DYNROT: A Matlab Toolbox For Rotordynamics Analysis: January 1994
DYNROT: A Matlab Toolbox For Rotordynamics Analysis: January 1994
net/publication/236592072
CITATIONS READS
3 2,949
2 authors:
Some of the authors of this publication are also working on these related projects:
Actually, I am a Consultant of the European Community • Deputy Chairman of the Technical Committee TC3 of the IMEKO (Force, Mass and Couples and italian member
of TC-17 for Robotics. View project
All content following this page was uploaded by Giancarlo Genta on 28 July 2015.
Joint Editors:
D.R.J.Owen
E. Hinton
-
-
Volume 13 Number 6 1996 . -
_ _
•4
11~
MCB 1/1"
University Prtss
---' -,--•••: ,--f•-•-• ,-.!•••-•/,••••, •;• ,.......• • •p-rz,.•rf• ■ •• ••• -t/..■••:•••,••••2,.• ■••:•--...”
. •-•[/•• •••••••• e...•••■n• rt;•••••■■••.,•, ■-•.• •,•• •
`-'' • •
• , 7•,.,,, ,?2,)rx.--7,-,.7.1-.7mT.Tft:4,7.7:4"?'..775`,. ,' 3'.1.5r7,17:..;‘,Z-;9nYZM'').272.
..7.7"-',7' ,Zit4'"?-7
- - • • ,kt, - • ,
-
• .T4
- - .•
o
......., ,...:•., ,, , t ,:.'-,:ti. - • • ,........ , -.. i. , , i , •:.:.,.. ,.:-..: •:-..:..,:...,:;,-., ,,,,,,,, ;...•,,:. •.,..,:. 7. , • ..,.,:-,-,,,-',, ,-- ..-,• ■ ,,i.:■ •••-•. ,, ,,, s
. ---,:- r.,-..72-: -i-- ,
.. 7—.-.--.."-lé•g7---
•', ,,,, ...., ".'
27, -7.-, - 4.*:". '
o,-
...'71
• ■-• tAr'f•o
't
o
EC DYNROT
13,6
A finite element code for rotordynamic
analysis based on complex
86 co-ordinates
Received May 1995 G. Genta, C. Delprete
Politecnico di Torino, Torino, Italy, and
D. Bassani
Merlo SIA, Cuneo, Italy
Introduction
The finite element method (FEM) is gaining popularity also in the field of rotor
dynamics, mainly for its ability of modelling intricate geometries in a simple
way, at least from the point of view of the user, and its possibility of preparing
general-purpose codes. Codes based on methods which were quite common in
the past, like the transfer matrix method, are now gradually substituted by
codes based on the FEM.
• The use of standard commerciai codes for structural dynamic analysis also
for the dynamic study of rotors is sometimes possible, but compels to resort to
some sort of "trick" to take into account the effects of the rotation of the system
which can have a strong influence on its flexural vibration. It can affect the
natural frequencies and couple the flexural motions in such a way that it is
more correct to speak of "whirling" motion than of vibrations.
To take into account this instance it is possible to force a "gyroscopic"
matrix, which affects mainly the mass matrix of the system, into any code
devised for dynamic analysis; but this is not a handy procedure and has the
disadvantage of allowing only the study of synchronous whirling. It is then only
possible to compute the criticai speeds and the unbalance response but not that
of the Campbell diagram (i.e. the plot of the whirl natural frequencies against
the spin speed), which is the basic tool for understanding the dynamic
behaviour of any rotor. Even more problematic is the study of the behaviour of
the rotor during an acceleration.
Only a purposely written code, which correctly takes into account the
presence of the "gyroscopic matrix" and perhaps of the "centrifugai stiffening
effect" can adequately fulfil the task.
Starting from the end of the 1970s, the development of a FEM code
specifically suited for rotor dynamics' computations was undertaken at the
Department of Mechanics of Torino Politecnico. It has been evolved through
various versions over 15 years. The present version is based ori the MATLAB
Engineering Computations, Vol. 33
No 6 1996, pp. 86309. C MCB
(MATLAB is a trademark of The MathWorks, Inc.) interactive software
- Univesity Press, 0264-4401 package and consequently can be used on any hardware on which MATLAB
:',ner.n.”'"..,•,:- —,,,,,I,,,C
,
,„.`i.•.•
• -..r.
S.,.
.V....: ''.,
.... rre
el.
....,'
.• /7., .
.
., ‘1. h• ì •
has been installed. Although most of the code can be used on any PC/AT DYNROT: a
machine, to deal with more complex problems in a reasonable time at least a finite element
80486/class machine is needed; stili larger problems can be dealt with using a
mainframe-based version of MATL.kB. code
The orig-inal code was written using HPL and then HP-BASIC language for
desktop HP 9800 computers. In those versions the rotor geometry was
considered axially symmetrical and - linear and the presence of hydrodynamic 87
bearings was not considered. Subsequent versions written in Fortran and C
languages were developed in the 1980s and finally the present version using
MATLAB package evolved. The originai assumptions were gradually released,
adding the possibilities of studying nonlinear rotors, in vv -hich either the stator
or the rotor or even both are unsymmetrical, while elements aimed to model
hydrodynamic or magnetic beatings were added. Routines for the study of the
torsional behaviour of generai rotors and reciprocating machines were also
added.
A comprehensive key to the symbols used in this article is given in the
Appendix.
Theoretical background « -
Generai considerations
Usually rotors are modelled as beam-like structures. This is the traditional
approach based on some form of "beam theory", with either the Euler-Bernoulli
or the Timoshenko approach[1,2] and the numerical approach based on some
form of the Myklestadt-Prohl transfer matrix method[3,4]. Most of the proce-
dures based on the FEM consider the structural parts as composed by beam
elements[5,6]; however in the literature it is possible to find models based on
other types of elements, as in [7]. Also for the study of the torsional behaviour
of shafts the structural parts are mainly modelled as beams.
Under the above-mentioned assumptions, if the beams which model the rotor
are straight, their axes are ali aligned with the spin axis and if the centre of
gravity and the shear centre of all the cross-sections lie on the spin axis, the
axial, flexural and torsional behaviours are uncoupled. The presence of a small
static or couple unbalance does not modify this feature substantially.
Clearly the assumptions leading to uncoupling do not hold in the case of
crankshafts, but in the study of the torsional vibrations of reciprocating
machines it is customary to resort to a so-called equivalent model which is
essentially a straight beam-like structure and, as a consequence, the uncoupling
is restored[8].
Flexural behaviour
When the system is completely axially symmetrical, the best choice for what
the generalized co-ordinates are concerned with is the use of complex co-
ordinates. As the structure is assumed to be beam-like, each node has four real
degrees of freedom, namely two lateral displacements and two rotations.
Assuming that the axis of rotation of the system coincides with the z-axis of an
e «
• •
. •-•
- • • o• • o:. •-;- •
EC orthogonal reference system xyz and using the assumptions which are
13,6 customary in the beam theory, the flexural displacement of any point of the
rotor axis and the rotation of the cross-sections can be expressed by the
complex displacement
Z =X+
(1)
88 -.
it is possible to demonstrate that this approach is suitable also when the rotor
or the stator lack axial symmetry[9]. -
The equation for the motion of a nonlinear rotor which is performing an
acceleration with a sta ted law co (t) is [10]:
[M],„{fl + ([C],.„ — co[G]){q} + ([K]„, — ico[C,],n ){q} + [Mn ]dW
+[M,Jd e2:0 + [CnId {:d.} + ([Cy ]d + 2iCOEMAd )e22.(9 {i} (2)
+[K]d {} + ([1C,]d + ico[Cr]4e2'e + {g(q i3 O ,t)}
= (w2 — ia){F,}e 78 +
where: -
• [M], [G], [C] and [K] are respectively the mass, gyroscopic, viscous
damping and stiffness matrices. They are all symmetrical.
• Subscripts m, d, r, n refer respectively to the mean and deviatoric
matrices and to the matrices referred to the rotating and the non-rotating
parts of the machine. Mean matrices without either subscript r or ?I refer
to the whole model.
• {q} and {F} are respectively the vector of the complex generalized co-
ordinates and that of the nodal forces. In this case subscript r designates
forces due to unbalance and subscript n the non-rotating forces as rotor
weight. All forces can be prescribed functions of time.
• The generic vector function {g(q 4•, e, t)} is introduced to take into
account the behaviour of the nonlinear part of the system.
• O, w = é and a = ci) are respectively the rotation about the spin axis, the
spin speed and the angular acceleration. They are all known functions of
time and are the same for the whole rotor; however, with simple
modifications to equation (2) to study the behaviour of multi-shaft
systems, the model can be subdivided in different substructures each one
having a different angular velocity and hence different laws 6(t), w(t) and
a(i).
The assumption that the angular acceleration is the same at all sections comes
directly from the uncoupling between the torsional and the flexural behaviour
and is usually referred to as "torsionally stiff rotor assumption". However, an
,;.- •.: ,.-,-:, .,.:_.. . '.',, '.-- ..', ",'.•.,; ..•.`. .:.,.;. *: t., e.., ...:, ..,..,-,: ‹,7;•,..";,,.."-,?,'..',':.._.7,-,,.;?..,-{2..n,-.‘,.,,,. .-,.' !.•,:tz,-;•.•:, 't.";' .-- -.-,.---..' , ,
■
«' :' ,. ',.'., -',-.....:......
--.
..- ',...' S
. -,
f
t
'.-.. ì ' ,.1
. 1,-;;
. !
...., . . - •
... , ,
equation allowing to link the rotational degree of freedom of the system with DYNROT: a
the driving torque can be easily obtained. It is a single scalar equation, owing to finite element
the assumption of torsionally rigid rotor, except in the case of multi-shaft code
systems:
M z = 9 ({F} T (3)
where the total polar moment of inertia of the rotor can take also into account
89
the effect of the different unbalances which are at any rate very small:
n
blot 2{n, X:2117n, — jd i cos(2fixi )]} (4)
Equation (2) is the basic formulation for the study of the flexural behaviour of
the system. Clearly its complete formulation cannot be solved in closed form
and the only possible approach is the numerical integration in time. Various
levels of simplification can be used in order to solve the problems which are
common in rotor dynamics: by setting the angular acceleration to zero it is
possible to study the behaviour of the rotor at constant speed; neglecting
deviatoric matrices it is possible to study the behaviour of isotropic systems;
neglecting the nonlinear parts the behaviour of linear rotors is dealt with and so
on. Obviously many of the simplified problems allow a closed form solution: for
instance the study of the critical speeds, of the Campbell diagram and of the
stability is performed through an eigenanalysis, as they are essentially linear
problems.
In equation (2) damping has been assumed to be of the viscous ty-pe, as the
time-domain equation cannot accommodate damping if it has been assumed to
be of the hysteretic type, at least in some parts of the system. However, this is
no more true when such solutions as circular whirling or harmonic vibration
are assumed. As a consequence, the model can be built using both forms of
damping: viscous damping being introduced in the form of damper elements
and hysteretic damping by stating the loss factor of the materials. When
hysteretic damping can be used directly, the equation of motion is accordingly
modified to allow its use[8]; when on the contrary this is impossible the user has
two alternatives: neilect hysteretic damping altogether or resorting to a form of
equivalent viscous damping. A way to compute the latter is by performing the
modal transformation based on the modes of the linearized, natura], isotropic
system (i.e. based only on the mass and stiffness mean matrices) of all hysteretic
damping matrices and reducing them to their generalized proportional
component[8] by cancelling all elements outside the main diagonal. The
equivalent viscous damping matrices can then be obtained by dividing the
elements on the main diagonal by the corresponding natural frequency and
then performing the inverse modal transformation. This procedure is clearly
approximated but, if the system is lightly damped, leads to acceptable results.
As there is no theoretical way to assess its adequacy, a check can be done by
comparing the steady-state unbalance responses of the models with hysteretic
•
• - , , • . . .
• -•
- '
'••• • , • • , . • • I .. • •, .• •
■
'•••••/• .,••'''', • ' • • -
•
'
EC and with viscous equivalent damping. Note that the equivalent damping so
13,6 computed is just a mathematical approximation and the structure of the
equivalent damping matrices has nothing to do with that of the starting
hysteretic matrices, nor does it retain any physical meaning.
In the study of the unbalance response, both at constant speed or during an
acceleration, after computation of the flexural behaviour equation (3) can be
90 used to compute the driving torque needed to maintain the spin-speed constant
or to follow the stated law w (t).
The main advantages of the use of complex co-ordinates can be summarized
as follows:
In the case of isotropic systems, the equations are all real and their
number is halved with respect to the approach based on real co-
ordinates. A substantial reduction of computation time is so achieved.
When the system is non-isotropic, elliptical or polyharmonic whirling is
expressed as the sum of forward and backward circular whirling
components. When a more complicated whirl pattern exists, the various
harmonic components are well separated and this allows a good insight
to the motion of the system in a simpler way. _
The model is built in terms of "mean" and "deviatoric" properties.
Different solutions can then be obtained one after the other: by
neglecting the deviatoric matrices the dynamic behaviour of a "mean"
systems — an equivalent symmetric system which retains many of the
features of the originai one — is obtained. Later, more realistic
simulations which take into account deviatoric matrices can be obtained
without building new models. The possibility of refining the solution
without the need of remodelling the system is a very interesting feature
of this approach.
where:
:sij ,:•.•
•
[M], [C] and [K] are respectively the mass, viscous damping and stiffness DYNROT: a
matrices. They are all symmetrical. finite element
• {g} and {Fi} (the latter only for the case of torsional dynamics) are code
respectively the vector of the complex generalized co-ordinates and that
of the jth nodal force. A number k of harmonic forcing functions, each
one with its frequency all multiples of a fundamental frequency of the —
excitation Al can be introduced into the model.
The study of the dynamic behaviour of reciprocating machines is performed
using the "equivalent system" approach described in detail in[8]. The so-called
inertia torques are directly introduced into the model, together with the forcing
functions defined by the user. They are directly computed through a numerical
FFT of functionsfi (e) andf2 (8) (see [8D instead of using the series expansions
found in the literature.
Alternatively, the variation in time of the moment of inertia of the cranck
system can be accounted for sowing, through a truncated form of Hill's infinite
determinant, the differential equation with periodic coefficients. A stability
study of the system can thus be performed.
Also here hysteretic damping can be directly introduced into the model only
in some of the cases. When this is impossible, an equivalent viscous damping
matrix can be computed following the procedure for flexural behaviour.
Use of modal transformation
DYNROT code works for most of the computations using directly physical
generalized co-ordinates. In particular, the torsional response to a polyharmonic
forcing function does not rely on any modal approximation, as the usual
procedure, which takes into account only the response related to the first or the
few first modes, can lead to unacceptable approximations.
However, there are cases in which the modal approach a]lows strong
reduction in computation time and consequently is worthwhile considering. In
the flexural behaviour two different modal approaches are possible: the first is
that of using as the transformation matrix the matrix of the eigenvectors of the
natura!, undamped, linearized, isotropic system as already seen.for the
computation of the equivalent viscous damping[11]. A second approach,
suitable for the study of the unbalanced response, is that of using the
eigenvectors of a system whose mass matrix is equal to (WHG1J[12]. In this
case the only difference from the usual modal uncoupling practice is that the
"mass matrix" can be non-positively defined and correspondingly some
imaginary critical speeds can occur. It can be shown that this does not introduce
any conceptual or computational difficulty and does not prevent the resort to
modal uncoupling provided that the modes with imaginary eigenfrequency are
also considered.
The modal approach (first of the above-mentioned procedures) is
compulsorily used for the computation of the acceleration response. To avoid
high frequency modes which would cause problems in the numerica!
• •••
• '--.- -. • l ' ',. ' ' -. • • ,.. • • ,!.-...01,:,,iti...-.;•:-. -• ..e..,,,,•••••A •?:-.
.;77:••••••"5-97..•
•• .' • .. ,' •••• .-".-...; ,>.•'
'':"'-;:'--:- , -:•,•,..-,s3.- ''' • :•'...■-, . .
:"."• ." '..:: -
; , 3: .g... ,, '
' - . ' ,•• • •• • .-.•.:' "..t.•••• .., 2 • ,,:.,:,-,vt;,i,;:-..,, ' • ',....:•••' •'; •;?> '"• • '. .."7.•;•.'i ' .:,•:::,`,...,..",2
,,,, ,-'-..- -7". . :; i-•";;',V,.. ‘■'<• , , • ■•-. . ' ..., , ..,,.:••••"-•
-•' -.‘, -•,, ' • ' •-'. • ,..3- :›sr.:.■1'•'••::•'•7‘,;••5;;,•:...3'
- -•,.. • ' .,..,•1'.••;•,:t...., • • - • ,. .0..., i,',=. i.".:•%•i•
,re- .29
•:1, ':.'••••••••;.-:•`.: ::..',:- • _ •.> •, .- :-.:-,;,....:,,•••••..,•••-,.....
.;•'•,•-•',••;-•.-.••9••';•,p;,.;,+''.-:?;',1.14,51/4 k.:',;i:%,...‘.:P.,,,;•.' 11 •-,2,,1`,, . e ..,,:,- t
+•!1•4,!).:1S;• ■ •••5."`;•,,•
., • ,:--••,- • •,,,•.,.. ,,,. ziii•'
■ ti-,•-:;•9!•.`..;•.?"•.0,g•:•,:‘,!,....,•:;.::,•.r.,
.,. . .. • , .. , . .. .
• - •., • •
_ EC integration procedure, only the modes with natural frequency not higher than
13,6 the maximum spin speed are considered.
Another case in which the modal approach is used is the study of rotors
including magnetic bearings. Here the user can decide the number of modes to
be considered, it being possible to chose all modes if the approximation due to
the modal approach is considered unacceptable.
92 In other cz.ises it is possible to resort optionally to-the modal approach.
Gyroscopic matrices and, in the case of darnped systemg, damping matrices
also can be easily subdivided into a "generalized proportional" and a "non-
proportional" part and the coupling betweén modes can be dealt with using a
simple, fast-converging, iterative technìque. _
; - •":
.
„
- •
• ...'...j5...›';;),••••.111.,
, • . ,
'.•••••• ';‘••
• . • ••
i • '-,••='"r''''..;','-'•'••••••• ••• •.
••.• r • " • ••
, . .
•
the various working cycles in the case of torsional behaviour of reciprocating DYNROT: a
machines, the data regarding the working cycles themselves and the harmonic finite element
components of the driving torque and the time histories of the spin speed and code
the unbalance used in the time integration of the equations of motion. The last
matrix contains the information needed to perform the Guyan reduction, i.e. the
status (master, slave or constrained) of each degree of freedom.
Once-the data file has been prepared, the user can choose to run the code in _ - 93
an interactive way, which is best suited when an installation on PC-type
machines is considered either in DOS or in WINDOWS environment, or to
prepare a driver file to run the code in a batch way, which is usually convenient
when a mainframe installation is used. In the first case a driver which operates —
with a number of subsequent menus is available. A third alternative for the
solution and the postprocessing routines is to cali directly the relevant functions
one by one from the keyboard.
Apart from the results related to the dynamic analysis, while building the
model, the inert properties (mass, moments of inertia and co-ordinates of the
centre of mass) of the whole model and of the various substructures are
computed and a drawing of the model can be supplied.
In the present version of the code either the central difference or fourth order
Runge-Kutta (adaptive or not) methods are is used when riumerical integration
is required; nonlinear algebraic equations are solved using the Newton-Raphson
algorithm; equations with time-depending coefficients are solved by resorting
to an approach similar to Hill's infinite determinant, obviously truncated at a
certain harmonic at the choice of the user, linear sets of equations and
eigenproblems are solved using the facilities which are standard in MATLA13
package. In ali cases, when complex arithmetic is needed, the ability of
MATLAB to deal with complex quantities is exploited, the only exception being
equations in which both a complex unknown and its coniugate are present and
which need to be treated by separation of the real from the imaginary part.
Elements
The element library of DYNROT code is particularly tailored for rotordynamic
applications; in the present version 21 mechanical elements and two control
elements are included (see Table I). _ _ _
All beam elements are based on a formulation of the type usually referred to
as "simple Timoshenko beam" with consistent formulation for mass and
gyroscopic matrices. Isotropic beams have a circular or annular cross-section. If
needed, an anisotropic beam can be used in ali cases; if it is actually isotropic
the deviatoric matrices are null and later are discarded.
To take into account the flexibility of discs an annular element has been
introduced. Its formulation is described in detail in [13]. The element has three
nodes, one which is in common with the shaft, another two at the inner and
outer radii of the ring; the latter have a number of degrees of freedom which is
different from that of the "normal" nodes of the shaft. A transition element to
connect the disk and the shaft is also provided. Disc elements include both
— —..r srr
' . ,._ *. v •g r, i,. , •
.. :,..• i„,• ,
_
, .1-t- <Ah ì•lb
EC Flexural degrees Torsional degrees Axial degrees
13,6 of freedom of freedom of freedom
Type Forrnulation (complex) (real) (rea»
2 2
Spring Isotropic 4 2 2
Anisotropic 4 2 2
Cubic nonlinear 4 2 2
Linear with clearance - 4 2 2
Damper Isotropic 4 2 2
Anisotropic 4 2 2
Mass Isotropic 2 1 1
Anisotropic 2 1 1
Disc Annular disc 10 3 5
Disc-beam transition 6 2 3
Blades Row-of-blades 12 5 5
Disc-biade transition 11 • - 4 5
Bearing Eight-coefficients 2
Eight-coefficients (speed
dependent) 2
Hydrody-namic 2
Magnetic (with sensors) 4
Magnetic anisot-opic 4
Table I. Specialized Crank with piston 1
Elements implemented Controllers PLD
in DYNROT code for AMB Generai
gyroscopic effect and stress stiffening due to both centrifugal and thermal
stressing. \Vhen disc elements are used, the stress field is computed first using
the finite difference procedure introduced by Manson[14] and the centrifugai
and thermal stress field can be plotted before starting the actual dynamic
computations.
A row-of-blades element, developed along the same lines, has also been
introduced, together with a disc-blade transition element. Spring and damper
elements can connect two nodes of the structure (e.g. to simulate joints, bearings
or dampers between t-wo different shafts or between a shaft and the stator) or
one node to a fixed point (e.g. to simulate an elastic support). 11 two nodes are
present their co-ordinates must be the same.
The restoring force of a cubic spring, useful to model ball bearings or ela-
stomeric springs, is assumed to be of the type
F — k x (1 + 1.4x2 ), (6)
where x is the relative displacement of the end nodes. The linear part of the DYNROT: a
restoring force must be entered separately as a linear spring element. finite element
The restoring force for a linear spring with clearance, which can be used to
model roller bearings, is assumed to be of the type code
1
F – k(x – c) if x > c
FO if–csxsc 95
(7)
F.–k(x+c) if x <–c,
where x is the relative displacernent of the end nodes and c is the value of the
clearance.
The magnetic bearing elements can be connected at both ends at two nodes
of the structure for the electromagnetic actuator and at two nodes for the sensor
or at one node only, the other end being fixed, i.e. connected with the "ground".
The characteristics of the bearing can be isotropic in xy; piane or anisotropic,
which allows the study of the cases of bearings which are geometrically and
electrically isotropic but work in different conditions in xz and yz planes due to
the presence of static forces. If two nodes are present their co-ordinates must be
the same.
The element includes the actuators, t-wo electromagnets in x andy directions,
which supply a force proportional to the square of the c-urrent and inversely
proportional to the square of the radiai displacement u:
F I u)2 (8)
and two displacement sensors measuring the displacements in the same
directions. The sensors can be located in nodes different from those in which the
actuators are located. A bias current can be given to linearize the characteristics
of the actuators.
As active magnetic bearings need a suitable control system, two controller
- elements are included. The first is a PLD controller ori error feedback with an
additional filter on sensor output. The input-output time relation of the ideal
PLD controller is:
l de(t) )
" 7
•, -• . ;
_ •• • " • ^. - • '• •
. •; .' s • 'In,. ' • • • . • . •' ■
').;t., • • C• •,.". ■
• 'li '•
: j:; r< • •• .1 • •-• • ',";
• •• • - • • •" • -i -• , • +.
' • • .- • • •' , • .>„•r!:e <
•
EC
F(s) - (11)
13,6 sr +1'
where ris its time constant.
Also a generai controller is included. Its characteristics are expressed in the
forrn of the coefficients of the polynomials at numerator and denominator of the
96- transfer function. The generai contoller transfer function C(s)has the following
structure:
bn s" + bn_isn-1 + b„_2s n-2 + + b,
(12)
C(s)
sn + an _is n-1 + an _9s n-2 + a,
As seen for magnetic bearings, the hydrodynamic radiai bearing element can be
connected also at both ends at t-wo nodes of the structure or at one node only, the
other end being fixed, i.e. connected with the "ground". The bearing is modelled
according to the "short bearing" assumption[8]. The film is assumed to be "fully
cavitated".
To model different types of hydrodinamic bearings two eight-coefficients
bearing models are present. The first one has speed-independent properties
while in the second one the coefficients are defined at a number of speeds and
then interpolated in the whole speed range.
An element which can be used to model the inert properties of a crank
mechanism for the study of crankshafts is included; it has only one node, as a
mass element. The elastic properties of the crank can be introduced in the form
of a spring element, having the correct equivalent stiffness, or of a beam
element with a suitable equivalent length.
Static analysis
The aims of this part of the code are to compute the static flexural deformations
of the system under the effect of static forces, to plot the inflected shape and to
compute the stressing of the various elements.
Only a linear isotropic rotor is considered. Because in this case also hysteretic
damping can be included in the mode], the equation to compute the static
inflected shape is:
([K] - ico[Cr] + [K",D {q} = {F}, (13)
where VC") is the imaginary part of the complex stiffness matrix.
As the response does not depend on the speed if viscous damping is not
included in the mode], two different solution routines are included for systems
without or with viscous damping. Two other routines allow the plotting of the
inflected shapes and another one to print the forces acting on the various
elements.
Critic.al speeds
Only linear systems can be dealt with as the very concept of critical speed is
linked with linearity.
• .• .•
, ..
■ "...- ' ''' '
The solution routines deal with isotropic systems, systems with either DYNROT: a
unsymmetrical stator or rotor or systems in which both are anisotropic. In the finite element
latter case an approximated solution is obtained by truncating the series
expansion for the solution at a number of harmonics decided by the user. When code
possible, the user is allowed to perform the computation along a generai
straight line = luo on the Campbell diagrarn to compute the secondary criticai
speeds. Two further routines allow the computation of criticai speeds of a rotor 97
supported by magnetic hearings.
Post-processing facilities allow the plotting of the mode shapes related to the
criticai speeds.
Campbell diagram
Various solution routines allow the computation of the Campbell diagram of
linear systems with or without damping. Also in this case the system can be iso-
tropic, and can include either an unsymmetrical stator or rotor or be completely
unsymmetrical. In the latter case an approximated solution is obtained by
truncating the series expansion for the solution at a number of harmonics
decided by the user. If the system is isotropic a routine based on an iterative
modal technique is included.
Only linear systems can be dealt with, with the exception of two routines
allowing the computation of the Campbell diagram and the decay rate plot of a
rotor on hydrodynamic bearings, with the stator either isotropic or
unsymmetric. In this case the static equilibrium position of the system is
computed first using the nonlinear model, as the position of the journal depends
on the speed and, in case more than two bearings are present, on the shaft
deformation. The behaviour of the system is then linearized about the
equilibrium position and the natura] frequencies are computed.
Two routines compute the natura] frequencies of isotropic rotors on magnetic
bearings, one with varying speed and the other at a chosen value of the speed,
varying the gains of the controllers of the hearings.
Generally speaking, the undamped analysis is far faster than the damped
one and should be performed as a first approach to the problem. Also hysteretic
damping can be introduced into the computation, even if this compels the
repetition of the computations several times for each value of the speed[8].
The post-processing subroutines allow the plotting of the Campbell diagram,
the decay rate as a function of the speed and the root locus. While in the
Campbell diagram the natural frequencies are plotted as functions of the spin
speed, and in the root locus the imaginary part of the Laplace variable s is
plotted as a function of its real part for various values of the spin speed. As s is
used instead of the complex frequency, its real part gives the decay rate
changed in sign and its imaginary part gives the whirl frequency.
In addition, the mode shapes, real or complex, can be plotted at varying
speed as well as the static inflected shape for rotors with hydrodynamic
bearings. A further routine allows the plotting of the root locus for rotors with
magnetic bearings by varying the values of the gains of the controllers of the
bearings instead of the spin speed.
; :. . -" . _ .- _. . : . : .. .. :.,,;;;•
;. ...-
.:-..; .. , ....;.',.:,:..:-:. ::'..,•.,:'.. '.' *:<./.z,,-..,;:-..,:,. :-..... -. ..*;ili;';'.....» .
- :. -
.':..,....:::'-,
. ,:',,,.,:'-.•...'..: l'.-:.. • ''..'.. -,..'.1.-..'.,.-....', , z:':?....
.
,. , . , : :':: .. • = • .• -, - .` ,'. ,..-•,',' • >.--' ..::
,
... , , . .
.",,..::..,',' ''.''''', ..- '-'• • - ...'"••••.;,. ' ',' ', ',',': -''-.• •,,' - -..',1V,,Z.t.;.-1?,•';11",-.:=-','.',
, . -.. •• , , _ - , ,
.... , ,. , , , - • , - • ... '-,-, • .. ..';' : : : •' . 1:--.'.•7:: ',.' .`-'2 . : .• , -.:•.;,;_:•.;,....
• •".-,.:;;'1.•',:;,',.‘..;;:,?.Yrs,,;,.'‘,',."-',-11't":1,3"'•' .'''
.7.‘-'.", ■ :;,•.l.5'.
-
EC Unbalance and acceleration response
13,6 A solution routine allows the computation of the unbalance response of systems
which can be either linear or nonlinear and include also non-isotropic elements.
A routine to study rotors on magnetic bearings and one based on an iterative
modal technique are included.
The routines which can be used for the computation of the acceleration
98 response can deal with the same systems seen for the Imbalance response at
constant speed. As they are based on the numerical integration in time of the
equation of motion, no hysteretic damping can be introduced into the
computation. If it is present, it can be either neglected or transformed into
equivalent viscous damping using a purposely written routine.
Equation (2) is revn-itten in the rotating frame, transformed to modal form
retaining a reduced number of modes and then integrated numerically in time
using central difference or a fourth order Runge-Kutta algorithm. As already
stated, the modal analysis is performed using the eigenvectors of the
undamped, natural (linearized, if the system is nonlinear) system. All modes
with natural frequency lower than ten times the maximum spin speed are
retained.
As the computation is based on numerical integration in time of the equation
of motion, there is no difficulty in varying the arnount of unbalance in time. A
variation of unbalance at constant speed, as the well-known problem usually
referred to as "blade loss" for turbine rotors can be studied with these routines
by assigning an angular acceleration equal to zero and an unbalance which
follows a step function.
Four post-processing routines allow the plotting of the displacement at any
node, the driving torque needed to maintain constant the speed or to accelerate
the rotor as a function of the speed and the inflected shape at any speed. Also
the forces acting on the various elements can be obtained for any value of the
speed.
Torsional and axial vibra tions
The torsional and axial analysis has been introduced only recently in DYNROT
code. At present only a few solution routines are present to compute the natural
frequencies of the rotor and the corresponding mode shapes and, for torsional
dynamics, to obtain the forced torsional response under polyharmonic loading.
The latter is mainly aimed to perform a torsional analysis of reciprocating
machines: if crank elements are present the so-called "inertia torques" are
accounted for, together with the driving torques. Alternatively, the linearized
equation with periodic coefficients can be dealt with using a reduced form of
Hill's infinite determinant. The number of harmonics to be retained is left to the
choice of the user.
In both cases the system is assumed to be linear and reciprocating machines
are studied following an "equivalent system approach", but no further
assumptions regarding a modal approach are made[8]. If the hysteretic
damping of the system has to be taken into account, it must be transformed in
0.2
0.15
0.1
0.05
o
—0.05
—0.1
—0.15 Figure 1.
Sketch of the rotor of a
—0.2. small gas turbine
—0.1 0 0.1 0.2 0.3 0.4 0.5 . 0.6
Elements 3, 4, 11 and 12 are used to model the compressor and turbine rotor:
they are very stiff, but massless (the density is set to zero). The inertia
properties of the compressor and the turbine are included as mass elements.
The damping of the material is introduced as loss factor of the material and the
same is true for the bearings. In the latter case the loss factor is referred only to
the linear part of the stiffness.
As a first computation the first three values of the criticai speeds of the
linearized model are 1,768, 3,753 and 31,363rad/s. The damped Campbell
Z-- ' - ' - ."7-' .— - : : r ' '"-; 7 « - - '-: 70''',77; ,t. ,':..,:■••'=" ...',',... :' -',-19. .tr: lk , , .t";•.,,', .,`.", 5 ..p•Mt.::-/-; ;',,;.?'::;:-:A,Y;.7" -V.r, •5,.: v -t ;•,,,e.,:•••11. • .• ,-:72,•=r737-1- ;17.-77",1-7-ZY,,Mr.w",.7‘rtirr, ,rr,-.F.:g2,"àfi?iiit.
•„^-,2-;•!2„;5 ,-;;.1,,te',..,(... .,, r.Z.."".'44:' ..::P ; ■., , ',"■ :,.1:4;,■,.>;,;'',,"1 !:;.ege.",,.., 1:,. gky
- • "- -
j{,,. , . _ . - ' • , ,, ',. - ' ::',, l',.-,;•",..2.{..-:.:z:é(.1..;i2',:i.,j,j'zs:3-.7.',.. 5"". :-!...,.`.". :'; .,..- .4.-",'::,';':"..•;,^.;....,-: '-i*:..'..-II •;-';:?-.',"
-
.k ': ',..',, `,..:«,',..'":"''':. ; ,
-
-,
' ' ', t .•,`,,:.-..,:.t3•4•Y.4..:••: •. :>,;,•-••::5-,,,•-•\'', , , -.1:'::-. , 4; , r••-•.,14,, .....?,,p-s_
=' : .- ' ' ' •- •'-`" '' • ' -• . ' • ': '- - , , ' .. .,;-i . '2'; s *.:•• • ,,,i'-‘‘z,'-:?7•i“ ,;•,t , . 1 '• .71:-, ,,:g.;,' : i :..t .-1.T.i5;!*
EC Elernent ID (mm) OD (mm) Length (mm)
13,6
1,2 15 30 30 Front bearing shaft
3 0 120 68.3 Compressor
4 0 120 34.2 Compressor
5, 6 55 63 15 Central shaft
100 _7-10 60 68 44 Centrai shaft --
11,12 0 150 30 Turbine
Table 13 30 40 40 Rear bearing shaft
Beam elements 14, 15 20 30 20 Rear bearing shaft
diagram has been computed and the results are shown in Figures 2-4. The real
and imag-inary parts of the complex frequency A are shown together with the
root locus where s = iÀ is reported. Note that the root locus is not symmetrical
with respect to the real axis, as usual in rotordynamics. The mode shape of the
first mode of the linearized rotor at various speeds from O to 5,000rad/s is
reported in Figure 5.
The unbalance response of the linearized rotor is reported in Figure 6. The
amplitude and phase of the orbit at node 4, i.e. the centre of gravity of the com-
pressor, is plotted as function of the speed. An unbalance due to an eccentricity
of 1,um of the compressor has been assumed; at a speed of 2,500rad/s it
corresponds to a balancing grade of the compressor G2.5 following ISO
standards. The driving torque needed to keep the speed constant is plotted in
Figure 7.
The acceleration response of the nonlinear model is reported in Figure 8.
Again the amplitude at node 4 is reported as function of time. The rotor is
performing an acceleration from standstill to 3,000rad/s in 6 s with an
_
...... 1.,,..-■-•". ..rt r./....m... .. .- -eq,-. , •+- ' • „.. - .., • ^ ,...••-‘-."''
•-r.- --,
, -,:-..,,Fre:t,,77,1,71n7,,,T,79,.2,-,,,,..„:„......,
""••"• ' ---
- - --`-----*--
4vp
•;-, ::••:.?$.0. .•, - - ~- ‹.,3 .,,:•4 t ,--"-''-- "' ' " '""""'
-*--,,,,,,
. • - .1, ,(y.1.,-,
•-
• • - . - --• •":, ,•, •:',
,- ''. , . , - ... :,' ..-. -,-•..':
.....~1,, , c," ,, ,,,--3,.
..‘..',...".‹. ,"., ',"'..."-.4.-• ".'pZ.,',;',,,.5.1"" •Z`..." ,Ktr.•4,,,,,,.. tgikr.,-...,34.,.L(...„..,,út,,,A,, >,„,:,.,;>›,n,„. , s ,'e,..,,,,
,
..:,..,1. ...e, .,-.
;.
!"R''''N
,,,.;,....,.
, .. .. ■
'...1. ,r-4. - 4
- .5 ,> - ', , ',.... 7 ; - ,../.,'..:ir,b•rt.z.., '.. - .-.1i.p , ;k„.-:■&1.,', "“rk.,'A>t,« ,..5:.•.,",MY:_,_
..‘,R,,1‹
. .. ';30:ly.l'.,-.
-, .-.. ._ . : .-,, .•:5:z.'",?■ ‘,2:Its ir, ‘•;p5,,.:;•;;;.;,, , - . -•,..- ; - ,,,. . ,, ",...., ' ' • . • ". ,1:•'•?,...`;;',:g.:•,1-,,"-til.;»,,tt::i..,;,:,5.4-'.5y;',g1,:in`")1.24,-,:*•'; i• 2: 1'.:::•:.P..- 1 ›•,.. 'sg,g1Alst•ì‘::4!;••ft-'1 .IP.'''.'il' ,'..p.42.
,":,...-,- . : .. - ' " '.'. ' ' ' '.- -: - '.. "*''''''''''''f ,;:', '.»-'''',.t• 3 1.,4',•::«44 l';:aP
-- ' • ',' ' : ' r.... ".:.'-' ,';'.'..'''','',;?•:r....13'3.',-" ,,A-i;.7,,,•4,7, -;;;,.. ,-.'?J"?.„:12;:s.;',:ì'. 1.3-`';,::.5:.'•;; III/"'"•ii`j -i':''•'' ''-‘
' ' - T • . ' ' - :" .1.4:?;"\gigs::‘0•C:' '4,:j4..‘2;"••.;'ii',.''',•': -
A'„,3r.;:'ìg'<tn,•••,11ìz,;i1:z•ii:17:, ...cgt..n. 41..1.-,,,,t;., ,,,, ^2
", , ,, ' - '-• • •, ." ." '•- ' - ' '••, . • ' .," • ••• ',•• •-•,;,•7.; f.ti,V;:•i'i'r.;." ' .,■. " :',-.'•:<1!Via•i.:Z.".`ii ' • :"e."i' .-•' , 11",'"i' ■ 'It're'r?:te? , ■ '..lt•,;,•,!XA':•‘",-
...: •-'.';',:f.r'et:Eti . \.,•%'*,,, ••••;`, ,• " ...." .• ,"• -• ".,• ," - , •_. :'.t'..' ,Z.,-,. ■ .4","..•,2..s.'i.'",•,•i"."'z'ii;•'• -1•2'ilt':.5.:,h
'
-'1,'Z',1-'!,41'.•(:`..,•';',..; ■ .`••-:'•«,d4
. 9$4,4'.9!''.1;•'?:,4.!:••r:-■‘•,,On't'-?'
, '
.■ . "
Lambda (rad/s) DYNROT: a
5,000 finite element
4,000
code
3,000
101
2,000 • ,
1,000
—1,000
—2,000
—3,000
Figure 2.
Campbell diagrarn of
—4,000 the linearized rotor: real
part of the complex
—5,0001 I r frequency A versus spin
0 500 1,000 1,500 2,000 2,500 3,000 3,500 4,000 4,500 5,000
speed
Omega (rad/s)
18-
16 -
14 - 41111111&
12 -
*I*111~Ii
ihmab..~
10 -
8-
6-
4-
Figure 3.
Decay rate of the
2-
linearized rotor:
imaginary part of the
O- complex frequency A
0 500 1,000 1,500 2,000 2,500 3,000 3,500 4,000 4,500 5,000
versus spin speed
Omega (rad/s)
" •
•
.
'
- ■
-
i •
• • t.4-1 '• r. 5
■3-9"R.- i 0.:,12:4, '"2
•
•
• ,
n•
', • • '
'4,1•,•;
,
l'11‘,:•:,15S*,' ■
‘
, . , t _ ,a
• • • • •
'••
EC 1(s) (rad/s)
13,6 5,000-
4,000- 51c
3,000- 300Ì.1.4
1,000-
0- __
- 1,000 -
o,
- 2,000-
-3,000-
- 4,000-
g.:N('
Figure 4. ).:k,
-5,000- i i r
Root locus of the -20 -15 -10 25 O 5
linearized rotor:
R(s) (i/s)
imaginary part of s = iA
vs rea l part Note: As usual in rotors, the focus is not symmetrical with respect to
the rea! axis
Figure 5.
Mode shape of the first
mode of the linearized
rotor at various speeds, -0.5-' I I I i i
-0.1 0 0.1 0.2 0.3 0.4 0.5 0.6
from O to 5,000rad/s
Mode shapes for the forward mode number 1
—200-i
560 1,600 1,500 2,600 2,500 3,600 3,500 4,600 4,00 5,600
Omega (rad/s)
Amplitude (m)
10-4 -
Figure 6.
10-6- Unbalance response of
the linearized rotor:
10-8- amplitude and phase of
the orbit at node 4
(centre of the
compressor). Unbalance
due to an eccentricity of
11.2m of the compressor
0 500 1,000 1,500 2,000 2,500 3,000 3,500 4,000 4,500 5,000
Torque (n m)
10-6-
1 0-2 -
1 0-4
i 0-8
1 0-10
1 0-12-
Figure 7.
Unbalance response of
10-14- the linearized rotor:
driving torque needed
to keep the speed
r constant
O 500 1,000 1,500 2,000 2,500 3,000 3,500 4,000 4,500 5,000
' • •
• - , - - • "S‘1,:t =5, 4,- ,s ;,5,117:,..t. ,('‘i'''.4.'ì?I'''').r9 1` '
' • '
, ,.. • -74 \*, • ; • ::-. 1•1••• PitIli d'i.n.ftt 4) .5 4 a,trkA .,,,:,•
, • ,
'
• :, • •• , t • • • :, . : • "'•'• •
EC Aryditude (m)1 0-5
13,6 2.51
2
104
1.5
1.
Figure 8.
Acceleration response of
the nonlinear model:
amplitude at node 4
(centre of the 0.5.
compressor) versus time
during an acceleration
from standstill to
3,000 racVs in 6 s O
(ai = 500rad/s2 O 1 2 3 e4 5 é
Time (s)
acceleration (1) = 500 rad/s2. Note that before encountering the first critical
speed the amplitude is very small while after passing it the roller bearing works
at its maximum clearance. Some vibrations are also present, particularly in the
supercritical range. The analysis has been performed using an equivalent
viscous damping and a total of four modes have been considered.
Torsional behaviour of a marine drivehie
A marine unit is made by four six-cylinder, four-stroke inline engines driving
two propellers through gear wheels. They work in pairs, the t-wo engines being
assembled with the fly-wheels one near the other. The t -wo pairs are located side
by side, driving a common shaft through gear wheels with a transmission ratio
of 0.9. Two propeller shafts are driven with a transmission ratio of 0.4 from a
common intermediate shaft. A sketch of a lumped-parameter model of the
system is reported in Figure 9.
The compliance of the gear wheels is neglected: in node 29 is located an
"equivalent" inertia equal to the moment of inertia of two pinion gears plus the
larger wheel, all reduced to a shaft revolving at the speed of the engines. In a
similar way, the "equivalent" inertia located in node 30 is equal to the moment
of inertia of the three gear wheels suitably reduced.
A dissipative damper is added to the front end of each engine. The casing of
the damper is connected to the first crank through a shaft while the seismic
mass is free to rotate. A total of 40 nodes and 79 elements (31 beams, four
• :'
..• , •.
7 14 DYNROT: a
Engine 1 Engine 2 Propeller 1
finite element
-I .1 iff n
1 U
[l pi i ,r1 i Lir _fin
2 LI 3 4 L 5 U 6 7 ,-;14j 1Ju12U11J1OL 9 u 8 J
1_1 ———
[
i
-
code
1 2 3 4 5 6 13 12 11 10 9 8 30 L
- 29 30 31
r - 29
105
_Ed n i hl P 11. f1
,
nir, [
springs, four dampers, 16 masses and 24 cranks) are used. The data are
reported in Tables V-VIII.
Apart from the first five frequencies equal to zero for the rigid body motions,
the values of the natural frequencies of the system are 8.2, 15.9, 56.9, 124, 250,
368, 375, 901, 1,443, 1,902, 2,315, 2,571 and 2,851Hz. By studying the
eigenvectors, the first mode considered is a motion in which the p .rop-ellers
oscillate in opposite directions while the whole system from node i to node 30 is
stili. The second natural frequency is related to a mode in which the whole
eng-ines move in a direction while the propellers move in the opposite one.
The results of the dynamic study are reported in Figure 10. The maximum
value of the dynamic stress in the engine within the working range occurs in the
1-5 10 45 82.5
6 0 45 71.5
7 0 40 300
29 - 0 65 4,000 Table V.
30 0 70 6,000 Beam elements
Node Mkgrni
7 0.0675
29 0.175
30 0.497
31 12
Damper case 0.004 Table VI.
Seismic mass 0.04 Mass elements
7:727, 2,
- y
'
,.!;: . •
•
;•• • .
, 3 s•`.
EC
13,6 Harmonic Ak Bk
1 0.330 O
2 0.330 0.140
3 0.290 •O
106
4 0.220 —0.040
5 0.160 o
6 0.110 —0.048
7 0.092 o
8 0.051 —0.051
9 0.058 o
10 0.048 o
11 0.040
12 0.034
13 0.029
14 0.025
15 0.022
16 0.019
17 0.017
18 0.015
- TabIe VII.
Harmonic components 19 0.014
of driving torque 20 0.012
. - -.... - •,. - -..........•••;,--c•:,-... -,:c.v.; ,,•:-.,-,;• , •••••? ,••r,••^-,;••••••-•••••, ••,• - •-•.• •••••-••••-• •••• :,:',..‘ ,.--s",•••••••--••••t"7:•••"••• •,,, c••••••••• • - ••••••• ■ -•••••••••••-••••—•—••••-•---•••• ' T'-'" • -
''''' — ' ''''r '-
' ..-. •.., :::•*:'• •, :• -i • 7 !"-•.;- .• ..-S '.t7,'••‘.' .-.-4 • - -5•••.p`t7-A-` A , • -• .3•,',"`$'
:: 54..i'i , ‘,y.-.
"-:5-3W•3'.3.551.'i''''''..":',"'>''''',7
a , r.z.,.r......-..2, , -, ,,::!...,:
,.2: ..:.:, ,,..,, ,,i '•-'.`.: ‘• ••,.-7n.e....% ;''.''•'''' •I'?':'''S .';'''''' , <,.••-'..,k--'''*.,','.- ;‘-:: ;?' -' ' ' '''' ("? .Z, `...-,''..,.,
.,..,,,,,,f.., , .,3,,•\...5,../ì); \ 1,-4
,—,
V.--•
,
,-;, c ,,...7,• • -,-.;•• ;;;,...•..2•2:04< ■ •••;;;'•,S1•87:
- .:;;,•• ■ .";i4i ,".Y;.:, ',-. ;!"■ •.. .r ."4",■ •sii;2- r..̀"...<;í>.-
> ,,.74 ‘ •-
,;•:',.:
P ,.'',19.F * f
,? „.
.• 1
•
.., ;!;-•> ;,•:, .
•;,,̀ 1 / 4 . , ' , , ., , ,t,....; >,' : , , , ,, ,,,,•,;.›1k-245;..z„.11..5
',..:."..—,•,•. .
. ..'2,',7,•' .k.k4
t‘"--‘,. . 17'' ‘' •4,...'' ,i't', .1•5. ' ,''‘,.-''.',5•. -• ;',."•,. .,
,, i ■
, 1/4..'"*S.4A
"' > 1
. 'f..)'
*'1'.7!›...A.g. ?•,,
' ,,
4 ‘ .;• ;..?_
1: -.
' •
1-
>;19 i'21.7•541-,'W‘Ai •ps:s.,••' , •tr,••: ,,1..•..•-•.•••',.),• -••• I : o' ''''' f-, Yi,,2 ••• '''';..1" '< ;,,,r• - •';‘
.,z•I.Y1t1-;:t‹...s+?'.;','Y•fi'':...,:', k.. ,..- 4" '1''' 1 . '
''( ,- , .'."-''.5'..".,:-.,,,S.:.
• ' '.• 2.b:' •':-..'.,..y.,
'••••):,,'" t.''',13>AVilky
,,,, ,,,11.._•{,•• ■ 9•:•..;p:t1,
..._.,,S1 - ,-5, .• '',.,;,t,,.,
,ze • ,.. -,-...frt„•:,....:
• •••,.14 •>••1• -k‹,,,e,...st<.,..:::
'•,.,:t.:,.•,,; i' •••••••.%«'' .,,.*: ":-. ty.Z.;‹, ,:>',
,..- , ,<.:,,4',.. t'''::•''r'. ,ee", ...Y.,
,,,t.rd,,,,,,,-.,:•,•!,5',..",;•;-.".-Z...,1,•.
,,,, •',.."••' -3.;e1 , i.41:1/2- .'": '. . : -:•.')•`; ' ,V.:'.. t ,•':-V,..-.'4,' - ''•- ■••::",ì:".,,..";.•••'.,?;"..._ .,-; - ::' .."Y• ,•.-}.. ",, ,.i;
t, .4,,,,
\ •,,F2-,:,15,.Yd.f,r;•
. < 05.,*,,,„,,, ,..,:`,1;•;21".,z . ,,,‘,•,.T.:-, -..".:..::;. , :" 1 ".• i",91,-..,':......' `.>„,,'' --•.'„,•;_"'," ''•''..:•: -0-'•,''' ,'?.,.,.- ' ,y't'rt';?•gít.j.•.;• :‘,,t ,",...„•'•1 .;',:
. .,.,....;;;.;.•::•',5.1•:-' 7,3';',55,,,.:.„, ,..;‘,..5,,`, ,, ,,..! ..: ..,•:,..,;-....`" :''''. . ■ '.. V 1..2.,• r....
1 , '', d .,...,3 ",'-',- " ...l - •4 , .: ';‘, ,n1,,,5 1:;‘,"',,ii
:" ■,:1,7,..,....;, ,'- ' -',, s-,rt.',, v;r5,,
-1-N. i ,, 7?, 4( , t , t- • ;•,, t-1:..-',
,,,,it',3,1:?n,C,......".,,,,,... .;•-,,, • i-N..; •
'•
... "".".;":,','ì-5,:',-ilt". 2 ,
- •,t‘,-,1t.e;••ì' , / ',"?.,%.,:.",r ; ...."'".•-• i ' "-,-,..,...4j.g.9.,1',,,,%•:•'.,,..,-,'":: ,-, ,V,,,
'" "-;.",--/,..:;‘,F.,.. '..."•,>,,....-•:1.0‘ti‘,(,-.'",f.'',',.' ;;L•. ,.....rz••ffi ,-
-..t.,..5'....•-• 5-;;7,--.7.,...";...,•;:lf,;.-1,..-C,,,..._. }: , ••••-•- 7 ,•,-• ..• . -, .± .•i•c:c?,•,-,7 ''...'•-, c:•••-.`-', ,, 5 - ;)E-.qc.tri 2. i, ::
- • - ''''''' "",• ; ,*...,-t-n,,,r-....1A7 ,•,4•.' ',1..r...tt;•.... •›•;. 2 - •; - ,' ' .- - - 4. -""--...
. , , " •'.4. ' I T---, - • - y. c( .-- • ,,,,,:-:,.-, • •••• ;--c .
' '.•• • - - •-, - ..- •-• • , ', -. - ;• -- - .• "., - :
[M N/M9 DYNROT: a
80
Element 4
7
finite element
code
60
107
20
30 Figure 10.
i- Dynamic stresses in the .
0-1 1
0 2,000 4,000 6,000 working range
w[rpm]
element number 4 at 4,87Orpm. The maximum value of the stress in the shafts
connecting the engines with the first gear wheels (elements 7, 14, 21 and 28)
occurs at 5,92Orpm, the other shafts are far less loaded.
Conclusions
DYNROT code allows study of the lateral and torsional axial dynamics of
rotating systems. In the first case the basic formulation is very generai, as it
includes the possibility of taking into account that both the rotor and the stator
are non-isotropic and that the angular velocity or the unbalance are not
constant. Clearly the equations of motion in their most generai form can be dealt
with only by numerical integration. If the rotor is linear, the steady state
analysis can on the contrary be solved in closed form.
The study of the torsional behaviour of reciprocating machines is reduced to
that of an equivalent system, obtained through using the usual assumptions.
Complex co-ordinates are used in the whole code and extensive use of
complex arithmetics is made. Some computations are dealt with using modal
co-ordinates with a reduced number of modes; also modal co-ordinates are
complex.
The element library contains 21 mechanical elements, including nonlinear
springs to model rolling element bearings, crank mechanisms, lubricated and - -
-,..,-- -,- . -.-vm,it . 5,..r.l.inv.r...v.,..ì",...-,----,- «...rn- ,r,.2.-• ,,,, ---„,,,....-.,tiv-r- , --..-,-,‘.zrzre -„re7.-,--7..,5"; ••• .....,7,
, _ -,-.:4,IV' pgil - "::"'
.•-• .....
'...%,:•• r. , ..'•;,'•:' ......:':« i .•,‘.• • ' ': 2-95.2:1T75",■1'51?3.17.:.t'.5.7:2 ;:::J^;'•:7':P•-s+-7. :• ',':::.-1,-,,:e-b.,:',7:.r,',.) . gr't › .;:!....--1i.;e'.1". e.)5--;.:35:-p,',..:*5;;;;P' ,,,--..,-.5(91,-.-...1-. .'":::t'int. '',, ----, '.'n ,..k9 •••ali; ‹-Xr' '••• 2 ,;^ k •fr, .(i eq,
',.:''?? ''- • ,,,:-.. -,-_:,, ....,"'"..- '. • ,;%;'•;1 .1•31f:; ,;.4.:>(...•';3 ., •., ' ,.. ••;-_,,e.,Kei`1,":".J`:,.•'),i,.-A-:-;:.:;,3,4 4.14.•..,:i.0 " .r sl'i'íjt,,•31.??.%5;,:t. - 415.,e-",1.,;• ~.,.:,,,,,,.("»,;;•'..,A.if ,4,>0..ti■.;',...1 ■ 1q,5.,•.-.;i:(,:••• ■.*1 -i ,::.'..9.4 A.-!› :•:-' ,Ii!".,.:1 .;.e• :::'. .
,:•.. . -...• : ' -:«f'. ':''' ..., ''.:t: , T7.10.,..iii' :. ,..-,•.'.., 5::, 'ke' • :::: .'•.' ....• :. :-.. :(19::: . :',3•:.!`A., SiV:',/,:tc'tlgtrst''S. ‘i...', •!;'. 5-•' :,l'e.,(11!^z,Z9r5„,,, :+d-15:;WO/.043,.,, st'Og'› c•.;•1/.4 1M;Ss.'.! >.:'z..3?`,"-0%- kr t; , -e ,1,.....,..-:..Pkt.k4;!.ttrq":171ii'&34i.
Z :- - • : - • • j- • ••.' • •' ,':''':•:.•';', ..52,.4.5 53;1.5';'-i',!,,k..?3:'2,' ,:,••••• 3g•.,•.;(::%V,>»,•'.;I•, •::>>.,Pr.1`,4 ,Z.,41^19=',--. '.-2'0 2•••, ;....,•%"';;2›.'11,i1-4, !-/-• ,r1-11%'7tt-,-/ Ilì2.•4 ' 1,,,i c',5,::y• • T''el,1..
_i .441;.1:., •:; ;',, :',i'&'-d."4,,,'4.,tr.i.:.;O: 11.4.b ., t
?..:r , I • • ,." •. . ,..• i *.:•'' •- •-r•-/L'" 1.55.-,.•!. 'y• ,A, • '',•ro';. ,. : . ;;&,,,',:•-,,',:•=5.„.5. ,j1.Z..r:M:::t0§,.•:..M.+FA4 14?, , •.•_•--,r i ••:'t i2,••=.9:....,55--,4`41., ì.,9.1: , :..',. ,,9,•-.,«;-,,,A-,-,,44,2,;>;•...,••:, ,,Se[1.‘1•:' •--.,`,Ue.,.,X; fli 3;.At;%-,:.,‘",»Aret...h,..1,:e. ,
.,.. - , . , •••.. ,.: • • ,",'.', ,.:.•+:....•,...t•,'
- i;„, •:'..,;.4-5.',,iYi...:,
, !. i , - •,,,1•:»":•i'; *, ',•:::;$:, 3/4^. ...-e . ,,„,?is •P;',•;g9,..:;; .,,,, t•,...-.'.... C..c5.;- s, :2.0.i.,•?:„„,-'...■Y' -í;.,..,.:„;-,',, ,,,,.<1.y. s. ;,-..;.s 'h‘‘.••;..' ,',:it't >i•-•:,!.,..t''''(•:1•,-:''.›.,,4 i,I ZIO.,',
...• ' ."•••'" ••• - -. " . ,••,:".: '',1'•;- ,:..;::' • •,,,•';‘,.',IR••;,, -,FV:„.1; i'....:75... ,z <„.',,‘,,:..,,,: ■-;.'j,•-• -.. r•A"^:'..'.:.•."-i••?...0.,....g..•345 ,..:•‘,., 5.;ì: . *9'.•5+-1;*0„4",55,...-",1•0.0kletir 7•<•+' .,q 0;";;A1^, ■ S.‘.?.%.,,:,::li.,)7s <1.'"(P.::•?,i,ff 0•49òLlYrgtR.'0.ìiit.i •
. . ,
EC magnetic bearings flexible discs and blades plus two different controllers for
13,6 magnetic bearings.
The code is based on the MATLAB interact-ive software package and, as a
consequence, can be used on any machine in which MATLAB is installed. It can
be run in an interactive or batch way and allows the results to be obtained in a
way which is readily usable for preparing reports or for further computations.
108 _
References
1. Dimentberg, F.M., Flexural Vibrations of Rotating Shafts, Butterworth, London, 1961.
2. Loevi, R.G. and Piarulli, V.J., Dynamics of Ratating Shafts, The Shock and Vibration
Information Center Naval Research Laboratory, Washington, DC, 1969. _
3. Prohl, M.A., "-A generai method for calc-ulating critiéar SipéedS of flealí -r-otors", ASME
Journal ofAppiied .Mechanics, Vol. 67, September 1945, pp. A142-A148.
4. Lund, J.W., "St2bility and damped critica] speed of a flexible rotor in fluid-film bearings",
Journal of Engineering for Industry, May 1974, pp. 509 17.
-
5. Nelson, H.D. and McVaugh, IM., "The dynanlics of rotor-bearing systems using finite
elements",Journal of Engineering for Industry, May 1976, pp. 593-600.
6. Genta, G., "Consistent matices in rotordynamics", Meccanica, Vol. 20, 1985, pp. 235-48.
7. Gerardin, M. and Kill, N., "A new approach to finite element modelling of flexible rotors",
Engineering Computation, Vol. 1, March 1984, pp. 52-64.
8. Genta, G., Vibration of Stnictures and Macl2ines, Springer, New York, NY, 1993.
9. Genta, G., "Whirling of unsymmetrical rotors: a finite element approach based on complex
co-ordinates", Journal of Sound and Vibration, Vol. 124 No. 1, 1988, pp. 27-53.
10. Genta, G. and Delprete, C., "Acceleration through criticai speeds of an anisotropic,
nonlinear, torsionally stiff rotor with many degrees of freedom", Journal of Sound and
Vibration, Vol. 180 No. 3, 1995, pp. 369-86.
11. Genta, G., "A fast modal technique for the cornputation of the Campbell diagram of multi-
degree of freedom rotors", Journal of Sound and Vibration, Vol. 155 No. 3, 1992,
pp. 335-402.
12. Genta, G. and De Bona, R, "Unbalance response of rotors: a modal approach with some
extensions to damped natural systems", Journal of Sound and Vibration, Vol. 140 No. 1,
1990, pp. 129-53.
13. Genta, G. and Tonoli„1., "A harmonic finite element for the analysis of flexural,
torsionalind axial rotordynamic behaviour of discs", Journal of Sound and Vibration
(forthc:oming).
14. Manson, S.S., Determination of Plastic Stresses in Gas Turbine aSks, NACA Report 871,
1947.
- _ . . •_ , •:,.f.; . ,r.,
. , .... .
•
... ..,., '-'•-''.q.
.. ,-., , ,. ., •
,... ....„...
. ...,...._,, _
u(f) control signal DYNROT: a
xYz fixed reference frame
complex co-ordinate (z = x + iy) finite element
C(s) controller transfer function code
force
M moment
N ratio between zero and pole
moment of inertia 109
Td prediction time
Ti reset time
[C] damping matrix
{F} , {f} complex force vector
[G], [g] gyToscopic matrix
[K], [k] stiffness matrix
[M], [m] mass matrix
{q} complex displacement vector (fixed frame)
f3 angle defining principal axes of inertia of the element
eccentricity
O rotation about
complex co-ordinate (0 = Oy - i
A natura] frequency
At coef:Ficient of nonlinearity
X angular misalignrnent
time constant
spin speed
imaginary part
complex conjugate
Subscrzpts
d deviatoric
m m ean
n non-rotating
r rotating
P polar
transversal
eq equivalent