Downloaded from SAE International by David Neihguk, Friday, March 09, 2018

A Particle Swarm Optimization Tool for Decoupling 2014-01-1687

Automotive Powertrain Torque Roll Axis Published 04/01/2014

Sameer U. Kolte, David Neihguk, Abhinav Prasad, Samir Rawte, and Aditya
Gondhalekar
Mahindra & Mahindra, Ltd.

CITATION: Kolte, S., Neihguk, D., Prasad, A., Rawte, S. et al., "A Particle Swarm Optimization Tool for Decoupling
Automotive Powertrain Torque Roll Axis," SAE Technical Paper 2014-01-1687, 2014, doi:10.4271/2014-01-1687.
Copyright © 2014 SAE International

Abstract passenger compartment, viz. the steering wheel, seats, floor,

gear knob, pedals etc. Hence, a typical powertrain mounting
A typical powertrain mount design process starts with
system is expected to provide good isolation in engine
performing the system calculations to determine optimum
operating conditions like key on, key off, idle shake, run up etc.
mount parameters, viz. position, orientation and stiffness
To satisfy the demand of future vehicles in terms of enhanced
values to meet the desired NVH targets. Therefore, a 6
occupant comfort and high power to weight ratio, effective
degrees of freedom lumped parameter system of powertrain
powertrain mounting systems are necessary. To achieve this,
and mounts is modelled in Matlab®. The approach is to
different types of engine mounts have been developed such as
decouple the torque roll axis mode from the remaining five rigid
rubber mounts, hydraulic mounts, active mounts etc. Different
body modes so that the response to the torque pulses is
optimization techniques have been implemented to find
predominantly ‘oscillations about Torque Roll Axis’. This is
optimum mount parameters viz. stiffness, damping, positions
achieved by optimizing the above mount parameters within
and orientation to minimize the transmitted vibrations. This
specified constraints so that ‘Rotation about the torque roll axis’
paper presents an optimization tool using a well known Particle
is one of the natural modes of vibration. The tool developed
Swarm Optimization algorithm[5] implemented to decouple the
here uses ‘Particle Swarm Optimization(PSO) algorithm’
powertrain torque roll axis.
because of its ease of implementation and better convergence
to the solution. The algorithm is programmed in TK solver®.
If the above NVH considerations of powertrain mounts are not
Further, for the given torque input, the harmonic response of
considered at the early design stage during vehicle architecture
the powertrain mounted on optimized mounts, is evaluated for
development, then the space protected for mounts is usually
comparison with NVH targets.
purely based on system interfaces. Later when the NVH
targets are set, it is difficult for the designer to achieve them
This paper presents the design criteria to develop the objective
with available package space, which leads to compromise in
function for optimization of automotive powertrain mounts. A
NVH performance. During the detailed design, repeating the
detailed procedure of determining problem constraints from
NVH testing on prototypes by changing the mount parameters
specified NVH targets and system interfaces is explained. The
is extremely tedious and a time consuming process. This gets
optimization is done for powertrain mounts of a 625cc single
even worse, when the designer wants to standardize the
cylinder Load carrier Diesel engine and a 1.2L three cylinder
powertrain mounts across different vehicle platforms. The
passenger vehicle Diesel engine. The results obtained from the
aforesaid put forward a requirement of a tool which can
developed tool are compared with NVH simulation results of
optimize the mount parameters with all available constraints to
baseline mount configuration defined by the supplier.
meet the NVH targets.

Introduction
Torque Roll Axis Decoupling
The automotive powertrain mounting system usually consists
When the torque is applied along an axis which is not
of either three or four elastic or visco-elastic mounts that are
coincident with any of the principle axes of a three dimensional
installed in between the powertrain and the body frame. The
rigid body which is either unconstrained or mounted on very
pulsating torque generated by the engine is one of the major
soft springs, it oscillates about a resultant fixed axis known as
sources of vibration in an automobile. Vibrations transmitted to
the Torque Roll Axis (TRA). In case of automotive powertrain,
the chassis are transferred to the tactile interfaces in
the torque is applied about the crankshaft axis which rarely
 Downloaded from SAE International by David Neihguk, Friday, March 09, 2018

coincides with any of the principle axes. The vibrations 2. A system will have a mode about the torque roll axis if and
transferred to the vehicle body can be substantially reduced if only if
the Powertrain is made to oscillate predominantly about the
Torque Roll Axis. This method is called Torque Roll Axis
decoupling. (2)

When a system is completely decoupled with respect to 6 rigid where [K] is the system stiffness matrix, [TRA] vector
body modes, then the excitation along a particular direction represents unit rotation around the torque roll axis and [T] is
produces response along the same direction. This method is input torque vector. [TRA] can be defined as
called elastic axis decoupling and can be best explained with
equation (1).
where xtra, ytra, ztra represent the direction of cosines of the
torque roll axis.

The significance of equation (2) is that the powertrain will have

mode about TRA (pure rotation about TRA) only if the
summation of reaction forces exerted by mounts on the
powertrain is equal to zero and the net reaction moment
exerted by mounts is parallel to the input torque T.

Further, T. Jeong and R. Singh [1] have proposed and proved

the following axiom to prove that TRA decoupling is the only
(1) method to obtain constant directional response.

The above equation represents a 6 degree of freedom, 3. If a rigid-body system excited by a force (torque or
undamped system. Since the stiffness matrix is diagonal, moment) with a constant direction, responds in a constant
hence the above equation breaks down to 6 single DOF translational (rotational) directional and varies only with the
equations each having response in the direction of the applied excitation frequency, then that response must be one of the
force. Therefore the system is said to be completely natural modes of vibration.
decoupled. Physical significance of diagonal [K] matrix is that
the elastic center coincides with the point of application of force Hence, TRA mode should be one of the natural modes of
which is CG. However, considering 3-D asymmetric shape of a vibration.
typical powertrain it is impossible to achieve the elastic axis
decoupling. TRA decoupling is a very effective practical Mathematically this can be expressed as equation (3),
alternative to this.

The primary load acting on the powertrain is the pulsating (3)

torque produced by the engine. It should be noted that at low
RPM, the amplitude of response produced by this torque is where [M] is the mass matrix and ω is the eigenvalue
high and can be felt. If this response is coupled i.e. the motion corresponding to rotation about TRA.
of powertrain is distributed along more than 1 DOF, then poor
NVH performance is obtained. Looking back at the definitions
Optimization Problem Formation
of the torque roll axis decoupling, it can be inferred that, for the
pulsating torque input, if the powertrain is made to oscillate From equation (2), [K][TRA] represents a vector of reaction
only about TRA, then it behaves as a free rigid body and forces and moments exerted by the mounts on the powertrain
vibrations transferred to the frame are considerably reduced. when the powertrain is oscillating about the torque roll axis. To
achieve complete TRA decoupling, all elements of [K][TRA]
should be equal to zero except the element representing
Methodology reaction moment parallel to input torque. The objective function
The main objective to achieve TRA decoupling is to adjust is formed on this basis.
powertrain mount parameters such that the powertrain
oscillates only about TRA when pulsating torque is applied.
Geck P. and Patton R. [2] have stated following axioms with Formation of Objective Function
necessary mathematical proofs. Since the mounts are generally not oriented along Vehicle
co-ordinate axes, the compressive and shear stiffness values
1. The torque pulses will excite one and only one mode if and of mounts are required to be transformed to vehicle co-ordinate
only if the mode shape of that mode is rotation around the system by using transformation matrix as shown in equation
torque roll axis. (4).
 Downloaded from SAE International by David Neihguk, Friday, March 09, 2018

The problem statement is to minimize fobj by optimizing mount

stiffness values, mount positions and mount orientations as
design variables.

(4)
Constraint Definition
where θ, Ψ and Φ represent the angles by which the axis of The important part of an optimization problem is to define the
powertrain mount is rotated about y, x and z axis respectively. constraints. For the problem under consideration, constraints
have to be imposed on design variables, viz. mount stiffness,
The stiffness matrix [Kg] for the ith mount in vehicle co-ordinate positions and orientations.
system is obtained as per equation (5).

Mount Position and Orientation Constraints

These are determined by the package space available for the
(5) mounts. The interfacing systems for the powertrain mounts
include the engine, the transmission, the exhaust system, the
The procedure to form the system stiffness matrix in vehicle cooling system, the chassis or the body frame, fuel and
coordinate system is explained by T. Jeong and R. Singh [1]. electrical lines, the battery etc. The lower and upper bounds for
The system stiffness matrix is given by equation (6), mount positions and orientation are defined such that adequate
clearances are maintained from above interfaces.

These constraints are also influenced by the effect of interfaces

on the performance of engine mounts. For example, consider
(6) that high temperature parts like the engine cylinder case, the
catalytic convertor or the exhaust muffler are near to the rubber
where [B]i represents the rotation matrix. mounts. If this causes the rubber engine mounts to operate
beyond the permissible temperature range, then it may cause
creeping of rubber and deteriorate the performance.

(7) Mount Stiffness Constraints

These constraints are defined by requirements of the natural
where xg, yg, zg represent distances of ith mount from frequency of modes, maximum allowable displacement of the
powertrain CG in vehicle co-ordinate system. powertrain, static deflection of powertrain, feasibility of rubber
design etc.
The reaction force and moment matrix is evaluated as

Static Deflection of the Powertrain

Static deflection of the powertrain is determined by the stiffness
offered by mounts in z-direction. Constraint equation (11) for
the same can be defined as,

(8)

where Rx, Ry, Rz represent the reaction forces and RMx, RMy, (11)
RMz represent the reaction moments exerted by mounts when
the powertrain oscillates about the torque roll axis. where kzi represent the z-stiffness of ith mount, Wt is the weight
of the powertrain and zst is the maximum allowable static
The objective is to minimize all elements of RM except the displacement of the powertrain.
element representing the reaction moment parallel to the input
torque in order to achieve TRA decoupling. i.e.
Angular Displacement about TRA
It is necessary to constrain the amount by which the powertrain
oscillates about the torque roll axis, if excitation frequency
(9)
coincides with the TRA natural frequency. Due to oscillations
The objective function (10) can be defined assuming input beyond permissible limit, the powertrain may interfere with the
torque about y direction as, interfacing systems during dynamic conditions. Also
performance of interfacing systems might get affected.

(10)
 Downloaded from SAE International by David Neihguk, Friday, March 09, 2018

The degree by which the powertrain oscillates is controlled by

defining a constraint on roll stiffness about TRA as shown in
equation (12).

(12)

For a 100% torque roll axis decoupled system, since the TRA
mode is one of the natural modes of vibration, the amplitude of
response at resonant frequency will be very high for an
undamped system. In order to predict the correct response of
the system, damping coefficients of mounts must be known. If
damping coefficients of mounts are taken into account, then
the equation of motion for the system can be written as,

Figure 1. Displacement Transmissibility

(13)
It can be observed that vibration isolation (Transmissibility < 1)
Equation (14) gives the amplitude of response, can only be obtained when frequency ratio (ω/ωn) is greater
than 1.414. Therefore, the frequency of TRA mode should be
kept less than or equal to ‘excitation frequency/1.414’ i.e.
7.91/1.414= 5.59 Hz for above example.
(14)
TRA mode being one of the natural modes of vibration, we can
where [To] is the amplitude of input torque vector. define constraint as per equation (16).

In this case, the response will be finite at resonance because

damping is considered. (16)

Frequency of TRA Mode Feasibility of Mount Design

Usually in all powertrain mount optimizations, focus is to place The minimum or maximum values of stiffness that can be
all the natural frequencies of the powertrain below or above the achieved also depend on limitations of rubber design. A
excitation frequency of engine. However, for a single cylinder requirement of feasible and durable rubber mounts may
or a two cylinder engine where the idle excitation frequency is impose necessity of certain ratio of compressive stiffness to
generally low, it may not be possible to keep all natural shear stiffness that needs to be maintained. Further, weight
frequencies below that value. In such cases, the mount and cost targets may put limitations on values of stiffness
parameters are adjusted such that none of the natural which can be achieved. One needs to keep these factors in
frequency lies within a certain range of the excitation mind while deciding lower and upper bounds for mount
frequency. For example, consider a single cylinder engine stiffness values.
having low Idle RPM equal to 950rpm. The excitation
frequency is calculated by using equation (15).
Particle Swarm Optimization Algorithm
Once the objective function for the optimization is generated in
Matlab, it is inserted into Particle Swarm Optimizer
(15) programmed in TK solver. The algorithm uses number of
variables called particles which update their positions based on
Since 1/2 order is the torque sensitive order for a single their personal best and global best position. [5]
cylinder engine; using equation (15), fexc=7.91Hz. Hence, none
of the natural frequencies of powertrain should lie within a 1. Initially a population of agents called particles is created and
certain range of 7.91Hz. random coordinates and velocities are assigned to each of
them.
The frequency of TRA mode is important because for a TRA 2. For every particle co-ordinate p, value of the objective
decoupled system, rotation about TRA is the natural mode and function (fitness) is calculated.
response of the system for all other modes is minimum(ideally 3. If fitness for p is better than the best fitness for that particle
zero). Consider the transmissibility curve for different damping so far, then pbest = p.
coefficients as shown in Figure (1). [10]
4. Best among pbest is determined and is set as ‘gbest’.
5. Particle velocities are updated according to equation (17a)
towards ‘pbest’ and ‘gbest’.
 Downloaded from SAE International by David Neihguk, Friday, March 09, 2018

6. Particle position is updated according to equation (17b). Also, the constraints on bounce stiffness and TRA mode
7. Above steps are repeated until stopping criteria is met. frequency were defined as equations (18a) and (18b),

(17a) (18a)

(17b) (18b)

where w is the inertia factor, c1 is the cognitive coefficient, c2 Table 4. 625cc 1-cylinder - mount position bounds (mm)
is the social coefficient, U1 & U2 represent a random number
between 0 and 1. v represents velocity, x represents position of
particle, pb and gb represent the personal and global best
respectively. Table (1) lists the values of parameters used for
the optimization.

Table 1. Particle Swarm Optimization parameters

Mount axes are oriented parallel to global vehicle axes.

Therefore, θi, Ψi and Φi are equal to 0. After running the
optimization, the optimized mount stiffness and positions were
obtained as shown in Table (5) and (6),

Examples Table 5. 625cc, 1-cylinder - Optimized mount stiffness (N/m)

The optimization was done for powertrain mounts of a 625cc

single cylinder transverse mounted engine of a load carrier and
1.2L three cylinder transverse mounted engine of a passenger
vehicle. The optimized parameters are compared with baseline
mount configuration proposed by powertrain mount supplier.

Table 6. 625cc 1-cylinder - Optimized mount positions(mm)

625cc Single Cylinder Engine Powertrain
The powertrain data is summarized in Table (2),

Table 2. 625cc 1-cylinder engine powertrain data

The figure (2) shows the amplitude of response of the

powertrain for the input torque. Since damping is not
considered, the amplitude of response has very huge value at
resonance. It is clear that for a given input torque the response
is pure rotation about TRA. Thus, TRA mode is completely
decoupled.

The upper and lower bounds on stiffness and positions were

Comparison with Baseline Mount Configuration
defined as shown in Table (3) and (4),
Modal analysis was done for both baseline and optimized
Table 3. 625cc, 1-cylinder powertrain - mount stiffness bounds (N/m) mount configuration. The natural frequencies of the system and
percentage modal kinetic energy distribution were calculated
as shown in Table (7) and (8).

The percentage modal kinetic energy distribution represents

the percentage distribution of total kinetic energy along
different DOFs for any given mode. It is calculated from mode
shape matrix by using equations (19a), (19b) and (19c).
 Downloaded from SAE International by David Neihguk, Friday, March 09, 2018

Figure 2. 625cc 1-cylinder - Amplitude of response for optimized mount

configuration, response is pure rotation about the torque roll axis.

Table 7. 625cc 1-cylinder - kinetic energy distribution(%) for optimized

Figure 3. 625cc 1-cylinder - comparison between baseline
mount parameters (TRA co-ordinate system), rot y represents the
positions(Blue spheres) and TRA optimized mount position cubes)
rotation about TRA.

1.2L Three Cylinder Engine Powertrain

The powertrain data is summarized in Table (9).

The upper and lower bounds on stiffness and positions were

defined as shown in Table (10) and (11),

Table 9. 1.2L 3-cylinder engine powertrain data

Table 8. 625cc 1-cylinder- kinetic energy distribution(%) for baseline

mount configuration (TRA co-ordinate system), rot y represents the
rotation about TRA

Table 10. 1.2L 3-cylinder - mount stiffness bounds (N/mm)

(19a)

(19b)

Also, the constraints on bounce stiffness and TRA mode

frequency were defined as equations (20a) and (20b),
(19c)
Table 11. 1.2L 3-cylinder - mount position bounds (mm)
where, KE represents the modal kinetic energy of system.
Using element by element matrix multiplication the kinetic
energy per mode is split into its contributors (KEDOF). KED is
the kinetic energy distribution.

Figure (3) shows comparison between baseline mount

positions and optimized mount positions.
 Downloaded from SAE International by David Neihguk, Friday, March 09, 2018

Figure (5) shows comparison between baseline mount

(20a) positions and optimized mount positions.

Table 14. 1.2L 3-cylinder - kinetic energy distribution (%) for optimized
mount parameters (TRA co-ordinate system), rot y represents the
(20b) rotation about TRA.

Mounts axis are oriented parallel to global vehicle axes.

Therefore,

(21)

After running the optimization, optimized mount stiffness and

positions were obtained as shown in Table (12) and (13),

Table 12. 1.2L 3-cylinder - Optimized mount stiffness (N/mm) Table 15. 1.2L 3-cylinder - kinetic energy distribution(%) for baseline
mount configuration (TRA co-ordinate system), rot y represents the
rotation about TRA

Table 13. 1.2L 3-cylinder - Optimized mount positions (mm)

The amplitude of response of the powertrain for the input

torque is shown in Figure (4). From graph, it is clear that for a
given input torque the response is pure rotation about TRA.
Thus TRA mode is completely decoupled.

Comparison with Baseline Mount Configuration

Modal kinetic energy distribution for optimized mount
parameters was calculated and compared with that of baseline
mount configuration as shown in Table (14) and (15),

Figure 5. 1.2L 3-cylinder - comparison between baseline mount

positions (Blue spheres) and TRA optimized mount positions (Red
cubes)

Conclusions
It is seen that complete TRA decoupling is achieved when the
reaction moment exerted by mounts is parallel to the input
pulsating torque and TRA mode is one of the natural modes of
vibration. This criteria is used to develop the procedure to
optimize mount parameters, viz. stiffness, positions and
orientations. Systematic mathematical formulations of the
problem constraints is done by considering NVH targets,
Figure 4. 1.2L 3-cylinder - Amplitude of response for optimized mount
system interfaces and feasibility of design.
configuration, response is pure rotation about the torque roll axis.
 Downloaded from SAE International by David Neihguk, Friday, March 09, 2018

Using Particle Swarm Optimization algorithm, the problem is 8. Erdelyi, H., Roesems, D., Toso, A., and Donders, S.,
solved for 625cc single cylinder engine and 1.2L three cylinder “Powertrain Mounting System Layout for Decoupling Rigid-
engine powertrain. The results of optimization are compared Body Modes in the Vehicle Concept Design Stage,” SAE
with performance of the baseline mount configuration. It is Technical Paper 2013-01-1706, 2013, doi:10.4271/2013-
observed that within the same package and stiffness 01-1706.
constraints, the optimization tool gives a completely decoupled 9. Wagh, N. and Deshmukh, R., “Engine Mounting
roll mode about TRA. Optimisation To Counter 0.5 Order Engine Vibrations,” SAE
Technical Paper 962492, 1996.
The results establish the usefulness of the tool to the
10. Liu C. Q., Huston Ronald L., “Principles of Vibration
powertrain mount designer at various design phases. The tool
Analysis, With Applications in Automobile Engineering”,
can be used at early design stages when usually the mount
SAE International, Warrendale, PA, ISBN 978-0-7680-3339-
supplier is not involved, to protect space for powertrain mounts.
7, 2011, doi:10.4271/R-395.
During later design stages, when assembly, service and
operator ergonomics aspects of vehicle design are faced, the 11. Rao, Singiresu S., “Mechanical Vibrations”, 4/E. Pearson
designer can perform iterations to determine optimum mount Education India, ISBN 978-81-7758-874-3, 2003.
parameters which would give the best decoupling possible
satisfying the above aspects. The tool has application in Contact Information
standardization of engine mounts across different vehicles. By Mr. Sameer Ulhas Kolte
keeping the stiffness values constant, the designer can Mahindra and Mahindra Ltd.
optimize the mount positions and orientations according to the [email protected]
packaging constraints of different vehicles.

References
1. Jeong T. and Singh R., “Analytical Methods of Decoupling
the Automotive Engine Torque Roll Axis”, Journal of Sound
and Vibration, 234(1), 85-114 (2000).
2. Geck, P. and Patton, R., “Front Wheel Drive Engine
Mount Optimization,” SAE Technical Paper 840736, 1984,
doi:10.4271/840736.
3. Bernard, J. and Starkey, J., “Engine Mount Optimization,”
SAE Technical Paper 830257, 1983, doi:10.4271/830257.
4. Park, J. and Singh, R., “Effect of Engine Mount Damping
on the Torque Roll Axis Decoupling,” SAE Technical Paper
2007-01-2418, 2007, doi:10.4271/2007-01-2418.
5. Eberhart, Russell, and Kennedy James, “A new optimizer
using particle swarm theory,” In Micro Machine and
Human Science, 1995. MHS'95., Proceedings of the Sixth
International Symposium on, pp. 39-43. IEEE, 1995.
6. Ford, D., “An Analysis and Application of a Decoupled
Engine Mount System for Idle Isolation,” SAE Technical
Paper 850976, 1985, doi:10.4271/850976.
7. Brach, R., “Automotive Powerplant Isolation Strategies,”
SAE Technical Paper 971942, 1997, doi:10.4271/971942.

The Engineering Meetings Board has approved this paper for publication. It has successfully completed SAE’s peer review process under the supervision of the session
organizer. The process requires a minimum of three (3) reviews by industry experts.

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical,
photocopying, recording, or otherwise, without the prior written permission of SAE International.

Positions and opinions advanced in this paper are those of the author(s) and not necessarily those of SAE International. The author is solely responsible for the content of the
paper.

ISSN 0148-7191

http://papers.sae.org/2014-01-1687