Bcg\`\mTg\ba bY 8aZ\aX @bhag 6[TeTVgXe\fg\Vf Hf\aZ

8kcXe\`XagT_"Ah`Xe\VT_ 4aT_lf\f

MAHMOUD S. FOUMANI
AMIR KHAJEPOUR
&GRCTVOGPV QH /GEJCPKECN 'PIKPGGTKPI 7PKXGTUKV[ QH 9CVGTNQQ 9CVGTNQQ 1PVCTKQ 0. )
%CPCFC 'OCKN CMJCLGRQWT"WYCVGTNQQEC

MOHAMMAD DURALI
&GRCTVOGPV QH /GEJCPKECN 'PIKPGGTKPI 5JCTKH 7PKXGTUKV[ QH 6GEJPQNQI[ 6GJTCP +TCP
(Received 8 August 2001; accepted 5 December 2001)

#DUVTCEV : In this paper an experimental/numerical technique is developed for engine mount optimization.
The method is general and can be applied to optimize active and passive vibration isolators or absorbers in
any mechanical systems or civil structures. Engine mount optimization techniques mostly rely on an accurate
mathematical model of the whole vehicle, which in most cases is not available or is too difficult to develop.
As a result, the current approach for selecting engine mounts for a vehicle is based upon trial and error which
is very time-consuming and expensive. The proposed technique counts upon experimental data for optimiza-
tion and does not require any mathematical model of the vehicle or its components. The required experiments
are similar to the current trial-and-error based experiments performed on a vehicle for mounts selection. The
method is evaluated experimentally using a quarter car model and the results corroborate the proposed opti-
mization method.

-G[ 9QTFU Engine mount, transfer path analysis, optimization, noise, vibration, harshness

$! <AGEB7H6G<BA

In the past 20 years, ride comfort and noise cancellation have found an especial importance
in automotive research and development. Since fuel efficiency is one of the fundamental
objectives in current automotive design, there is a strong trend toward weight reduction
and lighter cars. Powerful engines with lighter and more flexible bodies lead to increasing
vibration and noise in passenger cabins. Thus, design and optimization of anti-vibration
devices have been the subject of many recent studies in the field of noise, vibration, harshness
(NVH).
One of the most important vibration isolators in vehicles is the engine mount. It transmits
the engine vibrations to the body, and the road and chassis vibrations to the engine. Because
the nature of vibration isolation from the engine to the chassis is different than that of the
road to the engine through the chassis, designing engine mounts needs an especial attention.

,QWTPCN QH 8KDTCVKQP CPF %QPVTQN 9: 1121–1139, 2003 DOI: 10.1177/107754603030697

f 2003
? Sage Publications

Downloaded from jvc.sagepub.com at Selcuk Universitesi on December 30, 2014

1122 M. S. FOUMANI ET AL.

Engine disturbances resulting from unbalance forces and torques are in high frequency and
low amplitude while road disturbances are in low frequency and high amplitude. Therefore,
an engine mount must have different characteristics depending on frequency and amplitude
to be able to reduce undesired oscillations. Basically, the selection and design of an engine
mount is based upon trial and error and there is no systematic approach for optimizing mount
characteristics for a given vehicle. Because of increased rate of changes in automotive
industry, design and optimization must be performed in a short period of time and, as a result,
an automated design technique for finding an optimum engine mount is highly desirable (Yu
et al., 2001).
For optimization, an engine mount system is usually divided into three main parts:
engine, engine mount, and foundation. The engine is modeled as a rigid mass with six
degrees of freedom (Brach, 1997; Suh and Smith, 1997) and its equations are written in
Cartesian coordinates. It is common to consider engine mounts as the main parts that connect
the engine to the foundation (Brach, 1997; Suh and Smith, 1997). Few articles (Radcliffe
et al., 1983) consider other connections such as hoses, drivetrains, and the exhaust system as
important vibration paths. Each mount is usually modeled as a spring-damper system with
two parameters, stiffness coefficient and corresponding hysteresis or viscous damping. The
foundation is also modeled as a rigid body and its vibration is excluded from the optimization
algorithm. However, Lee et al. (1995) and Kim et al. (1994) discuss the effects of chassis
flexibility on the dynamic response of an engine mount system. Using analytical/simulation
approach they show that the effect of the chassis flexibility on the mounting forces transmitted
from the engine to the chassis is considerable when the vibration modes of the chassis are
coupled with the engine-mount system. It is mentioned that this situation is more probable
in low frequencies.
In an optimization problem, the most important parts are the selection of an objective
function and defining the system constraints. The usual choice for an objective function is
tuning the natural frequencies of mounts to some desired range to avoid resonances (Brach,
1997; Arai et al., 1993; Geck and Patton, 1984). Since, near a natural frequency, both the
engine displacement and transmitted force from the engine to foundation increase, some
authors express that minimization of the transmitted force has the same effect as tuning
the natural frequencies (Ashrafioun, 1993; Oh et al., 1991). Bornard and Starkey (1983)
have used an optimization technique to select the mount design variables including rate,
location and orientation to move the natural frequencies of an engine mounting system from
a selected region. The technique is based upon representing the engine mounting system
mathematically or experimentally, and minimizing a penalty function that includes the system
natural frequencies and the boundaries of the undesirable band.
Bretl (1993) presents a new simulated-based method which seeks minimal vibration
response regardless of the engine mounting system modes. He selects the vibration perceived
by vehicle occupants as its objective function without considering the powertrain rigid-body
modes. Minimizing power transmission vibrations is another objective function used by
Royston and Singh (1996). They consider vibratory power flow from the mount into the
support structure and minimize it over a frequency range.
Another part of an optimization problem is the definition of constraints. Two types of
constraints have been considered in the literature: (1) constraints for limiting parameters
range (Suresh et al., 1994); (2) constraints for limiting system outputs such as engine
displacement (Yu et al., 2001).

Downloaded from jvc.sagepub.com at Selcuk Universitesi on December 30, 2014

 OPTIMIZATION OF ENGINE MOUNT CHARACTERISTICS 1123

Suresh et al. (1994) proposed a nonlinear unconstrained optimization procedure to

optimum design variables of engine mounts for minimizing transmitted force to the chassis.
The method is capable of incorporating any stiffness, orientation/location and natural
frequency constraints. The method relies on a mathematical model of the engine and its
mounting system, and is limited to minimizing transmitted forces to the chassis through the
mounts.
Muller et al. (1994, 1996) have studied mount optimization to minimize noise inside the
passenger compartment using transfer path analysis. In this technique, the vibration transfer
paths from the engine to the mounts are measured in three directions using accelerometers
and the noise transfer paths from the mounts to the passenger compartment are measured
by an indirect technique using a sound source in the passenger compartment and measuring
the mounts accelerations. In this technique, the cross vibration and noise transfer paths are
ignored and it is assumed that the transfer paths are decoupled.
Matsuda et al. (1987) have investigated the optimum mount configuration in a diesel
engine for idling and cranking through a computer simulation. Using a rigid-body mass
for the engine and considering the force oscillation as the excitation force in the equations
of motion, they have optimized the mount configuration. Hata and Tanaka (1987) have
considered both rate and location optimizations of the mounts of a front wheel drive vehicle.
In their technique, the mounts are modeled with only springs and the transmitted force to
the chassis is approximated using the stiffness matrix of the engine mount system. The
transfer functions between possible locations on the engine and the seat track are measured
experimentally. A simple search technique is then used to find the best 15 locations for the
mounts. Another search is used within the 15 locations to find the optimum mounts rates.
The literature in this area indicates that there is not a wealth experience on optimization
of engine mounts considering all effective parameters. Most authors have ignored the
effects of chassis oscillations in their optimization. Also, their approaches rely on accurate
mathematical models for different vibration paths in a vehicle.
The main goal of this work is to introduce a method and procedure for optimizing
the characteristics of vehicle engine mounts using an experimental analysis. The proposed
method is independent of the mount type and any of rubber, hydraulic or even active mounts
can be considered in the optimization algorithm. In the following, we introduce a general
optimization technique for systems comprised of many subsystems in which some of them
can be changed. In a vehicle system for example, the engine mounts are the subsystems
whose characteristics can be changed or selected to minimize the noise inside the cabin or
vibration of the engine or chassis. We apply this procedure to an experimental one-quarter
vehicle model to find the optimized mount for given objective functions.

%! @4G;8@4G<64? @B78?<A:

In general, a linear system is comprised of some linear subsystems which interact with each
other or external systems as shown in Figure 1. The way a system is divided into subsystems
depends on the application, as different subsystems can be joined together to make a single
subsystem. All subsystems have some inputs and outputs with which they transfer power
among themselves or external systems. The number of subsystems that each subsystem can
interact with changes from one to the number of all subsystems. It is important to note that

Downloaded from jvc.sagepub.com at Selcuk Universitesi on December 30, 2014

1124 M. S. FOUMANI ET AL.

Figure 1. A general system with its building subsystems.

each subsystem must interact with at least one. Otherwise, it cannot be a member of the main
system. In Figure 1 all inputs from one subsystem to another are shown with a vector that
may have several elements. This rule is also applied to the outputs of subsystems.
To show the connections between the subsystems, we define the EQPPGEVKXKV[ OCVTKZ Fqq
where q is the number of subsystems. The Flm element of the connectivity matrix is set to
4 if subsystem l interacts with subsystem m, and it is set to 0 if it does not. It is clear that
Flm @ Fml and therefore the connectivity matrix is symmetric. The diagonal elements of the
connectivity matrix show how the main system in Figure 1 interacts with external systems.
The Fll element takes a value of 4 if subsystem l has an external output and 3 if it does not.
To define a formulation for optimizing the characteristics of adjustable or changeable
subsystems in the framework shown in Figure 1, we need to obtain the transfer functions
from all external inputs to all internal and external outputs. Note that any internal input to a
subsystem is an internal output from another subsystem.
To formulate the problem, we consider a portion of the main system as seen in Figure 2.
Using the superposition law, the equation for output vector R lm can be written as (see Figure 2)

[
q
R lm @ J lmn L ln > Fln 9@ 3 (1)
n@4

Downloaded from jvc.sagepub.com at Selcuk Universitesi on December 30, 2014

 OPTIMIZATION OF ENGINE MOUNT CHARACTERISTICS 1125

Figure 2. Inputs and outputs of a subsystem.

[
q
@ J lmn L ln . J lml L ll > Fln 9@ 3> n 9@ l (2)
n@4

where vector L ln and matrix J lmn indicate the input from subsystem n to subsystem l, and the
transfer function from input L ln to output R lm , respectively. It is clear that if subsystem n does
not interact with subsystem l, there is no input from subsystem n to l and therefore Fln @ 3=
As seen in Figure 2, we have

L lm @ R ml > l 9@ m= (3)

L ll and R ll are used to show external inputs and outputs for subsystem l, respectively.
Substituting equation (3) into equation (1), we obtain

[
q
R lm  J lmn R nl @ J lml L ll > Fln 9@ 3> n 9@ l= (4)
n @4

In equation (4) it is assumed that external input L ll and all J lmn are known and the output(s)
R lm is unknown.
Similar equations can be written for all internal and external outputs of the system to
result in enough equations for finding all the unknowns. The assembled equations yield to

Downloaded from jvc.sagepub.com at Selcuk Universitesi on December 30, 2014

1126 M. S. FOUMANI ET AL.

JR@L (5)

where
3 4 45 4q 4
^J `q4 q4 ^`q4 q5  ^`q4 qq
E 5 5q F
E ^J `q5 q5  ^`q5 qq F
J@E
E .. ..
F
F (6)
C V|p= . . D
q
^J `qq qq
3 4
4
^R `q4
E F
E ^R `5 F
E q5 F
R@E . F (7)
E .. F
C D
q
^R `qq

3 4
4
^L `q4
E F
E ^L `5 F
E q5 F
L@E . F= (8)
E .. F
C D
q
^L `qq

In the above equations ql is the number of internal and external outputs of subsystem l and
4
^J `, ^`, ^R `, and ^L ` can be obtained directly from equation (4). For example, ^J `q4 q4 is
3 4
4 J 445    J 44q
E 3 J 455    J 45q F
E F
@E F=
4
^J `q4 q4 E .. .. .. .. F (9)
C . . . . D
3 J q5    J 4qq
4

Note that in equation (9) it is assumed that subsystem 4 interacts at least with subsystems
5 and q, and also it has one external output. Please note that subsystem 1 does not need to
4
interact with all subsystems and the size of matrix ^J `q4 q4 depends only on the total number
of internal and external outputs. The 4s in the connectivity matrix in row l show which
l l l
subsystems should be included in ^J ` , ^R ` , and ^L ` .
4 4
Similarly, using equation (4), ^R `q4 and ^L `q4 can be written as
3 4 4 3 4 4
R4 J 44
E R5 F E J4 F
E 4 F E 54 F 4
^R `q4 @ E F E F
4 4
E .. F > ^L `q4 @ E .. F L 4 (10)
C . D C . D
R q4 J 4q4

Downloaded from jvc.sagepub.com at Selcuk Universitesi on December 30, 2014

 OPTIMIZATION OF ENGINE MOUNT CHARACTERISTICS 1127

provided as before, subsystem 1 interacts at least with subsystem 5 and q, and has an external
l
output and input. In general, we can express the no element of ^J `ql ql shown by j lno in a
more compact form as
; ;
? Flo 9@ 3
l
? J no o 9@ l
l
j no @ 4 n @ l> o @ l > Fln 9@ 3 = (11)
= =
3 otherwise n> o 5 ^4==q`

The elements of ^` in equation (6) are all zero except for the corresponding element of
45 4q
R lm for l 9@ m and Flm 9@ 3 in equation (4). For example, ^`q4 q5 and ^`q4 qq in equation (6)
are
3 4
3 3  3
E 4 3  3 F
45 E F
^`q4 q5 @ E .. .. . . .. F (12)
C . . . . D
3 3  3

3 4
3 3  3
E .. .. .. F
4q E . F
^`q4 qq @E . . F= (13)
C 3 3  3 D
4 3  3

lm
In general, we can express the no element of ^`ql qm shown by  lmno in a more compact form
as
;
 ? Flo 9@ 3
lm 4 n @ m> o @ l
 no @ > Fln 9@ 3 = (14)
3 otherwise =
n> o 5 ^4==q`
The solution of equation (5) is

R @ J 4 L (15)

which will be used later to minimize an objective function for selecting the adjustable or
changeable subsystems or their parameters. The homogeneous form of equation (5) is

J R@3 (16)

where it has a nontrivial solution if matrix J is singular or

mJ m @ 3= (17)

Matrix J is only a function of $ if all its elements are known at each frequency. The
solution of equation (17) provides the main system’s natural frequencies. When some of

Downloaded from jvc.sagepub.com at Selcuk Universitesi on December 30, 2014

1128 M. S. FOUMANI ET AL.

the subsystems transfer functions or their parameters are yet to be selected (such as engine
mounts in a vehicle), equation (17) can be used to remove or relocate main system’s natural
frequencies from a desired frequency range.
To this end, we have developed a formulation to relate the internal and external outputs
to the main system inputs, and individual subsystem transfer functions. The subsystem
transfer functions do not need to be in analytical forms and they could be represented in
experimental frequency response functions (FRFs). In practice however, separating a system
into its subsystems and performing experiments on each to obtain its FRFs is very time-
consuming and in some cases even impossible. For example, to optimize engine mounts a
vehicle can be divided into: (1) engine; (2) engine mounts; (3) the rest of the vehicle including
chassis, body and tires. For calculating all the point and cross FRFs of the chassis from the
wheels to each mount, and also from each mount to itself and other mounts, we need to
separate the mounts completely and excite the chassis from the mount locations in different
directions. This, in many cases, is difficult and imposes many limitations on the technique.
In Section 4, we introduce an experimental procedure to obtain subsystems FRFs without
separating them from each other.

&! B5=86G<I8 9HA6G<BA

The most important part in any optimization process is the selection of an objective function.
This function must be scaled and shows the desired behavior of the system that needs to be
optimized. Generally, an objective function for the system shown in Figure 1 depends on
internal and external outputs. Therefore, a general form of an objective function  can be
expressed as

 @ +R , (18)

where R is defined in equation (7). The outputs R are obtained from equation (15) and they
are functions of subsystems transfer functions including fixed and adjustable ones. As a
result, equation (18) can be written as
 
 @  J i l{ > J dgm > $ (19)

where J i l{ and J dgm represent fixed and adjustable transfer functions. The objective function
 can be calculated at any $ for known J i l{ and given J dgm. Therefore, equation (19) is, in
fact, a function of $ and J dgm or
 
 @  J dgm > $ = (20)

We can now minimize equation (20) at each frequency to find the FRFs of the adjustable
subsystems J dgm. Calculating J dgm by optimizing  at each frequency results in an optimum
response for all frequencies. This is a suitable approach for designing active subsystems
such as active engine mounts. In many cases, the structures of the adjustable subsystems
are known and with the proposed technique we can select their parameters for an optimum
response. Since the parameters of the adjustable subsystems are fixed the objective function

Downloaded from jvc.sagepub.com at Selcuk Universitesi on December 30, 2014

 OPTIMIZATION OF ENGINE MOUNT CHARACTERISTICS 1129

Figure 3. Quarter vehicle model for engine mount optimization.

cannot be minimized at each frequency and it should be redefined to represent the overall
response of the system for a range of frequency. For such cases, in which the mean value of
the objective function is important we define  as

] $5
 
 @  J dgm > $ g$
$4

] $5
@  +S4 > S5 > ===> Sn > $, g$ (21)
$4

where  is a measure of a desired response of the main system at each frequency, S4 > ===> Sn
are the constant parameters of the adjustable subsystems, and $4 and $5 are the desired
frequency range. For instance, to design an engine mount for a quarter vehicle model shown
in Figure 3, only the mount transfer functions or J 5 are adjustable and both the engine and

Downloaded from jvc.sagepub.com at Selcuk Universitesi on December 30, 2014

1130 M. S. FOUMANI ET AL.

chassis are fixed. Hence, optimizing  +J 5 > $, at each frequency to calculate J 5 results in
an optimum response for all frequencies provided an active mount is used. However, if the
mount is passive and only its stiffness, N, and damping, F, are unknown,  can be defined as
] $5
@  +N> F> $, g$ (22)
$4

where its solution results in optimum selection for N and F for the desired frequency range.
The function  in equation (22) can be, for example, a weighted summation of the vibration
amplitudes of the engine and chassis. When the maximum value of  is more important than
its mean value, we can define the objective function, , as

  
 @ pd{  J dgm > $

$ 5 ^$4 > $5 ` = (23)

'! 8KC8E<@8AG4? 9E8DH8A6L E8FCBAF8 9HA6G<BAF

Thus far, we have laid out a method for optimizing adjustable subsystems within a general
system. The optimization is performed at each frequency or a frequency range as shown in
equations (20) and (21) depends on the subsystem structure. The transfer functions of fixed
subsystems can be obtained from their mathematical models if they exist, or experiments
using modal analysis.
Optimization of a system when its mathematical model exists is not the focus of this
work, and here we only consider complex systems whose models do not exist or it is too
expensive to develop. Therefore, the FRFs in equation (5) need to be derived experimentally.
One simple solution is to separate all subsystems and derive all their FRFs. However, this
is not practical in most cases because of the complexity, cost, or time considerations. In this
section, we introduce a practical option for obtaining all transfer functions without separating
the subsystems. To explain the technique, we consider a quarter car model shown in Figure 3.
In this model, there are three subsystems: chassis, mount, and engine. The mount is
assumed to be adjustable and needs to be designed for an optimum response. Therefore, we
can assume that it can be separated easily from the system and tested to obtain its FRFs. Also,
the engine is considered rigid with known inertia matrix. Thus, only the FRFs of the chassis
are unknown. Since the chassis interacts only with the mount and has one external output
4
therefore, if ^J `q4 q4 in equation (9) represents the chassis, q4 becomes 5 and hence there
4 4
are two unknown FRFs in ^J ` and two in ^L ` , which need to be found before being able to
optimize the mount.
Note that, although we can separate the mount from the chassis, however, in practice,
exciting the chassis from the mount location in different orientations considering the small
space between the chassis and engine is not feasible. Instead, we test the complete system
in Figure 3 by considering a random or chirp input for the road disturbances and measure all
the internal and external outputs.

Downloaded from jvc.sagepub.com at Selcuk Universitesi on December 30, 2014

 OPTIMIZATION OF ENGINE MOUNT CHARACTERISTICS 1131

If equation (4) is written for the two outputs of subsystem 1 and if the equations are
rearranged for solving the four unknown FRFs, we will have two equations with four
unknowns assuming all the outputs are measured. To obtain two more independent equations
at each frequency the same test can be repeated if the mount is replaced with another one
or the engine inertia matrix is changed. As long as the overall dynamics of the system is
altered by different known adjustable subsystems or by any change in the inertia matrix of
rigid-body subsystems, independent equations can be obtained from the experiments. Note
that changing the engine inertia matrix is as easy as adding a block of mass to it. If change
of inertia technique is used, the same experimental arrangement can be used for obtaining
enough equations to find all the unknown FRFs. Both techniques (using known subsystems,
and changing inertia matrix of rigid-body subsystems) are usually applicable for finding
unknown FRFs. In summary, the procedure is as follows.
1. The system is excited through its external inputs and only the inputs and outputs of the un-
known subsystems are measured. In some cases indirect measurement can be performed
for some inputs and outputs. For example, in Figure 3 the input force to the chassis and
engine from the mount is equal. The force can be calculated indirectly from the mount’s
FRF by measuring the displacements at both sides of the mount.
2. The dynamics of the system is altered by: (1) replacing one of the adjustable subsystems
with another or (2) changing the inertia of rigid-body subsystems.
3. Parts 1 and 2 are repeated to have enough linearly independent equations to obtain all
unknown FRFs. The assembled equation for the outputs of the subsystem with unknown
FRFs using equation (4) can be written as

Dps [s4 @ Ep4 (24)

where p and s are the number of independent equations and unknowns respectively, vector
[ comprises of unknown FRFs, and matrix D and vector E include all the measured data.
4. If the number of equations is the same as the number of unknowns. i.e. p @ q, we have
a unique solution which is

[ @ D4 E= (25)

Otherwise, if p A q a pseudo inverse can be used to find the unknown [

[ @ D. E (26)

where D. is

D. @ +DW D,4 DW = (27)

To improve the accuracy of the results and reduce the effects of noise and errors in the
experimental data, the following considerations can be applied:

Downloaded from jvc.sagepub.com at Selcuk Universitesi on December 30, 2014

1132 M. S. FOUMANI ET AL.

Table 1. Experimental setup specifications.

P4 @ 4=863 Kg O4 @ ;=3 cm Width @ 6=3 cm
P5 @ 3=3<8 Kg O5 @ 48=8 cm Width @ 5=8 cm
P6 @ 3=6;3 Kg O6 @ 8=3  45=3 cm Width @ 6=3 cm

Table 2. Inputs and data collection specifications.

Data acquisition range 0–125 Hz
Source frequency range 5–37 Hz
Resolution 0.125 Hz
Source type Burst noise
Burst length 50%
Number of averages 15

1. increase the number of independent tests;

2. apply inputs that are similar to the actual inputs in terms of amplitude and frequency to
reduce the effects of nonlinearity in the results.

(! 8KC8E<@8AG4? FGH7L

In this section, we apply the technique to a quarter model of a vehicle shown in Figure 4. The
elements of this model include suspension, chassis, engine, engine mount, and steering wheel
that is modeled by a one-degree-of-freedom system. Since minimizing the steering wheel
vibrations is one of the objectives usually used in the design of engine mounts, its model is
included in the model. In this experiment, we assume that all elements of the system are
fixed and no modifications are allowed except for the engine mount that is represented by a
spring. Therefore, the purpose is to find an engine mount (N) that minimizes: (1) steering
wheel vibrations, and (2) chassis vibrations. The limitations in the selection of the engine
mount stiffness will appear as constraints in the optimization problem.
To realize the model shown in Figure 4, an experimental setup illustrated in Figure 5 was
fabricated. We used aluminum strips instead of springs for ease of stiffness change in the
setup. The lengths of the beams shown by O4 , O5 , and O6 correspond to the stiffness of the
suspension system N4 , steering wheel N5 , and engine mount N6 , respectively. The mass of
the chassis, steering wheel and engine are shown by P4 , P5 , and P6 . We assume that except
for O6 or N6 all other parameters are fixed. Table 1 shows the thickness, width and length of
the beams and the mass of P4 , P5 , and P6 used in the setup.
The road disturbances are applied to the system through base excitation [3 . The
displacements of the chassis, engine, and steering wheel are shown by [4 , [5 , and [6 and
are measured by accelerometers attached to the bodies. The specifications of the input (road
disturbances) applied to the experimental setup are shown in Table 2.
Since the steering wheel subsystem does not interact with the engine, we can divide the
system into three subsystems as shown in Figure 6. The engine is assumed to be rigid and
therefore only the four FRFs of subsystem 4 from road disturbances [3 and I 54 to [5 and
[ 54 are unknown. Using the change-of-subsystem technique described in Section 4, we can

Downloaded from jvc.sagepub.com at Selcuk Universitesi on December 30, 2014

OPTIMIZATION OF ENGINE MOUNT CHARACTERISTICS 1133

Figure 4. Vehicle quarter model.

Figure 5. Experimental setup.

Downloaded from jvc.sagepub.com at Selcuk Universitesi on December 30, 2014

1134 M. S. FOUMANI ET AL.

obtain the unknown FRFs by selecting different N6 for the engine mount. In this way, we can
eliminate the need for separating physically the system into its subsystems for measuring the
unknown FRFs. We further calculate the forces applied to P4 and P6 indirectly using the
displacements of the masses and the stiffness of the steering wheel N6 . In this way, we can
eliminate the need for any force sensors in the experiments.
 ==  two FRFs obtained from the experiments when O6 @ 8 cm. Figure 8
Figure 7 shows
 
illustrates how  [== 6  changes as O6 varies. As explained before, to reduce errors in calculating
[ 3
unknown FRFs the number of independent tests must be increased. In the experiment, instead
of two minimum independent tests (two different lengths for O6 ) we performed nine tests with
O6 varied from 8 to 45 cm. Using equation (26) we found the unknown FRFs of subsystem 4
and, to evaluate the accuracy of the calculated FRFs, we separated subsystem 4 and measured
its FRFs directly. Figure 9 compares the direct and indirect measurement of subsystem 4 FRFs
for mJ 444 m and mJ 454 m. The agreement between the direct and indirect measurements is very
acceptable and shows that indirect calculation of unknown transfer functions can be used in
the optimization procedure.
For the objective function , since the mount is passive and its structure is known
equation (21) should be used. The objective function that we select for the quarter vehicle
model is the integral of the amplitude ratio of steering wheel vibrations between $4 @ 8 Hz
and $5 @ 48 Hz or
] $5  == 
 [5 
@  ==  g$= (28)
$ 4  [3 

 == 
 
Having known all the FRFs in equation (15),  [== 5  can be calculated for any given O6 and,
[ 3
therefore, the optimization problem becomes

]  == 
$5  
 [5 
Min. @  ==  g$
$4  [3 
] $5
@  +O6 > $, g$ (29)
$4

Subject to: Oplq  O6  Opd{

where Oplq and Opd{ are assumed to be Oplq @ 8 cm and Opd{ @ 45 cm.
The global minimum of equation (29) can be obtained using simplex or any gradient-
based optimization techniques. However, for more complex systems with several free
parameters and constraints, a global optimization technique such as a genetic algorithm
should be used.
The solution of the optimization problem in equation (29) results in

O6 @ ;=6 cm= (30)

Downloaded from jvc.sagepub.com at Selcuk Universitesi on December 30, 2014

OPTIMIZATION OF ENGINE MOUNT CHARACTERISTICS 1135

Figure 6. Subsystems of the experimental setup.

Figure 7. FRFs when u ' D cm.

Downloaded from jvc.sagepub.com at Selcuk Universitesi on December 30, 2014

1136 M. S. FOUMANI ET AL.

Figure 8. Variation of FRFs for different u .

Figure 9. Comparison between direct and indirect FRF measurements.

Downloaded from jvc.sagepub.com at Selcuk Universitesi on December 30, 2014

 OPTIMIZATION OF ENGINE MOUNT CHARACTERISTICS 1137

Figure 10. Optimization results for steering wheel vibration minimization.

Figure 11. Optimization results for chassis vibration minimization.

Downloaded from jvc.sagepub.com at Selcuk Universitesi on December 30, 2014

1138 M. S. FOUMANI ET AL.

Figure 10 compares the effect of different mount stiffness (N6 or O6 ) on the vibration
of the steering wheel. As seen in the figure, the value of O6 obtained from the optimization
procedure shows significantly lower vibrations for the steering wheel.
Since all the transfer functions are known, other objective functions can be considered for
the mount optimization. For example, if the chassis vibrations are concerned, the objective
function can be defined as
] $5  == 
 [4 
@  ==  g$ (31)
$ 4  [3 

where its minimization with the same constraints on the length of O6 and frequency range
results in O6 @ :=3 cm. Figure 11 compares the effect of different mount stiffness (N6 or O6 )
on the chassis vibrations. As seen in the figure, the value of O6 found from the optimization
procedure shows significantly lower vibrations for the chassis.

)! 6BA6?HF<BAF

In this paper, we have provided a practical technique for mount optimization. The technique
has been developed in a general form that can be used in any system where some of its
elements are needed to be adjusted or fine tuned for a better response. Active and passive
vibration absorbers and isolators in any mechanical systems or civil structures can be designed
and optimized with this technique. We have applied the method to a quarter car model and,
without separating the system into its subsystems, we optimized the mount for two different
objective functions: minimizing (1) steering wheel vibrations and (2) chassis vibrations. In
both cases, the results showed a significant reduction in the level of steering wheel and chassis
vibrations. Currently, the technique is being evaluated on a real vehicle.

#EMPQYNGFIOGPVU 6JG CWVJQTU YQWNF NKMG VQ CEMPQYNGFIG VJG HKPCPEKCN UWRRQTV QH VJG 0CVWTCN 5EKGPEGU CPF
'PIKPGGTKPI 4GUGCTEJ %QWPEKN QH %CPCFC VJG VGEJPKECN UWRRQTV QH VJG 08* EQPVTQN U[UVGO ITQWR QH %QQRGT
5VCPFCTF #WVQOQVKXG CPF RQUKVKXG FKUEWUUKQPU YKVJ 2TQHGUUQT ) / )NCFYGNN

E898E8A68F
Arai, T., Kubozuka, T., and Gray, S., 1993, ‘‘Development of an engine mount optimization method using modal para-
meters,’’ 5#' 6GEJPKECN 2CRGT 5GTKGU 932898.
Ashrafioun, H., 1993, ‘‘Design optimization of aircraft engine mount,’’ ,QWTPCN QH 8KDTCVKQP CPF #EQWUVKEU 115, 463–
467.
Bornard, J. E. and Starkey, J. M., 1983, ‘‘Engine mount optimization,’’ 5#' 6GEJPKECN 2CRGT 5GTKGU 830257.
Brach, R. M., 1997, ‘‘Automotive powerplant isolation strategies,’’ 5#' 6GEJPKECN 2CRGT 5GTKGU 971942.
Bretl, J., 1993, ‘‘Optimization of engine mounting systems to minimize vehicle vibration,’’ 5#' 6GEJPKECN 2CRGT 5GTKGU
931322.
Geck, P. E. and Patton, R. D., 1984, ‘‘Front wheel drive engine mount optimization,’’ 5#' 6GEJPKECN 2CRGT 5GTKGU
840736.
Hata, H. and Tanaka, H., 1987, ‘‘Experimental method to drive optimum engine mount system for idle shake,’’ 5#'
6GEJPKECN 2CRGT 5GTKGU 870961.

Downloaded from jvc.sagepub.com at Selcuk Universitesi on December 30, 2014

 OPTIMIZATION OF ENGINE MOUNT CHARACTERISTICS 1139

Kim, J., Jho, S., and Yim, H., 1994, ‘‘Influence of chassis flexibility on dynamic behavior of engine mount systems,’’
5#' 6GEJPKECN 2CRGT 5GTKGU 942269.
Lee, J., Yim, H., and Kim, J., 1995, ‘‘Flexible chassis effects on dynamic response of engine mount systems,’’ 5#'
6GEJPKECN 2CRGT 5GTKGU 951094.
Matsudu, A., Hayashi, Y., and Hasegawa, J., 1987, ‘‘Vibration analysis of a diesel engine at cranking and idling modes
and its mounting system,’’ 5#' 6GEJPKECN 2CRGT 5GTKGU 870964.
Muller, M., Weltin, U., Law, D., Roberts, M., and Sieber, T., 1994, ‘‘The effect of engine mounts on the noise and
vibration behavior of vehicles,’’ 5#' 6GEJPKECN 2CRGT 5GTKGU 940607.
Muller, M., Eckel, H., Leibach, M., and Bors, W., 1996, ‘‘Reduction of noise and vibration by appropriate engine mount
systems and active absorbers,’’ 5#' 6GEJPKECN 2CRGT 5GTKGU 960185.
Oh, T., Lim, J., and Lee, S. C., 1991, ‘‘Engineering practice in optimal design of powertrain mounting system for 2.0l ff
engine,’’ in 2TQEGGFKPIU HQT VJG J +PVGTPCVKQPCN 2CEKHKE %QPHGTGPEG QP #WVQOQVKXG 'PIKPGGTKPI.
Radcliffe, C. J., Pickiemann, M. N., and Hlne, D. S., 1983, ‘‘Simulation of engine idle shake vibration,’’ 5#' 6GEJPKECN
2CRGT 5GTKGU 830259.
Royston, T. and Singh, R., 1996, Optimization of passive and active non-linear vibration mounting systems based on
vibratory power transmission,’’ ,QWTPCN QH 5QWPF CPF 8KDTCVKQP 194(3), 295–316.
Suh, C. H. and Smith, C. G., 1997, ‘‘Dynamic simulation of engine-mount systems,’’ 5#' 6GEJPKECN 2CRGT 5GTKGU
971940.
Suresh, N., Shankar, S., and Bokil, V., 1994, ‘‘Development of idealistic hydromount characteristics to minimize engine
induced vibrations using unconstrained minimization,’’ 5#' 6GEJPKECN 2CRGT 5GTKGU 941741.
Yu, Y., Naganathan, N. G., and Dukkipati, R. V., 2001, ‘‘A literature review of automotive vehicle engine mounting
systems,’’ /GEJCPKUO CPF /CEJKPG 6JGQT[ 36, 123–142.

Downloaded from jvc.sagepub.com at Selcuk Universitesi on December 30, 2014