AIM-91-1102-CP

OPTIMIZATION OF AIRCRAFT ENGINE SUSPENSION SYSTEMS

Douglas A. Swanson
Engineer, Corporate Research, Lord Corporation

Henry T. W U ~
Engineering Specialist, Aerospace Development, Lord Corporation

Hashem Ashrafiuon
Assistant Professor, Dept. of Mechanical Engineering, Villanova University
Villanova, PA 19085

Abstract transmitted forces by moving the system natural

frequencies away from an undesired frequency range. In
An interactive computer program has been developed to their simulation, the engine was modelled as a rigid body
design aircraft engine mounting systems used for vibration with six degrees of freedom, and the design parameters
Downloaded by Freie Universitaet Berlin on January 9, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.1991-1102

isolation. Mount design is largely driven by two were the stiffness and orientation of each mount. To avoid
competing criteria. Mounts must be soft enough to a trivial solution where the natural frequencies of the
provide vibration isolation, yet stiff enough to support the system approach zero, an augmented function was formed
engine without excessive motions. The Constrained where large design parameter changes were penalized. This
Variable Metric optimization technique is used to technique can be successful in moving the natural
determine the mount design parameters which minimize frequencies of the system away from an input frequency.
the transmitted forces in the mounts, subject to constraints However, it does not necessarily minimize the transmilted
on the maximum allowable deflection of the engine to forces in the mounts.
static forces. The design parameters are the stiffness and
orientation of each individual engine mount. The aircraft
engine is modeled as a rigid body that is mounted to a
rigid base representing the nacelle. An example is used to
show that the optimization technique is effective in
designing engine mounting system.

Introduction

Aircraft engine mounts have two main purposes, to

connect the engine and airframe together, and to isolate the
airframe from engine vibration. Vibratory forces are
mainly caused by rotational unbalances in the engine, and
result in increased stress levels in the nacelle, as well as
high noise levels in the cabin. Recently, the increased
utilization of more sophisticated and larger thrust engines
has resulted in larger unbalance forces and more diversified
operating conditions. This has increased the transmitted
vibration, and challenged design engineers to create even
more effective mounting system designs. One useful
method for helping the design engineer to develop more
effective mounting systems is through optimization Figure 1. A Typical Turboprop Engine Installation.
techniques.
In the case of aircraft engine mounting systems, the design
Previous studies on the optimization of engine mounting objectives are more versatile. Figure 1 shows the
systems have focused mainly on automobile applications. installation configuration for a typical turboprop engine.
Demic [l] optimized ride comfort and controllability of The engine is attached to the aircraft structure (nacelle) via
passenger vehicles by minimizing vertical and angular three elastomeric mounts. The power of the engine is
oscillations. In other studies, Bernard and Starkey [2], and transmitted to the reduction gear box and propeller through
Spieckerman, et a1 [31, attempted to reduce large pulleys and belts. These components and their unbalance
excitation operate at different frequencies. The rotational
unbalances create dynamic force in the mounts which are
f Member AIAA, 1635 W. 12th Street, Erie, PA 16514. directly transmitted to the airframe. Also, because the
available space is very small, the motion of the engine
Copyright@ 1991 by the American Institute of Aeronautics and
Astronautics. Inc. All riehts reserved.
 must be limited within a certain envelope. This places a Based on Eq. (I), the stiffness matrix of an elastic, or
constraint on the maximum allowable deflection of the viscoelastic, mounting in its local coordinate system can
engine to static, or very low frequency, forces. The two be written as:
design goals of minimizing the transmitted forces, and
limiting the engine deflection to maximum allowable
values, form the basis for the constrained optimization
problem studied in this paper. The problem is solved
using a recursive quadratic programming technique based
on constrained a variable metric approach [4].
This stiffness matrix must be transformed from the local
mount coordinate system to the global coordinate system
situated at the engine C.G. This is done through the
following linear transformation:
The equations of motion for the engine mounting system
model are derived in this section. The engine is modelled
as a rigid body with six degrees of freedom, which consist
of a single translation and rotation around each of the X-Y- where A, is the transformation matrix from the local
Z global coordinate axes. The engine is supported by an
arbitrary number of mounts, which have arbitrary coordinate system of the ith mount to the global
Downloaded by Freie Universitaet Berlin on January 9, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.1991-1102

positions and orientations with respect to the engine center coordinate system. The transformation matrix can easily
of gravity (C.G.), as shown in Fig. 2. The origin of the be formed from the orientation of the mount with respect
global coordinate system is located at the engine center of to the global coordinate system, which can be defined
gravity (C.G.). using Euler angles, Bryant angles, direction cosines, etc.
[5l.
To express the equations of motion in terms of the
displacements and rotations at the engine C.G., another
transformation is necessary. This transformation relates
the displacement of each mount to the displacements and
rotations of the engine C.G. Assuming that the engine
displacements are small enough that they can be considered
as infinitesimal displacements, the transformation of the
engine displacements to the translational displacements at
the mounting point i can be written as:

Forces and Torques

4
Figure 2. A Suspension System of an Aircraft Engine where u i is the translational displacement vector at the
+
mounting point i and r i is the position vector of mount i
The majority of mounts used in engine mounting + +
applications are of a rubber bonded to metal, or with respect to the engine C.G. Ut and Ur are the
elastomeric, construction. Elastomeric materials behave respective translational and rotational displacement vectors
viscoelastically, and for this reason, a complex spring of the engine C.G. In matrix form, Eq. (4) can be
stiffness is used to model the dynamic behavior of the rewritten as:
mount. The complex stiffness of a mount in the X, Y,
and Z directions of its local coordinate system is defined
by the equation:

where I is the (3 x 3) identity matrix, and Fi is the (3 x 3)

where k' and k" are the resistive and dissipative spring matrix defined for a vector cross product operation [ 5 ] . In
terms of the mount displacement, the viscoelastic force
rates, q is the loss factor, and j = fi Rotational vector vi transmitted from mount i to the engine is:
stiffness of the mounts are ignored. Another interesting
characteristic of elastomeric mounts is that they generally
possess a higher dynamic stiffness than static stiffness.
To account for this, a dynamic to static spring ratio equal
to k'lk is defined. A mount constructed from a purely
elastic material will have a loss factor of zero, and a
dynamic to static spring ratio of unity.
where KTis the dynamic complex stiffness matrix of
mount i. The moment vector, ti, that results from the
 application of the viscoelastic forces of mounting i on the
engine C.G. is:

Combining Eq.(6) and (7) yields

O O m O 0
where ei is defined as a generalized viscoelastic loading M = / 0 0 0 1- Ixy Iu
vector resulting for mount i that encompasses both forces
and moments at the engine C.G. in all six degrees of
freedom. By summing all of the viscoelastic loadings
Downloaded by Freie Universitaet Berlin on January 9, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.1991-1102

from a total of n mounts, the total viscoelastic loading

vector on the engine C.G. can be expressed as:
where m is the mass of the engine, and Iij are the elements
of the inertia tensor.

The static equilibrium equations of the engine can be

developed by setting the velocity and acceleration terms in
Eq. (12) equal to zero:

T T
Next, we define U=(U t ,U r)T as a generalized coordinate
vector which represents the X, Y, and Z displacements and where Kst is the static stiffness matrix of the mounting
rotations of the engine C.G. system, and fst is the generalized static loading.
For elastic mounts possessing viscous damping, a viscous
damping element is also included as an analysis option . Solution to the Eauations of Motioq
Using a similar assembly procedure to the mount loading
vector, the total damping loading vector on the engine Two types of excitation are considered, low frequency or
C.G. resulting from a total of m dampers is: static forces and torques, and dynamic forces and torques.

m o n s e to &tic and Maneuvering Loads

The deflections of the engine C.G. to static and

maneuvering conditions can be calculated from the
solution of Eq. (16):
where Di is the damping matrix of a viscous damper i.
With all of the component reactive forces derived in terms
of the generalized coordinates in Eq.(9) and (lo), the
dynamic equations of motion for the engine can now be Four types of static and maneuvering excitation conditions
expressed using Newton's second law: exist, forces due to engine weight, torques due to engine
rotation, "gee" forces due to aircraft maneuvers, and thrust
forces due to the engine's thrust. The equivalent
generalized static loading force at the engine C.G. for a
weight force W applied at the engine C.G. is:
where M is the generalized mass matrix, d is the
generalized damping force defined in Eq. (lo), and e is the
generalized mount force defined in Eq.(9). The vector f
represents the external excitation force vector applied to
the engine. Substituting Eq.(9) and (10) into Eq.(ll)
results in:

MIJ + DU + (K' + jK")U = f (12)

 where W g i , Wg2, and W g 3 are the direction cosine angles
between the gravitational direction vector and the engine
global coordinate system. The equivalent maneuvering
force for a torque applied to the engine is:
where cP s are the angles of rotary unbalance j with respect
to the X, Y, and Z axes, cj is the position vector of the
rotary unbalance j, Tj is the maximum amplitude of the
rotary unbalance j, 6.l is the excitation frequency, and g is
the gravitational constant.

In the case of force excitatim, for k forces fl ,f2, - .,fk

where 1 is the magnitude of the torque, and the direction having phase angles @ l , @ ~ , - - . ~ @ kwith respect to some
cosines of the torque with respect to the engine global time reference and acting on the engine at locations
coordinate system are equal to Vii, W12, and W13. For a Pi, P2,---,P k ; the total excitation is in the same form as
constant thrust forcef applied to the rigid body engine at in Eq.(23) except that:
Downloaded by Freie Universitaet Berlin on January 9, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.1991-1102

location p ,the generalized loading vector becomes:

f't = (;f)
The equivalent maneuvering force for a constant
acceleration excitation of "n" gees is:

nW cos yr,, \
nW cosyr,, The optimization process consists of determining a set of
nW cosyr,
elastomeric mount characteristics that minimize the
transmitted forces through all of the mounts when the
0 system is subjected to excitations of different amplitudes,
0 phases, and frequencies. The optimal mounting system
0 ) must also satisfy inequality constraints on the maximum
allowable static deflection at the engine C.G., and/or some
other critical locations.
EpS~onse to Dv Mathematically, the optimization problem can be
Unbalances expressed as:

The response of the engine C.G. to dynamic inputs can be minimize @(X> X E Rn (26)
calculated through the solution of Eq. (12). In the subject to C, (X) 2 0 k E I{l ,...,c} (27)
frequency domain, this can be found from the complex
matrix inversion:
where is the objective function, X is a design parameter
vector consisting of mount stiffnesses and/or orientations,
Ck(X) is a vector containing the inequality constraint
equations, and c is the total number of constraints.
The dynamic forces transmitted through the mount can
then be calculated using Eq. (6). Two types of dynamic The objective function in Eq. (26) is related to the
forces act on the engine, rotational unbalances and weighted sum of the magnitude of the transmitted dynamic
dynamic forces. In the case of a rotational unbalance forces, and encompasses the forces across all of the
acting at a single frequency, the generalized dynamic force mounts in all directions, for all forcing conditions.
vector is: Weighting factors are used to represent the importance of
each forcing condition, mount number, and mount
direction. The dynamic forcing conditions include cruise
and takeoff, and can represent different forces at different
frequencies and phases. The exact form of the objective
function is:

where,
 where ncond is the number of excitations, nm is the Table 1. Engine Inertia (lb-in-sec2)
number of mountings, Ss are the weighting factors, and fs
are the dynamic forces in each mount, in each direction, in
the global coordinate system. The constraint equations are
of the form:

where C i is the ith constraint equation, lUstli is the Table 2. Mounting Locations (in)
absolute value of the ith actual static deflection of the
engine, and (Umax)i is the ith maximum allowable static X Y z
deflection of the engine. Since both the objective function C.G. 0.00 0.00 0.00
and constraint equations are nonlinear, a nonlinear Mount 1 23.63 8.20 -11.79
constrained optimization technique must be used to solve Mount 2 23.63 -8.20 - 1 1.79
for the optimal solution. A non-linear optimization
program called CVMOl [6] has been used to solve the Mount 3 -11.66 -8.20 -10.74
above optimal design problem. , Mount 4 -11.66 8.20 -10.74
Downloaded by Freie Universitaet Berlin on January 9, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.1991-1102

Table 3. Mounting Orientation (Bryant Angles

in Degree)
The configuration of a two cylinder piston engine, with a
utility engine located in between the two cylinders, is ex 9~ ez
shown'in Fig.3. The engine weighs 1660 Ib, and its mass Mount 1 0 90 0
moment of inertias are shown in Table 1. The center of
Mount 2 0 90 0
gravity of the engine, as well as the utility engine, is
located at (0,0,0) of the global coordinate system. The Mount 3 1 0 90 0
center of gravity of the two cylinders are located at (0,1,0) , Mount 4 1 0 90 0
and (0,-1,0), respectively. The original design of the
suspension system consists of four identical axisymmetric Two maneuvering conditions, along with the weight, are
mounts with static spring rates of Kx=Ky=Kz=8757 lb/in. considered for the constraints on the maximum deflection
The dynamic-to-static spring rate ratio is 1.2, and the loss of the engine subjected to low frequency excitation. The
factor is 0.06. The location and orientation of each mount first maneuvering condition is a "gee" force acting at the
is listed in Tables 2, and 3, respectively. Bryant angles engine C.G. in the X-direction. The second maneuver is a
are used to represent the orientation of the mounts with torque of 1000 in-lbs in the negative X-direction. The
respect to the global coordinate system. The design maximum allowable deflections for these maneuvering
variables are the stiffnesses of the mounts and these four conditions are
mountings must be identical. It is also required that the
mounts to be axisymmetric, i.e. Ky = Kz in the mount's I~~110.03, IU,1<0.03. IuUl<0.08 (in)
local coordinate. (30)
[~,1<0.02. ~ 0 . 1 0 Iun1<0.02
, (rad.)

Two engine cruise modes are considered for the

Cyllnder 2 minimization of the transmitted forces. In the first cruise
mode, the engine is running at 40 Hz, while in the second
cruise mode the engine is running at 80 Hz. The utility
Cyllnder X engine runs with the same frequency as the engine.
-
Suecific details of the two dynamic forcing conditions are
listed in Tables 4 and 5.

Table 4. Forcing Condition of Cruise Mode 1

** 1 I Magnitude I Phase I Position I Direction

Figure 3. A Two Cylinder Engine

Utility 0.5 0 (O,O,O) (0,O.l)

Cylinder
 Table 5. Forcing Condition of Cruise Mode 2 one with very stiff mounts, were optimized. For the first
starting point, each mount had an axial static stiffness of
I Magnitude I Phase I Position 1 Direction 1 4000, and a radial stiffness of 2000 lblin. The initial and
optimized systems for this soft starting point are shown in
Table 8. Notice that the objective function for the
1 I I I I optimized system is actually larger than that of the
Cylinder 0.25 original system. The reason for this is that the starting
CI
0 (O.-l.O) (O,O,-l) stiffnesses are so soft that the static deflections at the
L

Utility 1.00 0
engine C.G. violate the constraints in Eq.(30). This
Cylinder
(O,O,O) (0,O.l) starting point is not feasible (it doesn't satisfy the
constraints). The optimized system has the lowest value
of the objective function that satisfies the constraints.
Table 6. Comparison Between Original and
Optimized Systems Table 7. Deflection at the Engine Center of
Gravity
Original Optimized
System System UU "ty ue Urx wry un
Objective Function 0.493 0.099 Static -0.0084 0.0000 -0.0435 0.000 -0.0429 0.0000
Downloaded by Freie Universitaet Berlin on January 9, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.1991-1102

KX (Axial Stiff.) I 8757 I 10635 0.16G in X 0.0055

1OOO in-lbs 0.0000
0.0000
0.0039
0.0013
0.0000
0.000 0.0129 0.0000
-0.020 0.0000 0.0005
K v & K 7 (Radial Stiff.) 1 8757 22322

From Table 8, it can be seen that while the axial stiffness

I
Natural
Frequencies (Hz) I 15.74
16.21
18.36
25.55
is almost identical as the first optimized axial stiffness,
the optimal radial stiffness decreased dramatically from
22,322 to 2418 lblin. Also, the optimized objective
function is lower than the first optimized system. From
Table 9, two constraints are active for this optimized
The weighting factor for the first dynamic forcing system, the rotation around the X-axis in the second
condition at 40 Hz. is 0.5, while the weighting factor for maneuvering condition, and the displacement in the X
the second dynamic forcing condition is 1.0. direction in the first maneuvering condition.

The original design and corresponding optimal solution is

shown in Table 6. From this table, it can be seen that the Table 8. Comparison Between Original and
objective function of the optimized system is Optimized Systems
approximately one fifth the objective function of the (Soft Starting Point)
original system. One reason for this dramatic
improvement is that one of the natural frequencies of the Original Optimized
original system is 39.94, which is very close to an input System System
forcing frequency of 40.0 Hz. The optimization Objective Function I 0.030 I 0.093
successfully lowered the dynamic forces in the mounts by Kx (Axial Stiff.) I 4000 10650
changing the mount stiffnesses. This resulted in a K y & K z (Radial Stiff.) 2000 I 2418
movement of the natural frequencies of the system away 5.10 6.58
from the resonant condition at 40 Hz. 6.46 1 7.82
Natural 7.94 9.18
In any constrained optimization, it is important to see
Frequencies (Hz) 9.82 21.28
which constraints limit the performance of the system.
These constraints are called active, or binding, and can be 13.63 25.24
located where the static deflection equals its corresponding 21.01 27.82
maximum allowable deflection value. Knowledge of the
active constraints is important because only these
constraints limit the system's performance. Small Table 9. Deflection at Engine Center of
changes in the values of the maximum allowable Gravity
deflection for the other constraints will have no effect on (Soft Starting Point)
the optimal solution. The deflections of the optimized
system at the engine C.G. are shown in Table 7. It can be
seen that the rotation around the X axis in the second
maneuvering condition is equal to its constraint value.

Because the objective function to be minimized is non-

convex, multiple local minima might be present. To see The second starting point had a very high axial stiffness of
if multiple local minima were present for this problem, 20,000, and a radial stiffness of 100,000 lb/in. The
two other starting points, one with very soft mounts and
 second starting point and its corresponding optimum
system are shown in Table 10. Also, from Table 11, it
can be seen that the rotation of the engine C.G. in the X The simulation and numerical optimization of aircraft
direction of the second maneuvering condition is the only engine mounting systems are demonstrated using an
active constraint. The interesting result here is that a very interactive computer program called Sixopt. The rigid
stiff mount can deliver equal performance to a very soft body engine is supported by elastomeric mounts on a rigid
mount (by objective function comparison). One problem, base. Vibratory forces transmitted through the elastomeric
however, with a stiffer mounting system is that the mounts under certain engine excitation conditions are
lumped parameter six-degree of freedom mounting system minimized by adjusting the stiffness and orientation of the
model may no longer be valid. As the stiffness of the engine mounts. Inequality constraints on the maximum
mounts increases, so does the natural frequency of the allowable deflection of the engine to static and
system, and wave effects become apparent due to the maneuvering conditions must also be satisfied. The
system behaving like a distributed system. constrained variable metric optimization technique solves
for the optimum design. The technique has been
demonstrated successfully on an example.
Table 10. Comparison Between Original and
Optimized Systems
(Stiff Starting Point)
Downloaded by Freie Universitaet Berlin on January 9, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.1991-1102

Original Optimized The authors of this paper would like to express their
appreciation to Prof. R.W. Mayne of SUNY at Buffalo for
his assistance in this work.

20000 10568
References
[l] M. Demic, "A Contribution to the Optimization of the
Natural I 25.81 I 18.92 I Characteristics of Elastodamping Elements of Passenger
Cars," Vehicle System Dynamics, Vol. 19, 1990, pp. 3-
Frequencies (Hz) 53.72 57.38
18.
65.14 68.71
124.01 131.24 [2] J.E. Bernard, and J.M. Starkey, "Engine Mount
Optimization," SAE paper No. 830257.

Table 11. Deflection at Engine Center of [3] C.E. Spiekermann, CJ. Radcliffe, and E.D. Goodman,
Gravity "Optimal Design and Simulation of Vibrational Isolation
(Stiff Starting Point) Systems," ASME Journal of Mechanisms, Transmissions,
and Automation in Design, Vol. 107, June 1985, pp. 271-
276.

[4] J. Zhou, and R.W. Mayne, "Monotonicity Analysis

and Recursive Quadratic Programming in Constrained
Optimization," ASME Journal of Mechanisms,
It can be seen from the three runs described above that the Transmissions, and Automation in Design, Vol. 107, Oct.
optimization technique is very reliable for this application. 1985, pp. 459-462.
A solution was found for each starting point that
successfully lowered the transmitted forces in the mounts, [5]E.J. Haug, Computer-Aided Kinematics and Dynamics
while satisfying the static deflection constraints. One of Mechanical Systems, Volume I: Basic Methods, Allyn
problem yet to be solved, however, is that of locating a and Bacon, Needham Heights, MA 02194.
global minimum instead of the nearest local minima. In
the hope of locating a global minimum, it is suggested [6] J.H. Wang, and H.Z. Wang, "The Theory and Analysis
that more than one starting point be med. of the Program CVMO1," Mechanical Engineering
Department, Hua Zhong University of Science and
For this numerical example, the two later cases yielded a Technology, Wuhan, China, 1985. (in Chinese).
better solution. For the starting point with a low mount
stiffness, the resulting optimum system had mounts with
a low radial stiffness, and lower natural frequencies. For
the starting point with a high mount stiffness, the
optimum system had mounts with a high radial stiffness,
and higher natural frequencies.