Engineering Science and Technology, an International Journal 20 (2017) 1610–1622

Contents lists available at ScienceDirect

Engineering Science and Technology,

an International Journal
journal homepage: www.elsevier.com/locate/jestch

Full Length Article

A combined nonlinear and hysteresis model of shock absorber for

quarter car simulation on the basis of experimental data
Vijay Barethiye, G. Pohit ⇑, A. Mitra
Department of Mechanical Engineering, Jadavpur University, Kolkata 700032, India

a r t i c l e i n f o a b s t r a c t

Article history: Modeling dynamic characteristics of an automotive shock absorber is a challenging task due to its com-
Received 20 June 2017 plex behavior. In the present paper, the nonparametric and hybrid approach is proposed to represent the
Revised 30 November 2017 nonlinear and hysteresis characteristics of the shock absorber. An experiment is carried out on a car dam-
Accepted 5 December 2017
per utilizing INSTRON to obtain force-velocity characteristics of the shock absorber. The experimental
Available online 13 December 2017
data is used to devise two different models, namely, piecewise linear model and hysteresis model, to cap-
ture the damping properties of the absorber and for consequent use in simulations. The complexity
Keywords:
involved due to certain physical phenomenon such as oil compressibility, gas entrapment etc. gives rise
Shock absorber modeling
INSTRON 8801
to hysteresis behavior and the present paper tries to model such behavior with the help of Neural
Nonlinear and hysteresis Networks. Finally, a combined (hybrid) shock absorber model (including the characteristics of both piece-
Neural network wise linear and hysteresis behavior) is developed in Simulink and integrated into a quarter car simulation
Quarter car to verify its feasibility. The results generated by the combined (hybrid) model are compared with linear as
Simulink well as piecewise linear model and the comparison shows that the proposed model substantially a better
option to study the vehicle characteristics more accurately and precisely.
Ó 2017 Karabuk University. Publishing services by Elsevier B.V. This is an open access article under the CC
BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction
The automotive suspension system has been a focus of atten-
tion to researchers over a long period of time as it enhances vehicle
stability, handling properties and comfort to the passengers. Due to
its multi-dimensional functions, it is regarded as one of the most
important item so far as the design of a vehicle is concerned. Since
the passenger cars have been historically one of the primary meth-
ods of ground transportation [1], it is essential to have a very good
vehicle suspension system that can reduce vibration of the chassis
to a great extent without compromising vehicle handling quality.
Suspension consists of the system of springs, shock absorbers
and linkages that connects a vehicle to its wheels. The main func-
tion of vehicle suspension system is to support the vehicle body
and provides riding comfort [2–3].
A shock absorber is the key element of the suspension system. It
is used to dissipate energy and hence reduce vertical oscillation of
the vehicle arising out of unevenness of the road surface. Comfort
and road-handling performance of a vehicle are mainly determined
by the damping characteristic of the shock absorbers [4]. The shock
⇑ Corresponding author.
E-mail address: goutam.pohit@jadavpuruniversity.in (G. Pohit).
Peer review under responsibility of Karabuk University.
absorbers generally used in motorcycle, car, light and heavy vehi-
cle exhibit the nonlinear and complex behavior [5–9]. The shock
absorber is typically characterized by the force-velocity curve also
called characteristics diagram. The modeling of the force-velocity
characteristics curve is not a trivial task due to the presence of hys-
teresis loop and nonlinear behavior of the damper [10]. The absor-
ber damping force is not only a strongly nonlinear function of
piston velocity but also exhibits asymmetric behavior in compres-
sion and rebound region.
In recent time, extensive researches have been carried out to
determine characteristics of the shock absorber. From the litera-
ture, it shows that the development era towards the modeling
can be categorized as theoretical modeling of physical system
[11–18], nonparametric modeling based on algebraic equations
[19–23], empirical modeling using various optimizations tech-
niques such as empirical, Neuro-fuzzy [24–27], and hybrid model-
ing comprising of theoretical-empirical and/or nonparametric-
empirical methods [28–34]. Several experiments have been carried
out to capture the dynamic properties of shock absorber including
its unsymmetrical hysteresis behavior.
Segel and Lang [11] developed the physical damper model hav-
ing a large number of parameters. However, it is not suitable for
dynamic simulation of vehicle performances due to its complex
nature. Wallaschek [12] suggested mathematical model using data

https://doi.org/10.1016/j.jestch.2017.12.003
2215-0986/Ó 2017 Karabuk University. Publishing services by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
 V. Barethiye et al. / Engineering Science and Technology, an International Journal 20 (2017) 1610–1622
Nomenclature
Symbol description
Cr1, Cr2, Cr3 damping coefficient for rebound (Ns/m)
Cc1, Cc2 damping coefficient for compression (Ns/m)
f damping force (N)
v velocity of piston rod (m/s)
Xn normalized data (hysteresis force)
Xm measured values (hysteresis force) from experimental
data
obtained from the experiment. He observed that a simple mathe-
matical model with a minimum number of parameters is not good
enough to capture the properties of the shock absorber. Besinger
et al. [13] developed semi-empirical damper model in which non-
linear spring is attached with the damper in series to describe the
nonlinear and hysteretic properties of the shock absorber. Duym
et al. [14] highlighted that two-slope or three-slope model of shock
absorber exhibited better result compared to the linear model.
In order to capture the complex behavior of shock absorber, Liu
and Zhang [19] developed a virtual prototype of a hydraulic shock
absorber in ADAMS environment. In the non-parametric category
of modeling, Calvo et al. [20], Yan Cui et al. [21] and Siradj [22]
developed models using mathematical formulation to describe
the nonlinear behavior of shock absorber. Gao et al. [26] developed
the strut model using a black-box model technique whereas Beghi
et al. [28] proposed the grey-box modeling technique for motorcy-
cle shock absorber; both techniques used artificial neural network
approach for the modeling purpose. Lion and Loose [29], and
Prancy et al. [30] developed a thermo-mechanically coupled shock
absorber model by considering heating phenomenon. Prancy et al.
[30] and Cui et al. [21] developed hybrid shock absorber model in
which the properties of shock absorber were represented by a
spline and polynomial algebraic function followed by neural net-
work technique. Castellani et al. [33] developed the hybrid model
of mono-tube shock absorber using physical and empirical model-
ing. In order to investigate dynamic characteristics of the automo-
tive shock absorber, researchers [34–44] performed experiment
along with modeling and analysis.
Calvo et al. [20] analyzed the dynamic behavior of a shock
absorber using damper test rig at different frequencies (0.25 Hz,
0.5 Hz, 1 Hz and 3 Hz). By using experimental results, author
generated a piecewise bilinear model but did not consider the hys-
teretic behavior of the damper. Lang [6] studied the frequency
dependent behavior of shock absorber. The hysteretic loop in the
force-velocity plot was supposed to be due to the compressibility
of oil and the presence of either gas or vapor phase at certain stages
of the cycle [11,19].
From the literature, it has been observed that various linear and
non-linear models have been proposed by different researchers to
model force-velocity and force-displacement relationships of the
shock absorber. However, little attempts have been made to cap-
ture the hysteretic property of shock absorber represented by the
area of the curve, as it is difficult to model it through parametric
modeling so that prediction of dynamic performance characteris-
tics of a vehicle using the model will be precise.
In this paper, an experiment is carried out on a commercial car
damper using INSTRON 8801 servohydraulic testing machine. With
the help of frequency and amplitude variation on the damper and
sinusoidal excitation, the force-displacement and force-velocity
characteristics curves are determined and presented in the paper.
The main aim of the work is, to build up the nonlinear and hystere-
1611
ms sprung mass for quarter car vehicle (kg)
mu unsprung mass for quarter car vehicle (kg)
xs displacement variables of sprung (m)
xu displacement variables of unsprung mass (m)
ks suspension system has stiffness (N/m)
kt stiffness of the tire (KN/m)
fd damper force (N)
sis characteristics of the absorber. Therefore, in order to construct
the actual force-velocity characteristic curve, a combined (hybrid)
shock absorber model is developed that can represent the hystere-
sis and nonlinear characteristics of the shock absorber. Nonlinear
properties of the absorber are modeled using the piecewise linear
curve, whereas hysteresis effect of the shock absorber is captured
by applying Neural Network technique. The model is purely based
on force-velocity characteristics phenomenon. The individuality of
the work is, modeled both properties successfully in Simulink and
incorporated the developed shock absorber model in a quarter car
suspension system.
The complete shock absorber model is obtained by combining
two different techniques. The model thus obtained is validated with
the results obtained from the experiment. Lastly, the combined
(hybrid) model of the shock absorber is incorporated with 2 degrees
of freedom quarter car suspension model to determine the response
of the hybrid model. The random road profile is considered as input
excitation to a quarter car model and performance characteristics of
the vehicle such as body acceleration, body displacement, and sus-
pension deflection is analyzed for linear, piecewise linear and com-
bined (hybrid) shock absorber model using Simulink.
The novelty of the present study is to develop a hybrid shock
absorber model to capture hysteresis characteristics of the damper
based on data obtained from an experiment carried out in the lab-
oratory. The model thus developed is subsequently utilized to pre-
dict the response behavior of the vehicle under different road
excitation.
2. Experimental set-up for car shock absorber dynamic analysis
In order to determine the dynamic properties of a shock absor-
ber, an experiment is carried out on an INSTRON 8801 servo-
hydraulic machine. The shock absorber is commercially used in
Ford car running on Indian roads. The experimental set-up, a com-
plete testing solution, is shown in Fig. 1. The shock absorber is
fixed between the dynamic load cell and fixture (see Fig. 1(a)).
The fixture (Fig. 1(b)) is fabricated as per the dimensions of the
damper. The INSTRON 8801 is supplied with data acquisition sys-
tem (Fig. 1(c)) that provides complete control over input and out-
put data.
Force-displacement properties of the shock absorber are
obtained with sinusoidal excitation having different amplitudes
and frequencies. The values of input data (Table 1) applied for exci-
tation purpose are chosen based on the experiments being con-
ducted in open literature and specification given in Gillespie [45]
and Wong [46].
2.1. Force-displacement and force-velocity characteristics curve
The Figs. 2 and 3 illustrate the nature of force-displacement
and force-velocity characteristics of the damper by varying
1612 V. Barethiye et al. / Engineering Science and Technology, an International Journal 20 (2017) 1610–1622

Table 1
Input data of sinusoidal excitation.

Frequency (Hz) Amplitude (mm)

0.5 5, 10, 15, 20, 30, 40
1.0 5, 10, 15, 20, 30
1.5 5, 10, 15, 20
2.0 5, 10, 15
2.5 5, 10

(a)

(b)

(c)

Fig. 1. Experiment on car damper: (a) experimental set-up, (b) clamping fixture Fig. 2. (a) Force-displacement plot and (b) force-velocity plot at 1 Hz frequency and
used during experiment, (c) data acquisition system. having different amplitude of excitations.
 V. Barethiye et al. / Engineering Science and Technology, an International Journal 20 (2017) 1610–1622
Fig. 3. (a) Force-displacement plot and (b) force-velocity plot at amplitude of 15
mm and having different frequency of motions.
displacement and frequency. It is observed that the damping force
is highly non-linear and asymmetric in nature at rebound and
compression region, being much higher in rebound compared to
compression region.
The plots also indicate that with an increase in excitation
frequency and amplitude, the damping force and hysteresis loop
is increased. The shock absorber properties depend on both
frequency and amplitude. However, the influence of excitation
frequency is more prominent than the amplitude. It can be
observed that beyond a limiting value of amplitude and fre-
quency, the hysteresis loop disappears leading to a saturated
phenomenon.
1613
3. Shock absorber modeling
As previously described, the characteristics of the shock absor-
ber are complex in nature. It is very important to capture the non-
linear as well as hysteresis properties of the shock absorber to
study the performances of the vehicle. Most of the researchers
focussed on modeling of force-velocity characteristics of shock
absorber using linear, polynomial, etc.
In the present paper, the nonlinear nature of the force-velocity
curve is modeled in piecewise linear manner, whereas the hystere-
sis property is captured with the help of neural network technique.
These two models are combined to form single hybrid nonlinear-
hysteresis model using Simulink.
3.1. Nonlinear modeling of shock absorber
In order to provide the damping in the suspension system, the
shock absorber dissipates energy in the compression stroke
(jounce) and extension stroke (bounce) whenever a vehicle moved
over a bump on a road. The positive force shown on the graph is for
the extension stroke (rebound stroke) and the negative force repre-
sents the compression stroke. Generally, the force in the rebound
region is more than the force in compression region [20].
Fig. 4(a) shows the force-velocity characteristics curve of the
shock absorber. The results are obtained by recording the force
using INSTRON machine depending on the input frequency of
2.5 Hz at 10 mm displacement for 10 full cycle operation contain-
ing rebound and compression stroke. The loop is divided into two
parts, the compression and rebound region, which is represented
by two different mathematical expressions to model the nonlinear
properties of the shock absorber. The piecewise linear fitting can
allow local control on the shape and close proximity towards the
data points.
The force-velocity relationships of the shock absorber during
rebound and compression strokes are represented by Eqs. (1) and
(2) respectively.
For rebound
8
< cr1 v; 0 m=s < v 6 0:05 m=s
>
f ¼ cr2 v; 0:05 m=s < v 6 0:1 m=s ð1Þ
>
:
cr3 v; v > 0:1 m=s
Here Cr1, Cr2 and Cr3 are damping coefficients for rebound regions.
For compression

cc1 v; 0:05 m=s 6 v < 0 m=s
f ¼ ð2Þ
cc2 v; v < 0:05 m=s
Here CC1 and CC2 are damping coefficients for compression regions.
The values of parameter Cr1, Cr2, Cr3, CC1, and CC2 are obtained by fit-
ting the curve shown in Fig. 4(b) and these values are presented in
Table 2.
3.2. Neural network model for hysteresis characteristics of shock
absorber
The neural network (NN) is an excellent optimization tool used
to model and control complex nonlinear system. In the past, much
work has been carried to model and identification of nonlinear sys-
tem by using a neural network (NN) technique [24,26] and this has
motivated to model the hysteresis force of shock absorber by using
Neural Network (NN). A neural network consisting of three main
layers; input, hidden and output layer, is an interconnected net-
work of artificial neurons with a set of weights (w) and bias (b)
adjustment (Fig. 5). Bias nodes are added to increase the flexibility
of the model to fit the data whereas, weights or strength is the
1614 V. Barethiye et al. / Engineering Science and Technology, an International Journal 20 (2017) 1610–1622

dynamic structure of an NN model that adjusts to perform the nec-

essary task. With the high capability of the neural network, the
nonlinear relation between input and output can be predicted.
In this paper, a feed-forward neural network with back propa-
gation method (Fig. 5) is employed to model the complex nature
of hysteresis force exhibited by the shock absorber.
The hysteresis force is obtained by subtracting the piecewise lin-
ear force obtained (Fig. 4(b)) from experimental data shown in
Fig. 4(a). The experiment results consist of three variable elements,
namely, displacement, frequency, and force. Since the damping
force is dependent on amplitude and frequency, both displacement
and velocity (obtained by differentiating displacement) of the pis-
ton are considered to be model inputs (x and v) and hysteresis force
as output (F). In the present case, a neural network with one hidden
layer is found to accomplish acceptable performance. In addition, a
number of layers may lead to computational complexity in the net-
work. In order to determine the number of neurons in hidden layer,
network training is performed randomly by increasing the number
of neurons from 1 to 10. The pattern is set as 70% training, 15% for
validation and rest 15% for testing purpose.
In the present case, in order to model the non-linear hysteresis
property of the shock absorber, the first layer is supposed to have
10 neurons, a sigmoid transfer function for hidden layer and linear
transfer function for output layer as shown in Fig. 5.
A set of test data is used to train the network. To reduce the
simulation time, all the data is normalized within the range ±1.
The formulation [26] to normalize data is given below,

X m  minðX m Þ
Xn ¼ 2  1 ð3Þ
maxðX m Þ  minðX m Þ

where Xn is expressed as a normalized data (Hysteresis force), Xm

has measured values (Hysteresis force) from experimental data,
min (Xm) and max (Xm) denote the minimum and maximum value
of data Xm respectively.
Performance of system during training, validation and testing
steps is determined by the mean square error (MSE) as given
below.

1 Xn
MSE ¼ ðYi; m  Yi; p Þ2 ð4Þ
m i¼1

where Yi,m and Yi,p are the ith measured output data (obtained from
experiment) and NN predicted output data respectively, and n is a
number of samples used in each step.

3.2.1. Results of neural network model for hysteresis characteristics of

shock absorber
In order to develop the neural network model, a proper training
is required for obtaining an acceptable output that can be subse-
Fig. 4. (a) Experimental force-velocity curve, (b) piecewise linear fit model. quently used for modeling purpose. After performing training, the
obtained results are shown in Fig. 6. Fig. 6(a) represents the Mean
Squared Error (MSE) variations for training and validation during
a number of epochs. It is observed from Fig. 6(b) that in the present
case mean square error value is 0.0054 for Neural Network model.
Table 2
The next step is to create a regression plot that shows a rela-
The values of damping coefficient.
tionship between the output of the network and the target. In
Parameter Description (Unit) Numerical value the present study, the measured value (target) is hysteresis force
Cr1 Damping coefficient for 2300 (0 m/s < v  0.05 m/s) which is determined by subtracting piecewise linear force from
rebound (Ns/m) the experimental force. The predicted value (output) is obtained
Cr2 Damping coefficient for 2100 (0.05 m/s < v  0.1 m/s)
from the best fit model for hysteresis force. The training and vali-
rebound (Ns/m)
Cr3 Damping coefficient for 2000 (v > 0.1 m/s)
dation results of present NN model are shown in Fig. 7(a) and (b),
rebound (Ns/m) respectively. The dashed line in each plot represents the relation-
Cc1 Damping coefficient for 660 (0.05 m/s  v < 0 m/s) ship between predicted value (output) and the measured value
compression(Ns/m) (target). The solid line represents the best fit between outputs
Cc2 Damping coefficient for 398 (0.05 m/s)
and targets. The coefficient of correlation, R-value is an indication
compression (Ns/m)
of the relationship between the outputs and targets. It measures
 V. Barethiye et al. / Engineering Science and Technology, an International Journal 20 (2017) 1610–1622 1615

Fig. 5. Feed forward neural network with back propagation method.

Fig. 6. Neural network modeling result (a) MSE variations during Epochs, (b) Error.

the direction and strength of linear the relationship between pre-

dicted value (output) and the measured value (target). In this case,
the value of R is close to unity indicating a very a good fit.
The values of hysteresis forces against velocities as obtained
from NN model are shown in Fig. 8. On the same plot, hysteresis
forces obtained from the experiment and piece-wise linear model
is also projected. It is observed that there is a very good agreement
with acceptable accuracy between NN model and experimental
results. Fig. 7. Regression results from neural network analysis: (a) training, (b) validation.
1616 V. Barethiye et al. / Engineering Science and Technology, an International Journal 20 (2017) 1610–1622
Fig. 8. Comparative analysis of hysteresis force of shock absorber.
3.3. Formation of combined (hybrid) shock absorber model
In the proposed hybrid model, the hysteresis force of the shock
absorber obtained from the neural network model (presented in
Section 3.2) is then combined with a nonlinear force of the shock
absorber determined from the piecewise linear model (presented
in Section 3.1) to obtain total damping force (fd) of the absorber.
It is shown in block diagram in Fig. 9(a).
The simulation of combined (hybrid) shock absorber model is
carried out in Simulink software. The accuracy of the hybrid model
will be enhanced vastly as it captures effectively both nonlinear
and hysteresis characteristics of the damper. Fig. 9(b) shows the
Simulink modeling of the combined model using piecewise linear
model and Neural Network model.
4. Validation of combined (hybrid) shock absorber model
The combined (hybrid) shock absorber model is validated with
the shock absorber characteristics curve obtained through an
experiment. Fig. 10(a) and (b) illustrates the comparison of force-
displacement and force-velocity characteristics curves obtained
from the hybrid shock absorber model and experimental results.
The figures show that there is a very good match between the pro-
posed model and experimental results.
Fig. 11 represents the comparison between damping force
obtained from a proposed combined hybrid model with that of
the experiment. The plots show excellent agreement. The calcu-
lated RMS error value of the damping force for the combined shock
absorber model is found to be approximately 2.85 N which is a
good improvement compared to the RMS error value of 13.28 N
when only pure nonlinear piecewise linear model is considered.
5. Quarter car suspension simulation
Simulation of a quarter car suspension model is a very effective
method to investigate the influence of damping forces on the
response characteristics of a vehicle under different road surface
inputs. Hence, the combined hybrid model is incorporated into
quarter car suspension system for subsequent dynamic analysis
of the entire system.
The quarter car model, represented by a 2-Degree of Freedom
spring-mass-damper system as shown in Fig. 12. The mass of the
vehicle is represented by sprung mass (ms) and unsprung mass
(mu). The vertical motions of the two masses are described by
the displacement variables xs and xu of the sprung and unsprung
mass, respectively.
The road excitation is given by (xr). The spring of the suspension
system has a stiffness (ks) whereas the damping characteristic of
the shock absorber is given by the force-velocity relationship of
the combined (hybrid) model as described in the preceding section.
The stiffness of the tire is represented by a second spring con-
nected to unsprung mass, having coefficient (kt). Both the springs
used in the system are assumed to be linear in nature with con-
stant spring coefficients whereas the tire damping is considered
to be negligible as compared to the system damping.
The equation for the quarter car model motions are found by
adding vertical forces on the sprung and unsprung masses [3]. They
are expressed as follows.
€s þ ks ðxs  xu Þ þ f d ¼ 0
ms x
ð5Þ
€u  ks ðxs  xu Þ  kt ðxr  xu Þ  f d ¼ 0
mu x
where fd is the damper force.
The quarter car parameter values used for simulation are given
in Table 3.
5.1. Road profile
The response of the first-order filter to white noise is taken as
road profile used by Refs. [16,47] (Fig. 13) for quarter car
simulation.
The following equation is used as road input to the quarter car
model.
x_ r ¼ avxr þ gðtÞ ð6Þ
Here, a is the road roughness parameter and is set as 0.151/m for a
smooth road, v is the speed of the car and is set as 20 m/s, and ƞ(t) is
white noise. With these values, the output xr(t) of the differential
Eq. (6) shall be taken as the road profile. The simulations are carried
out for a time period of 10 s and the results are established for 3 s to
make the comparative analysis.
Fig. 14 shows the modeling of the quarter car suspension sys-
tem with combined (hybrid) shock absorber model under Simulink
environment.
6. Results and discussion
The primary objective of suspension the system is to reduce the
vertical vibration of the vehicle which in turn offers a good quality
of road comfort to the passengers. From the shock absorber test
results it can be seen that the absorber damping force is a strongly
nonlinear function of piston velocity, being asymmetric in com-
pression and rebound region with an unsymmetrical hysteretic
loop. From the design point of view, the nonlinearity and hysteresis
properties of the shock absorber is notably important as it is used
to optimize stability and comfort by dissipating energy from the
system. Therefore in this study, an attempt is made to determine
the response of the proposed shock absorber model due to road
disturbance. Therefore, shock absorber damping force obtained
from the combined hybrid model is first evaluated using Simulink
(Fig. 14) and is compared with that of piecewise model i.e. without
considering the hysteresis effect of the shock absorber, input to the
system being road profile. It is observed from Fig. 15 that hybrid
model provides higher forces compared to the piecewise model
 V. Barethiye et al. / Engineering Science and Technology, an International Journal 20 (2017) 1610–1622
Fig. 9. (a) Block diagram and (b) Simulink modeling for combined (hybrid) shock absorber damping force.
due to the presence of hysteresis. The RMS value for the combined
model is about 110 N whereas that for piecewise model is only 80
N. Therefore, it is essential to consider the hysteresis force of the
absorber during modeling.
With the knowledge of the shock absorber forces, the dynamic
responses of the suspension system in the form of body accelera-
tion, body displacement and suspension deflection are determined
in Simulink platform. The block diagram is shown in Fig. 16.
The quarter car simulation results obtained are shown in
Figs. 17–19 for the body acceleration, body displacement, and sus-
pensions deflection respectively. From these figures, it is clear that
linear model is not at all adequate to represent the shock absorber
behavior and hence response of the vehicle against a particular
road excitation. It can be further highlighted that so far as the ride
comfort is concerned, combine hybrid model is the best option.
The analysis is carried out on three different models of the
absorber in the Simulink environment. In the first two cases,
namely, linear model and piecewise linear model, only shock
absorber force are considered. However, in the combined hybrid
hysteresis model, hysteresis force of the shock absorber is also
considered. The linear damper model used in the analysis is
1617
extrapolated as a viscous linear damping coefficient (Cs = 2000
Ns/m) from experimental data.
The statistics of the vehicle performance in terms of RMS value
and peak values of body acceleration, body displacement and sus-
pension deflection are also shown in Table 4.
On the basis of simulation results (Figs. 17–19) and statistical
data (Table 4), the following observations are made.
 The RMS and peak value of body acceleration of the combined
hybrid shock absorber model is smaller than those of linear
and piecewise models;
 The RMS value of body displacement of the combined model is
slightly larger than the linear model and similar to piecewise
model whereas the Peak value of body displacement of com-
bined model is smaller than the two other models;
 The RMS value of suspension deflection for the combined model
exhibits larger value, whereas the peak suspension deflection is
found to be smaller than those of other models.
The body acceleration response (Fig. 17) of simulated result
indicates that the damping coefficient value of the linear model
1618 V. Barethiye et al. / Engineering Science and Technology, an International Journal 20 (2017) 1610–1622

Fig. 11. Comparative analysis damping force plot obtained from experiment and
combined (hybrid) shock absorber mode.

Fig. 10. Comparative analysis plot for (a) force-displacement and (b) force-velocity
characteristics curve.

and piecewise model is smaller as compared to the combined

hybrid model. Fig. 18 shows the smaller overshot value of body
displacement response for the combined model compare to linear
and piecewise model. The suspension deflection value for the com-
bined model shows larger value (Fig. 19); as a result, the vehicle Fig. 12. Quarter car suspension system.
output shows small body acceleration. It can be expected that
the difference between the combined hybrid model and other
models (linear and piecewise linear model) would be more signif- Table 3
Quarter car parameters.
icant at a higher speed as at those speeds it is expected that higher
nonlinear and hysteretic characteristics would be manifested. Parameter Value Unit
From the simulated plots and Table 4, it can be seen that the lin- Sprung mass 280 kg
ear damping cannot be a choice to model the behavior of shock Unsprung mass 35 kg
absorber as the suspension responses predicted by such a model Suspension stiffness 17,600 N/m
Tyre stiffness 1,90,000 KN/m
deviates significantly from the response generated from experi-
V. Barethiye et al. / Engineering Science and Technology, an International Journal 20 (2017) 1610–1622 1619

Fig. 13. Road profile.

Fig. 14. Quarter car model in Simulink with combined (hybrid) shock absorber model.

Fig. 15. Shock absorber forces for combined (hybrid) and piecewise model.
1620 V. Barethiye et al. / Engineering Science and Technology, an International Journal 20 (2017) 1610–1622

Fig. 16. Quarter car model in Simulink for linear, piecewise linear and combine model.

Fig. 17. Body acceleration for linear, piecewise linear and combined model.

Fig. 18. Body displacement for linear, piecewise linear and combined model.

mentally obtained data. The linear model has also predicted larger
amplitude as compared to other models. Although piecewise linear
model shows better results compared to the linear model, In fact,
the combined shock absorber damping model captures the vehicle
behavior more closely and precisely.
The results obtained from simulation and from the statistics, it
can be observed that the linear and piecewise damper models are
not able to capture the vehicle performance accurately due to the
absence of hysteresis force. Hence, to understand and study the per-
formance of the vehicle and to offer the better passenger comfort,
the combined shock absorber model is preferable than the piecewise
and linear model. The combined shock absorber model captures
more accurately the complex behavior of shock absorber by combin-
ing the nonlinear characteristics and hysteresis property.
 V. Barethiye et al. / Engineering Science and Technology, an International Journal 20 (2017) 1610–1622
Fig. 19. Suspension deflection for linear, piecewise linear and combined model.
Table 4
Statistics for the vehicle characteristics.
Parameter Unit Linear model
Body acceleration (RMS) m/s2 0.4869
Peak body acceleration m/s2 1.3012
Body displacement (RMS) m 0.0076
Peak body displacement m 0.0167
Suspension deflection (RMS) m 0.0038
Peak suspension deflection m 0.0096
7. Conclusion
In the present paper, the work is carried out to develop a shock
absorber model of a passenger car. An experiment is carried out in
the laboratory using INSTRON 8801 to generate data on the force-
displacement and force-velocity characteristics of the damper. Dif-
ferent models of the shock absorber are developed to represent the
nonlinear and hysteresis force-velocity behavior of the damper.
Subsequently, the model is validated with experimental results.
It is found from the study that a combined hybrid shock absorber
model using neural network technique can only capture the exper-
imental behavior of the damper very closely. It is also evident that
the developed combined hybrid model describes the complex
behavior of shock absorber more effectively and precisely due to
the fact that hysteresis characteristics of the shock absorber are
incorporated in this model whereas such provision is not there in
case of linear and piecewise models.
In order to investigate the influence of damping characteristics
on vehicle performance, quarter car simulation are carried out for
linear, piecewise linear and proposed combined hybrid shock
absorber model in Simulink. Road profile is generated to provide
displacement excitation to the simulated car model and the vehicle
responses are compared to different absorber models. From the
simulated results and statistical analysis, it can be concluded that
the simple linear and piecewise linear model of the shock absorber
cannot accurately capture the vehicle performances and hence to
predict the responses of the vehicle due to road disturbances, the
combined hybrid shock absorber model is a better option. This
model has enough computational efficiency to apply in half car/full
car model to study the ride characteristics of vehicles.
Conflicts of interest
None.
This research did not receive any specific grant from funding
agencies in the public, commercial, or not-for-profit sectors.
1621
Piecewise model Combined (hybrid) model
0.4575 0.4414
0.8911 0.8668
0.0081 0.0081
0.0154 0.0132
0.0060 0.0068
0.0095 0.0073
References
[1] A. Belhocine, N.M. Ghazaly, Effects of Young’s modulus on disc brake squeal
using finite element analysis, Int. J. Acoust. Vib. 21 (3) (2016) 292–300.
[2] A.A. Aly, F.A. Salem, Vehicle suspension systems control: a review, Int. J.
Control Autom. 2 (2) (2013) 46–54.
[3] A. Agharkakli, G.S. Sabet, A. Barouz, Simulation and analysis of passive and
active suspension system using quarter car model for different road profile, Int.
J. Eng. Trend. Tech. 3 (5) (2012) 636–644.
[4] C. Lauwerys, J. Swevers, P. Sas, Robust linear control of an active suspension on
a quarter car test-rig, Control Eng. Pract. 13 (5) (2005) 577–586.
[5] L. Konieczny, Analysis of simplifications applied in vibration damping
modeling for a passive car shock absorber, Shock Vib. (2016).
[6] D. Ozcan, U. Sonmez, L. Guvenç, Optimisation of the nonlinear suspension
characteristics of a light commercial vehicle, Int. J. Veh. Tech. (2013).
[7] S. Li, Y. Lu, L. Li, Dynamical test and modeling for hydraulic shock absorber on
heavy vehicle under harmonic and random loadings, Res. J. Appl. Sci. Eng. Tech.
4 (13) (2012) 1903–1910.
[8] V. Cossalter, A. Doria, R. Pegoraro, L. Trombetta, On the non-linear behavior of
motorcycle shock absorbers, P. I. Mech. Eng. D-J. Aut. 224 (1) (2010) 5–27.
[9] A.M. Salem, W. Galal, Identification of characteristics of hydraulic shock
absorbers used in light weight tracked vehicles, in: Proceeding of 13th
International Conference on Aerospace Sciences & Aviation Technology (ASAT),
Military Technical College, Cairo, Egypt, 2009, pp. 26–28.
[10] J. de Jesus Lozoya-Santos, S. Aubouet, R. Morales-Menendez, O. Sename, R.
Ramirez-Mendoza, L. Dugard, Hysteresis modelling for a MR damper, in: 7th
EUROSIM Congress on Modelling and Simulation, Prague, Czech Republic,
2010, p. 71.
[11] H. Lang, Automotive Dampers at High Stroking Frequency (Ph.D. Thesis),
University of Michigan, US, 1977.
[12] J. Wallaschek, Dynamics of non-linear automobile shock-absorbers, Int. J.
Nonlinear Mech. 25 (2–3) (1990) 299–308.
[13] F.H. Besinger, D. Cebon, D.J. Cole, Damper models for heavy vehicle ride
dynamics, Veh. Syst. Dyn. 24 (1) (1995) 35–64.
[14] S. Duym, R. Stiens, K. Reybrouck, Evaluation of shock absorber models, Veh.
Syst. Dyn. 27 (2) (1997) 109–127.
[15] R. Basso, Experimental characterization of damping force in shock absorbers
with constant velocity excitation, Veh. Syst. Dyn. 30 (6) (1998) 431–442.
[16] A.K. Raj, C. Padmanabhan, A new passive non-linear damper for automobiles,
P. I. Mech. Eng. D-J. Aut. 223 (11) (2009) 1435–1443.
[17] Y. Chen, K. Guo, Y. Yang, Y. Zhuang, Physical modeling of shock absorber using
large deflection theory, SAE Int. J. Pas. Cars Mech. Syst. 5 (2012) 393–403.
[18] U. Ferdek, J. Luczko, Modeling and analysis of a twin-tube hydraulic shock
absorber, J. Theor. Appl. Mech. 50 (2012) 627–638.
[19] Y. Liu, J. Zhang, Nonlinear dynamic responses of twin-tube hydraulic shock
absorber, Mech. Res. Commun. 29 (5) (2002) 359–365.
[20] J.A. Calvo, B. Lopez-Boada, J.L. San Roman, A. Gauchia, Influence of a shock
absorber model on vehicle dynamic simulation, P. I. Mech. Eng. D-J. Aut. 223
(2) (2009) 189–203.
1622 V. Barethiye et al. / Engineering Science and Technology, an International Journal 20 (2017) 1610–1622
[21] Y. Cui, T.R. Kurfess, M. Messman, Testing and modeling of nonlinear properties
of shock absorbers for vehicle dynamics studies, in: Proceedings of the World
Congress on Engineering and Computer Science, vol. 2, San Francisco, USA,
2010, pp. 20–22.
[22] R. Siradj, F.J. Darsivan, M. Hamid, M.F. Abdullah, Nonlinear model
identification of a shock absorber dynamics, in: Automotive and Vehicle
Technologies Conference, Yildiz Technical University, Istanbul, Turkey, 2013,
pp. 1–14.
[23] Y. Mikhlin, S. Mytrokhin, Nonlinear vibration modes of the double tracked
road vehicle, J. Theor. Appl. Mech. 46 (2008) 581–596.
[24] J. Giacomin, Neural network simulation of an automotive shock absorber, Eng.
Appl. Artif. Intell. 4 (1) (1991) 59–64.
[25] A.J. Barber, Accurate models for bushings and dampers using the empirical
dynamics method, in: 14th European ADAMS Users’ Conference, Berlin,
Germany, 1999.
[26] B. Gao, J. Darling, D.G. Tilley, R.A. Williams, A. Bean, J. Donahue, Modeling of a
novel gas strut using neural networks, in: ASME 2004 International
Mechanical Engineering Congress and Exposition, ASME, 2004, pp. 137–143.
[27] Z. Sika, V. Bauma, M. Valasek, A. Bilkovsky, P. Kukula, P. Steinbauer,
Identification of phenomenological shock absorber model using neuro-fuzzy
method, Bull. Appl. Mech. 4 (15) (2008) 88–91.
[28] A. Beghi, M. Liberati, S. Mezzalira, S. Peron, Grey-box modeling of a motorcycle
shock absorber for virtual prototyping applications, Sim. Mod. Pract. Theor. 15
(8) (2007) 894–907.
[29] A. Lion, S. Loose, A thermomechanically coupled model for automotive shock
absorbers: theory, experiments and vehicle simulations on test tracks, Veh.
Syst. Dyn. 37 (4) (2002) 241–261.
[30] V. Pracny, M. Meywerk, A. Lion, Hybrid neural network model for history-
dependent automotive shock absorbers, Veh. Syst. Dyn. 45 (1) (2007) 1–14.
[31] A. Di Dino, F. Biral, P. Bosetti, Hybrid modeling of non-linear mechanical
systems: the case of a vehicle shock absorber, in: ASME 2011 International
Design Engineering Technical Conferences and Computers and Information in
Engineering Conference, ASME, 2011, pp. 1121–1130.
[32] Y. Cui, T.R. Kurfess, Vehicle parameter identification for vertical dynamics, J.
Dyn. Syst. 137 (2) (2015) 021013.
[33] F. Castellani, N. Bartolini, L. Scappaticci, D. Astolfi, M. Becchetti, Numerical and
Experimental Dynamics of a Monotube Shock Absorber, SAE Technical Paper,
2016.
[34] A.L. Audenino, G. Belingardi, Modelling the dynamic behavior of a motorcycle
damper, P. I. Mech. Eng. D-J. Aut. 4 (1995) 249–262.
[35] A. Simms, D. Crolla, The Influence of Damper Properties on Vehicle Dynamic
Behavior, SAE Technical Paper, 2002.
[36] Y. Liu, J. Zhang, F. Yu, H. Li, Test and Simulation of Nonlinear Dynamic Response
for the Twin-Tube Hydraulic Shock Absorber, SAE Technical Paper, 2002.
[37] S. Lu, T. Yang, S. Hwang, Dynamic analysis of a double-tube shock absorber for
robust design, JSME Int. J. C. Mech. Syst. 40 (2) (1997) 335–345.
[38] M.S.M. Sani, M.M. Rahman, M.M. Noor, K. Kadirgama, M.R.M. Rejab, Study on
dynamic characteristics of automotive shock absorber system, in: Proceedings
of the Malaysian Science and Technology Congress, Malaysia, 2008, pp. 623–
628.
[39] X. Ning, J. Shen, L. Wang, Modeling and experiment of vehicle double-tube
hydraulic shock absorber, in: International Conference on Electric Information
and Control Engineering (ICEICE), IEEE, 2011, pp. 2334–2337.
[40] Y. Lu, S. Li, N. Chen, Research on damping characteristics of shock absorber for
heavy vehicle, Res. J. Appl. Sci. Eng. Tech. 5 (3) (2013) 842–847.
[41] V.M. Barethiye, G. Pohit, A. Mitra, Simulation for the quarter car model
compliance with rubber bush and washers, in: Proceedings of the Sixth
International Congress on Computational Mechanics and Simulation (ICCMS),
IIT Bombay, India, 2016.
[42] D. Maher, P. Young, An insight into linear quarter car model accuracy, Veh.
Syst. Dyn. 49 (3) (2011) 463–480.
[43] C. Kim, P.I. Ro, An accurate full car ride model using model reducing
techniques, J. Mech. Des. 124 (4) (2002) 697–705.
[44] V. Pracny, M. Meywerk, A. Lion, Full vehicle simulation using
thermomechanically coupled hybrid neural network shock absorber model,
Veh. Syst. Dyn. 46 (3) (2008) 229–238.
[45] T.D. Gillespie, Fundamentals of Vehicle Dynamics, SAE Technical Paper, 1992.
[46] J.Y. Wong, Theory of Ground Vehicles, John Wiley & Sons, 2001.
[47] M.M. El-Madany, A. El-Tamimi, On a subclass of nonlinear passive and semi-
active damping for vibration isolation, Comput. Struct. 36 (5) (1990) 921–931.