Kiz = 0

where Ki, a constant (i = 1,2,3,4), is the focusing strength of the field. It represents a
rectangular, first-order, field model Ki = egi/po (where the ith magnetic potential is
equal to -gi . xz) in which focusing occurs in the xS-plane and defocusing in the Zs-plane.
The algebraic solutions of equations (2) are shown in the following equations:

x(out) xen - COS (K. As) + **(Ea) (3a)

|Ki

*(sout) -x(e) K.
- Sin (K. As) x(ºcos (K. As
+ (3.b)

*(sout) zºosh(K. As)

- + Z" sm) *(E As (3.c)

|Ki

*(sout) z(sn)K.
- Sinh (K. As) z'(ºcosh(K. As)
+ (3d)

As = Sout - Sn
The inverse case, namely, focusing in the Zs-plane and defocusing in the xs-plane, may
be implemented by using equations (2) as shown, but with the sign of Ki changed. i.
The drift-space translational equations

X" = 0
(4)
Z" = 0

also have algebraic Solutions:

x(soul) *(*) x(e)

- + AS (5a)

º
x(sout) x(ºn)
- (5b)

*(sout) zºn) z'(s)

- + AS (5c)

z'(sout) z'(s)
- (5d) :-

Throughout the entire translation along the s-axis, the position variables are sub- º
ject to the physical constraints imposed by the aperture size of each successive
quadrupole: -:

x2(s) + zº(s) - oft (6) º **

where of are the aperture radii for the four quadrupoles (i = 1,2,3,4). This constraint º
was imposed at the entrance and exit of each quadrupole as well as within the quadrupole. s
The locations of the extremum values of x and Z (sex and Se 2
2) within the focusing º
º
quadrupole are determined by

e,x \K. WKX(sn)

(7) is
s... ===tanhº"
* K.
#.
\Kız Sn) *
>
whºſe

= | s” = s” <
*X *ex (o - *e,x - As Se,z = °e,z
xk
(0 <
- °e,z
s
- As
AS (ºx As
x
> AS (*. As
×
>

Again, the equations presented are for focusing in the xs-plane, and a change in sign of
K is necessary for focusing in the Zs-plane.
In addition to the position constraints of equation (6), two types of parameter con
Straints may be present in the problem. The first type

-bi = Ki = bi (i = 1,2,3,4)
(8)

a1 = D2 = a2

is imposed by the physical properties of the quadrupoles to be used in the actual experi
ment, limiting the values of Ki to feasible ranges and the separation distance between
the doublets to values meeting the requirements of the experimental laboratory
timensions.

The second type of parameter constraint is introduced by the desire to use slits to
improve the beam quality. Slits are usually to be introduced at some focal point in either
texs- and Zs-planes. It is desirable that the focal points fall within the drift space
between the elements and still allow enough room for slit placement. It is also desirable
tº the focal points be separated from one another by sufficient distance for placement
ºf slits of different thickness at each focal point. These constraints are represented by

"ºun **ix **** (9a)

"ºmin s Piz s ** (9b)

P. 3 - Pfiz|= (AP), in (9c)

9
and the focal points are determined by solving equations (3) with initial conditions
x' = z' = 0, and x = z = Constant through the desired elements. The focal points are
then determined by the points of intersection with the S-axis.
Slits are implemented as position constraints:

9x
|x(s) <-à (the < r
> S >
<
*h,E)
-

(10)
|z(s)| < *z s ,, = S = S
2 ( v,B v.E) º

assuming no penetration (complete absorption), and can be introduced separately, jointly,

or not at all. With the above constraints imposed on the system, the problem is to find
a set of parameters which when placed in the system will yield a "best fit" to some pre
determined phase-space curve defined in the cost function. The algorithm chosen to per
form this optimization is explained in the section of this paper entitled "Monte Carlo
Method."

One primary aberration, chromatic aberration, was taken into account in the mathe
matical model. This aberration arises from the different values of momentum possessed
by the individual rays. In the equations presented previously, the momenta of the parti
cles are considered constant and equal throughout the transport system, that is,
Ki = egi/po. Usually an optimal case is obtained without the presence of chromatic aber
ration, and then the effect of the aberration is examined in the following manner: The
eg;
varying momentum is modeled as Ki = H. where
DA + Ap
Ap is a positive or negative
O

momentum perturbation. In the present paper, each ray of the beam is assigned a momen
tum perturbation based on Gaussian noise. The perturbation is then applied to each focus.
ing strength. In this manner, the momentum of an individual particle of the beam is con
stant throughout the focusing system, while individual momenta within the beam vary.
Optimization then takes place in the presence of chromatic aberration.

ANALOG SIMULATION

During the preliminary investigation of the problem, the equations of motion

(eqs. (2)) and the constraint equations (eqs. (8)), were programed on both the GPS 10000
repetitive analog computer and the EAI 680 analog-hybrid computer. The function Ki,
which represents the system parameters, was generated on both machines by the use of

10
 electronic comparators and available logic components. The variation of system param
eters was achieved by tandem potentiometers. Repetitive solutions of the differential
equations at a rate of 10 runs per second allowed oscilloscope displays of the trajectories
Z(s) and x(s) (beam traces) and phase-plane plots in both planes at any point in the
System.

A visual-resolution problem was encountered in the use of this method in that the
Operator found many promising solutions, or sets of system parameters, from which by
Visual means a "best case" could not be determined. To solve this problem, a digital
program was written for the solution of the differential equations with the predetermined
Sets of system parameters. The digital program, which also served as an independent
check, solved the resolution problem, but uncovered the fact that the analog-determined
System parameters did not fulfill the desired phase-plane end conditions. This fact will
be demonstrated in the "Results" section of this paper.

REAL-TIME SIMULATION SYSTEM

The inability of the analog program in meeting required phase-plane end conditions
led to the programing of the problem on the CDC 6600 Real Time Simulation (RTS)
Subsystem at the Langley Research Center. This system, which is described in refer
ence 7, provides both sufficient accuracy for the problem and man-machine communica
tion. In addition to the main computers, the RTS system contains program control sta
tions, an analog-to-digital and discrete input system, a digital-to-analog and discrete
Output system, a CDC 250 Series CRT display system, and peripheral units (time-history
recorders, x-y plotters, card readers, etc). The RTS software provides many "analog
like" features such as mode controls, subroutines for displaying and changing parameters,
and function-sense switches.

ſ: The application of the RTS system to the beam-transport problem allows the use of
an Optimization algorithm to determine an optimal set of system parameters while the
Operator maintains man-machine communication.

This communication link to the machine allows quick, effective use of the designer's
knowledge of the system. It enables him to examine different areas of the parameter
Space and determine the merits of particular parameter sets as to likely starting points
for the optimization algorithm, thereby reducing the time required for convergence. The
designer also has many options available through subroutines, function-sense switches,
and the RTS software from which he may change the model of the transport system,
position-variable and parameter constraints, the form of the desired output of the system,
and the form of the probability distribution that dictates the initial conditions of the beam
prior to magnetic-field entrance.

11
 Monte Carlo Method

The individual rays of the incident beam in the model are chosen to represent the |

physical beam for which the transport system is being designed. This is done by choos
ing a set of random initial conditions that satisfies the physical requirements. Equa
tions (3) are then evaluated to provide the values of the position variables at the entrance
and exit of each magnetic field, and also the extremum values within the magnetic fields
are evaluated (eqs. (7)). Any violations of the position-variable constraints are recorded
(solution of the equations of motion for an individual ray is terminated upon the occur
rence of a violation) and may be included in a weighted performance measure with the
deviation of the transmitted particles from the desired end conditions. This process is
then repeated and the deviations are accumulated to form the mean performance measure,
or mean cost function, over the desired sample size. The parameters of the system (four
focusing strengths and one separation distance) are then perturbed by the optimization
algorithm and the entire sequence is iterated until convergence is achieved.

Monte Carlo Algorithm

The optimization algorithm used in conjunction with the mathematical model was a
variation of the sequential random perturbation, or creeping random search, suggested by
Mitchell in reference 8. This algorithm was selected over deterministic methods that
are gradient dependent, such as steepest descent, conjugate gradient, Newton–Raphson,
and quasilinearization (see ref. 9), because of the presence of constraints and a stochastic
model. The algorithm is described in detail in reference 10.

A stochastic representation of the initial beam (beam prior to entry into the trans
port system) was advantageous in making the simulation of the system more realistic.
By considering the beam as a conglomerate of rays, each having individual characteristics
of displacement, slope, and momentum, position-variable constraints may be implemented
and the percentage of the beam striking the elements of the transport system may be cal
culated. The stochastic model also allows for realistic representation of chromatic aber
ration since each ray of the beam instead of the beam as a whole is assigned a momentum
perturbation.

The basic algorithm, devoid of any strategies, is best illustrated with the two
parameter contour plot shown in figure 2(a). In this simple problem, constant values of
the cost function are represented as concentric circles in the two-parameter plane.
From an arbitrary starting point, a random step is taken and the cost function is evalu –
ated. If the step has improved the cost function (a success), the present parameter val—
ues are used as a new starting point and a new step is taken. If the cost function was not
improved (a failure), the original parameter values are retained, and a new step is taken.
Thus a "walk" over the contour is generated.

12
 For explanation of the basic algorithm in more detail, attention is directed to fig
ure 2(b). As illustrated in the figure, there are nine possible directions that can be taken
from a given starting point in the two-parameter plane. These nine directions are
obtained by plus, minus, and zero perturbations of each parameter. (Note that one direc
tion is no movement at all.) The eight directions that involve movement are stored in an
array, which the algorithm will sample randomly without replacement. The array is
sampled until a success is obtained or all directions are exhausted, at which point the
array is reinitialized. If failures should occur in all possible directions from a given
point, a reduction in step size would be necessary in order to continue iteration. The
algorithm is stopped once the step size becomes smaller than the desired detectable dif
ference (see eq. (13)) between different values of the cost function.
In conjunction with the basic algorithm, two features are available to speed conver
gence. The first feature is a strategy implemented at the option of the designer by the
algorithm and simply requires that after each success the next try be made in the same
direction. The second feature involves interaction of the beam designer with the algo
rithm through the CRT displays and will be described later.

Parameter Constraints

The parameter constraints of equations (8) are implemented by simply not allowing
the algorithm to perturb parameters across boundaries. Thus these constraints affect
the problem only in that the parameters must always be within their feasible ranges. The
focal-point constraints, equations (9), are implemented by perturbing parameters, calcu
lating the focal points, and sensing for violations. Should a violation occur, the perturba
tion direction is deemed a failure, the equations of motion are not evaluated, and a new
direction is chosen by the algorithm. It should be obvious that the addition of the focal
point constraints greatly speeds convergence, since the evaluation of the equations of
motion is bypassed for all parameter sets which do not satisfy the constraints.

Statistical Test for Success

Assuming equal variances, a pooled tº test for equality of means, as described in

reference 11, is used to determine the success or failure of a perturbation direction. The
null hypothesis was chosen to assure control over the probability of accepting a failure as
a success (a 5-percent chance of occurrence was selected). The test is shown as follows:
Ho: Cp = C (Failure) HA: Cp > C (Success)
Cn – C
teal = 9 (11)
1 . 1 (ºp 1); + (n −
- 1)62
# #) np + n - 2

Reject Ho if tcal 2 t05(118) = 1.66

13
Where np + n - 2 = 118 degrees of freedom. Here, Cp is the value of the mean cost
at the last success; np is the sample size (60 rays/beam) at the last success; # iS

the variance at the last success; and C, n, and 6% are the mean cost, Sample size,
and variance, respectively, for the present parameter values being tested. It should be
noted that an F test was used periodically to validate the assumption of equality of
variances, as shown in the following equations:

H0: g; = gº HA: g; 4 gº
A = mºſº
Where (12)
cal # B = minº
Reject Ho if Feal - F.95(60,60) = 1.67

The null hypothesis was accepted in all cases, except for a few in which the step size of
the perturbations was large.

The sample size (60 rays/beam) was determined by using the iterative method
described in reference 12, pages 154–155,

n >
=
2(to 62tºº?
+
(13)

where

to t value associated with type I error; to - t.05(118) = 1.66

t1 t value associated with type II error (95-percent power for the chosen detectable
difference); tı = t 10(118) = 1.29

62 estimate of variance; 62 = 0.01

Ö detectable difference desired; 5 = 0.055

GRAPHIC DISPLAYS AND INTERACTIVE CAPABILITIES

Extensive use of the CRT displays is an integral part of program operation. In

addition to the algorithm displays, the following problem-related displays are available
to the analyst:
14
 (1) Phase plots for both the x and z variables at the entrance to the transport
system and at the exit of each quadrupole

(2) An xz-plane plot at the same locations

(3) Beam traces, both horizontal and vertical views, including the location and size
of each quadrupole, slit placement in the system, and the horizontal and verti
cal focal points

Thus the analyst has complete information of the transport system at important points
along the system. If the designer desires, based on his knowledge of transport systems,
algorithm operation may be bypassed, and with use of the displays, analog matching may
be carried out. This "hands-on" capability allows the designer to look at areas in the
parameter contour which he may deem interesting.

The algorithm displays are available to possibly speed convergence to an extremum

of the problem in the following manner: The designer may, during algorithm operation,
display on the CRT a history of the cost-function behavior versus the number of tries
(both successes and failures) and also histories of the value of each parameter versus the
number of successes. Based on this information, the designer may be able to determine
the sensitivities of certain parameters in the current region of operation of the cost
domain. If the parameter appears sensitive, the gain, or perturbation size of that param
eter, may be increased. On the other hand, if a certain value is required of a parameter
in order to obtain a success or if the parameter appears insensitive (varies randomly
about some mean), it may be made a constant. In this situation, the number of possible
directions is greatly reduced, as evidenced by table 1, and the array of directions is
regenerated for the reduced number of parameters. Therefore, increasing the gain of
Sensitive parameters and eliminating insensitive parameters should then speed conver
gence to an optimal set of parameters.

RESULTS

The major fields of interest in beam-transport design are transporting and trans
forming either a parallel source or a point source, at the exit of the accelerator, into a
parallel distribution or a point distribution at a desired location. Thus, four cases arise:
parallel to parallel, parallel to point, point to parallel, and point to point. All four are
presented to illustrate the versatility of the method.
Figure 3 shows a typical sampling from the initial distribution used in all parallel
source examples except the parallel-to-point case. Although 60 points were used in each
algorithm iteration, only 30 are shown in order to provide clarity in the plotting. The
distribution used to model the beam-entrance conditions was fixed along the envelope of

15
the phase-plane area, since interior points in the phase-plane area remain in the interior
throughout the transport system.

Four studies were made under the parallel-to-parallel case, all of which used the
equations

Cx = (14a)

Cz = —I- (14b)

C = Cx + Cz + |Cx - CzWait WNN (14c)

as the cost function defining the desired end conditions. (The desired end conditions are
represented as an ellipse in the appropriate phase-plane figures.) The four studies are
(1) best analog results, (2) best digital results, (3) addition of chromatic aberration to the
digital problem, and (4) addition of slits to the digital problem.
Table 2 together with figures 4 to 7 shows a comparison of typical results at the
end of the last quadrupole obtained from the analog program and the digital program using
the analog-obtained parameter values as a starting point for the digital program. As
exemplified in table 2, the digital program in all cases compared made changes in the
parameters which lowered the value of the cost function. Comparison of figures 5 and 7
reveals the visual-resolution problem mentioned previously. Figures 8 and 9 are results
obtained by adding 5 percent chromatic aberration to the digital system of figures 6 and 7.
Table 2 also lists pertinent information for this case. During this study it was noted that
in most cases the introduction of chromatic aberration did not affect the convergence of
the algorithm.

Table 3 and figures 10 to 13 illustrate an example of the introduction of a slit in one

plane to achieve Some desired end conditions which cannot be met by the system without
slits. In the example the requirement to eliminate all slopes greater than a certain value
in the vertical plane, as shown by the points outside the ellipse in the lower portion of fig
ure 10, has been placed on the system. Assuming it has been determined that the system,
subject to parameter constraints, cannot meet the requirement, the problem is solved in

16
the following manner: First, the additional constraint that the focal point of the first
doublet in the vertical plane lie within a certain area between the two doublets is intro
duced into the system (eq. (9b)). The algorithm is then used to determine a set of system
parameters Satisfying all constraints which yield end conditions best fitting the imposed
requirement. At this time a slit is placed at the vertical focal point of the first doublet,
and the slit opening is varied until the imposed requirement is met and the desired end
conditions are achieved (fig. 12).
The last parallel-to-parallel case is a comparison of computer-generated results
with those obtained from an actual experiment conducted at the NASA Space Radiation
Effects Laboratory. Figure 14(a) reflects the computer-generated results while fig
ure 14(b) depicts the results obtained from the actual experiment. (Fig. 14(b) shows the
proton beam passing through a spark chamber after it has exited the beam transport
system.)
In the point-to-parallel study, equations (14) are still used as the cost function since
the desired end conditions are the same as in the parallel-to-parallel case. Therefore,
the only change required to enable the program to go from the parallel-to-parallel case
to the point-to-parallel case is a change in initial conditions. The algorithm will then
determine a new set of parameters which will best meet desired end conditions. Fig
ure 15 shows a typical Sampling from the initial distribution used to simulate a point
source. Figure 16 reflects the end conditions obtained, and table 4 lists the set of system
parameters obtained by the program in addition to values of other variables for the point
to-parallel case. Figure 17 gives a good visual illustration of the point-to-parallel case.
The parallel-to-point and point-to-point cases are obtained by using the appropriate
initial conditions together with a cost function defining the desired end conditions as a
point. The following equations represent the cost function used in these studies:

m 2

- sº
Cx - +++ - (15a)

II] .. 2

Yºº

C = Cx + Cz +|Cx Cz(Waif
- + WNN (15c)

17
Figure 18 is a typical sampling of the initial distribution used for the parallel-to-point
case. Figure 19 and table 5 show the results obtained from the program. It should be
noted that the distance D4 between the end point and the beginning of the first magnet
is an additional parameter to be determined in both the parallel-to-point and point-to
point cases.

The final case, point to point, is shown in figure 20 with its pertinent data given in
table 6. The initial conditions used in this case are the same as for the point-to-parallel
case shown in figure 15.

CONCLUDING REMARKS

A modified creeping-random-search algorithm has been applied successfully to

Solving a particular beam-transport design problem. The method of application that
evolved in the process of solving this problem possesses the advantages of algorithm
operation with analyst interaction, or if desired, analog matching with digital accuracy,
and feasible solutions in the presence of both position-variable and parameter constraints.
In addition, realistic modeling of chromatic aberration has been included in the mathemat
ical model as affecting individual members of the beam, rather than the beam as an entity.
The ability of the method to handle all major cases has been demonstrated using a
particular mathematical model, although flexibility in the model is possible. Results
presented have shown that this method not only yields better solutions than those of a
standard analog program, but also yields solutions for systems, subject to constraints,
for which prior transport optimization programs give infeasible solutions. Also illus
trated by the results are the introduction of chromatic aberration and the use of a slit to
obtain desired end conditions. The results demonstrate that good agreement was obtained
between the results predicted by the program and actual experimental results obtained
with the 600 MeV cyclotron at the NASA Space Radiation Effects Laboratory.

Langley Research Center,

National Aeronautics and Space Administration,
Hampton, Va., November 12, 1971.

18
| REFERENCES

l, Steffen, Klaus G.: High Energy Beam Optics. Interscience Publ., c. 1965.
2. Good, Robert H.; and Piccioni, Oreste: Analog Computer for Charged Particle Tra
jectories. Rev. Sci. Instrum., vol. 31, no. 10, Oct. 1960, pp. 1035-1039.

3. Rampy, John M; and Berry, Donald T.: Determination of Stability Derivatives From
Flight Test Data by Means of High Speed Repetitive Operation Analog Matching.
FTC-TDR-64-8, U.S. Air Force, May 1964. (Available from DDC as AD 440 785.)
4. Colonias, John S.: TRACE: An On-Line Beam Transport Design Program. UCRL
18816 Preprint (AEC Contract No. W-7405-eng–48), Lawrence Radiat. Lab., Univ.
of California, Apr. 1969.
5. Baker, W. F.: A Computer Program To Optimize Magnets in a Beam Transport Sys
tem. Nucl. Instrum. Methods, vol. 20, Jan. 1963, pp. 55–57.
ſ 6, Hawkes, P. W.: Quadrupole Optics. Springer Tracts in Modern Physics, Vol. 42,
º G. Höhler, ed., Springer Verlag, 1966.
* 7. White, Ellis: Eastern Simulation Council Meeting. Simulation, vol. 12, no. 2, Feb.
- 1969, pp. 53–56.
8. Mitchell, Baker A., Jr.: A Hybrid Analog-Digital Parameter Optimizer for
ASTRAC II. AFIPS Conference Proceedings, Vol. 25 — Spring Joint Computer
Conference, Spartan Books, Inc., 1964, pp. 271-285.
9. Bard, Yonathan: Comparison of Gradient Methods for the Solution of Nonlinear
Parameter Estimation Problems. SIAM J. Numerical Anal., vol. 7, no. 1, Mar.
1970, pp. 157–186.
10. Parrish, Russell W.: Parameter Identification Using a Creeping-Random-Search
Algorithm. NASA TN D-6533, 1971.
11. Bowker, Albert H.; and Lieberman, Gerald J.: Engineering Statistics. Prentice-Hall,
Inc., c. 1959.
* Steel, Robert G. D.; and Torrie, James H.: Principles and Procedures of Statistics.
McGraw-Hill Book Co., Inc., 1960.

19
 TABLE 1.- VARIATION OF NUMBER OF POSSIBLE DIRECTIONS

WITH NUMBER OF PARAMETERS

Number of POS Sible directions

parameters (+)
2
8
26
80

:
* 3m - 1 possible directions.
242

728
2186

20
 TABLE 2.- PARALLEL-TO-PARALLEL CASE

Digital
Parameter Analog
NO aberration 5% aberration

K1, m-1 . . . . . . 0.9110 1.1525 1.1525

K2, m-1 . . . . . . –0.9398 - 1.5665 - 1.5665

K3, m-1 . . . . . . 0.8812 0.9125 0.9125

K4, m-1 . . -0.6596 –0.8250 –0.8250

D2, m . . . . . . . 3.21 4.0125 4.0125

9 . . . . . - - - - 3.2747 2.2222 3.2117

X's.) . . . . . . e 1.4965 × 10–3 1.0562 × 10–3 1.2758 × 10–3

7s) . . . . . . - 1.59914 × 10−3 1.0949 × 10−3 1.1208 × 10−3
\ _/
~~

*1 m . . . . . . . 0.7176

'2 m . . . . . . 0.7176

'3. m . . . . . . e 0.6858

'4 m . . . . . . . 0.6858

P1, m . . . . . . . 0.1969

P3 m . . . . . . . 0.2286

N, percent . . . . . 0

Wdif . . . . . . . . 1.0

WN . . . . . . . . 0

21
 TABLE 3.- SLIT IN VERTICAL PLANE

Parallel-to-parallel case; no chromatic aberration]

Parameter Without slit With slit

K1, m-' . . . . . . . . . . 1. 165 1. 165

K2, m-' . . . . . . . . . . - 1.554 – 1.554

K3, m** . . . . . - - - 0.938 0.938

K4, mº" . . . . . . . . - - –0.837 –0.837

D2, m . . . . . . . . . . . 3.747 3.747

C . . . . . . . . . . . . . 2.467 1.967

Fs) - - - - - - - 1.119 × 10−3 1.027 × 10−3

7(s) - - - - - - - - - 1.276 × 10–3 6.566 × 10-4

N, percent . . . . . . . . . 0 63.33
-V-

l1, m. . . . . . . . . . . . 0.7176
!2, m . . . . . . . . e 0.7176

13, m . . . . . . . . . . . 0.6858

l4, m . . . . . . . . . . . 0.6858

D1, m . . . . . . . . . - - 0.1969

D3, m . . . . . . . . . . . 0.2286

Wdif - - - - - - - - 1.0

r, m - - - - - - - - - - 0.3

Xsl; In - - - - - - - - - - - 0.045

22
 TABLE 4. – POINT-TO-PARALLEL CASE

[No chromatic aberration; no focal-point constraints

Parameter Value

K1, m-1 . . . . - - - - - - - - 1.265625

K2, m-1 . . . . . - - - - - - - - - – 1.065625

K3, m** . . . . . . . . . . . . . . 1. 1187.50

K4, m-' . . . . . . . . . . . . . . –0.934375

P2, m . . . . . . . . . . - - 4.28750

*"(Sf) . . . . . . . . . . . . . . . 1.1102 × 10-4

z7(st - - - - - - - - - - - - - - - 2.2458 x 10-4

‘’ . . . . . . . . . . . . . . . . . 3.14786

*1 m . . . . . . . . . . . . . . . 0.8128

22, m - - - - - - - - - - - - - 0.8128

*3 m . . . . . . . . . . . . . . . 0.5588

*4. m . . . . . . . . . . . . . . . 0.5588

*P1, m . . . . . . . . . . . . . . . 0.1016

P3, m . . . . . . . - - - - - - - - 0.0508

Yaiſ . . . . . . . . . . . . . . . . 1.0

WN • * - - - - - - - - - - - - - - 0

*CO), m . . . . . . . . . . . . . . 0.001

*" (O) . . . . . . . . . . • - 0.01

* (O), m . . . . . . . - - - - - - - 0.001

— ”’ (0) . . . . . . . . . . . . . . . . 0.01

23
 TABLE 5. – PARALLEL-TO-POINT CASE

|No chromatic aberration; no focal-point constraints

Parameter Value

K1, mº' . . . . . . . . . . . . . . . –0.2340

K2, mº' . . . . . . . . . . . . . . . 0.4 180

K3, m** . . . . . . . . . . . . . . . – 1.0650

K4, mº' . . . . . . . . . . . . . . . 1.4468

D2, m . . . . . . . . . . . . . . . . 4.7875

D4 m . . . . . . . . . . . . . . . . 9.0.063

C . . . . . . . . . . . . . . . . . . 4.2084 × 10-7

l1, m . . . . . . . . . . . . . . . . 0.5588

l?, m . . . . . . . . . . . . . . . . 0.5588

lº, m . . . . . . . . . . . . . . . . 0.8128

lA, m . . . . . . . . . . . . . . . . 0.8128

D1, m . . . . . . . . . . . . . . . . 0.0508

D3, m . . . . . . . . - - - - - - - - 0.1016

Wdif . . . . . . . . . . . . . . . . . 1.0

WN . . . . . . . - - - - - - - - - - 0

x(0), m . . . . . . . . . . . . . . . 0.055

X'(0) . . . . . . . . . . . . . . . . . 0.000235

2(0), m . . . . . . . . . . . . . . . 0.055

Z'(0) . . . . . . . . . . . . . . . . . 0.000235
 TABLE 6. – POINT-TO-POINT CASE

No chromatic aberration; no focal-point constraints]

*--

Parameter Value
|-----

K1, mº" . . . . . . . . . . . . . . . 0.4000

K2, m-' . . . . . . . . . . . . . - 1.3000

K3, m-1. - - - - - - - - - - - - 1.2000

K4, mº" . . . . . . . . . . . . . . . 1.2499

D2, m . . . . . - - - - 4.7000

P4, m . . . . . . . . - - - - - - - - 11.0656

° - . . . . . . . . - - - - - - - - - - 6.453 × 10-6

*1 m . . . . . . . . . . . . . . . . 0.7176

42, m - - - - - - - - - - 0.7176

&3, m - - - - - - - - - 0.6858

24, m • - - - - - - - - - - - - 0.6858

*P1, m . . . . . . . . . . . . . . . . 0.1969

**3, m . . . . . . . . . . . . . . . . 0.2286

Waif - - - - - • * - - - - - - - - - - 1.0

- - - - - - - - - - - - - 0

> (O), m - - - - - - - - - - - - - 0.001

*" (0) . . . . . . . . . . . - - - - - - 0.01

*CO), m . . . . . . . . . . . . . . . 0.001

z' (0) . . . 0.01

25
 Quadrupole magnetic fields

… / \, \,
/ \ \,
Fixed Fixed Variable
drift Variable drift space drift drift

space space space

* , 4–p, — a 4–p, —# , 4–p, — , .

|-- x

Da T

- PA is a variable in the parallel-to-point and point-to-point cases

Quadrupole axis system

Figure 1.- Beam transport system.

26
 º
L

7\N
\ \
ºº }
N– |
0
–
PT

K1
(a) Two-parameter contour plot.

(b) Parameter directions.

Figure 2.- Illustration of basic algorithm.

27
 .08

.06H-E-41 --

..O'H

X" - .00

- UT.DB = .05 - . 2 .OR .05 O X 10-?

, 10ſw

.08

.06

.0%

.02 --

Z" – .00
ſ
–.O2

— .OH

—.06 i.

— .08 `s
. 10
— ... 10 - .08 - .05 - .OR - ,02 - .00 - 2 - [[ q 5 . . 10 X 10-?

Figure 3.- Standard distribution for parallel initial conditions.

28
 - 26 × 10-2 [] l
- 20
15 P. D __º —

ſp2T
H- H
N
- 10
D
- O5 ſ
-

Fºr N
- OO –H [.

- - O5
\

- - LO
\| | E D J/
P
-- L5
++++ —PT
- - 2O D -

- - 25
re II.
[i][1 -

- - 69-.59-.49 -. 39-.29-.19 –.09 .01 .ll. .21 - 31 - 41 .5l -'6l × 10-1

- 27 x 10-4 X

- 22 [l £3.
- l'7 D

- liz -Hº- This

• OTP P N
Z"
-o- p
-o-1
T - O.S. \ Z

T-lis S- ºl.

T - L S. U

T - 23 D ab
T - es 4-.54–.44–. 34–.24 –. 14-.04 .06 . le . 26 .36 .46 .56 .66 x 10-4

**sure 4.- End conditions for analog case. K1 = 0.9110 m-1}

K2 = -0.9398 m-1; K3 = 0.8812 m 1; K4 = -0.6596 m-1;
and D2 = 3.21 m.

29
 EOTTOM VIEW

| |
SIDE VIEW

Figure 5.- Trace of analog case. K1 = 0.9110 m-1; K2 = -0.9398 m-1;

K3 = 0.8812 m 1; K4 = -0.6596 m. 1; and D2 = 3.21 m.

30
 .25

. 20

. 15

. 10

.05

.00

-. O5

- . 10

-- 15

- - 20

- - 25
l
—.69-.59–.49 –. 39-.29-. 19-.09 .01 . ll .21 - 31 .41 .5l .61 x 10T
x

- 28 ° 10-4 --- F------T---- I --- - - - - -

- 23

- l8

- la

- O8
.

- O3 –4–1-x--—!
D
- - O2 W

--oil–Nº.
T - Il 2 N- ---4-------4- +---

- - l7
El
in --- ----- - - ---- -------

T - 22 th -- - -
--- -
---- |---- —l
-.68 -.58–.48–.38 –. 28–. 18- .08 .02 . 12 .22 .32 .42 .52 .62 × 10

*igure 6.- End conditions for digital case. K1 = 1.1525 m-1;

K2 = -1.5665 m-1; K3 = 0.9125 m-1; K4 = -0.8250 m-1;
and D2 = 4.0125 m.

31
 BG T |G| | | | | ||

SIDF / IF i.

Figure 7.- Trace of digital case. K1 = 1.1525 m-1; K2 = -1.5665 m-1;

K3 = 0.9125 m-1; K4 = -0.8250 m-1; and D2 = 4.0125 m.

32
 . 29 x 10 |

. 22

[] [I]
. 15 2 F}+ → sº - H --
.08 S!

.01 [] . . [I EY
X" –.06
ºn
Vº [][]
*H)
} J

{} H N [...] →tº }
- . 13 - ~ HDH ~ |

—. 20 -

-. 27—Illil ſº ---

-. 34

-. 41 H
–.07 –.05 -.03-.01 .01 .03 .05 .07 .09 .11 . 13
X
.25 × 107%
[] {J
. 20

. 15
—t----
3–––– º
--- ſº-------
r
- -----

. 10 H.
Gł T
. 05 ––P. \ --------

Z" [] C
—.00 [...]
-
[]
— .05 []] ---4-----------. --- -- - | --- - - ---

r- |
- . 10 ~, +--AP3 —f-

- . 15 is ºf –1–

i-i-f
r- |

–. 20 {} [...] ------- -- -

—. 25 L | ſ
–.06–.04 –.02 - .00 .02 .04 .06 .08 . 10 . 12 . 14
Z

Figure 8.- End conditions for digital case with 5% chromatic aberration.
Ki = 1.1525 m-1; K2 = -1.5665 m-1; K3 = 0.9125 m-1;
K4 = -0.8250 m-1; and D2 = 4.0125 m.

33
 BG T TOM V [f W

- =TT –
- == ==TT--— –
-
===

|–| | – |

5 IDE VI tº

Figure 9.- Trace of digital case with 5% chromatic aberration. K1 = 1.1525 m-1;
K2 = -1.5665 m-1; K3 = 0.9125 m-1; K4 = -0.8250 m-1; and D2 = 4.0125 m.

34
 ſº

.24 × 10^
. 20

... lé -

PTHTES
. 12
n/ſ || | RöN -

. O8 -- - -

D
.04 #
X" D |

.00 d : .
- !ſ
. O4 . + :|

. 12
\l : H
As -

... lo +-4 Nºdºlº -

. 20 ſ i --- ---------- --

-. O7-. O5-. O3 -. O1 . Ol .03 - 05 . O7 . O9 - 11 . 13

A.

.26 × 10^* ----------- ----

2l [...]
WH
..lé E}+- [T]_

ll
J-T ---> -
afº .06 2^ + []
--1-
Z" .0l p + U +

* -.04
\ t t

1
*
––––4–
-

--------
H
+---
!
-

N - | s l

-.09 sºf-t-t-t- F- + - 3. - * -

- . 14 *— *- i + = *-i- -----

- . 19 [] . ----- +---- 4- + -----

{
-. 24 t ſºm ------ i. --|-- :

-.6l-.52- .43 -. 34-. 25- . 16 - .07 .02 . ll . 20 .29 .38 .47 .56 .65 x 10-4
Z

Figure 10.- End conditions for system prior to introduction of slits.

K1 = 1.165 m-1; K2 = -1.554 m-1; K3 = 0.938 m-1;
K4 = -0.837 mº"; and D2 = 3.747 m.

35
 BOTTOM VIEW

|
:
SIDE VIEW

Figure 11.- Trace of system prior to introduction of slits.

K1 = 1.165 m-1; K2 = -1.554 m-1; K3 = 0.938 m-1;
K4 = -0.837 m. 1; and D2 = 3.747 m.

36
 .19 x 10" r--~ +

..le N : - - -

. 13
... 10 7
Z ºx-- | D -

.07 /
i
.04 I- + - -- 5–
* *
--

; :

º X' . Ol

-.02

-. 05

-.08 \
1 l

-
-

|
1

|
-.ll \ I-7–1 ––––
-a, -\h A ||
..,-.07-.05-.03
| N -.ol .ol.os.os.oz.09.11 .13
-

. 18 x 10' + → - ---

. 15 º Tiss- |
. Q9 Z

…
|A
e
~
-
Z'
*H
#
.00 -H ---|---4----|--|--|--
|

-. 03 - # - | | | |
- |
1
*
|
-.06–A–1–
T —
º —
+ r—
|
N
-. 09 NH ; +: I
s i
- . 12 SD —l —l

- . 15 4
;
-
:
-r
2
*—
- . 18 º
F--T-----|--|--
—l l
-.61-.52-.43–.34 -.25-.16-.07 .02 .ll .20 - 29 .38 - 47 -56 - 65 * 10T
Z

Figure 12.- End conditions for system with slits. K1 = 1.165 m-1;
K2 = -1.554 m-1; K3 = 0.938 m-1; K4 = -0.837 m. 1; and
D2 = 3.747 m.
 BOTTOM VIEW

:
SIDE VIEW

Figure 13.- Trace of system with slits. K1 = 1.165 m-1; K2 = -1.554 m-1;
K3 = 0.938 m-1; K4 = -0.837 m-1; and D2 = 3.747 m.

38
: Horizontal
2
View

Spark Chamber

zzzzzzzzzzzzzzzº
-

Vertical
== *-Proton Beam
View E

Z
zzzzzzzzzzzzzzzzzz!"
(a) Computer-generated curves.
Figure 14.- Comparison of computer results and actual experiment.

39
 -
-

*o sº º:
A E. W.

V. E. RT I C A L - - º- p ROTC. N.

L–71–7119
(b) Results of experiment.
Figure 14.- Concluded.

40
 -*—iſh
. 10ſr-ru L-T

.08 U

.06|

X" -
.00 |
—T

.08

. 10
* - U - .
º
- DS-, OR - ,D - .00 JU2
—P
,0ſ. .05 .08 • 10 X 10-2

*—iſh
, 10ſº ITU L-T

.08

.06 --

..O'H

.02 |

Z" – *
-.02 D

— .OH

–.06

-.08

.1
9IUH =TE-DT-i-. O º º º 5 w B O X 10-?
Z

Figure 15.- Distribution for point-to-parallel case initial conditions.

41
 28 x 10T T- º
. 24 | A. +
:

• 20
*

i
\| _ !

. lé D HEVT
. 12
O | *H
:º
.08 -

I j I
| | ||
.04 TU —t +- –
-

* .00 E, +++---—
|
-.04 HG ! - l

-.
99 He -
H i
--4-H-----|--————
5| º |

º | i

-*-*H. L. H-----
-.l6 / H---- [] [[D D :

/ :
-. 20 +
-. 24 H / --

:
-. 28 ,
-. 07-. 05-.03 -.0l .01 .03 .05 - 07 . 09 . ll . l.2

.38 x 10"

.31

24
:
17 —
. . lo

.03 – 4---

-.04.

-. ll —

- . 18 —

-. 25
t

--32–1–1-- --L- D-1- Il L. -- . . k i - -

l
-.46-. 39-. 32-.25 - . 18–. ll-.04 .03 - 10 - 17 . 24 .31 - 38 .45 x 10
Z

Figure 16.- End conditions for point-to-parallel case.

42
 BOTTOM VIEW

|
| || | ||

SIDE VIEW
Figure 17.- Trace of point-to-parallel case. K1 = 1.265625 m-1;
K2 = -1.065625 m-1; K3 = 1.118750 m-1; K4 = -0.934375 m-1;
and D2 = 4.28750 m.

43
 —ſº
.26ſw-Tū

f º
--GHH
r
D ;;
:
§

- r) —J

&
º
• i
#
f
7
“ºtrº-HT-ºr-º-º-º:HE–F#–E–F–Au

Figure 18. – Distribution for parallel-to-point case initial conditions.

44
 BOTTOM VIEW

SIDF VI FW

Figure 19.- Trace of parallel-to-point case. K1 = -0.2340 m-1;

K2 = 0.4180 m-1; K3 = -1.0650 m-1; Ka = 1.4468 m-1;
D2 = 4.7875 m; and D4 = 9.0063 m.

45
 HOT TOM V [FW

SIDE VIEW

Figure 20.- Trace of point-to-point case. K1 = 0.4000 m-1;

K2 = 1.3000 m-1; K3 = -1.2000 m-1; K4 = 1.2499 m. 1;
D2 = 4.7000 m; and D4 = 11.0656 m.

46 NASA-Langley, 1971 — 19 L-7930

NATIONAL AER ON AUTICS AND SPACE A D M is TRAT I ON
WASHINGTON, D.C. 20546
POSTAGE AND FEEs PAid
NATIONAL AERONAUTICs AND
OFFICIAL BUSINESS
space Ad MinisTRATION
PENALTY FOR PRIvaTE use $300
FIRST CLASS MAIL

01 0 002 cl U 35 7 112 17 Soo 4 39H

UN IV ºn F I LL IND IS
LI PR ARY
ATT N: DncljMENT LIRR ARIAN
Up D ANA IL 6 1801

POSTMASTER . If Undeliverable
Postal Manual) Do(Section 1
Not Ret

"The aeronautical and ſpace activitieſ of the United States ſhall be

conducted ſo aſ to contribute . . to the expanſion of human knowl
edge of phenomena in the atmoſphere and ſpace. The Adminiſtration
shall provide for the wideſt practicable and appropriate diſſemination
of information concerning itſ activitieſ and the reſult, thereof.”
— NATIONAL AERONAUTICS AND SPACE ACT OF 1958

NASA SCIENTIFIC AND TECHNICAL PUBLICATIONS

TECHNICAL REPORTS: Scientific and TECHNICAL TRANSLATIONS: Information

technical information considered important, published in a foreign language considered
complete, and a lasting contribution to existing to merit NASA distribution in English.
knowledge.
SPECIAL PUBLICATIONS: Information
TECHNICAL NOTES: Information less broad derived from or of value to NASA activities.
in scope but nevertheless of importance as a Publications include conference proceedings,
contribution to existing knowledge. monographs, data compilations, handbooks,
sourcebooks, and special bibliographies.
TECHNICAL MEMORANDUMS:
Information receiving limited distribution TECHNOLOGY UTILIZATION
because of preliminary data, security classifica PUBLICATIONS: Information on technology
tion, or other reasons. used by NASA that may be of particular
interest in commercial and other non-aerospace
CONTRACTOR REPORTS: Scientific and
applications. Publications include Tech Briefs,
technical information generated under a NASA Technology Utilization Reports and
contract or grant and considered an important
Technology Surveys.
contribution to existing knowledge.

Details on the availability of these publications may be obtained from:

SCIENTIFIC AND TECHNICAL INFORMATION OFFICE

NATIONAL AERO NAUTICS AND SPACE ADMINISTRATION

Washington, D.C. 20546
/27/3 Pv6//
lº. 32.5% w
4 & 72–

NASA TEC H NICAL IN OTE NASA TN D-6572

. THE LIBRARY OF THE

JAN 11 1972
UNIVERSITY OF ILLINOIS
AT URBANA-CHAMPAIGN

THEORETICAL BUCKLING LOADS

OF BORON/ALUMINUM AND GRAPHITE/RESIN
FIBER-COMPOSITE ANISOTROPIC PLATES

by Christos C. Chamis
Lewis Research Center
Cleveland, Ohio 44135

MATIONAL AERONAUTICS AND SPACE ADMINISTRATION - WASHINGTON, D. C. - DECEMBER 1971

 . Report No. 2. Government Accession No. 3. Recipient's Catalog No.

NASA TN D–6572
4. Title and Subtitle THEORETICAL BUCKLING LOADS OF BORON/ 5. Report Date

ALUMINUM AND GRAPHITE/RESIN FIBER-COMPOSITE -*.*.*.* Code

ANISOTROPIC PLATES

7. Author(s) 8. Performing Organization Report No.

Christos C. Chamis E-6455

10. Work Unit No.
. Performing Organization Name and Address 134-14
Lewis Research Center 11. Contract or Grant No.

National Aeronautics and Space Administration

Cleveland, Ohio 44135 13. Type of Report and Period Covered

, Sponsoring Agency Name and Address Technical Note

National Aeronautics and Space Administration 14. Sponsoring Agency Code
Washington, D.C. 20546
15. Supplementary Notes

. Abstract

Theoretical results are presented for the buckling of anisotropic plates. The plates are sub
jected to simple and combined in-plane loading. The plates are made from fiber composite
material of boron/aluminum or high-modulus graphite/resin. The results are presented in
nondimensional form as buckling load against fiber orientation angle for various plate aspect
ratios. The results indicate that buckling loads of boron/aluminum plates are independent of
fiber direction if the plate aspect ratios are greater than about 1, and moderately dependent
when this ratio is less than about 1. In addition, the results indicate that the buckling loads
are independent of aspect ratio for plates with aspect ratios greater than about 2. Boron/
aluminum composite plates can resist buckling loads more efficiently than graphite/resin
composites on a specific buckling stress basis. The numerical algorithm and a listing of
the computer code used to obtain the results are included.

17. Key Words (Suggested by Author(s)) 18. Distribution Statement

Buckling loads, Anisotropic plates, Single and Unclassified – unlimited

combined loads, Fiber composites, Boron/

aluminum, Graphite/resin, Design data,
Numerical algorithm, Computer code
19 Security Classif. (of this report) 20. Security Classif. (of this page) 21. No. of Pages 22. Price"
Unclassified Unclassified 37 $3.00

For sale by the National Technical Information Service, Springfield, Virginia 22151
º:

s
 THEORETICAL BUCKLING LOADS OF BORONIALUMINUM AND GRAPHITEIRESIN

FIBER—COMPOSITE ANISOTROPIC PLATES

by Christos C. Chamis
LeWis Research Center

SUMMARY

Theoretical results are presented for designing with boron/aluminum composites

when these composites are buckling critical. The composites are assumed to be rec
tangular plates with four simply supported edges. They are subjected to single and com
bined normal and shear loads in the plane of the plate. The plates are made from a
unidirectional composite whose fiber direction is oriented at some angle to the load
direction.

The design data are presented in nondimensional form as buckling load against
orientation angle for several plate aspect ratios. The results indicate that the buckling
loads of boron/aluminum plates are independent of fiber orientation if the plate aspect
ratio is greater than approximately 1. The buckling loads are moderately dependent on
the orientation angle for plates with aspect ratios less than about 1. The buckling load
is independent of aspect ratio in plates with aspect ratios greater than about 2.
Comparison of buckling results for boron/aluminum composite plates and
Thornel-75/epoxy composite plates indicates that the boron/aluminum composite plates
resist buckling loads more efficiently than the Thornel-75/epoxy composite on the basis
of specific buckling strength. The results also indicate that the buckling loads of boron/
aluminum composite plates can be predicted using orthotropic theory if their aspect ra
tio is greater than about 1.
The numerical algorithm used to solve the buckling problem and listing of the cor
responding computer code through which the results were obtained are included.

INTRODUCTION

Feasibility studies for the space shuttle indicate that the use of advanced fiber com
posite structural components can result in a considerable increase in payload in the
shuttle system. Boron/aluminum and graphite/resin fiber composites are leading con
tenders for shuttle applications because these composites offer high stiffness-to-density
and high strength-to-density ratios. Panels made from these materials will have to meet
both material strength and buckling requirements. This report deals with a theoretical
investigation of the buckling of flat rectangular panels made from boron/aluminum and
Thornel-75/epoxy fiber composites. º

Several papers have appeared recently dealing with the buckling of anisotropic plates º

(refs. 1 to 8). However, design data are not available for flat panels subjected to com
pressive loads and made from advanced fiber composites such as boron/aluminum and º

graphite/resin. The method described in reference 3 has proved efficient in buckling -

studies of boron/epoxy plates. This method is used herein to generate design data for
boron/aluminum plates. Some data for Thornel-75 graphite/epoxy resin plates are also
generated for comparison purposes. Data were generated for aspect ratios of 1/2, 1,
2, and 4.
The panels considered are anisotropic and simply supported. They are subjected to
combined in-plane (normal and shear) load (fig. 1). The material is a unidirectional
composite with the fiber direction oriented at an arbitrary angle to the load direction
(fig. 1). The analytical algorithm used is the assumed mode method in conjunction with
the Galerkin method. A computer code was developed based on the Galerkin method, and
the code was used to generate the theoretical design data presented herein. A brief de
scription of the analytical method is given in the report. All symbols are defined in ap
pendix A. The numerical algorithm used to solve the resulting eigenvalue problem is
described in appendix B. Input data sample sheets with explanations are given in appen
dix C. A listing of the computer program with sample cases is given in appendix D.

2 FIBER
/ DIRECTION
(MATERIAL
1 AXIS)

x (STRUCTURAL
AXES)

Figure 1. - Fiber-composite plate geometry and loading - all four edges simply supported
(aspect ratio - alb).
 BRIEF DESCRIPTION OF UNDERLYING THEORY

The underlying theory for buckling loads of anisotropic plates is described in refer
ence 3 with pertinent discussions in references 6 to 8. Briefly, this theory consists of
expressing the potential energy of a plate in terms of displacement variables. Taking
the variation of the potential energy function yields the field equation and the correspond
ing boundary conditions. The resulting system then is solved by the assumed mode
technique in conjunction with the Galerkin method.
The equation resulting after the variation of the energy function is
al b
/ Aſ [Piaw. xxxx * 2(D12 + 2D33)w, xxyy * D22W, yyyy * 4D13W, xyxx

2N w, xy +N W, joway as
(Nºw
+ 4P23W, xyyy - (N.W., xx ++ 2Nzy yº yy

b a.

+ Aſ (D11W, xx + P12W, yy * 2D.13W, sh Öw, x dy

al b

(The notation is defined in appendix A.) The area integral represents the field equation,
and the line integrals represent the boundary conditions.
The assumed buckling mode described in reference 3 is represented by a Fourier
double sine series. This mode satisfies the imposed boundary conditions, but it does
not satisfy the natural boundary conditions if the material and structural axes do not co
incide. However, the mode is forced to satisfy the natural boundary conditions approx
imately through the Galerkin method as discussed in reference 3.
Substituting the assumed mode in equation (1), applying the Galerkin method, and
carrying out the algebra result in a set of linear equations which represent the eigen
value problem of the plate. This system is coupled for either a combination of shear
and normal loads and/or noncoincident material and structural axes.
The eigenvalue problem is solved by using the Power method, which is a highly ef
fective iterative numerical technique in seeking the largest eigenvalue of the system.
The indicial equations which were used to generate this system and the Power method
are given in appendix B in outline form.
 BRIEF DESCRIPTION OF COMPUTER PROGRAM

The numerical algorithm described in appendix B has been transformed into a com
puter code which is rather simple and can be generated from the information supplied
in appendix B. A FORTRAN IV compiled listing of a computer program based on the
algorithm in appendix B is given in appendix D, with sample cases and output. Input
data sample sheets are given in appendix C.
The inputs to the code are composite system identification, fiber volume ratio,
orientation angle, plate aspect ratio, and flexural rigidities. The outputs are the num
ber of terms in the assumed mode series expansion required for convergence, the rela
tive error between the last two iteration cycles, the buckling load, and topo-plot data
of the buckled shape of the plate normalized with respect to the largest deflection.
The algorithm described in appendix B runs into difficulty when the shear buckling
load of a plate is sought. In this particular case, the difficulty is bypassed by including
normal loads which are a very small fraction of the shear load. Further discussion on
why these difficulties arise is presented in reference 3.

THEORETICAL DESIGN DATA

The theoretical design data generated herein are based on the schematic illustrated
in figure 1. In this figure, the type of loading condition, the plate geometry, and the
fiber orientation are defined. The x-y coordinate reference system is referred to as
the structural axes system. The fiber direction coordinate system which is located at
the angle 6 from the structural axes system is referred to as the material axes sys
tem. The loading conditions are identified by N, N and Nxy as is noted in the
y’ y
figure.
The flexural stiffnesses required in calculating the buckling loads were calculated
by using the computer code described in reference 9. Typical values of the elastic con
stants of the plate along its material axes are given in table I for boron/aluminum and
Thornel-75/epoxy composites with a fiber volume ratio of 0.5. The flexural rigidities
are computed as functions of the orientation angle using the data in table I.
 TABLE I. - THEORETICAL UNIDIRECTIONAL COMPOSITE

PROPERTIES AT FIBER VOLUME RATIO OF 0.5

Property Boron/aluminum Thornel-75/epoxy

Longitudinal modulus, 24.2×10° (35.0×10°) 26.0×10° (37.8x10")

N/cm” (psi)
Transverse modulus, 16.8×10° (24.3×10°) | 6.9-10° (1.0×10")
N/cm” (psi)
Shear modulus, 8.0×10° (11.6×10°) 0.44×10° (0.63x10°)
N/cm” (psi)
Poisson's ratio 0.24 0.25

Density, g/cm" 2.62 (0.095) 1.55 (0.056)

(1b/in.”)

Buckling of Boron/Aluminum and Thornel-75/Epoxy Panels

Buckling loads for a single loading condition for panels made of boron/aluminum and
Thornel-75/epoxy are illustrated in figure 2, where the specific buckling stress has been
plotted as a function of the orientation angle for an aspect ratio of 2. The schematic in
the figure illustrates the type of load condition as well as the orientation angle. As can
be seen, in this figure, boron/aluminum composites are more efficiently utilized than

12x100

BORON/ALUMINUM º
l0

8H

º F-THORNEL
| 75/EPOXY
i.
:3
:: **
*
FIBER DIRECTION
4 - w

--~~~~ |-

...Hºº-º-
x -ī- N,
2H- === 2: X

| | |
0 30 60 90
ORIENTATION ANGLE, 6, DEG

Figure 2. - Specific buckling stress of two fiber-composite plates -

all four edges simply supported. Fiber volume ratio, 0.5; aspect
ratio alb, 2.
 1.4ſ

BORON/ALUMINUM

8–

THORNEL-75/EPOXY

- FIBER
Ny DIRECTION

30
ORIENTATION ANGLE, 9, DEG

Figure 3. - Buckling loads for two fiber-composite plates - all four edges simply
supported. Fiber volume ratio, 0.5, aspect ratio alb, 2.

Thornel-75/resin composites in structures which are critical in buckling as measured by

the specific buckling stress.
Buckling load comparisons where the panels are loaded in the y-direction are il
lustrated in figure 3. In this figure, it can be seen that the boron/aluminum composite
panel is considerably stronger in buckling than the corresponding Thornel-75/epoxy
panel. In this plot, the nondimensional buckling load parameter is plotted as a function
of the orientation angle for the fixed panel aspect ratio of 2.

Buckling Loads for Individual Loading Conditions

Design data for boron/aluminum panels which are subjected to compressive load in
the x-direction are illustrated in figure 4 as a function of orientation for various aspect
ratios. The important point to be noted from this figure is that the buckling load is in
dependent, or almost independent, of orientation angle in panels where the aspect ratio
is approximately greater than 1. Another point to be noted is that the buckling load de
pends only moderately on the orientation angle in panels of aspect ratio less than 1.
A cross-plot of figure 4 is illustrated in figure 5. The nondimensional load is
 ASPECT
RATIO,
alb

|- 1/2

- 1, 2, 4
y - FIBER
i DIRECTION
2|− 6

Nx
1H- X

| | | | _l
30 60 90
ORIENTATION ANGLE, 6, DEG
Figure 4. - Buckling loads for boron/aluminum composite plates, with all
four edges simply supported, subjected to normal (N,) compressive load.
Fiber volume ratio, 0.5.

ORIENTATION
ANGLE,
- 0.
| DEG
| --0
|-

90- \
\
y FIBER

l 2
ASPECT RATIO, alb
Figure 5. - Buckling loads for boron/aluminum composite
plates, with all four edges simply supported, subjected to
normal (N,) compressive load. Fiber volume ratio, 0.5.
plotted as a function of panel aspect ratio for various orientation angles. The insensi
tivity of the buckling load as a function of orientation angle in panels with aspect ratios
greater than about 1 is clearly illustrated in this figure. The dashed line symbol is used
to represent these curves to emphasize that buckling load values were computed only at
the aspect ratios 1/2, 1, 2, and 4.
Buckling loads for panels which are loaded in the y-direction only are given in fig
ure 6. As can be seen in this figure, the buckling loads are almost independent of the
orientation angle in panels with aspect ratios of 1/2 and greater. The buckling load, on
the other hand, is very sensitive to the aspect ratio in panels with aspect ratios of ap
proximately 2 or less.
Buckling loads for a boron/aluminum composite panel loaded with shear only are il
lustrated in figure 7. The points to be noted in this figure are the following:
(1) There is a mild buckling load dependence on the orientation angle for panel
aspect ratios of less than about 1.
(2) The buckling load is relatively independent of orientation angle for panel aspect
ratios of greater than about 1.
(3) The buckling load is very sensitive to the aspect ratio in panels with aspect ra
tios less than 2, and this dependence becomes rather insignificant for panel aspect ra–
tios greater than 2.

- ASPECI
=* RAI10,
º: alb ASPECT

º, 10H E 20
*
y
| N
112
º: `- RATIO,
alb

cº H. ,-FIBER º, N FIBER 1/2

<
3 8H 2. z * DIRECTION zº’
*-
16|- y xy z
,-

-1
$2
5: 6
º
*
º
2^e c.

3. l
--_-_--_-

b *}
12.É.
N
DIRECTION

x 2H- x

#
-

|
#
º

* º,
N
# y §
— - - -

3 AH tº
<
8H
l
#
co
l 3
vº)
2

# 2H tº 4E 4
~ 2 >
4 s
| | | | | 2 | | | | | |
0 30 60 90 0 30 60 90
ORIENTATION ANGLE, 9, DEG ORIENTATION ANGLE, 6, DEG
º t ..º.º.º.º.
3S, W - - -
º 7. ...'. loads for .." Composite
plates, with all four edges simply supported, subjected
to normal (N, compressive load. Fiber volume ratio, 0.5. to shear (No load. Fiber volume ratio, 0.5.
 Buckling Loads for Two Equal Simultaneous Loadings

Buckling loads, when the panel is loaded with equal loads in the x- and y
directions, are shown in figure 8. In this figure, the nondimensional buckling load pa
rameter is plotted as a function of orientation angle for various panel aspect ratios.
The results in this figure show that the buckling load is slightly dependent on the

E 5T
ck’
+
‘A ASPECT
2: 4 y RATIO,
c. | Ny- x alb
5 TTITI FIBER 1/2
—

g F- º z
<\; * -

no
s: %HN
-

s 2.É.]". ,
co 2– | | || ||
=
à l
vºn

à l'H ;
>
º

3
2 | | l l 1–
0 30 60 90
ORIENTATION ANGLE, 6, DEG

Figure 8. - Buckling loads for boronlaluminum composite

plates, with all four edges simply supported, subjected
to combined normal compressive (N, - Ny loads.
Fiber volume ratio, 0.5.

orientation angle for panel aspect ratios less than 1, and practically independent of
orientation angle for aspect ratios greater than 1. The buckling load is sensitive to panel
aspect ratio for aspect ratios less than or equal to 2. This dependence becomes insig
nificant for panel aspect ratios greater than 2. The curves of the buckling load as a
function of the independent variables indicated in figure 8 parallel the curves of the
buckling loads indicated in figures 4 and 6, for the individual loadings.
Buckling loads for the case when the panel is loaded in the x-direction combined
with shear are shown in figure 9. The curves of the buckling load for this loading con
dition are parallel to those of the individual cases (figs. 4 and 7). Buckling loads for
the case when the panel is loaded in the y-direction combined with shear are illustrated
in figure 10. The buckling load in this figure seems to be practically independent of
orientation angle for the aspect ratios investigated. However, it is quite sensitive to
the panel aspect ratio for aspect ratios less than 2.
 |- ASPECT
RATIO,
alb
|- 1/2

}
- 4
y No N, ,-FIBER
Hå - -- X5 DIRECTION

- b

Ø% "...,
| | | | | |
30 60 90
ORIENTATION ANGLE, 9, DEG
Figure 9. - Buckling loaus for boronlaluminum composite
plates, with all four edges simply supported, subjected
to combined normal (N,) and shear (Nxy N,) - loads.
Fiber Volume ratio, 0.5.

ASPECI

l0 - RATIO,
alb
T- 1/2

- y F-HBER
| Ny DIRECTION
-
====4,| -\g…'

b º -

-
º | No
+-I-T-3–x
N,

2
4
| | | | | |
30 60 90
ORIENTATION ANGLE, 6, DEG
Figure 10. - Buckling loads for boron/aluminum composite
plates, with all four edges simply supported, subjected to
combined normal (N) and Shear (Nxy Ny loads.
- Fiber
volume ratio, 0.5.

10
 Buckling Loads for Three Equal Simultaneous Loadings

Buckling loads for panels which are loaded in the x- and y-directions combined with
shear are shown in figure 11. The schematic in this figure indicates the type of loadings
and their respective ratios. The nondimensional buckling load is plotted as a function of
Orientation angle for various panel aspect ratios.
Comparing corresponding curves from figures 8 and 11, it is seen that the addition

HBER
DIRECTION
2* *

0 30 60 90
ORIENTATION ANGLE, 9, DEG
Figure ll. - Buckling loads for boron/aluminum composite
plates, with all four edges simply supported, subjected to
Combined normal (N, - N,) and shear (No. - N,) loads.
Fiber volume ratio, '0.5.

of the shearing load decreases the buckling load of the panel only slightly. The point to
be noted then is that a panel subjected to compressive loads in the x- and y-directions
will resist almost an equal amount of shearing load for approximately the same buckling
load.

Buckling Loads for TWO or Three Unequal Simultaneous Loadings

Buckling loads for panels which are subjected to unequal loads in the x- and y
directions are shown in figure 12. The type of loading condition and respective loading
magnitudes are illustrated in the sketch given in the figure. In this figure, the nondi
mensional buckling load parameter is plotted as a function of the orientation angle for
various aspect ratios. The curves of the buckling load for this type of loading condition

11
 ASPECI
RATIO,
alb

~ 9-ſº
DIRECTION
1/2

N,
—x

| | | | | |
0 30 60 90
ORIENTATION ANGLE, 9, DEG

Figure 12. - Buckling loads for boron/aluminum composite

plates, with all four edges simply supported, subjected to
COmbined normal (N, (12 N,) loads. Fiber volume ratio,
-

0. 5.

ASPECT
3 5r- RATIO,
“. alb

‘A 1/2
2:” 4H
c- --- FIBER
35
-1
2. Yg birtchon
g "H N,
Sz - — X
9 EF Hy, = (1/2)N
* 2|− XY_ x l

:
O
4
_-
2
vº, 4
à l'H 2
>
=
2 | | | | | |
0 30 60 90
ORIENTATION ANGLE, G, DEG
Figure 13. - Buckling loads for boronlaluminum composite
plates, with all four edges simply supported, subjected to
Combined normal (Ny = (1/2)N.) and Shear (No. - (1/2)N,)
loads. Fiber volume'ratio, 0. §

parallel those of the cases with equal loading condition, as was previously discussed (see
fig. 8). One additional point to be noted is that the buckling loads of panels with aspect
ratios of greater than about 2 remain almost invariant as a function of aspect ratio when
the orientation angle is greater than about 45°.
Buckling data for panels loaded with unequal combined loading conditions are illus
trated in figure 13. The loading condition for the panel is illustrated in the schematic
in the figure. The buckling load is plotted as a function of orientation angle for various

12
aspect ratios. A point to note in this figure is that, at some orientation angles, panels
with aspect ratios greater than 2 could have greater buckling loads than panels with
aspect ratios of 2. As can be seen, the buckling load for a panel with an aspect ratio of
approximately 2 is lower than for the panel with an aspect ratio of 4 when the orientation
angle is approximately less than 459
All the buckling data presented and discussed indicate that the buckling load is insen
sitive to orientation angle for panels with high aspect ratios. The buckling load is mildly
sensitive to the orientation angle in panels with low aspect ratios. This observation
leads to the important conclusion that the buckling loads of boron/aluminum composite
anisotropic plates can be approximately determined by using classical orthotropic theory.
This conclusion is indeed a useful one, since the buckling of orthotropic plates has been
extensively treated in reference 5.

POSSIBLE EXTRAPOLATIONS OF DESIGN DATA

The design data presented and discussed were based on a fixed fiber volume ratio of
0. 5. The data presented herein can be used to extrapolate buckling loads for plates made
from composites with different fiber volume ratios. The results will be very close if the
variation of the fiber volume ratio is within approximately +20 percent of the 0.5 value
which was used in generating the design data.
This extrapolation is justified since the buckling load is nondimensionalized with
respect to both composite longitudinal modulus and thickness. It is well known that both
composite modulus and thickness depend on the fiber volume ratio (ref. 10), and that this
dependence is approximately linear in the fiber-volume-ratio range 0.4 to 0. 6. In this
sense, then, the extrapolation using the design data presented herein for fiber volume
ratios within +20 percent of 0.5 should yield reasonable results.
The computer code appended in appendix C can be slightly modified to compute the
first natural frequency of anisotropic boron/aluminum composite panels. In reference 5,
the vibration problem is discussed, and the analogy in computing the buckling load
and the natural frequency is made.

CONCLUSIONS

The discussion of the theoretical design data presented leads to the following con
clusions:

1. Design data for the buckling of unidirectional boron/aluminum panels with fibers
oriented at any angle to the load direction have been generated and are reported herein.

13
 2. Specific buckling stress comparisons showed that, in general, boron/aluminum
composite panels are more efficient than high-modulus graphite/resin composite panels.
3. The buckling load of boron/aluminum unidirectional panels is practically inde
pendent of fiber direction at high aspect ratio values. At these aspect ratios, the plate
can be assumed to have its material axes coincide with its structural axes. Conse

quently, the classical buckling theory of orthotropic plates can be used to predict the
buckling load.
4. The buckling loads of boron/aluminum unidirectional panels are only moderately
dependent on fiber direction at plate aspect ratios less than 1.
5. Boron/aluminum composite panels loaded by normal in-plane loads which are
near the critical load can carry considerable shear load before they buckle.
6. The buckling loads of boron/aluminum panels are practically independent of
aspect ratio at aspect ratios greater than about 3.
7. The buckling loads of panels with fiber volume ratios within approximately
+20 percent of 0.5 can be extrapolated from the design data presented herein by using
the appropriate panel thickness and the appropriate composite longitudinal modulus.

Lewis Research Center,

National Aeronautics and Space Administration,
Cleveland, Ohio, September 3, 1971,
129-03.

14
 APPENDIX A

SYMBOLS

panel dimension, x-direction

panel dimension, y-direction

#: D matrix of flexural rigidities, eq. (1)
h thickness
*:::
K plate stiffness matrix, elements given by eqs. (B1) and (B4)
!{ L load matrix, elements given by eqs. (B1) and (B4)
summation index limit on m
:
m Summation index
ſº
N summation index limit on n; with subscripts, applied load
N parameter in eq. (B1)

Cr buckling load
n Summation index

p Summation index

Q Summation index

W vector of coefficients in the displacement mode expansion, eq. (1)

W displacement in z-direction

X, Y, Z panel structural axes

1, 2, 3 panel material axes

Ö Variation

t",é" convergence tolerances

§ orientation angle of material axes with respect to structural axes

\ buckling parameter
0 ratios defined by eq. (B5), also weight density
J Stress and matrix of stresses

Subscripts:
Cr
critical or buckling condition
i row index

15
 column index

undirectional composite property

buckling modes, x-direction, axes
buckling modes, y-direction

iteration cycle

X, y, Z directions associated with the respective structural axes

1, 2, 3 directions associated with the respective material axes

16
 APPENDIX B

DESCRIPTION OF NUMERICAL ALGORITHM

The numerical algorithm seeks the eigenvalue of the following matrix equation:

[K]{w} = NUL|{W} (B1)

where [K] and [L] are (M × N) × (M × N) square matrices, (W) is a column matrix con
|aining the Wmn, and N is defined subsequently. The (i,j) elements of the [K) and [L]
matrices are given by the following indicial expressions:

m = 1(1)M; n = 1(1)N

p = 1(1)M; q = 1(1)N (B2)

i = (m - 1)N + n; j = (p − 1)N + q

~
4
_ II* I b a -4
Kii = + * mºdu 2
+ – In "In
2 2
(D12 + 2D33) ++ n D22
4 a3 ab b3
p = m and q = n
2
b - 2 a -2
11 #(.
Lºs = - 1–|B mºp
*
+= n p
b ) º B3
)

Kii = Lii = 0 p # m or q + n

2 2 2 2
Ki = - 87°mnpg 2m + 1|D13 - 87°mnpg 2n + 1)D23
*a* - n2\\,,2 - ...? b”(p.” - m”)\g” - n”
m + p and n + q odd
La --—“*—p (B4)
(p” - m”)(q* - n”)

KH = Lij = 0 m + p or n + q even

17
where

N N N
px ==;
N
p,y ++,
N
psyy ==}}
N
(B5)

where N is the buckling parameter. Here N. N. and » 2

N,y carry their algebraic
signs, whereas N is always taken as a positive antity.
The buckling loads are determined by finding the largest eigenvalue in equation (B1)
using the Power method. Several methods are available for finding eigenvalues of linear
systems (ref. 3). The Power method is relatively easy to program and is applicable to
nonsymmetric matrices which have real eigenvalues. The method yields the largest
eigenvalue and the corresponding eigenvector. To apply the Power method, equation (B1)
is expressed in the following form:

x [w] = [KI'ſ Lºw} (B6)

where X = 1/N. The solution is obtained in an iterative fashion as follows:

- -1
^r, 1 (W) rºl = [K] [L]{W}r (B7)

where {W} r is normalized relative to its largest element. Thus

{w} = {w}r (B8)

^r

where

Ar = max(w.) (i = 1 (1) M × N.) (B9)

The iteration process stops when

^r 11 *r - e."
—H-H (B10)

where e' is usually taken 10-4 < e' = 10-6. This procedure converges rapidly as long
as the next largest eigenvalue is not close to X. The Power method runs into difficul

18
ties when the shear buckling load is sought because for this case the magnitude of the
two largest eigenvalues is the same. This difficulty is easily overcome by using very
small values for one or both of the normal loads in combination with the shear load as
is described in reference 3.
The buckling load is obtained from the relation

; - ^r 11

and

N =–1–
X.
= NCr (B11)
r:#1

Thus the smallest buckling load is obtained since ^r-1 is the largest eigenvalue of
equation (B6).
The buckling load of the plate then is determined by incrementing M and N in
equation (B1) and applying the Power method to compute N until the convergence
criterion

N(M + 1)(N + 1) & e." (B12)

N(M × N.)

is satisfied. The parameter e" is usually chosen as an acceptable percentage of N.

19
 APPENDIX C

INPUT DATA

Explanations

Explanations appear in parentheses leadered to corresponding card.

, * * * |* | * * * * * * Hus it is te in ". zo * * * * * * * * * *d ** * * * **, we * * * * *o a 4-4.5 as as at a ºak.9 so * *z as **** * * * * * •ºl. i

ez by ea tº 66iºn 68 tº 10 h ºr ** **, r* me ºr rs tº as

(CºMPOSITE IDENTIFIC I

Eſ: Ron/ALUMINUM AT, | | | |

F-G#. of ſoap conſ ºf- III.
9 : + 4 + + -ī- ------4 4-------4-- # *---4----4-4-4------4----4----------

L -- - - - - - + --- + +--- ----- ------- --------- +---4----

* -- - - - + --- - ------- -------- -------

- - - - - - + + - -

10
NUMBER of LošDIN ; conDIT ION SETS
+ +----4---- Hº

* * * * P x, – "º-1
- 1.0 - 1.0
l, .0
| . . .-.0 -103. . . .- 0.03. 1...O.
- - - - - -- - - +------

t—(oNE + +------ ---

-- - ------ ---

–( FL ---------- ... (ORIENTATION ANGLE)-

D23 D32 D33 ) (j.
24 O ... O 7 . 77 O .. O
O. O.

(continue WITH FLEXu F. OTHER ANGIES)

. . . . . . --------------
- - - -- --- - - - - - - - --- ------------ - + + +--------—— --—i-- ---

— l l
I
2 * * * º 1 * 9 to it º: tº, tº tº i7 tº
• to 2 × 2: 2-2s we ?, ?, ?, * *z as ** **, ** * * -o ** *2 = < * * * * * * * * º so sº sº ss sºlº ses, ss so solo ºr ex 6. e5 sº 67 as 69 no ºi nº 13 ** 75 he 77 he rs -
NASA-C-856 (REw 9-11,-59) ºt-stry

20
 Sample Input Data for 0.02-Inch-Thick Boron/Aluminum Plate

I
a y < * tº 20 21 22 tº ra zº, 26 ºn 28 29 x * 52 x3 + 35 ye ** ** 59 .0 as ºries al. 4.5 at 4.7 ± 8 |aw 50 *l ºr ºs sa * 56 ºr sº sº to ** **, *, *, * r *s tº 8.
*| * * * * * * * * * * * **** ** * **
I
+-------4----- +--- —----- --4----

/ALUMIN AT - 5

* - - - --------4---

9 5
+ + + -t- *—4---------- +------|--|-- ----------- +------------ - - - -- ------------ ------- +----------- + →--------- *

-4------ +--- + ----- * +-4----4--

-
------4--

-------------- +----|--

+ +--- -
O. +---Hº-º-º-º-º-º-º-º-

-------- +
1 .. O

+-4---- ------ - - H-i-t-t-i-

+-----4----- |-º- +-t:

24 49
---------

* + + ---

------4----------------- --4-- ----------|---, -, +------|--|--|------ +---4--------

* ------i-kº---i-4-4- * + + --- — —— ——— — — — — — — — — — — — — — — —— —— —— —

- * - -------------- -- --- º -- ---4- +–H–4–4–4.

----- + -4------ 4----- - - - - -- - ------------------- -----|--|--|-- —f --——

* * * - ---H Hº--------------4---4- 4 - + +-----4-----------|----4 +-- | -------- -----------

-- I - + -------- ---------4------- + --------4---- *-* +---- + -- - - ------------------

------------ - -----4-4 ------------4--- 4–- 4–4 -4------ +--- +

------ +----------- ---4--- --- -------

------- --- +--- ------

- * ------4-------- --- +--- - - +---- *-i- +--- *— — — — —

* * * * * :* : * s 9 to a als 14, is tº in is tº 20 21 ºr 2x *:: * * 28 zºo you ºl 32 35 54, 55 ya lyn a8 59 wo ki *** ****** -, -ºxos 52 ss sºjºs sº ºf 58 so so. 64 & 63 64, as eeler 68 tº no di zai 75 7. 1s 6 ºf 18 is 8

NASA-C-856 tREw 9-a-59) * *

21
 APPENDIX D

COMPUTER OUTPUT

C- - - - BUCKL ING OF AN IS C T RO p IC PLATES

CCM M C N / E I GVF C/ W (81 )
CCM MCN / A R F I Ll/ C (3, 3)/ ARF IL 2/RHO ( 3 ) / ARRAY/X ( 81, 81 )
REAL N l ; N 2, LAMB2
D IM ENS I CN V ( 10, 1 O ) , T ( 10 ) , RST (20, 3 } , T IT LE ( 3 )
D A T A P I / 2. 14 1592 7/
MCC E = 1
R E A D ( 5, 22 ) TITLE
REAC (5 , 1 2) NP, MU
READ ( 5 , 13 ). A , B
RE A C ( 5 , 14 ) ( ( R S T ( I J), J = 1, 3 ) , I = 1 , NP)
R E A C ( 5 , 25) ( ( D ( I J ) , J- 1 - 3) , I = 1 , 3 ) , THE T A
DC l l I I = 1 , NP
WR IT E ( 6, 15)
DC 2 J- 1 - 3
RHC (J) = R S T ( II , J )
N 1 = O. C.
M= 1
CALL T I ML FT ( Q1)
M M = M + M.
CALL MFI L L 1 ( M, A, B)
CALL M I NVA ( MM, K S I G )
IF ( KS I G. E. Q. 2) GO TO 10
CALL MFI L L 2 ( M, A, B)
CAL L E I GF IN ( MM, LAMB2)
IF ( K S I G. E. C. 3) I FL AG = 1
N2 = 1 . O/L A. MB2
EPS = APS ( ( N2–N 1) / N2)

IF ( M. EQ - MU) GO TO 4
SPS= EPS
M= M + 1
N1 = N 2
GC TC 3
WR ITE ( 6, 16 )
I F ( I FLAG - E Q. 1) WRITE (6 , 17)
CALL TIM L FT ( Q2 )
I Q = ( Q 1-Q 2 ) / 60.
WR IT E (6 , 24 ) TITLE, THE T A
WRITE ( 6, 18 ) M., SPS, EPS, I Q
VM A X = 0 . O
DC 7 I = 1 , 10
DC 7 J– 1 , 10
V( I J }=0.O
DC 6 K = 1, M
DC 6 L = 1, M
I J- ( K-1 ) + M + L
V ( I , J } =V ( I J } +w ( I J) tº S IN ( FLO AT ( K & I ) . P I / 10 - O ) & S IN ( FLO AT ( L & J ) + PI / 1 0. )
VM A X = A M AX 1 (VMAX, A BS ( V ( I , J ) ))
DC 8 I = 1, 1 O
DC 8 J- 1 , 10

22
 V ( I v J } = V ( I v J } /VMAX
WRITE (6 - 19 ) A, B , N2 , MODE, RHO, D
DC 9 I = 1, 10
J– 1C– I + 1
T ( I ) = C. O
WRITE (6 - 20 ) J., T ( I ) , ( V ( K, J ) , K= 1, 10 )
I =0
XI = 0 . C
WRITE (6 - 20 ) I X I V ( T (J), J- 1 , 10 )
WRITE ( 6, 21 )
GC T C 11
10 WR IT E (6 : 22)
ll CCNT INUE
GC TC 1

12 FC R M AT (2 I2)
13 FC R M AT ( 2 F 10. 4.)
14 FC R M AT (9 F 8. 3.)
15 FCR M AT ( 1 H 1 )
16 FCRM AT ( 2 3 H MAXIMUM INDEX AT TA IN ED )
17 FC R MAY (41 H I LL - CONDIT IONED MATR IX - RESULT IN DOUBT )
FC R M AT ( 5 X • 2 HM= , I l ; 5X , 2 OHPREVIOUS REL. ERROR = , 1 DE 1 C. 3., 5 x , 1 6HL AS T R
1E L. ERROR = , 1 PE10.3, 5X,33H TIME REQUIRED FOR THIS CASE WAS , I 3, 8H S
2 ECCNCS)
FC R M AT ( / / / 7X, 1H A , 13X •.1 HB , 11 X, 3 HNCR , 11 X , 4 HMOD F , 9X, 5 HRHO-X, 9X, 5 HRHO
1- Y, 9X, 6 HRHC-XY// 3 (2X, 1 P E 10. 3, 2X ) , 6.x, I 1 , 7.x, 3 (2x , 1 PE 10.3, 2X) / / /37H T
2HE C– AR R A Y COLUMNW ISE IS AS FOLLOWS / / 9 (2X , 1 P E 10. 3, 2X) / / / / 50 X , 13 H BU
3CK L E C SHA PE // 6X, 1HY)
20 FORMAT ( / / 5X, I2, 1 1 ( 5X, F5. 2.) )
2I FC R M AT ( / / 15X , 1 HO 2.9X , 1H 1, 9X, 1H2, 9X, 1H3, 9X, 1H4, 9X, 1H 5, 9X, 1H6 , 9X , 1 H7
1, 9 X , 1 H8, S X , 1H 9, 8 x , 2H 10, 6 X , 1 HX)
22 FCRM AT ( 27H M ATP IX SINGULAR - PROCESS TERM IN ATFD )
FC R M AT ( 3 A 6 )
24 FC RM AT ( /20 H COMPOS IT E SYSTEM = , 3A6, 12H AT THE TA = , F 5. 1 )
25 FCRM AT ( 1 OF 8. 3.)
£NC

SUBRCUT IN E M FILL 1 ( MM, A, B)

CCM MCN / A R F IL 1/ C (3, 3)
CCM MCN / A R RAY / x (8.1, 3 l )
INT FC ER p , C
DAT A P I / 2. 141592.7/
DC 3 wal, M M
DC 3 N= 1, M M
DC 3 p = 1, M M
DC 3 Q = 1 , M. M.
I = (M-1) + M M + N
= (P - 1) + M M + C
IF ( I. NE. J.) GO TO 1
I F (P. NE. M. C.R.Q. NE. N.) Gn TD 2
* { I J) = P I + + 4 + (P + FLſ, AT (M**4 ) + C ( 1, 1) / A * * 3+2.0% Fu O AT (w = M*N*N) + ( D ( 1, 2)
*:::::tº,” (“"). A-r narrº... •o 2,2) re...”..o
2. ſ 3

23
1 IF (MCſ) ( ( M + p \ w, (N + Q ), 2). FO. O ) GO TO 2
x ( I , J } = -8 . O & PI # * 2 * D ( 1, 3 } + F LOAT ( M*N*P* Q & ( 2 * M* M/ (P + P-M+ M) + 1) / (Q+0–Nº.
1N) ) / A * * 2-8. O & ſº I + + 2* D (2, 3) * FLſ) AT ( M*N*P* Q+ (2 *N*N/ ( Q+ Q-N*N) + 1) / ( P × P – M
2 + M) ) / R x, * 2
GC TC 3
X( I J ) = 0.0
: CCNT INUE
RETURN
ENC

SUB R CUT IN E M I NVA ( M., KS IG)

C. . . . . M A T R IX I N V E R S Iſ)"N BY GAUSS-JORDAN FL I M IN AT ION
CCM M CN / A R R AY / X ( 8 1, 81 )
D IM E N S I CN Y ( 8 1, 8 l ) , Z (2,81, 81 ), K ( 81 )
E QU I V AL E N C E ( Y, Z )
DCU P L E PR FC ISION A., R, PROD, Y
DATA C E L T , F PS, Lſ) CPS / 0.0001, l. OE-8, 2/
KS I G = C
I F ( M. N.E. 1 ) GO T C 3
PRCT = x ( 1, 1)
IF ( p ROC. NE. O. O.) GO TO 2
l KS I C = K S I C + 2
RETURN
2 X ( 1 , 1 ) = 1.0 / PROD
RETURN
3 PR C C = 1... O
MM=M- 1
DC 4 I = 1, M
K( [ ] = I
DC 4 J = 1, M.
4. Y ( I J}=X ( I J)
C. . . . . PEG IN R Y F IND ING LARGEST PIVOT AL E L E MENT
DC 9 I = 1, M
A=0 . O
DC 5 J– I, M
I F (A PS (Y ( J , 1)). L. E . A ) GO TO 5
A = A P S (Y ( J , 1))
L=J
5 CCN T INUE
I F ( A. E. Q . O. O.) GO Tſh 1
C. . . . . RE ARRANGE R CWS AND ſl RDER ARR AY
N= K ( L )
K ( L ) =K ( [ ]
K ( I ) =N
DC 6 J– 1 , M
A= Y ( I J )
Y ( I J }=Y ( L , J )
6 Y( L , J} =A
C. . . . . R F D U C E PIV C T A L R CW
A=Y ( I , 1 )
nC 7 J– 1, M M
7 Y ( ( , J ) = Y ( [ . J4 1) / A
Y ( I M ) = 1 ... O / A

24
 ..., - a2a338-- 38 w V NI ONI SM08
JJ 6 =l l w"
I 3 °53°1) 1 ( 09 1 0 6
=w A ) T I* (
OO 8 N = w°l w
)A T N' | = A ) (1+N'T - V At ) N*I (
1 + ) SBW A ) T N* ( ( * 11 * ) S8w ) )A “T +N I ( ( + da (S ( A 1) Nº ( = 0 ° 0
6. A N'l ( = v- * A I twº
3• O No N11 3ſ
. . . - 3.18ww80SNn 1 3AN & Q3.1 W W 81 XI
Od I & I = “I a
1 = x) (1) 53° ° 1 ( 09 01 £1
IO JQ Ol =ſ I * w
I 3 X) ſ.) ( O3" " I ( 09 01 II
I I OS) JI I
JQ 2I T = I * w
V As !“l] !
I z . \, {1 = A (f'i)
. \, | f | = w
I E . x=(t) I (

5- -* NL1,w}} 30
. NIV185. > ti 8 J 8 IVW & I X
DQ 81 N = I * 00T Sd

I * jº ºr “i-º
z 2 I “1” (ſ x* } I • (T
# 1) − H tº º,
wins;
) “2
º, XV I ) “1S31 S3W ) 8 ('[
JQ

# 3} I 9 Tº = “I w
...!? 2 > I " I " , ( + z *z) “T ſ. (
l§ "ºI ;* i: s= s
r 1| =* ZT 3‘I,
G" fºr
liº 1 (; w:
09 01 I6
I N
-5i. =& 3.I 3) + ©
Sl ... 81, XV
*

QQ ==$$. *i. : NI TV I AN Słł3 3

Q& JQ §: R = “I w
K T- = I * w
-???; trººz
JWs

\S S
º. RS. a as w – I TT 2 ) w “w °W 9 (
y Sºº / W & 3 l l /2 OHö ) & (
| " SI N. / W & V8 A / X ). 8 I * 8 (l
St.
l
SX iss,
Th
S9 l ) 8 I 8* I ( “ O ) 8 “I 8 I (

§ , , *śs **
<3 I / ? " I l', 1269 /
G3
 DC 3 M = 1, M M
DC 2 N= 1, M M
DC 3 P = 1, MM
CC 3 Q = 1, M M
I = ( M-1 ) & M M +N
J = ( p – 1) + M M + Q
IF ( I. NE. J.) GO T C 1
I F ( p , N.F. M. C. P. Q. N F = N ) GO TO 2
L ( I , J } = - F I + + 2 + ( B + FLOAT ( M + M) & RHD (1) / A+ A* FLOAT (N*N) *RHO (2) / B ) / 4.0
GC TC 3
I F ( M CD ( ( M4 P) * (N +Q ), 2). EQ. 0 ) GO TO 2
L ( I , J } = -R HC (3) + F LOAT (8*M*N*p *Q/ ( ( P + P-M*M) + ( Q+ Q-N*N) ))
GC IC 3
L ( I v J } = 0. 0
; CCNT INUE
NN = M M + MM
DC 5 I = 1, NN
DC 5 J– 1 , N N
SUM = C. O
DC 4 I J = 1 , NN
SUM = SU M + X ( I I J) & L ( I J , J )
: C ( I , J ) = SU M
DC 6 I = 1, NN
DC 6 J– l ; N N
X. ( I J ) = C ( I J }
RETURN
END

.
SUBRCUT IN E E I GF IN ( M.M., L AMB2 )
REAL LAMP 1, L AMB2
CCM M C N / E I G VE C/ W ( 8 1 )
CCM MCN / A R RAY / X (81, 8 l)
D IM ENS I ON V (81)
L AMP l =0 . C
W ( 1) = 0 - 5
I F ( M M . E.C. 1) GO TO 2
DC l I = 2, M M
W ( I ) =C - 5
D.C. 4 I = 1, M M
SU M = C - O
DC 3 J– l ; N M
SUM = SU M+ X ( I J) *w (J)
V ( I ) = SUM
LA M P 2 = A BS ( V ( 1) )
I F ( M. M. E. Q , 1) GO TO 6
DC 5 L = 2, M M
LAM P2 = AM Axl ( L AMB 2, ARS (V ( I ) ))
DC 7 I = 1, M M
; W ( [ ] = V ( I ) / L. A MR2
IF ( A P S ( ( L A M P 2–L AM R 1)/L AMP 2) . L. F. l. () E-4 ) RE TURN
L. A M P 1 = L AM B2
GC T C 2
EN C

26
 27
8l 9*7£ 1D
• O •O • O• O •O • O• O• O •O • O• O
00° O 1C * 0 - • I • C-1 l º 0 9, 1 *0- 90 º 0-60 * 0» Z * O I© º O 2 Z * O• 0
00 * 0 •I •O 1 Z * O-€ £ º O 8Z *O– I Į * O –8 1 * 09 ſy º O 69 ° O z • • O• O
00 * 0 0，2 * 0– 1 € º O-9. *) * 0– 6 % *0- • I • O-9 2 * 09,9 ° 0 I8*0 1 g * 0• O
00° 0 •Z *O *>.*; * O-* g * 0– 9 ºy º O 9 I º O –Iº º OL.L * 0 96 * 0 9.9 ° O• O
00° O 92 * 0– 1+7 ° C -9 g º O Lºy º 0 9 I * O –º £ º O28 º O OO * I 69 ° O• O
00° 0 9Z *O– g * * O-* g * O gº º O • I • O-€ £ º O6.1 ° O 96 * 0 9,9 ° 0• 0
OO • O ZZ * O 6 € º O-9 ºy º O 8% º0- 1. I º O-6.2 ° O19 ° O 08 º Q 9,9 ° 0• O
00° 0 9l ºO 8 Z * O-*> £ º O 1. Z * O 80 ° 0 -ZZ * O09 ° 0 8.9 ° 0 6 € º O• 0
OOº O 6C *O 9. I º O-8 I • O •I •O �O * O –Z I • O92 • O I£ºO O Z * O• O
00° 0 OC * 0 OO ^ OO C ° O 00 ^ O OO ^ O00 º O-OO • O OO • O 00 * 0 -• 0 Ol
BdVHS 03 TX2008
00+3002 º 8 00+3 C 1 € º la C0+3018 º 1 00+30.41% º 1 lO + B 9, º 8 * I00+306 ſy º º 00+3018 º 1 00+ 3O6 → * */ 1 0 + 3 6 12 * Z
SMOT TO 3 S v S | BS I MN alſ 10 0 A w 8 8V - J BH.1
•O •O OO + 3000 * I -1 00+ B 1 I l º 8 I O + B（OOO * I I O + 3O 3 D * 2
A-0H8 x-0H8300 w 80N 8 v
AX-OH：
9. № N
S CN003S l l SW M 3 SV 0 SI H1 802 0.3 8 1 003 o Bw. I 1 ș0-3996 º € = 80\，\!} º T3 & 15 yı »O-3 » g gº l = 80 883 * n 3 × S(\O I N 3 8 d.
0 * 09. = W 13 H 1 1 W wnN1 w01 w / N0 80 Q = w 31 SAS 3 1 1 S J dwÚ0
0BN 1 v 1. La x 3 J N I w^* 1x vºw
 … - - - - ----
9*»€.2
•O •O •O •O •O • 0• O• 0• O •0
00° O - 6 C° 0 81*O 9. Z * 0 6.2 * 0 I £ º OO £ º O92 º 06. I º O O1 *O •0
OO • O 8 1 *O 1* *O 9. 9. *0 69 ° 019 ° O6 * * 09€ * 0 61 *0 •O
00 ° C - *2 *O § 9º0 1 8 • O8.1 * 099 º O6 ºy º O 92 º 0 •O
00° 0 6.2 ° O 9,9 ° C 9.1 ° O 0,6 * 0 96 * 0I 6 * 081* 019 ° O •O
00° 0 1€*0 89, º 0 l 8* O 96 * 0 OO • I96 * 01 8 • O69 ° 0 •O
00 * 0 - 6.2 ° O 9g ºQ 1 1 *0 0,6 * 0 96 * 006 * 09.1 ° O9 g * 0 *U
00 * 0 - 92 *0 8 ** * 0 9 9 *O L.L * 0 18 ° O9.1 ° 09,9 ° 01+7 ° 0
00° 0 - 8 1 *0 9% * 0 8•• O 99, º 0 69 ° 09 g * 01+7 ° O*78. ^ O
00° 0 - O1 *0 8IºC 9; 2 * 0 0& *0 1 & * 06.2 ° O• 2 • O8 l º O
00 * 0 OC * Q 00 * 0– OC *O 00 ^0- 00 ^0-00 * 0–00 * O –OO * 0– •O 01
B dVHS Q 3 T × 01 || 8
00 + 3002 * 8 00 + 3C 1 & * \, 00+ B 0.16 * I 00+30.1% º la 1 0 + 39 478 º 100+ 3 O 6 ** * */00+ B O 18 º l OO+ 3O6 +7 * #7 1 0 + 36 I 2 * 2
S MOTTUB Sv SI 3 S 1 M N w ſł 10 O A w 8 d.w - J 3 H1
• 0 00+3000 * I - • OI00 + 398.6 ° 2 1 0 + 3000 * I 1 0 + 30 O D * 2
A X-U H \} A -OH \} x-0H™）300 w8 ON
S CINC 03 S 9 S vM 3SW D S1H1 803 QJ B \; I OČ) B & Bw I 1 90 - 3 ， 99, º € = 80 8 8 3 º T 3 J 1 S W T*70-38 6 2 º 8 = 80 883 * T3 & S {\J I N 3 8 d. €= W oO
0 * 0€ = V1 BH1 1 V w ßN I w QT w / NJ 808 w 3 1 SA S 3 1 1 S D dw 0 0 cN
 29
XOl68l99»€.21D
• O• 0• C• O• O• O• O• O• O• O• 00
0 0 * 0–6 C * 01 l º 0£ 2 * 08 2 * 01 € * 0O £ º O12 * 002 * Ol l º O• 01
00 * 0 -9 l º 02 £ º O9 % * 0*p g * 06 g * 08.9 ° 0i 9 * 08 € * 002 * 0* 02
OO * O –€ Z * O*> *, * 0Z 9 • O* 1 * 0I 8 ° O6.1 ° 069 ° O29 ° 08 Z * 0• O€.
00 * 0 -1 2 ° CZ 9. * 0£ 1 * 08.8 ° 096 * 0£6 * 0I 8 ° OO 9 * 02 € * 0• O}}
00 * 0 -6.2 * 09,9 ° CL 1 * 0£ 6 * 000 ^ I86 * 09 8 ° 0€ 9 ° O€ £ * 0• O9
00 º 0 -1 2 ° C,€ 9. * 0* 1 * O8.8 ° 096 * 0£6 * 00,8 ° 069 ° 01 & * O• O9
00° 0-ș Z * O9. * * 0£ 9 * 09.L * 0l 8 * 08.1 ° 08 9 ° O09, º 01. Z * O• OL
00° 0-All º O€ £ º O9 % • O9,9 ° 069 ° 0Lg º O6 • • O9€ º O6 1 * 0• O8
00° 0-6 C° 01 1 * 0* 2 * 06 Z * OI £ º OO 2 * 09.2 ° O6 1 * 0O I º O• 06
00 * 0OC º Q-OO * 0–O C º O-OO * O -00 º 0 -OO • O-OO • Q-OOº O-00 ^O-• O0 1
A
3 d'W HS Q 3 TX1009
OO + 3OC Z * 800 + 3C 1€ º 10C + B O 18 * I00+30.1% º la10 + 39 478 º I00+30,6 * * *OO + 3018 º 100+ B 06 * * */1 0 + 361 Z * 2
SMOT TU 3 SW S I 3S1 MNaſh 100 A w 88 w - J BH1
• oOO + 3000 * I -00+3000 º l -100+ 36.8 € º ZI O+ B（OOO * I1 0 + 30 00 * 2
Ax-UH adA-0H &x -OH ！300 w8 ON€.v
uB º T3 8 1 S w Tº 0-386 Z * 8 = 80 883 º nº » Sn0 1 A 3*
Q S I HI 803 03 J 1003 & Bw 1190-39Z L * g = 8|-* 0€ \} da€ = w
-·08.
S C NU 03 S 9 Sv M 3= Sw
SV 00 = w + 3H u u vw[\^ l wn i'w / NJ 800 = w 3 u SA S 3 1 1 S D d woo
 - - - - - -… — • • • • • •• • • • • ••-- --------·
- - - - - --- - - - - -• •• • • • •--- -• • •
68199•%2u.0
•O • O• O• O• O• O• O• O• O• O• 00
00° O 6 CºO -I Z * O-1 & * O-1 & * O-8 l º 0 -£0 * 0£ Z * OI £ º OZ Z * O• OI
00° 0 O 2 *0-£ * * 0 -O 9 * O-19, º 0 -6 Z * O -€ I • O1+7 ° 069 ° 0O 7 * O• O2
00° 0 Z & * 0–* 9 * Q-£ 8 * O-21 º 0 -6 Z * O -6.2 * 01 1 * 008 • O1 g * 0• O€.
00° 0 € £ º 0-l 8 * 0–16 * 0–9.1 ° O-6 I ° Q -1 * * 00,6 * 0I 6 * 09 g * 0• O�
CO º 0 1 g º O-0,6 * 0–OC * I -8 9 *0-£0 * 0 -£ 9 ° O66 * 026 * 0€ 9. * 0• 09
00 * 0 € 9. * C –6 8 ° C-I 6 * 0–€ 9. * 0–Z I º O21 º 016 * 0£ 8 ° 09, 47 ° O• CQ
00 ^ O 09. * 0 -6.1 * 0–£ 1 * 0–*> £ * 0–€ 2 * 00.1 * 0º 8 ° O9.9 ° O*> £ º O• 0L
00 * 0 6 % * 0–8 9 * 0–O 9 * 0–1. I º O-9 2 * 09,9 ° OI 9 * 0gº º OI Z * O• O8
00 * 0 Z Z * O-1 & * O-• 2 • O-90 º 0 -9 I • O0£ * 0Z £ º OZZ * OO 1 * 0* 06
00° 0 0C * 000 * 00 C * 000 * 000 º 0 -00 * 0 -00 * 0 -00 º 0 -00 º 0 -• O0 1
A
3 d. V HS Q 3 m X 0018
00 + 30 02 * 8 00+ BC 1 & * I00+3018 * I00 + 30.1 € º la1 0 + 39 478 * I00 + 30,6 % * */0 0 + 30 18 º la00+ 2O6 +7 º 7l 0 + B 6 I 2 * 2
SMO T TO B Sv SI 3S 1 MN×0 10 0 A v 8 &. w - J BHI
CO + 3 000 º l -• 000 +-3000 * I -l0 0 + + 9 8 I º 9\ 0 + 300 C * I[ 0 + 3 0 0 ) * &
A X-UH ajA-0HTM）X-UH\}B 0 0 w8 ON8V
S CINC 0:3 S l l SvM 3 SV 0 S I HI HU 3 0 3 & 1 (\O B & Bºw I 1€0-30 M L * I = 808 8 3 * T B & 1 SW T20-30 I 6 * I = 8 U 8 8 3 * T B 8 S (10 I A 3 8 d.9 = w
0 * 0% = w 1 BH1 1 ww[\N I wºn TV / NO 800 = n. 3 I S AS 31 I S J dw 0 0
O
0BN I W 1.J. W X3 JN I wºn w IX w. w cro
 31
• O* 0• 0• O• O• O• O• O• O• 0• O0
00 * 0 -1 C ° 0* I * 0I Z * O1 2 º 00 £ º QI £ º O8 & * 0I Z * 02 I • O• 0I
O O º 0 -• I • O8 2 * 01 * * O29 ° O8.9 ° 089 * 029. * Jo £ * 0I Z * O• C2
0 0 * 0 -O Z * O047 ° 06 g * 0º 1 * 0O 8 * 06.1 ° 069 ° O29 ° O8 Z * O• O£
O C * O –92 • Ol 9. * O£ 1 * 08.8 ° 096 * 026 º 06.L * 019 ° O0 £ º 0• 0»
00° 0 -6.2 ° C19. ^ OO 8 * 096 * 00 0 * I96 * 00,6 * 019 ° O6.2 * 0• 09
00 º 0 -0 £ º 019 ° 06 1 * 0Z 6 * 096 * 068 ° 0£ 1 * 0l 9 * 09.2 ° O• O9
OO * O –12 * 0i 9 * 06 9° 06.L * 0I 8 ° O* 1 * OO 9 * 0I • • OO 2 * 0• UL
00 º 0 -I Z * O69 ° 02 g * 089, º 08 9 *0€9 ° OZº º O8.2 ° O• I • O• O8
00° 0 -2 I º QI Z * O8 Z * O1 & * OI £ º O1 2 ° OI Z * 0• I • O1 0 * 0• O6
00 * 0O C * O –00° 0-O C º O-00 º 0-00 º 0 -00 * 0–00 º 0 -00 º 0 -OO º 0 -• 00 1
A
B dV HS 03 TX209
00 + B002 º 800+ B C 1 & * ICO + B 01.8 * I00+301€ º l10 + 39 478 ºli00+30,6 * * */00+ B 0.18 ° U00 + 306 * * */1 0 + 3 6 1 2 • Z
SM0T l03 SW S I 3S I MN alſ 100 A w 8 ww. -J EH1
OO + 3 000 * I -00+3000 º l -• oI00+ Bº 98 º 2l'O+ 3000 º 11 0 + 3 0 0 0 * ?
AX-OH\\A - OH ！x -OH ！300 w► ON8v
S C NU 03 S Z I S v M 3 SV 0 S 1 H 1 8030 3 8 1 003 & 3 W l 1• O -3.8 € 1 º Z = 80 ± 83·ıą1S w T€ 0-3 € 09 º l = 80 ± 8.3 ° n.B » Sno I A 38 a9 = N
0 * 0€ = v 13 H I 1
vw[\N l w[\\w / NJ 800 = w + 1 S AS 3 1 1 S J &au 3
03N I W I J w xB JN 1 wºn w 1 x w w
 … - …… --
xO l. 99.*7£
•O •O •0 •0 • 0• O• O• O •0 •O •0
00 º 0 - 1C * 0 •I • O I Z*O 1.2 * 0O £ º O1 2 ° O12 º 0 IZ *O ZI•O
OOº O * 1 *0 82 * 0 Z• •O Z g * 08 9 * 089 * 029 ° 0 6€ * 0 IZ*O • O
OOº O OZ *O I•• O 6g*0 £ 1 * 008° 06.1 ° O69 ° 0 29 ° O 82 *0 •O
00° 0 - 92 *0 09 * 0 Z 1*0 8.8 ° 096 * 026 ° O6.1 ° 0 89. *0 l£ºO •O

cº,
OOº O 6 2 *0 9g * 0 * 6 * 0OO * 196 * 0I 8 * O 8g * 0
OO º 0 6.2 * 0 99 ° C 16 * 096 * 006 * 09.L * 0 € 9. * O
00 * 0 - 92 * 0 09. ^ 0, 8.L * 0l 8 * 09. 1 * 029 ° O £7 *0
00° U - O2 * 0 8£ * 0 O9*0 19 ° 069 ° 0ºg º O€ £ º O 62 * 0
00 * 0 - ll ºO OZ * J 12* 0 0 & * 0I £ º O8.2 ° O2.2 * 0 9. I º O • O
00 * 0 00 * 0 - 00 * 0– OC *O OO • O-00 * 0 -OO • O-00 * 0 -00 * 0- -• 0 OI
3 d. V HS Q3 T × 000
00 + 30 O Z * 8 CO + B C 1 € * I C0 + 3 O 18 * I 00+301 € * I1 0 + 39, * 8 * l.00+306 7 *** 0 0 + 30.18 ° l. 00+ 306 47 * #7 1 0 + 3,6 l 2 * 2
S MOT TO-3 Sv S1 B S I MN NO 10 O A w 8 8 w - J B HI
CO + B C 00 * I - 00+ 3 000 * l-OO + 3000 * I -T 00+ B 02 @ * 21 0 + 3 000 * ltI O + 300 ) * 2
Ax-UH ae A -OH \}X-OH \;300 w \J ON8w
S C N C C BS I I SVM BSv 0 S 1 H1 dC → Q 3 & I Qð B & Bw I 1 *> 0-3 l º £ * I = \JU U 8 B *TB & 1Sv T € 0-3192 * I = 80 öð 3 * TB 8 S 0 0 1 A 3 8 d.9 = w
0 * 0% = W 13 H I 1V.WON 1 w QT w / NJ 800 = w 3 1 SA S 31 I S DdwJ 0
cN
Q 3N I W 1 1 7 XB J N i w ^ w I XV w cro
 33
• O• C• O• O• O• O• O• O• 0• 0• 00
00 ° C -1 C ° 0• I • OI Z * O1. Z * 00 £ º Uu 2 * 08 Z * O2.2 * 02 į * 0• OI
00 ° C -• I • O12 * 0į • • O29 ° 08 9 * 089 * 029 ° 0O ſy º O2.2 * O• O2
00° 0 -O 2 * 06 % * 08 g * 0Z 1 º 008° 008 ° 0l l º 0€9 ° 06 Z * O• O€
00 º 0 -9 2 * 06 ºy º O0.1 * 018 ° 096 * 0£6 * 028 º QO 9 * OZ £ º O• 0�
00° C –1. Z * O*p g * 09.1 ° O26 * 0OO • įL6 * 0º 8 ° OI 9 ° O2 € * 0• O9
00° 0-8 2 * 0º 9 * 09, 1 º 068 ° 096 * 026 • O81* 099, º 06 2 * 0• O9
00° 0 -g. Z * 0147 ° C9, 9 ° OAL L * 0I 8 ° OALL º Oș 9 • Ogº? º O£ 2 * 0• Ol
00 º Q-6. I º O9 € º 08 %, º Q19 ° 069 ° O9,9 ° 09 × ° OZº º O9 į * 0• 08
00° 0-O 1 * O6 I • O92* 0O £ º OI £ º O6.2 ° Oș Z * O9 I • O80 ° O• 06
OO • OOO • O-00° U-O Cº O-00 º 0-OO * O -OO • O-OO • O-OO • O-00 º 0-• O0 1
Å
BdVHS 03 TX1009
00 +3002 º 800+ BC 1， * 100+3018 º 1OO + 30,1 % * II O+ E gº 8 * IOO + B 06 ºy º */OO + 3018 º 1OO + 306 * * */I O + 36 I Z • Z
SMOTT03 SW SI 3 SI MN NO 10 0 A w 88 v -Q. BH1
10-3 000 º g-10-3000 º g-00+3000 * i -I00+396.8 ° €1 0 + B（OOO * II O + BOO D • Z
Ax -OH （d.A-0H8x -OH ！300W8 DN8v
S v T€ 0-3 2 00 * I
1
S CNU 03 S0
|-I Sv M a Sw.) S I H I 8U-4 0.3 × 1 mð 3 & Bw 1 1€0-36 l g * 2 = 808 à 3·ığș1.
* 0€ = w 1 BH || 1
= 80 × 8.3 ° na » Sno I a. 3 u.a.9 != 'w
ww^N i wſhīw / NO 800 = w B 1 SA S B I l s bawo o
O 3N Ì w I J w X3JN1 wn+ 1xww
 • O• O• O• O• O• O• O• O• O• O• OO
00° 0 -1 C ° O£ 1 ° U6 1 * 09; 2 * 06.2 ° OI £ º Q6.2 ° 0£ 2 ° O£ I º O• Ol
Q U * O -£ 1 º 09. Z * 01 € * 01+7 ° O9 g º O69 ° 09 g º O£ 47 ° O• 2 • O• CZ
00 º 0 -1. I º O* º * Ql 9 * 099 º OL.L * 01 8 • O9.1* 069 ° O€ £ º 0• CĶ
00 * 0–I Z • O1 & * 0O 9 * O1.L * 00,6 * 096 * 068 ° 069 ° 08 £ º O• 0*7
00 * 0 -Z Z * O£ º * U* 9 * O28 *096 * 0OO * I€ 6 * 0Z.1 ° OO 7 • O• O9
00 * 0 -I Z * OI 47 ° 0l 9 * 08.1 ° 0I 6 * 096 * 088 ° 08 9 ° O1 € * 0• O9
00 º 0 -8 1 * 09 %, º CZ 9. * 019 * 0L.L * 0I 8 • O9, 1 º 089, º 0Z £ º O• Ol
OO * O -£ 1 * 09.2 ° C8 € * 06 7 * 09 g º O69 ° Oºg º OZ » º O€ Z * O• 08
00 * 0 -1 C ° O• I • OO 2 * O9,2 * 0Og º OI £ º O8 Z * O22 ° O2 1 * 0• O6
00° OOC • O-OQ * C –O Q * O-00 * 0–00 * 0–OO • O-00 º 0 -00° 0-00 * 0 -• 00 I
Å
3 d. VHS Q 3 T × 00 €
00 + 3 0 0 2 * 80C + B C L { * ICO + B 0.18 ° l.00+301 € * I10 + 39 478 º 100+306 47 * #700+30 18 º la00 + 306 47 º 71 0 + 36 I 2 * 2
SMOTTO-, Sw SI 3 S 1 M N w ſì il D :) A w Ł ł w - J + H I
* O-10-3 000 º 9 -00+30 00 * I -I0 0 + B2 86 º €Į O + 3 OO O * I1 0 + 3000 º Z
AX-0H8A -OH \;X -OH \;B 0 0 W80N8W
S C NU CBS g i Sv. M B S w Q S I H 1 803 QB & l [lō3 & B w 1 190 - B 88 l º 9 = }}U \j \j B º T3 & 1 S W T*70 - 386 2 º 8 = 80 8 8 3 * T B 8 S (TO I ^ B 8 d.€ = N<rſ
0 * 0€ = w 1 BH1 1 Ww[] N I Wºn TV /N0}}08 = w + 1 S AS 3 1 I S Jd w00 co
 9*7£2lD

35
•O *U •O •O •O • O• O• O• O• 0• 0
00° 0 £C * 0 - 80 ° 0 O IºO 470 º O O I º O»Z * OI £ º OLZ ° Us l ° 0• 0
00° 0 1C °0- gI ºO 9, 1 *0- 1 0 * 0– 9,2 ° 009 * 069 ° O9 × ° Og 2 * 0• 0
O0 ° 0 II •O 61 *O * I *O Ol. * O 8 ſy º 09.1 ° O8.1 ° C)19 ° 012 ° O• C
00° 0 * I *C– 6 1 * 0– 9C ºO 62º0 l l º O• 6 • O9.8 ° 0ºg º O€ Z * O• 0
00 * 0– º l º 0– * I * 0– 6 C° 0 6 ºy º O 8.8 ° 0OO * IO 8 * 0ș ș• Og ( * 0• O
00 º Q 0 1 *0- 9. O º 0 +, Z * O •9 °O £6 * 026 * 0ș9 ° O6.2 ° OLO * 0• O
00 º 0 - 9C ºO 470 ° 0 *> £ º O 19 ° O £8 ° 02.1 ° O£<; • Ogl º 01 0 * 0• 0
00° 0 - 20 ° C - 60 * 0 € £º O 99 º O I 9 ° O1ły º O47 Z * O9,0 ° 0Z O º 0 -• O
00° 0 - OC * 0 80 ° O I Z*O Z 9. * OZZ • O60 ° OIO * OI O * O –• O
00° 0 OC °O O Qº Q OC^0 00 º O-00 º O-OO º 0 -00° 000° O• C OI
3 d'WHS 03 T X O(\ 9
00 + 3002 º 8 00+301 € º la 00+3018 º la 00+30.1% º la 10 + 39 478 º 100+ B O 6 * * */00+3018 º 1OO+ 306 ºy º º 1 0+ 36 1 2 º 2
S M 0 Tl TU-J S W S I 3 SI MN alſ 100 A V 8 8V - 0 3 H 1
OO + 3 OQO * I - 20-3000 * *?-- 20-3000 * * -TI O + 39 O Z * II O + 3000 º I 1 0+ 3 0 0 ) * 2
Ax-OH （d A -OH & x -OH »300 w8 9N8 y
S CINC O 3S I I S v M 3S v J S 1 H 1 803 J 3 8 1 003 & 3 & l \ ç 0-380 £ º Z = 808 83 ºn 3 8 1Sv T20+3 & b Z* \ = 8088.3 ° na » SnJ I A 3 （ ） 9= N
0 *0€ = v.1 → H i u vw f（\MN i wſh Tv /N0 M08
* w 3 \ S AS 3 1 I S D dwJ 2
Q 3N 1 w I J w X3J N1 × n ~ I xvw
 REFERENCES

Ambartsumyan, S. A. : Theory of Anisotropic Plates. Technomic Publishing Co.,

1970.

Ashton, J. E.; and Whitney, J. M.: Theory of Laminated Plates. Technomic

Publishing Co., 1970.
Chamis, Christos C. : Buckling of Anisotropic Composite Plates. J. Structural
Div., ASCE, vol. 95, no. ST10, Oct. 1969, pp. 2119–2139.
Kicher, T. P. ; and Mandell, J. F. : A Study of the Buckling of Laminated Compos
ite Plates. AIAA. J., vol. 9, no. 4, Apr. 1971, pp. 605–613.
Lekhnitskii, S. G. : Anisotropic Plates. Gordon and Beach, Science Publ., 1968.
. Wang, James T. -S. : On the Solution of Plates of Composite Materials. J. Com
posite Mat. , vol. 3, July 1969, pp. 590–592.
Hsu, T. M.: Buckling of Anisotropic Plates, Discussion J. Structural Div., Am.
Soc. Civil Eng., vol. 96, no. ST7, July 1970, p. 1604.
. Chamis, C. C. : Buckling of Anisotropic Plates, Closure and Errata. J. Structural
Div., Am. Soc. Civil Eng., vol. 97, no. ST3, Mar. 1971, p. 960.
Chamis, Christos C. : Computer Code for the Analysis of Multilayered Fiber
Composites - Users Manual. NASA TN D-7013, 1971.
10. Chamis, Christos C. : Characterization and Design Mechanics for Fiber
Reinforced Metals. NASA TN D-5784, 1970.

36 NASA-Langley, 1971 — 32 E-6455

º:
N AT I ONAL AERO NAUTICS AND SPACE A D M ISTRATION
WASHINGTON, D.C. 20546
Post AGE AND FEEs PAid
NATIONAL AERONAUTICs And
OFFICIAL BUSINESS space Ad Minist RATION
PENALTY FOR PR v ATE use $300
FIRST CLASS MAIL

0 1 1 002 C1 U 35 71 1203 SOO u 39H

UNIW OF ILLINOIS
LIBRARY
ATTN: DOCUMENT LIBRARIAN
URBAN A IL 6 1801

POSTMASTER . If Undeliverable
Postal Manual) Do(Section 15
Not Retu

"The aeronautical and ſpace activitieſ of the United States ſhall be

conducted ſo as to contribute . . . to the expansion of human knowl
edge of phenomena in the atmoſphere and ſpace. The Administration
shall provide for the wideſt practicable and appropriate disſemination
of information concerning its activities and the reſultſ thereof.”
— NATIONAL AERONAUTICS AND SPACE ACT OF 1958

NASA SCIENTIFIC AND TECHNICAL PUBLICATIONS

TECHNICAL REPORTS: Scientific and TECHNICAL TRANSLATIONS: Information

technical information considered important, published in a foreign language considered
complete, and a lasting contribution to existing to merit NASA distribution in English.
knowledge.
SPECIAL PUBLICATIONS: Information
TECHNICAL NOTES: Information less broad derived from or of value to NASA activities.
in scope but nevertheless of importance as a Publications include conference proceedings,
contribution to existing knowledge. monographs, data compilations, handbooks,
sourcebooks, and special bibliographies.
TECHNICAL MEMORANDUMS:
Information receiving limited distribution TECHNOLOGY UTILIZATION
because of preliminary data, security classifica PUBLICATIONS: Information on technology
tion, or other reasons. used by NASA that may be of particular
interest in commercial and other non-aerospace
CONTRACTOR REPORTS: Scientific and applications. Publications include Tech Briefs,
technical information generated under a NASA Technology Utilization Reports and
contract or grant and considered an important
Technology Surveys.
contribution to existing knowledge.

Details on the availability of these publications may be obtained from:

SCIENTIFIC AND TECHNICAL INFORMATION OFFICE

NATIONAL AERO NAUTICS AND SPACE ADMINISTRATION

Washington, D.C. 20546
 629.13
Un325tn ENGIN
+ 43-73

NASA TECHNICAL NOTE | NASA TN D-6573

:
FULL-SCALE WIND-TUNNEL TESTS
OF A SMALL UNPOWERED JET
AIRCRAFT WITH A T-TAIL

by Paul T. Soderman and Thomas N. Aiken

THE LIBRARY OF THE
Ames Research Center
and NOW 29 1971
U.S. Army Air Mobility R&D Laboratory *ś
Moffett Field, Calif. 94035

NATIONAL AFRONAUTICS AND SPACE ADMINISTRATION . WASHINGTON, D. C. - NOWEMBER 1971

 1. Report No. 2. Government Accession No. 3. Recipient's Catalog No.
NASA TN D-6573
4. Title and Subtitle 5. Report Date
FULL–SCALE WIND-TUNNEL TESTS OF A SMALL UNPOWERED November 1971
JET AIRCRAFT WITH A T-TAIL 6. Performing Organization Code

7. Author(s) 8. Performing Organization Report No.

Paul T. Soderman and Thomas N. Aiken A-3 135

10. Work Unit No.

9. Performing Organization Name and Address 126-13-01-43-00-21
Ames Research Center, NASA
Moffett Field, Calif., 94035 11. Contract or Grant No.

13. Type of Report and Period Covered

12. Sponsoring Agency Name and Address Technical Note

National Aeronautics and Space Administration -

Washington, D.C., 20546 14. Sponsoring Agency Code

15. Supplementary Notes

16. Abstract

The aerodynamic characteristics of a full-scale executive type jet transport aircraft with a T-tail were investigated in
the Ames 40- by 80-Foot (12.2- by 24.4-meter) Wind Tunnel (subsonic). Static longitudinal and lateral stability and
control characteristics were determined at angles of attack from -2° to +42°.

The aircraft wing had 13° of sweep and an aspect ratio of 5.02. The aircraft was tested power off with various wing
leading- and trailing-edge high-lift devices. The basic configuration was tested with and without such components as engine
nacelles, wing tip tanks and empannage. Hinge-moment data were obtained and downwash angles in the horizontal-tail
plane location were calculated. The data were obtained at Reynolds numbers of 4.1X10° and 8.7X10° based on wing
mean aerodynamic chord.

The model had static longitudinal stability through initial stall. Severe tail buffet occurred near the angle of attack for
maximum lift. Above initial stall the aircraft had pronounced pitch-up, characteristic of T-tail configurations. A stable trim
point was possible at angles of attack between 30° and 40° (depending on c.g. location and flap setting).

Hinge-moment data showed no regions that would result in adverse effects on stick force. Comparisons of
wind-tunnel data and flight-test data are presented.

17. Key Words (Suggested by Author(s) ) 18, Distribution Statement

Deep stall Unclassified – Unlimited

T-tail aircraft
General aviation

19. Security Classif. (of this report) 20. Security Classif. (of this page) 21. No. of Pages 22. Price"
Unclassified Unclassified 99 $3.00

"For sale by the National Technical Information Service, Springfield, Virginia 22151
 NOTATION

wing span, 10.40 m (34.1 ft)

wing chord measured parallel to the plane of symmetry, m (ft)

2 /
mean aerodynamic chord, 5 J." * cº dy, 2.14 m (7.04 ft)
-

drag
- - -

drag coefficient (wind axes), JTag

docS
- - - lling -*.
-

rolling-moment coefficient about stability axis,

- - -

Tolling moment
q2Sb
3Cl
T lateral stability parameter, per deg

0C1 . -

55. aileron effectiveness parameter, per deg

d

- - - - lift
lift coefficient (wind axes), —-
docS
ôCL
5T flap effectiveness parameter, per deg

- - - - - - Č --- itching moment

pitching-moment coefficient about - (stability axes), pitching moment
4 (locSC
3C - - - - - - -

ſno. + longitudinal stability parameter, per deg

Cn yawing-moment coefficient about moment center shown in figure 2(a) (stability axes),
yawing moment
qooSb

0C
# directional stability parameter, per deg

ôCn
5- rudder effectiveness parameter, per deg
r

side force
side-force coefficient (wind axes),
qo-S

horizontal-tail incidence angle, deg

dynamic pressure, N/m3 (1b/sq ft)

Vood
Reynolds number, —

wing area, 21.50 m? (231.77 ft”)

free-stream velocity, m/sec (ft/sec)

iii
y spanwise distance perpendicular to the plane of symmetry, m (ft)

Cy angle of attack of fuselage, deg

6 angle of sideslip, deg; positive — nose to left

ða trailing-edge aileron deflection angle, deg; positive – left aileron trailing edge down

Öe elevator deflection angle, deg; positive – trailing edge down

ðf trailing-edge flap deflection measured from wing chord line, deg

ör rudder deflection angle, deg; positive - trailing edge left

ôs spoiler deflection angle, deg

e average downwash at the tail location with respect to free stream, deg

m wing Semispan station, --

b/2

Ac/4 sweep angle of quarter-chord line, 13°

19 free-stream kinematic viscosity, m*/sec (ft”/sec)

l leading-edge contours defined on figure 2(d)

Subscripts

L left

IIla X maximum

R right

t tail

Ul uncorrected

A change

co free stream

Hinge Moments

Positive hinge moments tend to move the control surface in the direction of positive
deflection. The average chord aft of the hinge line was used for the reference length.
iv
Aileron

_hinge moment
Cha where = 0.544 m2 (5.85 ft”)
qSača
= 0.38 m (1.24 ft)

Rudder

_hinge moment
Chr where = 0.609 m2 (6.56 ft”)
qSrCr
= 0.46 m (1.51 ft)

Elevator

_ hinge moment where

Che Se = 0.635 m2 (6.83 ft”)
qSeče
Ce = 0.29 m (0.96 ft)

Horizontal stabilizer

_ hinge moment where

Chh Sh = 5.02 m2 (54.0 ft2)
qShēh = 1.17 m (3.83 ft)
Čh

Note: Se is the area of the right or left elevator; Sh is the total area of the horizontal stabilizer.
 FULL–SCALE WIND–TUNNEL TESTS OF A SMALL

UNPOWERED JET AIRCRAFT WITH A T_TAIL

Paul T. Soderman and Thomas N. Aiken

Ames Research Center

and
U.S. Army Air Mobility R&D Laboratory
Moffett Field, California 94035

SUMMARY

The aerodynamic characteristics of a full-scale executive type jet transport aircraft with a
T-tail were investigated in the Ames 40- by 80-Foot (12.2- by 24.4-m) Wind Tunnel (subsonic).
Static longitudinal and lateral stability and control characteristics were determined at angles of
attack from -2° to +42°.

The aircraft wing had 13° of sweep and an aspect ratio of 5.02. The aircraft was tested power
off with various wing leading- and trailing-edge high-lift devices. The basic configuration was tested
with and without such components as engine nacelles, wing tip tanks, and empennage.
Hinge-moment data were obtained and downwash angles in the horizontal-tail plane location were
calculated. The data were obtained at Reynolds numbers of 4.1X 10° and 8.7X 10° based on wing
mean aerodynamic chord.

The model had static longitudinal stability through initial stall. Severe tail buffet occurred near
the angle of attack for maximum lift. Above initial stall the aircraft had pronounced pitch-up,
characteristics of T-tail configurations. A stable trim point was possible at angles of attack between
30° and 40° (depending on c.g. location and flap setting).

Hinge-moment data showed no regions that would result in adverse effects on stick force.
Comparisons of wind-tunnel data and flight-test data are presented.

INTRODUCTION

Most small aircraft, including executive jet transports, are designed with a minimum of
wind-tunnel data. Furthermore, flight tests are likely to be qualitative rather than quantitative. As a
result, the designer has little opportunity to verify his design predictions.

Therefore, to aid designers the present investigation was conducted to determine the static
longitudinal and lateral stability and control characteristics through deep stall of a full-scale
executive jet aircraft. The deep stall testing was conducted to see if the aircraft exhibited
unfavorable characteristics at high angles of attack because of its T-tail. Some of these problems and
related research can be found in references 1 through 4. Unfortunately, it cannot be determined
from wind-tunnel tests of unpowered aircraft whether the powered aircraft can become locked in
deep stall.

AIRCRAFT AND APPARATUS

In figures 1(a) and (b) the model is shown mounted in the Ames 40- by 80-Foot (12.2- by
24.4-m) Wind Tunnel. Pertinent dimensions of the basic model configurations are given in
figures 2(a) and (b).

Wing

The wing had a quarter chord sweep of 13°, an aspect ratio of 5.02, a taper ratio of 0.507, and
a dihedral angle of 2.5°. The airfoil section was an NACA 64A-109 modified by increased camber
and chord at the leading edge (fig. 2(d)) which was minimum at the root and maximum at the
wing-tip tank junction.

High Lift Devices

Flap details- The basic wing had a single slotted, extendable (Fowler) flap (fig. 2(c)) located
from the edge of the fuselage at 7.1 percent to 61.2 percent m. Maximum flap angle was 40° at the
lower Reynolds number and 38° at the higher Reynolds number because of air load effects. A
center section flap that extended under the fuselage was tested (fig. 1(b)). There were no gaps
between the sides of the center section flap and the main flaps.

Leading-edge contours- The drooped leading edges of the basic wing were removed, part way
through the test, and replaced by various leading-edge contours (fig. 2(d)). The dimensions of the
leading edges varied linearly from root to tip.

Wing planform modification- In an attempt to delay the stalling of the wing tip region, fences
were placed first on the tops and then on the sides of the tip tanks (fig. 2(e)).

Lateral Controls

Ailerons- The ailerons (fig. 2(b)) had relatively blunt leading edges and balance tabs to
decrease stick force. As the ailerons were moved, the balance tabs moved in the opposite direction
such that

otab = -(5/6)6a

where otab is the tab angle relative to the aileron chord and 6a is the aileron angle relative to the
wing chord.
 Spoilers- The chords of the spoilers were 10.25 percent of the wing chord at midspoiler and
were located from 22.2 percent to 49.4 percent semispan (see fig. 2(b)). Spoiler angles ranged from
0° to 42°. In addition to the basic wing spoilers, dummy spoilers were tested outboard (fig. 2(f)).

Tail

The geometry of the horizontal and vertical tails is described in figure 2(g). Pitch control was
provided by an all-movable tail that had an available deflection range of 0.4° to -7.0° and by a
32 percent chord elevator with balance horn. The elevator angle was variable from 15° to -15°. The
rudder (25 percent chord) had a deflection range of 30° to -30° and had a trim tab that was locked
at 0°. The horizontal stabilizer was used for aircraft trim.

Nacelles

Engine nacelle detail and location are shown in figure 2(h). A constant-area circular duct was
installed in each nacelle to allow mass flow conditions of 4.81 kg/sec (10.6 lb/sec) of air at standard
conditions, similar to that of the jet engines for idle airflow at a Mach number of 0.2. Static and
total pressures were measured with rakes at the aft ends of the ducts to determine the actual
dynamic pressure of the nacelle flow and the internal nacelle drag (which was removed from the
data). The nacelles were removed from the pylons during a part of the test.

Tip Tanks

Wing tip tank detail and location are shown in figure 2(b). All data are presented with the tip
tanks on unless stated otherwise.

TESTING AND PROCEDURE

Forces and moments were measured for the model through an angle-of-attack range from -2°
to 42°. Pitching-moment data were computed about a moment center location at 25 percent é. The
center-of-gravity range for this aircraft is 16 percent C to 31.5 percent é. Tests were conducted at
Reynolds numbers of 4.1X 10° and 8.7X 10° based on a mean aerodynamic chord of 2.14 m
(7.04 ft) and speeds of 27.8 m/sec (54.2 knots) and 59.0 m/sec (115.0 knots), respectively. These
speeds correspond to dynamic pressures of 478.8 N/m3 (q = 10 psf) and 2156 N/m3 (q = 45 psf).

Tests were conducted with the basic configuration' at several tail incidences with variable
elevator, rudder, aileron, spoiler, and flap settings. Data were also obtained with landing gear down.
The maximum angle of attack at R = 8.7X 10° was 16° (tail on) because of tail buffet load
limitations. Most data, tail on, at angles of attack higher than 16° were taken at a Reynolds number
of 4.1 X 10°.

'Basic configuration refers to the airplane as shown in figure 1(a) with engine nacelles, tip tanks, and
empennage on model. Control surfaces were at zero angle unless stated otherwise.
 DATA ACQUISITION AND REDUCTION

Forces and moments were measured on the wind-tunnel six-component balance. Torque tubes
in the elevators and rudder were gaged to provide hinge-moment data.

All data were corrected for strut tares, nacelle internal flow drag, and wind-tunnel wall effects.
Nacelle internal flow drag was calculated from pressure measurements in the nacelle ducts, and
ACD = 0.0005 cos o was subtracted from model drag. Corrections added for wind-tunnel wall
effects were

Aq = 0.506 CLu

ACD = 0.0088 CLu”

ACm = 0.0171 CLu (tail on runs only)

ACCURACY OF MEASUREMENT

The various quantities measured were accurate within the following limits, which include error
limits involved in calibrating, reading, and reducing the data.

Angle of attack +0.2°

Angle of sideslip +0.5°
Free-stream dynamic pressure +0.5 percent
Control surface settings +0.5°

Coefficients at
Force or moment R = 8.7x 106

Lift +22.4 N (+5 lb) +0.0005

Drag + 13.4 N (+3 lb) +.0003
Side force + 13.4 N (+3 lb) +.0003
Pitching moment +271 J (+200 ft-lb) +.0027
Yawing moment + 136 J (+100 ft-lb) +.0003
Rolling moment +475 J (+350 ft-lb) +.0010

RESULTS

Table 1 is the index to the figures. The longitudinal data are presented in figures 3 to 18 and
the lateral data in figures 19 to 30. Downwash and hinge-moment data are given in figures 31 and
32, respectively.
 DISCUSSION

Longitudinal Characteristics

Flap effectiveness- The longitudinal characteristics of the basic airplane at R = 8.7X 10° with
three flap settings are shown in figure 3(a). The flap effectiveness parameter, CL5, was 0.015/deg
for the 20° flap setting and 0.013/deg for the 38° flap setting. A theoretical flap effectiveness
estimate was made using the simplified lifting-surface theory of reference 5, which gave the value of
CL5 as 0.022/deg, almost 60 percent higher than measured. This discrepancy was probably due to a
nonoptimum gap setting for the single-slotted, Fowler type flaps. A comparison of small-scale with
full-scale wind-tunnel data to be discussed in a later section shows that small-scale flap effectiveness
is closer to the theoretical value. This suggests that the flap gap choice was based on small-scale test
data and not corrected properly for full-scale Reynolds number effects.

Maximum lift- Figure 3(a) shows the basic stall characteristics of the aircraft at R = 8.7X 10°.
Because of severe buffet on the tail as it penetrated the wing wake, the tail was guy-wired as shown
in figure 1(a).” In addition, some of the data were taken at a reduced Reynolds number of
4.1 X 10°. The tail buffet acted as a strong stall warning. Figure 3(b) shows the longitudinal
characteristics at R = 4.1X10°. Increasing the Reynolds number from 4.1X 10° to 8.7X10° caused
an increase in maximum lift coefficient of 0.19 (flaps down) and 0.20 (flaps up) as shown in
figure 4. The high Reynolds number condition is closer to actual flight conditions. Observation of
tufts on the left wing indicated that a region of separated flow developed near the wing leading edge
tip tank junction at 8° angle of attack (this did not happen with tip tanks off). As angle of attack
was increased the region of separated flow spread aft and inboard. Near CLm aX the wing root began
to stall. Both regions grew with angle of attack until most of the wing stalled and lift dropped.

Static stability- A study of the variation of the stick-fixed pitching-moment coefficient with
angle of attack (fig. 5(a)) shows that the airplane was stable through maximum lift (even for aft c.g.
limit of 31.5 percent d). Above maximum lift, the classic deep stall situation occurred that will be
discussed later. The data presented for c.g. at 25 percent & give Cmo = -0.0186/deg. At stall the
aircraft experienced a slight nose down pitching moment. The stick-free static stability
characteristics, determined from hinge-moment and pitching-moment curves, are shown in
figure 5(b) (data are shown for c.g. at 25 percent C). Freeing the elevator reduced the stability, but
the aircraft did not become unstable. For the aft c.g. case (31.5 percent C), 6f = 0°, o = 0°, Cimo was
reduced from -0.0185 to -0.005/deg.

Deep stall- As illustrated in figure 6, the airplane was unstable above maximum lift
(stick-fixed) with the center of gravity at the quarter chord, and maximum nose-down trim until an
angle of attack of 32° was reached at which point static stability was again attained. Furthermore,
the pitching moments became zero or slightly positive above o' = 28° flaps down. Thus it may be
possible (at low Reynolds number) for the airplane to reach a region of positive pitching moment
and pitch up to o = 32°, a trim condition (power off) if the pilot does not take corrective action.
However, aircraft rolloff may preclude this possibility, as will be discussed in a later section. As
shown by the axes superimposed on figure 6(b), at forward c.g. the pitching moments do not
become positive, but at the aft c.g. the aircraft would reach the positive pitching-moment region
*The wires had very little effect on the data.
sooner and could pitch up to trim at a = 39° (flaps up or down) while completely stalled.
Figure 6(c) shows that while the effect of sideslip was beneficial, 8° of sideslip changed the pitching
moment only 0.06 at 0 = 32°.

With the flaps up, elevator effectiveness was maintained at all angles of attack but
pitching-moment increment due to full elevator deflection at angles of attack greater than 24° is
approximately one fourth that at angles of attack below stall (see fig. 7(a)). Therefore, recovery
from deep stall (flaps up, c.g. at 25 percent C) would be possible using the elevator, but the time it
takes to rotate the nose down may be long. With the c.g. in the aft location there is insufficient
elevator effectiveness to recover from deep stall. With the flaps full down (fig. 7(b)) there was an
almost complete loss of elevator control power above o = 24°. Since the data (flaps up and down)
were taken with the horizontal stabilizer leading edge full up, any movement of that control surface
would only make the pitching moments more positive.

Figures 8 and 9 illustrate the effects of horizontal stabilizer incidence and removal of the
empennage, respectively, on the longitudinal characteristics.

Effect of wing tip tanks, engine nacelles, and landing gear- Figure 10 shows that the wing tip
tanks caused an increase in lift coefficient and lift curve slope primarily because of the increased
wing area and aspect ratio (reference area was not changed). The drag change was small up to
‘Limax The addition of the tip tanks made the aircraft slightly more stable in pitch.
The engine nacelles caused a decrease in lift, especially with flaps down (fig. 11). This decrease
was probably due to interference with flow around the wing that reduced wing lift since the nacelles
did not develop negative lift or reduce tail lift. This explanation is substantiated by the increase in
nose-down pitching moment with the nacelles on the aircraft. If the nacelles had developed negative
lift or if the tail lift had been reduced, the pitching-moment change would have been nose up. The
fact that the wing tips were probably not affected by the nacelles accounts for the nose-down
pitching-moment change (i.e., the lift loss was inboard).

The landing gear effect on the longitudinal characteristics is small (fig. 12).

High-lift devices- The effects of four wing leading edges are given in figures 13(a) and (b), flaps
up and down. For the flap down case the leading edge la , which had the greatest droop, increased
maximum lift beyond the value achieved by ly, the basic configuration leading edge.

In an attempt to improve the CLimax of the airplane, fences were placed on the tops and,
later, sides of the tip tanks to alleviate flow separation at the junction of the tip tank and wing.
Fences on the sides of the tip tanks caused an increase in lift due to the increased wing area and
aspect ratio (fig. 14). In no case was the flow separation alleviated near the tip.

The center body flap (fig. 1(b)) caused a very small reduction in lift and drag of the model and
a very slight change in pitching moment (fig. 15). The reason for the reduction in lift and drag is
unknown.

Drooping the ailerons 13.7° increased maximum lift coefficient by 0.1 (fig. 16). Since
drooped ailerons reduce roll control, outboard spoilers were tested. These will be discussed in the
lateral control section.

6
 Effect of spoilers- Runs were made with various right and left spoiler deflections (see
figs. 17(a) and (b)). The deflection of one or both spoilers 42° caused a nose-down
pitching-moment change probably because of an induced increase of tail angle of attack. This
Supposition checks with figure 17(c) that shows very little change in pitching moment with
outboard spoiler deflection. It was expected that the flow field of the tail would not be affected
greatly by deflection of the outboard spoilers. The drag was increased 80 percent with full spoiler
deflection.

Comparison of wind-tunnel and flight-test data- A comparison of Ames 40- by 80-Foot (12.2-
by 24.4-m) Wind Tunnel data, Wichita State University 7- by 10-Foot (2.1- by 3.1-m) Wind Tunnel
data and Lear Jet flight-test data is made in reference 6. Two figures from that paper are presented
in this report as figures 18(a) and (b). Results show good agreement between full-scale wind-tunnel
and flight-test data. Reynolds number effects account for most of the difference between
Small-scale and full-scale results.

Lateral and Directional Stability and Control

The lateral characteristics of the airplane are shown in figures 19 to 23, and lateral and
directional control effectiveness in figures 24 to 29. Stability derivatives Cng and Cl6
are plotted
versus angle of attack in figure 30. These data show that the airplane had positive effective dihedral
(-Cla) over the normal operating range and was directionally stable statically (positive Cng). With

the tail removed (fig. 22) the nonzero rolling moment and side force at 3 = 0° were probably due to
flap misalinement. The flaps had been removed and reinstalled on the model prior to these runs.
The data in figure 23 show that as the model stalled with flaps up, it tried to roll left (left wind
down) and with flaps down, it tried to roll right (right wing down). The change in roll direction at
Stall was probably caused by asymmetric deflection of the flaps. The rolling moment, flaps down,
was greater than that produced by full opposite aileron deflection. This severe rolloff in stall would
complicate recovery, but it might prevent a deep stall condition.

Control effectiveness- Aileron roll power was fairly constant below stall but decreased rapidly
in stall (fig. 24). The airplane had slight favorable yaw due to aileron above 6° angle of attack.
Figures 24(b)–(d) show the control power due to one aileron. The nonzero side force was probably
due to model misalinement in the test section. Figure 25 is a summary plot of °ls,
versus angle of
attack.

Rudder deflection affected the longitudinal characteristics very little. Figure 26 shows the
lateral effects of rudder deflection. The rudder was capable of holding the airplane in sideslip
between -15° s 3 s 15°.

The control power of the basic spoiler is shown in figure 28(a) as plots of Cy, Cn, and Cl
versus left spoiler angle (right spoiler full down). Figure 28(b) shows the effectiveness of dummy
outboard spoilers S1, S2, and S3. These spoilers were more effective than ailerons or inboard
spoilers for lateral control. The lateral characteristics of the airplane with the landing gear extended
are shown in figure 29. Comparison with the results in figure 19(a) (landing gear retracted) indicates
that the landing gear had a small effect on Cy vs. 3 but only a slight effect on Cn and C1 vs. 3.
 Downwash at the Horizontal Tail

An average downwash angle at the horizontal stabilizer was estimated from curves of Cm vs. 0.
for several tail incidence angles. The intersection of the tail-on curves with the tail-off curve are
points where tail lift is zero; and for a symmetrical horizontal stabilizer

e = 0 + it

Figure 31 shows the results of the above calculation, which were identical for both Reynolds
number cases.

Hinge Moments

Typical curves of hinge-moment coefficient Ch versus angle of attack and Ch versus control
position are presented in figures 32(a)–(h) for aileron, elevator, rudder, and horizontal stabilizer.
The data were obtained at R = 8.7X 10° to approximate actual flight conditions. These results show
no control force reversal for any of the controls within the normal operating range.

CONCLUSIONS

A full-scale wind-tunnel investigation was made of a small jet aircraft with a T-tail to
determine the longitudinal and lateral stability and control characteristics through deep stall, power
off. The following conclusions were drawn from the results of the investigation:

1. The airplane had stick-fixed static longitudinal stability at angles of attack up to stall for
the full c.g. range. With the stick free, stability was reduced but the aircraft did not become
unstable.

2. Before stall the tail experienced severe buffet as it penetrated the wing wake, and, in stall,
the airplane tended to roll right wing down or left wing down depending on flap angle. The tail
buffet acted as a strong stall warning, that might prevent deep stall entry during actual flight
conditions. However, the rolling moment in stall, flaps down, was greater than that produced by full
opposite aileron deflection.

3. Above stall, the airplane was unstable in pitch, and the pitching moments could become
positive, depending on c.g. A trim condition in deep stall (o = 39°) with a large reduction in elevator
control was possible. With the c.g. in the aft position, elevator control and horizontal stabilizer
control were insufficient for recovery from deep stall trim.

4. The airplane was directionally stable, below stall, and had positive effective dihedral.

Ames Research Center

National Aeronautics and Space Administration
Moffett Field, Calif., 94035, July 6, 1971
 REFERENCES

. Aoyagi, Kiyoshi; and Tolhurst, William H., Jr.: Large-Scale Wind-Tunnel Tests of a Subsonic Transport With Aft
Engine Nacelles and High Tail. NASA TN D–3797, 1967.

2. Ray, Edward J.; and Taylor, Robert T.: Effect of Configuration Variables on the Subsonic Longitudinal Stability
Characteristics of a High-Tail Transport Configuration. NASA TM X–1165, 1965.

. Trubshaw, E. B.: Low Speed Handling With Special Reference to the Super Stall. J. Roy. Aeronaut. Soc., vol. 70,
no. 667, pp. 695–704, July 1966.

. Thomas, H. H. B. M.: A Study of the Longitudinal Behavior of an Aircraft Near-Stall and Post-Stall Conditions.
R.A.E. TM Aero 953, 1966.

. DeYoung, John: Theoretical Symmetric Span Loading Due to Flap Deflection for Wings of Arbitrary Plan Form
at Subsonic Speeds. NASA Rep. 1071, 1952.

. Neal, Ronald D.: Correlation of Small-Scale and Full-Scale Wind-Tunnel Data With Flight—Test Data on the Lear
Jet Model 23. Paper 700237 presented at SAE National Business Aircraft Meeting (Wichita, Kansas), March
1970.
 TABLE 1.- INDEX TO FIGURES

The model in the wind tunnel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Geometric details of the model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Longitudinal characteristics with flap deflections . . . . . . . . . . . . . . . . . . . . . . . . .
Reynolds number effect on longitudinal characteristics . . . . . . . . . . . . . . . . . . . . .
Variation of pitching-moment coefficient with angle of attack . . . . . . . . . . . . . . . . .
Longitudinal characteristics through deep stall . . . . . . . . . . . . . . . . . . . . . . . . . .
Elevator effectiveness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Longitudinal characteristics with horizontal tail deflection . . . . . . . . . . . . . . . . . . .
Longitudinal characteristics with empennage removed . . . . . . . . . . . . . . . . . . . . . .
Longitudinal characteristics with tip tanks removed . . . . . . . . . . . . . . . . . . . . . . .
Longitudinal characteristics with engine nacelles removed . . . . . . . . . . . . . . . . . . . .
Longitudinal characteristics with landing gear down . . . . . . . . . . . . . . . . . . . . . . .
Effect of four leading edges on longitudinal characteristics . . . . . . . . . . . . . . . . . . .
Effect of fences on longitudinal and lateral characteristics . . . . . . . . . . . . . . . . . . . .
Effect of center section flap on longitudinal characteristics . . . . . . . . . . . . . . . . . . .
Effect of drooped ailerons on longitudinal characteristics . . . . . . . . . . . . . . . . . . . .
Longitudinal characteristics with spoiler deflection . . . . . . . . . . . . . . . . . . . . . . . .
Comparison of wind-tunnel and flight-test data . . . . . . . . . . . . . . . . . . . . . . . . . .
Lateral characteristics versus sideslip angle 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Lateral characteristics versus angle of attack o' . . . . . . . . . . . . . . . . . . . . . . . . . .
Lateral characteristics versus lift coefficient . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Lateral characteristics versus 3, empennage off . . . . . . . . . . . . . . . . . . . . . . . . . .
Lateral characteristics versus O., empennage off . . . . . . . . . . . . . . . . . . . . . . . . . .
Lateral characteristics with aileron deflection . . . . . . . . . . . . . . . . . . . . . . . . . . .
Aileron effectiveness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Lateral characteristics with rudder deflection . . . . . . . . . . . . . . . . . . . . . . . . . . .
Rudder effectiveness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Lateral characteristics with spoiler deflection . . . . . . . . . . . . . . . . . . . . . . . . . . .
Lateral characteristics with landing gear down . . . . . . . . . . . . . . . . . . . . . . . . . .
Lateral stability derivatives Cng and Cl6 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

Downwash at tail . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Control surface hinge moments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

10
 Basic
(a)
configurati on.
E Figure
The
1.-
model
mounted
in
Ames
the
40-
80-Foot
by
(12.2-
24.4-m)
Wind
Tunnel.
 (b) Center section flap, nose and tip booms on model.

Figure 1.- Concluded.

l2
 - Horz. Vert.
Wing ºf ºf
Aspect ratio 5.02 || 4 O -

Toper rotio O.5O7 |O469 (O493

21.5| || 5.02 || 34.48
Ared, sq m (23.77)|(54.00) (34.48)

Sweep, deg
(25% C |3 || 25 | 40
Airfoil section 64A-IO964AOO864AOIO
All dimensions in meters (feet)

(19.2)5.86 —-
- 2.47
(8, IO)

4.47
- (4.67)

Moment center - w \
C
c/4. Line

c/4. Line

N
c/4. Line 4.72
-—(15.50)
Verticol foil C

I
|60(5.25) O

Moment center’ (3.34)|O2

| 7 4.93
--553---—(léº--
12.50 (41)
|3.18(43.25) *-

(34. IO)
(a) General arrangement of model.

Figure 2.- Geometric details of the model.

l3
 Hoza,023
O 5
( )

78% C
|
C/4 Aileron
Line j hinge line

|3°

5.2O
(7. O5)

O.24
(O. 78)

Spoiler

73%, C
Flap leading
edge -
|
|--|-->
Gºgº.43
M. (O.93)(1.39)
Ç. Aircraft l º
|- 2.75
F (9.O2)

(b) Basic wing detail.

Figure 2.- Continued.

ll.
 O.268 C

Flop angle, deg Gap, cm (in) d,cm (in)

O O
2O 1.57 (O.62) 13.97 (5.5)
38 2.46 (O.97) 18.55 (7.3)
40 2.46 (O.97) |8.8O (7.4)

(c) Trailing-edge flap.

Figure 2.- Continued.

lj
 l
12 (STD NACA (blunt)
- - - - - 64A-IO9) L6% Chordline
Wing and tip tonk junction, cm (in) of STD NACA

x = 7.62 (3.0) R = 1.52 (0.6)

x2=8.13 (3.2) R2=1.27 (0.5)
X3=10.91 (4.3) R3=1.52 (O.6)

Hinge point
configuration) f 1
of 14
lA
Wing and fuselage junction, cm (in)
n
(sºdrooped º:o)
NACA
30° L6% Chordline
x1=8.64(3.4) R1 =2.28(O.9) of STD NACA
l
x2=13.71 (5.4) R2 = 1.78(O.7)
l2 and l3

| cm
H

(d) Leading-edge modifications.

Figure 2.- Continued.

l6
 º | 6|
(5.28)
_-T f
|
O.18(O.58)
354;
Area =(5.81 ft“)
O 48
(i. 58)
—
| Side view

O. 49
(1.6 O)

(e) Fence located on side and top of tip tank.

Figure 2.- Continued.

l?
 2.7°, −:
(3.00

--~ - -
L;
— —-m-

-TET--

Symbol "A" dimension, (in) cm

Sl (2.25).5.7.
S2 (5.5O)|3.98
S3 (6.75)|7.15
(f) Outboard spoiler.

Figure 2.- Continued.

l3
2. *i Q
1.43
|-- (#33)---—t:33,
Fuselage

© Aircraft

O.25 (O.75) 2.24

(7.35)
(g) Horizontal and vertical stabilizer.

Figure 2.- Continued. l9

O.76 O.63 O.42
(2.50)(2.08) ( .38)
º º%2RN

Cross-section at engine inlet Chord line"

Cross-section at wing leading edge-fuselage junction

(h) Fuselage-wing cross section at two locations.

Figure 2.- Concluded.

2O
 CD
Cm
deg
a,
Lhoanrfgaicgtuerdiastniaoclns
3.-
cFigure
basic
the
of
angles.
flap
three
for

ſ If 8.7x10°,
R
0(a)
=
it
.4°

/ |&#
TIMZ

I//
Tºr
||
º

|
|

ſº
 s

| #ess
O D o

*
J’ ~4–)
‘Lºſ V

22
 A-2
&Q
-

f T-F-ta
~~-H,

&^
Ø\
-
*N.
| N. NO

|!/ R
xIO-6
4.
|O
[]
5.8
KX
8.7

.
.3
O
.4
8
|
42
.2
|6
- .2
-.4
CD
deg
Cm
Cl,
(a)
8f=0°

Č3 4.-
Reynolds
of
Effect
longitudinal
on
characteristics;
0Figure
=number
it
.4°.
º

ſº

/
/
deg
8,
xIRO-6
4
|4
O.O
38
8.7
[]

–4
–2
.5
.4
.3
.
.2
|6
Cm
|
8
4
O
–42
CD
deg
a,

down.
Flaps
(b)

Concluded.
4.-
Figure
 8t, deg
O O
D 20
KX 38

a, deg

(a) Stick fixed.

Figure 5.- Variation of pitching-moment coefficient with angle of attack; R = 8.7X 10°, it = 0.4°.

25
&
T

O
=8f º
N
0
f=8 °

T--
T
T-G
.2O
== TS
T
3
=st8°
f8=380
.30 |.6
|
. . .2
8
.4
|6
|2
8
O
CL
deg
a,

free.
Stick
(b)

Concl
5.-
Figureuded.
 72

64

56

48

4O
:
Ç
32
deg
8t,
O
O
38
D
24

.6
.8
I.O
O
|O
2O
3O
4O
CD
a
,deg
R
(a)
= .7x10°,
8.4°
0it

ſº Figure
Longitudinal
6.-
characteristics
through
stall.
deep
% 64
|.6
k -
lo
S$
56
\f |.4
lo
5 IO
}H3
S$r
48
|.2 -
-

<
3|
*P*
-
R~~
-
co
4O
I.O
-

+\c
32
.8
-

deg
8f,
§
9
e
OO
CL
4O
&[] c
Hr
24
.6

*
|6
.4
\|
8
| .2
Q
O
O

–8
-.2
-
..4
4O
3O
2O
|
|.2
I.O .4
. 8O
.O
4
Cm
deg
a,
CD

=0it
R .4°
.1x10°,
4(b)

Continued.
6.-
Figure
56 48 4O 24 6
|

0.1X10°,
6f
4(c)
0R
=
it
sideslip
in.4°,
°,

fºr: 32

Concluded.
6.-
Figure

f
t

4 ()
O []
— —8

NS
 56

Ś
48

4O
lº
O
-

32
º -

24
| Cn
$
s
6
|
-

}| deg
8e,
8
-
|5
[]
O
O
1—5
()
O

||
–8
&8
-16
.8
I8
.
4.O
.O26 5O
4O
3O
2O
|
O O
-IO
deg
CD
a,

=
it
R .1x1
0°,
f = 0°,
8.4°
04(a)

effe
7.-E re
Figuctiv tors.
levaenes

~
-
a-
/
Z
 |.6
64
Å
|4
56

#wº
§=2
# LO
ſºl
G)
KX
4O

.8
g32 -

to
|i CL
8.
deg
8e,
[]
º .624 ----
-

O
fº
–|5
K»

. &4|6l
G

#
O
O

–25
-8;
.
.8
46
|.O
|.2
|.4
–.4
—.8
|-IO
CD
OO
2O
3O
5O
4O
Cm
deg
a,

4(b)
=
6f 0°,
.1X10°,
0R.4°
it

Figure
Continued.
7.-

º
º

W|| i)
|*NY
ſº
|\|
Ž
Nº
-
:
%| |
-

|
|||
|}sº de
8e,

%’ |||}}}
|
|||Z%
/ |

it=R.1x10°,
°,
6f.4°
04(c)
Continued.
7.-
Figure
 33
ºpºnuņuoO -- L 9 InáA
„L- = }} · 90IX I’v = H ‘40 = }g (p)
uo 6æp º o
O 2 2 |8t»O
O G – O|- Q | –
 ！ LITÈ,
*№， FH）
*<！[]（）|）ºļ.
<>）T°）+o）ļ9

-T° +→| gº
<< ·O

"№t-FĒ| }

}}

.1x10°,
<£
40°,
6

-R7°
4it
=
§

(e)
8f=

№ss
）
71A，
<S.
ſĒŠș！<

do <r. N o do co < c\!

 35
'pºpnĮouOO -- L 9 InáJ
90IX L'8 = H ‘40 = }g ‘AL- = }} (})
+++++#
||||||||ſię
||||||||||//
||||||||//*
|||||||||||||||/}
Î//////|// O’]
//|/º3/} 2"|
†;"|
3. |.2

§
Nº
s .8
:º
##–8
|*G.º

I#
.6
|||| CL
,dIfeg
O
Toil
off
-

O.4
[]
-
Ó2.6 f|
4
N
-.5
-
A7.O

§
-

–.2
HS
-

—.4
2.O
3
.
O
|2
.4
.2
—
O
–4—
4
8
6 2
4O
. .2
CD
deg
a,
Cm

40R(a)
°,
=6f
.1x10°
Figure
Effect
8-
horizontal
of
stabilizer
incidence
on
longitudina
characteristlics.
 37
·pænuņuoO -'8 9Iná_I
90IX L'8 = H '40 = 39 (q)
6øp ‘odo
„929 | 2 |8t»O

O D C <!
// ±{440 ||0.1
6øp º 4 !
O’I
2" |
t»’I
%

|\
\;=#| || ||
-

deg
it,
ºfoff
.4
!-2.6
–4.5
–7.O
->
A

-.2
O
2
.6
.5
.
34
2
6
|2
8
4
O
—O
Cm
deg
a,
CD

4R =40°,
=f(c)
8 .1x10°

Concluded.
8.-
Figure
 ſ C
m

56 48 4O

5O

CD
|4O
O
-IO
3O
2OO
Rgº
cſ:
--GO
| A-A
—º. - -

lhoanrgaciteuridsitncasl.
on
empennage
of
Effect
9.-
cFigure

deg
a,

f.1X10°
8
4(a)
=
R
0°,
=

deg-O.4
it,
OO
4 n
50
LO
|.2
|.O
..6
O
.4
.82

|
|| |
°L
Toil Off
[]

|.2
lſº º

.2

§
à 64

56

|.2 48
G-

º -

I--
-
I —aprº-d
El -
[...]
EU
ºt.
-
4O
|| I.O

32
.8-
s
-

.6
P
-

Toil
.4
-Odeg
n
it,
O.4
Off
[]

.2

+2
|.4
|.2
I.O
62
48
.O
4O
3O
2O
|
OO
C
deg
a,

4R
=.1x10%
0°,
6f
(b)
Continued.
9.-
Figure

/
/
-
Z*
/- --->
-
.--
A--
/
º
 \!.
crºo

5
J|

8.7X
=
R
(c)
104

Figure
Concluded.
9.-
 O
—O
2
6
2
|
8
4

8.7x106.
=
R
off;
tail
tanks,
tip
wing
removing
of
Effect
10.-
Figure

deg
a,
%

Z
-

ſ
deg
8t,
tank
Tip

AN
/\\
On []
O Off
O

- On
KX
38
O38
A ff

/
º,

§
 M’
/
/ Z. |
/ |
-

ſZ4 Nocelles
deg
8f,
On
O
Off
O
[]
38
KX
On
-

| 3
A 8
Off

–4
O|8
||6
8
42
deg
a,

R
=
on .7x10°,
characteristics;
8.4°.
longitudinal
nacelles
engine
of
Effect
11-
0itFigure
d;
 lhoa.7X10°,
on
doors
and
gear
landing
of
Effect
12.-
3cFigure
f
86
0R
=
it
nr.4°.
a8°,
gciteruidsitncasl;

ſ
- ..3
+
O
.2
-.4
|
.4 CD
Cm
2 |
8
4
|6
O
— 2

Goors
dD
-
down
open
ear
~7
ºw

own-do rs closed
dKX
Gear

Æk,
Ä
KVX

| Gear
O
up

deg
a,

f
 Leading
edge
lD
| 1©2
lO
3
l4A

.2
O
-.34
.2
.2
4
–
CL
O
8
|42
|6
2O
24
Cm
deg
a,

(a)
=06f
°
Figure
13.-
Compariso
of
leading-edg
=four
.7×10°,
08modificatio
itR4". nens;
J.
 'pºpnĮouo0-ºg I 9 InãIJ
„88 = }Q (q)
ôæp ‘o
9||2||8t»O
3
[] O O
º'

Đôp3 õupoºT
O’I
Z’I
9"|

|6
 §

B model
Basic
of
tops
on
Fences
tonks
tip
of
sides
on
Fences
KX
tonks
tip

|6
8
O42
|–4
.3
.2
deg
a,
CD
data.
Longitudinal
(a)

=
R .7x106.
°,
6f
.4°,
it
tanks;
tip
0fences
on
814-
2Effect
of
Figure
#
 .24

.2O
O Basic model
D Fences on

.O8

.O4

Y.
.O4 L.J

Cn

–.O4 C}

.O4

–.O4
–2O —|6 – |2 4 8

(b) Lateral characteristics with fences on tops of the tip tanks; o' = 0°.
Figure 14.- Concluded.
 /
Sf
O
flop
tondard
l-J
.3
.O
.4
.2
|
-
–4 Cm
CD
2 |8
O
—
4
|6
2
8.7X10°,
=
R
flap;
section
center
without
and
with
flap
standard
of
Effect
15.-
Figure
f
8
38°.=

Center
D
section

added
flop

deg
a,

'N

ſ
-

S.
§

model
Bosic
O

oileron
Drooped
D

Cm
|
I6
8
O
42
—
CD
deg
a,

R
8f.7x10°,
13.7°,
drooped
ailerons
both
=8with
16-
characteristics
Longitudinal
Figure
3it28°,
.4°.
 %:ãº

% /

°ss
8s...
deg
O
O
3//
[]
20
O
A
4
O2
()
42
-

*H
g/

4;
O
- 3
.| .2
–4
O
4
8
|
—2
|6
2O
CD
Cm
a,
deg

f(a)
R =
.8x10°,
it=50°,
8.4°,
basic
0spoiler.
Figure
17.-
Effect
spoiler
of
longitudina
on
characteristlics.
Y.
/º
/-
£e

6f
(b)
 ſ * ºf
2.
A

model
B
Oosic
[]
S3
on

O
84
||
6
2—2O
CD
,d
aeg
0(c)
=8R3f°,
.7x10°,
outboard
spoiler
on
S,
left
wing.
S.
Figure
17.-
Concluded.
 2.8 I I
I | I |
Flops up
O Flight test
2.4
T — Ames 40 × 80,
R = 86 x IO6
2.O — — — — Wichita St. Univ. 7×10 — .--T
R = | 4 x |O6
22,
/J- Az'
7

2×3, -40°
7

2% 24°ſ 8, = O’
º
| / O Model 23
2,2' Z" ſ flight test
— Ames 4 O X 8C
22 T ——— Wichito St. Univ. T.
7x|O

4
JºO 4 8 |2 |6 O4
|
O8
|
|2 |6
|
.2O .24
a , deg CD

I-I- Z.
|
Flops |
20° 26 .*
--- Flops full down /
,”

£1. 3/
Żóſ
/

O / ſ
i O Flight test H O Flight test
— AmeS 40 X 80 — Ames 4O X 80
--- Wichito St. Univ.– — --- Wichito St. Univ.
7x|O 7x|O
l | I | | | | 1
O4 O8 || 2 .16 .2O .24 .O4 O8 || 2 || 6 .2O .24

CD CD

(a) Lift and drag characteristics.

Figure 18.- Comparison of wind-tunnel and flight-test data (taken from ref. 6).

5||
 O.2 I
Toil – On Flops up

- F=-
–O.2

…” JP
O.2
Toil-Off

Cm O = r. _ --rº- aſ --

— Ames 40x80,
R = 8.6 x 108
O.2 - Flops full down --- Wichita St. Univ.
-- Toil - On 7x10,
o R = 1.4 XIO6 —
Tss's
Cm – O.2 F=-|--|--
--" Dº

—O4

O
Toil-Off ~

— — —”
Cm - O.2 ~~~~- - --

- O.4

(b) Pitching-moment characteristics.

Figure 18.- Concluded.

 .24

.2O

. 16

. 12

Cy
.O8 -

.O4

O "A

8t, deg
-.O4 O O

[] 38

.O4

O ~
Cn
-.O4 º

-.O8 .O4 Tºt

Cl

-.O4
–2O -I6 -12 –8 -4 O 4
B, deg
(a) o = 0°, R = 8.7X10°, flaps 0° and 38°

Figure 19.- Lateral characteristics of basic configuration in sideslip.

56
 .24

.2O
a, deg

. º:

.O8

. O4

– O4

.O8

.O4

–2O - |6 – 12 –8 —4 O
B, deg

(b) 6f = 0°, R = 8.7X10°, a = 0°, 8.4°, and 14.6°.

Figure 19.- Continued.
57
 a , deg

|2
.O8

Cy .04

24 (
–.O4

32 (
.O4 –.O8

a, deg
Cn O
33 &

—.O4
;
.O4
a, deg
24, 32, 16 (
O
|2
C1 O

–.O4
–2O - |6 —|2 –8
B, deg
(c) 6f = 0°, R = 4.1X10°, a = 0°, 12°, 16°, 24°, 32°
Figure 19.- Concluded.
58
 O4 B–G–El TºS,º
Cy
o: *— ºt, o—o-o-º-o-º-º

O4 Sº I
–. S
.O2 – O8
*> K»

O
-
C-sizeYo G
9-5-º-º-º:
[]—E)
o” So

Cn
-O2
IT
•--- Nſ
---> H–C
/ *—l-f
B,Ceg
-O4 .O6 ~|~/ 2 :
Ó –8
—
.O4 R ºs
C1
* O2 s—ºf [] /
g/
**

º-l
~|~1. st
s E
O r N &r-o-tºe ºf *~o—o

-02; 4 8 *—#–E–F#–E–F#–E,
a, deg

(a) 6f = 0°, R = 4.1x10°, 6 = 0°, -4°, -8°

Figure 20- Lateral characteristics of basic configuration versus angle of attack.

59
 .O2

º
– O2

\ O
D
8 f, deg
O
20
– O4 N/
Ó 38

O2

—.O2

O4

O2

– O2
O 4 8 |2 |6 2O 24 28
a , deg
(b) R = 8.7x10°, 3 = 0°, 8f = 0°, 20°, 38°
60 Figure 20.- Concluded.
 .O4

ſh8JTY'
-

k
}†
At
Q
9I.O
Lateral
21.-
Figure
basic
cof
lift
versus
46f
0R
=
o
ho.1X10°,
anerfaicf°.
°,
gtiuecriasetnitoc;ns

4
–4
[]

-8
A

J//|Z
|

º
}|
!}5
-

\CL
.6 |
*}
!
|
.28, O

©
 .O8

O4

– O4

.O4
.

—. O4

O8

.O4

–04; — |6 – |2 –8 —4 O 4 8

B, deg
Figure 22.- Lateral characteristics of the model in sideslip with the empennage removed;
R = 8.7x 106.
62
 .O4

Cy

.O4 –.O6

.O2

Cn
O

—.O2 O6
Tail 8t, deg
O on O
.O4 C off O

|
|-

D on
off
40
40

.O2

Cl

§3–5% S O-O

—.O2 Ri
Nºm
O O
–.O4 |2 |6 2O 24 28 36 4O

a, deg

Figure 23.- Lateral characteristics versus angle of attack, tail on and off; 6f = 0° and 40°,
R = 4.1X 10°.

63
 O4

-15,415

.O2 2% {-100 -

Cy &=- 1.39 -15,O

o Bºz-H Něžiosº
.O2 – O2

(a) 6f = 0°, R = 4.1X10°

Figure 24.- Aileron effectiveness.

6||
 O2

V-IO

O
/
2%
[...]

zº –5
Nº -15

-A

Cy +|O s N +
– O2 TS Yst 5
Szás

—O4

.O2

Cn O
wºn

- #E=#
^–

lºsskº

– O2 O6

O4 -

Bas; deg
+|5|A -

O2 +Iok-tº--&
- + 5b- N - N A
-A

O –5 KF 4)——& :

-IO V– &– :

– O2 NO

–04; O 4 8 |2 |6 2O 24
a, deg
(b) 6f = 40°, R = 4.1x10°
Figure 24.- Continued.
65
 .O3

.O2

Ol'

.O2

.O |

.O3

.O2

.O |

(c) 6f = 0°, R = 8.7x10°

Figure 24.- Continued.

66
 .O2

* –H

—. O2

.O2

Cn o; "L

—y F}= –S

–.O2

.O4
a, deg
O .3
D 8.7
Ó 12.9
.O2

Cl

– 15 — |O O |O |5

8as deg
(d) 8 f= 38°, R = 8.7x10°
Figure 24.- Concluded.

67
 .OO4 |
:
8t, deg
O O
[] 38

.OO2
Tl

5H5++ *~

°ls, /deg
N
O b.

—.OO2
O 5 |O 15 2O
a, deg

Figure 25.- Cls for right aileron versus angle of attack (measured at ÖaR = 0°); R = 8.7x10°.
d

63
 .O8 A

__2~
-
O4 _-z-T
>+”
..~~ ~~

y __^ d, deg

- - O O
.O4) - [] 8.4
- O |6.4
— .O8

.O4

.02)

Cn O º

– .O2 ^sº
===
S

—.O4 TNº.

O2

O –––––––.
Cl
– O2 -— — — — — — ——-Q

—.O4
–3O –2O — |O O |O 2O 3O
Sr, deg

(a) 8 f= 0°, 3 = 0°, R = 4.1x10°

Figure 26.- Effect of rudder deflection on directional characteristics.

69
 |2 ~3
O8 ---
21-T *
23- _-T
Cy
O4
__T--
_º
O
3–ſ -

-
O
a, deg
O
.044- D 8.4
© 14.6
—, O 8

.O4

, O2

Cn
– ,O2 ~.
-.O4 s
~

—. O6

– 30 –2O - |O O |O 2O 3O

Sr., deg
(b) 6f = 0°, 6 = -4°, R = 8.7x106
Figure 26.- Continued.

70
 2O

_- + --~~

|2
L-2-1-T
=-- ...~" H
º
C
y
O8
|2=º
.
- ~~_{T _-T
--

.04, y

a, deg
O O
O D 8.4
K» [4.6
—.O4

.O22

– .O2 L

n *Hº S
—. O4 >>-
Sis
-.O6 sº
—.O8

.O6

.04&= -
Cl
.O2H —ſ
–––E=====
R=—=
-

236 –2O -IO O |O 2O 3O

8, deg
(c) 3 f = 0°, 6 = -8°, R = 8.7x10°
Figure 26.- Continued.

7l
 24 -

L^
-r
_-Tº-T ~%
| 6 S-T-7 --~
-" ...~r

Cy _2^_-
.O8
L^_*
f

O
a, deg
O
.O4 7 D 8.4
© 14.6
O

. O2

ots==
—. O2

Cn `ss
- O4 ==s
S
S
-. O6 >=
SS
-. O 8 sº
— . O8

Cl

–3O –2O -IO O |O 2O 3O

Sr, deg

(d) 6f = 0°, 3 = -12°, R = 8.7×10°

Figure 26.- Continued.
72
 .32 2.
.28 ~

...~"
J–~ . -

…~~ _-T
P"

_-T
L-T
_-T

a, deg
O
8.4
|4.5

`ss
—.O8
TSs
sº

— . [O

O8

2––––––––.

–3O –2O — |O O |O 2O 3O

8, , deg
(e) 8 f= 0°, 3 = -16°, R = 8.7X10°
Figure 26.- Continued. . 73
 .O4 _L-T--~~
Cy –3–2–
O }-
a, deg

-.O4 º D 8.7

-
KX 16.6
-.O8

.O4

.O2 —;
Cn
O
Fºss-s -

S
`S- ——
T-J ſ
-.O2
*-

-.O4

.O4

.O2

- |O O |O 2O 3O

8, , deg

(f) ºf = 40°, 3 = 0°, R = 4.1x10°

Figure 26.- Continued.

7||
 | 2
zzº
21-7
.O8 ==+==
_."

O4 =--- __--" - ~ •
|----f
Cy •
_-- T.
O !---
a, deg
-.O4\ -
O .3
/ [] 8.7
Ó | 2.8
—.O8

.O4

.O2 S

O -

—.O2 ->+=
*-
*s, S >
`S$ss
~sss
—.O4
*

—.O6

.O6

.O4 —º. - * = <!-- * * * * !----,

8, deg

(g) 6f = 38°, 3 = -4°, R = 8.7X10°

Figure 26.- Continued.

75
 . 16
•

|2 __{T _T

- _--T _-TT _2?

—ſ _* _*
O8 -* ~~~
Cy _2~ _P-T
.O4
Pºiº
—
---
O
a , deg
.3
O D 8.7
K» 12.8
-. O4

.O2C

Cn
-.O2 S

-.O4
SFSs
s
SSS

-.O6
=sº -

loaº
‘. . .
l —ſ"
_|_+––E===-
-

./
-
_---T

*º--
O
–3O – 20 - |O O |O 2O 3O

8, , deg *

(h) 8 f= 38°, 3 = -8°, R = 8.7x10°

Figure 26.- Continued.

76
-.O2

-.O4

—.O6

-.O8

–20 - |O O |O 2O 3O

8, , deg

(i) 8 f= 38°, 3 = -12°, R = 8.7x10°

Figure 26.- Continued.

77
 .28 D
__*
L^ [...]

|-T L-1
.2O --~~ L-T
~~ =}

|6 –C
PT -

H-T
_--T
_-T
.* --~~
.*

. I2
r-r __ →

a, deg

.O8

.O 4.
: º:

— .O4 Tss |
Sis S
N. S.S.
–.O6 >s
N- SS
—.O8
~ Ss
Sºss

-. |O
sº
.O8

.O6
-

---------,
O4 | __---→
.02+
–2O - |O O |O 2O 3O

8, , deg
(j) 6f = 38°, 3 = -16°, R = 8.7x10°
Figure 26.- Concluded.

78
 AE, deg
O | O |
8t, deg 8 f, deg
–.0005H,Rnx°,-g
|O
— Rn 40
× IO-6
4. | 4. |
—.OO || O P

—.OO || 5
*— _* !—lº
C
"3r B, deg
O | -4 |
8t, deg 8t, deg
–.oOos H 9 _ 38
Rn × IO-6 Rn × IO-6
8.7 8.7
-.OO |O

—O-H-O
-.OO | ...t== --"
-

IO 15 2O
Q
O 5 |O 15 2O
- O 5
a , deg a , deg

(a) 3 = 0° and -4°

Figure 27.- Cnór versus angle of attack for flaps up and down; R = 4.1X 10° and 8.7X106.

79
 3, deg
O I - 8 |
8t, deg 8t, deg
— .oOo.5 - 0 — 38
Rn × IO-6 Rnx IO-6
8.7 8.7
— ,OO | O

—.OO || 5
{- 2T→o {- No

0ns, B, deg
O | – 12 |
8t, deg 8t, deg
—.oOO5 H 0 — 38
Rn × IO-6 Rn x O-6
8.7 8.7
-. OO |O

— . OO || 5 O
5 |O 15 2O O 5 IO 15 2O
a , deg a, deg

(b) 3 = -8° and -12°

Figure 27.- Concluded.

8O
ſº * ºf if
º .O4
&=&#3
O—O-Otº
O
Cy

—.O4

.O4

Cn

—.O4
model
B
Oasic
Sl
SDpoiler
.O4
S2
SÓpoiler
S3
SApoiler

-
-
O
-

Q-33.32*
—.O4

—.O8
2
|6
8
O2
4
—O
–8
d
,aeg

6f
=8R°.
.7x10°,
wing;
left
on
spoilers
Outboard
0(b)

Concluded.
28.-
Figure
 N a, deg
-
O -2

.O8 N
N. Ó
[] 6
2
A |O

Cy
.O4

O Y

– O4

.O4

–O4

.O4

Cl
O
>

-O4
–|6 – 12 – 8 —4 O 4 8 |2
B, deg

Figure 29.- Lateral characteristics with landing gear down; R = 8.7X 10°, 6f = 38°.

83
 .OO3 _-O

!---T
V

.OO2

Rn × IO-6
Cn O 8.7
£3 D 4. I

.OO |

—.OO |

Cig -002 N Tº

— OO3
N. +.

—.OO4
O 4 8 |2 |6

a, deg

Figure 30.- Stability derivatives Cng and Cl6 versus angle of attack; 6 f = 0°.
84
 ...” Stollº
of
Effect

L-Tº
*sº
ho.1X10°
at
angle
downwash
Average
31.-
4Figure
=
R
tail;
8.7X10°.
and
rizontal

_T

6
&
LT
|-rº.upO.4|Flops

down
Flaps
O.29

/da
de

===
8
~5
i
4

3.
 8t, deg
. IO O O
[] 20
O 38

.O8

Che
.O6

.O4

.O2

O
–4 O 4 8 |2 |6

a, deg

(a) Right aileron hinge-moment coefficient versus angle of attack for three flap angles, ðaR = 0°.

Figure 32.- Hinge-moment coefficients; R = 8.7X10°.

86
 .4 8f , deg

Ch 0 °

— .O.5
|5 |O 5 O –5 -IO – 15
8q, deg
(b) Right aileron hinge-moment coefficient versus aileron angle for 6f = 0° (top) and 6 f = 38°
(bottom).

Figure 32.- Continued. 87

3.

|
.O

—.O2

-.O3
1
8
42
O 2O
deg
a,

0=°.
6e
angles;
flap
three
for
attack
of
angle
versus
coefficient
hinge-mom
elevator
Right
(c) ent

Continued.
32.-
Figure
 . IO

8t, deg
O

a, deg
Ch - left O O
e D 8.4

C 14.6
- O5 V

- . [O N
- . 15

. IO

8t,deg
38

a, deg
Ch., right O O.4
D 8.7
Ó 12.9
- O5

- . [O

- . 15
– 15 —IO –5 O 5 |O 15
Se, deg
(d) Elevator hinge-moment coefficient versus elevator angle for 6f- 0° (top) and 6f = 38° (bottom).

Figure 32.- Continued.

89
 . 3O

d; deg
O O
D 8

sº

— . 20

— .3O
–3O –2O - |O O |O 2O 3O

8r, deg

(e) Rudder hinge-moment coefficient versus rudder angle at two angles of attack; 8 f = 0°.
Figure 32.- Continued.

90
 .2O
B, deg 8t, deg
O – 8 O

D – 8 38

KX –|6 38
. |O A – || 6 O —

O
r’S

*hr
- NS
><

\\
NJN
N. N.
N. N.

—.2O
NNN
NY.
N

-393.5 –2O -IO O |O 2O 3O

8r deg
y

(f) Rudder hinge-moment coefficient versus rudder angle at two sideslip angles and two flap angles; o' = 0°.
Figure 32.- Continued.

91
§
.O2

deg
8t,
O
O
20
D
38
Ó
-.O4

-.O6
I6
O
|2
—
8
4

angles.
flap
three
at
attack
of
angle
versus
coefficient
hinge-moment
(g)
stabilizer
Horizontal

Continued.
32.-
Figure
 hionegfe°.
stabilizer
Horizontal
0c(h)
=
f
6
angles;
incidence
tail
two
at
angle
elevator
versus-imcoiment

deg
,
it
7.O
—
O.4

O Ó
^
Concluded.
32.-
Figure

deg
8e,
1–5
–|5O
-
O

.2O .2O
—

- * §
NAT I ONAL A E R ON AUT ICS AND SPACE A D M IST RAT I ON
WASHINGTON, D.C. 20546
POSTAGE AND FEEs PAid

NATIONAL AERONAUTICs And

OFFICIAL BUSINESS
sPACE ADM 1 Nist RATION
PENALTY FOR PRI v ATE USE $300
FIRST CLASS MAIL

0 2 1 002 C1 U 35 71 1029 SO 01, 89 H

UNIW OF ILLINOIS
LIBRARY
ATTN : DOCUMENT LIBRARIAN
URBAN A IL 61801

ruS 1 MASTER . If Undeliverable

Postal Manual) Do(Section 15
Not Retu

"The aeronautical and ſpace activitieſ of the United States ſhall be

conducted ſo aſ to contribute . . to the expanſion of human knowl
edge of phenomena in the atmoſphere and ſpace. The Administration
ſhall provide for the wideſt practicable and appropriate diſſemination
of information concerning its activitieſ and the reſultſ thereof.”
— NATIONAL AERONAUTICS AND SPACE ACT OF 1958

NASA SCIENTIFIC AND TECHNICAL PUBLICATIONS

TECHNICAL REPORTS: Scientific and TECHNICAL TRANSLATIONS: Information

technical information considered important, published in a foreign language considered
complete, and a lasting contribution to existing to merit NASA distribution in English.
knowledge.
SPECIAL PUBLICATIONS: Information
TECHNICAL NOTES: Information less broad derived from or of value to NASA activities.
in scope but nevertheless of importance as a Publications include conference proceedings,
contribution to existing knowledge. monographs, data compilations, handbooks,
sourcebooks, and special bibliographies.
TECHNICAL MEMORANDUMS:
Information receiving limited distribution TECHNOLOGY UTILIZATION
because of preliminary data, security classifica PUBLICATIONS: Information on technology
tion, or other reasons. used by NASA that may be of particular
interest in commercial and other non-aerospace
CONTRACTOR REPORTS: Scientific and
applications. Publications include Tech Briefs,
technical information generated under a NASA Technology Utilization Reports and
contract or grant and considered an important
Technology Surveys.
contribution to existing knowledge.

Details on the availability of these publications may be obtained from:

SCIENTIFIC AND TECHNICAL INFORMATION OFFICE

NATIONAL AER ON AUTICS AND SPACE ADMINISTRATION

Washington, D.C. 20546
 629.13 ---

Un325tri ENGIN

# 4s 7%

NAS A T E C H IN I C A L IN OTE NASA TN D-6574

: The Lºſ CF THE

FEB 101372
uº.
Aſ u º

EXPERIMENTAL BOUNDARY-LAYER EDGE

MACH NUMBERS FOR TWO SPACE SHUTTLE
ORBITERS AT HYPERSONIC SPEEDS

by George C. Ashby, Jr.

Langley Research Center
Hampton, Va, 23365

NATIONAL AERONAUTICS AND SPACE ADMINISTRATION - WASHINGTON, D. C. - FEBRUARY 1972

1. Report No. 2. Government Accession No. 3. Recipient's Catalog No.
NASA TN D-6574
4. Title and Subtitle 5. Report Date
EXPERIMENTAL BOUNDARY-LAYER EDGE MACH NUMBERS February 1972
FOR TWO SPACE SHUTTLE ORBITERS AT HYPERSONIC 6. Performing Organization Code
SPEEDS

7. Author(s) 8. Performing Organization Report No.

George C. Ashby, Jr. L–8030
10. Work Unit No.
9. Performing Organization Name and Address 171-07-01-01

NASA Langley Research Center 11. Contract or Grant No.

Hampton, Va. 23365

13. Type of Report and Period Covered
12. Sponsoring Agency Name and Address Technical Note

National Aeronautics and Space Administration 14. Sponsoring Agency Code

Washington, D.C. 20546
15. Supplementary Notes

16. Abstract

Pitot-pressure profiles and surface pressure measurements have been obtained at

four axial stations along the center lines of space shuttle straight-wing and delta-wing
orbiters at angles of attack from 20° to approximately 60° and Mach numbers of 20.3 (helium)
and 6.8 (air). The results show that the Mach number at the edge of the boundary layer is
best predicted by tangent-cone theory up to shock detachment.

17. Key Words (Suggested by Author(s)) 18. Distribution Statement

Hypersonic boundary layer

Unclassified – Unlimited
Aerodynamic heating
Thermal protection system
High angle of attack
19. Security Classif. (of this report) 20. Security Classif. (of this page) 21. No. of Pages 22. Price"
Unclassified Unclassified 31 $3.00

"For sale by the National Technical Information Service, Springfield, Virginia 22151
 EXPERIMENTAL BOUNDARY-LAYER EDGE MACH NUMBERS

FOR TWO SPACE SHUTTLE ORBITERS

AT HYPERSONIC SPEEDS

By George C. Ashby, Jr.

Langley Research Center

SUMMARY

Pitot-pressure profiles and surface pressure measurements have been obtained at

four axial stations along the center line (from approximately 11 nose radii to 86 nose
radii) of a space shuttle straight-wing orbiter and a delta-wing orbiter at angles of attack
from 20° to approximately 60°. The data were obtained at Mach 20.3 in helium for the
straight-wing orbiter and at both Mach 20.3 in helium and Mach 6.8 in air for the delta
wing orbiter. The free-stream Reynolds numbers per unit nose radius for Mach 20.3
and 6.8 were 2.77 x 10° and 1.93 × 104, respectively.

On the lower-surface center line, for the foregoing test conditions the boundary
layer edge Mach number was best predicted by a local tangent-cone approximation up to
shock detachment.

The surface static pressures were reasonably well predicted by tangent-cone theory
up to shock detachment and by modified Newtonian theory and generalized Newtonian
theory (where appropriate) beyond shock detachment.

INTRODUCTION

The space shuttle is a reusable space transportation system with a primary goal of
substantially reducing the high cost of earth-to-orbit launches. (See, for example,
ref. 1.) Because of the relatively high sensitivity of payload to inert weight for a
reusable system as compared with the sensitivity for expendable launch systems, it is
imperative that the inert weight requirements of each element be accurately defined. A
major component of the inert weight of a shuttle system is the thermal protection system
(TPS). Since little of the flight test or operational space missions entry data had direct
application to shuttle-type vehicles, there has been a diverse spread in the techniques
and assumptions used to calculate the TPS weight requirements. The results of refer -
ence 2 indicate that a primary factor in the determination of the heating environment is
the Mach number at the edge of the boundary layer.
 The purpose of the present study was to determine experimentally the Mach numbe.
at the edge of the boundary layer for hypersonic free-stream conditions. Pitot-pressure
surveys and surface pressure measurements were made at four stations along the center
line of a straight-wing orbiter at Mach 20.3 in helium and a delta-wing orbiter at both
Mach 20.3 in helium and Mach 6.8 in air. Angles of attack were varied from 20° to
approximately 60°. This report presents the measured conditions at the edge of the
boundary layer and compares them with values estimated by several theoretical methods.
Summary plots of some of these data are presented in reference 3.

SYMBOLS

°p,stag stagnation pressure coefficient (behind normal shock)

K Newtonian constant

Me Mach number at edge of boundary layer

M., free-stream Mach number

Pt. 2 free-stream total pressure, N/m2

Pt,3 pitot pressure behind body shock, N/m2

Ps local static pressure, N/m2

R. free-stream Reynolds number per unit nose radius

Tn nose radius of model, mm

X axial distance from model nose in nose radii

y distance normal to model surface, mm

Oy angle of attack, deg

Ö slope of surface relative to free stream, deg

 APPARATUS AND TESTS

Tunnels

Most of the tests were conducted in the Langley 22-inch helium tunnel at a Mach
number of 20.3 and a Reynolds number per unit nose radius of 2.77 × 104. Operational
characteristics of the facility and details of the contoured nozzle flow characteristics are
available in reference 4. The remainder of the tests were conducted in the Langley
11-inch hypersonic tunnel at a Mach number of 6.8 in air at a Reynolds number per unit
nose radius of 1.93 × 104. This facility is described in reference 5 and its calibration
is given in reference 6.

Models

Models of the North American Rockwell 130B straight-wing low-cross-range

orbiter (fig. 1(a)) and the Martin Marietta delta-wing high-cross-range orbiter (fig. 1(b))
were used in the investigation. The two configurations are described in detail in refer
ences 7 and 8. Since the portions rearward of the pressure orifices and the vertical
fins of the two configurations had no influence on the flow in the selected survey regions,
they were removed to enhance the tunnel operating characteristics. The portions
removed are shown by the dashed lines in figure 1. The four pressure survey stations
along the model center line are also shown in the figure.

Instrumentation

Multiple-range electrical pressure transducers were used to sense the pressures

on the model surface and at the survey probes. The static–pressure-orifice size is
given in figure 1 and the survey-probe designs are shown in figure 2. The larger pitot
probe (probe 1) was used primarily for the continuous boundary-layer sweep, whereas
the smaller probe (probe 2) was used to determine probe interference effects close to the
model surface. The static-pressure probe was designed according to the guidelines
given in reference 9. The data were recorded on strip charts in the Langley 11-inch
hypersonic tunnel and on magnetic tape in the Langley 22-inch helium tunnel. Errors in
the measured pressures normalized by the free-stream total pressure are less than
0.048 x 10-3 at M. = 20.3 and less than 0.0128 × 10-2 at M. = 6.8. The error in Mach
number based on these pressure errors would be less than 0.015.

Tests and Methods

Pitot-pressure surveys and surface pressure measurements were made on the

straight-wing and delta-wing models at Mach 20.3 in helium at nominal angles of attack
of 20°, 40°, 50°, and 56°. The nominal angles of attack were 20°, 30°, 40°, 50°, and 60°
for the delta-wing model at Mach 6.8 in air. To compare the surface static pressure
with the pressure in the boundary layer, measurements at two points within the boundary
layer at the rear survey station were made for 209 angle of attack at Mach 20.3. Fig
ure 3 is a sketch of a typical test setup in the Langley 22-inch helium tunnel, showing the
straight-wing orbiter, traverse probe, survey orifice stations, and shock system observed.
The angles of attack were measured for each set of tests. The pitot-pressure sur
veys were conducted from the model surface outward. A fouling light indicated probe
departure from the surface and a calibrated slide-wire potentiometer measured survey
distances. Data acquisition was not started until departure of the probe was indicated.
The initial data point was then taken to be zero to compute survey position. The rate of
traverse in both tunnels was slowed to within the observed pressure-lag rate. In the
Langley 11-inch hypersonic tunnel (strip charts) the traces were continuous; in the
Langley 22-inch helium tunnel (magnetic tape) the data were sampled 20 times per second
resulting in a spacing between the data readings of less than 0.050 mm. Because of this
close spacing, these data are also plotted as a continuous curve. All traverses were
made normal to the free-stream flow direction, and the survey position of the probe was
converted to normal to the model surface by the cosine of the surface slope. It should
be noted that although there was no lag in the boundary-layer pitot pressures, the probe
position relative to the model can be off as much as 0.254 mm because of the momentun
of the probe drive after the fouling light goes out. The pressures obtained with the large
and small pitot probes were essentially the same; therefore, the larger probe was used
because of the reduced settling time. Representative values from the small probe are
shown in figure 4.

RESULTS AND DISCUSSION

Pitot-Pressure Surveys

Pitot pressures measured during the tests are presented in figures 4 to 6. The
x locations shown in the figures are the probe locations when the probe is at the sur
face. From previous experience, a laminar boundary layer was expected for all the
tests and the pitot-pressure profiles are typical of laminar flow. (See, for example,
refs. 10 and 11.) There was some scatter in the measurements near the body surface
(within 0.4 mm) which is presumed to be due to probe interference. Each pitot-pressure
profile was faired to the surface static value, ignoring the scatter. The profile between
the wall and the initial probe measurement is indicated by the dashed portion of the curve.
The boundary-layer edge was assumed to occur on the flat portion of the pitot profile as
shown in the figures. The particular edge locations were selected by using the typical
profile of reference 11 as a guide. In general, the selected locations, within the survey
position accuracy (0.254 mm), conform to the constraints that the boundary-layer thick
ness increases along the body center line, the edge pitot pressure is either continuously
increasing or continuously decreasing with increasing x depending on the angle of
attack (this variation results from the combination of expanding flow and decreasing
entropy moving rearward on the body), and the thickness at a given station decreases
with increasing angle of attack. The reasonable leeway from the selected edge location
along the semiflat portion of the profile would affect the edge Mach number by a maximum
of 0.2 and, in general, less than 0.1. The boundary-layer thickness selected for the
straight-wing orbiter was in good agreement with the computed thickness for M. = 20.3
in helium and O = 20°, assuming two-dimensional flow and constant normal-shock entropy

(fig. 4(a)). These calculations were performed by using the technique presented in refer
ence 11. It should be noted that the numerical results presented in figure 4(a) are used
only to compare the boundary-layer thickness and should not be used to infer profile shape
within the boundary layer since flow divergence and variable entropy effects are neglected
in the calculations. The agreement between the measured and calculated boundary-layer
thicknesses indicates that the measured profiles are primarily boundary-layer profiles
rather than vortical profiles.

The two static–pressure measurements near the boundary-layer edge at the rear
survey station at q = 20° are also shown in figure 4(a). The excellent agreement
between the wall pressure and these two pressures indicates that the wall static pressure
could be used with the pitot pressure at the edge of the boundary layer to determine the
edge Mach number. Reference 10 also showed this result for hemisphere cylinders at
a = 0°.

Surface Static Pressures

The measured surface static pressures are compared in figure 7 with the values
estimated by modified Newtonian theory (kCp Stag instead of
-
y
2.0)
and tangent-cone

theory; good agreement is obtained with both sets of calculations at the lower angles of
attack. The modified Newtonian values underpredict the measured values as or
increases; nevertheless, they are in reasonable agreement. Generalized Newtonian
theory (ref. 12), wherein the ratios of pressure coefficients from station to station along
the body surface are equal to the ratio of the square of the sine of their respective slopes,
was applied to the delta-wing orbiter at angles of attack near 60°. For the delta-wing
model at that angle of attack, the forward portion of the orbiter has a slope 5 greater
than 61°, therefore, stagnation pressure occurs on the surface (as shown in ref. 13).

This condition established a theoretical surface pressure and a known slope from which
the other surface pressures could be computed by using the ratio of slopes.
 Boundary-Layer Edge Mach Number

The Mach number at the edge of the boundary layer determined from the measured
pitot pressure at the edge and the measured surface pressure is plotted as a function of
surface slope in figure 8. Values computed by various methods are also presented in the z
figure for comparison. The data show for both types of orbiters over the hypersonic
Mach number range that from the most forward survey station (approximately 11 nose >

radii) to the trailing edge, the edge Mach number agrees closely with the tangent-cone º

estimate up to shock detachment and is considerably higher than the estimate using
normal-shock entropy. Reference 4 shows this same result for a delta wing and a
straight body at o = 40° and Mach 8 in air. Beyond conical shock detachment, the
boundary-layer edge Mach number on the forward portion of the configuration agrees
with the estimate assuming Newtonian surface pressure and normal-shock entropy. As
the flow moves rearward on the body, the high entropy is progressively absorbed in the
boundary layer and the edge Mach number approaches oblique-shock values. A lower
bound of oblique-shock entropy, obtained by assuming that the shock is parallel to the
body surface, is shown for comparison.
A closer look at the trend of edge Mach number with angle of attack for both the *~.

straight-wing and delta-wing configurations reveals a shift from agreement with tangent
wedge theory at the lower angles of attack to agreement with tangent-cone theory at
about o = 30°. This trend is consistent with previous shock-angle and surface-pressure
results for delta wings which show tangent-cone theory to be valid at 0 = 30° and above
for hypersonic speeds. (See ref. 14.)

Implications of Results

Because the vehicle shock varies from normal at the nose to oblique along the body,
it has not been clear prior to the present tests which type of shock properties dominated
conditions at the edge of the boundary layer. Studies like references 2 and 15 investigated
the effect of having oblique-shock properties instead of normal-shock properties and found
that for turbulent flow, temperature levels on the lower surface of space shuttle configu
rations were typically 150° C to 200°C higher for oblique shock. The present results
clearly indicate that oblique-shock properties best predict the variation in edge Mach
number for hypersonic flow at the tunnel test conditions. Many questions remain
unanswered, such as What are the real-gas effects, the variable-entropy effects, and the
highly viscous flow effects during the initial portion of entry? Nonetheless, on the basis
of the present results and the importance of preventing inert weight growth in the terminal
stages of vehicle development, oblique-shock properties should be utilized in the deter
mination of boundary-layer edge conditions until these questions are answered.
 CONCLUDING REMARKS

º Pitot-pressure profiles and surface pressure measurements have been obtained at

four axial stations along the center line of a space shuttle straight-wing orbiter and a
delta-wing orbiter at angles of attack from 20° to approximately 60°. The data were
: obtained at Mach 20.3 in helium for the straight-wing orbiter and at both Mach 20.3 in
º, helium and Mach 6.8 in air for the delta-wing orbiter. The Mach number at the edge of
the boundary layer was computed by using the measured pitot pressure and the surface
static pressure.

On the lower-surface center line, for the foregoing test conditions the boundary
layer edge Mach number was best predicted by a local tangent-cone approximation up to
shock detachment.

The surface static pressures were reasonably well predicted by tangent-cone theory
up to shock detachment and by modified Newtonian theory and generalized Newtonian
theory (where appropriate) beyond shock detachment.

Langley Research Center,

National Aeronautics and Space Administration,
Hampton, Va., December 13, 1971.

:
 REFERENCES

. Mueller, George E.: The New Future for Manned Spacecraft Developments.
Astronaut. & Aeronaut., vol. 7, no. 3, Mar. 1969, pp. 24-32.
. Masek, R. V.; and Forney, J. Alan: An Analysis of Predicted Space Shuttle Temper
atures and Their Impact on Thermal Protection Systems. Vol. I of NASA Space
Shuttle Technology Conference, NASA TM X-2272, 1971, pp. 75–96.
. Johnson, Charles B. (With appendix B by George C. Ashby, Jr.): Boundary-Layer
Transition and Heating Criteria Applicable to Space Shuttle Configurations From
Flight and Ground Tests. Vol. I of NASA Space Shuttle Technology Conference,
NASA TM X-2272, 1971, pp. 97-156.
. Arrington, James P.; Joiner, Roy C., Jr.; and Henderson, Arthur, Jr.: Longitudinal
Characteristics of Several Configurations at Hypersonic Mach Numbers in Conical
and Contoured Nozzles. NASA TN D-2489, 1964.
. McLellan, Charles H.; Williams, Thomas W.; and Bertram, Mitchel H.: Investigation
of a Two-Step Nozzle in the Langley 11-Inch Hypersonic Tunnel. NACA TN 2171,
1950.

. Bertram, Mitchel H.: Exploratory Investigation of Boundary-Layer Transition on a

Hollow Cylinder at a Mach Number of 6.9. NACA Rep. 1313, 1957. (Supersedes
NACA TN 3546.)
. Stone, David R.: Aerodynamic Characteristics of a Fixed-Wing Manned Space Shuttle
Concept at a Mach Number of 6.0. NASA TM X-2049, 1970.
. McCown, James W.: A Two-Stage Fully Reusable Space Transportation System.
M-69–19, Martin Marietta Corp., Sept. 15, 1969.
. Cronvich, L. L.: Pressure Distributions Over a Cylinder With Conical or Hemi
spherical Head at Supersonic Velocities. CM-528 (NOrd Contract No. 7386), Appl.
Phys. Lab., Johns Hopkins Univ., Feb. 9, 1949.
10. Crawford, Davis H.; and McCauley, William D.: Investigation of the Laminar Aero
dynamic Heat-Transfer Characteristics of a Hemisphere-Cylinder in the Langley
11-Inch Hypersonic Tunnel at a Mach Number of 6.8. NACA Rep. 1323, 1957.
(Supersedes NACA TN 3706.)
11. Harris, Julius E.: Numerical Solution of the Equations for Compressible Laminar,
Transitional, and Turbulent Boundary Layers and Comparisons With Experimental
Data. NASA TR R-368, 1971.
12. Love, E. S.: Generalized-Newtonian Theory. J. Aero/Space Sci., vol. 26, no. 5,
May 1959, pp. 314-315.
13. Ashby, George C., Jr.; and Goldberg, Theodore J.: Application of Generalized
Newtonian Theory to Three-Dimensional Sharp-Nose Shock-Detached Bodies at
Mach 6 for Angles of Attack up to 25°. NASA TN D-2550, 1965.
14. Keyes, J. Wayne; and Ashby, George C., Jr.: Calculated and Experimental Hinge
Moments on a Trailing-Edge Flap of a 75° Swept Delta Wing at Mach 6. NASA
TN D-4268, 1967.
15. Hamilton, H. Harris: Turbulent Heating on Space Shuttle Orbiters During Reentry.
Space Transportation System Technology Symposium, NASA TM X-52876, Vol. I,
1970, pp. 463-483.
 .
were
portions
Dashed
stations.
survey
pressure
showing
models
of
Sketches
1.-
Figure

/
straight-wing
Rockwell
American
North
(a)
(130B).
orbiter
-

/ 82.2rn

tests.
these
for
removed

rin—-
4
59-

orifices)
(l,
D.
I.
mm

|-4–
ira-,
E-23.
ºr -n

1.016

_s=

- -

10.8rn}+-

3
 ~
--T
-
a2
/
orifices)
(l,
D.
I.
mm
/1.016
/
/
| |-

—é
@

\
\
\
\
\ *
~
S--\
<-
>
- *-
*~
~
S
S.
-
>

mm
~7
rn=1.78
z
>*
| /
| / |
/
/
|

rn—-
85.8

line).
base
1969
(September
delta-wing
orbiter
Marietta
(b)
Martin
Modified
Concluded.
1.-
Figure
:
§

—F-
I
T
E
-
Ia

Probe
-v
O.D.
0.635
1
O..09
tube
D.
flat
tened
to
I.D.
0.318

º:
0 0.D.
2 .508
O.
tube
D.
flat
tened
to 0.25),
o
D4rifices
90°
p
a art
25.
A
1H2.7
H-46 .5 12.7
Hºt
typical łż=>e
— —

175
3.
0.
D.
º
24
tub
p
in
nose
TAL

v=
=
−

Wedge
fairin
p We
dige
fairing

(a)
Pitot-pressure
probes. (b)
Static–pressure
probe.
Figure
Sketches
2.-
survey
of
probes. All
dimension
in
are
millimeterss.
 straight-wing
with
setup
test
typical
of
Sketch
3.-
Figure
orbiter.

25.lrn
location
orifice

8
...
n
r
and

station
Survey

a
 Fnlat-plate
(–
theory
entropy)
ormal-|
shock

2rn
=
x
l5. Bioturonatdi-agprhyte-slwaiuynreg
4.-
spFigure
on
stations
survey
four
the
at
profiles
o

/Static
p P,
pressure,
”
—-co
,
20.3
=
M.
at
orbiter
layer.
boundary
of
edge
assumed
identify
marks
Vertical
helium.
in
- 3
U
S|2-
2 |
H
2 |
O
|
L
|
l
!
L
l

21.5
=
§–3

|
==Measured my,
2
ym
mm

28.269
=
ô edge
Assumed

—
8
H
L.
B.
of

5
3
2
O
l
3
2
l

200.
=
or
(a)

...”
*
K-

’
–
6
–
6 Pt,”
L l,
l,-

1l()x10.8
=
1x
*
10
ra
00

/
./

º
 x1818xra 0−3
x=107%
0.8 lrn
25.
X -

lig.
§=
66° l;
§=
5.9°

fºre
/p
t,oo /
|

lII
Ill|O
OIl
l
3
2
O

mm
y,
,m
y m

–3
x1
x
8 0
9.l.,
5ra=39.l.
x=rn
||
Ś=
l;9
3.

p
== |
Pt,”
/
/

H-.
L
l|1
O
l3lO2
}. .y,
mm

by
shown
are
probe
small
from
values
Typical
+(b)1.4°.
o=4symbol.
Continued.
4.-
Figure

3.
3. 107
5x
x=10.8
rp
ô
=669
. –
_ 59.66
H.
l,

/
–/ 3
25.l
=
|rm
|§=55.9°
2

l |-

||II1l
OlI
l,3l
O 2

y,
mm
min
-

10–3
x
5

3 -

/rn
x=
5rm
39.l.,
9.l.
1
= 9.1°
Ś965/3.1,
7 2

1F

|0llI_IIl
,
2
l
l3O

y
,ynum
-mm.

a=51.4°.
(c)
Continued.
4.-
Figure
 17
'pºpnĮouoo – 'ſ 3.Inºț¢I
'689= 20(p)
umII º AC1u
mr i º AC
9.2Ofi82
TO
-r--→{0 £I|IT|I0
TT
dI
ooº 4
22 ºr
18 ***
|
|
Hº9
g_0Ix ſ_—−.fi
99 =o09 = Q
Ou
·
fi º 69 = x
ºa tº 64 =a
umu “A，umII º AC
9.20fi92TO
III-I-1 0 1TI|TTI| 0
TT
•“ ºg
2z grīņſ
•
| 29.T9.
__）*
-g
_0
69 º 29 = Q-t *fig_0 t *fi
392 ’99 = 9
ºr tº ga = xu r g º OT = x
 o
|
L
T
|
|
—l
l
1
I
—l
Bioetulontd-apr-yewsliauynreg in
-
5.
dpFigure
on
stations
survey
four
the
at
profiles 20.3
=
M.
at
orbiter
layer.
boundary
of
edge
assumed
identify
marks
Vertical
helium.
l
|
1
l O
—1–
|
|
l
1–

1x
2x
=
ra
18.6
1.3
x
10
0−3
07° 8x
=
ra510–3
1ra
=5.8
0−38
7.2
x
26
=
&
6.79
6.7° my
-
2
y
munum |
2
y
Iy
2
mnn.m.
20°
=
&

l,
3
O
l
2
Ol l,
3
2
l
O2
3
l

H
8
H H.
l,
|-

—
6
- 2
--
-
l,
- 2
-

20°.
=
a
(a)

—l
I6

/
L-T-
|-

10

3.
 l
=
r
x
2 8.6
=•l.3 18xTn
0−3
n
l;6.79
h6.79
=
6-

l,

—l
1
l
l –
l|
Il|-
O
2
l,O
l
2
3
Y
>mm.
>Y
mm

8x
10T 10–3
x
8
57.2
-
rn
85.8
=
x
rn
1=
|S
0°
=|,09

|
fix;
1–
l,
Pt.co
>

0|—|→
|l1–
|lI
O
2
l
,
3
2
l3
y
2mm
m
2y m

4=(b)
of0°.
C5.
–Figure
ontinued.

§
§ 8.6
=2l.3
lrh
=ra
=
5
&5 6.79
=6.7°
6
07°
1x| 56.7 -
-

2–
2––
3
/ /

o.
P.
2
fº
2
Pt,”
l
l

||l—
O l-
II|—||Il-
O lI
2
l
l,
O
3
l
O2
mm
y,
,mm
y

1x07°
07355

/F
3 8
=
x 5.8
rn
z/§=50°
2

H
l
H
l .

I
lIL
lIIL
O
O
l,
32
O
l
2
l
O
mm
y,
mm
y,

5=(c)
o0°.
C ontinued.
-5.
Figure
rm
28.6 ra
85.8
58.25°
6l.95°

Concluded.
–
5.
Figure
O
0

|
|Ol,
|
J 3
2O l,
3
O
2

58.25°.
=
a
(d)

|
1–O
|
|
ſ
Pt,”
5Tn
=
x7.2

mm
y, mm
,
y

ll.3
=
rm -

58.25°
=
S -
10=
x
l;6l.959

10T
X
l,
T/ H
2
fix; F
a
fº,2
-

Pt,•

s
§

3rn
ll.
-

3|-
26.5°
S=
|26.5°
|
|
°F)
I
l|-|
/

mm
y,
-

20°.
=
or
(a)
delta-wing
orbiter
on
stations
survey
four
the
at
Boundary-layer
profiles
pitot-pressure
6.-
Figure
layer.
boundary
of
edge
assumed
identify
marks
6.8
=
Vertical
air.
in
M.
at
 my
2
2
y
m.
mm.

|
|
|
|
lOl,
O
_l
l
1
|
|
| 3
2
l
O

Continued.
6.-
Figure

30.33°.
=
a
(b)

5Tn
=
x7.2

Pt,3
30.33°2
=
ô

mm
V,

3
2
O
l

10-2
x
l,
3 l

Pt,”

§
º

1l,x| 07°

ſ
|
-

Pt,3
2
2—l- .3
=
xll
lrn
-
ſ Pt,”
/
119
§=7.

|||I
O
O
3
2
O
mm
y,

0-8
1|x1l,x07°

Pt,3
/2
2
5ra
=7.2
x
//Pt,”
o
-
lô=;0.9
z

I|1
O
O
O
3
2
O

mm
y,

4=(c)
a0.9°.
Continued.
6.-
Figure
 2x - 57.2
rn 85.8
2x
-
*n
2 & -
o l,x151.6
&07° 51.6°

fºra
al-/
/
2H
|Pt,”
/

I
I|I
|O
OI
O
2
l
3
l

2mm
y

(d)
a=51.6°

=
57.2
*n
2.5°
|x1&=60^*

Pt,3
-
——
—2–
2
Pt,”
,”

L
||I|
O
O
l
2
3

mm
y,

(e)
a=62.5°.
Figure
Concluded.
6.-
§
§
Measured
theory
T
— angent-cone
1
=
(K.76)
theory
Newtonian
M
:—
–odified

.00l.
.00l.

=
2
q0°
p .002
- S
OO2
"[ \

||_l l
II|l—l
O
80
60
l:0
2
O0
80 O
60
l,0
2
OO

radii
nose
x,
X,

.00l.
5
= 8°
al.
519
=
Q
-

p .002
.OsO2
*t,”

|II1I
|L
l
O1–
ll|I
80 O
60
80
l,0
60
O 0
l;0
20
2
O
radii
nose
x,
radii
nose
X,

04.
.77
2x1=R.
helium;
in
0.3
M.
orbiter;
Straight-w
(a) ing
attack.
of
angles
various
at
pressures
static
-S
7. urface
Figure
 Measured
theory
T
—
— angent-cone
1(K
=
theory
Newtonian
M
- odified
- .76)
-
6
2
& 1°)
at
(C
theory
Newtonian
T-
Generalized.
“
-
stag
p,

.00l.
.00l.
2
Q; 0°
lOo
=
O

Ps
Pt,”

radii
nose
x,
radii
Se
no
X,

.00l.
.00l.
| Q0°
5
= 8.25°

Ps
Pt,”

_l
l|
O
60
30
l,0
80
20
90
10
50
70 90
80
70
60
50
radii
nose
X,
radii
nose
X,

R.
2M. 04.
.77
0.3
x1=in
helium;
orbiter;
Delta-win
(b) g
Continued.
7.-
Figure

s:
 .OO8
§ 2
= 0°
o,
.006
M
— easured
theory
——
Tangent-cone
\–
—
.00l. 1(K
—
—.81)
=theory
Newtonian
M
—
.—
—odified
*—\!
e) 1°)
(C
×6at
theory
Newtonian
eneralized
G–
| p,stag
:—
– -
-
-
-
-
--

.002

radii
nose
x,
.Olo ſ .0ll, -

=
1q,0°
=
3
q0°
.008 .Olſº
-
\– — — — —

—
:M.–
.006 .Old
-

Pt.co .008-
-
-
.00l.
-

.002 H
.006
-

.001,
|||l|!!|
||l
90
30
50
70
l,0
50
l
l;0
60
90
80
1
30
0O0o
20
20
X,
radii
nose
x,
.016 F
.020
H
5
=0°
a
—
—O—
\—
60°
=
O
H
.Ol&
-—
.Olli -

g—- -
-

*
.Ol2 — — — *; >
|--- — . . — . . — . . —
-- - - –
TOH
. ll,
–
.Old

-
.-
—
—
L– -—

.008 .
!|||||I|Il| 012
80
60
90
l;0
30
70
20
O
80
70
60
l
90 C)
l;0
50
20
30
lOC)
nose
radii
X,
radii
nose
x,

×=1R.
air;
M. in04.
.93
.8
6orbiter;
Delta-win
(c) g
C
7.oncluded.
–Figure
 radii
nose
x,

10.8
O
D
25.1
T
5 39.l.
KX
59.l.
A
A -
*

** Measured
—
\pitot-pressure
suryey
with
surface
pressure
l,H.
Tangent-cone
theory
Tangent-wedge
theory
Newtonian
with
pressure
surface
normal
–
shock
Prandtl-Meyer
and
entropy
expansion
Newtonian
pressure,
surface
H
shock
63
=
angle

|-

shock
Conical
# 2
H
detachment

H
Il
`--...

|l|
0.
O
l,0C)
l2O
30
60
50
70
deg
6,

(a)
Straight-wing
20.3
=
M.
orbiter;
helium.
in
Boundary-layer
8.-
Figure
Mach
edge
angles
at
number
to
20°
from
attack
of
58°.

§
3.
x,
nose
radii
.3
ll.
28.3
57.2
85.8

Measured
pitot-pres
survey
with sure
surface
pressure
Tangent-cone
theory
Tangent-wedge
theory
Newtonian
surface
pressure
with
normal
shock
entropy
Prandtl-Me
and
expansion yer
Newtonian
surface
pressure,
shock
angle
6=

/Conical
shock
ydetachment

lC) 2O
30 l,0
60
50
70
deg
6,

(b)
Delta-wing
M.
=orbiter; 20.3
in
helium.

Figure
8.-
Continued.
 >
X nose
radii

: ll;
.3
28.3
57.2
85.8
S -

pitot-pres
Measured sure
survey
with
surface
pressure
** Tangent-co
theory ne
Tangent-we
theory dge
Newtonian
surface
pressure
with
normal
shock
entropy
and
Prandtl-Me
expansion yer
Newtonian
surface
pressure,
shock
angle
6=

Conical
shock
*
N detachment

Y
Ses
``s
`s N -

-
Sºjº
|||I|lI||
lC)
2O
l,O
30
60
50
70
6,
deg

Delta-win
(c)
orbiter; g M.
=.8
in
6air.

: 8.-
Concluded
Figure .
NATIONAL AERONAUTICS AND SPACE A D M is TRATION
WASHINGTON, D.C. 20546
POSTAGE AND FEEs PAid
NATIONAL AERONAUTICs And
OFFICIAL EUSINESS space ADMinistration
FIRST CLASS MAIL
PENALTY FOR PRivate use $300

02 c1 U 35 7.20121 S00
OF ILLINOIS

|
. If Undeliverable (Section 158
POSTMASTER: ;... Manual) Do Not Return

"The aeronautical and ſpace activitieſ of the United Stateſ ſhall be

conducted ſo as to contribute . . to the expansion of human knowl
edge of phenomena in the atmosphere and space. The Administration
ſhall provide for the widest practicable and appropriate dissemination
of information concerning its activities and the results thereof.”
— NATIONAL AERONAUTICS AND SPACE ACT OF 1958

NASA SCIENTIFIC AND TECHNICAL PUBLICATIONS

TECHNICAL REPORTS: Scientific and TECHNICAL TRANSLATIONS: Information

technical information considered important, published in a foreign language considered
complete, and a lasting contribution to existing to merit NASA distribution in English.
knowledge.
SPECIAL PUBLICATIONS: Information
TECHNICAL NOTES: Information less broad derived from or of value to NASA activities.
in scope but nevertheless of importance as a Publications include conference proceedings,
contribution to existing knowledge. monographs, data compilations, handbooks,
sourcebooks, and special bibliographies.
TECHNICAL MEMORANDUMS:
Information receiving limited distribution TECHNOLOGY UTILIZATION
because of preliminary data, security classifica PUBLICATIONS: Information on technology
tion, or other reasons. used by NASA that may be of particular
interest in commercial and other non-aerospace
CONTRACTOR REPORTS: Scientific and applications. Publications include Tech Briefs,
technical information generated under a NASA Technology Utilization Reports and
contract or grant and considered an important
Technology Surveys.
contribution to existing knowledge.

Details on the availability of these publications may be obtained from:

scienTIFIC AND TECHNICAL INFORMATION office

NATIONAL AERONAUTICS AND SPACE ADMINISTRATION
Washington, D.C. 20546
 629.13 ENGIN
UnS25trf -

z: 4S 73

NASA. T E C H IN I C A L IN OTE

:
THE LIBRARY OF THE

JAN 3 1972
UNIVERSITY OF ILLINOIs
AT URBANA-CHAMPAIGN

SUMMARY OF SPIN TECHNOLOGY

AS RELATED TO LIGHT
GENERAL-AVIATION AIRPLANES

by James S. Bowman, Jr.

Langley Research Center
Hampton, Va. 23365
NATIONAL AERONAUTICS AND SPACE ADMINISTRATION - WASHINGTON, D. C. - DECEMBER 1971
 1. Report No. 2. Government Accession No. 3. Recipient's Catalog No.
NASA TN D-6575
4. Title and Subtitle 5. Report Date
SUMMARY OF SPIN TECHNOLOGY AS RELATED TO LIGHT December 1971

GENERAL-AVIATION AIRPLANES 6.. Performing

Perf -
Organization Code
- - -

7. Author(s) 8. Performing Organization Report No.

James S. Bowman, Jr. L–7952

10. Work Unit No.

9. Performing Organization Name and Address 136–62-02-03

NASA Langley Research Center 11. Contract or Grant No.

Hampton, Va. 23365

13. Type of Report and Period Covered
12. Sponsoring Agency Name and Address Technical Note

National Aeronautics and Space Administration 14. Sponsoring Agency Code

Washington, D.C. 20546

. Supplementary Notes

16. Abstract

A summary has been made of all NASA (and NACA) research and experience related
to the spin and recovery characteristics of light personal-owner-type general-aviation air
planes. Very little of the research deals with light general-aviation airplanes as such, but
many of the airplanes and models tested before and during World War II were similar to
present-day light general-aviation airplanes with regard to the factors that are important
in spinning. The present paper is based mainly on the results of spin-tunnel tests of free
spinning dynamically scaled models of about 100 different airplane designs and, whenever
possible, includes correlation with full-scale spin tests. The research results are dis
cussed in terms of airplane design considerations and the proper use of controls for recovery.

17. Key Words (Suggested by Author(s)) 18. Distribution Statement

Spinning Unclassified – Unlimited

General aviation
Light-airplane spinning
Tail design for recovery
Factors affecting spins

19. Security Classif. (of this report) 20. Security Classif. (of this page) 21. No. of Pages 22. Price"
Unclassified Unclassified 34 $3.00

For sale by the National Technical Information Service, Springfield, Virginia 22151
 CONTENTS

P
*
INTRODUCTION . . .

SYMBOLS . . . . . .

THE SPIN . . . . . .

SIGNIFICANT FACTORS . .
Mass Distribution .
Relative Density . .
Tail Configuration . e - - - - -

Criterion for spin recovery . . .

Rudder effectiveness 10

Elevator effectiveness
Antispin fillets . . . 12

Wentral and dorsal fins 12

External Wing Tanks 13

Aerodynamic effects 13
Mass effects . . - - - - - e. e. 13

Wing Trailing-Edge Flaps and Landing Gear . . . . . . . . . . 13

Wing Position . . . 14

Tail Length . . . . 14

Center-of-Gravity Position . 15

Power . . . . . . . . . . . . - - - - 15

CONCLUSIONS . . . . 16

REFERENCES . . . . 18

FIGURES - e e e e e e 20

iii
 SUMMARY OF SPIN TECHNOLOGY AS RELATED TO

LIGHT GENERAL-AVIATION AIRPLANES

By James S. Bowman, Jr.

Langley Research Center

SUMMARY

A summary has been made of all NASA (and NACA) research and experience related
to the spin and recovery characteristics of light personal-owner-type general-aviation
airplanes. Very little of the research deals with light general-aviation airplanes as such,
but many of the airplanes and models tested before and during World War II were similar
to present-day light general-aviation airplanes with regard to the factors that are impor
tant in spinning. The present paper is based mainly on the results of spin-tunnel tests
of free-spinning dynamically scaled models of about 100 different airplane designs and,
whenever possible, includes correlation with full-scale spin tests. The research results
are discussed in terms of airplane design considerations and the proper use of controls
for recovery.

Three factors are found to be of almost overriding importance in spinning for this
type of airplane. These factors are the relative distribution of the mass between the wing
and fuselage, the density of the airplane relative to that of the air, and the tail design.
The mass distribution and relative density determine the tail-design requirements and
the control movements required for recovery. An empirically determined design factor
is available as a guide for the design of the tail to insure good spin recovery. The rud
der is generally regarded as the primary recovery control. The elevator can be very
effective in some cases, such as positive (wing-heavy) loadings or recovery during the
incipient spin, but it might prove to be ineffective for fully developed spins, flat spins, or
cases in which the mass distribution or center-of-gravity position has been changed.

INTRODUCTION

The technology of spinning seems to receive little attention from most people asso
ciated with airplanes — from design to operation — because it is not a normal part of the
operation of most airplanes. Most general-aviation airplanes are no longer required to
be able to recover from a fully developed spin (ref. 1), and spin training is no longer
required for a private pilot's license. These factors, and many more, have led to a gen
eral lack of understanding of the basic principles of Spinning. Consequently, a crisis
usually develops when a new design is involved in a spin crash or when an old design has
a series of spin accidents. In either case, the design is usually so fixed that the optimum
design change to improve the spin-recovery characteristics involves so much time and
money that it is ruled out in favor of a minimum, less expensive modification which is
less desirable.

The purpose of the present paper is to summarize findings of the NASA (and NACA)
research that relates to the spinning of general-aviation aircraft. This summary is
intended to be sufficiently detailed to help the designer build safer airplanes by giving
adequate treatment to spin recovery early in the design stage, and yet sufficiently gen
eral to help pilots and operators have a better understanding of spinning so that they may
better cope with spin problems that occur with their airplanes. Most of the applicable
research was performed before and during World War II and was not performed on
general-aviation airplanes as such, but many of the airplanes and models tested during
this period were similar to present-day general-aviation airplanes with regard to factors
that are important in spinning. From these tests the effects of many pertinent design
features were determined. This work is analyzed herein with regard to present-day light
general-aviation airplanes and is updated with more recent spin experience applicable to
this class of airplane, practically all of which is fragmentary and unpublished. The class
of airplane toward which this summary report is directed is the personal-owner aircraft
of less than about 1800 kg (4000 pounds) gross weight. The analysis is made, however,
in terms of nondimensional parameters so that it may be more broadly applicable.

SYMBOLS

b wing span, m (ft)

F force, N (lb)

Ix, Iy moments of inertia about X- and Y-axis, respectively, kg-m? (slug-ft2)

IX - IY . .. -

- inertia yawing-moment parameter

mb2

L distance from center of gravity of airplane to centroid of fuselage area SF,

m (ft)

L1 distance from center of gravity of airplane to centroid of rudder area SR1,

m (ft)
 distance from center of gravity of airplane to centroid of rudder area SR2,
m (ft)

airplane mass, kg (slugs)

spin radius, m (ft)

wing area, m” (ſt?)

fuselage side area under horizontal tail, m2 (ft2)

unshielded rudder area above horizontal tail, m2 (ft2)

unshielded rudder area below horizontal tail, m2 (ft2)

TDPF tail-damping power factor

TDR tail-damping ratio

URVC unshielded-rudder volume coefficient

weight, kg (lb)

X,Y,Z longitudinal, lateral, and vertical body axis of airplane, respectively

angle of attack, deg

relative-density factor, m/pSb

air density, kg/m3 (slugs/ft3)

angle between Y body axis and horizontal measured in vertical plane,

positive when right wing is down for erect spins, deg

airplane spin rate, turns/sec

THE SPIN

The spin has been defined as a motion in which an airplane in flight at some angle
of attack between the stall and 900 descends rapidly towards the earth while rotating about

3
a vertical axis. (See ref. 2.) The spinning motion is very complicated and involves
simultaneous rolling, yawing, and pitching while the airplane is at high angles of attack
and sideslip. Since it involves separated flows in the region beyond the stall, the aero
dynamic characteristics of the airplane are very nonlinear and time dependent; and hence,
at the present time, the spin is not very amenable to theoretical analyses.
The overall spin maneuver can be considered to consist of three phases: the incip
ient spin, the developed spin, and the recovery. An illustration of the various phases of
the spinning motion is given in figure 1.

The incipient spin occurs from the time the airplane stalls and rotation starts until
the spin axis becomes vertical or nearly vertical. During this time the airplane flight
path is changing from horizontal to vertical, and the spin rotation is increasing from
zero to the fully developed spin rate. The incipient spin usually occurs rapidly for light
airplanes (4 to 6 seconds, approximately) and consists of approximately the first two
turns. As indicated by full-scale tests and by the model tests of reference 3, the typical
incipient-spin motion starts during the stall with a roll-off. Then, as the nose drops,
the yawing motion begins to build up. About the half-turn point, the airplane is pointed
almost straight down but the angle of attack is usually above that of the stall because of
the inclined flight path. (See fig. 1.) As the one-turn point is approached, the nose comes
back up and the angle of attack continues to increase. As the airplane continues to rotate
into the second turn, the flight path becomes more nearly vertical, and the pitching,
rolling, and yawing motions become more repeatable and approach those of the fully
developed spin.

In the developed spin the attitude, angles, and motions of the airplane are some
what repeatable from turn to turn, and the flight path is approximately vertical. The
spin is maintained by a balance between the aerodynamic and inertia forces and moments.
The spinning motion is made up of rotation about the airplane center of gravity plus trans
latory motion of the center of gravity; however, it is primarily a rotary motion and is
affected mainly by the moments acting on it. A typical example of an airplane spinning
motion and the forces in a spin is illustrated in figure 2.

The third phase, the recovery, is caused by a change in the moments so as to upset
the balance between the aerodynamic and inertia moments. Such a change in the moments
is obtained by deflecting the controls of the airplane. The specific control movements
required in any particular airplane depend on certain mass and aerodynamic characteris
tics, which are discussed in the subsequent sections of this paper.

SIGNIFICANT FACTORS

Reference 2 is a summary paper in which many of the factors that affect spin and
recovery are discussed. It affords much useful background information which is of

4
interest with regard to the present problem, but it is oriented mainly toward modern high
performance military airplanes. The present paper, on the other hand, identifies and
discusses the factors that are of particular significance with regard to the light airplane.
The picture that will evolve in the discussion is that three principal factors are of
almost overriding importance in the Spinning of light general-aviation airplanes: the
relative distribution of the mass of the airplane between the wing and fuselage, the den
sity of the airplane relative to the density of the air, and the tail configuration of the air
plane. The relative density is generally fixed by performance considerations and cannot
be accommodated to spin requirements. Of the other two factors, mass distribution is
very important because it determines the control movements required for recovery and
together with the relative density, it determines the tail-design requirements for recov
ery. The tail design is important because it must have certain features to provide the
aerodynamic moments required for recovery and to damp the spinning rotation and also
because it is the factor that can most easily be controlled by the designer, particularly
in the latter stages of the design and development of the airplane.

Mass Distribution

The way in which the mass of an airplane is distributed between the wing and fuse
lage is the most important single factor in spinning because it determines the way in
which the airplane, while spinning, responds to control movements, especially to eleva
tors and ailerons. An airplane rotating in a spin can be considered to be a large gyro
scope. Since there are mass and angular rotation about all three axes, inertia moments
are produced about all three axes. In addition, aerodynamic forces and moments are
acting on the airplane because of its motion through the air. The inertia forces and
moments are opposite in sign to the aerodynamic forces and moments and are both equal
and opposite for an equilibrium spin condition. An example of the aerodynamic and iner
tia moments balanced in pitch is illustrated in figure 3. Perhaps the clearest example
of this balance is that for a wing-level spin, the nose-down aerodynamic pitching moment
must be exactly balanced by the nose-up inertia pitching moment. In order for the air
plane to recover from the spin, the equilibrium must be broken, and this is normally
accomplished by changing the aerodynamic moment by moving a control or combination
of controls that can cause the greatest antispin moment.

The mass distribution of all airplanes (general aviation, military fighters, bombers,
etc.) can be grouped into three general loading categories, as shown in figure 4. The
Iv - I
X - 17
mass distribution of the airplane is evaluated in terms of the parameter , which
mb?
has been found to be a normalizing factor and which is nondimensional so that it is inde
pendent of the size and weight of the airplane. This parameter is important in deter
mining the inertia yawing moment, which is a controlling factor in a spin, and is
commonly called the inertia yawing-moment parameter. When the weight of the airplane
is distributed mainly along the wing, the moment of inertia in roll is greater than that in
pitch, and the value of this mass-distribution parameter is positive. This situation is
referred to as a positive or wing-heavy loading, and features such as wing-mounted
engines and tip tanks contribute to such a loading. Conversely, when the weight of the
airplane is distributed mainly along the fuselage, the moment of inertia in pitch is greater
than that in roll, and the value of the mass-distribution parameter is negative. This situ
ation is referred to as a negative or fuselage-heavy loading, and features such as fuselage
located engines, fuel, luggage, and cargo contribute to such a loading. Almost all light
general-aviation airplanes actually fall into the zero loading category of figure 4, where
the moments of inertia in roll and pitch are about equal. However, there are some excep
tions, especially when heavy tip tanks are installed on the wings. The zero loading range
is generally considered to be the range between values of -50 × 10-4 and 50 x 10-4 for the
inertia yawing-moment parameter. When the difference between the rolling and pitching
moments of inertia is this small, the inertias contribute little, or nothing, to the recovery.
The loading of the airplane dictates the control movements required for recovery.
(See refs. 2 and 4 to 7.) Deflection of the rudder to oppose the spinning rotation directly
is always recommended, but in many cases, it is not adequate to provide recovery. For
the wing-heavy loadings, down elevator is the primary recovery control. For fuselage
heavy loadings, the aileron is the primary recovery control; the aileron should be deflected
with the spin, for example, stick right for a right spin. For the zero loading, the rudder
is always an important control for spin recovery. Therefore, any airplane in this loading
condition should have a rudder designed for effective spin recovery. Determining the
elevator effectiveness in the zero loading range is difficult, especially where the "zero
loading" tends toward fuselage-heavy loadings. However, in a subsequent section on tail
design, the effectiveness of the rudder and, to some extent, the elevator is discussed, and
the conditions under which the rudder and elevator would be more effective for a given
loading are also discussed.

Relative Density

The relative-density factor is an indication of the density of the airplane relative

to the density of the air in which it is flying. The formula for computing the relative
density factor is m/pSb. The relative density is fixed by design requirements and varies
according to changes in the gross weight and altitude, which are generally very small for
light general-aviation airplanes. Therefore, it cannot be adjusted to accommodate spin
requirements. However, the particular value of relative-density factor of an airplane is
significant and does have an appreciable influence on the spin recovery. The variation
of relative density with wing loading and wing span is given in figure 5 for sea-level air
density. The values would be about 50 percent higher for an altitude of 4000 meters

6
(13 000 feet). Airplanes with high relative-density factors normally require more rudder
and elevator effectiveness for spin recovery than airplanes with low relative-density fac
tors (other factors being equal). The normal change in the gross weight due to passengers
and fuel does not usually change the relative-density factor significantly. On the other
hand, a steady increase in the gross weight over a period of years, due to an increase in
performance and load-carrying ability of a particular airplane model, could appreciably
increase the relative-density factor. Such an increase could cause greater rudder effec
tiveness to be required for spin recovery. On the basis of the results of model tests to
determine the effects of relative density (refs. 8 and 9), the number of turns for recovery
would be expected to increase as the relative density increases for a given airplane.
Therefore, if there is an appreciable increase in the gross weight of an airplane due to
periodic model changes, the tail design should be reexamined to determine if it is still
adequate for satisfactory spin recovery.

Tail Configuration

Criterion for spin recovery.- The tail configuration is a very important factor in
the spin and recovery characteristics of airplanes, especially for light airplanes that are
in the zero or near-zero loading range, where the rudder is a primary recovery control.
(See refs. 9 to 15.) A relatively large moment is needed to recover an airplane from a
spin, especially a flat spin; therefore, it is important that the airplane control surfaces,
particularly the rudder, be effective at spin attitudes. The special problem is that during
a spin, much of the rudder usually is in the stalled wake of the horizontal tail and some
times the wing, over which the dynamic pressure is low or abnormal airflow conditions
exist.

A sketch illustrating the factors which are important in the tail configuration for
spin recovery is given in figure 6. This figure illustrates the dead-air region over much
of the vertical tail, which is caused by the stalled wake of the horizontal tail and which
seriously decreases the effectiveness of the rudder. In order to have good rudder effec
tiveness, a substantial part of the rudder must be outside the horizontal-tail wake.
Another important, but less obvious, consideration is that the fixed area beneath the hori
zontal tail be sufficient to damp the spinning motion, since it has been found (ref. 10) that
this area contributes much of the damping of the spinning rotation.
The criteria for tail design for spin recovery were determined many years ago from
spin-tunnel tests of about 100 different models. In setting up the tail-design requirements,
the factors considered were the inertia yawing-moment parameter, the airplane relative
density factor, and the tail-damping power factor. Both the inertia yawing-moment
parameter and the airplane relative-density factor have previously been discussed. The
tail-damping power factor is an empirically determined parameter based on various
geometric properties of the vertical and horizontal tail which have been found to relate
to the observed spin and recovery characteristics. Its value is an indication of the effec
tiveness of the overall tail configuration in terminating a spin.

The method of computing the tail-damping power factor is given in reference 9 and
is discussed with particular reference to light airplanes in reference 10. For the reader's
convenience, the tail-damping power factor is discussed again in the present report, and
illustrations are given in figure 7 to show the method of computation. As indicated in fig
ure 6, the rudder must have a substantial amount of area outside the horizontal-tail wake
in order to be effective, and also, the fuselage must have a substantial amount of area
under the tail in order to provide damping of the spinning rotation. When converted to
coefficient form, the unshielded rudder area multiplied by its moment arm from the cen
ter of gravity is referred to as the unshielded-rudder volume coefficient, and the fuselage
side area under the horizontal tail multiplied by the square of its moment arm is referred
to as the tail-damping ratio. These two coefficients are used to calculate the tail-damping
power factor. When the concept of tail-damping power factor was being formulated, some
method had to be devised to define the position and extent of the wake of the horizontal
tail (fig. 7). An analysis of the model results at that time showed that if the tail-damping
ratio coefficient was less than 0.019, the spin angle of attack (relative wind) could be
assumed to be 459 and a wake boundary could be assumed to be defined by the 30° and 60°
lines of figure 7. If the tail-damping ratio coefficient was greater than 0.019, the spin
angle of attack (relative wind) could be assumed to be 309 and the wake boundary could be
assumed to be defined by the 15° and 45° lines of figure 7.
A particularly important point brought out by the form of the equation for tail
damping power factor is that both the fixed area beneath the horizontal tail and the
unshielded rudder area are required to give significant values of this parameter. The
reason for this situation is that the damping provided by the fixed area is required to
steepen and slow the equilibrium spin, and rudder power is required to provide the change
in moment necessary to effect a recovery.
A summary of the tail-design requirements for insuring satisfactory recovery is
presented in figure 8. This figure gives the boundaries for satisfactory spin recovery
for aircraft which have relative-density factors for values from 6 to 35 and for a range
of inertia yawing-moment parameters from -280 × 10-4 to 120 × 10-4. These criteria
should not be used for airplanes that have values of relative-density factor or inertia
yawing-moment parameter outside these limits. Regions of satisfactory and unsatisfac
tory recovery characteristics are given for recovery by rudder reversal alone and for
recovery by simultaneous rudder and elevator reversal. The recovery characteristics
for a given airplane are considered unsatisfactory if the tail-damping power factor falls
below the boundary line for the relative-density factor for that airplane.
 For very lightweight airplanes, the aerodynamic contribution to the recovery moment
can be much larger than for corresponding heavier airplanes, and consequently, a smaller .
tail-damping power factor may be required for spin recovery. The tail-damping power
factor required for satisfactory spin recovery for these aircraft is plotted to a larger
scale in figure 9. This figure gives the boundaries for satisfactory spin recovery for
airplanes which have values of the relative-density factor of 6 and 10 and for a range of
values of the inertia yawing-moment parameter from -120 × 10-4 to 120 x 10-4 (ref. 10).
As the inertia yawing-moment parameter increases in the positive direction (weight
increased along the wings), the required tail-damping power factor increases for satis
factory recovery by rudder alone. However, at the same time, the effectiveness of the
elevator increases for recovery. Therefore, if recovery is attempted by rudder reversal
followed by elevators down (for zero and wing-heavy loadings), the required tail-damping
power factor could be smaller.
Caution should be used in relying on the elevators alone for spin recovery. It is
important to point out that most of the models used in the tests to determine the bound
aries in figure 9 had large elevators with large trailing-edge down deflections (elevator
leading edge at about 50 percent chord of the horizontal tail and down deflections of 159
to 20°). Therefore, when correlation of present-day airplanes is made with the bound
aries given in figure 9, the elevator size and down deflection should be considered.
Another factor to consider is the effect of center of gravity. Experience has shown that
in most cases, an airplane will spin flatter as the center of gravity is moved rearward.
When the elevator alone is relied on to provide recovery (for fuselage-heavy loadings),
the elevator effectiveness usually decreases at the flatter spin attitudes and in many cases
has been demonstrated to be completely ineffective for spin recovery. Because of the
rather indeterminate nature of some of these factors relating to the effectiveness of the
elevator for recovery, it is recommended that, as a factor of safety, a sufficiently large
tail-damping power factor be provided in the original design so that recovery by rudder
alone can be obtained without the use of the elevators.

In some cases, it has been found that simultaneous reversal of the rudder and eleva
tor gives unsatisfactory recovery characteristics, whereas the reversal of the rudder
alone with the elevator up gives satisfactory recoveries. This result is believed to be
due to the rudder being shielded by the downward movement of the elevator. For this
reason, it is recommended that, when both the rudder and the elevator are reversed for
recovery, the rudder should be reversed first, and then about one-half to one turn later,
the elevator should be deflected down. This technique was proved in full-scale spin tests
conducted on several airplanes by the NACA in 1935 (ref. 14), and the results from these
tests led to the so-called NACA recommended spin-recovery technique: Briskly move
the rudder to full against the spin; after the lapse of appreciable time (approximately
one-half turn), briskly move the elevator to approximately full down, and hold these con
trols until the recovery is complete. It is important to note that when these results were
obtained in 1935, the airplanes of that day probably were in the zero loading condition
previously discussed and today this recovery technique would apply only for airplanes
that have similar loadings. As previously pointed out, the control technique required for
spin recovery is primarily dictated by the mass distribution in the airplane. Therefore,
for airplanes of different loading conditions, this control technique recommended in 1935
would probably not apply.

Rudder effectiveness.- In general, two types of rudders are used on general-aviation

airplanes today: full-length rudders and partial-length rudders. Full-length rudders
extend to the bottom of the fuselage, whereas partial-length rudders generally terminate
at or above the top of the fuselage. (See figs. 6 and 7.) Regardless of the design, how
ever, the rudder should provide an adequate tail-damping power factor for good spin
recovery. In general, the optimum horizontal-tail position to provide the maximum
unshielded rudder area is different for partial-length and full-length rudders.
For full-length rudders, the part of the rudder below the horizontal tail provides
most of the unshielded area. Therefore, high and forward positions of the horizontal tail
alºe usually the most effective configurations for spin recovery for designs employing

full-length rudders.
For partial-length rudders, all the rudder is above the horizontal tail, and the top
part of the rudder provides most of the unshielded area; therefore, low and rearward
positions of the horizontal tail are most effective to provide the needed unshielded rudder
area. For example, in reference 16, test results indicated that spin-recovery character
istics would become worse as the center of gravity moved rearward, but just how much
worse depended on the tail-damping power factor and the position of the horizontal tail
on the vertical tail. Low values of tail-damping power factor and high horizontal-tail
positions (for partial-length rudders) were adverse to recoveries and had about the same
effect as moving the center of gravity rearward. It is believed that the low horizontal
tail positions unshielded more of the rudder and were thereby favorable to recoveries.

One particular point that should be recognized with regard to tail design is that with
a low horizontal tail and a sweptback vertical tail, it is possible that almost the entire
vertical tail, including the rudder, might be in the stalled wake of the horizontal tail.
Such a tail design is characteristic of some modern light general-aviation airplanes and
would have approximately zero tail-damping power factor. This does not imply that
recovery from the spin would be impossible, since the elevator would have some effect,
particularly in the incipient spin. But the certainty of recovery would be jeopardized
because of both the foregoing and the following qualifications with regard to the use of the
elevator for recovery.

10
 Elevator effectiveness. - For airplanes with partial-length rudders and often for
full-length rudders with a low horizontal tail, the rudder is usually mostly shielded by the
horizontal tail and is, consequently, ineffective for spin recovery. Therefore, the eleva
tor is relied on for most of the spin recovery. Even so, an almost universal control
technique suggested for recovery is rudder reversal followed by deflection of the elevators
to neutral or down. Because most light general-aviation airplanes are required in the
spin demonstrations of reference 1 to rotate only one turn before recovery attempt, this
technique is usually successful, provided that recovery is attempted before one turn is
completed. However, in many cases, it would be disastrous for the airplane to inadver
tently wind up more than one turn because this technique may not recover the airplane if
the spin has developed to two or more turns. The widespread random success of using
the elevator as the main recovery control has led many persons to a false sense of secur
ity. Down elevator is almost always assumed to be able to recover. Consequently, the
vertical-tail and rudder designs required for good spin-recovery characteristics are sel
dom considered. The reason that the effectiveness of the elevator for spin recovery
decreases as the spin progresses beyond one turn involves many factors. Normally, a
light airplane does not attain an equilibrium or balanced spin condition until after approxi
mately two turns. During this time (before the two-turn spin point) the spin is somewhat
slower and the average angle of attack is lower, both of which lead to the type of spin
mode from which recovery is easier than from faster rotating or flatter spins. There
fore, the consequences of relying on the elevators alone may be fatal in a marginal situa
tion. If the airplane loading is assumed to be near zero or wing heavy, where the eleva
tors should be effective for spin recovery, the actual effectiveness of the elevators depends
on such factors as the angle of attack, the tail length, the elevator size, the maximum down
deflection angle, the spin rate, and the tail-damping power factor. As the angle of attack
increases, the effectiveness of the elevator to produce a nose-down moment decreases
(ref. 17); while at the same time, the amount of nose-down moment required for recovery
increases because the spin is flatter.

Results from tests of full-scale airplanes have shown that the spin attitude can have
a pronounced influence on the effectiveness of the elevator for spin recovery. In several
documented spin test programs, good and rapid recoveries were obtained by rudder rever
sal and down elevator from spins that were steep and typical of median or forward center
of-gravity positions. However, poor recoveries, or no recoveries at all, were obtained
from the flatter spins resulting from rearward center-of-gravity positions. It was shown
in one case that the rudder was completely ineffective for spin recovery from any spin
mode, steep or flat, because the rudder was shielded, and that the elevator was serving
as the primary recovery control. This condition was satisfactory for steep spins but was
ineffective for flatter spins. These results again illustrate the importance of a good
rudder in a tail design for spin recovery.

11
 Antispin fillets.- The purpose of antispin fillets is to increase the damping of the
tail, which causes the spin rate to decrease and thereby cause the airplane to spin steeper.
The characteristics of a typical antispin fillet are shown in figure 10. The effectiveness
of antispin fillets for improving recovery characteristics of a given airplane generally
depends on the tail-damping power factor, the relative density, and the mass distribution
of the airplane. Generally, when an improvement is seen, the antispin fillets cause the
airplane to spin at a steeper angle, where the recovery characteristics are better. On
the basis of the results presented in reference 18, however, the addition of antispin fillets
seems to offer only a slight improvement in the recovery characteristics, regardless of
the relative density and the mass distribution. Therefore, if the recovery characteris–
tics are on the borderline between satisfactory and unsatisfactory, the addition of antispin
fillets might make a noticeable improvement in the spin and, consequently, the recovery
characteristics. On the other hand, if the tail-damping power factor is well below that
required for satisfactory recovery, any small improvement offered by the antispin fillets
is not expected to be noticeable, and the recovery characteristics may still be unsatisfac
tory. Therefore, if a large improvement is needed in the recovery characteristics of a
given airplane, the use of antispin fillets is not expected to offer any appreciable
assistance.

The presence of antispin fillets in a tail configuration is important in computing the

tail-damping power factor. In order to compute the tail-damping ratio, the length of the
fillets is used in the determination of the fuselage side area beneath the horizontal tail.
However, the fillets are assumed not to be wide enough to affect the wake above the hori
zontal tail and, therefore, are not considered in computing the unshielded-rudder volume
coefficient.

Ventral and dorsal fins.- A typical ventral and dorsal fin configuration is shown in
figure 10. The effectiveness of a ventral fin in improving the spin and recovery charac
teristics of airplanes generally depends on the tail-damping power factor and the relative
density. In general, the ventral fin causes the airplane to spin slightly steeper because
of the increased tail damping caused by the increased fixed area beneath the horizontal |
tail and, therefore, causes some improvement in the recovery characteristics. On the
basis of the results obtained in references 16 and 19, the use of a ventral fin can be effec
tive in improving the recovery characteristics if a small improvement is required to |
make the airplane recovery satisfactory. However, if a large improvement is needed, the
addition of a ventral fin is expected to offer little or no help. Of course, if the basic prob
lem is little or no unshielded rudder, the addition of a ventral fin is not expected to offer |
any improvement since the ventral fin affects only the tail-damping ratio. In such a case,
if the ventral fin increases the value of the tail-damping ratio, even by a large factor, |
little or no effect is expected.

12
 The addition of a dorsal fin to the vertical fin is expected, from the concept of
tail-damping power factor, to have little or no effect on the spin and recovery character
istics, and such has been found to be the case. Tests conducted on the effects of dorsal
fins on spin recovery are given in reference 18.

External Wing Tanks

External wing tanks can have two effects on the spin and recovery characteristics
of an airplane. One is the aerodynamic effect that may occur because of the size and
shape of the tanks, and the other is a mass effect which is due to the weight and location
of the tanks and fuel.

Aerodynamic effects.- In general, the aerodynamic effects of a tank on the spin and
recovery are small and are not noticeable unless the tanks are very large in comparison
with the airplane. However, some effects have been seen on the spin-entry characteris
tics of military aircraft, especially those with underslung wing tanks. The effects have
been observed in flight tests and are evidenced by a decrease in stability, which causes
the airplane to be more prone to enter a spin. This same type of effect might be expected
on light airplanes with tanks mounted under the wings. The aerodynamic effect of tip
tanks is even less well established; but, in any event, the airplane may be more prone to
enter a spin if the tanks, regardless of position, cause a decrease in stability.
Mass effects.- The mass effects of the tanks and fuel can be very pronounced,
especially if the tanks are on the wing tips. Additional weight on the wings can change
the loading and the technique needed for recovery. As previously pointed out and illus
trated in figure 4, the primary recovery control is dependent on the loading distribution.
As indicated in the figure, for the zero range loading, the rudder is the primary recovery
control. However, as the weight increases along the wings (loading changes in the posi
tive direction), the elevators become the primary recovery control. Note that the effec
tiveness of the rudder decreases (tail-damping power factor required for recovery
increases) and the effectiveness of the elevator increases as weight increases along the
wings. The recovery characteristics of an airplane with tip tanks can, therefore, change
markedly with the fuel load, and particular caution should be taken to note possible large
changes in the loading from the negative to the positive range (fig. 4), where the primary
control for spin recovery would change from the rudder to the elevator. If wing tanks
cause the airplane to be loaded in the positive direction, the elevators should be large
enough with adequate down deflection to provide satisfactory spin recovery.

Wing Trailing-Edge Flaps and Landing Gear

On the basis of research conducted in references 17 and 20, the use of landing flaps
would be expected to have an adverse effect on the spin and recovery. The extension of

13
flaps usually causes the spin to be flatter and the spin rate to be slightly slower. In
addition, the results of reference 17 show that the effectiveness of the rudder for spin
recovery can decrease when the flaps are down. The wake behind the wing is believed
to be larger when the flaps are down than when they are up, and thereby the tail is more
likely to be in a region of reduced air velocity. These results are for low-wing airplanes
and are not necessarily expected to apply to high-wing airplanes, where the tail surfaces
are farther from the wake of the wing.
Extension of the landing gear usually has little effect on the spin and recovery
characteristics (ref. 20), but slight adverse effects have been seen from lowering the
landing gear on some airplanes. Therefore, it is generally recommended that the gear
be kept in the retracted position when possible.

Wing Position

The position of the wing (high or low) is believed to have some influence on the spin
and recovery characteristics of airplanes. There are no documented data to provide a
technical analysis of what the effects may be, but the history of stall/spin problems asso
ciated with high- or low-wing airplanes indicates that a high-wing airplane is expected
to have better spin and recovery characteristics than a low-wing airplane, all other fac
tors being equal. The reason for the apparent improvement in spin characteristics of
the high-wing airplane is believed to be related to the higher dihedral effect caused by
the high-wing position and to improvement in the wake characteristics of the wing in the
vicinity of the tail. The wake from a high wing is believed to pass above the tail so that
the tail surfaces are not appreciably affected and the rudder and elevators are more
effective in the spin recovery.

Tail Length

Tail length can have an appreciable effect on the spin and recovery characteristics
of an airplane. Tail length is generally expressed nondimensionally as the ratio of the
distance between the center of gravity and the rudder hinge line to the wing span. On
the basis of the results of studies made in references 17 and 19, the recovery character
istics of an airplane are influenced to a much greater extent by the tail length than is
indicated by the increase in the tail-damping power factor due to tail length. In one case,
for example, the recovery characteristics for a long-tail model were satisfactory for a
tail-damping power factor of 395 × 10-6, whereas for the same model with a shorter tail,
the tail-damping power factor had to be increased to about 520 × 10-6 before satisfactory
recoveries could be obtained. The general effect of increasing the tail length is to cause
the airplane to Spin at a lower angle of attack and at a higher rate of rotation, whereas
the shorter tail will cause the airplane to spin flatter and at a slower rate of rotation.

14
The results of tests conducted in reference 17 indicate that the effectiveness of the rud

der in producing a yawing moment is much greater at lower angles of attack than at higher
angles of attack. Also, the effectiveness of the horizontal tail in producing a nose-down -

moment increases when the horizontal tail is moved rearward (increased tail length).
The implication of these results is that strong consideration should be given to designing
an airplane with as long a tail as possible in order to begin with a basic design most con
ducive to good recovery characteristics.

Center-of-Gravity Position
The center-of-gravity position can significantly affect the spin and recovery char
acteristics of an airplane. (See refs. 15 and 21.) Usually, the effect is unpredictable and
is dependent on the tail-damping power factor and other characteristics of the airplane.
For this reason, tests are normally required to determine the effect of center-of-gravity
position for a given airplane. In general, however, the airplane usually spins flatter
as the center of gravity is moved rearward. (See ref. 16.) This result is, of course,
adverse since the control effectiveness normally decreases as the airplane spin angle of
attack increases. Whether or not the controls become ineffective for recovery of the air
plane in the flatter attitudes cannot be determined by any empirical methods known by the
author and must be determined experimentally for a given airplane.

Power

The effect of applying symmetrical power in a spin is believed to be insignificant.

A number of varied observations have been made through the years by many people, and
the conclusions regarding power effects on spins range from favorable to adverse. Since
there has been no systematic study of the effects of applying power during a spin, random
visual observations and sparse results constitute all the data available.

In some of the observed results, pilots reported a definite aid to recovery when
applying symmetrical power (single-engine airplanes), and in other reports, pilots have
observed a definite adverse effect when applying power. In almost all cases, the results
were not obtained under controlled conditions. Therefore, the type of spin, the center of
gravity, the angle of attack, the spin rate, and the line of thrust with respect to the center
of gravity were not identified.
In a few cases where the effect of thrust has been measured under controlled con
ditions (ref. 22), the application of thrust had no effect unless the thrust axis was dis
placed from the center of gravity and thereby produced a moment. In these tests, both
favorable and adverse effects were observed, depending on the type of moment produced
and the loading condition of the model being tested. Since thrust effects can be adverse

15
and unpredictable, it is generally recommended that for a single-engine airplane, the
throttle be retarded to the idle position during a spin.

For asymmetric power for a twin-engine configuration with the engines mounted on
the wings, power from only one engine can produce a large asymmetric yawing moment,
which will be favorable or adverse to the spin and recovery, depending on the direction
of the moment. Both model and full-scale spin-test results of multiengine airplane
designs have shown that power on the outboard engine (e.g., the right engine in a left
spin) can create a large prospin yawing moment, which can cause a flatter and faster
spin. On the other hand, power on the inboard engine can create an antispin moment to
aid spin recovery. Normally, the manipulation of thrust can be confusing and disastrous
if the power is applied to the wrong engine. Therefore, unless asymmetric power is
necessary to aid recovery, it is generally recommended that for a multiengine airplane,
the throttle be retarded to the idle position on all engines during a spin.

CONCLUSIONS

A summary has been made of all NASA (and NACA) research and experience related
to the spin and recovery characteristics of light personal-owner-type general-aviation
airplanes. Very little of the research deals with light general-aviation airplanes as such,
but many of the airplanes and models tested before and during World War II were similar
to present-day light general-aviation airplanes with regard to the factors that are impor
tant in spinning. The following conclusions were drawn from a summary and analysis of
all the information related to light general-aviation airplanes, and it should be noted that
they do not apply to heavy, high-density airplanes, such as small transports and jet
airplanes:

1. Three factors are of almost overriding importance with regard to spin and
recovery characteristics:

(a) The relative distribution of the mass of the airplane between the wing
and fuselage, which is commonly expressed in terms of the inertia yawing-moment
parameter, a nondimensional factor relating the rolling and pitching moments of
inertia

(b) The tail configuration, which must provide damping for the spinning
rotation and the rudder power for recovery and which is commonly evaluated in
terms of an empirically determined tail-damping power factor
(c) The density of the airplane relative to the density of the air, which is
commonly expressed in terms of the relative-density factor

16
 2. The mass distribution and the relative density determine the tail configuration
requirements and the control technique required for recovery. The relative density is
generally fixed by performance requirements and cannot be adjusted to accommodate the
spin.

3. An empirically determined factor, called the tail-damping power factor, based

on tests of over 100 designs is available as a guide for the design of the tail to insure
good spin recovery.

4. The rudder is generally the principal recovery control, but for positive (wing
heavy) loadings or for recovery during the incipient spin, the elevator can also be an
important recovery control and can reduce the rudder power requirements. Experience
has shown, however, that relying on the elevator is dangerous because it might become
ineffective for fully developed spins, flat spins, or cases in which the mass distribution
has been changed or the center of gravity has been moved behind the normal rearward
limit because of changes in loading of the airplane due to growth or operational factors.
5. Significant secondary factors which might affect the spin are center-of-gravity
position, wing position (high or low), tip tanks, and asymmetric power.

Langley Research Center,

National Aeronautics and Space Administration,
Hampton, Va., November 5, 1971.

17
 REFERENCES

. Anon.: Airplane Airworthiness – Normal, Utility, Acrobatic, and Restricted Purpose

Categories. Pt. 3 of Civil Air Regulations, Civil Aero. Board, U.S. Dep. Commerce,
Nov. 1, 1949.
. Neihouse, Anshal I.; Klinar, Walter J.; and Scher, Stanley H.: Status of Spin Research
for Recent Airplane Designs. NASA TR R-57, 1960. (Supersedes NACA
RM L57F12.)
. Stone, Ralph W., Jr.; Garner, William G.; and Gale, Lawrence J.: Study of Motion of
Model of Personal-Owner or Liaison Airplane Through the Stall and Into the
Incipient Spin by Means of a Free-Flight Testing Technique. NACA TN 2923,
1953.

. Neihouse, A.I.: The Aileron as an Aid to Recovery From the Spin. NACA TN 776,
1940.

. Neihouse, A.I.: A Mass-Distribution Criterion for Predicting the Effect of Control

Manipulation on the Recovery From a Spin. NACA WR L-168, 1942. (Formerly
NACA ARR, Aug. 1942.)
. Neihouse, Anshal I.: The Effect of Variations in Moments of Inertia on Spin and
Recovery Characteristics of a Single-Engine Low-Wing Monoplane With Various
Tail Arrangements, Including a Twin Tail. NACA TN 1575, 1948.
. Bowman, James S.: Airplane Spinning. Astronaut & Aeronaut., vol. 4, no. 3, Mar.
1966, pp. 64–67.
. Seidman, Oscar; and Neihouse, A. I.: Free-Spinning Wind-Tunnel Tests of a Low
Wing Monoplane With Systematic Changes in Wings and Tails. W. Effect of Air
plane Relative Density. NACA Rep. 691, 1940.
. Neihouse, Anshal I.; Lichtenstein, Jacob H.; and Pepoon, Philip W.: Tail-Design
Requirements for Satisfactory Spin Recovery. NACA TN 1045, 1946.
10. Neihouse, A.I.: Tail-Design Requirements for Satisfactory Spin Recovery for
Personal-Owner-Type Light Airplanes. NACA TN 1329, 1947.
11. Bamber, M. J.; and Zimmerman, C. H.: Effect of Stabilizer Location Upon Pitching
and Yawing Moments in Spins as Shown by Tests With the Spinning Balance. NACA
TN 474, 1933.
12. Seidman, Oscar; and Neihouse, A. I.: Free-Spinning Wind-Tunnel Tests of a Low
Wing Monoplane With Systematic Changes in Wings and Tails. I. Basic Loading
Condition. NACA TN 608, 1937.

18
13. Bihrle, William, Jr.: Floating Characteristics of Rudders and Elevators in Spinning
Attitudes as Determined From Hinge-Moment-Coefficient Data With Application to
Personal-Owner-Type Airplanes. NACA TN 2016, 1950.
14. McAvoy, W. H.: Piloting Technique for Recovery From Spins. NACA TN 555, 1936.
15. Zimmerman, C. H.: Effect of Changes in Tail Arrangement Upon the Spinning of a
Low-Wing Monoplane Model. NACA TN 570, 1936.
16. Klinar, Walter J.; and Wilson, Jack H.: Spin-Tunnel Investigation of the Effects of
Mass and Dimensional Variations on the Spinning Characteristics of a Low-Wing
Single-Vertical-Tail Model Typical of Personal-Owner Airplanes. NACA TN 2352,
1951.

17. Stone, Ralph W., Jr.; Burk, Sanger M., Jr.; and Bihrle, William, Jr.: The Aerody
namic Forces and Moments on a 1/10-Scale Model of a Fighter Airplane in
Spinning Attitudes as Measured on a Rotary Balance in the Langley 20-Foot Free
Spinning Tunnel. NACA TN 2181, 1950.
18. Gale, Lawrence J.; and Jones, Ira P., Jr.: Effects of Antispin Fillets and Dorsal
Fins on the Spin and Recovery Characteristics of Airplanes as Determined From
Free-Spinning-Tunnel Tests. NACA TN 1779, 1948.
19. Klinar, Walter J.; and Snyder, Thomas L.: Influence of Tail Length Upon the Spin
Recovery Characteristics of a Trainer-Type-Airplane Model. NACA TN 1764,
1948.

20. Gale, Lawrence J.: Effect of Landing Flaps and Landing Gear on the Spin and
Recovery Characteristics of Airplanes. NACA TN 1643, 1948.
21. Seidman, Oscar; and Neihouse, A.I.: Free-Spinning Wind-Tunnel Tests of a Low
Wing Monoplane With Systematic Changes in Wings and Tails. IV. Effect of
Center-of-Gravity Location. NACA Rep. 672, 1939.
22. Lee, Henry A.; and Libbey, Charles E.: Incipient- and Developed-Spin and Recovery
Characteristics of a Modern High-Speed Fighter Design With Low Aspect Ratio as
Determined From Dynamic-Model Tests. NASA TN D-956, 1961.

19
 |

l Incipient
spin

Spin axis — Developed

y Spin

-N

Spin radius —- º

l Recovery

9| Nº.
KL) *s.

(a) Complete spin, stall through recovery.

Figure 1.- Illustration of spinning motion.

20
 21
-
J
·dd
-- -I
IO (q)
**

*
 2–
Sº
tº

*
*

-–
- #

y
--
SS
* 4. t

-- - –––

(c) Developed spin phase.

Figure 1.- Continued.

22
 A - -- - - - - - -

º Hº

". k

s
-
(d) Recovery phase.
Figure 1.- Concluded.

23
 Rodius of spin
-Spin axis

4–wind direction

L-70-7552
Figure 2.- Balance of forces in a spin.

24
 Aerodynamic
moments moments
Inertia

Spin
Spin
axis
axis

Aerodynamic Inertia
moment,
pitching moment,
pitching
down
nose nose
up

Airflow

3.-
Figure
aerodynamic
of
Balance and
inertia
moments
pitching
spin.
a
in

§
§

Ailerons
With
ElevatorS
DOWn
Rudder
Against
Plus
Plus
Followed
By
Rudder
Against
ElevatorS
DOWn
/
-

|
|

. |

-
|

|
-
|

Loading
Fuselage-Heavy
Zero
Wing-Heav
Loading y
-

~(Roll
>
PInertia
Inertia)
Kitch

Figure
Primary
4.-
recovery
controls
as
determine
mass
by
distributiodn.
 160

140
–
28

120
H
24

i# H
20
100

H
12 60

40

20

12
1
911
8
70
m
b,
Span,
||J||I
40
38
36
34
32
30
28
26
24
22

ſt
b,
Span,

personal-
light
for
span
and
loading
wing
with
factor
relative-d
of
Variation
5.- ensity
Figure owner-ty
density.
air
Sea-level
airplanes.
general-aviation
s:
 shielded
º-Rudder

regionDead-air
Dead-air
region Thairalc-tdereisitgcns
6.-
cFigure
recovery.
spin
for

axis
Spin
-

not
Rudder
Shielded Airflow

§
 Stabilizer
wake

L
L2-
×.019
0TDR
<.019
of
Angle
ofAngle
(q)
wind
relative
Wind
45°
be
to
aSSumed
30°
C.g.
TO
º C.9.
airplane
of
airplane
of
60°
00
|50

-Relative
wind
c.g.
Relative
wind
T0.04.
.To -

L
of
airplane

SF UDR
=RVC
L2XTTDPF
L.,
So,
S
+
L
RI
TI
°2
R2
TDR
—
=
2
-
-
URVC
=
L+ H
SR
|2
*R2
SF
s(b/24
S(b/2)
TDPF
=
X
2
S(b/2)

Figure
Method
7.-
computing
of
tail-damping
factor.
power

§
 Key
Satisfactory
Recovery by rudder alone
W
- - - - - - Recovery by rudder and elevator
Unsatisfactory
1000
x ſo-6

ll-OO

l2OO

É
lCOO

8OO

600 20 < u < 55

15 - 1 - 20 +
liOO

10 < 1 < 15

2OO

–280 –210 –200 —160 —l2O –80 –0 O liO 8O 120 x 10+

IX - IY
Inertia yawing-moment parameter. 2
mb

Figure 8.- Tail-design requirements for airplanes having relative-density

factors from 6 to 35.

30
 Satisfactory
Key
%
rudder
by
Recovery
alone
—
Recovery
—
rudder
by
Unsatisfactory
elevator
and
800
x10-6

700

600

500

200

100

x
Tºo
0-"
8O
160
100
10
-60
-10
-80
2 0
-100
-l2O
-20
O

IX
-Iy
yawing-moment
Inertia
parameter,
º
Tail-design
9.-
Figure
requirements
relative-density
having
airplanes
for
of
factors
6and
10.

:
 Antispin fillet N

DOrSal fin N

Ventral fin

Figure 10.- Typical tail designs showing antispin fillets

and dorsal and ventral fins.

32 NASA-Langley, 1971 – 2 L–7952

NATIONAL A E R ON AUT | CS AND SPACE A D M is TRAT I ON
WASHINGTON, D.C. 20546
Post AGE AND FEEs PAid
NATIONAL AERONAUTics And
OFFICIAL BUSINESS
space: A D Minist RATION
PENALTY FOR PRiva TE use $300
FIRST CLASS MAIL

020 on 2 cl U 35 7 112 l 7 SO 04 89H

UNIV (YF I LL INſ) IS
L I BRARY
AT T N : Dnct JMENT LIBRARIAN
UP RANA IL 6 1801

POSTMASTER:. If Undeliverable
Postal Manual) Do ;:
(Section 1:
N.

"The aeronautical and ſpace activitieſ of the United States ſhall be

conducted ſo as to contribute . . . to the expanſion of human knowl
edge of phenomena in the atmoſphere and space. The Adminiſtration
shall provide for the wideſt practicable and appropriate diſſemination
of information concerning its activitieſ and the reſult, thereof.”
— NATIONAL AERONAUTICS AND SPACE ACT OF 1958

NASA SCIENTIFIC AND TECHNICAL PUBLICATIONS

TECHNICAL REPORTS: Scientific and TECHNICAL TRANSLATIONS: Information

technical information considered important, published in a foreign language considered
complete, and a lasting contribution to existing to merit NASA distribution in English.
knowledge.
SPECIAL PUBLICATIONS: Information
TECHNICAL NOTES: Information less broad derived from or of value to NASA activities.
in scope but nevertheless of importance as a Publications include conference proceedings,
contribution to existing knowledge. monographs, data compilations, handbooks,
sourcebooks, and special bibliographies.
TECHNICAL MEMORANDUMS:
Information receiving limited distribution TECHNOLOGY UTILIZATION
because of preliminary data, security classifica PUBLICATIONS: Information on technology
tion, or other reasons. used by NASA that may be of particular
interest in commercial and other non-aerospace
CONTRACTOR REPORTS: Scientific and
applications. Publications include Tech Briefs,
technical information generated under a NASA Technology Utilization Reports and
contract or grant and considered an important
Technology Surveys.
contribution to existing knowledge.

Details on the availability of these publications may be obtained from:

SCIENTIFIC AND TECHNICAL INFORMATION OFFICE

NATIONAL AERO NAUTICS AND SPACE ADMINISTRATION

Washington, D.C. 20546
629.13
UnS25tri

* As 7%

NASA. T E C H IN I C A L IN OTE

: THE LIBRARY OF THE

JAN 3 1972
UNIVERSITY OF ILLINOIS
AT URBANA-CHAMPAIGN

ROTATIONAL-TEMPERATURE
DETERMINATION IN FLOWING
NITROGEN USING AN ELECTRON BEAM

by D. C. Lillicrap and L. P. Lee

Langley Research Center

Hampton, Va. 23365

NATIONAL AERONAUTICS AND SPACE ADMINISTRATION - WASHINGTON, D. C. . DECEMBER 1971

 ROTATIONAL-TEMPERATURE DETERMINATION IN FLOWING

NITROGEN USING AN ELECTRON BEAM

By D. C. Lillicrap” and Louise P. Lee

Langley Research Center

SUMMARY

The measurement of rotational temperature by the electron-beam fluorescence

technique is investigated for temperatures between 78 K and 300 K by using the 0–0 band
of the nitrogen first negative system. Existing methods for calculating the rotational
temperature from electron-beam spectra result in values which are higher than the true
temperature and the error increases with density and electron-beam current. The
errors are believed to arise through neglect of the effects of secondary electrons.
Excitation by low-energy secondary electrons can increase the population of the sparsely
populated rotational ground states at the expense of the abundantly populated states. This
secondary excitation is proposed as the cause of the error in the calculated rotational
temperature and a theoretical study determines that the density and beam current govern
the excitation rate. Rotational-temperature measurements are made in a flowing stream
of nitrogen at a known temperature for a range of densities and beam currents. The
product of density and beam current correlates the present data, and least-squares cubic
polynomials are fitted through the temperature data as a function of this correlating
parameter. An iterative procedure based on the polynomials allows more accurate rota
tional temperatures and, consequently, densities to be obtained with the electron-beam
fluorescence technique. This procedure can be applied to measurements in some high
speed flow regimes.

INTRODUCTION

As flight speeds are extended into the hypersonic regime, high recovery tempera
tures are encountered and can result in low-density regions within the flow field. The
study of these regions with conventional probes may lead to unacceptably large errors.
For instance, in high-speed low-density flows, corrections to pressure measurements
for probe interference and orifice effects are large and difficult to apply, particularly in
boundary layers close to the surface. Consequently, electron beams are used extensively
*NASA-NRC Research Associate, presently affiliated with Imperial College,
London, England.
to study this flow regime since they have no appreciable effect on the flow, and measure
ments can be made to within a few mean free paths of the surface. (See ref. 1.)
Interference-free measurements of this kind would be particularly valuable in the study
of high Mach number turbulent boundary layers. In the low-density region close to the
surface, it has been found (refs. 2 and 3) that the static pressure appears to increase
significantly as the surface is approached. Wallace (ref. 4) used an electron-beam probe
to determine mean and fluctuating densities in a turbulent boundary layer but the results
were preliminary and temperature profiles were not measured. The potentially greater
accuracy of interference-free electron-beam measurements of both density and tempera
ture could provide a better understanding of the low-density regions of turbulent boundary
layers.

The electron-beam fluorescence technique (ref. 5) consists of directing a collimated

beam of moderate energy electrons through the flow field and, by use of a suitable optical
system, of observing the fluorescence produced. This paper considers only density and
rotational-temperature measurements in pure nitrogen. The intensity of the first nega
tive bands can be related to the density and, if the rotational fine structure of a band is
resolved, a rotational temperature can be obtained from the relative line intensities. For
most conditions of interest, the rotational temperature equals the translational tempera
ture and, in principle, the state of the gas can be determined at any point in the flow with
out appreciably disturbing the flow. (See ref. 1.) In practice, the rotational temperature
calculated by using the equation derived by Muntz (ref. 5) is too high and the discrepancy
increases with density and the number of spectral lines used (refs. 6 and 7). This dis
crepancy also leads to errors in determining the number density for conditions where
quenching of the excited states through collisions is appreciable. (See ref. 8.)
The objective of this paper is to examine the possible causes of the high calculated
rotational temperatures and hence remove some of the uncertainties associated with den
sity and rotational-temperature measurements, particularly at the upper end of the density
range where electron-beam techniques are applicable, on the order of 1023 molecules
per meter*. Excitation by low-energy secondary electrons is investigated, since it could

result in a redistribution of the molecules among the ground-state rotational levels. An

excitation model is proposed and the parameters affecting the calculated temperature are
identified. The equations describing this excitation model involve unknown constants, and
an experimental calibration of the technique for temperatures between 78 K and 300 K is
performed for a range of densities and beam currents. An iterative procedure based on
tabulated experimental results is developed to obtain more reliable values of rotational
temperature and consequently density, provided certain conditions are satisfied.
 SYMBOLS

A,B,C constants

Bv; rotational constant for v. vibrational level

Co,C 1,02-03 cubic polynomial coefficients

C mean molecular velocity

CE experimental constant

C speed of light

F excitation function

F(K") defined by equation (19)

f(v) secondary-electron velocity distribution function

G defined by equation (2)

g(v,42)k, excitation cross section for secondary electrons

H defined by equation (7)

h Planck's constant

I(k,k)-I(K)-(K.K.) intensity of K' rotational line

*(0,0) intensity of R branch of 0–0 band

i beam current

J defined by equation (8)

j current density

K rotational quantum number

K' rotational quantum number for N.B22 State

K" rotational quantum number for N2x12 State

K. rotational quantum number for Nºx2x. State

Kmax maximum number of spectral lines used

k Boltzmann constant

N nitrogen number density

Nk. number density of K" level

N;B” upper electronic state of nitrogen ion N:

N;x? ground state of nitrogen ion N;

Nax' ground state of nitrogen molecule N2

ns secondary electron number density

Pp relative rotational transition probability for P branch

PR relative rotational transition probability for R branch

p beam diameter or field-of-view width

QR rotational State sum

Q. V" • V'1 Franck–Condon factor for transition between V' and V',1 vibrational levels
of two electronic states

R(T) = (*).º (ºft:- K

R}… net rate of secondary-electron excitation of K" level

 rate of excitation of K" level by one transition

temperature

rotational temperature

wall temperature (equals static temperature)

vibrational quantum number for N;B”: State

vibrational quantum number for Nax's State

electron velocity

reference electron velocity

X(K") defined by equation (17)

A(K) change in rotational quantum number

A(K") change in upper state rotational quantum number

A(K") change in ground-state rotational quantum number

AI(K") difference between experimental and theoretical rotational line intensities

rarefaction parameter

defined by equation (9)

secondary-electron mean free path

wave number of emitted radiation

qb rate of excitation defined by equation (5)

q = Bvſh; k
Subscripts:

calc calculated values defined by equation (11)

exp experimental values

M Maxwellian distribution

Inax maximum value

min minimum value

|
1,2 iteration approximations

THEORETICAL EXCITATION MODELS

Previous Models

To obtain the rotational temperature, Muntz (ref. 5) used a theoretical model to

relate the measured rotational line intensities for the N; first negative bands to the
undisturbed ground-state rotational distribution. He argued that primary electrons obeye.
the optical dipole selection rules (AK = +1) in exciting the ground-state molecules directl
to the N;B%. state. This state decays spontaneously to the N;x*x state and gives
rise to the first negative system. If it is assumed that no other processes significantly
affect the relative rotational line intensities, the excitation model leads to the equation

I(K", K") = Constant v4K'Ge -K (k+1)/TR d

where

G = (K' + 1)e -2(K+1)*/TR + K'e 2K" d/

(b/T
R (2)
2K' + 1

An iterative procedure is used to solve these equations for TR, but the values
obtained are too high. Robben and Talbot (ref. 9) first derived a correction curve for
TR) as a function of the temperature but subsequently, it was shown that the error
calc
n (TR) also increased with density. (See refs. 6 and 7.) Secondary electrons,
calc

roduced when the primary electrons ionized ground-state molecules, were suggested
refs. 10 and 11) as the cause of the high calculated temperatures. Some secondaries
ave sufficient energy to excite ground-state molecules to the N;B%. state, but for
hese low electron energies, transitions for which AK = +1, +3, +5, . . . can occur. This
ondition results in high calculated temperatures, but order-of-magnitude calculations
how that there are insufficient secondaries with the required energy to account for the
arge errors in (TR) . Furthermore, any appreciable direct excitation of the
calc
2
1.B 2 states by secondaries would result in a nonlinear variation of band intensity
with density, even when quenching is negligible, as can be seen from reference 10. For
ow densities, less than 1022 molecules per meterº, the band intensity varies linearly
with density, and Maguire (ref. 7) used this variation to argue against secondary electron
xcitation in favor of preferential quenching of the rotational states. For this model to
work, the quenching cross sections for the rotational states have to be a complicated
unction of the temperature and some states need to be strongly quenched even for den
ities less than 1022 molecules per meter*. Yet another approach to this problem is
he double excitation model proposed by Hickman (ref. 12) which allows transitions for
AK = +1 and +3. Although the calculated temperatures are lower, this model offers no
‘eal advantage over Muntz's model.

Present Excitation Model

When a beam of moderate energy electrons passes through nitrogen, some of the
round-state molecules are ionized. The ions and secondary electrons diffuse out of the
!eam region, but the electrostatic field set up around the beam retards the electrons and
‘esults in high concentrations of secondary electrons. Beinkowski and Harbour (ref. 13),
ising a formalism similar to that of Self and Ewald (ref. 14) for a discharge column,
alculated values several orders of magnitude larger than those derived previously with
he electrostatic forces neglected. For a 10-3 meter beam radius and a 10-8-ampere
Yeam current in a gas at a density of 3.2 × 1022 molecules per meterº, the estimated
secondary-electron concentration was 4 × 10 17 per meterº. Harbour, Bienkowski,
und Smith (ref. 15) subsequently measured concentrations close to the predicted values
und an electron temperature of about 1 eV. For electrons of this energy, the cross sec
ion for excitation of ground-state rotational levels with a selection rule AK" = +2 is
approximately 10-19 meter2. (See ref. 16.) In view of the large secondary-electron
concentrations and excitation cross sections, the effect on the ground-state rotational
distribution function must be considered. The population of a particular ground state K"
is influenced by the net result of four transitions between it and the levels K" = +2, as
illustrated in figure 1. These secondary transitions will cause a redistribution of the
molecules among the ground-state rotational levels; thus, the population of the sparsely
populated levels is increased at the expense of the more abundantly populated levels.
This condition produces the effects observed by Maguire (ref. 7) without the need to
assume that the abundantly populated N;Bºx states are preferentially quenched. The
redistribution of the molecules will not appreciably change the total intensity of the first
negative bands, and at low densities (less than 1022 molecules per meterº), this model
predicts a linear variation of intensity with density, as observed experimentally. There
fore, the objections to earlier modifications of Muntz's model (ref. 5) do not apply to
this model, but further comparison with experiment requires a detailed study of the effect |
of the secondary electron excitation on the relative line intensities. |
Figure 1 illustrates the complete excitation model by showing the transitions |
affecting the seventh rotational line. The K' = 7 level is populated by primary elec
trons from the K" = 8 and K" = 6 levels. Each of these levels is influenced by tran- |
sitions excited by secondary electrons between them and adjacent even-numbered lines.

UPPER STATE
K" |
N.'Bºz V - 0

GROUND STATE
N2 XX V" - 0
K" | ` AK-4]

|0 PRIMARY ELECTRONS
AK=#|

SECONDARY ELECTRONS
- - - -- AK"=+2

*-i-

N(K")

Figure l. - Excitation model showing transitions for I(K' = 7).

To include the secondary-electron excitation, the perturbation of a particular K" level
is assumed to be proportional to the net rate of excitation of that level RK". Then,

k = (ºk), + CR." (3)

where C is a constant and the subscript M denotes the population for a Maxwellian
distribution.

Order-of-magnitude calculations based on equation (3) predict an overpopulation of

the higher rotational states compared with a Maxwellian distribution. Calculations are
performed for unit volume occupied by a beam of diameter p, with the assumption that
molecules enter and leave the beam only by thermal diffusion through the edge of the
beam. The difference between the rate of diffusion out of the beam for a particular
K" level, and the rate of diffusion into the beam for the same level is the rate of excita
tion of that level per unit volume per second by the secondary electrons. It can be shown
that

from which

77 t N. 1.

#– 1-#– (4)
*(Nk), (Nk),
The expression for Rk.
is developed later and depends on the density, secondary elec
tron concentration, secondary electron energy, and corresponding excitation cross sec
tions. Cross sections for 1 eV electrons were estimated to be two orders of magnitude
greater than the values listed in reference 16 for 0.14 eV and are of the order of
10-19 meter2. By using a secondary electron concentration ns of 4 × 1017 per meterº
from reference 13, equation (4) predicts a considerable overpopulation of the high rota
tional states. For example, at 78 K an overpopulation of approximately 30 percent is
predicted for the eleventh rotational level. These order-of-magnitude calculations
demonstrate that a more complete study of the proposed secondary-excitation mechanism
is required.

To study the effect of secondary electrons in detail, the excitation by low-energy

secondaries must be included in the equation for the relative rotational line intensities.
Based on the Muntz method (ref. 5), the rate of excitation by primary electrons from
a. v. vibrational band to a K' level of the V' band of N;B*: is

p(V",K') - FN). N., . P., + N., . P

K"11* P T *"K"-1" R q(v.V.)
1
(5)

vº
*7

2(ºr + Nk-IPR)
K'
* r

per second per unit volume. In equation (5), F is the excitation function for an electron
defined so that if all molecules are in the ground state, then FNi molecules are excited
to the N;B*: states per second per unit volume. (In the previously used notation
(ref. 8), F was defined so that Fi molecules were excited to the N;B*: states per
second per unit volume. As a result, the so-called constants in the equations of ref. 8
involve the density.) For a Maxwellian distribution

(8k),
" ... \ =
N isºr),
2k" , 1 e
-K"(K"+1)*/TR

where QR (TR) 1-
is the rotational State sum for the v; level. Therefore, from
v;
equation (3)

-K"(K"+1)(b/T
CR."
1. QR (TR)
vº

Substituting for Nk, 1 and Nk, -1 in equation (5) and rearranging the equation yields

- Nvº
H + J (6)
*(vºk)-Hº)v; (vºv)=#|
QR(TR) |

per second per unit volume. In equation (6)

H = (K' + ... [ºr] …-ſk-º/Tº + (7)

q = 2.88

10
for the ground state of nitrogen and

J = C(Rø, 1PP + RK -1PR) (8)

In reference 5, 9 was introduced as a normalizing factor and is given by

•-20%, ºr ‘Nk-PR)
r
(9)

The number densities in 6 will not have a Maxwellian distribution among the ground
states but this quantity is still a valid normalizing factor for a particular set of
conditions.

At low temperatures there is no appreciable excitation of the vibrational levels of

Nox"2 other than the v'. = 0 level so that N ... nV = N. Therefore, for the
2 1 (vº-0)
0–0 band, equation (6) becomes

* QR(TR),
Vi

which for a given temperature and beam configuration, can be written as

NH
Ö(V" = 0,K') = Constant i + J

QR(TR),
1

where the constant is FN ſº 0 per second per unit beam current and is independent of

N since 6 varies as N. By equating the rate of photon emission to the rate of excita
tion, Muntz (ref. 5) obtains the emission-time intensity from

(**), 0) = constant P(K.K.) +(v' = 0,K) v4

Substituting P(K',K.) = PR = K/(2k + 1) into this equation and treating v4 as a con
stant gives for the 0–0 band

-K" (K'4-1) (b/T t

I(K.K.). -c —º-roº +1)*/ *:::: (10)

p QR(TR) 77 + 1

11
where G is given by equation (2) and CE is a constant relating the experimental
intensities to the theoretical intensities for a particular experiment. Define a function

I(K', K.) ...-koe -K" (K'4-1)

(K'+ b/T/ R (11)
Call C

Then equation (10) can be written as

I(K.K.)
(K", º = ceil—º-I(K.K.
Foſſ.) ( 'al * 2K7.K' I J
nº
12
(12)

The reader is referred to Muntz's paper (ref. 5) for more discussion of the steps
leading to equation (12) since the present derivation is similar. The effect of secondary
electrons on a particular K' level is introduced through J and depends on the rate of
secondary excitation to the K" = K' + 1 levels. The total intensity of the R or
P branch of the band is not changed by the secondary excitation since one level is
always populated at the expense of another level. Therefore, the sum over all K" of
the terms involving J must be zero and from equation (12)

***).]-ººº...]
K" "R), K"

Let

I l -) I(K',K,
Boole,"; "“”),
and

Foo|
I
calc
=) I(K",K.
(
K" ºals
Then

Fool exp = CEi St. Foo calc

12
Normalizing equation (12) by ſo o yields
lexp
K"J T 17

º
*(0,0)
º
*(0,0) calc
... *"ºv
1)NRoo)
(2K" + calc
(13)
exp

where only the upper state quantum number K' denotes the rotational line.
To evaluate J, values of Rºn are required for each ground-state rotational level.
The selection rule for transitions produced by the low-energy electrons is AK" = +2
and Rk" is the net value for the transitions between levels K" and K" + 2. The
cross sections are denoted by g(v,i.2)K. where the subscript denotes the level from
which the transition arises and (+2) is AK".
When the excitation of the K" level from the K" - 2 level is considered, the rate
of excitation by secondary electrons is

rk" - nsºk"-2 \, g(v,42)k, .2f(v) V dv

per unit volume per second. To avoid the integral over the unknown f(v), the cross sec
tion g(v,42)K"-2 is written in terms of the value for a reference velocity vo and all
the cross sections are assumed to vary in a similar manner with velocity. This variation
does not seem to be unreasonable from the values given in reference 16. Then

g(v,42)K'_2
rk" - nsNº. –25(Vo,42) K"-2 ſ —º-t–
t
f(v) v dv
g(Vot?)K"-2
V

Within the accuracy of this assumption, the integral is the same for all K", can be
treated as a constant, and allows the calculation of relative values for rk". Thus,

rR" = Constant nsNk"-23 (vo,42)k, .2 (14)

per unit volume per second.

The rate of excitation required is that due to transitions within the observed gas
volume. For a field of view of the optics larger than the beam diameter, the beam size
determines this volume. For large or scattered beams, the field of view of the optics
usually determines this volume. If p is the beam diameter or field-of-view width, the
dependence of rk" on number density is different for the two limits As = p and
As - p. For As & p, the excitation rate is given by equation (14) but for As > p; the
13
probability of a secondary electron colliding within the observed volume increases as N
and rR" * NnsNº'-25(Vot?)K"-2. The excitation rate Ric, is the net result of ric
for four transitions and for As 2 p and is

RK" = Constant Nns (ºkº (Vo-?)K".2 + Nk"-25(Vot?)K"-2

-*[Coºk'. Co-ºr)
Substituting values of RK" into equation (13) and writing the result in terms of K'
gives for As 2 p

º º + R}v;
An K'Q, , (T
nsk'QR( (K 4: 1)[x(K. 4, 1]
I
(0,0) I (0,0) [ool...(***) º t
2K" + 3
exp calc

, KTX(k -)
2K" - 1
(15)

and, similarly, for As ‘p

BnsK'QR(TR)
ſº
I
- º
I
+
N|I
S R\* RJrrit

|
V. ſk' . . 1)[x(K. 11)
2K' + 1 2K" + 3
(0,0,…, |0.0), ale NHoolºl.” “” +

, K'i X(K" –
“Lººk - 1)
1)(K (16)
2K" –

where A and B are constants and

X(K) = Nk -29 (vo,+2) K'_2 * Nk,29 (Vol -%)k, .2

- Nkſ:(voºk *Coºk + (17)

A first approximation to X(K') can be obtained by substituting Maxwellian values

for Nk"; this substitution gives

14
 x(K)=–8– F(K) (18)

where

F(K) = g(Woº), ºr - 3)e

-(K'-2) (K'-1) */ Tº
+ g(vo-?), ºr + 5)e -(K'4-2) (K'4-3
(K'+2)(K'+ Yº/ T º
+

º
- (2K' + 1)e
-K'(K+1)*/TR Fºº, **). + (19)

Substituting for X(K') in equations (15) and (16) gives for As 2 p

ſº s |I(K)
*(0,0) exp *(0,0) calc

+
An
shK"
Fools.” D2K 3
r
t
-
ſk ++ F(K 1) + šil:* º 7

2K" - 1
* (20)

and for As “p

ſº º
'09)|ex, '09|ale2:

+ Bng K'
Foo) ...(ºk
ſk i nE(k: ; ), kºrſk -
DLT2K 3 2K" - 1
i (21)

For the proposed excitation model, equations (20) and (21) predict that the differ
ence between the measured line intensities and values calculated from Muntz's equation
(ref. 5) varies as Nns at low densities (^s > p) and tends to a dependence on ns at
high densities (A sº p). The transition from one regime to the other occurs for As = p

15
and therefore at a density which depends on the geometry of the electron beam and
optical system. At the present time ns cannot be reliably evaluated as a function of
the gas density. Estimates can be made of the increase in ns due to the electrostatic
field set up around the beam (ref. 13), but the quantitative effects of beam spreading and
the recombination of electrons with ions are unknown. In addition, the constants A
and B involve the integral over the secondary electron distribution function and cannot
be reliably evaluated.
The value of equations (20) and (21) is the predicted dependence of the normalized
line intensities, and hence the calculated rotational temperature, on the gas density,
secondary electron number density, and beam geometry. Experimental measurements
are needed to confirm these predictions and to provide a calibration of the technique for
a range of temperatures and densities.

MEASURED ROTATIONAL TEMPERATURES

The experimental measurements were made in a flowing stream of nitrogen at a

known temperature using a 30 keV electron beam and beam currents between 0.3
and 1.5 mA. The impurity level in the nitrogen was 50 parts per million. Figure 2
shows the heat exchanger used to cool the gas and the position of the electron beam.

| GAS IN
DR
LIQUID N2 |N 2CONTROL VALVE #.
CONCENTRIC º
BAFFLES P
º

º
*
*
N *
N *
N ,”
N *
N *
*
*

T *
N º
º
N*
*
*
VACUUM -

CARBON FARADAY CUP

COOLANT BATH TO VACUUM PUMPS
(2-METHYL BUTANE)

Figure 2.- Schematic diagram of test chamber.

16
The heat exchanger is enclosed in a bath of 2-methyl butane which is thermally insulated
by the surrounding vacuum. A controlled flow of liquid nitrogen through a coil immersed
in the 2-methyl butane permits the cooling of the heat exchanger to any desired tempera
ture between 300 K and 113 K. Alternatively, liquid nitrogen is used in place of a
2-methyl butane to obtain temperatures close to 78 K. The heat exchanger consists of
two chambers separated by a valve which controls the pressure in the first chamber.
This chamber acts as a reservoir of cooled gas for the main chamber to insure that the
gas is cooled to the temperature of the heat exchanger, even at the lowest pressures.
The gas temperature is assumed to be the mean value recorded by thermocouples on the
last baffle and along the exit pipe from the heat exchanger. The diameter of the exit pipe
was 32 mm and the pumps maintained a maximum velocity through it of approximately
50 m/sec which is sufficient to prevent convective heating of the gas by the hot drift tube
and mixing with the warmer gas from the surrounding chamber. (See ref. 8.)
A calibrated pressure transducer recorded the test chamber pressure through an
orifice in the chamber wall. The wall temperature remains close to ambient and thermal
transpiration along the tubing is negligible. The density of the flowing gas is calculated
from the measured pressure and temperature.
The electron beam enters the test region through a narrow orifice in the drift tube.
For pressures below 130 N/m2, the orifice diameter is 0.1 cm but for higher pressures
is 0.05 cm. Both drift tube and electron gun are insulated from ground potential and a
grounded shield leaves only a small circular area around the drift tube orifice exposed.
To measure beam current, the whole test chamber served as the collector and the carbon
Faraday cup reduced the backscatter of primary electrons from the chamber walls. This
procedure avoids errors due to beam spreading that arise when only the Faraday cup is
used as the beam collector.

Measurements showed that at high beam currents the nitrogen first negative band
intensities did not vary linearly with beam current, particularly for the higher densities
and smaller drift tube orifice. The nonlinearity is thought to be due to an interaction
between the drift tube and the plasma produced by the beam in and around the end of the
drift tube orifice. The drift tube is slightly positive with respect to ground potential and
draws a net electron current from the plasma. Unbalanced positive ions left in the plasma
diffuse to the test chamber walls and thus reduce the indicated electron-beam current
below the true value. This problem is discussed in reference 8 and leads to errors for
high beam currents at high densities, but these errors do not seriously affect the conclu
sions of this study.
An optical system formed a 1:1 image of the beam on the entrance slit of a spec
trometer which resolved all the lines in the R branch of the 0–0 band. The slit length,
oriented perpendicular to the length of the beam image, determined the width of the field

17
of view of the optics. This width was 0.1 cm on either side of the beam center line,
unless otherwise stated. The slit width determined the resolution along the length of the
beam so that the volume of gas observed depends on the slit area and the depth of focus
of the optical system. At low densities, when beam spreading is small, the fluorescence
is observed from a cylindrical volume defined by the intersection of the field of view with
the beam. For higher densities some of the electrons are scattered beyond the field of
view (ref. 8). Fluorescence is then observed from a section through the beam of almost
constant width determined by the slit length. Measurements were made at selected tem
peratures for a range of densities by scanning the rotational lines while measuring the
total fluorescence intensity to monitor for changes in pressure or beam current.

Figure 3 shows the influence of beam current and density on the measured rota
tional temperature at 125 K. The symbols denote the values for a fixed density with
variable beam current, and Ni correlates the ratio (TR) ſº
for the limited
calc

range of beam currents. The solid line is a cubic polynomial fitted through the points by
the least-squares method. It is noted that (TR) ſº
does not approach 1.0 as
calc
Ni tends to zero and it increases to an almost constant value for Ni > 6 x 10% micro

amperes X molecules per meterº.

3
Calc H DENSITY/METER
'w A o 1.4x102 to 9.5x10%
|ºo D 2.86x10°
A 6.2xl.”

1.0 1 l l | I | L I l

0 2 4 6 3 oxid”
N, MICROAMPEREs, MOECUIESINEER’
Figure 5. – Calculated rotational temperature against Ni. *max = 15, Tw = 125 K.

18
 The present results for other temperatures are similar to those presented in fig
ure 3. The ratio (TR) |Tw
increases with density and reaches an almost constant
calc

level at high densities. This level decreases with increasing temperature.

DISCUSSION

Equations (20) and (21) predict that Aitk/ſo o **/J Calc

varies as nsN at low
densities (*s p)
> and tends to a dependence on ns at high densities (^s p).
< Figure 4

Shows ſkyo o for two rotational levels at 111 K plotted against Ni.
y exp
At high

densities ſk)/0,0] exp' Aitk/ſolo) calc' varies linearly with

and hence Ni and

at low densities it varies approximately N times faster, as predicted by equations (20)

and (21). The transition between the two regimes occurs for a number density of
approximately 1.5 × 1022 meterº, corresponding to a total collision cross section
per
of 3 × 10-19 meter” which is the same order of magnitude as that given by Takayanagi
et al. (ref. 16) for electrons of 1 eV energy. Furthermore, I(3) originating from the
abundantly populated rotational levels is reduced and I(13) originating from the

0. I? I- ...)

Faired curve

—a O C
I

0 5 10x10°
A A —A-4) —ZY
0.02

( ...)
| Faired Curve
(0,0) exp (º
.0l t- (0.0%alc | |

0 5 3 10x10”
Ni, MICROAMPERES X MOLECULES/METER
Figure H. - Normalized line intensities. T = lll K.

19
sparsely populated levels is increased. Similar trends observed for all the present data
strongly support the assumption of low-energy secondary electron excitation of the
ground-state rotational levels.
The ratio ſkyo o ”* I exp
does not tend to ſkyo o is as
* 2 J CalC
N tends to

zero for the present measurements. This condition could be due to some unaccounted
for secondary effect or to the variation of ns with density. Dunn and Self (ref. 17)
state that for N < 3 × 1018 molecules per meterº, the beam-generated plasma is
typically of about the same density as the beam. At higher densities, the electrostatic
forces become important and references (13) and (15) show that ns is increased by
several orders of magnitude for N = 3 × 1019 molecules per meter*. Therefore, ns
must increase rapidly with density for 3 × 1018 × N < 3 × 1019 molecules per meter 3.
All the present results are for higher densities and would therefore appear to tend to the
wrong limit when plotted against Ni.
The potential distribution in the beam-generated plasma governs the distribution
of nS (refs. 13 and 17) but the magnitude of ns depends on the rate of production of
secondary electrons and ions by the primary beam. Although the total beam current does
not change greatly, the distribution of current density j changes with beam length and
density because of the spreading of the beam through collisions. Therefore, unless the
distribution of ns is independent of the distribution of j, the present measurements at
a distance of 2.8 cm along the beam should reflect the variation of j with density. The
ratio I(K)/Hoo) varies approximately as N2i, tends to Ni at high densities, and

thus implies that over the observed area of the beam, ns does not depend on the distri
bution of j but only on the total rate of production of secondary electrons which is pro
portional to Ni. The solution of Bienkowski and Harbour (ref. 13) implies that this
concept should be approximately true since the secondary electron number density at the
beam center tends to become independent of beam diameter for large ratios of chamber
to beam diameter.

This point is of major importance in any attempt to correct rotational temperature

measurements by using curves of (TR)
cal
ſºw
against Ni. If ns is not independent

of the distribution of j, then the ratio (TR) calcſº would vary along the length of the

beam. The reasonable agreement of the present results (figs. 5 and 6) with those of
other investigators (refs. 6 and 18) and the evidence of figure 4 suggest that ns is
approximately independent of the distribution of j at least for moderate beam lengths.
Figure 4 shows that there are three density regimes to consider. For
N < 3 × 1019 molecules per meterº, a rapid increase in ns with density is thought to

20
 Cubic polynomial "—
1.4|- O
O
A
(R), ''Fa SLIT LENGTH, N, 3
Calc % ºe METERS MOLECULES/METER
'w e Ix IO-3 t
| 2H,
[] []
eP —
A
2x 10−3
5x IO-3
| variº ResultS

. . my In”
I. I |
A
- * - - - - --

g-ºff| 5 §§
2.j, loº
º Results

|. 0 | 1 |
0 | 2 3x10”
3
Ni, MICROAMPERES X MOLECULES/METER
Figure 5. - Comparison with Williams' result at 78 K. Kmax = 15.

l. 3 –
(T ºak O PRESENT RESULTS

'w |.. 2
O ASHKENAS’ RESULTS

I. I H

|.0 l | | | | _l

0 2 4 6 8 10 2x 10°
N. MOIECULESIMETER’
Figure 6.- Comparison with Ashkenas' results at 78 K using lj rotational lines.

21
be responsible for the large values of A1(k)/io o y
calc
, and hence high (TR) lºſºw
Ca

even at the lowest densities reached in the present experiment. For higher densities,
ns increases approximately linearly with Ni and the mean free path of the electrons
varies as 1/N. Therefore, when As 2 p, the number of secondary electrons colliding
within the observed volume increases as N, and results in A1(k)/ſo o ’’’ J Calc
increasing

approximately as N*i. For As & p, all the electrons are observed and A1(k)/io O) >

increases as Ni. This result explains the variation of (TR)

calc
ſº which should be

a function of Ni when As & p and a function of N2i when As > p. The correlation
of the data in figure 3 with Ni for the complete density range is not understood, but the
range of beam currents is not large.

When As & p, the geometry of the electron beam and optical system are less
important since practically all the secondary electrons produced within the observed vol
ume collide within that volume and (TR) 1
Ca10
|Tw increases only very slowly with Ni.

However, when As = p, the geometry determines how many secondary electrons collide
within the observed volume and hence the rate at which (TR)
calc
ſº approaches the

almost constant value for As & p; therefore, for low densities, increasing the slit length
should increase (TR) ſº
calc
Figure 5 compares (TR)
calc
ſw measured at 78 K

for different spectrometer slit lengths. The limited amount of data and the experimental
scatter precludes any definite conclusion, but (TR) |Tw
appears to increase slightly
caic
with slit length. Petrie et al. (ref. 19) observed a similar effect of slit length in a
Mach 11 flow with static temperatures for 20 K and 35 K. Williams (ref. 18) also made
measurements at 78 K but in his case the slit dimension, in the direction perpendicular
to the beam length, appears to be 2 × 10−5 meter. The experimental scatter is too large
to draw any conclusion about the effect of the small slit dimension but the results are
scattered about the present data and increase with Ni.

Ashkenas (ref. 6) measured rotational temperatures at 78 K and room temperature

but did not state what beam currents were used. The maximum beam current for his
data (ref. 6) appears to be 3.5 × 10-4 ampere, whereas for the present results it was
a maximum of 1.5 × 10−3 ampere and decreases to about one-fifth of this value at high
densities. When As ‘p, (*R), alſº w is insensitive to Ni and therefore these dif

ferences in beam current should not be important. This result is confirmed in figure 6
which compares the results for 78 K plotted against density only. At low densities
(^s p), the experimental
> scatter prevents different trends in the two sets of data from
being detected.
22
 In principle, equation (15) or (16) allows the relative values of AI(K")/ ſo o **/J Calc

to be evaluated, but the secondary-electron excitation redistributes the molecules among

the ground-state rotational levels and Nº. is not known. A first approximation can be
obtained by assuming a Maxwellian distribution which leads to equations (20) and (21).
With equation (19), either equation (20) or (21) gives values of AI(K)/AI(3) as a func
tion of K' only, and these values are compared with experimental values in figures 7
and 8. Quantitative agreement is not expected because of non-Maxwellian distribution,
but the qualitative agreement is very good and sparsely populated levels are populated at
the expense of the more abundantly populated levels. Qualitative agreement at higher
temperatures is equally good and, as the peak in the rotational distribution moves toward
higher K", the depopulation of successively higher rotational states is predicted and
agrees with the data. At temperatures of several hundred K, the low rotational states
have relatively low populations and secondary electron excitation would increase their
population at the expense of the higher states. This increase leads to low values of
(*R)ale and could explain Hunter's (ref. 20) results at high temperatures, but the
report does not state which rotational lines were used.

l. Oſ

PRESENT
THEORY

H
0 e; I l

Al(K") O A
-AI(3) 3
Q N, MOLECULES/METER
ZN o 0 19 x 10%
º A 1.13 x 102?
à D 6.00 x 1022
-l. 0}– + 1000 x 1022

| | |
0 5 10 15
K"

Figure 7. - Relative intensity error at 79 K.

23
 1.0 T

PRESENT
THEORY

0 T

º N. MOIECUIESINEER’
O 0.67 x 1022
A ºf x 1%
-1.0 H +
à D 60 x 10%
+ 9.5 x 1022

l | I
0 5 10 15
K"

Figure 8. – Relative intensity error at lºj K.

Aerodynamic Measurements
An important use of the electron-beam fluorescence probe is the study of high
Mach number low-density aerodynamics and consideration must be given to the effect of
high flow velocities. The present results strongly support the hypothesis of excitation
of ground-state rotational levels by low-energy secondary electrons. The qualitative
effect of high flow velocities on this model can be deduced by reference to the rarefaction
parameter 6 defined by Bienkowski and Harbour (ref. 13) as the ratio of the beam
radius to a modified mean free path. High velocities can only affect the "R) al.|Tw
through the distribution of ns since the rate of production by primary electrons is inde
pendent of velocity. Collisions between fast-moving neutral molecules and ions in the
beam will sweep the ions downstream and the electrostatic attraction between ions and
electrons causes a redistribution of the secondary electrons. Bienkowski and Harbour
(ref. 13) concluded that for 6 s 1, there should be little effect of velocity, but for
ô -> 1, the distribution of ns, and hence (TR) Tw, is affected. A value of 5 of 1
calc
corresponds to approximately 10 N/m2 at room temperature for many electron-beam
situations and most of the present results are in the regime where 5 = 1.
Comparison of the present results with wind-tunnel measurements for known flow
conditions and similar beam parameters would determine the effect of flow velocity.

24
Measurements are available in a Mach 6 flow (ref. 21) for a static temperature of 71.5 K
and 5 of order 1. Extrapolating the present results to this temperature gives
(TR) calJºw = 1.13 + 0.03 compared with 1.09 + 0.03 for the wind-tunnel data; thus,

it is confirmed that there is little effect of the velocity for 5 = 1. For all other known
wind-tunnel measurements, there are either geometric differences or the beam currents
are not given. In addition, there is some doubt about Ni as a correlating parameter.
If a beam current of 4.5 × 10-8 ampere is assumed, the Mach 9 data of Williams,
Hornkohl, and Lewis (ref. 22) agree with the present data to within 6 percent. Their
Mach 10 data are considerably higher, possibly because of uncertainties in the flow con
ditions. The Mach 3.92 data of Robben and Talbot (ref. 9) for a beam current of
10-2 ampere are very much lower than the present results for the same Ni but show
reasonable agreement when compared only in terms of the density. Maguire's free-jet
measurements at 154 K (ref. 7) cover a range of densities for 5 × 1, but the beam cur
rents were not given and the slit was parallel to the beam image. The effect of the dif
ferent geometric arrangement is less important at high densities, and for
N = 1.23 × 1023 molecules per meterº, Maguire's result is 4 percent lower than that of
the present study. This difference shows that there is no strong effect of flow velocity
On ("R), al. Such a result is to be expected since "R) alsſ", is insensitive to n
S

at high densities (^s << p) and therefore to the influence of flow velocity on the distribu
tion of ns. For lower densities, the discrepancies increase to 10 percent for
N = 1.32 × 102” molecules per meter”, but no conclusion can be drawn because of the
different geometries.

Correction Procedure for (TR)

calc

Many variables influence (TR) and simple correction procedures will apply
calc

only for specific conditions. The present measurements used beam currents of approxi
mately 10 –3 ampere and a field of view of the optical system of 0.1 cm on either side of
the beam center line. The drift tube orifice was 0.1 cm or 0.05-cm diameter, but the
beam spreads beyond this at the measurement station which is 2.8 cm from the drift tube.
A correction procedure using curves of (TR) ſº
as a function of Ni will apply
calc
only for beam currents for which the correlating parameter Ni is valid. The results
might be different for beam lengths considerably greater than 2.8 cm, and at low den
sities (*s p), the field of view is important.
>

For each set of data at a fixed temperature Tw, a least-squares cubic polynomial
was fitted through the values of (TR)
calc
ſºw
as a function of Ni. The polynomial

25
 TABLE I. - POLYNOMIAL COEFFICIENTS FOR (TR) ſt
calc

Coefficients for T of –

79 K 111 K 125 K 167 K 210 K 237 K 258 K 300 K

Kmax = 15

CO 1.226 | 1.103 1. 110 1.047 1.028 | 1.054 1.002 1.014

cı X 1027| 15.11 9.86 2.24 2.66 4.20 2.42 8.85 11.69

c2 x 1058 -22.69 |-16.0 1.30 0.174 || -4.71 || -1.83 –25.81 –37.52

c3 x 1079| 11.65 8.85 –2.37 -2.07 1.95 0.246 25.12 36.2

Kmax = 14
Co 1.193 | 1.091 1. 106 1.037 1.025 | 1.054 1.007 1.015

c1 x 102"| 18.36 || 8.84 0.597 || 3.40 4.46 2.26 7.7 11.15

c2 x 105* -22.71 -15.3 4.82 –2.15 —5.36 || -0.850 | -20.61 –37.52

ca x 1079| 12.74 9.23 –4.64 –0.224 2.30 | –0.9.20 18.58 33.72

Kmax = 13
Co 1.162 | 1.105 1.093 1.035 1.025 | 1.063 1.023 1.016

cı x 1027| 12.81 6.14 1.45 3.51 4.64 1.77 5.23 10.68

C2 x 108°, -24.06 || -9.46 2.35 -1.44 –5.73 || -0.373 || -11.20 –34.55

ca x 1079| 14.45 5.41 –2.89 || -1.19 2.51 -0.623 8.78 34.18

Kmax = 12
CO 1. 142 | 1.126 1.103 1.020 0.992 | 1.054 1.033 1.022

c1 x 1027 | 11.81 4.00 0.345 3.78 7.82 1.66 5.13 10.22

c., x 10%| -23.07 | -6.03 0.438 || -2.46 || -13.23 0.898 || -14.05 –34.28

cs x 10° 14.06 3.88 –4.09 || –0.217 7.58 -2.27 13.55 34.77

Kmax = 11
CO 1.121 | 1.126 1.075 0.995 0.978 || 1.042 0.996 1.007

cı X 1027 | 11.77 2.83 2.54 5.11 7.67 2.22 8.81 11.04

c2 x 1058 -24.34 -2.07 -0.581 || -4.82 | – 12.7 2.65 –24.97 –34.06

cs x 1079 15.37 0.401 || -0.954 1.26 7.55 -5.69 23.40 32.22

Range of Ni, microamperes X molecules

meter*

(Ni).max 1 × 1026 |7.7 x 1025 |9 × 1025||7.2 × 1025|8 x 1025 |7.2 × 1025|6 x 1025 ||6 x 1025
(Ni)min 7.8 x 1023| 1.5 x 1024 |2 x 1024 |2 x 1024 |2 x 1024 |1 x 1024 |1.3 × 1024|6.4 x 102?
26
>efficients are given in table I, where Kmax is the number of spectral lines used
<cept for 300 K, when the first two lines were neglected because of overlapping by
le P branch. Within the range of Ni given at the bottom of this table for each tem
2rature, values of (TR) ſº
can be calculated from
calc

~2 A3
(T"allTu, = Co0 + Ci1 Ni + Co.(Ni)"
-

W 2
+ C3(Ni)
3

alues of (TR)
calc
|Tw calculated from these polynomials are plotted in figure 9 for

elected values of Ni.

Ni MICROAMPERES X MOLECULES
1.5r- 3
METER

O 0.5 x 1025
l, 4– [] l.0 x 1025
KX 4.0 x 1025
A 6.0 x 10”

—Fak Faired Curves

W l. 2H

l. 1 H

1.0 | | | | | |
0 50 100 150 200 250 300

T, K
Figure 9. - Mean values of (TE) calc|. computed from polynomials. Kmax = 15.

The calculated rotational temperature depends on the density, but to measure den
sity when quenching is appreciable, the temperature must be known and an iterative pro
:edure is required to correct the measurements. (See ref. 8.) First estimates are made
if T 11 and N11 and equation (1) used to calculate (T
( *%ale and (T
( *%alef W
is

:omputed from the polynomials by substituting Nii. There is some advantage in norma
lizing (T
( Real ſt W
by the value at 300 K, that is, ( TWaleſt W 300 K , since the

normalized ratios are insensitive to small changes in Ni. A polynomial is fitted through

27
the normalized temperature ratios as a function of temperature and stored on the com -
puter for use in successive iterative steps. If R(T) denotes the normalized temperatur
ratio at temperature Tw, then (**) ſº l
- R(T) (**) ſº l
is

obtained by substituting T1 and N1i into the relevant polynomials. A better estimate
of the temperature is TR = (TR)
calc
(TR)
calc
ſ", T
. with N1 fixed, the loop is

repeated by substituting this improved estimate into the polynomials until the calculated
temperature converges to a constant value T2. A second estimate of the density N2
can then be obtained by using T2 and the quenching cross sections given in reference 8.
Repeating this process gives successive estimates of temperature and density. If the
estimate N1 was reasonable, the original polynomial for R(T) can be used for each
iterative step without making serious errors. This procedure involves more computation
than that given previously (ref. 8) but allows a greater range of Ni to be covered at low
temperatures.

Other correction procedures were investigated and one consequence of the proposed
secondary-electron excitation model is that X
K"
*/[. ol l
* * Zl CalC
= 0. This summation

varies slowly with the error in the temperature. Combined with large errors in AI(K'),
this condition makes it impractical to correct the temperatures by minimizing the sum
mation. A simple and apparently successful procedure developed earlier (ref. 10) was
based on the incorrect assumption of direct excitation of the N;B% states by secondary
electrons. Subsequent study shows that the equations in reference 10 are approximate
functions fitted through Ashkenas' data. (See ref. 6.) The successful application of this
procedure in reference 21 is the result of similar beam currents and geometry for the
two experiments. Correcting this shock-layer data (ref. 21) by the present method did
not change the temperatures sufficiently to affect the conclusions of that paper.

CONCLUDING REMARKS

Rotational temperatures measured by the electron-beam fluorescence technique

increase with density and beam current and depend on which spectral lines are used.
The secondary mechanism responsible for the high rotational temperatures cannot be
conclusively identified, but excitation of the ground-state rotational levels by low-energy
secondary electrons can account for all the observed results. This excitation mechanism
redistributes the molecules among the ground-state rotational levels and thus increases
the population of the sparsely populated levels at the expense of the more abundantly
populated levels.

28
 For this excitation model, the spatial distribution and number density of the sec
ondary electrons and the field of view of the optical system determine the rotational
temperature error for a given density. High flow velocities can influence the measure
ments through collisions with ions, and thus result in a different spatial distribution of
electrons and ions around the beam. For low densities, when the secondary-electron
mean free path is large compared with the beam diameter or field-of-view width, there
is little influence of high flow velocities. At high densities, when the secondary-electron
mean free path is relatively small, the calculated temperature is insensitive to the
secondary-electron number density, and hence to all but strong effects of the flow on the
distribution of the secondary electrons in the observed gas volume. For these two
regimes the present correction scheme should be reliable, but at intermediate densities
the flow will reduce the secondary electron concentration in the observed gas volume
and the temperature will be underestimated.

Further studies are needed to establish reliable correlating parameters for the
calculated temperatures and to determine more accurately the effect of high flow veloc
ities, particularly at intermediate densities.

Langley Research Center,

National Aeronautics and Space Administration,
Hampton, Va., October 26, 1971.

29
 REFERENCES

. Lillicrap, D.C.; and Berry, C. J.: Experimental Model for High-Speed Rarefied
Flow Over a Sharp Flat Plate. Phys. Fluids, vol. 13, no. 5, May 1970,
pp. 1146–1152.
. Fischer, Michael C.; Maddalon, Dal V.; Weinstein, Leonard M.; and Wagner,
Richard D., Jr.: Boundary Layer Surveys on a Nozzle Wall at M = 20 Including
Hot-Wire Fluctuation Measurements. AIAA Paper No. 70-746, 1970.
. Beckwith, Ivan E.; Harvey, William D.; and Clark, Frank L. (With appendix A by
Ivan E. Beckwith, William D. Harvey, and Christine M. Darden and appendix B
by William D. Harvey, Lemuel E. Forrest, and Frank L. Clark): Comparisons of
Turbulent-Boundary-Layer Measurements at Mach Number 19.5 With Theory and
an Assessment of Probe Errors. NASA TN D-6192, 1971.
. Wallace, J. E.: Hypersonic Turbulent Boundary-Layer Measurements Using an
Electron Beam. Compressible Turbulent Boundary Layers, NASA SP-216,
1968, pp. 255-308.
. Muntz, E. P.: The Electron Beam Fluorescence Technique. AGARDograph 132,
Dec. 1968.

. Ashkenas, Harry: Rotational Temperature Measurements in Electron-Beam Excited

Nitrogen. Phys. Fluids, vol. 10, no. 12, Dec. 1967, pp. 2509-2520.
. Maguire, Bernadette L.: Density Effects on Rotational Temperature Measurements
in Nitrogen Using the Electron Beam Excitation Technique. Rarefied Gas
Dynamics, Vol. II, Leon Trilling and Harold Y. Wachman, eds., Academic Press,
Inc., 1969, pp. 1761–1782.
. Lillicrap, D. C.: Experimental Determination of Density and Rotational Temperature
by an Improved Electron Beam Technique. AIAA Paper No. 71-605, June 1971.
. Robben, F.; and Talbot, L.: Measurements of Rotational Temperatures in a Low
Density Wind Tunnel. Phys. Fluids, vol. 9, no. 4, Apr. 1966, pp. 644–652.
10. Lillicrap, D. C.; and Harvey, J. K.: Electron-Beam Rotational Temperature Mea
surements Including the Effect of Secondary Electrons. AIAA. J., vol. 7, no. 5,
May 1969, pp. 980-982.
11. Smith, Ronald B.: N2 First Negative Band Broadening Due to Electron Beam
Excitation. Rarefied Gas Dynamics, Vol. II, Leon Trilling and Harold Y. Wachman,
eds., Academic Press, Inc., 1969, pp. 1749–1760.

30
12. Hickman, Roy Scott: Rotational Temperature Measurements in Nitrogen Using an
Electron Beam. AFOSR 66-2509, U.S. Air Force, Sept. 1966. (Available from
DDC as AD 645 113.)

13. Bienkowski, G. K.; and Harbour, P. J.: Structure of Electron-Beam Generated

Plasma. Rarefied Gas Dynamics, Vol. II, Leon Trilling and Harold Y. Wachman,
eds., Academic Press, Inc., 1969, pp. 1651–1658.
14. Self, S. A.; and Ewald, H. N.: Static Theory of a Discharge Column at Intermediate
Pressures. Phys. Fluids, vol. 9, no. 12, Dec. 1966, pp. 2486-2492.
15. Harbour, P. J.; Bienkowski, G. K.; and Smith, R. B.: Influence of Secondary Electrons
on an Electron-Beam Probe. Phys. Fluids, vol. 11, no. 4, Apr. 1968, pp. 800-803.
16. Takayanagi, Kazuo; and Takahashi, Tan: Behavior of Slow Electrons in Atmospheric
Gases. Part III – Collision Cross Section of Molecular Nitrogen. Report of
Ionosphere and Space Research in Japan, vol. 20, no. 4, Dec. 1966, pp. 357-373.
17. Dunn, D. A.; and Self, S.A.: Static Theory of Density and Potential Distribution in a
Beam-Generated Plasma. J. Appl. Phys., vol. 35, no. 1, Jan. 1964, pp. 113-122.
18. Williams, William D.: Laboratory Verification Studies of Rotational and Vibrational
Temperature Measurements by the Electron Beam Technique. AEDC-TR-68–265,
U.S. Air Force, Feb. 1969. (Available from DDC as AD 683 001.)
19. Petrie, S. L.; and Boiarski, A. A.: The Electron Beam Diagnostic Technique for
Rarefied Flows at Low Static Temperatures. Rarefied Gas Dynamics, Vol. III,
Leon Trilling and Harold Y. Wachman, eds., Academic Press, Inc., 1969,
pp. 1685–1701.
20. Hunter, William W., Jr.: Investigation of Temperature Measurements in 300°
to 1100°K Low-Density Air Using an Electron Beam Probe. NASA TN 4500,
1968.

21. Lillicrap, D.C.; and Berry, C. J.: Rotational Temperature Measurements in Low
Density Flows. AIAA. J., vol. 8, no. 11, Nov. 1970, pp. 2078-2080.
22. Williams, W. D.; Hornkohl, J. O.; and Lewis, J. W. L.: Electron Beam Probe for a
Low Density Hypersonic Wind Tunnel. AEDC TR-71–61, U.S. Air Force, July
1971.

NASA-Langley, 1971 — 12 L-8011 31

NATIONAL AERO NAUT ICS AND SPACE A D M ISTRATION
WASHINGTON, D.C. 20546
POSTAGE AND FEEs PAid
NATIONAL AERONAUTics and
OFFICIAL BUSINESS sPACE ADMINISTRATION
FIRST CLASS MAIL
PENALTY FOR PRIvarE use $300

02 1 002 C1 U 2 5 71 1217 SnO4 a 9H

UN IV OF I LL IND IS
L I BRARY
ATT N: DnCUMENT LIBRARIAN
UR BANA IL 61 an 1

POSTMASTER:. . If Undeliverable
Postal Manual) Do(Sectiºn º
Nºt ºft

"The aeronautical and ſpace activities of the United States ſhall be

conducted ſo aſ to contribute . . . to the expansion of human knowl
edge of phenomena in the atmoſphere and space. The Adminiſtration
shall provide for the wideſt practicable and appropriate diffemination
of information concerning its activitieſ and the reſultſ thereof.”
— NATIONAL AERONAUTICS AND SPACE ACT OF 1958

NASA SCIENTIFIC AND TECHNICAL PUBLICATIONS

TECHNICAL REPORTS: Scientific and TECHNICAL TRANSLATIONS: Information

technical information considered important, published in a foreign language considered
complete, and a lasting contribution to existing to merit NASA distribution in English.
knowledge.
SPECIAL PUBLICATIONS: Information
TECHNICAL NOTES: Information less broad derived from or of value to NASA activities.
in scope but nevertheless of importance as a Publications include conference proceedings,
contribution to existing knowledge. monographs, data compilations, handbooks,
sourcebooks, and special bibliographies.
TECHNICAL MEMORANDUMS:
Information receiving limited distribution TECHNOLOGY UTILIZATION
because of preliminary data, security classifica PUBLICATIONS: Information on technology
tion, or other reasons. used by NASA that may be of particular
interest in commercial and other non-aerospace
CONTRACTOR REPORTS: Scientific and
applications. Publications include Tech Briefs,
technical information generated under a NASA Technology Utilization Reports and
contract or grant and considered an important
Technology Surveys.
contribution to existing knowledge.

Details on the availability of these publications may be obtained from:

SCIENTIFIC AND TECHNICAL INFORMATION OFFICE

NATIONAL AERO NAUTICS AND SPACE ADMINISTRATION

Washington, D.C. 20546
 $29,13 ENGIN
(ſn225tn
ºil z 23: 72
ſº

NASA. T E C H IN I C A L IN OTE NASA TN D-6577

;
|\

\ERODYNAMIC CHARACTERISTICS OF
AN ALL-BODY HYPERSONIC AIRCRAFT
XONFIGURATION AT MACH NUMBERS
ROM O.65 TO 10.6
THE LIBRARY OF THE
y Walter P. Nelms, Jr., and Charles L. Thomas NON 22 1971
Y OF ILLINQS
lmes Research Center ºśćs
---

Moffett Field, Calif. 94035

ATIONAL AERONAUTICS AND SPACE ADMINISTRATION - WASHINGTON, D. C. • NOWEMBER 1971

 1. Report No. 2. Government Accession No. 3. Recipient's Catalog No.

NASA TN D-6577
4. Title and Subtitle 5. Report Date
AERODYNAMIC CHARACTERISTICS OF AN ALL-BODY HYPERSONIC November 1971
AIRCRAFT CONFIGURATION AT MACH NUMBERS FROM 0.65 TO 10.6 6. Performing Organization Code

7. Author(s) . Performing Organization Report No.

A–401.7
Walter P. Nelms, Jr., and Charles L. Thomas
. Work Unit No.
Performing Organization Name and Address
722-01-10-02-00-21
NASA Ames Research Center
. Contract or Grant No.
Moffett Field, California 94035

. Type of Report and Period Covered

. Sponsoring Agency Name and Address
Technical Note
National Aeronautics and Space Administration . Sponsoring Agency Code
Washington, D. C. 20546

Supplementary Notes

. Abstract

Aerodynamic characteristics of a model designed to represent an all-body, hypersonic cruise aircraft are presented for
Mach numbers from 0.65 to 10.6. The configuration had a delta planform with an elliptic cone forebody and an afterbody
of elliptic cross section. Detailed effects of varying angle of attack (-2° to +15°), angle of sideslip (-2° to +8°), Mach
number, and configuration buildup were considered. In addition, the effectiveness of horizontal tail, vertical tail, and
canard stabilizing and control surfaces was investigated. The results indicate that all configurations were longitudinally
stable near maximum lift-drag ratio and the configurations with the vertical tails were directionally stable at all angles of
attack. Trim penalties were small at the hypersonic speeds for a center-of-gravity location representative of the airplane,
but because of the large rearward travel of the aerodynamic center, trim penalties were severe at transonic Mach numbers.

17. Key Words (Suggested by Author(s) ) 18. Distribution Statement

Hypersonic aircraft Unclassified – Unlimited
Hypersonic flight
Lift-drag ratio
Aircraft stability
Aerodynamic characteristics
Wind tunnel models
19. Security Classif. (of this report) 20. Security Classif. (of this page) 21. No. of Pages 22. Price"

— Unclassified Unclassified 100 $3.00

For sale by the National Technical Information Service, Springfield, Virginia 22151
 NOTATION

The longitudinal force and moment coefficients are referred to the stability-axis system and
the lateral-directional coefficients are referred to the body-axis system. Unless otherwise noted, the
moment reference center is located on the body center line at 32.5 percent of the mean
aerodynamic chord, which corresponds to the 55.0 percent point of the body length.

d. C. longitudinal aerodynamic center defined at (L/D)max, percent C

*T semimajor axis of maximum cross section

Ab model balance cavity area

b span (measured between body tips)

br semiminor axis of maximum cross section

C local chord length of airfoil section

C mean aerodynamic chord (; )

C
A
axial-force coefficient, CAtotal – CAb

(p. – ph)Ab
C Ab balance cavity axial-force coefficient, —s-

A total measured axial-force coefficient

total

- - drag
CD drag coefficient, qS

CDo drag coefficient at zero lift

- - - lift
CL lift coefficient, qS

CLo lift coefficient at zero angle of attack

O CL
“La lift-curve slope at zero lift, 5. , per deg

- rolling moment
- -

Cl rolling-moment coefficient, — FH
qSb
0 Cl
°is lateral-stability parameter, 0.3 ° per deg

itching moment
Cm pitching-moment coefficient, pitching moment
qSČ
 pitching-moment coefficient at zero lift
- - normal force
normal-force coefficient, —s—
yawing moment
yawing-moment coefficient,
qSb
0Cn
directional-stability parameter , per deg
' 03
side force
side-force coefficient,
qS
ÖCY
side-force parameter , per deg
06

K boundary-layer trip (grit) size, cm

lift-drag ratio
#
maximum lift-drag ratio
( #) max

body length

forebody length (measured from nose to maximum cross section)

free-stream Mach number

model balance cavity pressure

free-stream static pressure

free-stream dynamic pressure

unit Reynolds number, per m

Ib
reference area (body planform area defined as –5)
maximum cross-sectional area

maximum thickness of airfoil section

airfoil thickness-to-chord ratio

longitudinal coordinate, measured rearward from model nose

spanwise coordinate, measured from body center line

iv
 vertical coordinate, measured from body center line

angle of attack (referred to body center line), deg

angle of sideslip (referred to body center line), deg

The following code is used to designate the various components of the model:

body

canard (trailing edge down for positive deflection)

horizontal tail (trailing edge down for positive deflection)

rudder

vertical tail

Subscripts

R right (looking upstream)

L left (looking upstream)

Number subscripts (either plus or minus) indicate deflection angles of the control surface in degrees.
 AERODYNAMIC CHARACTERISTICS OF AN ALL–BODY HYPERSONIC AIRCRAFT

CONFIGURATION AT MACH NUMBERS FROM 0.65 TO 10.6

Walter P. Nelms, Jr., and Charles L. Thomas

Ames Research Center

SUMMARY

An experimental investigation was conducted at Mach numbers from 0.65 to 10.6 of the
aerodynamic characteristics of a model designed to represent an all-body, hypersonic cruise aircraft.
The configuration had a delta planform with an elliptic cone forebody and an afterbody of elliptic
cross section. Detailed effects of varying angle of attack (-2° to +15°), angle sideslip (-2° to +8°),
Mach number, and configuration buildup were considered. In addition, the effectiveness of
horizontal tail, vertical tail, and canard stabilizing and control surfaces was investigated.

The results indicate that for the Mach number range of the test, all configurations were
longitudinally stable near maximum lift-drag ratio, (L/D)max, and the configurations with the
vertical tails were directionally stable at all angles of attack. At Mach numbers below 2, the lift and
pitching-moment curves were essentially linear up to (L/D)max for the complete configurations.
For these same configurations, the lateral-directional characteristics were nearly linear for the
angle-of-sideslip and Mach number ranges of the tests. Trim penalties were small at the hypersonic
speeds for a center-of-gravity location representative of the airplane, but because of the large
rearward travel of the aerodynamic center, trim penalties were severe at transonic Mach numbers.

INTRODUCTION

Results of a number of performance studies typified by those presented in references 1

through 4 have shown that hydrogen-fueled, hypersonic aircraft configurations with airbreathing
propulsion systems are potentially suitable for both cruise and boost missions. These studies were
based largely on estimated aerodynamic characteristics because of the lack of experimental data on
configurations having large fuselage volumes necessitated by the storage requirements of low-density
hydrogen fuel. Therefore, a program was undertaken at Ames Research Center to provide this
needed data over a broad Mach number range, as well as to assess the adequacy of various
theoretical procedures for use on these types of configurations. To date, experimental and
theoretical results have been obtained for two wing-body designs and a blended wing-body concept.
References 5 through 7 present a portion of the results from these studies. The next phase of the
program was to investigate the aerodynamic characteristics of an all-body concept. An all-body
shape represents the obvious limit in wing-fuselage blending to obtain the large volumes required for
fuel storage. This report includes results from wind-tunnel tests of an all-body hyperSonic aircraft
configuration.
 The experimental investigation was conducted in the Ames 6– by 6–Foot Supersonic and the
3.5 Foot Hypersonic Wind Tunnels at Mach numbers from 0.65 to 10.6. The Reynolds number was
held constant at 8.2X 10°/m for most of the tests; at Mach numbers 2.00 and 10.6, the Reynolds
number was limited to 4.9X10°/m. Angles of attack ranged from -2° to +15° and angles-of-sideslip
ranged from -2° to +8°.

MODEL

A drawing of the complete model with pertinent dimensions is shown in figure 1 (a), and
details of the stabilizing and control surfaces are presented in figure 1(b). Figure 2 presents
photographs of the model.

The model is representative of a hypersonic cruise vehicle derived from the analytical studies
of references 8 through 11. These studies considered an all-body design, featuring an air-breathing
propulsion system with liquid hydrogen as a fuel. This particular all-body configuration was selected
for wind-tunnel testing both because of its geometrical simplicity, which simplifies theoretical
estimates, and because of the mission studies that had been accomplished on this shape
(refs. 8–11). The model was designed to allow a complete buildup of the various configuration
components during the wind-tunnel tests. Effects of the propulsion system on the aerodynamic
characteristics were not investigated experimentally.

The body had a delta planform with leading edges swept back 75°. The forebody was an
elliptic cone, and the afterbody had elliptical cross sections (fig. 1(a)). The maximum cross-sectional
area of the body was located at the break point between the forebody and afterbody at 2/3 of the
body length from the nose (la/l = 0.6667). The ratio of the maximum cross-sectional area to the
body planform area (S/S) was 0.0935, and the major-to-minor axis ratio of the maximum cross
Section (ar/br) was 4. Since the forebody was an elliptic cone, it had a major-to-minor axis ratio of
4 at all stations. The ellipticity continuously increased with increasing body station for the elliptic
cross sections of the afterbody which terminated in a straight-line trailing edge. Removable
outboard tips were provided so that the body could be tested alone as well as with aft stabilizing
surfaces. The model span (b) was defined with these tips in place (body alone), as indicated in
figure 1(a).

Horizontal tails, twin vertical tails, and a canard surface were provided for the wind-tunnel
model (fig. 1(b)). The horizontal tails, mounted on the body center line, had 55° of leading-edge
sweep and symmetrical wedge-slab airfoil sections with the ridge lines at 50-percent chord locations.
The maximum thickness-to-chord ratio (t/c) was 4 percent, and the total exposed area of both the
right and left horizontal-tail surfaces was 12.5 percent of the body planform area. For control, the
right and left horizontal tails could be rotated either symmetrically or differentially about a point
corresponding to the longitudinal location of the centroid of the tail areas (fig. 1(b)).

The outboard-mounted vertical tails had 60° swept-back leading edges with an unsymmetrical
wedge-slab airfoil section (the inboard sides were flat). Ridge lines were located at 50-percent chord
on the outboard sides, and the thickness-to-chord ratio was 4 percent. The combined plan area of
the two vertical tails was approximately 16.9 percent of the body planform area. The major portion
of the vertical tail surfaces was above the horizontal tail and a small portion was below. Screw-on

2
wedges were added to the aft 50 percent of the vertical tails (upper portion only) to simulate
deflected rudders either in an outboard or inboard direction. The rudder hinge line was along the
50-percent chord line at a 50.4° sweepback angle (fig. 1(b)). This deflection could be either
symmetrical (rudder flare) or differential.

A canard, with 50° swept-back leading edges, could be mounted on the fuselage center line as
shown in figure 1(a). The combined exposed plan area of the right and left canard surfaces was
4 percent of the body planform area. The canard airfoil was a symmetrical wedge slab section with
ridge lines at 50-percent chord and with a maximum thickness-to-chord ratio of 6 percent. For
longitudinal control, the canard could be rotated (symmetrical deflection only) about a point
corresponding to the longitudinal location of the centroid of the canard area (fig. 1(b)).

TESTS

Data were obtained in air in two Ames wind tunnels at Mach numbers from 0.65 to 10.6. The
6– by 6–foot supersonic tunnel is a closed-circuit, continuous-flow facility with a sliding block
nozzle and a slotted wall test section; in this tunnel, the Mach number was varied from 0.65 to 2.00.
Mach numbers of 5.37, 7.38, and 10.6 were obtained in the 3.5-foot hypersonic tunnel, which uses
interchangeable nozzles; this tunnel is a blowdown facility in which incoming air is preheated by a
pebble-bed heater to prevent liquification of air in the test section. The stagnation temperature was
maintained at about 720° K for Mach numbers of 5.37 and 7.38 and at about 1050° K for Mach
number 10.6. Data were obtained at a constant Reynolds number of 8.2X10°/m at all Mach
numbers except 2.00 and 10.6 where the Reynolds number was limited to 4.9X 10°/m because of
wind-tunnel limitations.

The model was sting-mounted through the aft upper surface of the body; this method of
support was used so as to maintain a smooth lower body surface for testing at hypersonic speeds.
Force and moment measurements were made with an internally mounted, six-component
strain-gage balance. Test angles of attack ranged nominally from -2° to +15°, and angles of sideslip
ranged nominally from -2° to +8° at about 5° angle of attack. Additional tests were conducted for
the model in pitch at a constant angle of sideslip. The angles of attack and sideslip were corrected
for wind-tunnel-flow misalinement and for balance and sting deflections caused by the aerodynamic
loads. Balance cavity pressure was measured and the drag data were adjusted to a condition
corresponding to free-stream static pressure in the cavity.

Generally, boundary-layer transition was not fixed on the model, but grit was used in some
studies at several of the lower Mach numbers to provide an all-turbulent boundary layer as a basis
for data evaluation. At the hypersonic speeds, no effective method was found for fixing transition
near the leading edges of the model components to achieve fully turbulent flow. Studies utilizing
sublimation techniques and Reynolds number variation indicated the hypersonic boundary layers to
be nearly all laminar with possible small areas of transitional flow. The results of the grit and
Reynolds number variation studies are presented in a later section of the report.

Based on repeatability of the data and known precision of the measuring equipment the test
Mach numbers 0.65–2.00 and 5.37–10.6 are considered accurate within +0.01 and +0.05,
respectively; the corresponding dimensionless aerodynamic coefficients are considered accurate
within +2 and +3 percent, respectively. The angles of attack and sideslip are considered to be
accurate within +0.2”.

RESULTS AND DISCUSSION

The experimental results are presented in figures 3 through 20. The contents of these figures
are summarized in table 1 which lists the configurations and briefly notes the purpose of each
figure.

Component Buildup

Longitudinal characteristics – The longitudinal aerodynamic characteristics of the body alone

and in combination with horizontal tails, vertical tails, and a canard are presented as a function of
lift coefficient (fig. 3) and of Mach number (fig. 4). Table 2 is a tabulation of the data used in
figure 3 for selected Mach numbers. The balance cavity axial-force coefficients subtracted from the
drag measurements of the body alone configuration are listed in table 3 (similar corrections were
applied to the data of the other configurations). The lift curves for the body alone configuration
were nonlinear (particularly above an angle of attack of about 5°) at all Mach numbers of the test
with increasing lift-curve slope for increasing lift coefficients. Also, the body alone configuration
had the lowest lift-curve slope for all test Mach numbers (fig. 4). The addition of horizontal tails not
only produced essentially linear lift curves through Mach number 2, but the additional lifting area
substantially increased lift coefficient for a given angle of attack at all Mach numbers of the test.
The vertical tails and canard generally had no significant effects on the lift characteristics of the
model. However, at M = 1.10 and 1.30, the vertical tails reduced the lift coefficients, and at
M = 10.6, the canard increased the lift coefficients at positive angles of attack.

The addition of model components resulted in increases in drag coefficient at zero lift (CD.)
as shown in figure 4. The horizontal tails reduced the drag due to lift associated with the body alone
configuration (fig. 3) at all test Mach numbers.

Adding the horizontal tails increased the maximum lift-to-drag ratio ((L/D)max) above those
of the body alone configuration, particularly for Mach numbers from about 1 to 5 as indicated in
figure 4. The vertical tails caused fairly large losses in (L/D)max at superSonic Mach numbers.
Generally, the addition of model components had only small effects on (L/D)max at Mach numbers
above about 5. Values of untrimmed (L/D)nax varied from about 4.1 at M = 5 to about 3.2 at
M = 10.

In general, the horizontal tails tended to increase longitudinal stability, the canard tended to
reduce longitudinal stability, and the vertical tails had only small effects on longitudinal stability at
all Mach numbers of the test (fig. 3). For most Mach numbers below about 2, the pitching-moment
curves were reasonably linear for a range of lift coefficients beyond those for (L/D)max. At the
highest test angles of attack, some indication of pitch-up was exhibited at the subsonic Mach
numbers (fig. 3(a)–(c)). All configurations tested had positive stability near (L/D)max for the
Selected moment reference center of 0.325& (0.550l). From mission studies involving this

4
: configuration (refs. 8–11), it appears that the center of gravity would be located aft of this point at
: approximately 0.4456 (0.630l). As shown in figure 4, the aerodynamic centers moved forward with
increasing supersonic Mach numbers and gradually moved rearward at the higher hypersonic speeds.
The overall travel from the most aft to the most forward location was about 16 to 20 percent of the
mean aerodynamic chord depending on the configuration.

As previously indicated, the sting support exited from the aft, upper surface of the model in
order to provide an undisturbed lower body surface for testing at hypersonic speeds. The sting
exiting in this manner tended to produce a region of higher pressure on the aft upper surface than
would be obtained without the sting. This increased pressure resulted in a slightly negative CLo and

a small positive Qmo at the lower speeds as seen in figure 3 for the body-alone configuration. As
anticipated, this effect essentially disappeared at the hypersonic Mach numbers of the test.

Lateral-directional characteristics— The effects of component buildups on the

lateral-directional aerodynamic characteristics are presented as a function of 3 in figure 5 and
summarized as a function of Mach number in figure 6. The effects of angle of attack are shown in
figure 7. In general, the lateral-directional characteristics were nearly linear over the angle-of-sideslip
range of the test at all Mach numbers (fig. 5). The body-alone configuration had negative directional
: stability at transonic speeds and essentially neutral stability at all other Mach numbers. Adding the
vertical tails produced positive directional stability at all Mach numbers and angles of attack of the
test. Making the vertical tails smaller would probably increase (L/D)max. The canard had no effect
on Cn, except at M = 1.3 where adding this surface slightly increased directional stability
* (fig. 5(c)). There were only small effects of angle of attack on Cn (fig. 7).

: Near o = 5°, all configurations had positive effective dihedral (-CI.) at all test Mach numbers
and, in general, adding model components had little effect on this parameter (fig. 6). The variation
of C with angle of attack (fig. 7) indicates the increase in effective dihedral for angles of attack
beyond that in figure 5.

Except for the body-alone results near M = 1, all configurations had negative values of CY,
Also, there were only small effects of angles of attack on CY (fig. 7).

Horizontal-Tail Deflection

Symmetrical deflection – The effects of negative horizontal-tail deflections on the longitudinal

aerodynamic characteristics for the configuration with the canard off are presented as a function of
lift coefficient (fig. 8) and Mach number (fig. 9). Deflecting the horizontal tails negatively had no
effect On “La in the test Mach number range (fig. 9), but at a given angle of attack, CL was

reduced, as expected. At all speeds, CDo increased, but at Mach numbers of 2 or less, deflecting the
horizontal tails reduced the drag due to lift. Except for the -6.5° deflection at Mach numbers of 0.6
through about 2.5, deflecting the horizontal tails reduced (L/D)max.

At supersonic Mach numbers, deflecting the horizontal tails had little or no effect on
longitudinal stability, but at subsonic and hypersonic speeds the aerodynamic center moved forward
(fig. 9). For the selected moment reference center, about -6° of horizontal tail deflection was
required to trim near (L/D)max at the higher Mach numbers, causing a loss of about 0.5 in L/D. A
more aft center-of-gravity location (representative of the airplane used in the mission studies of
refs. 8–11) would result in small trim penalties at the hypersonic Mach numbers. On a more refined
configuration, a small amount of negative camber could be provided in the forward part of the
fuselage which should essentially eliminate longitudinal trim penalties at the hypersonic speeds.
Because of the large aft movement of the aerodynamic center at transonic Mach numbers (fig. 9),
there was insufficient control power available to trim near (L/D)max for the selected moment
reference location. A more aft center-of-gravity location or greater control deflections or both
would provide trim nearer (L/D)max, but the deflections would result in severe trim penalties. Thus
it appears that a program of fuel management or a fuel transfer system would be necessary in order
to reduce the large trim penalties at transonic speeds.

The effects of deflecting the horizontal tails on the longitudinal aerodynamic characteristics
for the complete configuration with the canard are presented in figures 10 and 11. These results are
similar to those for the canard-off results, but the trim penalties would be slightly less since adding
the canard moved the aerodynamic center approximately 5 percent farther forward.

Differential deflection – The effects of differential deflection of the horizontal tails on the
hypersonic lateral-directional aerodynamic characteristics as a function of angle of attack are
presented in figure 12 for Mach numbers 7.38 and 10.6. Positive deflections of the left horizonta
tail produced positive rolling moments, but also caused large values of adverse yaw, which worsened
with increasing angle of attack. Equal but opposite deflections (plus left and negative right) of the
horizontal tails produced positive rolling moments and improved yawing moments, which however,
became adverse at angles of attack above 3°. These results indicate that an upward deflection of the
right horizontal tail would probably provide positive roll with acceptable yaw. The effectiveness in
producing rolling moments by differential deflection of the horizontal tails indicates that elevators
on the horizontal tails might suffice for roll control; however, for these tests, the model was not
provided with this type of control.

Canard Deflection -

The effects of deflecting the canard on the longitudinal aerodynamic characteristics are
presented in figures 13 and 14 as functions of lift coefficient and of Mach number, respectively.
Deflecting the canard had only minor effects on CL.
Cy
(fig. 14) and, except at the highest Mach
numbers, the lift curves were essentially linear. As expected, the canard when deflected
increased CDo and decreased (L/D)max at all Mach numbers. In contrast to deflecting the
horizontal tail, the canard had little or no effect on the drag due to lift at Mach numbers below 2.
Deflecting the canard had essentially no effect on the aerodynamic center location at any of the tes:
Mach numbers. Like the horizontal tail, the canard was capable of trimming the vehicle near
(L/D)max at the hypersonic Mach numbers for the selected moment reference point. However, aſ
the lower Mach numbers, the canard was even less effective than the horizontal tail in trimming the
configuration.
 Rudder Deflection

Rudder flare— The effects of rudder flare on the hypersonic longitudinal aerodynamic
characteristics at Mach numbers 7.38 and 10.6 are presented in figure 15. Both outboard and
inboard deflections are considered. Except for the 30° deflections, flaring the rudders had only
minor effects on the lift curves. Rudder flare increased the CDo and reduced the (L/D)max dS

expected. Flaring the rudder produced positive values of Cmo' and for the highest deflection, a
slight increase in stability was obtained. These effects may be attributed primarily to the 50.4° of
sweepback of the simulated rudder hinge line (fig. 1(b)), where a downward component results
from the rudder loading. With exception of CDo at Mach 10.6, there was little difference between
the effects of outboard and inboard rudder flare on the longitudinal aerodynamic characteristics.

The effects of outboard and inboard rudder flare on the lateral-directional characteristics are
presented in figure 16 for Mach numbers of 7.38 and 10.6. Outboard rudder flare increased
directional stability, but flaring the rudders inboard had very little effect. The 15° rudder flares
(both outboard and inboard) had only small effects on Cl, the 30° outboard flare reduced Cl, to
zero. Rudder flare had only minor effects on CY, which increased slightly with outboard
deflections. 6

Individual rudder deflection— The effects of outboard or inboard deflection of individual

rudders on the lateral-directional aerodynamic characteristics in sideslip are presented in figure 17.
The rudder was effective in changing the yawing moments at nearly all Mach numbers,
with Cn becoming more negative with increasing left rudder. The inboard deflections were
generally not as effective in varying Cn as were the outboard deflections, and inboard deflections
produced adverse rolling moments. Rudder deflections changed C1, and the direction of this change
(in a positive or negative sense) depended first on whether the inboard or outboard surface was
deflected and second on the downward component of the rudder load caused by the 50.4°
sweptback hinge line.

Figure 18 presents the effects of individual rudder deflections on the lateral-directional

aerodynamic characteristics in pitch. At the lower Mach numbers, the effects of rudder deflections
were relatively constant with angle of attack up to about 8°. At Mach number 7.38, the
effectiveness of individual deflections of the rudder decreased significantly with increasing angle of
attack.

Boundary-Layer Studies

Studies were conducted to determine the nature of the boundary layer on the wind-tunnel
model. Some of the results of these investigations are presented in this section.

Grit studies— Grit-type boundary-layer trips were used to establish the drag level at Mach
numbers 0.90, 1.30, and 1.99 for the wind-tunnel model with an all turbulent boundary layer. The
model configuration used for these tests was the body with horizontal and vertical tails. Grit was
applied around the model nose, along the top and bottom of the body near the leading edges
(similar to the procedure used on delta wings), and near the leading edges of the tail surfaces. A drag
polar was then obtained for six different grit sizes at each Mach number, and the drag level for an
all-turbulent boundary layer was determined by the procedures described in reference 12. These
results are presented in figure 19.

For the subsonic Mach number of 0.90, the drag coefficient versus grit size is plotted in
figure 19(a) for various lift coefficients. The plateau on the curve defines the grit-free drag level
(ref. 12) for an all-turbulent boundary layer for particular lift coefficients. The drag coefficients
defined in this manner are plotted in the lower part of figure 19(a) in the form of a grit-free drag
polar for an all-turbulent boundary layer. Figures 19(b) and (c) present the results obtained for the
supersonic Mach numbers of 1.30 and 1.99. Here the drag coefficient is plotted versus the square of
the grit dimension for various lift coefficients. A linear extrapolation of the data to zero-grit size
(ordinate) defines the drag levels for an all-turbulent boundary layer for each lift coefficient
(ref. 12). The grit-free drag polars for an all-turbulent boundary layer are plotted at the bottom of
the two figures. The results obtained from these grit studies are summarized in figure 19(d) in the
form of CD, and (L/D)max as a function of Mach number. For comparison, the data for the body
with horizontal and vertical tails and with untripped boundary layer (from fig. 4) are also presented
in figure 19(d). As can be seen, extrapolating to all-turbulent boundary-layer conditions on the
model resulted in a small decrease in (L/D)max.

Reynolds number variation – There was no effective method of fixing transition near the
leading edge of the model components at the hypersonic Mach numbers. On the basis of
sublimation studies on similar models in the 3.5 foot hypersonic facility (ref. 5), it was concluded
that the boundary layer on the present model would be mostly laminar with possible small areas of
transitional flow at the hypersonic Mach numbers of the tests. To support this conclusion, Reynolds
number variation studies were conducted similar to those of reference 5. The configuration
consisting of the body with horizontal and vertical tails was used for these tests. Three unit
Reynolds numbers were investigated at M = 5.37 and 7.38; the resulting drag polars are plotted at
the top of figures 20ſa) and (b). Since it was suspected that the boundary layer on the model was
laminar, the values of CD, and drag coefficient at (L/D) Ina X from the polars at different Reynolds
numbers were plotted versus the parameter 1/VRe, which is representative of a drag-coefficient
variation associated with a laminar boundary layer. These results are shown at the bottom of figures
200a) and (b). An extrapolation of the resulting straight lines back to the ordinate (infinite Re), as
represented by the dashed lines, indicates the pressure drag of the configuration, which agrees well
with the calculated pressure drag for the model. These theoretical estimates were based or
tangent-wedge theory for all windward surfaces of the model using the method described in
reference 13. A Prandtl-Meyer expansion was employed on the leeward or expansion surfaces. Thus,
this analysis and the previous sublimation studies indicate that the boundary layer on the model was
mostly laminar at the hypersonic Mach numbers of this investigation.

CONCLUSIONS

An experimental investigation of the aerodynamic characteristics of a model representative of

an all-body, hypersonic cruise aircraft was conducted at Mach numbers from 0.65 to 10.70. The
configuration had a delta planform with an elliptic cone forebody and an elliptic cross-section
afterbody. The effects of varying angles of attack and sideslip, Mach number, and configuration
buildup were considered. In addition, the effectiveness of horizontal tail, vertical tail, and canard
stabilizing and control surfaces was investigated. The following conclusions are drawn from these
results:

1. For Mach numbers below about 2, the complete configuration exhibited essentially linear
lift and pitching-moment curves. At all Mach numbers of the test, the lateral-directional
characteristics of the complete configuration were nearly linear over the angle-of-sideslip range.

2. Values of untrimmed maximum lift-drag ratio, (L/D)max, at hypersonic speeds varied from
about 4.1 at Mach number 5 to about 3.2 at Mach number 10.

3. For the selected moment reference center, all configurations tested were longitudinally
stable near (L/D) max at all Mach numbers of the test.
4. The configurations with the vertical tails had positive directional stability for the Mach
number and angle-of-attack ranges of the test.

5. The aerodynamic centers moved forward with increasing supersonic Mach numbers and
gradually moved rearward at the higher hypersonic speeds. The overall travel from the most aft to
the most forward location was about 16 to 20 percent of the mean aerodynamic chord, depending
on the configuration.

6. Trim penalties were small at the hypersonic speeds for a center-of-gravity location
representative of the airplane; but because of the large rearward movement of the aerodynamic
center, trim penalties were severe at transonic Mach numbers.

7. The horizontal tails provided marginal longitudinal trim capability except at the hypersonic
Mach numbers. The canard was even less effective in providing longitudinal trim.

Ames Research Center

National Aeronautics and Space Administration
Moffett Field, Calif., 94035, August 30, 1971

REFERENCES

1. Gregory, Thomas J.; Petersen, Richard H.; and Wyss, John A.: Performance Tradeoffs and
Research Problems for Hypersonic Transports. J. Aircraft, vol. 2, no. 4, July-Aug. 1965,
pp. 266–271.

2. Petersen, Richard H.; Gregory, Thomas J.; and Smith, Cynthia L.: Some Comparisons of
Turboramjet-Powered Hypersonic Aircraft for Cruise and Boost Missions. J. Aircraft, vol. 3,
no. 5, Sept-Oct. 1966, pp. 398–405.
 Jarlett, F. E.; Performance Potential of Hydrogen Fueled, Airbreathing Cruise Aircraft. General
Dynamics/Convair Diyision Rep. GD/C–DCB-66–004, vols. 1–4 (Contract NAS2–3 180),
May and Sept. 1966.

. Morris, R. E.; and Williams, N. B.: Study of Advanced Airbreathing Launch Vehicles With
Cruise Capability. Lockheed-California Company Rep. LR21042, vols. 1–6 (Contract
NAS2–4084), Feb. 1968.

Nelms, Walter P., Jr.; and Axelson, John A.: Longitudinal Aerodynamic Characteristics of
Three Representative Hypersonic Cruise Configurations at Mach Numbers From 0.65 to
10.70. NASA TM X–21 13, 1970.

Nelms, Walter P., Jr.; and Axelson, John A.: Effects of Wing Elevation, Incidence, and Camber
on the Aerodynamic Characteristics of a Representative Hypersonic Cruise Configuration at
Mach Numbers From 0.65 to 10.70. NASA TN D–6049, 1970.

Nelms, Walter P., Jr.; Carmichael, Ralph L.; and Castellano, Charles R.: An Experimental and
Theoretical Investigation of a Symmetrical and a Cambered Delta Wing Configuration at
Mach Numbers From 2.0 to 10.7. NASA TN D–5272, 1969.

Gregory, Thomas J.; Ardema, Mark D.; and Waters, Mark H.; Hypersonic Transport
Preliminary Performance Estimates for an All-Body Configuration. AIAA Paper 70–1224.
Presented at the AIAA 7th Annual Meeting and Technical Display, Houston, Texas.
October 19–22, 1970.

Williams, Louis J.: Estimated Aerodynamics of All-Body Hypersonic Aircraft Configurations.

NASA TM X–2091, 1971.

10. Gregory, Thomas J.; Williams, Louis J.; and Wilcox, Darrell E.: The Airbreathing Launch
Vehicle for Earth Orbit Shuttle – Performance and Operation. AIAA Paper 70–270.
Presented at the AIAA Advanced Space Transportation Meeting, Cocoa Beach, Florida,
Feb. 4–6, 1970.

11. Gregory, Thomas J.; Wilcox, Darrell E.; and Williams, Louis J.: The Effects of Propulsion
System – Airframe Interactions on the Performance of Hypersonic Aircraft. AIAA Paper
67–493. Presented at the AIAA 3rd Propulsion Joint Specialist Conference, Washington,
D. C., July 17–21, 1967.

12. Braslow, Albert L.; Hicks, Raymond M.; and Harris, Roy V., Jr.: Use of Grit-Type
Boundary-Layer-Transition Trips on Wind-Tunnel Models. NASA TN D–3579, 1966.

13. Gentry, Arvel E.: Hypersonic Arbitrary-Body Aerodynamic Computer Program. Douglas Rep.
DAC–56080, vols. 1 and 2, 1967.

10
 TABLE 1.- SUMMARY OF FIGURES

Figure Model configuration Purpose of figure

3, 4 B, BH, BHV, BHVC Effect of component buildup on longitudinal
characteristics

5, 6, 7 B, BHV, BHVC Effect of component buildup on lateral-directional

characteristics in sideslip and in pitch

8, 9, 10, 11 BHV, BHVC Effect of horizontal-tail deflections on longitudinal

characteristics

12 BHV Effect of differential horizontal-tail deflections on

lateral-directional characteristics in pitch

13, 14 BHVC Effect of canard deflections on longitudinal

characteristics

15 BHVR Effect of rudder flare on longitudinal characteristics

16 BHVR Effect of rudder flare on lateral-directional

characteristics in sideslip

17, 18 BHVR Effect of individual rudder deflections on

lateral-directional characteristics in sideslip and in
pitch

19 BHV Boundary-layer transition results

20 BHV Reynolds number variation results

11
 TABLE 2. — TABULATED DATA

(a) M = 0.65; Re/m = 8.20X 10°

o: CL CD Cm CN CA L/D o: CL CD Cm CN CA UD | –
B BH | –
-0.38 -0.030 0.0056 || 0.0075 |0.030 0.0054 -5.36 -0.40 |40.038 |0.0073 || 00120 0.038 |0.0070 $2.
.68 || -.002 .0054 .0044 || -.002 .0054 -.33 .74 .004 0070 0021 .004 .0070 58 |

1.81 .027 .0058 .0011 .028 .0049 4.76 1.86 ,044 .0077 || -.0076 ,044 .0062 sm -

3.93 .080 .009 || || --0050 ,081 .0035 8.86 3.86 .123 .01.34 || -.0271 .124 0051 |921
5.99 .139 .015 1 || -.01.23 .139 .0005 9. 19 6.03 .208 0235 | -,0478 209 0016 || 8.83 º

8.13 .204 .0259 || -.02.11 .206 || --0032 7.87 8.13 295 .0397 || -,0677 .298 || -0025 14. º
10.31 .296 ,0492 -.0331 .300 || --0046 6.02 10.82 438 .0782 || -,0972 445 || -0054 || 560 |
12.55 .399 || 0846 -.0469 | 408 || --0041 || 4.72 12.48 .518 .1095 || -.1099 || 530 -005) || 414
14.91 ,507 .1293 || -.0606 .524 -.0056 3.92 …'

15.84 .549 .1495 || -.0652 .569 || --0061 3.67 .

BHV BHVC – , -

-0.18 -0.045 0.011 1 || 0.0146 -0.045 0.01 || 0 || -4.07 -0.42 |40,057 |0.0136 || 0.0151 || 0.057 00132 |4.1%
.73 || -.010 .01 12 .0062 | -.010 .0113 -.88 .72 -.013 0.135 .0064 - 013 || 0137 || -94
1.93 .036 .0118 || -,00.48 .036 .01.06 3.03 1.82 .028 .0143 || --0007 028 || 0135 | 192
3.99 .118 .0172 -.0251 .119 .0089 6.86 4.01 .118 .0196 || -.0184 .119 | 01:13 || 601
6.14
8.20
.205
295
.0280 -.0469
.0440 || -.0683
.207
.298
.0059
.0015
7.33
6.70
6.17
8.31
209
.305
.0307 || -,0359
.0486 || -.0540
.211 0080
309 || 0040 6.18
º -

10.43 401 .0719 - 0863 .407 || --0019 5.58 10.48 .395 .0726 -.0688 402 | .0005 | 544
12.74 .509 .1090 -.1044 .521 -.0060 4.67 12.62 493 .1079 || - 08:12 .504 |-0024 457 !-
14.81 ,598 .1494 || -.1 158 .616 || -.0084 4.00 15.12 .594 .1554 || --0876

(b) M = 0.90, Relm = 8.20x10°

B BH
-1.75 -0.071 0.0178 0.0155 -0.072 0.0156 -4.01 -0.60 -0.059 0.0185 0.0199
-.45 || -.036 .0164 .01.00 || -.036 .0161 || -2.17 .36 -.019 .0172 .0090
.46 || -.011 .0158 .0066 || -.011 .0159 -.70 1.42 023 .0172 | -0021
1.54 .018 .0168 .0024 ,019 0163 1.08 2.53 .074 .0190 || -.0167
3.67 .077 .0.197 -.0062 .078 10148 3.88 3.61 .124 .0253 || -.0307
5.81 .141 .0280 || -.0162 .143 .0.136 5.03 6.32 .245 ,0404 || -.0635
8.16 .225 .0423 -.0306 .229 .0098 || 5.33 7.93 .324 || 0562 || -.0851
10.23 .313 .0661 || -.0423 .320 .0094 || 4.74 9.27 .393 .0742 || -.1021
12.46 415 ..1004 || -.0584 .427 .0085 || 4.13
14.75 .523 .1451 || - 0753 .543 .0072 3.61

BHV BHVC
-0.58 || -0.072 || 0.0219 || 0.0241 || -0.72 || 0.0212 | -3.29 -0.63 || 0.075 0.0246 0.0222
.57 || -,024 || 0220 | |011 1 || -,024 || 0222 || -1.11 .61 -,023 || 0244 || 01.05
1.49 .013 .0226 .0012 .014 .0223 .59 1.46 .011 .0248 .0033
3.83 | .118 || 0281 || -.0285 .120 | 0201 | 4.22 3.77 .1 18 || 0320 -.02.13
6.01 .217 .0405 || -.0558 .220 0.175 5,37 5.98 .222 .0444 -.0439
8.22 | .326 | 0607 || - 0860 | .332 || 0134 5.38 8.32 .337 || 0675 | -,0682
10.38 .421 .0891 || -.1032 430 .0118 4.73 10.38 424 .0937 - 0863
12.64 .507 .1243 || -.1 154 ,522 .01.03 4.08 12.89 .513 .1326 || -,0904
14.93 ,589 .1655 - 1222 .612 0083 3.56 15.00 ,599 .1743 || - 0.981

12
 TABLE 2.- TABULATED DATA – Continued.

- (c) M = 1.10: Re/m = 8.20X 10°

a || C CD | Cº. CN |CA | ID | < | c. CD on CN CA UD
– B BH

_| -2.72 || 0.104 |0.0411 |0.0284 |40.105 || 0.0361 | -2.52 -1.66 |-0.096 | 0.0410 || 0.0360 || 0,097 | 0.0382 | -2.34
-1.48 || -.069 || 0388 || 0214 || -,070 || 0370 -1.77 -.82 -.059 || 0394 || 0231 || -.060 | 0386 -1.50
-.70 || -.046 || 0373 || 0160 | -.046 || 0368 || -1.23 .38 -.006 || 0385 || 005 | | -.006 || 0385 | -.15
, .39| -005 || 0372 || 0053 -005 || 0373 -.14 1.60 ,044 || 0390 -.01.19 || 045 || 0378 || 1.13
| | 1.61 | 029 || 0374 |-0026 030 || 0365 .76 3.56 .133 || 0443 -.0395 || 135 | 0360 2.99
|| 3.83 || 090 || 0411 |-0141 || 093 || 0350 || 2:20 5.69 220 || 0548 -.0666 224 || 0327 | 402
4.95 | 1 16 || 0441 || -.0161 .119 || 0339| 2.63 7.18 .282 || 0663 -.0860 | .288 . .0306 | 4.25
*| 5.75 .140 | 0473 || -0212 144 || 0330 2.96
*| 8.61 | .242 || 0653 || -0429 || 249 || 0284 || 3.70
| 10.26 .320 | 0863 || - 0589 .330 | 0279 || 3.71
12.42 || 4 || 0 || 1 166 || -.0775 425 || 0257 3.52
14.75 || 5 || 3 | .1575 || -,0986 .536 || 0217 | 3.26

- BHV BHVC
~| 0.50 |-0.064 || 0.0456 || 0.0270 |-0.064 || 0.0451 -1.39 -1.53 -0.106 || 0.0494 || 0.0350 | -0.107 || 0.0465 | -2.14
.51 || -.022 || 0450 | .0143 || --021 || 0452 || -.48 -.47 || -.066 || 0473 || 0256 || -.067 || 0468 || -1.40
1.48 || 016 | .0444 || 0023 || 017 | 0440 | .35 .52 || -.020 || 0461 || 0 130 | -.020 | 0463 || -.44
3.75 | 101 || 0488 -,0238 .104 || 0421 2.07 1.64 .012 || 0463 | |0069 || 013 || 0459 || 25
5.96 .191 || 0592 || -,0520 | .197 || 0390 3.24 3.88 .103 || 0507 || - 0.165 | .106 || 0436 || 2:03
* 8.01 || 288 | 0737 || -0843 295 | 0328| 3.91 5.97 .191 || 0608 || -.0387 | .197 || 0405 || 3.15
"| 8.78 .323 || 0807 |-0957 || 331 || 0305 || 4.00 8.26 .304 || 0804 || -.0676 .312 || 0359 || 3.78
jº 9.15 .364 || 0922 -,0890 .374 || 0331 || 3.95

(d) M = 1.30; Relm = 8.20x10°

~ B BH

_| 2.79 |0.091 |0.0358 |0.0225 |0.092 |0.0314|| 2.53 2.68 |0.111 || 0.0380 |0.0366 |0.113 |0.0328|292
-1.61 | -.058 || 0334 || 0 155 || -.059 || 0318 || -1.74 || -1.74 || -.074 || 0356 || 0250 -.075 || 0333 || -2.08
* -42 | -.025 || 0319 || 0085 -,025 | 0317 | -.78 -.99 || -.038 || 0347 || 0 139 -.038 || 0340 || -1.09
.44 -.002 || 0317 | |0044 -.002 || 0318 -.07 .31 .003 || 0337 || 0016 || 003 || 0337 || 08
| 1.50 | 028 0324 -0019 || 029 || 0317| 86 1.60 .050 || 0346 || --0130 || 051 || 0332 | 1.43
| 3.75 | 090 || 0370 || --0140 || 092 | 0310 2.44 3.59 .122 || 0402 || -.0347 .124 || 0325 || 3:03
* 5.95 || 153 || 0451 -.0254 .157 || 0290 3.39 5.83 .207 || 0515 -.0594 || 2 || 1 || 0302 || 4.01
| 8.02 . .221 || 0585 -.0375 .227 || 0271 3.78 7.98 291 || 0688 -.0818 .298 || 0277 || 4.23
| 10.27 | .299 || 0808 || -0496 | .308 || 0262 | 3.70
|12.76 | 384 | 1109 || - 0623 .399 || 0232 | 3.47
|14.82 || 457 | .1431 || -0731 || 478 || 0215 || 3.19
BHV BHVC

^2.53 [0.2] Toomas T00424 |0,123 Toossoſ 274 T262 Tops Tooap Toosss Toºg TooalsT26,
* -1.58|-087 || 0418 || 0317|-088 | 0394 | 2:07 || -1.69 | -089 || 0442 || 0279 -090 || 0416 || 2:01
-47 || -047 || 0397 || 0201 || -048 || 0393 | -1.19 -.62 | -.053 || 0424 || 0.195 || -.053 || 0418 -1.24
65 -,006 || 0391 || 0082 | -006 || 0391 || -.16 .52 || -,014 || 0416 || 0103 || -014 || 0418 || -.34
; 1.64 | 028 0397 -0019 | 029 || 0389 .71 1.63 .024 || 0420 || 0012 || 026 || 0413 | .58
3.90 .111 || 0453 -,0266 .113 | 0377| 244 3.83 .108 || 0468 || -.0188 .111 || 0395 || 2.30
5.93 .186 || 0548 || -0494 | .191 || 0352 340 6.06 .198 || 0582 -,0396 || 203 || 0370 3.40
8.17 | 279 || 0724 |-0751 287 || 0320|| 3.86 8.29 .289 || 0769 || - 0600 | .297 || 0344 || 3.76
, 10.30 .366 | 0963 || -.0961 | .377 | 0293 || 3.80 | 10.58 .383 || 1042 -,0791 .395 || 0322 || 3.67
13.23 || 490 .1455 -.0997 || 510 || 0295 || 3.37
v–

13
 TABLE 2. — TABULATED DATA – Continued.

(e) M = 2.00; Re/m = 4.90X 10°

a Clſ CD Cº. CNICA LID | < |Clºſ CD cm | CN | CA ||LD

B BH

-3.27- 0.077 0.0248 |0.0145 |0.078 |0.0204 || 3.09 -2.22 || 0.063 || 0.0241 || 0.0178 -0.064 0.02 17 | -2.63
-2.19 || -.054 || 0229 || 0 1 07 || -.055 || 0208 || -2.35 -1.52 -,036 || 0233 || 0 1 11 || -.037 || 0223 |-1.56
-1.26 || -.033 .02 18 || 0073 || -.033 || 0210 || -1.51 -.33 -.004 || 0225 .0030 | -.005 .0225 -.19
-.11 || -.009 || 02:13 0037 -.009 .02 13 -.42 .51 0.18 .0226 -.0025 .018 .0224 S0
.88 .013 .02 12 || 0005 .013 .0210 .61 2.91 .076 .0252 -0173 077 .02 13 || 3.01
2.97 .058 .0235 | -.0068 ,059 ,0205 2.48 5.15 .137 .0330 -.0324 .140 .0206 || 4.15
5.22 .108 ,0298 || --0146 .110 || 0 198 3.63 7.11 .191 .0435 | -.0442 .194 .0.196 || 4.38
7.28 .158 || 0390 || -.02 19 .162 || 0186 || 4.06 9.27 .250 .0594 || -.0556 .257 .0183 || 4.21
9.31 .206 || 051 1 || -.0276 .212 .0170 || 4.04 11.39 .310 .0800 || -.0658 .319 || 0173 || 3.87
11.36 | .256 || 0675 -.0335 .264 || 0 158 || 3.79 13.51 371 ..1062 || -.0767 .385 .0166 3.49
13.57 .311 ,0905 || - 0400 .324 || 0 150 || 3.44

BHV BHVC

-3.07 || 0.097 || 0.0300 || 0.0278 || 0,099 | 0.0247 -3.25 -3. 19 || 0.101 || 0.0324 || 0.0227 |-0.103 || 0.0267 || -3.13
-2.07 || -.069 .0275 || 0206 || -,070 || 0250 -2.5| -2.18 -,072 || 0299 || 0172 || -073 || 0271 || -2.42
-1.07 || -.041 .0259 .01.35 | -.041 .0252 -1.58 -1.05 -.042 .0280 .0113 || -,042 .0272 |-1.50
.01 || -.013 .0254 | |0066 || -.013 .0254 -.50 -.06 -014 .0272 .0058 || -014 .0272 -.51
1.09 .016 .0254 || --0004 || 0 || 6 || 0251 .61 1.01 .013 .0271 .0006 .013 .0269 47
2.96 || 065 .0278 || --0130 || 067 .0244 || 2.36 3.19 ,073 .0303 || -.01 12 075 .0262 2.41
5.22 || 128 || 0354 || -.0285 | .131 .0236 || 3.62 5.30 .133 || 0378 || -.0226 .136 || 02:53 || 3.53
7.37 .190 || 0467 || -.0426 .194 .02.20 || 4.06 7.33 .191 .0490 || -.0329 .196 .0242 || 3.90
9.44 .247 .06 16 || -,0543 .254 || 0202 || 401 9.38 .252 .0650 || - 0434 260 .0230 || 3.88
11.35 | .301 ,0796 || -.0647 .311 .0187 3.79 11.60 .317 || 0874 || -,0533 .328 .02 19 || 3.62
13.52 .363 .1056 -.0759 .378 .0177 || 3.44 13.63 .375 .1129 -.06.17 .391 0214 || 3.32

(f) M = 5.37; Relm = 8.20x10°

B BH

-2.16 || 40.028 || 0.01 12 || 0.0005 || 0.028 0.0102 || 2.46 -2.28 0.035 |0.01.33 || 0.0045 ||0.035 | 0.01.20 -2.59
-1.18 -.017 | |0,106 || 0009 || -.017 || 0 103 || -1.60 -1.18 -.020 || 0123 .0027 | -.020 || 0 1 19 || -1.61
.10 || -.001 .01.05 .0004 || -.001 .01.05 -.07 -.24 -.006 || 0119 || 0012 || -.006 .01 19 || - 54
1.00 || 011 .0.108 || -.0002 .01.1 .01.06 1.02 .82 .009 0121 || -,0008 .010 .01.20 .78
2.18 || 027 . .01.16 - 0010 || 027 | |0,105 2.82 1.77 .024 || 0127 | -.0026 025 .01.20 || 1 90
3.91 .050 || 0140 -.0020 || 051 .01.05 || 3.59 3.80 ,057 || 0158 || -.0068 || 058 .01.20 || 3.58
5.83 .076 || 0.184 || -.0031 .077 | 0 106 || 4.14 5.73 ,087 .02 10 || -.0109 ,089 .0122 || 4.15
7.82 | 104 || 0252 || --0047 | .106 || 0108 || 4.14 7.64 .119 .0286 || -.0155 .122 .01.26 4.16
9.81 .134 || 0343 -.0067 || 137 .01 11 3.89 9.61 .154 || 0394 || --0212 .158 .013 1 || 3.91
11.67 .163 .0452 || -.0090 | .169 .01 12 3.61 11.53 .191 .0529 || -.0274 | 198 .01.36 || 3.61
13.42 .194 .0580 || -.0117 .202 || 0 1 14 || 3.34 13.43 230 || 0694 || -.0343 240 .01 42 || 3.31
BHV BHVC
-2.27 || 0.035 | 0.0150 |0.0052 |-0.036 || 0.0136 2.36 -2.23 || 0.039 || 0.0157 |0.0040 |-0.040 || 0.0142 -2.50
-1.21 | -.021 .01 39 .0035 | -.022 .01.35 | -1.52 -1.08 -.022 10144 || 0025 | -.022 .0 140 |-1.54
-.12 || -.006 || 0 135 .0018 -.006 || 0 135 -.41 -.22 -.009 0.139 .0015 -.009 .01 38 || -.62
.84 .009 || 0137 -.0002 .009 .01.36 .67 .64 .006 .0139 0001 .006 .01 38 45
1.99 || 027 || 0 144 || -.0025 || 027 | |0135 1.85 1.76 .025 | 0147 | -.0015 ,025 .0139 | 1.69 |
3.80 || 057 || 0173 || -.0065 ,058 ,0135 3.27 3.83 .060 | |0,178 || --0050 .061 0.138 3.38
5.85 .089 || 0228 -.01 11 ,091 .01.36 3.91 5.60 ,090 0229 || -,0081 ,092 .01 40 3.95
7.72 .121 .0304 || -.0158 .124 || 0 139 3.97 7.74 .130 || 0326 -.01.24 | .133 .01.48 || 3.98
9.57 .154 || 0405 || - 0212 .158 10144 || 3.79 9.62 .166 .0442 || --0169 | .171 01:58 3.76
11.47 .190 .0537 -.0273 .197 .01.49 3.54 11.52 .206 ,0591 || --0221 .214 .0167 3.49
13.39 .230 || 0706 || -.0343 .240 .0154 3.26 13.33 .248 .0769 || -,0274 259 0.178 || 3.22

14
 TABLE 2.- TABULATED DATA – Concluded.

(g) M = 7.38; Relm = 8.20x10°

a || CL CD | Cm | CN | CA | LD * | CL || CD | Cm | CN | CA Tº -

BH
-2.62 || -0.023 0.0095 -0.023 0.0084 || -2.44 -2.61 -0.027 || 00108 || 0.0008 || 0.028 0.0095 || -2.54
-1.52 || -.014 || 0087 -.015 0083 | -1.67 -1.07 -.013 | |0097 || 0004 || -.013 | |0095 || -1.34
-.54 || -.006 || 0085 -.006 || 0085 -.74 .50 .002 .0098 .0001 .002 ,0098 21
.62 | .004 || 0086 .004 || 0086 47 1.35 .01.1 0103 || -.0004 || 011 .01.00 | 1.08
1.29 || 010 || 0089 .01.1 .0086 | 1.16 3.47 ,036 || 0126 || - 0026 || 037 || 0104 || 2.85
3.49 || 033 || 0110 .034 .0090 2.98 5.62 .062 || 0170 || --0050 || 063 || 0109 || 3.63
5.38 || 053 || 01:44 .054 || |0094 || 3.68 7.32 ,085 || 0224 || -,00.73 || 087 || 0 1 15 3.77
7.37 .076 0198 ,078 || 0099 || 3.83 9.46 .115 || 0317 | -.01.10 | .119 || 0124 || 3.63
9.40 | 101 .0275 .104 || 0 107 || 3.66 11.26 .144 || 0421 || -.0149 | .150 | 0132 || 3.42
1 1.26 | .126 || 0367 .131 .0114 || 3.43 13.18 ..176 || 0552 - 0.190 .184 || 0137 || 3.18
13.16 .152 || 0476 .159 || 01 17 | 3.20
BHVC
-2.61 | -0.027 | 0.01.23 -0.028 || 0.01.10 | -2.23 -2.54 -0.031 0.01.27 || 0.0005 || 0.032 || 0.0113 || -2.44
-1.43 -.016 || 0 1 13 -.017 || 0108 || -1.46 -1.37 -.019 || 0 1 16 || 0006 || -.019 || 0112 || -1.63
-.46 -.007 || 01:11 -.007 || 0110 -.66 -.58 -.01.1 .01 12 || 0006 || -.011 .01 11 || -.95
.66 .005 || 01:11 .005 || 0 1 11 46 .35 .001 .01.10 || 0005 .001 .01.10 || 05
1.44 || 013 | .0114 .013 || 011 1 | 1.11 1.44 .014 || 0 1 18 || 0003 || 014 || 0 1 14 | 1.18
3.43 || 036 || 0 136 .037 || 0 1 14 || 2.65 3.40 .038 || 0.142 - 0012 || 039 || 0119 2.70
5.23 .059 0171 .060 || 0 1 17 | 3.43 5.27 .062 || 0181 || - 0026 || 063 || 0124 || 3.40
7.40 || 088 || 0238 ,090 .01.23 || 3.68 7.25 ,089 || 0246 || --0046 || 091 .01.32 || 3.61
9.33 .115 || 0321 .1 19 || 0 130 || 3.59 9.20 .119 || 0337 || --0071 .122 || 0 143 || 3.52
11.09 | .143 || 0421 .149 || 0138 || 3.40 11.09 .149 || 0450 | - 0100 | .155 || 0154 || 3.32
13.06 .176 || 0556 .184 || 0 143 || 3.17 13.04 .185 || 0599 || -0136 .194 || 0166 || 3.09

(h) M = 10.61; Re/m = 4.90x 10°

BH
-2.50 -0.015 || 0.0063 -0.015 0.0056 || -2.36 -2.49 || 0.019 |0.0085 0.0008 || -0.019 || 0.0077 -2.18
- 1.36 || -.010 | .0056 -.010 || 0053 || -1.72 -.50 -.005 0073 || 0001 || -.006 || 0072 -.75
-.45 || -.004 || .0059 -.004 || 0058 -.72 .51 .002 || 0076 || 0003 || 002 || 0076 .28
.59 || 002 | .0062 .002 || 0062 .37 1.46 ,010 || 0082 || 0007 || 010 || 0080 | 1.17
1.46 || 007 || |0067 .008 || 0065 | 1.12 3.51 .026 || 0 106 || 0014 || 027 | |0090 2.46
3.67 || 022 || 0086 .022 || 0072 2.53 5.35 ,041 0138 || 0014 || 042 | |0099 || 3:00
5.56 || 033 || 0112 ,034 || 0079 || 2.96 7.35 .062 | .0193 || -.0002 || 064 || 0112 || 3.24
7.39 || 050 || 0157 .051 .0092 || 3.15 9.29 ,091 .0276 || -.0039 || 094 || 0126 || 3.29
9.39 || 071 .0221 ,074 || 0103 || 3.21 11.25 .124 || 0390 -.0085 .129 || 0141 || 3.17
11.24 || 094 || 0301 .098 || 0 1 12 || 3.12 13.12 .159 || 0529 || --0140 .167 || 0154 || 3.01
13.11 .122 || 04:11 .128 || 0123 2.97

BHVC
–2.45 |-0.021 || 0.01.16 -0.022 || 0.0106 || -1.85 -2.50 | -0.024 || 0.0118 -0.0002 || 0.024 || 0.0108 || -2.01
-.46 || -.006 | |0098 -.006 || 0097 -.56 -1.31 -.013 || 0100 || --0001 || -,013 | |0097 || -1.31
.88 || 004 || 0102 .004 || 01.01 41 -.59 -.007 || 0099 || 0001 || -,008 || 0098 || -.75
3.61 .027 | 0129 .027 | 0112 || 2.07 1.35 .010 || 0103 || 0008 || 011 0100 | 1.01
5.68 || 045 || 016.5 .046 || 0120 || 2.71 1.93 .016 || 01:11 .0010 || 0 || 6 || 0 106 | 1.42
7.62 || 065 || 0219 ,067 || 0131 || 2.97 3.45 .030 || 0137 || 0014 || 031 .01.19 || 2.18
9.65 || 095 || 0306 ,099 || 0 143 || 3.10 5.49 .051 .0182 || 0007 || 053 || 0132 2.82
11.12 .118 || 0387 .123 || 0152 || 3.05 7.32 073 | 0235 | -.0004 || 075 || 0141 3.10
13.24 .158 || 0541 .166 || 0 || 65 2.91 9.28 ..100 || 0318 || -.0029 || 103 || 01:53 || 3.14
11.19 .129 || 0423 || --0060 || 135 || 0 165 3.05
13.11 .16.1 .0556 || --0095 .170 || 0176 2.90

15
TABLE 3. – BALANCE CAVITY AXIAL–FORCE COEFFICIENT (CAL)

B Configuration
O. M = 0.65 0.80 0.90 1.10 1.30

-2.00 0.0002 0.0001 |0.0008 0.0019 |0.0013

-1.00 .0002 .0001 || 0010 0019 || 0013
0 .0002 .0001 | .0011 .0020 || 0014
1.00 .0003 .0002 | 0011 .0020 || 0014
2.00 .0003 .0002 || 0011 .0020 0015
4.00 .0003 | . .0002 || 0010 .0020 || 0015
6.00 .0004 .0003 || 0008 || 0021 0015
8.00 .0005 .0003 | .0009 || 0023 0016
10.00 .0006 .0004 | 0012 0026 .0018
12.00 ,0008 .0005 || 0016 || 0029 || 0019
14.00 0010 .0007 || 0021 0.031 || 0020
o, M = 1.60 2.00 5.37 7.38 10.61

-2.00 || 0.0008 || 0.0004 || 0.0001 || 0 -0.0002

- 1.00 .0009 .0005 || 0001 | 0 -.0002
0 .0009 .0005 | 0001 | 0 -.0002
1.00 .0009 .0005 || 0001 | 0 -.0002
2.00 .00 10 .0006 | 0001 || 0 -.0002
4.00 .00 10 .0006 .0002 || 0001 || -.0002
6.00 .0010 .0007 | .0002 || 0001 || -.0002
8.00 .0011 .0007 | 0002 || 0001 || -.0002
10.00 .00 12 .0008 .0002 || 0001 || -.0002
12.00 .0013 .0009 .0002 || 0001 -.0003
14.00 .0014 .0009 || 0002 || 0001 -.0003
 33.31

Menodel
-
l.
Figure
are
dimensions
all
cdrawings;
square
in
areas
and
timeters.

geometry
Model 6667
17 .86
. 38 .62 .
. 16 .
35
O6 O935

21.21
Sr/Slaſt
S

l2.93
(b/2)
configuration
Complete
(a)
Horizontal
rotation
(l)
bit
ſt
a (lx)
|6.7l
O tail
(#)
l,8.26=
l+
9.6826.5|
t

Removable

\
tº body
N Section
Cross

Maximum

|
7sections
cross
N

El iptical
-
(O.3255)rotation
39
section
cross
maximum
for
Equation

6||
Canard -

32.17)
(station
32.17
Station

º
 section
cross
Typical
~ cross
Typical
c
O6
O.

–4LTIT
Rotation Canard
2
axis
c
5
O.

surface
inboard

hotail,
of
Details
(b)rizontal
canard
and
vertical
Flat

-F 9
60°
5.
c
0.0l

1 O8
ll.

-sº H–1–- lo
50.
Body
tail
Wertical
Concluded.
-
l.
Figure

c
0.5 – lO.2l

|
t

6.03
55

º
-— C
| l6
2.
-—

-T c
Hº- section
cross
Typical
--
H—
3.7l
Horizontal
tail

Rotation
axis
*-i-- | | –D
c
5
O.

3.
 (a) BHVC configuration

(b) BHV configuration

Figure 2. - Model photographs.

19
 |O

L/D

O, deg

- . 16 -.08 o .08 .16 .2|, .32 . HO .18 .56 .6),

CL

(a) M = 0.65

Figure 3. – Effect of addition of components on the longitudinal aerodynamic

characteristics.

20
 lO

L/D

. Oli

Cm

O, deg

-.16 -.08 o .08 .16 .2, .32 ...o is .56.6.

(b) M = 0.80

Figure 3.- Continued.

21
 . Oli

. Ol

.O8

. 12

C., deg

(c) M = 0.90

Figure 3. – Continued.

22
 6

l,

L/D
O

-l;

.0l.

Cm —. Ol.

-.08

- . 12

C., deg

- . 16 -.08 O .08 .16 2, .32 ho

(d) M = 1.10

Figure 3. – Continued.

23
 r-+
Q\]
!

C., deg

- . 16 -.08 O . O8 . 16 .211 .32 -. HO . 18 .56 .6l.

(e) M = 1.30

Figure 3. – Continued.

24
L/D

-l;

. Oli

-.08

-. l.2

C., deg

-.16 -.08 o .08 .16.2, .32 ..ho...a .56 .é,

(f) M = 1.60
Figure 3. – Continued.

25
 L/D

-l;

. Oli

-. Oli

-.08

C., deg

-. 16 -.08 0 .08 . 16 .2|, .32 . HO . 18 .56 .6||

CL
(g) M = 2.00
Figure 3. – Continued.

26
L/D

. Ol

C., deg

CL

(h) M = 5.37
Figure 3.- Continued.
 L/D

. Ol

—. Ol

—. O2

O, deg

CL

(i) M = 7.38
Figure 3.- Continued.

28
 L/D

. Ol

—. Ol

—. O2

C., deg

CL

(j) M = 10.6
Figure 3. – Concluded.

29
(L/D) Imax

º*

Figure li. - Variation with Mach number of the effect of addition of compo
nents on the longitudinal aerodynamic characteristics; B = 0°.

30
 –6 –) -2 o' 2 6 8 lo 12
f}, deg

(a) M = 0.65; d. = 5.3°

Figure 5. – Effect of addition of components on the lateral-directional

aerodynamic characteristics in sideslip.

31
 5, deg

(b) M = 0.90; 0 = 5.3°

Figure 5. – Continued.

32
 f}, deg

(c) M l. 30; O. = 5.1°

Figure 5. – Continued.

33
 Cn

—6 -l. o 2 6 l.O

B, deg

(d) M = 2.OO; o = 5.2º

Figure 5.- Continued.

34
 .00l.

-.OOl

-6 -l. –2 O 2 l 6

5, deg

(e) M = 7.38; d = 1.8°

Figure 5. – Continued.

35
 . Ol 2

.OO8

Cn .00l.

-.00l.

.00l.

–. OOl.

. O2

—. O2

-. Oli
O 2 l 6 lO

B, deg

(f) M = 10.6; G = li. 7°

Figure 5. – Concluded.

36
 .008

.006

—. OO2

-. OOl;

.OOl;

-. OOH

-.008

-. Ol2

Figure 6. – Variation with Mach number of the effect of addition of components

on the lateral-directional aerodynamic characteristics; O. = ||.T * to 5.1 °.

37
 Cn

—l –2 O 2 || 6 8 lo l2 li ló

O, deg

(a) M = 0.65

Figure T. - Effect of addition of components on the lateral-directional

aerodynamic characteristics in pitch, B = 1.0°.
38
Cn

2 || 6 8 lo l2 ll. l6

Q, deg

(b) M = 0.90
Figure 7. – Continued.

39
 2 l; 6 8 lC) l2 ll. l6

C., deg

(c) M = 2.00

Figure 7. – Continued.

40
 . Ol2

.008

Cn .004

–. OOH

. O2

–. O2

-. Oli
—li –2 O 2 l 6 8 lo 12 ll: 16

Q, deg

(d) M = 7.38
Figure 7. – Concluded.

41
 L/D

. O8

. Oli

. Oli

.O8

. l.2

(a) M = 0.65

Figure 8. – Effect of horizontal-tail deflections on the longitudinal aero

dynamic characteristics; canard off; B = 0°.

42
L/D

.08

. Oli

-. Oli

-.08

. 12

C., deg

O .08 .16.2, .32 ..is . 56 .6l.

(b) M = 0.80
Figure 8. – Continued.

43
 L/D

. O3

. Oli

.0l.

. O8

. 12

O, deg

.08 .16.2, .32 .io .

CL

(c) M = 0.90
Figure 8. – Continued.

44
L/D

. O8

. Oli

-. Ol;

-.08

- . 12

O, deg

2.16 -.08 o .08 .16 .2. .32 . HO . 18 .52 .6l.

CL

(d) M = 1.10

Figure 8. – Continued.

45
 . O8

. Oli

–. Oli

-.O8

O, deg

—ll -

- . 16 -.08 O -
.08 .16 .24 .32 . HO . . h9 .56 .6l.

CL

(e) M l. 30

Figure 8. - Continued.

46
 . O8

.0l.

-. Oli

-.08

O, deg

-l -- -

- . 16 -.08 O .08 . 16 .2} .32 . HO . 18 .56 .61

CL

(f) M = 1.60

Figure 8. – Continued.
 OO
C)
©

-. Oli 16

-.08

O, deg

- . 16 -.08 O
.08 .16 2, .32 ...o
CL

(g) M = 2.00
Figure 8. – Continued.
48
 . O7
L/D
. O6

.05

.o.

CD

. Ol

. Oli

. O2

.02

.0l. l6

C., deg

CL

(h) M = 5.37
Figure 8. – Continued.

49
 L/D

. Ol.

. O2

C., deg

CL

(i) M = 7.38

Figure 8. – Continued.

50
L/D

C
O
C\]

–. O2

-. Oli

O, deg

(j) M = 10.6

Figure 8. – Concluded.

51
 (L/D)nax

- - 9 La

- t
º --

t -

-
º

1. -

CDo

--

! º
ſ
º
-

- -

3, . C a . C.
• * * x l
% G %

Figure 9. – Variation with Mach number of the effect of horizontal-tail

deflections on the longitudinal aerodynamic characteristics; canard
off; B = Oº.

52
 L/D

. O8

. Ol;

O, deg

—l;
- . 16 - 08 0 .08 .16 .211 .32 .10 . , 8 .56 .64
CL

(a) M = 0.65

Figure 10. – Effect of horizontal-tail deflections on the longitudinal

o
aerodynamic characteristics; canard on, f = 0 °.
53
 L/D

.08

.01

- -
Cm
- -. Oli

- . 12

O, deg

—l. - - -

-.16 - 08 0 .08 .16 .24 .32 . HO .48 .56 . 6;

(b) M = O. 90

Figure 10. – Continued.

54
L/D

-2

-li

.08

. Ol

Cm
–. Oli

–. O8

-.l2

C, deg

-li
-.l6 -. C8 O .08 . l6 .2li .32 . li O .li8 .56 . 6li
CL

(c) M = l. 30

Figure l'O. – Continued.

55
 . Oli

Cm

. Oli

.O8

O, deg

- . 16 -.0 O .08 .16 .24 .32 .110

CL

(d) M = 2.OO
Figure 10. – Continued.

56
L/D

Cm
O
Q\]
|

–. Oli

C., deg

(e) M = 5.37

Figure 10. – Continued.

57
 L/D

. Oli

-.02

O, deg

CL

(f) M = 7.38

Figure lC). - Continued.

58
 . Ol;

Cm

C., deg

(g) M = 10.6

Figure 10. – Concluded.

59
1.

Figure ll. - Variation with Mach number of the effect of horizontal-tail

deflections on the longitudinal aerodynamic characteristics; canard
on; B = 0°.

60
 . OO8

.00l.

-.00l.

Cn -.008

—. Olº

-. Oló

-. O20

-. 0211
. 02h

.02O

. Oló

. Olſº
Cl
.OO8

.00l.

-.00l.
.O2

CY O

—. O2 – . | | | | l
—l –2 O 2 || 6 8 10 12 ll, 16
O, deg

(a) M = 7.38
Figure 12. – Effect of differential horizontal tail deflections on the
lateral-directional aerodynamic characteristics in pitch; 5 = O9.
 .OO8

. OOH

.00l.

coe
. Ol 2

. Oló

. O20
.032
.02
.O28
.O28
.02h

. O20

. Oló

. Ol2

.OO8

.00l.

.02
–. OOH
Cy

—. O2
2 || 6 8 lo
O, deg

(b) M = 10.6

Figure l?. - Concluded.

62
 L/D

. O8

. Ol

O, deg

-H
- . 16 -.08 O .08 .l6 .211 .32 . 10 . H8 .56 .64
CL

(a) M = 0.65

|Figure l?. - Effect of canard deflections on the longitudinal aerodynamic

characteristics; B = 0 °.
 .10

.09

.08

. Oli

Cm
. Oli
l6
.08
l2

.12

C., deg

*H — H l

the -os o os .16.2, .32 . HO .118 .56

CL

(b) M = 0.80
Figure lº . – Continued.

64
L/D

.08

. Oli

Cm
. Oli

. O8

.12

O, deg

(c) M = O. 90

Figure lº. - Continued.

35
-

.08

. Oli

Cm
-. Oli

–. O8

–.l2

C, deg

-li
. -lº -. O3 O .08 .l6 .2li .32 . l O .56
CL

(d) M = l. 10

Figure lº . – Continued.

66
 .08

. Ol;

—. Oli

-.08

- . 12

C., deg

-.l6 -.08 O .08 .16 .211 .32 . HO . H8 .56 .6||

CL

(e) M = l. 30
Figure lix. - Continued.

67
 L/D
！
± − × ± •

C., deg

-.16 -.08 O .08 .16 .24 .32 .10 . H8 .56 .61

CL

(f) M = 1.60

Figure liº. – Continued.

68
L/D

.08

. Oli

Cm O

–. Oli

-.08

O, deg

l; -

-.16 -.08 0 .08 .16 .211 .32 .lio .48 .56 .64
CL

(g) M = 2.00
Figure lix. - Continued.

69
 L/D

. O2

Cm

. Oli

O, deg

(h) M = 5.37

Figure l?. - Continued.

70
L/D
rd
ÒS

b0
Cl)
•

(i) M = 7.38
Figure lº. - Continued.

71
 ----
----
-+-+-+
-+-+-+
-+ - ----

Figure l?. - Concluded.

(j) M = lo. 6
L
—. O2
L/D

C., deg
Cm

72
{
----
- - - - -
----
،، ، ،
-- - - -----
- - ~~~ ~ ~ ~） _■ ******
* -
(I/D) wax

°Io

CDo

a •C-2 8 - C

% G % l

Figure ll. - Variation with Mach number of the effect of canard deflection
on the longitudinal aerodynamic characteristics; P = 0°.

73
 –2

BHV
BHVR1s (outboard)
CD - - BHVRao(outboard)
BHVR1=(inboard)

Cm O

O, deg

(a) M = 7.38

Figure lj. – Effect of rudder flare on the longitudinal aerodynamic

characteristics at hypersonic speeds; B = 0°.

74
 l,

L/D 2

O .06

–2 . O5

. Oli

.03 BHV
CD -
BHVR12(outboard)
. O2 - -
BHVRao(outboard)

. Ol
BHVR, (inboard)

. O2 O

. Ol

Cm O

-. Ol

—. O2 lé

l2

O, deg

CL

(b) M = 10.6
Figure lj. - Concluded.

75
 .O2);

. O20

. Oló

. Olſº

.008

.001

BHV
-.001
BHVR, (outboard)
BHVRagſ outboard)
-.008
BHVR1-(inboard)
.00l.

-.001

-. Oli

-.06

£, deg

(a) M = 7.38; d = ||.8°

Figure 16. – Effect of rudder flare on the lateral-directional aerodynamic

characteristics at hypersonic speeds.
 .02h

. O20

. Oló

. Ol2

Cn .008

.OOli º
BHV
-.00l.
BHVR1 (outboard)
-.008
BHVRao(outboard)
BHVR, (inboard)
.001

–.008 . O2

f, deg

(b) M = 10.6; d = ||.7°

Figure l6. – Concluded.
 O BHV
D BHVRR_o
L=15(outboard
A BHVRR_o
L=30(outboard
- - - |Q BHVRR=15(inboard)
- L=O

–6 -li –2 O 2 || 6 8 lo lº

£, deg

(a) M = 0.65; d. = 5.3°

Figure lºſ. – Effect of individual rudder deflections on the lateral

directional aerodynamic characteristics in Sideslip.
78
 -
H
l
O BHV
|- D BHVRe_
R=O
–H L-is(outboard
BHVRR-o
L=30(outboard
BHVRR=ls(inboard)
L=O

(b) M = 0.90; O. = 5.3°

Figure l7. – Continued.

79
 Cn

-. O2 O BHV
D BHVRR-o
t=is(outboard)
-. O3 |-A BHVRR-o
H i =30(outboard
Ol -. Oli o BHVRR-15(inboard)
L=0

. Oli

–. Oli

–. O6 - -
--- - l - l
*H
1–1–1–1–1–1–
–6 —l; –2 O 2 | 6 8 10 12

(c) M = 1.30; O. = 5.1°

Figure l7. – Continued.

80
 O BHV

D BHVRR-o
L=15(outboard
BHVRR-o
L=30(outboard
. Ol PHWRR=ls(inboard)
L=O

Cl
-. Ol

—. O2

Cy

-6 —l; –2 O 2 l; 6 8 10 12

£, deg

(d) M = 2.00; O = 5.2°

Figure lºſ. – Continued.

81
 . Ol?

.008

.001

Cn
-.001

O BHV
-.008 TD BHVRR-o
- - -
L=15(outboard
- - —. Ol2 *A BHVRR-o
: - -
L=30(outboard
-. Olć *W*- 5(inboard)

! .001

Cl O

–. OOl;

º --

. O2

O
CY —. O2

-6 -, -2 o' 2 l; 6 8 lO l2

£, deg

(e) M = 7.38; O. = ||.8°

Figure lºſ. – Concluded.

82
 H
A BHVRE-2
|- L=is(outboard)
BHVRR=1s(inboard)
L=O

-l –2 O 2 l; 6 8 lC) l2 ll. l6

O, deg

(a) M = 0.65

Figure 18. – Effect of individual rudder deflections on the lateral

directional aerodynamic characteristics in pitch; f = 0°.

83
 Cn

Cl

- D BHVRR a -

L=15(outboard)
O PºWRR-15(inboard)
L=o

Cy -

- -2 o’ 2 || 6 8 no 12 in 16
C., deg

(b) M = 0.90

Figure 18. – Continued.

84
 D BHVRR_o
L=15(outboard)

• *-ºſ incara)
=O

Ill 2 l; 6 8 lC) 12 ll. 16

O, deg

(c) M = 1.30

Figure 18. – Continued.

85
 Cn

+
D BHVRR-g
L=Is(outboard)
© BHVRR_o
L=30(outboard)

–2 O 2 l; 6 8 10 l2 l! l6

Q, deg

(d) M = 2.00

Figure 18. – Continued.

86
 D BHVRs a
L=is(outboard)
A BHVRR-o
L=30(outboard)
KX PFWFR=1s(inboard)
L=O

C., deg

(e) M = 7.38

Figure l8. - Concluded.

87
 CL = 0.11 K., cm
O. Ol35
0.0lS3
0.
Drag level for all-turbulent O º
boundary layer defined by 0.0460
plateau on curves olosso

CD

K., cm

Grit-free, all-turbulen
CL boundary-layer polar

To .01 .02 .03 .01 .05 .06 .07 .08 .09

CD

(a) M = 0.90

Figure lº). - Determination of the drag level for an all-turbulent boundary

layer using variable size roughness; BHV configuration.

88
 Drag level for all-turbulent
boundary layer

O 10 2O 30 |O
50×107*
K*, Sq cm

Grit-free, all-turbulent
boundary-layer polar

O . Ol .02 .03 . Oh, .05 .06 . O7 .08 .09 .10

CD

(b) M = 1.30
Figure lº). - Continued.

89
 . OC

|T
. O3H-\{ *-i-T—H
—t
H
. O'ſ -
+ | +

+– |-
Drag level for all-turbulent
.06 boundary layer

.. O

CD

O3L

O
[]
A
O K»
D
O
O lC) 2O 30 l;0 5Oxlo'º
K*, sq cm

CL
Grit-free, all-turbulent
boundary-layer polar

To .01 .02 .03 .01 .05 .06 .07 .08 .09 .10
CD

(c) M = l.99

Figure l9. – Continued.

90
 O Extrapolated to all-turbulen
boundary layer
Untripped boundary layer

(I/D) wax

(d) Summary

Figure l9. – Concluded.

91
 Rexlo'º
O 3.9
D 7.5 Experiment
^ 12.1

«» Pressure
drag |Theory

O l 2 3 l, 5

l
×lO4
V Re

(a) M = 5.37

Figure 20. – Variation in unit Reynolds number; BHV configuration.

92
 .2O
Rexlo'º
l6 O l; .9
- D 8.5 Experiment
A ll.2
. l.2 © Pressure
drag | Theory
CL .08 -

. Oli

–. Oli
O .008 . Oló .02|| . O32 . OliO . Ol. 3

. O32

. 02h

CD . Oló

.008

l
VRC ×lO4

(b) M = 7.38

Figure 20. – Concluded.

93
AsA-Langley, 1971 — 2 A-4017
NATIONAL AERO NAUTICS AND SPACE A D M is TRATION
WASHINGTON. D.C. 20546
POstage And FEE's PAid
NATIONAL Ae. RONA uTics and
OFFICIAL BUSINESS space Ad M I N is TRAT icon
PENALTY FOR PRI v ATE use $300
FIRST CLASS MAIL

NAS

0.22 002 C 1 U 35 71 1029 S004.89H

UN IV ſh;F I LL INO IS
L I BRARY
ATIN: DOCUMENT LIBRARIAN
URBAN A IL 6 1801

POSTMASTER:
.. If *-
E.M.'ſ Nº.

"The aeronautical and ſpace activities of the United States ſhall be

conducted ſo aſ to contribute . .. to the expanſion of human knowl
edge of phenomena in the atmoſphere and ſpace. The Administration
ſhall provide for the wideſt practicable and appropriate disſemination
of information concerning its activitieſ and the reſultſ thereof.”
— NATIONAL AERONAUTICS AND SPACE ACT OF 1958

NASA SCIENTIFIC AND TECHNICAL PUBLICATIONS

TECHNICAL REPORTS: Scientific and TECHNICAL TRANSLATIONS: Information

technical information considered important, published in a foreign language considered
complete, and a lasting contribution to existing to merit NASA distribution in English.
knowledge.
SPECIAL PUBLICATIONS: Information
TECHNICAL NOTES: Information less broad derived from or of value to NASA activities.
in scope but nevertheless of importance as a Publications include conference proceedings,
contribution to existing knowledge. monographs, data compilations, handbooks, WO
sourcebooks, and special bibliographies.
TECHNICAL MEMORANDUMS:
Information receiving limited distribution TECHNOLOGY UTILIZATION
OF
because of preliminary data, security classifica PUBLICATIONS: Information on technology
tion, or other reasons. used by NASA that may be of particular
interest in commercial and other non-aerospace
CONTRACTOR REPORTS: Scientific and
applications. Publications include Tech Briefs,
technical information generated under a NASA Technology Utilization Reports and
contract or grant and considered an important lall.
Technology Surveys.
contribution to existing knowledge. K

H(lh
Details on the availability of these publications may be obtained from:

SCIENTIFIC AND TECHNICAL INFORMATION OFFICE

NATIONAL AER ON AUTICS AND SPACE ADMINISTRATION

Washington, D.C. 20546
 29.13
325tn ENC

<3-2 3'-

NASA. T E C H IN I C A L N OT E

:
--

THE LIBRARY OF THE

JAN 3 1972
UNIVERSITY Q'. ILLINOIS
AT ÜßANA-CHAMPAIGN
:

VOLTAGE-PROGRAMED CONTROL
OF TURBOALTERNATOR LOAD

by Richard N. Young and Eugene L. Kelsey

Langley Research Center
Hampton, Va. 23365

| NATIONAL A ERONAUTICS AND SPACE ADMINISTRATION - WASHINGTON, D. C. • DECEMBER 1971

 1. Report No. 2. Government Accession No. 3. Recipient's Catalog No.
NASA TN D-6578
4. Title and Subtitle 5. Report Date
VOLTAGE-PROGRAMED CONTROL OF December 1971
TURBOALTERNATOR LOAD 6. Performing Organization Code

7. Author(s) 8. Performing Organization Report No.

Richard N. Young and Eugene L. Kelsey L-7963
10. Work Unit No.
9. Performing Organization Name and Address 113-34-22
NASA Langley Research Center 11. Contract or Grant No.

Hampton, Va. 23365

13. Type of Report and Period Covered
12. Sponsoring Agency Name and Address Technical Note

National Aeronautics and Space Administration 14. Sponsoring Agency Code

Washington, D.C. 20546

15. Supplementary Notes

16. Abstract

A power source and load-control concept was designed and tested for use with a
Brayton cycle turboalternator. The concept provides power that adjusts to a range of
loads while controlling alternator speed and isolating load-switching transients. The con
cept is also applicable to power-generating sources other than the Brayton cycle.

17. Key Words (Suggested by Author(s)) 18. Distribution Statement

Brayton cycle Unclassified – Unlimited

dc distribution

Space power control

Parasitic speed control
19. Security Classif. (of this report) 20. Security Classif. (of this page) 21. No. of Pages 22. Price"

Unclassified Unclassified 11 $3.00

"For sale by the National Technical Information Service, Springfield, Virginia 22151
--• •►æ~♥~- - -*--（ ：-（ ：-（ ：-*
 VOLTAGE - PROGRAMED CONTROL OF TURBOALTERNATOR LOAD

By Richard N. Young and Eugene L. Kelsey

Langley Research Center

SUMMARY

A power source and load-control concept was designed and tested for use with a
Brayton cycle turboalternator. The concept involves full rectification of the alternator
butput, forming a dc bus which provides power that adjusts to a range of loads while con
rolling alternator speed and isolating load-switching transients. The system was found
:o be highly stable and efficient, and it allows a simple method of providing overcurrent
and fault protection. The concept is also applicable to power-generating Sources other
han the Brayton cycle.

INTRODUCTION

Studies of long-term manned missions such as Manned Orbital Research Laboratory

und Space Station/Space Base have identified electrical power requirements ranging from
.0 to 100 kW. Candidate systems capable of producing this range of power have also been
dentified and are currently under development. For near-future application the leading
:ontenders are solar-array systems, Rankine cycle turboalternator systems deriving heat
rom a nuclear reactor source, and Brayton cycle turboalternator systems deriving heat
rom a nuclear reactor or radioisotope source. While each of these systems has individ
tal application problems and unique advantages, disadvantages, and constraints, inherent
eatures common to all power systems must be considered in the design of the control
ogic System.

To date work has been done in the area of speed control through the use of variable
lternator loading (refs. 1 and 2). However, these systems utilize the alternator (ac) out
ut directly, which means the loads may have to operate at frequencies as high as 1200 Hz.
The Manned Orbital Research Laboratory studies (MORL) of reference 3 indicate
nat a do system may be preferable to an ac system for spacecraft. The conclusions of
his study show that an electrical power system should be sized for average power
equirements and that variations should be furnished and absorbed by an energy-storage
evice. Such a system was designed and built as a combined project of the Langley
‘esearch Center and Manned Spacecraft Center (ref. 4). A breadboard load logic and
ontrol system simulating partial spacecraft loads was integrated with the Brayton Cycle
Demonstrator. Parasitic loading, mandatory real loading, permissive real loading, and
conditioning equipment were integrated into the control method, and distribution and load |
priority considerations were included.
The system described in reference 4 uses the rotational speed of the alternator as
the main control input to the load and power management computer. Such a control phi
losophy requires that all load transients generated in the system be reflected through the
alternator and measured as speed changes. Thus the response of the system is greatly
influenced by the rotating mass of the alternator. The combination of time delays inher
ent in such a system makes stability difficult to achieve, and system response to pertur
bations is slow.

The system described in the present paper eliminates the single major source of
time lag. The system control input is no longer the alternator speed, but rather the alteſ:
nator terminal voltage, which is more sensitive to changes in load level. This approach
eliminates a large portion of the power conditioning equipment necessary in the original
approach, while improving the system response characteristics.

OPERATIONAL CONCEPT

The general design of any spacecraft power conditioning system must include the
subsystems in figure 1. In practice, other subsystems may also be necessary. When
the power source is a Brayton or Rankine turboalternator, the full generated electric

Power Source

Power —- Energy
conditioning storage

Main electric buses

Figure l. — Basic load-control and conditioning system.

 power must be consumed at all times in order to prevent the rotating unit from exceeding
its design speed. Thus, such a system will require a parasitic load that absorbs excess
energy when electric usage is below the output capability of the source and full use has
been made of the energy-storage device.
r Figure 2 shows a block diagram of the system that was designed specifically for a

Brayton or Rankine power source. The system can, however, be readily modified for
other power sources. The full output power of the turboalternator is rectified to form a
dc bus. This voltage is then applied directly across the battery, parasitic load, and any
#

Field
Turboalternator kºº
regulator

3. ! |-- Speed
SensOr

3. Rectifiers

— |-

Battery

Parasitic
load

dc-to-dc
converter

dc-to-ac
converter

Figure 2. - Voltage-controlled load and conditioning system.

inverters and convertors that may be required. As a minor control loop, the alternator
speed is sensed and used to control the field regulator. The field regulator then varies
:he alternator field current to adjust output power and thus speed.
Since the output of the alternator is likely to be high voltage (100 volts or more),
iirect rectification will result in high-voltage do buses. These buses provide certain
advantages: (a) a convenient tie point if two or more alternators are to be operated in
parallel and (b) a voltage level suitable for high-power transmission over longer distances.

3
 It may be desirable to have filters on the output of the rectifiers (choke input) to
reduce the rms current to its dc value. This addition can reduce the alternator internal
losses and potentially improve efficiency. Because do grounds are not required in this
system, shock hazards are minimized, and single-order shorts to ground will not cause
a loss of power. The alternator neutral is likewise not required, and thus the problems
of single-point failures are further reduced.

Energy storage in this system is provided by nickel–cadmium batteries. This

selection was made on the basis of cost and immediate availability. Other candidates for
a flight system are reverse-cycle fuel cells or any efficient battery. The test system
was designed for a 90-volt dc bus, again because of the availability of the energy-storage
elements.

The parasitic load consists of a bank of power resistors driven by an active circuit
that is placed directly across the bus, and its action is similar to that of a large zener
diode. It functions to limit the bus voltage from rising above the end-of-charge voltage
of the battery. The parasitic load will then draw whatever current is necessary to limit
the bus voltage to this value. This scheme allows the battery to charge or fully support
the high-voltage bus with no additional power-conditioning equipment or logic circuitry
for either function. Using the natural characteristics of the battery and parasitic load in
this manner does allow the bus voltage to fluctuate with load. This feature should not be
detrimental to the low-voltage do or the inverter-supplied ac buses. Inverters and con
verters can be designed to yield regulated output when the input varies over a moderate
range. The de-to-dc converter designed for use in the investigation of reference 4 oper
ated satisfactorily with input voltage variations between 90 and 150 volts dc. The need
to design for a variable input voltage sacrifices very little in efficiency. In general, how
ever, the higher the input voltage, the more efficient the conversion process.
The main dc converter and the ac inverter were not supplied in this demonstrator
system. Equivalent loading was simulated through the use of switched resistance.
Two alternator parameters are sensed and used as system controls. They are the
rotational speed and the terminal voltage. The speed sensor can be of a simple linear
type and is used to control the field voltage. An increase in speed produces an increase
in field, and thus more power is generated and the alternator tends to slow down; and con
versely for a decrease in speed. The terminal voltage, after rectification, is applied
directly across the battery. Also connected across the battery is the parasitic load.
When the battery voltage exceeds a preset value, the parasitic load begins to conduct cur
rent, simulating the action of a large zener diode. Thus, the battery voltage is prevented
from rising above its end-of-charge voltage. In this manner each parameter is indepen
dently controlled. Cross-coupling of control loops is kept to a minimum and system
instability is prevented.

4
 THEORY OF OPERATION

The operation of this system is analogous to that of a rechargeable battery in par

allel with a zener diode and an input that is constant in power; that is, a reduction in volt
age results in an increase in current from the source so that the product of voltage and
current is always constant. The loading is then simply resistive across the battery. The
schematic diagram of figure 3 shows this simplified model.
-Q ! ! -
8 8
l 2
Constant
power ſº
8Ource

l
-

Battery D

Ri R2

© - - -

Figure 5. - Control-system equivalent circuit.

The battery consists of a pair of 45-volt nickel–cadmium wet cell batteries con
nected in series to establish a 90-volt bus. Figure 4 is a typical curve of the relation
between the nickel–cadmium cell voltage and state of charge. As charge is increased
there is an extensive plateau at about 1.4 volts. At a charge of approximately 85 percent
a "knee" is observed where the voltage rises to another plateau of about 1.6 volts. The
I

2H
|-
|
on |- |
f
O
:-
|-
_A |
J. F- |
º |
#
O
1F -
|
|
:-

T |-
q)
o |
|
I
L l l —l-
O 25 50 75 LOO

State of charge, percent

Figure H. - Charge characteristics of nickel–cadmium battery.

100-percent charge is assumed to be on the second plateau. However, the higher voltage
region results in inefficient charging because of increased heat generation and excessive
outgassing of the cells. In order to eliminate this type of problem, the maximum cell
voltage was limited to 1.47 volts, which is about halfway up the knee. In this range of
charge the battery is reasonably efficient (approximately 85 percent), and no problem
exists with outgassing or heat.

The zener diode labeled D in figure 3 is a symbolic representation of the parasitic

load. The conduction voltage was adjustable and was set at 90 volts. Figure 5 shows the
relation between current and voltage of this parasitic load.

lO t

ºn
q)
H

3.
#
., 5 H.
+2
ſ:
Q1)

l l l _l L 1.
O 20 l;0 60 8O lOO l2O

Woltage, volts

Figure 5. - Relation between current and voltage of parasitic load.

The resistors R1 and R2 in figure 3 are representative of the real or useful loads
which are switched in and out. As the system is started, the power source is turned on
and the energy is absorbed by the battery. After the battery becomes charged, the bus
voltage rises to 90 volts. When this point is reached, the parasitic load D begins to con
duct, preventing the voltage from rising further. The battery current then decreases and
finally reaches some trickle-charge value which is quite low — less than 0.5 ampere in
this system. The system is now in a state where the parasitic load is absorbing nearly
all the power generated by the source. If S1 is closed, current will flow through R1, which
represents a load of 50 percent. In order to equalize the system, the parasitic load will
decrease its conduction current by an amount equal to the current drawn by R1. The bus
voltage has not yet varied more than 0.5 volt. The battery is still fully charged and con
suming a trickle charge. The system load is being satisfied by the source, and the extra
energy is being absorbed by the parasitic load. If S2 is also closed, current will flow
6
a through R2. If R2 has been sized to draw approximately the full load (100 percent of
* power-source energy), then the combination of R1 and R2 presents an overload. In order
2 to satisfy the current through R1 and R2, the parasitic load will turn fully off. This, how
# ever, is not enough, and the bus voltage will start to drop. Because of its natural charac
* teristics, the battery will supply current to the external circuit when the terminal voltage
is reduced. In this condition the system is using about 150 percent of full load capacity;
º the power source is providing 100 percent and the battery the remaining energy. This
condition can continue until the extra load is dropped or the battery discharges to its set
limit. If S2 is opened after a period of time, the load demanded by the "real load" returns
again to 50 percent and the excess energy from the power source will be absorbed by the
battery. The battery will accept this excess until its terminal voltage again returns to
90 volts. When this happens the battery is nearly fully charged, and the parasitic load
will begin to draw current to prevent the voltage from rising further. The battery cur
rent will then slowly reduce to a trickle charge and the system is fully recovered.

The important thing to note is that at no time will the source be called upon to pro
vide more than its steady-state energy. Thus, the source has not experienced any power
transients. The only effect that has been transmitted back to the source is a small and
relatively slow variation in terminal voltage. If the reduced terminal voltage results in
speed variations, the speed sensor will control the field voltage through the field regulator
and maintain correct speed and a fixed output power. Since the source terminal voltage
has varied only slightly, small corrections in field voltages are all that are required to
maintain speed, and no large or rapid transients are noted in alternator speed. It is this
feature that is so important when system stability is considered.

In the design of this demonstrator system several areas were purposely omitted.
The protective, monitoring, and priority logic circuitry necessary for mission control
and reliability were not studied in depth, nor was specific conditioning equipment. Such
problems as exist in these areas are common to all control philosophies. Protective cir
cuits, however, become simpler because of the large fault-clearing capacity of the battery,
which is unhampered by the current limitations of power-conditioning circuitry.

RESULTS AND DISCUSSION

The system described was built and tested with a simulated Brayton cycle turbo
º
alternator. The simulator consisted of a 24-volt automotive-type alternator driven with
º a separately excited do motor. The motor supplied a nearly constant torque to the alter
º nator. This is a close approximation to the combined rotating unit that is typical of a
| Brayton cycle machine operated with a very small speed deviation about its maximum
* power point.
;
ſ
 Figure 6 shows a portion of a recording of one test run. The operating power level
from the alternator in this run was about 450 watts with a peak real load requirement of
970 watts, more than a 100-percent overload. There were no transients or perturbations
in the alternator output current; there was some variation with bus voltage and "noise"
which was traced to brush arcing in the driver motor. Load switching was not visible in
the output of the alternator. In this simulation the maximum bus voltage was 90 volts.
One minute after a 115-percent overload the bus voltage had dropped to 78 volts, or nearly
15 percent below maximum. The parasitic load is observed to be operating in the manner
predicted; it draws current only after the battery has been charged and the bus voltage
has risen to 90 volts. The parasitic load then draws only enough current to hold the volt
age at this level, and the battery current remains at a low trickle-charge level. It should
be noted that these tests were conducted with wet cell nickel–cadmium batteries. For

spacecraft, sealed cell batteries would be used and the voltage variations on the bus would
be greatly reduced.

Variations of alternator speed could not be detected with the installed sensor and
so were not included in these recordings. Thus, a very simple speed sensor is adequate
for this type of system. -

The battery current can be seen to absorb any load-switching transients when the
total load demand is in excess of alternator output power or during the battery charge
mode.

The tests represent an unusually severe case of power switching. In any actual
spacecraft, the relative load changes would be much smaller. Power changes of less
than 1000 watts may be quite frequent, but in a 100-kW system they amount to only 1 per
cent of the connected load and would not be visible on these recordings.

With the system described here, up to two-thirds of the power conditioners used in
the system of reference 4 can be eliminated. Figure 7 shows block diagrams of the two
systems. The biggest single difference in system complexity is the elimination of the
battery charge and discharge converters. This, combined with the simpler control logic,
yields a system which is lighter, simpler, more efficient, and inherently more stable.

CONCLUDING REMARKS

A system was developed that fully controlled the rotational speed of a Brayton cycle
turboalternator while providing useful electrical energy. The system is simpler and
inherently more efficient than those resulting from earlier efforts. A single parasitic
load and battery provide a controlled bus for distribution. From this principal bus, con
verters can be connected in each module and the resulting interchangeability of equipment
between modules minimizes spares and enhances reliability. The advantage of having

8
 Real-load
Parasitic-
voltage,
Bus
Alternator current,
Battery load current
current,
volts y
amperes
current,
amperes
amperes
amperes
&
39
TS3
o, & + -

ON
RSOc
O
\x. O N
CA)
IIi | II

§ H -
H.
|-

o
S2
ty
on
g |M-º- *
|–|
3
|- d|:º
|6
E
Oº
d

: 5,

§. H. |-

. H–

§.
 Alternator
Field Frequency
regulator ~* sensor IT T.
A *
High voltage do | Load
controller
U-J I

Hº-P Parasitic load

- - - - - - - -
— ———
C -> Storage |
|
|
C ———— — ———
|-- Real |
C loads |

- - - - - - - - - - - - - l

Speed-controlled system [c] Power converter

•- Power flow

•--- Sensing & control

r—— — — — — — — — — — — — — — - -]
Yº Alternator !

Field - * Frequency L_
sensor

High voltage de
l—ſ

H Parasitic load

–- Storage

Real
° Hºliosis
Woltage-controlled system

Figure 7. – Comparison of Brayton control concepts.

10
 several small converters delivering practical currents rather than one huge converter is
obvious. A de-to-do converter in the 1000-ampere range is not off-the-shelf equipment.
The variation allowed in the high-voltage de bus need not be an objectionable char-,
acteristic. Compared to the system in reference 4, this concept provides the same con
trol in a more exacting manner with far less equipment. It offers better system response
with increased efficiency and reliability, and does so without perturbations of alternator
Speed and power.

Langley Research Center,

National Aeronautics and Space Administration,
Hampton, Va., December 3, 1971.

REFERENCES

1. Engineering Staff: Test Report — AiResearch Brayton Cycle Demonstrator Test Pro
ºf
gram. APS-5270-R (Contract NAS 9–6116), AiResearch Manufacturing Co. of
Arizona, Garrett Corp., Nov. 15, 1967. (Available as NASA CR-65924.)

|: 2. Word, John L.; Fischer, Raymond L. E.; and Ingle, Bill D.: Static Parasitic Speed
Controller for Brayton Cycle Turboalternator. NASA TN D-4176, 1967.
3. Anon.: Report on the Design Requirements for Reactor Power Systems for Manned
Earth Orbital Applications. DAC-57950 (Contract NAS 1-5547), Missile & Space
Systems Div., Douglas Aircraft Co., Inc., Jan. 1967. (Available as NASA CR-66254.)
4. Kelsey, Eugene L.; and Young, Richard N.: Power and Load Priority Control Concept
for a Brayton Cycle Power System. NASA TN D-6478, 1971.

ASA-Langley, 1971 — 3 L-7963 11

 WASHINGTON, D.C. 20546
Post AGE AND FEEs PAi D
NATIONAL AERONA UT I Cs and
OFFIC I AL E U S I N ESS
space ADMIN is TRATION
PENALTY FOR PRI v ATE Use $300
FIRST CLASS MAIL

022 002 cl U 35 71 lz l 7 Snoq 39H

IJN T V D F I LL INO I S.
L I BRARY
AT T N : Dſh CUMF NT LIBRARIAN
UR BANA T L 6 1301

POSTMASTER:. If Undeliverable (Section

N. *ºs15°

! Postal Manual) Do

"The aeronautical and ſpace activities of the United Stateſ ſhall be

conducted ſo as to contribute . .. to the expanſion of human knowl
edge of phenomena in the atmoſphere and ſpace. The Adminiſtration
ſhall provide for the wideſt practicable and appropriate diffemination
of information concerning itſ activitieſ and the reſult, thereof.”
— NATIONAL AERONAUTICS AND SPACE ACT OF 1958

NASA SCIENTIFIC AND TECHNICAL PUBLICATIONS

TECHNICAL REPORTS: Scientific and TECHNICAL TRANSLATIONS: Information

technical information considered important, published in a foreign language considered
complete, and a lasting contribution to existing to merit NASA distribution in English.
knowledge.
SPECIAL PUBLICATIONS: Information
TECHNICAL NOTES: Information less broad derived from or of value to NASA activities.
in scope but nevertheless of importance as a Publications include conference proceedings,
contribution to existing knowledge. monographs, data compilations, handbooks,
sourcebooks, and special bibliographies.
TECHNICAL MEMORANDUMS:
Information receiving limited distribution TECHNOLOGY UTILIZATION
because of preliminary data, security classifica PUBLICATIONS: Information on technology
tion, or other reasons. used by NASA that may be of particular
interest in commercial and other non-aerospace
CONTRACTOR REPORTS: Scientific and
applications. Publications include Tech Briefs,
technical information generated under a NASA Technology Utilization Reports and
contract or grant and considered an important
Technology Surveys.
contribution to existing knowledge.

Details on the availability of these publications may be obtained from:

SCIENTIFIC AND TECHNICAL INFORMATION OFFICE

NATIONAL AERO NAUTICS AND SPACE ADMINISTRATION

Washington, D.C. 20546
 629.13
Ungºstn
42–77

NASA. T E C. H. Nic A. L. No TE \ NASA TN D-6579

THE LIBRARY OF THE

MAR 22 1972
UNIVERSITY of |LLINOIS
AT URBANA-CHAMPAT&N

DETERMINATION OF STABILITY
DERIVATIVES FROM FLIGHT DATA
USING A NEWTON-RAPHSON
MINIMIZATION TECHNIQUE

by Kenneth W. Iliff and Lawrence W. Taylor, Jr.

Flight Research Center
Edwards, Calif. 93523

NATIONAL AERONAUTICS AND SPACE ADMINISTRATION - WASHINGTON, D. C. - MARCH 1972

*** •• • •
（~~~~ wae，
• •• • • •• •
 1. Report No. 2. Government Accession No. 3. Recipient's Catalog No.

NASA TN D-6579
4. Title and Subtitle 5. Report Date
DETERMINATION OF STABILITY DERIVATIVES FROM FLIGHT DATA March 1972
USING A NEWTON-RAPHSON MINIMIZATION TECHNIQUE 6. Performing Organization Code

7. Author(s) 8. Performing Organization Report No.

H–626
Kenneth W. Iliff and Lawrence W. Taylor, Jr.
10. Work Unit No.
9. Performing Organization Name and Address 126–62-01-02-24

11. Contract or Grant No.

NASA Flight Research Center
P. O. Box 273
Edwards, California 93523 -

13. Type of Report and Period Covered

12. Sponsoring Agency Name and Address Technical Note

14. Sponsoring Agency Code

National Aeronautics and Space Administration
Washington, D. C. 20546
15. Supplementary Notes

16. Abstract

A modified Newton–Raphson or quasilinearization minimization technique for

determining stability derivatives from flight data was developed and compared
with simple—equations, analog-matching, least-squares, and Shinbrot methods
of analysis. For the data analyzed, the solutions computed by using the estimates
obtained from the Newton–Raphson technique fit the data and determined coeffi
cients adequately. A further modification to include a priori information was
found to be useful.

A model statistically similar to the flight data was analyzed using the same
methods (excluding analog matching), and the Newton–Raphson technique was
found to yield superior estimates. An approximate Cramér–Rao bound was
compared with the error covariance matrix of the model and was found to provide
information about the reliability of the individual estimates obtained.

The technique was successfully applied to data obtained from a light airplane,
a large supersonic airplane, and a lifting body vehicle. It was shown that the
reliability of the estimates of a given coefficient obtained from these vehicles
depends upon the data analyzed.

17. Key Words (Suggested by Author(s)) 18. Distribution Statement

Identification
Unclassified – Unlimited
Nonlinear minimization

Stability derivative determination

19. Security Classif. (of this report) 20. Security Classif. (of this page) 21. No. of Pages 22. Price"

Unclassified Unclassified 57 $3.00

-

For sale by the National Technical Information Service, Springfield, Virginia 22151
**--
•• • • • •• • •
• • ►•• • ••• • ••
 DETERMINATION OF STABILITY DERIVATIVES FROM FLIGHT DATA

USING A NEWTON-RAPHSON MINIMIZATION TECHNIQUE

Kenneth W. Iliff and Lawrence W. Taylor, Jr.

Flight Research Center

INTRODUCTION

Identifying a system which will accurately and reliably determine the stability and
control derivatives (i.e., coefficients of the differential equations of motion) of air
planes from flight data remains difficult and time consuming, despite the considerable
effort that has been devoted to it over the years. There is no lack of methods to deter
mine stability derivatives from flight data; on the contrary, many methods have been
tried, and under idealized conditions most have been successful. Unfortunately,
however, inadequate aircraft excitation, instrumentation and measurement noise, and
departures of the actual airplane from the model used for the airplane dynamics.
generally cause considerable error in the resulting estimates.

Among the methods available to solve this problem, the simplified—equations,

time-vector (ref. 1), analog-matching (refs. 1 and 2), and regression methods
(refs. 3 and 4) have been used most extensively. The simplified—equations and time
vector methods are often severely limited because the set of aerodynamic coefficients
Dbtainable is incomplete and the types of applicable responses are restricted. The
analog-matching method is also limited because estimates resulting from it vary with
:he skill and technique of the operator. The accuracy of the coefficients estimated by
..he regression methods is unreliable if measurement noise on the response is exces
sive or maneuvers are poorly conditioned.

Of most interest is the possibility of obtaining a linear model which results in a

3omputed response that best fits the measured data. Analog matching imitates this
approach, but the estimates reflect the operator's judgment. Regression methods
minimize the integral-squared error in fitting the measured data to a linear model of
he system point by point in an algebraic sense, ignoring the time-sequential nature
»f the aircraft and model responses. Many of these difficulties can be circumvented
intuitively by minimizing the weighted integral-squared error of the difference between
he computed and measured response. The standard gradient technique (ref. 5) can be
ised for this minimization, but this technique has been found to converge much too
slowly. Consequently, a modified form of the Newton–Raphson technique (ref. 5) was
nvestigated as a means of obtaining more rapid convergence. This method could also
je viewed as an application of quasi-linearization (ref. 6).

This report compares the results obtained by using the modified Newton–Raphson
echnique of minimization with the results obtained from simplified—equations,
inalog-matching, and regression methods for a representative lateral-directional
1ight maneuver. The proposed method is applied to obtain stability and control
 derivatives for vehicles encompassing the geometric and flight extremes of modern
aircraft. An example of the application of the method to the longitudinal equations of
motion of an airplane is also presented.

SYMBOLS

Physical quantities in this report are given in the International System of Units (SI)
and parenthetically in U. S. Customary Units. The measurements were taken in
U. S. Customary Units. Factors relating the two systems are presented in reference 7.

stability matrix (P × P)

null matrix except for the i-jth element which equals 1

(P × P)

ith row of the stability matrix (1 × P)

i-jth element of A

linear acceleration along the X-, Y-, and Z-axes,

respectively, g units
:
control matrix (P × Q)

null matrix except for the i-jth element which equals 1

(P × Q)

ith row of the control matrix (1× Q)

i-jth element of B

augmented A and B matrices (P. X (P + Q))

ith row of the C matrix (1× (P + Q))

vector of unknown coefficients (m X 1)

ith element of the c vector

vector of a priori estimates of the unknown

coefficients (m × 1)

vector of actual values of unknown coefficients (m X 1)

weighting matrix for observation vector (R XR)

 weighting matrix for a priori estimate vector (m × m)

i-jth elements of D1 and D2, respectively

expected value

matrix functional of 2 for Shinbrot method (1 × (R-P))

matrix functional of zs for Shinbrot method (l × P)

method function vector (l × 1)

jth element of method function vector

partition of matrix relating the state vector to the

observation vector ((R-P) × P)

acceleration due to gravity, m/sec.2 (ft/sec.2)

vector of observation biases (R × 1)

ith element of g
partition of matrix relating the control vector to the
observation vector ((R-P) × Q)

f vector functional of ? for modified least-squares

method ((R-P) × 1)

vector functional of zs for modified least-squares

method (P × 1)

vector functional of u for modified least-squares method

(Q X 1)

identity matrix

cost functional or weighted mean-square-fit error

scalar weighting factor (gain) for a priori weighting

matrix

rolling moment divided by the moment of inertia about

the X-axis, rad/sec.2

number of time samples

 pitching moment divided by the moment of inertia about
the Y-axis, rad/sec2
number of unknowns in c vector

yawing moment divided by the moment of inertia about

the Z-axis, rad/sec.2

vector of measurement noise (R × 1)

ith element of n

P number of state variables

roll rate, rad/sec or deg/sec

number of control variables

pitch rate, rad/sec or deg/sec

number of observation variables

yaw rate, rad/sec or deg/sec

total time, sec

intermediate or incremental time, sec

matrix trace operation (sum of diagonal elements)

control vector (Q X 1)

velocity, m/sec (ft/sec)

covariance matrix of the noise

longitudinal force divided by mass, m/sec.2 (ft/sec.2)

state vector (P × 1)

ith and jth component of x

side force divided by mass and velocity, rad/sec

observation vector (R X 1)
ith element of the y vector

normal force divided by mass and velocity, rad/sec

ºf ZS measurement of state variables (PX 1)

? measurement related to derivatives of state variables

((R-P) × 1)

measurement of observation vector (R X 1)

angle of attack of X-axis, rad or deg

angle of attack of principal X-axis, rad

angle of sideslip, rad or deg

weighting matrix for hs used in modified least-squares

method (P × P)

weighting matrix for hu used in modified least-squares

method (Q X Q)

increment

first variation

aileron deflection, rad or deg

elevator deflection, rad or deg

rudder deflection, rad or deg

constant control deflection, rad or deg

relaxation coefficient

damping ratio

pitch angle, rad or deg

probability density function of (...)

approximate estimate for the variance determined with

the Cramér–Rao inequality of the i-jth term
estimated variance of wind-tunnel estimates for the
i-jth element of D2
estimate for the variance determined with the
Cramér–Rao inequality
variance of the estimates from an experimental model

5
T auxiliary time variable, sec

Cramér–Rao bound for the error covariance matrix

bank angle, rad or deg
frequency, rad/sec

gradient of (...) with respect to c

Vg(-) gradient of ( ) with respect to g

Subscripts:

d Dutch roll

e extended

ith row or component

jth column or component

iteration index

me measured

0 a nominal or constant value

Öa, Ör,0e, 30, partial derivatives with respect to subscripted

variables
P, Q, T, V, C2,
B, Q, 6

Superscripts:

i,j index representing time of sample

T matrix transpose

A dot over a quantity denotes the time derivative of that quantity. Principal axes are
used throughout.
Boldface type indicates a vector.

STABILITY-DERIVATIVE DETERMINATION

The task of determining all the stability and control derivatives can be greatly
simplified if it is realized that some of the dynamic modes may be uncoupled for most
aircraft configurations. The longitudinal modes are usually separated from the
lateral-directional modes because the resulting error is small. The bulk of this report
6
deals with the lateral-directional modes, because this estimation problem is more
complex than that for the longitudinal modes. After the proposed technique is fully
developed and evaluated for the lateral-directional modes, an example is given to show
that the method is equally successful for a longitudinal mode.

The model most often used to describe lateral-directional airplane dynamics can
be expressed as a system of linear, constant-coefficient, differential equations in the
following form (refs. 8 and 9):

x(t) = Ax(t) + Bu(t) (1)

where

p(t) p(t)
Öa(t) -

r(t) - r(t)
u(t) = | 5r(t) x(t) = x(t) = | .
ôo(t) 8(t) B(t)

Q(t) Ö(t)

Pp Pr Iº O *6, *6, *6,

N N N O
,- || “p ºr ºf B = | Nöa Nör Nöo
Cy -1 Y Y
0 6 *g, Yo, Yö, Yö,
1 0 0 0
0 0 0

It should be noted that no state noise is considered in this analysis. The last column
in the B matrix represents the effect of an uncertain bias on p, r, and 3.
Next zs(t) and 20t) are defined as the noise-contaminated measurements of x(t)
and x(t), respectively; that is, zs(t) is the noise-contaminated measurement of the
A -

State vector x(t), z (t) is the noise-contaminated measurement of x(t), and the con
trol inputs u(t) are considered to be noise-free. For the remainder of this report,
the time argument of the variables is omitted where the time dependence is clear.
- - A - -

The problem addressed is: Given z's, 2, and u, determine certain unknown
elements of the A and B matrices. In airplane dynamics these coefficients of the
linearized equations of motion are the stability and control derivatives which result
from the Taylor's series expansion of the aerodynamic forces and moments.
 REVIEW OF PREVIOUS METHODS

10

Öa, deg 0 |- To examine previous methods of determin

ing stability and control derivatives, one typ
-10 |→ ical flight maneuver was selected and several
methods were applied to analyze the data
10ſ– obtained from the maneuver. This maneuver
Ör, deg 0 was chosen because all the measurement
r, signals appearing in zs, 2, and u were
-10 1–1–1–1– available, which is needed in some methods
50 T- to estimate all the coefficients. Except for
this stipulation, the maneuver was arbi
Pme. trarily selected and was not known to favor
deg|Sec 50 | | | | any of the methods.

10 I- The maneuver selected was for the X-15

research airplane (ref. 10) at a hypersonic
ſº. OH ==== flight condition (figs. 1 to 13). Time his
deg|sec |→l—l-l- tories of the maneuver are shown in figure 1.
-10 The control-input time histories apply to all
10 T- the data in figures 2 to 8, 10, and 12 and
therefore are not repeated. The 3 me data,
Éme, deg 0 _T- which are used in the subsequent analysis,

-10 | | | | | | were not measured directly in flight, but were

computed from other flight variables by using
40 r- the following equation:

º:
Wine, deq 0 –-
| N__-- Éme = *pme - "me * (ayme * sin ºme)wg
- -

60ſ- The least-squares and Shinbrot methods

Pme. 0A ZT (discussed later) use 3 because the formu
deg'secº
-60
\U/|--|--
N_2^ - - ºme ºr
lation is more convenient with Éme than

with a and the resulting estimated pa

10 F- y me
‘me ... O _*-*. ~T- rameters are identical.
deºx2 N-D- - -

-10 H-I-I-I- Stability derivatives are usually deter

mined first from wind-tunnel data. However,
. 1 [- because of unavoidable differences between
Éme
-
0 /N ATN the wind-tunnel
wind-tunnel testshould
values and flight conditions,only
be considered the

as initial estimates. Figure 2 compares time

6 histories obtained from flight with those com
t, Sec puted by using wind-tunnel values. The dif
ferences shown offer motivation for deter
Figure 1. Time history of X-15 maneuver. mining coefficients from flight data.
 Simplified-Equations Method

For selected types of responses the effect

of only a few coefficients dominates, thus
— Flight
permitting the use of simple expressions to
determine these coefficients. Some of the
— — — Computed
expressions (ref. 1) that are used for this
purpose and the types of responses that are
required are as follows:
Af
No, ~ A5. (rudder pulse)

A Ap . .
Lô, is #. - "pää, (aileron pulse)

- Ar Ar A

3. deg 0
Nóa *- Å5. - Nrā; - Nºż. - Nº. 5 a
-

(aileron pulse)

Yº > # Aay
Aºſ (Dutch roll oscillation)

2
Nº s wa + o Lº (Dutch roll oscillation)

dör dôa
Lp s-Lôr Tää -Lôa T (steady sideslip)

Aöa
Lp ~ +6, TAF (aileron step)

Nr s -2' dººd - Yº (Dutch roll oscillation)

The notation has been changed from that of

reference 1 to be consistent with the principal
axes system used in the preceding equations
of state. The time-vector method, also dis–
cussed in reference 1, serves as a basis for
the equations in which oscillatory responses
are used.

A partial set of coefficients was obtained

Figure 2. Comparison of X-15 time histories by using equations, such as the preceding,
measured in flight and computed by using from the example set of X-15 flight data.
predictions based upon wind-tunnel tests. These values (predictions based upon wind
tunnel values were used to complete the set)
were substituted into the equations of motion,
and time histories were computed. These
 time histories are compared with those
measured in flight in figure 3, which shows
differences at least as great as for the time
histories computed by using the wind-tunnel
coefficients.
— Flight
— — — Computed

Analog-Matching Method

The equations of motion can be readily

solved by using an analog computer. This
would suggest fitting the measured responses
****
with the computed responses by manually
adjusting the potentiometers of the analog
computer mechanization, in which the settings
correspond to particular values of the coeffi
cients. The operation, referred to as analog
matching, is described in detail in refer
3. deg 0 –Rs –- ences 1 and 2. Analog matching was applied tº
the X-15 flight data, with the results shown in
-10–1–1–1–1–1– figure 4. Because analog matching is a man
40 T--
ual operation, the operator can somewhat
constrain values of some of the coefficients
% deg 0 H,- on the basis of his knowledge of the quality of
-40 | NS |
the wind-tunnel values and his experience,
* ~~ - - that is, a priori information.
60 r- *-

By comparing the results shown in figures

deg'sec2 0 3 and 4, it is apparent that the fit has been
improved considerably. Nevertheless, it
-60 would be desirable to automate the fitting proc
10 T
ess, perhaps improving it further as well as
making it more efficient. Regardless of the
f. method used, the same match can be obtained
deg'secº 0 --

with analog matching, because it solves the

same equations, but the effort involved may
be prohibitive.

Least—Squares Method

One approach to determining aerodynamic

derivatives is to minimize the integral square
of the state equation error by substituting the
measured values for the state and its deriva
Figure 3. Comparison of X-15 time histories tives, that is, minimize the following cost
measured in flight and computed by using functional:
coefficients obtained from flight data by the
method of simple equations. T

ºſ (? - Azs - Bu)" (? - Azs - Bu) dt (2.

0

10
 To derive an expression for the values of A
and B that minimizes equation (2), let

C = | * |
|

|
— Flight
50 – — — — Computed then

* 0
ſº. 4'-[º]. [...] c'eſ;).
-50 | | | | |
10 I
-ſ tº a ſºleſ?]- ſºjºſº.
r,
deg/sec 0 -> E

-10
- “[ſ";" |- alſº jº'er “ſ"Fºlſº j'eee!
10 r
where
fl, deg 0 ---> -->
-10–1–1–1–1–1– tr(AB) = tr(BA)

% deg 0 ||-- The minimization is achieved by taking the

first variation of the preceding expression,
*LIN_r= that is,

* --ºſ'ſ #jº's cºſ" [+][*] caseſ

and putting

ÖJ = 0

1 I
The resulting solution requires that, for an
arbitrary 5C,

-1–1–1–1–1–1–
0 1 2 3 4 5 6
t, sec

Then
Figure 4. Analog match of X-15 time histories
measured in flight.
cº-ſ"|*||*||al'ſ' +|: '... (3)

11
 This equation is the solution to the least-squares problem. It has the advantage of a
compact form, but it disguises the independence of each of the equations to be minimized.
This limitation can be illustrated in the following manner. Considering only the first
state equation and minimizing the same cost functional,
T
J =ſ (Pine - Hºpme - Prºme - Hºme - 16,9a - Lö, ºr - 16%)
- 2
dt
0

where the subscript me identifies the measured value. A more convenient form may
be obtained by making the following substitutions:

A1 = [ºp Lr Ig º
! B1 = [º, Lör lº
I

Then
T || 7
-ſ
-
T
Pme
- -
91 | |
–5–
Ul
2
dt

Once again taking the first variation and setting the resulting expression equal to zero

" *{ſ}|ºſ. [*].

Now C1 is the first row of the C matrix, and *me is the first element of 2, which
makes it apparent that the elements of the first row of the C matrix are independent of
all the elements of ? except the first, pne. A similar relationship is easily shown
for the other rows of the C matrix. This independence is one of the drawbacks of the
least-squares method, in that only one of the measured state derivatives is used in
determining a given row of the C matrix. If one of the signals has not been measured,
the least-squares method does not provide an estimate of the derivatives related to
that signal. This independence also illustrates that the estimate of one row of the
C matrix is obtained independently of the other rows, and no "trade-off" can be made
between elements in different rows to improve the match.

Figure 5 shows the fit of p, r, and § obtained with the least-squares method.
The fit is considered to be good. Since only p, r, and 3 were fit, these are the only
time histories shown. It should be noted that the fit of the time histories is not a
solution of the differential equations of motion. It is merely a minimization of the
integral-squared error of the fit of the measured data to a linear model of the system
(that is, point by point in an algebraic sense, ignoring the time-sequential nature of
the data). In other words, only measured state variables and their derivatives are
12
used in the minimization because the differential equations are never solved for the
computed state.

— Flight
1.0 ––– Computed least-squares fit
5 -
P,
radisecº
o
-.5 |- N_A"2^T\s Y-_^
__
-1.0 | | | | |
.08 T
- .04 H - -

r, 0 2.-->s- _T==~
rad/secº-.0% v/ NG”
-.08 | | | | | |

.08 r
: .04 H
fl, 0 2^ 2–~s
rad/sec -.04 H 2< SS -
-.08

t, Sec

Figure 5. Least-squares fit of X-15 flight time histories.

The least-squares values of the coefficients were substituted into the A and B
matrices, and a time history was computed and compared, in figure 6, with that meas
ured in flight. The fit of p, r, and 3 is still good, but other variables have drifted
apart, especially q), p, and r. These are typical results obtained by using the least
Squares method. The poor fit of the state variables should not be surprising, inasmuch
as the computed solution is not included in the minimization.

Shinbrot Method

In reference 4, Shinbrot showed the existence of similarities in a variety of regres

Sion methods and introduced a general method function, f(t). This can be shown by
starting with the state equation with the measured values zs and 2 denoting the state
and its derivative, then multiplying by a method function and integrating. Thus

T, T | zs (t)
2 (t)f(t)T dt = C / |--E-- f(t)T dt
0 0 u (t)

The method functions used are those suggested by Shinbrot in reference 4 and have the
form

2Tjt
, Leº j = 1, 2, . . ., l

13
 where the maximum of j must be greater
than or equal to the number of unknown ele
ments in C. Then define

^T
F * =
T.z f T dt
— Flight
50 T- — — — Computed 0
p,
deglsec 0 zz Y S --> and
-50– 1–1–1–

lO [-
r,
deg|Sec 0 --> *er

-10–1–1–1–1–1– It follows that

10 I
A

F = Fs c'
-10 ––––––
For this transformation, the solution may be
40 r obtained in the form

Ø deg 0 ~L T T – T \
–40 | | * = == -- ~ - | Fs' Fs C" = Fs' F
60 r- ~

(4)
a...? PA €C
-60 \L/
^eº
>
^. *

CT = | Fs' º -1
Fs" f
A

10 I
It can be readily shown that this method has
r,
deg|Secº 0 V Se
- -
the same row independence of C illustrated
for the least-squares method. This method
-10 –1–1–1–1–1– is similar to the least-squares method and
- .. l I yields nearly identical results, as shown in
figure 7, in that the computed time histories
B,
/
AN
* w
^e ~.”
also drift away from the measured values.
*
0
|\,A
1 2 3
|
4 5
|
6
|
Modified Least—Squares Method
t, Sec
Some of the difficulties encountered with
the least-squares method can be overcome by
Figure 6. Comparison of X-15 time histories
measured in flight and computed by using
modifying it to include the state vectors in the
coefficients obtained from flight data with
error minimization, thus combining the
the least-squares method. standard least-squares method with the inte
grated least-squares method. The integrated
least-squares method is merely the least
squares formulation with the integral of each
of the variables replacing the variables.

14
 By defining

h s(t) = ºſ 0
z's (1) dr + zs (t)
— Flight
— — — Computed
50 r
D, t

deg/sec 0 zz *~--> hu (t) = ºſ 0

u (1) dr + u (t)
-50 - | | | |
10 F
and
r,
deg/sec 0H------ t

-10–1–1–1–1–1– horſ 0
2 (1) d 1 + 2 (t)
10 r

where rs is a (P × P) diagonal matrix and

-10–1–1–1–1–1– ſu is a (Q X Q) diagonal matrix which pro
vide weighting among the various signals
40 T
representing the relative confidence in the
q, deq 0 H signals, then

* N-F-
deg/sec
60 T-

%.2 oz. S_2

^-

,”
ſº-cſ:}{i-cº!.
...[ \Z Nº or the complete formulation for the minimum
becomes, as in least squares,
- 10 ſ—
r, -

dejsecº "N7 *S-

-10–1–1–1–1–1–
.1 ſ−

§ 0|^. 7.

radisec
-. 1 |Nº|| || >=
cº-ſ’ |||}|..} ſºjº',
Figure 7. Comparison of X-15 time histories Once again it is easily shown that this method
measured in flight and computed by using has the same row independence of C shown
coefficients obtained from flight data with the
Shinbrot method. by both the least-squares and the Shinbrot
method.

15
 The fit of the computed time history to
that obtained in flight in figure 8 is still not
acceptable. It would, of course, be desirable
to minimize the differences in the time his—
tories. Although this can be done, it will be
— Flight shown that the identification problem becomes
50 — — — Computed nonlinear, requiring some iterative technique.
p,
|-
deg/sec 0
|\}^i TT Modified Newton–Raphson Technique
–50

10
Once the mathematical model for the sys
tem has been selected, two items remain to
ſ, be specified to identify the system: (1) the
~=-
deg/sec 0 cost functional to be minimized, and (2) an
-10
H–1–1–1– algorithm to minimize the cost functional. In
some of the methods analyzed these items are
10
not readily specified. In analog matching,
the cost functional is obscure in that what
B. deg 0
constitutes a good fit is primarily a function
-10 of the judgment of the operator. The operator
minimizes his subjective cost functional by
40
adjusting potentiometers until, in his opinion,
the fit is attained. This procedure can be
Q deg 0
time consuming, and the results vary greatly
-40 from operator to operator. The regression
methods discussed alleviate some of the prob
60 lems of analog matching in that a cost
5, functional is specified and is readily mini
deg|secº 0 mized because the formulation is linear in the
-60 parameters to be estimated. Unfortunately,
the resulting fit to the measured data is
10 unsatisfactory. Ultimately, a linear model
r, is desired with a computed response that best
fits the measured data. It has been demon
deg|secº
-10 strated that the fit obtained with analog match
ing is superior to that obtained with the other
.l
methods. This suggests that the cost
3, functional should reflect the difference
rad/sec 0 between the computed response based upon
-. 1 the coefficient estimates and the measured
response. Therefore, more satisfactory
results would be expected if the problem were
posed as one of minimizing a quadratic cost
functional of this difference. Many other
criteria could be proposed, but the quadratic
Figure 8. Comparison of X-15 time histories measured is the most thoroughly analyzed form and
in flight and computed by using coefficients obtained has the most desirable mathematical char
from flight data with the modified least-squares method. acteristics.

To be more specific, consider the

16
following model:
x = Ax + Bu

where A and B are defined as in equation (1) and

L5 L L
Ip Ir Lg 0 a Ör 90
G = N N N H = N N
p r g 0 T | *ča 8, No 0
0 0 Y 0

r 0

The non-zero elements of G and H are also elements of A and B, respectively.

The vector y is the set of "output" response quantities:
- - T

y = [p r a o f : a,
Let z denote the measurement of the actual aircraft response quantities corresponding
to y. Although no state noise is assumed, z would not be exactly the same as
because of measurement errors and differences between the actual and the assumed
linear model (eq. (6)). A reasonable description of the relationship of z and y is

z(t) = y(t) + n (t) + g

where n (t) represents errors with zero bias, referred to as "noise" and g represents
all bias errors. In summary, x is the model response, y is the model observations,
and z is the actual measurement.

The objective is to minimize the difference between z (t) and y(t) in some sense,
so that an appropriate cost functional is
T T

J -ſ 0
|-0 sº
- D1 |:0 so- dt

where D1 is a weighting matrix reflecting the relative confidence in the measurements

(similar to the inverse of the covariance of noise). Now that the cost functional has
been completely specified, the only remaining step in the systems identification problem
is to choose an algorithm for minimizing this cost functional.

17
 Gradient Method

Probably the simplest minimization technique is the gradient technique, which

changes the coefficients to be estimated by proceeding in the direction of the greatest
decrease of the cost functional. Notationally, it is convenient to define a column
vector, c, of the unknowns to be estimated. The elements of c are some or all of
the unknown elements of A and B (hence, G and H) of the initial conditions x (0)
and of the noise biases g, that is,
c = c [aij, bij, gi, xi (0)]

* *-*.
The gradient of J with respect to the vector c can be expressed in terms of the
gradient of (z – y) as
T T

vo J -*{ſ (z - y)' D1 ve •-9) dt (7)

Thus, all that is needed to specify ve J in addition to terms already defined is

we (z = y), as follows:

-}.} #}} { | | }} |##. H jº.

I w I
We (z - y) = -
G -
c
x -
Vc (z - y)
H G

(8)

The gradient of z is affected only by the bias terms. The quantity Vc (z – y) can be
expressed in terms of various partial derivatives. These partial derivatives with re
spect to the individual coefficients of c (aij, bij, gi, xi (0)) are
| 0

*H |}
-0

I O 0

fºl,
8 bij
ſº
lò
X =
0
:

()

where X; appears in the i + P row of the vector

oſ-i:
- - -
H u = ||.
0:
X =
Q
:
0

18
 8 bij u = "j
i
i

a [+]
8 £i (0)
u =
i i
where uj appears in the i + P row of the vector. Thus, only the gradient of x with
respect to c and the gradients of (z = y) with respect to the bias terms g remain,
where

6p ôp ôp
8 C1 a C2 T 6 Crn

*r- ºr- - - -- +
6C1 6C2 ôCm
Vc X =
25 ºf - - - - 23
ôC1 ô Co 8 Cºn

** **— 8%
ôC1 6C2 T T 8 Cn

A column is needed for each element of c, which includes unknown elements in A and
B, initial conditions, and bias terms.

The elements of the gradient x can be determined in the following manner. If the
state equation is differentiated with respect to aij, for example,

0 0

#- #1J
- -
1J
- -
a# ij #3 nº
y- yº (9)

19
 The solution for #. can be expressed as
ij
8 X (t) -ſ t e Aſt – ' ) Aa. x (f) dr
baij 0 l]

A similar procedure for coefficients in B gives the expression

- *-*.
8x o-ſ’ • *" '' Bº a (, )d.
0

and, when used for elements of the initial conditions x(0), results in

Q
8 x (t :
*** = e At |i
8xi(0) :
0

where the 1 appears in the ith row. The corresponding partial derivatives with respect
to constant bias errors in measuring the state variables can be expressed as follows:

Q
6 (z - y) o
ôgi *
Ö

where the 1 appears in the ith row.

Thus, all the terms in equation (8) have been defined; We J can now be evaluated by
using equation (7). To obtain a gradient solution that minimizes J, the following recur
sive relation is used:

* k + 1 * ck – e Ve J

where the subscript k is the index of the iteration. The value of e can be chosen in
many different ways. One way is to use 70 percent of the value that minimizes the fit
error in one iteration. The gradient method was successful for only a few unknowns
but failed otherwise. Theoretically, each iteration reduces the cost or fit error, but
in practice the reductions become almost infinitesimal. Unfortunately, although the fit
20
error is reduced considerably before it stalls, the corresponding values of the unknown
coefficients are not determined with sufficient accuracy, because the minimum is never
reached.

Many methods have been developed for nonlinear minimization, such as direct
search techniques and accelerated gradient techniques (ref. 11), but for this particular
application a modified Newton–Raphson technique was chosen.

Modified Newton–Raphson Minimization Technique

Minimization based only on flight data. - The Newton–Raphson technique is an

iterative method for finding a zero of a nonlinear function of several parameters, or,
in this instance, a zero of the gradient of the cost functional, that is,

Vc J = 0
Consider a two-term Taylor's series expansion of Vc J about the kth value of ck:

(Weſ). 1 = (ve)) . * (*). Ask, (10)

with

**k + 1 * (* k + 1 “k)
where (v3 J). is the second gradient of the cost functional with respect to c, or the
Hessian matrix, at the kth iteration. If equation (10) is a sufficiently close approxima
tion, the change in c on the (k+ 1) iteration to make (ve J)k + 1 approximately
Zero is
-1

Ack = - ſe: •), (Vc J)k (11)

which is the Newton–Raphson algorithm, as shown in the following sketch:

| Tangent line at cK
|

Vc Jk (c) |
|
`--slope of tangent line = Vc(Vc J )k
| |
l C

°k + 1 °k

This method is much more efficient than the gradient method because it attempts to
predict where the local minimum point is and to step directly to it rather than merely

21
 stepping in the local downhill direction. However, it is much more complex because of
the computation of the second gradient matrix. This complexity can be reduced signifi
cantly by an appropriate approximation to the second gradient matrix which results in
a method termed either modified Newton–Raphson or quasi-linearization. It is most
clearly derived by the method of quasi-linearization.

The difference between the measured and computed responses, z(t) – y(t), can be
represented as quasi-linear with respect to a change in the unknown coefficients, that is,

[.0-0) - |z0 sºlº- + v.[.() - , (), as (12)

Using this approximation in the cost functional results in the following first and second
º gradients:
T T T

We J = 'ſ 0
(z = y); D1 […]."
T T

véJ = eſ
0
|v. (z - 9]. D1 |v. (z -9]. dt

Now the modified Newton–Raphson algorithm becomes

T T - T T

Ack – -|ſize -y]. D1 |v (, - »)." ſ 0

|v (, -y]. D1(z - y) k dt (13)
0

All the terms in equation (13) involve only the first gradients of (z – y), which were
derived previously and are readily computed. Thus equation (13) involves no second
gradients of (z = y) which would appear in the true (vá J).
This greatly reduces the
computation time, and the approximation improves as the solution is approached. All
the preceding terms have been derived for the gradient techniques, so the solution is
readily obtained.

Because the minimization by the Newton–Raphson technique is done in the discrete

case by a digital computer, the discrete approximation transforms the integrals into
summations. Equation (13) becomes

T -1 T - -

*:: |v (, -y). bº ("-39)

i= 1
# |v, (2)-y)' D, (z) - yº)] (14)
j=1

where the superscripts i and j are the indices indicating the time sample, and l
is the total number of samples.

22
 Although the difference in notation might disguise the fact, the preceding solution is
the same as that obtained with the Newton–Raphson method if the term involving the
second gradient multiplied by the noise is neglected as suggested by Balakrishnan in
reference 5. The savings in computation is considerable.

All subsequent references to the Newton–Raphson method will be based upon the
modified Newton–Raphson approach to minimization of the integral-squared error.

To keep computer usage to a minimum, no noise biases or initial conditions were

determined for the X-15 data used, thus

* -(1, Lr Hº Lô, L6, L6 o No Nr Na Nö, No, Noo Yp Yº

r Yôo)
Figure 9 shows the rapid reduction in the weighted fit error, J, when the Newton
Raphson method is used. For the example shown, only four iterations were necessary.
This represents a computation time of about 7 minutes on the XDS 9300 digital computer
(ref. 12). The Newton–Raphson method has been found to be superior to the gradient
method both in terms of the number of iterations and the computation time required for
the application under discussion.

1600

120 H

100 H.

Fit ºw. 80 H.

60 HT

40 H.

IterationS

Figure 9. Convergence of fit error for the modified Newton-Raphson method.

23
 Figure 10 shows the computed state time
histories compared with the measured time
histories. The corresponding weighted mean
square-fit error is 0.2 percent of that obtained
with the least-squares method. Most of the
— Flight reduction is attributed to the improved fit of
50 r- — — — Computed
q2. The fits of p and r are also improved,
p, as expected, because the cost functional was
deg|Sec chosen on the basis of optimizing with respect
-50 – | | to the integral squared error of the difference
between measured and computed data.
10 I
rr ~– Minimization including a priori informa
deg|Sec 0 tion. - Often, independent estimates of the
values of the coefficients are available from
-10 ––––.
wind-tunnel data or previously obtained flight
10 T data. It is desirable to incorporate this a
priori information into the flight data. The
use of the a priori values in the Newton
-10 –1–1–1–1 Raphson method resembles the procedure
often followed in analog matching in which,
40 T initially, the predictions based upon wind
tunnel values are used and changes made to
Q deg 0 improve the fit are weighted against the depar
ture from these values. It seems reasonable
then to use the a priori feature with the Newton
Raphson method, making use of all the infor
p mation available to obtain the estimates and
r 0
deg|Secº Nel.” insuring that no change is made in the deriva
...[\!/ tives unless there is sufficient information in
the flight data. The procedure used is to ex
10 T pand the cost expression to include a penalty
for departure from the a priori values. The
r,
deg'secº 0 NZ sº-> expanded cost becomes
T
J -ſ (z-y)' D1(z - y) dt # (c-co)" K D2 (c-co)
(15)
where co is the vector of a priori estimates
of c.

Recognizing the solution obtained by using

Newton–Raphson to be of the form
-1

Figure 10. Comparison of X-15 time histories

Ac = - ſº Vc J
measured in flight and computed by using
coefficients obtained from flight data with the and the contributions to the first and second
modified Newton-Raphson method. gradients due to the additional term to the
cost expression to be D2 (c – co) and D2,
respectively, the new expressions become
24
 T T

Vc J (ſ 0
(z-y)* D1 […]*) + 2 K D2 (c-co)

T
T

If the a priori information is based upon previously analyzed flight data, the selec
tion of the D2 matrix can be based upon the relative amounts of data involved and the
expected variance of co. If the a priori values are obtained from wind-tunnel tests, it
becomes necessary to estimate the variance of these values, as discussed in the next
Section.

Figure 11 is an example of how the fit error changes as the weighting of the a priori
values is changed. The weighting was varied by changing the weighting factor, K (a
scalar), multiplied by D2. For a weighting of zero the a priori values are ignored, and
for an infinite weighting the flight data are ignored. Shown in the same figure are two
-15 - |
l6 –20 H |
–25 C | | | | | |

Fit error,
J
J = 2 x (minimum J)
| | | | | A
0 10 100 1000 10,000 100,000 Co

Figure 11. Effect of a priori weighting on coefficients and the fit error.

coefficients as a function of the same weighting factor. It is noteworthy that the coeffi
cient Nr changes immediately to its a priori value, but the coefficient Lg does not.

It would appear that the estimate of the coefficient Nr is weaker in the sense that it
could vary over a wide range with little effect on the fit error.
 A weighting factor was used that doubled
the fit error. Although the fit error is nec
essarily increased when a priori information
is used, figure 12 shows that the agreement
Flight of the computed and measured time histories
50 — — — Computed is still good. The use of a priori information
in estimating the unknowns has been found to
reduce the scatter of the estimates (especial
0
deg|Sec ly when the flight data contain little informa
-50– |-|-- tion on a given coefficient) and to reduce the
number of iterations to convergence. How
10 I ever, more computer time may be required
in selecting the overall gain, K, of the a
degked 0H-----→ * priori weighting.
-10–1–1–1–1–1–
Weighting Matrices
10 T

B. deg 0|-r--- Since the cost functional in equation (15)

contains the weighting matrices, D1 and D2,
-10–1–1–1–1–1– it should be expected that these matrices play
40– major roles in the value obtained for each of
the coefficients. There are many ways of
(0 deg 0
determining the D1 and D2 weighting
-40 N----- matrices; the following discussion describes
the method found to be most useful in this
60ſ– -
study.
5, 0|A 2^
deg'secº A reasonable selection of the weighting
-60 \L/ iT matrix, D1, can be based upon knowledge of
l0 T
the statistical properties of the instrumenta
tion noise. If the noise were stationary,
Gaussian, and white,
ove-------...~~~
-lolº-I-I-I-I-I- d1;; = —7-
1

l r— 11 E {*}
OH---e->===- would be the obvious weighting for mean
square minimization. In practice the best fit
-il–1–1–1–1–1– "obtainable" for a particular quantity was
used as an indication of its noise level, in the
following manner. An estimate of the noise
was made for each variable, and the matrix
was assumed to be diagonal. As more experi
ence was acquired through using the data, the
Figure 12. Comparison of the X-15 time histories
measured in flight and computed by using coeffi diagonal elements were changed if a particu
cients obtained from flight data with the modified lar fit of a given variable produced an average
Newton-Raphson method with an a priori weighting error smaller than the original estimate. The
which doubles the unweighted fit error. new value of the element would be the recipro
cal of the new minimum integral error squared
in the measurement of that variable. Grad–
ually, a new D1 matrix based on the minimum
26
º
average error evolved, and further changes were unnecessary.

It should be stressed that much more research is needed in determining a repre

sentative nondiagonal D1 weighting matrix. This can be done adequately only by com
º:
pletely analyzing the instrumentation used to obtain these measurements. Primarily,
such an investigation would supply a covariance matrix of the noise in the output of a
given instrumentation system. In addition, such information as phase shift, amplitude
attenuation, and nonlinearities of the instrumentation dynamics, both electrical and
mechanical, is critical to any analysis of the data. The more representative the data
are of the true observation y, the more satisfactory the results by any method of esti
mation. The elements of the D1 matrix can readily be determined along with the other
unknowns; although the instrumentation remains the same, a different noise covariance
matrix will result for each maneuver.

The values used for D1 are listed in table 1. If part of the data is unusable, the
corresponding element of D1 should be set to zero. This procedure offers an advan

.!
tage over other methods, such as the least squares or Shinbrot, which require a com—
plete set of data. The integral-squared-error method can determine all unknowns,
even when the data used are incomplete.

TABLE 1.-VALUES USED IN THE D1 AND D2 WEIGHTING

MATRICES FOR THE X-15 AIR PLANE

º i dlii yi d2ii Ci

1 1, 530 | p 0.1445 || L6a

2 | 1,040, 000 || r . 1445 || Lor
3 507, 000 || 3 0 || L50
4 31, 700 || @ 522 || Lp
5 635 | p 5. 22 || LF

6 46, 800 | r 1445 L,

7 5,333 ay 14.45 | Noa
8 || — — — — — — - 14.45 | Nör
9 I — — — — — — - 0 | Nö0

10 || - - - - - - - 815.5 | Np
11 I — — — — — — - 815. 5 || Nr

12 l — — — — — — - 14.45 N,
13 || - - - - - - - 0 || Yö0

14 — — — — — — - || 13,010 || Yp
15 || - - - - - - - || 13,010 || Ye

The D2 matrix for weighting wind-tunnel estimates would ideally be the inverse of

27
the error covariance matrix of the test data from which the estimates were obtained,
inasmuch as this would reflect the relative confidence in each of the a priori values.
Since the covariance matrix was not available, the following method, which is only one
of many justifiable procedures, was used to obtain the D2 matrix. No attempt will be
made to show that this method is preferable to other methods. The estimates are
assumed to be independent, which results in D2 being a diagonal matrix. Next, on
the basis of experience with wind-tunnel estimates, a wind-tunnel variance, or 2ii? Was "

selected for each of the unknown coefficients. An attempt was made to use the same
variance for similar coefficients (Lóa, Lör. and Ig, because they are determined
in a similar manner in the wind tunnel). Now D2 was determined by taking the recip
rocal of each variance, +
Or T.
and inserting it as the appropriate element along the diag
11

onal of D2. (These values are shown in table 1.) This fixed the relative weighting
among the individual coefficients. The overall weighting of the wind-tunnel coefficients
to the flight-data-determined coefficients was changed by varying the values of K.

Comparison of Methods

From a comparison of the time histories shown previously, it is evident that the
Newton–Raphson method without the a priori feature provides the best fit of computed
data with the flight time history, and there is only a slight change in the quality of the
fit with the addition of the a priori feature. In addition to the time history fit, it is also
of interest to compare the stability derivatives obtained with the various methods.
These stability derivatives are shown in figure 13 together with derivatives predicted
from wind-tunnel data. Because of the uncertainties in the wind-tunnel predictions as
well as in the flight-determined coefficients, the only definite conclusions that can be
reached regarding the superiority of any of the methods is that the Newton–Raphson
method with the a priori feature provides the closest match of both the flight data and
the a priori values. When the derivatives determined by the Newton–Raphson method
with and without the a priori feature have approximately the same values, it can be
concluded that these are the strong, well-defined derivatives. Thus L6 and L6a are
strong derivatives. For Nr and Nóa there is a large difference between the Newton
Raphson method with and without the a priori feature, which indicates a weakly defined
derivative; similarly, the results from the other methods show little agreement with
one another. In this instance there is not sufficient information in the flight data to
strongly define Nr, and the wind-tunnel estimate is as good as any other estimate.
To make a more rigorous comparison, a statistical model was constructed and used
to indicate the relative values of variance in the estimates obtained by using the least–
squares, Shinbrot, and Newton–Raphson methods. Specifically, a time history was
computed, and actual measured flight data were simulated by adding shaped, unbiased
white noise. The noise was approximated as the error, n (t), for a best fit "obtainable"
for each quantity in the actual flight data previously analyzed. The power spectral
densities were determined for each quantity; however, they were approximately the
same, so only one statistical representation of the noise was used for all the quantities.
Five time histories were generated which differed only in the specific noise added and
28
were analyzed by each of the methods. Estimates of the variance and average value for
each coefficient are shown in figure 14, in which the Newton–Raphson method provides
-. 6 3 –30 30 30 .008
D Re
O

-. 4 2 -zolº 20 g 20 3.00%
lp P Lr LB Lö, & Lö, Goy Np D

-2 P o 1 - 10 10 10 0|- D7
gº-P O o A priori (K = oo)
0 0 [PU 0 0 0 - .004 P. Simplified equations
D Shinbrot
D. Analog matching
© Least squares
D Newton-Raphson (K–0)
4 3 1.2 - - P. Modified least squares
3 D 3 06 o Newton-Raphson with a
p° gº priori (K =57)
2 2 .8F- D –2 tº P .2 -.04 H- GD

0 Po l 4 O -l l -.02
D

-.2 L V6 0 0 0 0 0

Figure 13. Comparison of coefficients determined from X-15 flight data by using several methods of analysis
and predictions based upon wind-tunnel tests.

-.6 4 20 - 60 - 60 .02

º
-.4 2 ſº 0H 40 H. 40 Ol
Ho } Lr B *6, Lör Np
-.2 0 } -20 pºrto “20 | 20 p- o 0

O Lº
KY º
–2 -40 L-
Fºes
0 L-
x->

0
Kołº

-.01
# O Actual
© Least squares
D Modified least squares
Dº Shinbrot
.1 6 1.5 0 T- .3 F- 0 o Newton-Raphson

0 4 1.0 –2 H .2 H. -.05 I Standard deviation

Nr NB Nö, N6, Yp Y5
-. 1 2 .5 H- -4 H. .1 perº - 10

-.2 0 0 -6 L- 0 L- - . 15

Figure 14. Comparison of results of applying several methods of analysis to computed time histories to which
noise has been added.

29
the value closest to the actual value for every coefficient. Although the variance for the
Newton–Raphson method is generally larger than that for the least-squares and Shinbrot
methods, all the actual values are included within the +10 E band about the mean.

Cramér–Rao Bound

It is of considerable interest in parameter estimation to assess the validity of the

estimates of the unknown parameters. If the actual covariance matrix of the estimated
parameters were known, it would indicate which parameters had been most reliably
estimated. The error matrix of the estimates can be defined as

q = E | (c-cr)(c -º'ſ
where c is the current estimate and cºr is the "true" value of the parameter. The
derivation in the appendix shows that this error matrix can be bounded from below by
the matrix Cramér–Rao bound (ref. 5). In other words, the bound provides the minimum
variance with which any of the parameters may be estimated for a given set of data. It
is on this basis that an approximation of the Cramér–Rao bound is used in conjunction
with the Newton–Raphson estimation to assess the amount of confidence to be placed in
the various parameters estimated.

More specifically, by assuming that the estimates obtained by the Newton–Raphson

method are asymptotically unbiased and the noise is a stationary Gaussian white noise
process, a useful form of this bound may be obtained for a direct comparison with the
results of the Newton–Raphson method. These assumptions simplify the expression
for the matrix bound to

/ ...] T ..] ) - 1
• { v.[,]
i=1
D1 Vc [...] (16)

This expression is derived in the appendix. Although the specific data being analyzed
do not conform exactly with the preceding assumptions, a comparison of this bound
with the variance determined from the statistical model of the last section would give
some insight into the relative reliability of the estimates. The assumption of asymp
totically unbiased estimates can be shown to be reasonable because, through use of the
Newton–Raphson method, the first term of equation (14) changes only slightly from
one iteration to another. If no change occurs, this term can be deemed to be independ
ent of the noise. Thus, if #tº = 0,
1
the noise is unbiased and Ac is a linear trans

formation of the noise. So, if the noise is unbiased, the estimate of the coefficient is
unbiased. The assumption of the statistics of the noise is not met by the data, but the
actual noise may be such that the Cramér–Rao bound will give additional insight into
the validity of the estimates obtained.

Because the first term in equation (14) is identical to the term derived for the
Cramér–Rao bound determined under the somewhat idealized assumptions, the approxi
mation of the Cramér–Rao bound is available in the computation of the Newton–Raphson

30
minimization. Figure 15 compares the standard deviations obtained by using the
Cramér–Rao bound from the Newton–Raphson computation with those obtained by analyz
ing the statistical model discussed in the preceding section and shown in figure 14. It
.8 –

18, 15, lº l, 16 No, Nº, No Nr NB Yº Yº

Figure 15. Comparison of the standard deviation indicated by the approximation of the Cramér–Rao bound
and by experiment.

should be noted that figure 15 shows that the standard deviation of the experimental
model is greater than that predicted by the application of the Cramér–Rao bound. The
O.
ratios of are fairly well centered about 0.35. Thus, the theoretical result pro
OE
vides insight into the relative confidence that might be placed on the estimates.
If the coefficients of the a priori feature are to be used, it might be anticipated that
the same computation may not represent an estimate of the Cramér–Rao bound. How
ever, the same type of information may be obtained with the a priori weighting by inter
preting the calculation in the following way: If the total probability p(z, c T) were used
instead of p(z |cT) in the derivation in the appendix, an expression similar to that for
the Cramér–Rao bound would be readily found. The cost functional associated with
using the total probability is that used with the a priori feature.

The estimates without a priori information can be considered to have the D2 ma

trix equal to the null matrix. The Cramér–Rao bounds of the variances obtained are
used to assess the reliability of the estimates of the coefficients. That is, if the vari
ance is large, less confidence is placed in that estimate. In this instance, the variances
obtained are compared with the inverse of the D2 matrix, which is infinite for all
coefficients. Therefore, the larger ("closer to infinity") the bound, the less confidence
is placed in that estimate.

Generalizing to the case where D2 is a nonsingular matrix, once again the bound
of the variance should be compared with the inverse of D2. So the computation

31
analogous to the Cramér–Rao bound ["aii] will satisfy D5" > [...] To make proper

use of this information, the values of the bounds closer to the corresponding element
-1
of D2 will correspond to the least reliable estimates. In the limiting case, the values
would be identical. This would indicate that no additional information is contained in
the data. Although the comparison is not made as readily for a non-null D2 as for a
null D2, the information is still contained in the same computation. The result would
not properly be called the Cramér–Rao bound, but the relative reliability of the esti
mates can still be ascertained.

RESULTS FOR THREE CLASSES OF AIRCRAFT

The Newton–Raphson technique of minimizing the integral-squared error was

applied to aircraft of widely varied geometry and flight regions to demonstrate the
versatility of the method. A light general aviation airplane, a large supersonic air
plane, and a lifting body vehicle were selected, in addition to the X-15 airplane dis
cussed previously, to represent the extremes of modern aircraft. In all the vehicles
a variable bias in measurement was allowed for the first three elements of the state
vector (p, r, B). Variable bias was not used on the remaining measurements, nor
were any variable initial conditions used, because the increased accuracy realized
from these additional variables did not justify the computer time needed to solve for
more unknowns. Not all the measurements available on the X-15 airplane were obtain
able on every other vehicle. Two separate time histories were analyzed simultaneously
on the XB-70 airplane to demonstrate how several time histories at the same flight
condition may be analyzed together.

The D1 (table 2) weighting matrices were generated in the manner discussed

earlier. However, the D2 weighting matrix was the same as that used previously,
and only the overall scalar weighting, K, was changed for each vehicle. The overall
weighting used was that which doubled the mean square error attained with zero weight
ing. As in the X-15 data previously analyzed, the a priori values used were the pre
dicted values based upon wind-tunnel tests.

Light General Aviation Airplane

To show that the method can be applied to a low-speed, light, conventional airplane.
a rudder maneuver for a light twin-engine (ref. 13) airplane was chosen. In figure 16
a flight time history of this airplane is compared with a time history computed from
the a priori values based primarily on wind-tunnel data. It should be noted that bank
angle, q, was not measured and no aileron input was made. Thus, the aileron deriv
atives were not determined, but the match was still attempted. To obtain the match,
the element of D1 corresponding to p was set to zero.
In figure 17 the measured data are compared with the Newton–Raphson solution with
no a priori weighting. It is obvious that the Newton–Raphson solution more closely
approximates the flight data than do the wind-tunnel estimates. The most significant

32
 TABLE 2. —VALUES USED FOR THE D1 WEIGHTING MATRICES FOR A LIGHT
AIR PLANE, THE XB-70 AIR PLANE, AND A LIFTING BODY VEHICLE

Light airplane XB-70 Lifting body

Lateral – Directional

i Vi d1i d1ji d1ji

1 p 103,000 13,400 14, 300
2 r 33, 300 562, 000 353, 000

3 B 294,000 208, 000 289, 000

4 Ø 0 75, 500 9,900

5 p 25, 300 0 0

6 r 57, 100 0 0

7 ay 13, 100 27, 700 24. 000

Longitudinal
i yi - - - - - d1ji - ----
1 q I — — — — — 20, 000 || – — — — —
2 Oy - - - - - 3, 431 || – — — — —

3 V | — — — — — 0 | — — — — —

4 6) I — — — — — 0 l — — — — —

5 q | ---- - 800 || — — — — —

6 az I - - - - - 200 || – — — — —

7 ax - - - - - 0 | — — — — —

33
 departure is in the ay trace.

Figure 18 compares the flight

— Flight data with the time history com—
10 - — — — Computed puted by using the Newton–Raphson
method with an overall a priori
Öa, deg 0 weighting of K = 1.0. The fit is
| | | | | |-|--|- improved somewhat over that in
-10
figure 17 in some of the signals
10 but is poorer in others, including
a...
Ör, deg 0
Nſ
|\ |-|--|--|-
y

Figure 19 is a summary of
-10 V_J the coefficients obtained by the
10 three different methods. It should
be noted that the coefficients used
_2^ 2= _2===
for the a priori values of Lp, Lr:
dejsec NS-27
| | |--|--|--|-- were computed from
-10 Np. and Nr
10 the vehicle geometry and other
aerodynamic characteristics and
deſsed
A 2=---- not obtained from wind-tunnel
\ V
1–1–1–1–1–1– data. The wind-tunnel fit and the
a priori weighted fit agree fairly
well except for Ig, Lr, and Ya.
ATS The coefficient Lg is particu
B. deg 0 T-_--"
larly interesting, in that it was
forced away from the a priori
value in order to improve the
correlation with the other coeffi
cients. It appears that there is a
significant amount of information
in the flight data to indicate that
Yº should not actually be at the
a priori value. Also, it is inter
/\ z-Tº-S~~~~-*=–- esting to note that Lp, Np, and

|\ºf 1–1–1–1–1–1– Nór are at the a priori value

when weighted and show a signi
ficant departure when unweight
ed. This supports the conclusion
that little information on these
| | |→l– coefficients is contained in the
0 1 2 4 5 6 7 8 9 10
data. Although this might be attrib
t, Sec uted to the relative weighting in
the D2 matrix, which certainly
is a primary factor, L5 and
Figure 16. Comparison of light airplane time histories measured
in flight and computed by using a priori values. K = • , Np exhibited the opposite be
havior for the X-15 airplane

34
 — Flight
10 T- — — — Computed

Öa, deg 0 =

-10 | | | | | | | | | |
10 F

ðr, deg 0 V

*D | TTTTT
VJ

10 I
P, 0 2^ __*~ =-

deglsec N--~~
-10 |→l—l-l-l-l-l-

deq/
p T/\
0 ^YS->~-— * - - -

* . V 1–1–1–1–1–1–
10 —

B. deg 0 E=- ^s T

...tº ºr TTT
20 T- ,
py 0 %
dealsecº *

* …, || | \ſ | | | | |
- 40 T

r,
deg|Secº 0
-40 H–1–1–1–1–1–1–
.2 r

ay, 9 0 • --> - --

Figure 17. Comparison of light airplane time histories measured in flight and computed by using coefficients
obtained from flight data with the modified Newton-Raphson method without a priori weighting. K = 0.
 Flight
10,– — — — Computed
Öa, deg 0

-10 | | | | | | | | | |

Ör, deg 10–

0
ſ\ V

NI TTTTTTTT
U

10 I
p, _^ St- ====ma

deglsec 0 <=-->
-10 |→
10 —

r,
deglsec
l/A\!//*NN__*__*** -
-10 | \, |-|--|--|--|--
10 T

B. deg 0 2^ ~:=
...DY-Z S-2T T.
20 T
p, a - -

deg'secº 0
–20 | |→

40 r

r,
deglsecz 0
-40 – 41–1–1–1–1–1–1–1–
.2 I

ay, 9 0 Hºe ~ :--> ~~~~< - - --

-.2 | |→l
0 1 2 3 4 5 6 7 8 9 10

Figure 18. Comparison of light airplane time histories measured in flight and computed by using coefficients
obtained from flight data with the modified Newton-Raphson method with an a priori weighting which doubles
the unweighted fit error. K = 1.0.
36
 D

-8 8 –D -4 8 -10
O D
Ho H D ºr |B Lör o"p D

4 O C. 4 Ó -

2 O 4 -

0 0 0 0 0 O
o A priori (K = oo)
D Newton-Raphson (K = 0)
-. 6 6 -6 -, 3 o Weighted Newton-Raphson (K = 1)

- 4 O 4 4 -, 2 D C

- 2 2 D -2 O -. 1 O

0 L- D 0 0 0

Figure 19. Comparison of light airplane coefficients determined from flight data by using a modified
Newton-Raphson method and from a priori estimates.
data which had the same relative weighting. This points up once again the need for
more information on the covariance matrix of the wind-tunnel data which are intuitively
the values desired for the D2 matrix. Of course, the covariance matrix of wind
tunnel estimates would be different for each vehicle.

Large Supersonic Airplane

The large, supersonic-cruise XB-70 airplane (ref. 14) was chosen to represent
this extreme in aircraft. The maneuvers selected consisted of an aileron, Öa, maneu

ver followed by a rudder, Ör, maneuver at a later time in the same flight at essentially
the same flight condition. The two maneuvers were analyzed simultaneously. Fig
ures 200a) and 2006) compare the flight time histories of the two maneuvers with the com—
puted time histories based upon wind-tunnel estimates modified for elasticity effects
(a priori). The p and r signals were not measured, but the match was obtained by
setting the corresponding elements of D1 equal to zero (table 2). In contrast to the
fit for the light airplane, the agreement is poor. There is considerable drift in all the
signals largely because of the unknown bias in the controls, except p and r, and the
amplitudes are vastly different on the p and B signals..

The ability to match both time histories simultaneously would be a distinct advan
tage over some of the current methods of obtaining coefficients. With these techniques
the control derivatives Lör. Nór, and Yö r could not be determined from the time
:
37
 — Flight
— — — Computed
4r
Öa, deg 0 ſ

-4 | ––––––––

ðr, deg 0 U---

4–1–1–1–1–1–1–1–1–1–1–

. 1 [-
ay, 9 0 =e *A*-SS <==>–==

-1–1–1–1–1–1–1–1–1–1–1–
0 2 4 6 8 10 12 14 16 18 20 22
t, Sec

(a) First maneuver.

Figure 20. Comparison of the XB-70 time histories measured in flight and computed by using a priori values. K = ce.

38
 — Flight
4 – ––– Computed

& deq 0}
–4 |-|--|--|--|--
4 I

Ör, deg 0 A—
T-ſi T, TTT

.1 ſ−
2^\. 22-SS eº ->

dy, 9 0 - SNL/ |-- --

-.1 ––––––––––
0 2 4 6 8 10 12 14 16 18 20
t, Sec

(b) Second maneuver.

Figure 20. Concluded.

39
 history in figure 20(a), so these
values had to be fixed at some a
priori value while the estimates
were obtained. A similar situa
tion exists for L5 , N5, , and
Flight a a
— — — Computed Yôa in the time history of fig

ure 20(b). The net result of

these techniques is a different
set of control derivatives for
two maneuvers known to be at
the same flight condition. This
could be alleviated rather labo
riously by repeating the proce
dure several times, each time
updating with the most recently
obtained control derivative.

The problem can be elimi

nated with the Newton–Raphson
method by recognizing in equa
tion (14) that the summations
may be continued for more
points if corrections are made
for the initial conditions at each
of the discontinuities where a
new time history is started.
This method was used to obtain
the comparison of flight data
with data computed by using the
Newton–Raphson method without
a priori weighting in figures 21(a)
and 21(b). The match is consid—
ered to be good, and little could
be suggested to improve it.

Figures 22(a) and 22(b) com

pare the two flight time histories
with the computed time histories
- 1 0=2 −1–––––––––.
4 6 8 10 12 14 16 18 20 22
based upon the estimates ob
tained by the Newton–Raphson
t, SeC method with the a priori weight
ing set at K = 80. The fit would
still be considered good, al
though some phase shift is now
(a) First maneuver. apparent, particularly in the
r, 3, and ay traces.
Figure 21. Comparison of the XB-70 time histories measured in flight
and computed by using coefficients obtained from flight data with the
modified Newton-Raphson method without a priori weighting. K = 0. Figure 23 is a summary of
the estimates obtained by using

40
 — Flight
— — — Computed

1–1–1–1–1–1–1–1–1–

_^a 2. TS _*~ _*

H––––––––
2 4 6 8 10 12 14 16 18 20
t. SeC

(b) Second maneuver.

Figure 21. Concluded.

41
 — Flight
— — — Computed

Ör, d eg 0

-1–1–1–1–1–1–1–1–1–1–1–
0 2 4 6 8 10 12 14 16 18 20 22
t, Sec

(a) First maneuver.

Figure 22. Comparison of the XB-70 time histories measured in flight and computed by using coefficients obtained
from flight data with the modified Newton-Raphson method with an a priori weighting which doubles the unweighted
fit error. K = 80.

42
 — Flight
- — — — Computed

Ča, deg
| | | | |

2 I
ſ:... O ATN /N ATN 2-->
deg/sec S.--

2–1–1–1–1–1–1–1–1–1–
2 I

B. deg 0
2– 1–1–1–1–1–1–1–

10 T- 2^N.

10–1–1–1–1–1–1–1–1–1–
. 1 [-

dy, 9 o-º-º-º-----
-. 1 1–1–1–1–1–1–1–1–1–
0 2 4 6 8 10 12 14 16 18 20

t, sec

(b) Second maneuver.

Figure 22. Concluded.

43
the three different methods. It is significant that Ip, Pôa. Ng, Nór, Yp, and Yº
are approximately the same for all three methods, but I-3, Lör, and Nóa are rela

tively unaffected by the a priori weighting. In contrast, in figure 13 for the X-15 air
plane it was noted that N6,a was readily changed to the a priori value and in figure 19
for the light airplane Lör was easily changed to equal the a priori value. Thus, differ
ent combinations of derivatives appear to be strong or weak for each maneuver analyzed.

-1.2 .4 -12 3 1.5 - . 15

D
Ó

-.8 2| P -8 2 1.0 -. 1
O O
p Lr OLB [] ©l6,
‘’d lº D
Ö
O r D
oN p []

- 4 0 –4 l 5 05 —
- O O

- 0 0 0 0
0 .. 2 o A priori (K = oo)
D Newton-Raphson (K = 0)
-

.3 3 -
08 -1.2
l .. 3 -

. 12 © Weighted
(K = 80) Newton-Raphson

-.2 D 2 .04 8 [] 2 - .08

Nr NB N6, Nör op YB
D O D C
-

|o o ||5 © 0
-

4 l -.04|Q D Q

D O
0 0 - .04 0 0 0

Figure 23. Comparison of XB-70 coefficients determined from flight data by using a modified Newton-Raphson method
and from a priori values.

Lifting Body Vehicles

Lifting body reentry vehicles (ref. 15) represent a wingless aircraft with somewhat
unusual aerodynamic characteristics. The lifting body maneuver selected for analysis
is a combination aileron and rudder maneuver.

Figure 24 is a comparison of time histories obtained from flight data and computed
from the predictions based upon wind-tunnel (a priori) estimates. The p and r sig
nals were not measured. The drift of the computed signals is severe, probably largely
because of biases in the control measurements.

Figure 25 compares flight data with data computed by using the Newton–Raphson
method without a priori weighting. The fit is good on p, q), and ay, but there is some
phase shift in r and 3.

44
 — Flight
8 r- ––– Computed

-8–1–1–1–1–1–1–1–1–1–1–1– ſ

4 r- ^=
4–1–1–1–1–1–1–1–1–1–1–1–1–1–
20 —

º
eg/sec "E=-4A-A-A-7-7
-20TS---|- A \ l A JVC Aſ
`-- /* z- > - ~~~
| S-- || ||
/ \ , Nvº
4 – \ | \ſ \-/

Figure 24. Comparison of lifting body vehicle time histories measured in flight and computed by using
a priori values. K = co.
45
 — Flight
— — — Computed

-8–1–1–1–1–1–1–1–1–1–1–1–1–1–

P,
dejsec 0 Lea-’
/\\ a . . />
-20 —
4 I-
|\/\/Y. V
| | | |

deglsec
al-A
-4 WNZNZS-T-/~
1–1–1– | | | |

Figure 25. Comparison of lifting body vehicle time histories measured in flight and computed by using coefficients
obtained from flight data with the modified Newton-Raphson method without a priori weighting. K = 0.
46
 In figure 26 flight data are compared with data computed by using the Newton
Raphson method with an a priori weighting of K = 22. The fit is still fairly good, but
some of the maximum excursions fail to match.

— Flight
––– Computed

t, sec

Figure 26. Comparison of lifting body vehicle time histories measured in flight and computed by using coefficients
obtained from flight data with the modified Newton-Raphson method with an a priori weighting which doubles the
unweighted fit error. K = 22.

47
 Figure 27 is a summary of the coefficient estimates obtained by using the three
methods. For this vehicle Lp, LB, Ng, Nör, Yp. and Yg are approximately the

same for all three methods. The coefficients Np and Nóa appear to remain near the

unweighted estimates in spite of the weighting.

-1.2 - -1.2 r- -120 r- 40 T- 12 – .3 r

-.8H 7.8 p -80 H. 30 H. *F. 2 H.

lp o p <> |r |B D Ó | 6a Lör Np
O O D O O Po
-, 4 H -. 4 H -40 H 20 ſo 4H O l H

K» O

0L 0L- 0 L- 0 L- 0L- 0L o A priori (K = oo)

D Newton-Raphson (K =0)
-.8 ſ− 30 I- 3 r- -12 r- . 3 I- -.3 [- © Weighted Newton-Raphson
(K = 22)

-. 4 H. D. 20 H 2 H –8 H 2H -.2 H.
Ó D o Do O
Nr O NB N6, o Nör OD > Yp Yp DO
O Ó
-.2|− 10 — D 1 |O -4 – l H -. 1 —

0 L- 0 L- 0 L- 0 L- 0 L- 0L

Figure 27. Comparison of lifting body vehicle coefficients determined from flight data by using a modified Newton
Raphson method and from a priori values.

Summary of Results for Various Vehicles

Even though the same relative weighting in the D2 matrix was used for all four ve
hicles, different coefficients in each case are affected by that same weighting. Although
this does not indicate that the D2 matrix is adequate, it does show that the information
in the data primarily determines the estimate of the coefficient. This is a desirable
characteristic of the technique, inasmuch as it departs from the a priori value only if
the measured data contain information contradictory to the a priori coefficients.
In summary, the application of the Newton–Raphson method to the four vehicles
resulted in (1) a good fit of the computed data with the flight data, (2) a set of reason
able derivatives, and (3) the a priori option which indicated the coefficients to place
confidence in.

APPLICATION TO SHORT-PERIOD LONGITUDINAL MODE

I I
• , - *

With x = Ax + Bu and y -F#, + Fiju as the constant coefficient differential

48
 equation, the Newton–Raphson method was used on the so-called short-period longitu
dinal aircraft mode where

6 Q § -:-

6 V V az
0 6 -
6 a
X

A = B =

Xq Xo, XV Xo X5, X60

1 0 0 0 0 0

G = 0 Z Oy 0 0 H = Z Öe Z Ö0

0 Xo, 0 0 Xöe X60

The coefficients in G and H are also in A and B. Because only the short-period
mode was analyzed, only the elements of the first two rows and columns of the A and
B matrices and the biases on the state variables q and Cy were variable.

The XB-70 (ref. 14) flight data were used for this part of the analysis. Figure 28
is a comparison of the flight-measured data and data computed from the wind-tunnel
predicted (a priori) estimates. Only the q, o, q, and az signals were used, and the
- corresponding elements in the D1 (table 2) matrix were set to zero for the other sig
nals. The fit is fairly good, although some departure is noted in all the signals.
Figure 29 compares the flight data with the data computed by using the coefficients
obtained from the Newton–Raphson method without a priori weighting. The fit is ex
tremely good for all the signals; very little could be suggested to improve it.

Figure 30 compares the flight data with the computed time history based on coeffi
cients obtained with the Newton–Raphson method with an a priori weighting of K = 0.03.
The fit is still good, but the q signal shows a small departure near the maximum
excursions. The coefficients are now weighted toward the a priori prediction, thus this
is a satisfactory compromise.

49
 The coefficients obtained by the three methods are summarized in figure 31. The
coefficients agree well except for Md, Möe, and Zöe. The a priori weighting tends to

bring these quantities toward the wind-tunnel values. The initial peak on az and q in
figure 28 showed the computed values to be larger, so it is obvious that the data contain
information contrary to the wind-tunnel values. One apparent advantage in determining
a smaller number of unknowns is that a visual interpretation of the time histories can be
maintained for the effect of each of the coefficients. This gives some confidence that the
a priori feature of this method is performing the same function for more unknowns.

— Flight — Flight
–––. Computed ---ºne
12 – 12

Öe deg 6H- \" - Öe deg H \"

0–1–1–1– 0–1–1–1–
2 ſ− 2 I
Q, 0 A Q, 0
deg|Sec sy-\z--- deglsec

*F--------- * --~
q, deg 0 G. deg 0H
-12–1–1–1– -12–1–1–1–
"H = -
-->=- 10 C-T-

9, deg OH- 6, deq 0H

-10–1–1–1– -10–1–1–1–
- 4 - a 4 -

"..., 0 d
deg'secº 7- dejsecº 0
-4 | | -4 | "Y Ll

OF- 0 –

d7, 9 -1H-->==- 37, 9 -l —-

–2 — | –2 —— |
0 2 4 6 8 0 2 4 6 8
t, Sec t. SeC

Figure 28. Comparison of XB-70 longitudinal time Figure 29. Comparison of XB-70 longitudinal time
histories measured in flight and computed by using histories measured in flight and computed by using
a priori values. K = Co. coefficients obtained from flight data with the
modified Newton-Raphson method without a priori
weighting.
50
 — Flight IMPROVING MANEUVERS
— — — Computed
12 I
Öe deg 6 – V In future studies of coefficient determi–
nation, better maneuvers should be defined
0–––– in order to maximize the amount of infor
mation contained in the data. The data
2 I- o analyzed in this report were generated
Q, specifically for existing analysis techniques.
deglsec 0 This generally entailed separating data
-2-1-1-1– into various segments, such as one in which
only one control was moving and another
with no input so the zero input dynamics
appeared clearly. Because the stability
and control coefficients are needed for
sophisticated research vehicles which only
briefly maintain a given flight condition,
these types of maneuvers are difficult to
obtain.
9, deg 0
-10 –––– Heuristically, certain types of maneu
vers can be defined that would increase the
4 I information content of the data. If no a
priori information about the coefficients
sº 0 were available, white noise would be desir
-4 able as the control input to insure that all
frequencies were represented in the input.
0 r— Knowledge of the coefficients implicitly pro
vides knowledge of the frequency of the
*z, 9 -1 -->- modes being investigated. It might be
assumed that the control input should resem
-2 0––––
2 4 6 8
w
ble a frequency sweep in the frequency
range of interest. In addition, the ampli
t, Sec tude of the resultant maneuver should be in
Figure 30. Comparison of XB-70 longitudinal time
histories measured in flight and computed by using
the linear range of the coefficients, but the
coefficients obtained from flight data with the motions should be large to minimize the
modified Newton-Raphson method with an a priori effect of the noise. In summary, intuition
weighting which doubled the unweighted fit error. suggests a frequency-sweep control input,
" K = 0.03.
*
-1.2 I- -1.5 ſº
O D O
-1.5 T-
O
- 1.2 - - . 15 T
O

O
O
-.8 H -1.0 H. -1.0|- P - .8 o -.1 H o A priori (K = oo)
º O

D
°25, D Newton-Raphson (K = 0)
© Weighted Newton-Raphson
-.4 T D -.5 H -.5 H - .4 H. - .05 H P (K = 0.03)

0 - 0– 0 L- 0 - 0 L

º Figure 31. Comparison of XB-70 longitudinal coefficients determined from flight data by using a modified Newton
º
Raphson method and from a priori values.
51
:
large motion in the maneuver, and linear responses.

CONCLUDING REMARKS

Stability derivatives or coefficients of the state equations were determined by using

several methods. An example set of X-15 lateral-directional flight data was analyzed by
using simplified-equations, analog-matching, least-squares, and Shinbrot methods. The *
same set of flight data was analyzed by using a gradient method to find the set of coeffi
cients that would minimize the integral-squared error between the measured state and *
that computed by using the coefficients obtained with the same control input. Difficulties
in convergence were encountered when there were more than a few unknowns. §:

A modified Newton–Raphson (quasi-linearization) method of minimization was used

with success to solve the convergence problem. The resulting fit of the flight data was
clearly superior to that of the least-squares and Shinbrot methods. One important ad
vantage of the Newton–Raphson method is that it is not necessary that all components of
the state variables and their time derivatives be measured.

The use of a priori values (for example, values predicted from wind-tunnel data) in
the Newton–Raphson method resulted in estimates that made full use of all available
information. When the a priori option was used, the coefficients that were only weakly
determined by using only flight data approached their a priori values and the fit error
increased slightly. This was found to be an effective way of utilizing previously ana
lyzed data together with new data to update the estimates of the coefficients.

A model statistically similar to the flight data was constructed and used, and the
values for the coefficients were determined by using the least-squares, Shinbrot, and
Newton–Raphson methods. The Newton–Raphson method was superior to all other
methods for the example considered.

An expression was derived with the Cramér–Rao inequality to obtain a bound for the §
error covariance matrix for an idealized case. The results of this bound were comparº tº
with the variances obtained from the statistical model. The Cramér–Rao bound was
found to give insight into the reliability of the estimates obtained with the modified
Newton–Raphson method.

To demonstrate that the Newton–Raphson method was generally applicable to a widº

range of modern aircraft, it was applied also to a light general aviation airplane, a
large supersonic airplane, and a lifting body vehicle. The method proved to be satist
factory for each of the maneuvers analyzed and for two maneuvers made at the same
flight condition which were analyzed simultaneously. |

The Newton–Raphson technique was applied to longitudinal equations of motion of an *:

airplane and was found to be successful for the short-period longitudinal mode analyzed
To obtain better estimates of the coefficients of the state equations more research
is needed to define the weighting matrices used in the cost functionals and the types of
maneuvers needed to maximize the information contained in the data.

Flight Research Center,

National Aeronautics and Space Administration,
Edwards, Calif., October 4, 1971.

52
 APPENDIX

CRAMER-RAO BOUND
---

º: In an identification problem it is useful to have a bound on the error covariance

... matrix. By making several assumptions about the noise and the estimates of the coeffi
is cients, some indication of the reliability of the estimates can be obtained by applying
... the Cramér–Rao inequality (ref. 5). By assuming the noise to be Gaussian and defining
... an extended noise vector, ne, formed by stacking subsequent (R × 1) noise vectors
º: sequentially with increasing time, the probability density function can be represented as
ºt 1 - T ...,-1
º *(*)= -RT. 1/2|ng w ne
-: 2T 2 lw
where

º
ſº

n".
º n?
º n – :
e - -

£: n!-1
f n!

º and the superscript i represents the time of the sample, l is the total number of
: Samples, and W is the covariance matrix of the noise, ne.
ſ:
t For the noise, ne = ze-ye, the conditional probability density function can easily
be shown to be

-1

p(ze |ct) = ——
R
,-1/2 (ze - ye)"w (*e - ye)
#m |1/2
2Tſ 2 w
where ze is the sampled data vector and ye is the sampled true state vector as
follows:

53
 APPENDIX

and where cT is the true value of the estimate and the superscript i represents the
time index of the sample. If, in addition, the estimate of c is assumed to be asymp
totically unbiased, the Cramér–Rao bound can be written as

* = E{(-en)(e-et)}
| T |-
2 * We log º [. º ſ

where the expectation E{-} is with respect to ze

-1

E |. ye' W -1
T
(ze ye)(ze ye)
- -
T
w-lve
-

w
Substituting ne = ze – ye,

-1

- ** ye" wil T
ne ne' W-1 Vc *
By further assuming that the noise is also white, the representation for W-1 becomes
block diagonal such that

F- : -

| | | : ] |
D*| 0 || 0 | : 10 | 0
- T - T - - - - -- - - --
0 Dº! 0 || - -'º -- 0.
| | 3 : | |
0 – 0 – D – – f – 9 — — —9–
| | |
– º –
W-1 = .....! - - - - - - !......] ...i....!..... ...!- - - - - - -

| | : | |

0 1 0 1 0 1-i-p'º' 0.
| | | W.
|
0 L 0 || 0 || – ; – 10 – | D:
| :

: |
| | | |

54
 APPENDDX

where Di is (R × R) and W is (l R × l R). With this assumption the last equation

simplifies to

–1
/ l" ºr , *T . -

q 2 K E |XX, we | yi Di nº n* D've | yi
i =1

Interchanging the summation with the expected value and noting that only the noise ni
is a function of z”,

–1
..]" .T
/
q 2 K XX Vc y! | Di
i=1
E mini } . .T
Dl
-

v.[y] -

Noting for the assumption of white noise that

. . T ...] –1
E (nºw } =[p]
then

T -1
&
q 2 K XX
i =1
v.[y] D1 ve [...]
- I - -

If, in addition, stationarity of the noise is assumed, then

D = Di = D.) for all i and j

Thus, the desired result for the matrix Cramér–Rao bound, p, of the error covariance
matrix becomes

& T —1

• *}#, “[y] ºv.[,]

1 F

55
 REFERENCES

Wolowicz, Chester H. : Considerations in the Determination of Stability and Control

Derivatives and Dynamic Characteristics From Flight Data. AGARD Rep. 549–
Part 1, 1966.

Rampy, John M.; and Berry, Donald T. : Determination of Stability Derivatives

From Flight Test Data by Means of High Speed Repetitive Operation Analog
Matching. FTC-TDR-64–8, Air Force Flight Test Center, May 1964.

Howard, J. : The Determination of Lateral Stability and Control Derivatives From

Flight Data. Canadian Aeronautics and Space Journal, vol. 13, March 1967,
pp. 127-134.

Shinbrot, Marvin: On the Analysis of Linear and Nonlinear Dynamical Systems

From Transient-Response Data. NACA TN 3288, 1954.

Balakrishnan, A. V., ed.: Communication Theory. McGraw-Hill Book Co., c. 1968,

Bellman, Richard E. ; and Kalaba, Robert E. : Quasilinearization and Nonlinear

Boundary-Value Problems. American Elsevier Publishing Company, Inc., 1965.

Mechtly, E. A. : The International System of Units – Physical Constants and Con

version Factors. NASA SP-7012, 1969.

Anon. : Dynamics of the Airframe. Bu. Aer. Rep. AE-61–4 II, Northrop Corp.,
Norair Div., Sept. 1952.

DeRusso, Paul M. ; Roy, Rob J. ; and Close, Charles M. : State Variables for
Engineers. John Wiley & Sons, Inc., c. 1965.

10. Yancey, Roxanah B. : Flight Measurements of Stability and Control Derivatives of

the X-15 Research Airplane to a Mach Number of 6.02 and an Angle of Attack of
25°. NASA TN D-2532, 1964.

11. Powell, M. J. D. : A Survey of Numerical Methods for Unconstrained Optimization.

Studies in Optimization 1, Soc. for Industrial and Applied Math. , c. 1970,
pp. 43-61.

12. Anon. : SDS 9300 Computer Reference Manual. 90 00 50E, Scientific Data Systems,
July 1966.

13. Fink, Marvin P. ; and Freeman, Delma C., Jr. : Full-Scale Wind-Tunnel
Investigation of Static Longitudinal and Lateral Characteristics of a Light
Twin-Engine Airplane. NASA TN D-4983, 1969.
14. Andrews, William H. : Summary of Preliminary Data Derived From the XB-70
Airplane. NASA TM X-1240, 1966.

56
 15. McTigue, John G. ; and Ryan, Bertha M. : Lifting-Body Research Vehicles in a
Low-Speed Flight Test Program. Annals of the New York Academy of Sciences,
vol. 154, art. 2, Nov. 22, 1968, pp. 1014-1032.

º:

NAsA-Langley, 1972 — 19 H–626 57

NATIONAL AERON AUT ICS AND SPACE A D M ISTRATION
WASHINGTON, D.C. 20546
PostAGE AND FEEs PAID
NATIONAL AERONA utics and
OFFICIAL BUSINESS
space: A DM 1 Nist RATION -
PENALTY FOR PR v ATE use $300
FIRST CLASS MAIL

005, 002 c1 U 35 7202 18 SO 0 || 39H U

UNIw of ILLINors
LIBRARY
ATTN: Docum ENT LIBRARIA Nº
URB AN A I 1801

POSTMASTER:. If Undeliverable
Postal Manual) Do(Section
Not Re1

"The aeronautical and space activities of the United States shall be .

conducted ſo as to contribute . .. to the expansion of human knowl
edge of phenomena in the atmoſphere and ſpace. The Administration
shall provide for the wideſt practicable and appropriate diſsemination
of information concerning its activitieſ and the reſults thereof.”
— NATIONAL AERONAUTICS AND SPACE ACT OF 1958

NASA SCIENTIFIC AND TECHNICAL PUBLICATIONS

TECHNICAL REPORTS: Scientific and TECHNICAL TRANSLATIONS: Information

technical information considered important, published in a foreign language considered
complete, and a lasting contribution to existing to merit NASA distribution in English.
knowledge.
SPECIAL PUBLICATIONS: Information
TECHNICAL NOTES: Information less broad derived from or of value to NASA activities.
in scope but nevertheless of importance as a Publications include conference proceedings,
contribution to existing knowledge. monographs, data compilations, handbooks,
sourcebooks, and special bibliographies.
TECHNICAL MEMORANDUMS:
Information receiving limited distribution TECHNOLOGY UTILIZATION
because of preliminary data, security classifica PUBLICATIONS: Information on technology
tion, or other reasons. used by NASA that may be of particular
interest in commercial and other non-aerospace
CONTRACTOR REPORTS: Scientific and applications. Publications include Tech Briefs,
technical information generated under a NASA Technology Utilization Reports and
contract or grant and considered an important
Technology Surveys.
contribution to existing knowledge.

Details on the availability of these publications may be obtained from:

scienTIFIC AND TECHNICAL INFORMATION office

NATIONAL AER ON AUTICS AND SPACE ADMINISTRATION
Washington, D.C. 20546 -
 629:13
UnS25tn

* 4 3 22
\

NASA. T E C H IN I C A L IN OTE NASA TN D-6580

:
DRAG OF A SUPERCRITICAL BODY
OF REVOLUTION IN FREE FLIGHT
AT TRANSONIC SPEEDS AND
COMPARISON WITH WIND-TUNNEL DATA

by J. W. Usry and John W. Wallace

Langley Research Center

Hampton, Va. 23365

WATIONAL AERONAUTICS AND SPACE ADMINISTRATION - WASHINGTON, D. C. • ‘DECEMBER 1971

 DRAG OF A SUPERCRITICAL BODY OF REVOLUTION IN FREE FLIGHT AT

TRANSONIC SPEEDS AND COMPARISON WITH WIND-TUNNEL DATA

By J. W. Usry and John W. Wallace

Langley Research Center

SUMMARY

The forebody drag of a supercritical body of revolution was measured in free flight
over a Mach number range of 0.85 to 1.05 and a Reynolds number range of 11.5 × 106 to
19.4 × 106 and was compared with wind-tunnel data. The forebody drag coefficient for a
Mach number less than 0.96 was 0.111 compared with the wind-tunnel value of 0.103. A
gradual increase in the drag occurred in the Langley 8-foot transonic pressure tunnel at a
lower Mach number than in the Langley 16-foot transonic tunnel or in the free-flight test.
The sharp drag rise occurred near Mach 0.98 in free flight whereas the rise occurred near
Mach 0.99 in the Langley 16-foot transonic tunnel. The sharp rise was not as pronounced
in the Langley 8-foot transonic pressure tunnel and was probably affected by tunnel-wall
interference effects. The increase occurred more slowly and at a higher Mach number.
These results indicate that the drag measurements made in the wind tunnels near Mach 1
were significantly affected by the relative size of the model and wind tunnel.

INTRODUCTION

A free-flight test was made to provide data at high Reynolds numbers on an aero
dynamic configuration designed to fly in the sonic or near-sonic speed regime. The flight
test provided sting-free, wall-interference-free data on the transonic drag characteristics
of a low-drag body of revolution. These data along with transonic wind-tunnel data mea
sured at the same Mach numbers and Reynolds numbers are being used as part of the data
base in applications such as advanced technology aircraft. One approach to the design of
these types of aircraft is to define the aerodynamic configuration that allows the aircraft
to cruise near sonic speeds without large penalties in performance due to the drag-rise
characteristics. This type of design would require a low subsonic drag configuration with
a drag divergence Mach number near 1.
Wind-tunnel studies in the Langley 8-foot transonic pressure tunnel (TPT) have pro
duced several bodies of revolution with low drag which appear to have the desired drag
divergence characteristics. Acquisition of accurate reliable data near Mach 1 is difficult,
however, because of effects due to the tunnel boundaries. For this reason, a free-flight
test of one configuration tested in the Langley 8-foot transonic pressure tunnel and in the
Langley 16-foot transonic tunnel (TT) was made for data correlation purposes.
The purpose of this paper is to present the results of the flight test and to compare
the data with wind-tunnel results. Specifically, forebody drag and base pressure coeffi
cients are presented and compared with wind-tunnel data over a Mach number range from
0.85 to 1.05 and a Reynolds number range from 11.5 × 106 to 19.4 × 106. A comparison
of the drag-rise trends near Mach 1 determined from free-flight and wind-tunnel data is
presented and the effects due to test-environment differences are considered. Also his
tories of the flight-test environment parameters and onboard measurements are pre
sented. The test was conducted at the NASA Wallops Station, Wallops Island, Virginia.

SYMBOLS

ax accelerometer reading along body X-axis, g units

CA,b base axial drag coefficient, -Cp,b }

CA,fb forebody axial-force coefficient, CA,t - CA,b

ACA,fb incremental forebody axial-force coefficient, CA.fb - (C Atºm-0.9

Wa
CAt
A, total axial-force coefficient of model, -º
q.S

CD.f
>
skin-friction drag coefficient, based on S

prº - 9
Cp,b base pressure coefficient, +
C chord, centimeters

db diameter at base of model, centimeters

dmax maximum diameter of model, centimeters

g acceleration due to gravity, m/sec”

h altitude, meters

ha airfoil thickness, centimeters

relative humidity, percent

length of model, centimeters

Mach number

Reynolds number,

base pressure, N/m2

saturation pressure, N/m3

total pressure, N/m2

vapor pressure, N/m2

free-stream static pressure, N/m3

differential pressure, N/m2

differential pressure in pitch plane, N/m2

differential pressure in yaw plane, N/m2

free-stream dynamic pressure, N/m2

gas constant for air

body radius, centimeters

reference area based on maximum diameter of model, **k, meters 2

reference area based on diameter at base of model, id:/4, meters

temperature, K

time, seconds

earth relative velocity, m/sec

Vs local speed of sound, m/sec

Vw wind velocity, m/sec

W.o free-stream velocity, m/sec

W weight of model, 36.35 kgf or 356.47 newtons

X,Y,Z body-axis system

X,y model coordinates, centimeters

Xi any variable listed in table III

Oy angle of attack, degrees

6 angle of sideslip, degrees

1/2
m resultant angle, (~2 62)
+ , degrees

6 elevation angle, degrees

|loo free-stream viscosity, N-sec/m?

Pºo free-stream density, kg/m3

O standard deviation of variable denoted as subscript

V azimuth from true north of payload, degrees

b' = b - lºw, degrees

*w azimuth from true north of wind, degrees

RESEARCH MODEL DESCRIPTION

A sketch of the model with dimensions and other details is presented as figure 1 and
photographs of the model on the aircraft are shown as figure 2. The aircraft support rig
was designed to carry the model so that no protrusions or indentations appeared on the
surface of the model. The model was a smooth aerodynamic body of revolution. The
4
exterior shape was defined to provide low drag characteristics with drag divergence near
M = 1. The body radius as a function of longitudinal body station is presented as table I.

The length of the model was 114.30 centimeters and the maximum diameter was
12.70 centimeters. The diameter of the model at the base was 3.28 centimeters. The
model was aerodynamically stabilized during the flight by using fins which had a biconvex
airfoil section with a thickness-chord ratio of 0.03. The fins were swept 45° and the root
chord projected to the body center line was 17.20 centimeters in length. The area rule
was used to provide a smooth distribution of the longitudinal cross-sectional area in the
vicinity of the fins. Fin dimensions and a set of typical airfoil coordinates are listed in
table II. The center of gravity was located 53.46 centimeters aft of the nose tip (46.7 per
cent of the model length). Differential-pressure ports in the pitch and yaw planes were
located on the model as shown in figure 1. A total-pressure port was located at the
nose tip and a base-pressure port was located midway between two fins at a distance of
1.32 centimeters from the model center line. A transition strip was located 2.54 centi
meters aft of the nose. The strip was 0.127 centimeter wide in the stream direction and
consisted of No. 120 carborundum grit particles in a concentration of approximately
20 per centimeter. Transition strip location and grit characteristics were the same
as those used on the wind-tunnel model and were determined by using the method of
reference 1.

RESEARCH MODEL INSTRUMENTATION

An FM/FM telemeter transmitted six channels of information to ground receiving

stations. These transmissions included total pressure, two differential pressure mea
surements on the nose, two longitudinal accelerations, and base pressure. The following
table lists the parameters measured and the range of each instrument used. The expected
accuracy (10 deviation) of each measurement after data reduction to engineering units was
assumed to be +1 percent of the range of the instrument. An error analysis is presented
in the appendix. The orientation of the body-axis system for the results presented in this
paper is shown in figure 3.

Measurement º
Total pressure, kN/m3 . . . . . . . . . . . . . - 0 to 172.4
Differential pressure (two), kN/m2 . . . . . . . . 0 to 68.9
Longitudinal acceleration, g units . . . . . . . • (; to -1.00
0.00 to -0.50
Base pressure, kN/m2 . . . . . . . . . . . . . . . 0 to 103.4
 TEST ENVIRONMENT

Trajectory Parameters

The test environment was achieved by dropping the research model from an aircraft
at an altitude of 9411 meters and a velocity of 182.9 m/sec. The mission profile is shown
in figure 4 as a plot of altitude against horizontal range. The time at release from the
aircraft was taken as t = 0. Time after release and the Mach number range of interest
are indicated on the plot. These data were calculated from the best available radar track
of the model.

Histories of the altitude, velocity, flight-path angle, Mach number, dynamic pres–
sure, and Reynolds number are shown in figure 5 for the prime data period. Altitude and
flight-path angle were calculated from FPQ-6 radar positional data. Free-stream veloc
ity was obtained from FPQ-6 radar positional data, FPQ-6 radar Doppler data, and
Jimsphere wind data. Mach number, dynamic pressure, and Reynolds number were cal
culated from free-stream velocity, atmospheric speed of sound, temperature, and density.
Mach number was also calculated by using the tables of reference 2 and the ratio of the
total and atmospheric pressure. Accuracies of these parameters and the method used to
estimate the accuracy are presented in the appendix.

Atmospheric Parameters

Variations of the atmospheric temperature, pressure, density, speed of sound, and

relative humidity with altitude are presented in figure 6. The temperature, pressure, and
humidity were measured by use of a standard radiosonde balloon launched at the same
time that the model was released from the aircraft. The measurements, therefore, were
made within 25 minutes after release. The other parameters were calculated from these
measurements. In addition, the parameters from the 1966 U.S. Standard Atmosphere
Supplements (ref. 3) are shown for comparison. Accuracies of the measured quantities
are based on those in reference 4.

Variations of the atmospheric wind velocity and direction with altitude are shown in
figure 7. These data were measured by using the FPS-16 radar-Jimsphere balloon sys
tem and have a root-mean-square accuracy of 0.5 m/sec (see ref. 5). The Jimsphere
balloon was launched about 25 minutes after the model was released from the aircraft.
Therefore, these data were obtained within 50 minutes after model release. These data
were used in the calculation of free-stream velocity, Mach number, and dynamic pressure.

RESULTS AND DISCUSSION

Histories of the longitudinal acceleration, total and base pressures, and the differ
ential pressures measured onboard the model are shown in figure 8. The measurements

6
made by the two longitudinal accelerometers were essentially the same and the curve
in figure 8(a) represents both measurements. The primary data period was from
t = 22 seconds to t = 36 seconds corresponding to M = 0.85 to 1.05 and Reynolds
numbers from 11.5 × 10° to 19.4 × 10°.

Angle-of-Attack Effects

A model geometrically similar to the flight model was tested in the Langley 8-foot
transonic pressure tunnel and the Langley 16-foot transonic tunnel to determine the vari
ation of the differential pressures with angle of attack and Mach number. These varia
tions were linear with angle of attack for a given Mach number for the range of angles
and Mach numbers of interest. The slopes of these curves (differential pressure coeffi
cient per degree) plotted against Mach number are shown in figure 9. The difference in
*(Ap/a-)ho between the two sets of wind-tunnel data at M= 1 is equivalent to a dif
ference in angle of attack of 0.05% for the maximum angles obtained in flight; thus, the
faired curve of figure 9 was used with the differential-pressure data of figure 8 measured
in flight to estimate the angles of attack and sideslip.

Time histories of the angles of attack, sideslip, and resultant angle are shown in
figure 10. These data show that the model flew at a trim angle of attack which decreased
with time or increasing Mach number. As the model accelerated through M = 1, the
trim angle increased to 29 but continued the same decreasing trend after passing through
M = 1.0. (The estimated standard deviation of the resultant angle at M = 1 was 0.4°.)
The reason for the model flying at these trim angles is not apparent and cannot be deter
mined from the limited measurements made.

The variation of axial-force coefficient with angle of attack was measured in the
wind tunnels at an angle of sideslip of 0° and is shown in figure 11. These data were used
to obtain the variation of axial-force coefficient with Mach number at an angle of attack
of 09 for comparison with the free-flight test results. In addition, these data are pre
sented to show trends of C A,fb with angle of attack and to illustrate that CA,fb
decreases with increasing angle of attack for small angular deviations.

Comparison of Flight Data With Wind-Tunnel Data

The forebody axial-force and base pressure coefficients at the flight trim angles
are compared with wind-tunnel data at an angle of attack of 0° in figures 12 and 13. The
flight data were calculated as follows:

CA.fb = CA,t - (Base axial drag coefficient)

y 3.

_ _Wax, (Pb Pºp

q.S q., /S
where

and

Sb
CA, b = -Cp,b #

The 16-foot-tunnel data in figures 12 and 13 were measured at Reynolds numbers

from 15.6 x 106 at M = 0.85 to 17.04 x 10° at M = 1.01 and the 8-foot-tunnel data were
measured at a Reynolds number of 16 × 10° for all Mach numbers. The flight Reynolds
numbers varied from 11.5 x 106 at M = 0.85 to 19.4 x 10° at M = 1.05. (See fig. 5.)
The CA,fb from free-flight data (see fig. 12) for M × 0.96 was 0.111 whereas
the value measured in both wind tunnels was 0.103. The error analysis presented in the
appendix shows that one-half of this difference could be attributed to the estimated error
in CA,fb. The Reynolds number varied during the flight but was essentially constant for

both wind-tunnel tests. Also, sting-interference effects could have caused the pressure
distribution on the wind-tunnel model afterbody to be considerably different from that on
the flight model. Both of these effects could have contributed to the drag level differences
for M × 0.96.

Figure 13 shows the base pressure coefficient as a function of Mach number with
the wind-tunnel data shown for comparison. The base pressures were higher in both tun
nels for M × 0.96 and indicated that the pressures on the afterbody of the wind-tunnel
model could also have been higher. This condition would make the wind-tunnel CA.fb
lower than the free-flight values since these pressures were positive. These differences
in base pressure alone, however, represent 2.4 percent of the total drag at M = 0.85 and
become practically insignificant above M = 0.96. The sting configurations used in the
wind-tunnel tests, however, were designed to minimize sting interference effects by use
of the method of reference 6.

Differences which can be attributed to Reynolds number effects are indicated by fig
ure 14 where a theoretical estimate of the skin-friction drag coefficient is shown as a
function of Mach number. These coefficients were calculated by using the Sommer-Short
T' method and the Karman-Schoenherr flow equation for a turbulent boundary layer and
the Blasius-Sutherland equation for a laminar boundary layer. (See refs. 7 and 8.) For
this analysis the flow was assumed to be laminar forward of the transition strip and turbu
lent aft of the strip. Flight conditions were used to obtain the solid curve in figure 14
whereas the broken curve was obtained by using a constant Reynolds number and stagnation
temperature for all Mach numbers and thus simulates the wind-tunnel conditions. Both
curves at the top of the figure were calculated by assuming turbulent flow on the body
(aft of transition strip) including the fins. The broken curve in the bottom of the figure
was obtained by using flight conditions and laminar flow over the fins. These curves
indicate that the drag coefficient obtained in flight should be slightly higher than the wind
tunnel values for M × 0.96 and slightly less for M > 0.97, provided the flow was turbu
lent over the fins. The actual flow probably was mixed since local Reynolds numbers
based on fin chord varied from about 1 × 10° at the fin-body juncture to 0 at the fin tip at
the lower Mach numbers and from about 2 × 106 to 0 for the larger Mach numbers. Also,
parts of the fins were immersed in the turbulent boundary layer over the body. The
model tested in the wind tunnels was geometrically similar and should have had the same
flow characteristics as the flight model had, provided the boundary and sting interference
effects did not affect the nature of the flow. These data do indicate, however, that con
siderable differences between the flight and wind-tunnel data could be attributed to
Reynolds number and flow character effects.
A comparison of the incremental axial-force coefficients is shown in figure 15 as
ACA,fb plotted against Mach number where ACA.fb is the difference in CA,fb from
the value at M = 0.9. The gradual increase in drag obtained in each of the three tests
is probably due to the initial formation of weak shocks on the aft end of the body and the
resulting upstream pressure adjustment through the subsonic part of the boundary layer.
This gradual increase began at a lower Mach number in the 8-foot transonic pressure tun
nel than in either the free-flight test or the 16-foot transonic tunnel and was probably due
to boundary interference effects. The ACA,fb data from the free flight and 16-foot
transonic tunnel agree very well up to M = 0.98. At this point the free-flight drag data
increased sharply whereas the sharp rise in the 16-foot-tunnel data occurred near
M = 0.99. The sharp rise in the data from the 8-foot transonic pressure tunnel was not
as pronounced. The increase in drag occurred more slowly and at a higher Mach num
ber. The generally good agreement between the flight data and the 16-foot-tunnel data
and the fact that the 16-foot- and 8-foot-tunnel data were obtained under very nearly iden
tical conditions (except for tunnel size) indicate that the drag data obtained near Mach 1
On this model may be significantly affected by factors related to the relative size of the
model and the tunnel.

CONCLUDING REMARKS

Drag characteristics of a body of revolution with cruciform fins were determined

in free flight over a Mach number range of 0.85 to 1.05 and a Reynolds number range of
11.5 × 106 to 19.4 × 106 and were compared with data measured in the Langley 8-foot
transonic pressure tunnel and the Langley 16-foot transonic tunnel. The free-flight
model was dropped from an altitude of 9411 meters at an initial velocity of 182.9 m/sec.
Differences in the flight and wind-tunnel data could be attributed to the combined or
individual effects of Reynolds number differences between the flight test and the wind
tunnel tests, flow quality conditions, boundary effects, and instrumentation differences.
Further wind-tunnel studies and experimental tests are needed to determine the effects
of each of these parameters on the drag.
The axial-force coefficient for a Mach number less than 0.96 was 0.111 in free
flight compared with 0.103 for the wind tunnel. A gradual increasing trend in the drag
occurred in the Langley 8-foot transonic pressure tunnel at a lower Mach number than
in the Langley 16-foot transonic tunnel or in the free-flight test. The sharp drag rise
occurred near Mach 0.98 in free-flight data and near Mach 0.99 in the Langley 16-foot
transonic tunnel data. The sharp rise was not as pronounced in the data from the Langley
8-foot transonic pressure tunnel. The increase in drag occurred more slowly and at a
higher Mach number. These results indicate that the drag measurements made in the
wind tunnels near Mach 1 were significantly affected by the relative size of the model
and wind tunnel.

Langley Research Center,

National Aeronautics and Space Administration
Hampton, Va., October 27, 1971.

10
 APPENDIX

ERROR ANALYSIS OF FREE-FLIGHT DATA

The root-sum-square (RSS) method for combining errors from independent sources
was used to estimate accuracies of the free-flight data. The major parameters of inter
est are the forebody axial-force coefficient CA.fb, Mach number M, and Reynolds num
ber NRe. The assumption was made that the individual error source contributions to
the parameters are statistically independent and normally distributed with mean zero.
The RSS method may be expressed as follows:
1/2
gy = Gº + . . . -- (; º + . . . -- (; º (A1)

where

Gy one standard deviation in y = f(xi)

9xi one standard deviation in independent variables xi (i = 1, 2, 3, . . ., n)

# partial derivative of y =f (xi) with respect to xi

The study performed here is concerned with the accuracy of axial-force coefficient, Mach
number, and Reynolds number.
The forebody axial-force coefficient is a function of several independent variables
and may be expressed as follows:

2RETWax-(po-p-)s.
A, = -
CA.fb 2

4. whºse
- 18. + 3.537 X loºſa 459.67) -
2.245
}*.e. tly – ºw)cos d
(A2)

By applying the RSS method (eq. (A1)) to equation (A2), the standard deviation of CA.fb
can be estimated as follows:

2 2 2

°CA,fb -

- (:: º (# º
9CA
+
8CA
+ . . . -- (* w)
6CA
(A3)

11
 APPENDIX – Concluded

Histories of the standard deviations were calculated by using the chain-rule method
illustrated in figure 16. Values of the standard deviations used in these calculations are
listed in table III. The expected accuracy (10 deviation) of each onboard measurement
after data reduction to engineering units was assumed to be +1 percent of the range of the
instrument.

Standard deviations of the wind data ſo.

Vw and 0
Ww and atmospheric data (orT,
°hr, and op ) were obtained from references 4 and 5.
oo
Standard deviations for radar

data (ºv, 9, and O. 0) were obtained from a radar study conducted at Wallops Island,
Virginia.

Table IV presents the values and the estimated standard deviations of CA,fb) M,
and NRe at M = 1. The values and estimated standard deviations for the intermediate
variables in these estimations are also listed. The standard deviation of CA,fb is

about 2 percent of the calculated value at M = 1.00. The standard deviation is less than
0.2 percent for M and less than 0.4 percent for NRe.
Figure 17 shows the variation of the error contribution of four of the basic vari
ables and GCA,fb with M. The four variables make the largest error contributions
to °CA.fb and the error contribution of the ax measurement is an order of magnitude
y

larger than the other three error sources and clearly illustrates the need to improve the
accuracy of the complete data acquisition system.
Figure 18 shows the variation of CA.fb and CA.fb + ocA.fb
> y 2
with M. Also

indicated in the figure is M + GM at M = 0.97. The error in CA,fb at M = 0.80

is 8 percent of CA,fb and decreases with increasing M. At M = 1, the error is

2 percent.

12
 REFERENCES

1. Braslow, Albert L.; and Knox, Eugene C.: Simplified Method for Determination of
Critical Height of Distributed Roughness Particles for Boundary-Layer Transition
at Mach Numbers From 0 to 5. NACA TN 4363, 1958.
2. Anon.: Equations, Tables, and Charts for Compressible Flow. NACA Rep. 1135,
1953.

3. Anon.: U.S. Standard Atmosphere Supplements, 1966. Environ. Sci. Serv. Admin.,
NASA, and U.S. Air Force.
4. Anon.: Environmental-Measuring Equipment Used by Air Weather Service in Support
of Air Force and Army Operations. AWS Pamphlet 105-3, U.S. Air Weather Ser
vice, Nov. 10, 1967. (Available from DDC as AD 822 895.)
5. Scoggins, James R.: Sphere Behavior and the Measurement of Wind Profiles. NASA
TN D-3994, 1967.
6. Cahn, Maurice S.: An Experimental Investigation of Sting-Support Effects on Drag
and a Comparison With Jet Effects at Transonic Speeds. NACA Rep. 1353, 1958.
(Supersedes NACA RM L56 F18a.)
7. Sommer, Simon C.; and Short, Barbara J.: Free-Flight Measurements of Turbulent
Boundary-Layer Skin Friction in the Presence of Severe Aerodynamic Heating at
Mach Numbers From 2.8 to 7.0. NACA TN 3391, 1955.
8. Chapman, Dean R.; and Rubesin, Morris W.: Temperature and Velocity Profiles in
the Compressible Laminar Boundary Layer With Arbitrary Distribution of Surface
Temperature. J. Aeronaut. Sci., vol. 16, no. 9, Sept. 1949, pp. 547–565.

13
 TABLE I. - COORDINATES OF BODY

/
—e-f -

=-
H.
x, cm r, cm X, cm r, cm x, cm r, cm

0.0 0.0 22.86 5.30 78.74 6.00

.25 .83 25.40 5.47 81.28 5.91
.51 1.14 27.94 5.62 83.82 5.80
.76 1.37 30.48 5.75 86.36 5.68
1.02 1.55 33.02 5.86 88.90 5.54
1.27 1.71 35.56 5.96 91.44 5.38
1.52 1.85 38.10 6.05 93.98 5.20
1.78 1.98 40.64 6. 12 96.52 5.00
2.03 2.10 43.18 6. 19 99.06 4.77
2.29 2.20 45.72 6.24 100.33 4.64
2.54 2.30 48.26 6.28 101.60 4.50
3.81 2.73 50.80 6.31 102.87 4.33
5.08 3.07 53.34 6.34 104. 14 4.17
6.35 3.35 55.88 6.35 105.41 3.86
7.62 3.60 58.42 6.35 106.68 3.56
8.89 3.82 60.96 6.34 107.95 3.23
10. 16 4.02 63.50 6.33 109.22 2.85
11.43 4.20 66.04 6.30 110.49 2.47
12.70 4.36 68.58 6.26 111.76 2.09
15.24 4.65 71. 12 6.21 113.03 1.77
17.78 4.90 73.66 6. 16 113.66 1.67
20.32 5.11 76.20 6.09 114.30 1.64

14
 TABLE II. - FIN DIMENSIONS

the coordinates of any airfoil section in the chordwise direction are given by

where a - oo:

— c = 17.20

Typical coordinates of air foil section, c = 12.70 cm

X, cm y, cm x, cm y, cm x, cm y, cm x, cm y, cm

0.0000 0.0000 3.3020 0.1466 6.6040 0.1902 9.9060 0.1308

.2540 .0149 3.5560 . 1537 6.8580 . 1892 10. 1600 . 1219
.5080 .0292 3.8100 . 1600 7.1120 .1877 10.4140 . 1125
.7620 .0429 4.0640 . 1659 7.3660 .1857 10.6680 . 1024
1.0160 .0561 4.3180 . 1709 7.6200 . 1829 10.9220 .0917
1.2700 .0686 4.5720 . 1755 7.8740 .1796 11. 1760 .0805
1.5240 .0805 4.8260 . 1796 8.1280 . 1755 11.4300 .0686
1.7780 .0917 5.0800 . 1829 8.3820 . 1709 11.6840 .0561
2.0320 . . 1024 5.3340 .1857 8.6360 . 1659 11.9380 .0429
2.2860 . 1125 5.5880 . 1877 8.8900 . 1600 12. 1920 .0292
2.5400 . 1219 5.8420 . 1892 9. 1440 . 1537 12.4460 .0149
2.7940 . 1308 6.0960 . 1902 9.3980 . 1466 12.7000 .0000
3.0480 . 1389 6.3500 . 1905 9.6520 . 1389
TABLE III. - ASSUMED VALUES OF STANDARD DEVIATIONS (10) FOR CALCULATIONS

gax, g . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0.005
°pp. kN/m" . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . - - - - - - - - 1.034

9p,” kN/m" . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.724

t
9Wu,” m/sec . . . . . . . . . . . . . . . . . . . . - - - - - - - - - - - - - - - - - - 0.5000

°il/w’ rad . . . . . . . . . . . . . . . - - - - - - - - - - - - - - - - - - - - - - - 0.01745

ov, m/sec . . . . . . - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 0.3048

y, “
G.I. rad . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . e 0.00015

ag, rad - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 0.00015

GT, K - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 0.70
°hr for —

273.15 K P T = 233.15 K . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0.033

T = 273.15 K . . . . . . . . . . . . . . . - - - - - - - - - - - - - - - - - - - - - 0.017

°p. for —

h “ 3048 m . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0.023
3048 m × h : 6096 m . . . . . . . . - - - - - - - - - - - - - - - - - - - - . . . . 0.033
6096 m × h & 9144 m . . . . . . . . . . . - - - - - - - - - - - - - - - - - - - - - 0.040

16
 TABLE IV.- CALCULATED STANDARD DEVIATIONS FOR M= 1

Parameter Value Standard deviation

CA,fb . . . . . . . . . . . . . . . . 0.191 0.004

M . . . . . . . . . . . . . . . . . . 1.000 0.002

NRe . . . . . . . . . . . . . . . . . 16.29 × 106 0.06 × 106

ps, kN/m3 . . . . . . . . . . . . . 0.21 0.01

p, kN/m" . . . . . . . . . . . . . 0.06 0.01

l', rad . . . . . . . . . . . . . . . 0.33 0.02

p_, kg/m" . . . . . . . . . . . . . 0.729 0.002

V., m/s . . . . . . . . . . . . . . 325.2 0.4

q., kN/m" . . . . . . . . . . . . . 38.6 0.1

CA,t . . . . . . . . . . . . . . . . 0.181 0.004

CA,b . . . . . . . . . . . . . . . . 0.011 0.002

Vs, m/s . . . . . . . . . . . . . . 325.2 0.4

u..., N-sec/m” . . . . . . . . . . . 1.664 x 10-8 0.004 x 10-8

17
&

iconvex
foil
air
B
section,
thick
3%

ports
5 ressure
strip
PTransition
-I
| 3.28-de
|

34.39
".
port
pressure
Base

1114.30

model.
research
1.-
of
Sketch
Figure centimeters.
in
are
Dimensions
L-70–140

Photographs
2.-
Figure
aircraft.
on
model
of

view.
quarter
Front
(a)

3
'pºpn ſouoo - "Z arnºſ
“Aaa! A requenb reºſ (q)

20
 21
'uoņequaļJo tuæ)sÁs-sțxe-Apog - ºg arnº! ¡

W
 Time, see

8 9 10 1 1 12 13 ll! 15

Horizontal range, k"

Figure 4.- Variation of altitude with horizontal range based on radar data.

22
 |- 360 – 8

-
Velocity
-66 H 7
Altitude Flight-path angle

–6]+

– 5 #
º: H
-x

§3

| -60

–
260

240 H
–
--

º
2
lº

3 |

- 220 - l ---

– 200 – 0 –
22 2|+ 26 28 30 32 34 38

Time, see

(a) Time histories of altitude, velocity, and flight-path angle.

º
2 Figure 5. - Time histories of test conditions.

23
 60 2OX 10
Reynolds number

56 19

52 18 Dynamic pressure

1.08 r- lº& 17

Mach number

1.Ok lil: 16

ov

1.00 l:0 15

: .96 H

.92 |
36

32
| 14

, 88 H 28 12

.81; H 21+ 11

.80 – 20
22 2|+ 26 28 30 32 3! 36 38

Time, sec

(b) Time histories of Mach number, dynamic pressure, and Reynolds number.
Figure 5. - Concluded.

24
 #EEE
Radio son de Radio son de
— — — 1966 Standard — — — 1966 Standard

i
22O 230 240 250 260 270 280 30 l:0 50 60 70 8O 90

Temperature, ok Pressure, kN/.”

(a) Temperature. (b) Pressure.

Radio son de – — Radio son de

— — — 1966 Standard — — — 1966 Standard

i
2 -

.* .5 .6 .7 .8 .9 1 -O 3OO 310 320 330 340 35O 360

Density, ka/* Speed of sound, s/sec

(c) Density. (d) Speed of sound.

;
2 10 20 30 l{O 50 60 70

Relative humidity, per cent

(e) Relative humidity.

Figure 6.- Atmospheric properties as a function of altitude.

25
 ;

“o 2 * 6 8 10 12 14 16 18 20 22 2+
t ! I 1 l l
* ind velocity, */s •c
240 260 280 300 320 340

wind direction, deg

Figure 7. - Variation of wind velocity and direction with altitude.

26
 23 2- 25 26 27 28 29 30 31 32 33 3* 35 36
Wi-e, -ec

(a) Longitudinal acceleration.

22 23 2- 25 26 27 28 29 30 31 32 33 3h 35 36
Uime, sec

(b) Total and base pressures.

Time, see

(c) Differential pressures in pitch and yaw planes.

Figure 8.- Time histories of data from onboard instruments.

27
 .08

: D 8-Fit TPT

.07
# 2, 16-ft. II. H

:i .06

.05 H

.80 .84 - .88 .92 .96 1.00 1.0l.

Figure 9.- Variation of differential pressure coefficient per degree angular

displacement with Mach number.

28
 36

35

34

angle.
resultant
and
sideslip,
attack,
angles
of
histories
Time
10.-
Figure
33
32
31
30

sec
Time,

28
24
27
26
25
29

;
§
 8-Fit TPTH
16-Ft TTH H
Faired H H
.*H

.20;

**, r*.16 H

...H

• 10 - It -

Figure 11.- Variation of forebody axial-force coefficient with angle of attack from the
Langley 8-foot transonic pressure tunnel and the Langley 16-foot transonic tunnel.

30
 .28

O **, r*(free flight)|

.2, H P --- **, r*(8-ft (PT)
H C –– º, reſić-FT II)
*a,b (free flight) H

• 20

. 16 H

• 12 H -

; . O8 H
-

.0l. H

--0.5, .8l4 .88 .92 .96 1.00 1.04

Figure 12. - Variation of forebody axial-force and base drag coefficients

with Mach number.

31
 HD---8-Fit TPT
HC –-16–Ft TT

.80 .84 .88 .92 .96 1.00 1.0+ 1.08

M

Figure 13.- Variation of base pressure coefficient with Mach number.

32
 ... l l

. . 10 E Turbulent ==EH ==Free-flight case :

F fins --

H.--Fi
Wind-tunnel case

.08

.07 EHEE -

.80 .8l4 .88 .92 .96 1.00 1.0l. 1. O8

Figure 14.- Variation of theoretical skin-friction drag coefficient with Mach number.

33
 • 12

o–free flight
[]---8-Fit TPT

• 10
Ö--16–Ft TT

.08

A9A, fb.06

. Olt

.02

- .94 .98 1.02 1.06

Figure 15.- Comparison of incremental axial-force trends for data

from tests in the Langley 8-foot transonic pressure tunnel, the
Langley 16-foot transonic tunnel, and in free flight.

34
 ["I
|–
"a “O
H
COOp
ſſ
CO
sOſ
"p,
yG
V
G number
Reynolds

coef icieMach
Drag nt
number M
0
W
l
*A,
º
"co
fb
A,CO

sº
W
°
T
o|g F---
––
W

Falclow
-
16.
cFigure
for
chart
ulations.

x
Oſi
Pb
Oſ
Y
-—
Oſ

PvOſ

Poo Pb

Oſ
f

Goo
*

Oſ

3.
 . OO7

.006 H

.005 ||

.OOh H

• OO2 H :::::::::::::::::

.001

.8l4 .88 .92 .96 1.00 1.Olt 1.08

Figure 17. – Variation with Mach number of contribution of variables to GCA’

36
 .80 H; .92 .96 1.00 1 - 1.08

Figure 18.- Variation of drag coeff icient and standard deviation with Mach number.

NASA-Langley, 1971 1
L-8019
37
- - - -
NAT I ONAL AERO NAUTICS AND SPACE A D M ISTRAT I ON
WASHINGTON, D.C. 20546
Post AGE AND FEEs PAID
NATIONAL AERONA uTics and
OFFICIAL BUSINESS
space Ad Minist RATION
FIRST CLASS MAIL
PENALTY FOR PRIvaTE use $300

o 12 002 c 1 II 35 71 1203 S00 ly 89 H

UNT v of ILLINOIS
LIBRARY
ATTN: Docu MENT LIBRARIAN
U P B A N A IL 6 1801

POSTMASTER:. . If Undeliverable
Postal Manual) Do(Section 15
Not Retu

"The aeronautical and ſpace activities of the United States ſhall be

conducted ſo aſ to contribute . . . to the expanſion of human knowl
edge of phenomena in the atmoſphere and ſpace. The Adminiſtration
ſhall provide for the wideſt practicable and appropriate diffemination
of information concerning its activitieſ and the reſults thereof.”
— NATIONAL AERONAUTICS AND SPACE ACT OF 1958

NASA SCIENTIFIC AND TECHNICAL PUBLICATIONS

TECHNICAL REPORTS: Scientific and TECHNICAL TRANSLATIONS: Information

technical information considered important, published in a foreign language considered
complete, and a lasting contribution to existing to merit NASA distribution in English.
knowledge.
SPECIAL PUBLICATIONS: Information
TECHNICAL NOTES: Information less broad derived from or of value to NASA activities.
in scope but nevertheless of importance as a Publications include conference proceedings,
contribution to existing knowledge. monographs, data compilations, handbooks,
sourcebooks, and special bibliographies.
TECHNICAL MEMORANDUMS:
Information receiving limited distribution TECHNOLOGY UTILIZATION
because of preliminary data, security classifica PUBLICATIONS: Information on technology
tion, or other reasons. used by NASA that may be of particular
interest in commercial and other non-aerospace
CONTRACTOR REPORTS: Scientific and
applications. Publications include Tech Briefs,
technical information generated under a NASA Technology Utilization Reports and
contract or grant and considered an important Technology Surveys.
contribution to existing knowledge.

Details on the availability of these publications may be obtained from:

scieNTIFIC AND TECHNICAL INFORMATION office

NATIONAL AERO NAUTICS AND SPACE ADMINISTRATION
Washington, D.C. 20546 -
 \
• →-- ~~~~ ·
 UNIVERSITY OF ILLINDIs-URBANA
629.13UN325in C DD1
TECHNICALNOTE:WASHINGTON
658 -

lſ|ill
T|i| 008547058