Research Article

Advances in Mechanical Engineering

2016, Vol. 8(4) 1–12
Ó The Author(s) 2016
Analysis of modal frequency DOI: 10.1177/1687814016643640
aime.sagepub.com
optimization of railway vehicle car
body

Wenjing Sun1,2, Jinsong Zhou2, Dao Gong2 and Taiwen You2

Abstract
High structural modal frequencies of car body are beneficial as they ensure better vibration control and enhance ride
quality of railway vehicles. Modal sensitivity optimization and elastic suspension parameters used in the design of equip-
ment beneath the chassis of the car body are proposed in order to improve the modal frequencies of car bodies under
service conditions. Modal sensitivity optimization is based on sensitivity analysis theory which considers the thickness of
the body frame at various positions as variables in order to achieve optimization. Equipment suspension design analyzes
the influence of suspension parameters on the modal frequencies of the car body through the use of an equipment-car
body coupled model. Results indicate that both methods can effectively improve the modal parameters of the car body.
Modal sensitivity optimization increases vertical bending frequency from 9.70 to 10.60 Hz, while optimization of elastic
suspension parameters increases the vertical bending frequency to 10.51 Hz. The suspension design can be used without
alteration to the structure of the car body while ensuring better ride quality.

Keywords
Railway vehicle, car body, modal frequency optimization, sensitivity analysis, equipment suspension parameter

Date received: 5 January 2016; accepted: 16 March 2016

Academic Editor: Rahmi Guclu

Introduction lightweight railway vehicle car bodies. The result shows

Lightweight design of railway vehicles has been com-
monly applied as means to conserve energy. However,
using light materials and altering mechanical structures,
lightweight design can often result in deterioration of
rigidity of the car body. This reduction in the rigidity of
car body is one of the primary causes of low modal fre-
quencies. Li et al.1 established two finite element (FE)
models to compare the effects of lightweight design on
the modal frequencies of the car bodies of railway vehi-
cles. Results indicate that most modal frequencies
decreased to varying degrees as a result of lightweight
design principles. Lower structural modal frequencies
induce strong, flexible vibrations and contribute to
worse ride comfort of passengers.2 Tomioka et al.3 ana-
lyzed three-dimensional (3D) flexural vibrations of
that the vibration of lightweight car bodies with low
modal frequencies occurred in a high-sensitivity range
for human, leading to poor ride comfort for railway
passengers. Therefore, qualitative regulations regarding
the eigenfrequencies of railway vehicle car bodies have
been reported in UIC 5664 and EN 12663,5 and many
countries have specific frequency requirements. For
example, the first vertical bending frequency of a car
1
School of Mechanical Engineering, Tongji University, Shanghai, China
2
Institute of Rail Transit, Tongji University, Shanghai, China
Corresponding author:
Jinsong Zhou, Institute of Rail Transit, Tongji University, No. 4800 Caoan
Road, Shanghai 201804, China.
Email: jinsong.zhou@tongji.edu.cn

Creative Commons CC-BY: This article is distributed under the terms of the Creative Commons Attribution 3.0 License
(http://www.creativecommons.org/licenses/by/3.0/) which permits any use, reproduction and distribution of the work without
further permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/
open-access-at-sage).
2 Advances in Mechanical Engineering

body under service conditions should not be lower than

10 Hz6 according to the technical standards of ICE in
Germany.
Many papers have emphasized on the influence of
low modal frequencies and car body flexibility on the
subsequent vibration of railway vehicles. For example,
the influence of car body flexibility on the vehicle ride
quality was investigated by Zhou et al.,7 including the
first two modes of a Euler–Bernoulli beam model.
Results indicate that high bending frequency can allevi- Figure 1. Original car body FE model with suspended
ate the resonance of the car body. And some previous equipment.
research aims to reduce the flexible vibration of car
body with different control methods, such as Kamada
et al.8 used piezoelectric actuators to suppress bending Table 1. Mass of car body and equipment.
vibration of high-speed railway vehicle and Schandl
Parameter Mass (kg)
et al.9 improved passengers comfort through designing
an active vibration reduction system for flexible car Car body in service condition 39,444
body. However, few studies to date have investigated Electro-pneumatic brake unit 1041
the means by which to improve modal parameters of Charger 350
car body. Wu et al.10 increased the first two modal fre- Inverter 700
Battery container 1164
quencies of automobile car bodies according to sensi-
tivity analysis, which attempts to improve dynamic
parameters by modifying mechanical structures.11
The objective of this article is therefore to investigate Table 2. Material parameters of car body.
specific methods by which to increase the modal fre-
Parameter Value
quencies of railway vehicle car bodies. This article pro-
poses modal sensitivity optimization of railway vehicle Elastic module (MPa) 20,600
based on sensitivity analysis and a new method of Poisson ratio 0.3
equipment suspension design for underneath the car Density (t/mm3) 7.80E29
body. In the latter case, coupling exists between the
flexible car body and the equipment. Simulation studies
and experiments were conducted, and the effectiveness It is assumed that all materials are linear, and that the
of the proposed methods was confirmed. Two methods mass of the car body is equally distributed. The car body
were also compared to determine the most suitable comprises low-carbon steel. The material parameters used
method for use in various situations. in the simulation model are listed in Table 2.
Modal frequencies and shapes of car body are
obtained by modal analysis with block Lanczos
FE model and modal analysis of car body method, which is conducted under free boundary con-
The studied railway vehicle is a passenger car. The car ditions. The three typical mode shapes of the first verti-
body has an integral bearing cylindrical structure with- cal bending, the first lateral bending, and the first
out a central sill and is composed of an end wall, side torsion are shown, respectively, in Figure 2, in which
wall, chassis, head, and an apron board. A car body in the first vertical bending frequency is only 9.15 Hz, the
service condition includes a bearing structure, interior first lateral bending frequency is 11.05 Hz, and the first
materials, and equipment beneath the chassis. The torsion frequency is 13.29 Hz.
equipment primarily includes an electro-pneumatic
brake unit, charger, inverter, and battery system. First,
a 3D FE model was established with FE software Modal test and model correction
HYPERMESH according to its 3D geometric model, Modal test model and equipment
as shown in Figure 1.
Four-node shell elements were used here. All types The FE simulation model requires confirmation by
of equipment are considered to be large mass points experimental data. Therefore, a modal test of the stud-
which connect rigidly to the bottom of the car body at ied car body is implemented to correct the model.
various positions. The mass parameters of the car body Equipment and sensors used in the modal test are listed
and equipment are listed in Table 1. This model has a in Table 3. The data collection system is made up
total of 650,181 elements and 813,117 nodes. of acceleration sensors-BW14596, data collection
Sun et al.
Figure 2. Modal shapes of railway vehicle car body: (a) first
vertical bending, (b) first lateral bending, and (c) first torsion.
front-end-LMS SCR09, and collection software LMS
Test.Lab. There are 72 channels and 1 CAN-bus and
GPS in the collection front-end. The data transmission
rate is 8 M/s. Acceleration sensors all work at three
directions with the sensitivity of 50 mv/ms2. The sam-
pling frequency is 256 Hz.
The following steps are implemented during the test:
1. The geometric size of the car body is measured,
and the positions of data collection points are
determined;
2. Acceleration sensors are fixed at each measure-
ment point in the Y- and Z-directions;
3. Excitation positions are determined and exciters
are installed. The excitation devices connect to
the car body via mandrills, force transducers,
and impedance heads;
Figure 3. Measurement sections of car body.
3
4. The test model and test parameters are input
into the computer running LMS Test.Lab
software;
5. The entire modal test system is debugged;
6. The force excitation signals and acceleration
response signals of each point are recorded in
the time domain during the formal test;
7. Data preprocessing is performed with LMS
Test.Lab. Modal analysis and identification are
conducted to obtain the structural modal fre-
quencies, shapes, and damping ratios of the
studied railway passenger car body.
The measurement points must be selected near the
main body frame. A total of nine sections are identified
along the length of the car body, as shown in Figure 3.
There are eight acceleration measurement points
located in each of the nine sections, which are num-
bered as shown in Figure 4. The directions of each sen-
sor are defined according to the entire vehicle
coordinate. The acceleration values in both the vertical
and lateral directions are recorded.
There are four modal exciters, all of which are
located along the bottom side beam of the car body.
Exciters are installed at two points of car body head
and at two points of car body rear as shown in Figure
3. The four exciters are close to points A1, A3, I1, and
I3, respectively. Figure 5 illustrates one of the modal
excitation systems installed at the head of the car body
(close to A1).
Figure 6 illustrates installation of the acceleration
sensor in detail. All signals from the acceleration sen-
sors and from the force transducers of the exciters are
transferred to a data collection system in the computer
through numerous cables, as shown in Figure 7.
Test conditions
The car body is supported by four very soft rubber
springs with the vertical stiffness of about 320 kN/m in
order to realize free boundary conditions; the modal
test employed is based on the mechanisms of multi-
point excitations and responses. The test is divided into
4 Advances in Mechanical Engineering
Figure 4. Acceleration response point distribution along the car body.
Figure 5. Modal excitation equipment.
Figure 6. Acceleration sensor installation.
Figure 7. Data collection system.
Table 3. Modal test equipment and sensors.
Name Amount Type Precision (%)
Data collection 1 LMS SCR09 0.2
front-end
Acceleration sensor 44 BW14596 1
Force transducer 4 PCB 208C03 1
Modal exciter 4 HEV-1000 –
two parts, in order to account for the vertical and lat-
eral directions. A frequency sweeping input method is
employed in order to excite the car body. All test exci-
tation conditions are listed in Table 4.
Modal test results
The modal identification method based on polymax
method is applied with LMS software in this modal
test. This polymax method is capable of obtaining clear
stabilization diagrams, even in the presence of large
damping and dense modals, so as to achieve a given
physical modal order.
The polymax method requires frequency response
functions (FRFs) at each measurement point. All FRF
amplitudes are then summed in order to centralize the
peak information in a single system. The integrated
FRFs sum to achieve modal parameter identification
of a steady diagram. Poles are then selected based on
the modal selection indicator functions (MIFs) and all
local minimums at all resonance frequencies. Finally,
modal frequencies, modal shapes, and damping ratios
of the body are obtained according to modal vector
estimation of the selected modal poles. Figure 8 shows
the frequency response results of the modal test, which
can be used for obtaining the modal frequencies later.
Table 5 depicts the obtained modal frequency
results. Some difference was observed between simula-
tion and test results, likely because FE modeling
ignores certain interior decoration masses in the service
condition, which must be uniformly distributed in the
Sun et al. 5

Table 4. Modal test conditions.

Method Frequency (Hz) Excitation condition

Burst random excitation 2–40 Vertical and lateral vibrations

Vertical harmonic sweep at two ends 2–40 Vertical vibration
Vertical harmonic sweep at two sides –
Lateral harmonic sweep at two ends 2–40 Lateral vibration
Slow lateral harmonic sweep at two ends –

Figure 8. Test results in frequency domain.

Table 5. Modal analysis correction (Hz). employed. In this case, the modal frequencies optimiza-
tion of a railway vehicle car body based on sensitivity
Modal shape Original Test Correction analysis theory is proposed. The local parts, which pri-
First vertical bending 9.15 9.71 9.70 marily contribute to structural stiffness, are selected as
First lateral bending 11.05 11.69 11.71 the design variables in order to increase the rigidity of
First torsion 13.29 14.20 14.09 the structure.
The dynamic equation of the car body structure is
established as follows
car body. According to the test results, the original
model can be corrected by increasing the equivalent (K  v2i M)ui = 0 ð1Þ
mass and stiffness.
Table 5 indicates good agreement between the new where K is the global stiffness matrix; M is the global
FE model and the modal test, with a maximum error mass matrix; and vi and ui are the i order modal fre-
equal to 0.5%. The newly developed model will be used quency and modal shape vector of the structure,
for the following calculations. respectively.
As indicated by the above results, the first vertical The derivative of equation (1) is calculated as follows
bending frequency of the car body in service condition
is less than 10 Hz, which fails to meet the standard (K 0  2vi v0i M 0 )ui + (K  v2i M 0 )u0i = 0 ð2Þ
criterion.
where v0i is the partial derivative vi sub-optimal vari-
able b (beam thickness); u0i is the partial derivative ui
Modal sensitivity optimization sub-variable b; K0 is the partial derivative K
sub-variable b; and M0 is the partial derivative M
Modal sensitivity analysis sub-variable b.
When dynamic properties are unable to meet related The variable uTi is used to do left multiplication to
requirements, dynamic structural modification is equation (2). Two terms are produced, as follows
6 Advances in Mechanical Engineering

Table 6. Results of frequency and mass sensitivity analysis.

Order Variable Lower limit (mm) Upper limit (mm) Frequency sensitivity Mass sensitivity

1 Longeron 1.00E + 00 1.00E + 01 22.32E206 5.58E202

2 Roof camber beam 1.00E + 00 1.00E + 01 8.24E206 3.32E202
3 Camber beam 1.00E + 00 1.00E + 01 7.77E205 1.23E201
4 Roof boundary beam 1.00E + 00 1.00E + 01 5.71E206 1.73E202
5 Jambe 1.00E + 00 1.00E + 01 8.33E205 2.00E201
6 Window longeron 1.00E + 00 1.00E + 01 5.78E206 1.24E201
7 Upright (thicken) beam 5.00E + 00 1.30E + 01 1.90E207 3.39E203
8 Upper boundary beam 2.00E + 00 1.00E + 01 1.05E205 1.91E202
9 Chassis boundary beam 8.00E + 00 1.50E + 01 3.98E204 3.99E202
10 Chassis boundary beam 6.00E + 00 1.50E + 01 5.40E204 6.21E202
11 Battery hanging beam 6.00E + 00 1.00E + 01 1.42E208 8.09E203
12 Brake hanging seat 6.00E + 00 1.00E + 01 26.95E208 2.58E203
13 Hanging beam 4.00E + 00 8.00E + 00 4.16E204 6.63E203
14 Rib plate 2.00E + 00 6.00E + 00 8.08E208 3.24E203
15 Charger hanging beam 6.00E + 00 1.00E + 01 8.62E207 9.03E203
16 Motor hanging beam 2.00E + 00 7.00E + 00 21.04E207 5.89E203
17 Stiffening beam 4.00E + 00 8.00E + 00 8.13E206 2.30E202
18 Lateral beam 2.00E + 00 6.00E + 00 6.05E202 1.11E201

uTi (Kvi M)u0i = 0 ð3Þ

uTi Mui = 1 ð4Þ

Then, v0i can be obtained as follows

1 T 0
v0i = u (K  v2i M 0 )ui ð5Þ
2vi i
In the simulation model, K and M are determined by
the stiffness matrix Ke and mass matrix Me of each ele-
ment, respectively. The variable e represents the elemen-
tal number
X
K= Ke ð6Þ
X
M= Me ð7Þ

The derivative of equations (6) and (7) sub-variable

b is calculated as follows
X
K0 = K 0e ð8Þ
X
M0 = M 0e ð9Þ

Then, equation (5) can be written as follows Figure 9. Optimal design variables of modal sensitivity analysis.

1 T
s0e, i = u (K 0  v2i Me0 )ue, i ð10Þ
2vi e, i e
the objective function s0e, i . The added mass must be
where s0e, i
is the objective function, representing the controlled within a certain range to ensure the light-
sensitivity value of the i order modal frequency vi sub- weight character of the car body.
variable b of the element e.12 Most stiffness properties of a car body are deter-
The values of car body mass and stiffness depend on mined by its bearing structure. First, 18 components of
the material properties and its specific structural the bearing framework of a car body are chosen to rep-
makeup. Based on the above modal frequency sensitiv- resent optimal variables, as shown in Figure 9. And
ity theory, the first vertical bending frequency is set as the specific meaning of variable is shown in Table 6.
Sun et al.
Figure 10. Sensitivity value of each component in reference to
first vertical bending frequency.
The sensitivity of each part thickness on the first verti-
cal bending frequency is then calculated. Finally, modal
frequency sensitivity analysis is conducted in order to
determine which component has the greatest effect on
increasing modal frequencies.
Modal sensitivity optimization
First, both the sensitivity of the first vertical bending
frequency and of the car body mass are calculated.
Table 6 depicts modal frequency sensitivity and mass
sensitivity values as they relate to numerous variables.
As shown in Table 6, some components exhibit large
sensitivity values as related to frequency, while others
primarily influence mass. These values represent differ-
ent degrees of influence of the changing thickness of
different beams on frequency and car body mass.
Figure 10 depicts the modal frequency sensitivity
values of all studied components. A positive sensitivity
value indicates the addition of thickness to increase fre-
quency and vice versa. If the sensitivity value is close to
zero, the effect of that particular component can be
ignored. Increasing the beam thickness can effectively
increase the bending frequency of the car body; thus,
this article focuses on the improvement of modal fre-
quencies. A small number of components that exhibit
high-sensitivity values are selected from among the 18
studied components. Component number 18, the lat-
eral beam, demonstrates the greatest sensitivity value,
while number 13 (hanging beam) and numbers 9 and
10 (chassis boundary beam) also exhibit a significant
positive effect on the first vertical bending frequency.
In order to guarantee the addition of less weight due
to changes in variable thickness during the optimiza-
tion process, a new Sr index is constructed. The relative
modal sensitivity value Sr is defined as follows
sv
Sr = ð11Þ
mv
where sv is the modal frequency sensitivity to beam
thickness, and mv represents the variable mass. The
7
Table 7. Relative sensitivity values.
Order Relative sensitivity Order Relative sensitivity
1 24.16E205 10 8.70E203
2 2.48E204 11 1.76E206
3 6.32E204 12 22.69E205
4 3.30E204 13 6.27E202
5 4.17E204 14 2.49E205
6 4.66E205 15 9.55E205
7 5.60E205 16 21.77E205
8 5.50E204 17 3.53E204
9 9.97E203 18 5.45E201
relative sensitivity value Sr depends on both the modal
frequency and the variable mass. According to the rela-
tive sensitivity results displayed in Table 7, the four
beams represented by numbers 9, 10, 13, and 18 are
identified as the optimal design variables.
The limit thickness ranges of the four optimal
design components are then set as constraint condi-
tions, and modal optimization is conducted. The
memory consumption of the computer is 32,000 MB
and the computational time is about 20 min for one
variable condition. There are four variable conditions,
so the computational time is 160 min in total. As
shown in Table 8, the first vertical bending frequency
increases from 9.70 to 10.60 Hz while the weight of the
entire car body increases from 39.44 to 40.37t. The
thickness of most design variables reaches their upper
limits, which indicates the effectiveness of modal sen-
sitivity optimization with the cost of additional struc-
tural mass.
The first lateral bending and the first torsion fre-
quencies also increase with different levels, as shown in
Table 9. In the initial stages of railway vehicle design,
modal sensitivity optimization can be conveniently
applied. However, as most vehicle parameters have
already been determined and altering the thickness of
components in the car body is difficult to realize, this
method will be limited in its implementation. Based on
this situation, another method that does not require
alteration of any internal structures of the car body is
proposed.
The design of equipment suspension
parameters
The coupled dynamic model of car body and
suspended equipment
Vibration of a railway car body includes rigid body
modes, which normally operate in a low-frequency
range. The car body also exhibits higher frequency flex-
ible modes consisting of twisting and bending deforma-
tions of the entire car body.13 A simplified equivalent
dynamic model with 2 degrees of freedom is
8 Advances in Mechanical Engineering

Table 8. Results of relative modal sensitivity optimization.

Description Original value Lower limit Upper limit Optimal result

Objective function First vertical bending frequency (Hz) 9.70 – – 10.60

Constraint condition Car body mass (t) 39.44 – 42.00 40.37
Optimal variable No. 9 Chassis side beam (mm) 10.50 8.00 15.00 15.00
No. 10 Chassis side beam (mm) 9.00 6.00 15.00 13.50
No. 13 hanging beam (mm) 6.00 4.00 8.00 8.00
No. 18 lateral beam (mm) 4.00 2.00 6.00 6.00

Table 9. Modal frequency results (Hz). When elastic suspension replaces the rigid connec-
tion between the equipment and the car body, they con-
Modal shape Original Optimization stitute a model with 2 degrees of freedom. The
vibration response of each component occurs as a
First vertical bending 9.70 10.60
First lateral bending 11.71 12.16 superposition of the lower frequency vibration mode
First torsion 14.09 14.25 with v1 and the higher frequency vibration mode with
v2. The vibration responses of the car body and equip-
ment can be calculated as follows

zc = A1 eiv1 t + A2 eiv2 t
ð14Þ
zeq = B1 eiv1 t + B2 eiv2 t

where A1, A2, B1, and B2 are amplitudes of the two

modes.
The binomial theorem is then applied in order to
obtain the ratios between the amplitudes of the two
natural modes, as follows
8
(1 + m)v2eq
Figure 11. Equivalent dynamic model. < A1 ’
B1 (1 + m)v2eq + v2c
v2c
ð15Þ
: A2
established, as shown in Figure 11, to determine the B2 ’ 1  m  v2eq
effects of equipment suspension on the modal para- pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
meters of the car body. where m = (meq =mc ), veq = (keq =meq ) = 2pfeq ,
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
In this model, the flexibility of the car body is repre- vc = (kc =mc ), and feq is the frequency of the indepen-
sented by the equivalent vertical bending stiffness dent suspended equipment (Hz).
kc = 134.4 MN/m, mc is the car body mass, zc is the Suspension static deflection is selected as the design
vertical displacement of the car body, meq is the equip- variable. Relationship between static deflection d and
ment mass, keq is the suspension stiffness, which is a its stiffness keq, as well as its frequency feq, can be
design variable here, and zeq is the vertical displacement expressed as follows
of equipment.
meq g
The differential motion equations of the system can d= ð16Þ
be expressed as follows keq
 g
mc€zc + (kc + keq )zc  keq zeq = 0 d= ð17Þ
ð12Þ 4p2 feq2
meq€zeq + keq zeq  keq zc = 0
The parameters of the car body are substituted into
According to the solutions to equation (12), the nat- equations (13) and (16) in order to obtain the natural
ural frequencies can be expressed as follows frequencies of the system, which change in response to
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi static deflection d, as shown in Figure 12. As the static
8  rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
  ffi
>
> deflection increases, both the independent equipment
>
> 1 keq keq + kc keq keq + kc keq kc
< v1 = 2 meq + mc  meq + mc 2  4 meq mc
feq and both natural frequencies decrease.
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 r ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
>
>   When the static deflection is approximately equal to
>
> 1 keq keq + kc keq keq + kc
2  4 meq mc
keq kc
: v1 = 2 meq + mc + meq + mc 0 mm, the car body is rigidly connected to the equip-
ment, forming a one-body system with the mass sum of
ð13Þ its bearing structure and the suspended equipment. In
Sun et al.
Figure 12. The relationship between natural frequencies and
static suspension deflection.
the original model, the first vertical bending frequency
is 9.7 Hz. Then, the static deflection of equipment sus-
pension is increased to a nearly infinite value, resulting
in the decoupling of the car body and the equipment.
At this time, the mass of the equipment is subtracted
from the car body, and the first vertical bending fre-
quency of car body is close to 10.97 Hz. The system can
be considered as one with 2 degrees of freedom with
elastic equipment suspension. The vertical bending fre-
quency falls between 9.70 and 10.97 Hz, as shown in
Figure 12. As long as the suspension deflection is posi-
tive, the frequency exceeds the 9.70 Hz demonstrated
by the rigid model, indicating improvement achieved
by elastic suspension. Elastic spring elements are incor-
porated into the FE model in order to determine spe-
cific suspension parameters in the next section.
Parameter design of elastic suspension
In this model, 3D rubber spring elements are utilized to
represent elastic connections employed to hang equip-
ment under the chassis of the car body. For example,
the details of rigid and elastic connections between the
electro-pneumatic brake unit and the car body are
shown in Figure 13. There are four rubber springs
under the car body to suspend the bake unit mass.
The stiffness and damping coefficient of rubber
springs can be expressed as follows
keq = 4p2 feq2 meq ð18Þ
pffiffiffiffiffiffiffiffiffiffiffiffi
c = 2e meq k d ð19Þ
where kd is the dynamic rubber stiffness, which changes
with temperature, excitation frequency, and material; c
is the elastic element damping coefficient; and e is the
damping ratio.
9
Figure 13. Equipment suspension in FE model: (a) rigid
connection and (b) elastic connection.
The bounce frequency of a bogie is always approxi-
mately 5 Hz14 and the first vertical bending frequency
of a car body is approximately 10 Hz; thus, the inde-
pendent suspension frequency of equipment feq should
avoid these frequencies in order to prevent resonance.
Based on resonance avoidance, 6, 7, and 8 Hz are,
respectively, selected as feq values to perform modal
analysis of a flexible car body in service condition. The
results are shown in Table 10. The memory of com-
puter is 32,000MB and the computational time is about
1.5 h for one condition of flexible suspension. There are
three conditions, so the total computational time is
4.5 h.
As shown in Table 10, all modal frequencies of a car
body increase for all studied suspension frequencies, as
compared to the original value. Thus, no positive corre-
lation is observed between equipment and the suspen-
sion and modal frequencies of the car body. In this
case, feq = 7 Hz demonstrated optimal improvement to
the car body modal parameter, resulting in a first verti-
cal bending frequency of 10.51 Hz. The elastic suspen-
sion parameter design should follow the principle that
independent suspended frequency of equipment must
not be equal to the natural frequencies of the car body
or bogie.
Ride quality of railway vehicle
The random vibration accelerations of vehicles at an
operation speed of 160 km/h through on-track field
tests are considered to be input signals of modal transi-
ent response calculations. Ride qualities are compared
10 Advances in Mechanical Engineering

Table 10. Modal frequencies of car body with different suspension methods (Hz).

Modal Shape Original model Elastic (6 Hz) Elastic (7 Hz) Elastic (8 Hz)

First vertical bending 9.70 10.43 10.51 10.28

First lateral bending 11.71 12.39 12.42 12.38
First torsion 14.09 14.24 14.25 14.20

Table 11. Vibration test equipment.
Name Amount Type
Analysis software 1 DASP-V10
CPCI case 1 INV3020D
Data acquisition card 8 INV3018G
ICP acceleration sensor 10 INV9828
CPCI: Compact Peripheral Component Interconnect; ICP: Internal
electronics piezoelectric.
after improving the modal parameters according to the
two method proposed methods. The sampling fre-
quency is 2048 Hz, and the utilized test equipment is
listed in Table 11.
Figure 14 shows the test train and data collection
system inside of the car body. Vibration acceleration
sensors are located at the air spring cover plate under
the bolster beam of the car body.
Figure 15 illustrates the vertical and lateral vibration
excitation signals transmitted from rail/wheel contact
in the ride quality analysis test. Ride quality at the car
body center is calculated in both the vertical and lateral
directions.
There are a variety of standards and criteria accord-
ing to which the ride quality of railway vehicles may be
evaluated. For the long-term evaluation of ride comfort
of railway vehicles, the evaluation procedure based on
ISO 236115 and that proposed by Sperling16 are often
employed. The Sperling index is an alternative method
of analysis based on Wz, a frequency-weighted root
mean square (RMS) value of acceleration that is evalu-
ated over defined time intervals or over a defined track
section.
Ride quality implies that the vehicle is assessed
according to the effect of mechanical vibrations on the
occupants. The equations for the Wz factor are written
as follows17
sffiffiffiffiffiffiffiffiffiffiffiffiffiffi
10 a
3
Wz = 0:896  F(f ) ð20Þ
f
where a is the RMS acceleration (cm/s2), f is the
response frequency (Hz), and F(f) is a frequency-
dependent factor used to express human sensitivity to
vibration. The above equations were constructed based
on vibration tests with human subjects and were then
supplemented by other test results.
The acceleration measured on the car body does not
exhibit a single frequency, but a frequency spectrum in
which natural frequencies of the vehicle are particularly
pronounced. The Wzi ride factor is determined for each
individual frequency. The total ride quality index Wztot
is determined by a combination of Wzi values exhibited
at different frequencies
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Wztot =
10
Wz110
+ Wz2 10
+    + Wzn 10 ð21Þ
Table 12 shows the classification of ride quality for
railway vehicles used in China. If the Sperling index
Wztot is greater than 3, the vehicle is not accepted as
demonstrating adequate ride quality.
Figure 16 shows the Sperling index obtained from
different models. According to Table 12, all obtained
values meet the standardized requirements for ‘‘good’’
ride quality, except for the vertical component of the
sensitivity optimization model. As shown in Figure 16,

Figure 14. Vehicle vibration test.

Sun et al.
Figure 15. Vibration acceleration input data.
Table 12. Classification for ride quality.
Class Sperling index Assessment
1 \2.5 Good
2 2.5–2.75 Satisfactory
3 2.75–3.0 Acceptable
Figure 16. Railway vehicle ride quality index.
the elastic suspension design of equipment improves
ride quality in both the lateral and vertical directions.
However, the vertical ride quality deteriorates after
modal sensitivity optimization, due to the additional
car body mass.
The Sperling index is at its minimum when the sus-
pension frequency of equipment is equal to 7 Hz. At
this moment, the vertical bending frequency is
10.51 Hz, which is bigger than that of other suspension
conditions. Therefore, as long as the equipment suspen-
sion parameter is set reasonably, it can improve car
body modal frequencies and the ride quality of railway
11
vehicles simultaneously. Additionally, this method does
not require alteration of any car body internal struc-
tures, making it more applicable to practical situations.
Conclusion
Modal sensitivity analysis is applied to optimize modal
frequencies by modifying the structural thickness of car
body frames with high sensitivity. The first vertical
bending frequency of a car body in service condition
increases by 9.2%, while the body mass increases by
2.35% as the rigidity of the car body is increased.
Modal sensitivity optimization can improve the modal
frequencies of car body. However, subsequent increase
in the vehicle sprung weight may result in deterioration
of dynamic performance.
The connection mode between equipment under the
chassis and the car body is altered based on the coupled
model. When the independent suspension frequency of
equipment is equal to 7 Hz, the resonance frequency
ranges of both the car body and the bogie may be
avoided, and the first vertical bending frequency
increases to 10.51 Hz. Modal frequencies can be effec-
tively improved by setting reasonable elastic suspension
parameters, which allows the vehicle to meet relevant
modal frequency requirements (.10 Hz) while simulta-
neously optimizing ride quality.
Although both proposed methods are capable of
achieving the desired results, the second method does
not require structural alteration of the car body and
thus increase the vehicle weight, which is beneficial for
implementation in existing engineering applications.
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with
respect to the research, authorship, and/or publication of this
article.
Funding
The author(s) disclosed receipt of the following financial sup-
port for the research, authorship, and/or publication of this
article: This work was sponsored by the National Science and
Technology Support Program of China (no. 2015BAG19B02)
and Project funded by China Postdoctoral Science
Foundation (no. 2015M571597).
References
1. Li Y, Zhou L and Li X. Lightening design and research
on steel trailer car bodies of high speed trains. Roll Stock
2002; 40: 13–21.
2. Nagai M, Yoshida H, Tohtake T, et al. Coupled vibra-
tion of passenger and lightweight car-body in consider-
ation of human-body biomechanics. Vehicle Syst Dyn
2006; 44: 601–611.
12
3. Tomioka T, Suzuki Y and Takigami T. Three-dimen-
sional flexural vibration of lightweight railway vehicle
car body and a new analytical method for flexural vibra-
tion. Q Rep 2003; 44: 15–21.
4. UIC 566:1990. Loadings of coach bodies and their
components.
5. EN 12663:2000. Railway applications: structural require-
ments of railway vehicle bodies.
6. Munich Research Center of German Federal Railway.
Technical specs for intercity express train ICE. Chengdu,
China: Southwest Jiaotong University Press, 1993.
7. Zhou J, Goodall R, Ren L, et al. Influences of car body
vertical flexibility on ride quality of passenger railway
vehicles. Proc IMechE, Part F: J Rail and Rapid Transit
2009; 223: 461–471.
8. Kamada T, Hiraizumi T and Nagai M. Active vibration
suppression of lightweight railway vehicle body by com-
bined use of piezoelectric actuators and linear actuators.
Vehicle Syst Dyn 2010; 48: 73–87.
9. Schandl G, Lugner P, Benatzky C, et al. Comfort enhance-
ment by an active vibration reduction system for a flexible
railway car body. Vehicle Syst Dyn 2007; 132: 835–847.
10. Wu G, Shi W and Liu W. Structural optimization of a
light bus body-in-white based on modal sensitivity analy-
sis. J Vib Shock 2013; 32: 41–45.
Advances in Mechanical Engineering
11. Choi K and Kim N. Structural sensitivity analysis and
optimization 1: linear systems. New York: Springer,
2006.
12. Kodiyalam S, Yang R and Gu C. The multi-disciplinary
design optimization of a vehicle system in a scalable, high
performance computing environment. Struct Multidiscip
O 2004; 26: 256–263.
13. Sun W, Zhou J, Thompson D, et al. Vertical random
vibration analysis of vehicle–track coupled system using
Green’s function method. Vehicle Syst Dyn 2014; 52:
362–389.
14. Gong D, Zhou J and Sun W. Analysis on the coupled
vibration between flexible car body and bogies of high
speed railway vehicles. J Traffic Transport Eng 2011; 11:
41–47.
15. ISO 2631-4:2000. Mechanical vibration and shock eva-
luation of human exposure to whole-body vibration. Part
4: guidelines for the evaluation of the effects of vibration
and rotational motion on passenger and crew comfort in
fixed guideway transport systems.
16. Ni C and Wang Y. A brief discussion for ride index and
comfort. Railway Locomotive & Car 2003; 6: 1–3.
17. GB5599-85:1985. Railway vehicles—specification for
evaluation of the dynamic performance and accreditation
test.