See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/293330196

Impact simulation and experiment on rubber anti-vibration systems

Article in Polymer Testing · February 2016

DOI: 10.1016/j.polymertesting.2016.02.001

CITATIONS READS

15 924

1 author:

Robert Keqi Luo

Central South University
80 PUBLICATIONS 941 CITATIONS

SEE PROFILE

All content following this page was uploaded by Robert Keqi Luo on 12 October 2017.

The user has requested enhancement of the downloaded file.

Polymer Testing 50 (2016) 335-342
Impact simulation and experiment on rubber anti-vibration
systems
Robert Keqi Luo 1, 2*

1
Department of Engineering and Technology, Trelleborg IAVS, Leicester, LE4 2BN, UK
2*
Department of Railway Engineering, School of Civil Engineering and Architecture, Central South University,
Changsha, Hunan, China, 410075

Abstract: There is very limited literature regarding impact analysis on solid rubber anti-
vibration systems as the true damping characteristics of rubber materials are very complex and
difficult to define. Viscoelastic approach is a usual method and has only achieved limited
success. In this paper, an integrated quasi-static and impact analysis with validation on an anti-
vibration mount is presented. The Rayleigh damping has been introduced for rubber hysteresis.
The impact responses from both simulation and experiment have been compared and shown
very good agreement in real time domain. In addition, it has revealed that real geometry and
elasticity of an impact object have to be included in simulation in order to obtain accurate
response. It has been shown that the proposed approach is reliable and can be used for an
appropriate design stage to evaluate an impact/dynamic response of rubber anti-vibration
systems. The key points to use this approach are also provided.
Keywords: Impact; rubber damping; rubber hysteresis; solid rubber model; dynamics

1. Introduction
Rubber springs are essential anti-vibration components for industrial applications. Rubber
components are widely used in order to minimize vibration levers generated from dynamic
environment. However, the real time-domain analysis for rubber structures under dynamic
loading are not well covered, especially for solid rubber components. There are many papers
dealing with quasi-static behaviour of rubber material using hyperelastic models. Luo et al
predicted the load-deflection response [1-4] and subsequent lifespan of rubber components [5-
7]. Sharma [8] calculated responses on an anti-vibration rubber component using three
hyperelastic models. It was found that the error for the stiffness varied about 3% to 40%
between the test and the simulation and suggested to correct the errors based on the experiment
data. Verron et al [9] predicted the continuous volumetric change in rubber including the
damage into a hyperelastic model based on the strain energy density. Lillbacka et al [10]
evaluated strain of rubber components using several hyperelastic models and indicated that
most of models gave good prediction up to 100% strain.
For many dynamic analyses, real rubber components have been converted into simple
individual elements in engineering applications. Magalini et al [11] simplified rubber
components as lumped springs and dashpot elements using concentrated parameters in a
frequency domain. The models used were either serial or parallel linked elements to calculate
the stiffness in different directions. Grassie [12] employed a lumped system with two degrees
of freedom to evaluate the effective dynamic stiffness on a rubber pad using a resonance

1
apparatus and made a good comparison between a laboratory test and real rail tracks.
Thompson et al [13] considered a rubber pad a non-linear spring to obtain a noise level when
wheel sets ran through a rail track section using a continuous supported model and a discrete
point supported model. Luo et al [14-17] performed rail vehicle dynamic evaluations and
service assessments in which rubber suspensions were simulated as damping elements.
For a dynamic simulation on the behaviours of a solid rubber structure, viscoelastic approach
has received intensive attention over many years. Kelvin-Voigt, Maxwell and fractional
Kelvin-Voigt models were investigated [18-21]. Busfield et al [22] employed a free oscillation
technique to measure the dynamic storage and the loss moduli of carbon black filled natural
rubber materials. These tests were performed with small oscillations that were superimposed
on a range of tensile pre-strains. It was observed that the dynamic storage and loss moduli did
not depend strongly on the pre-strain at small strains. Austrell et al [23] used a three-layer one-
dimensional model, called overlay method, to evaluate a response from a stationary cyclic load
on a rubber bush. However, the viscoelastic models have issues either in accuracy or in
complexity [21]. On the other hand, the Rayleigh damping has been widely used in structure
dynamics involving linear materials, including impact simulation on composite materials [24-
26]. Luo et al [27] has tried to introduce the Rayleigh damping into a solid rubber component
to perform a steady state dynamic analysis and the results were compared with the experimental
data.
There is very limited literature regarding impact analysis on solid rubber anti-vibration
components used in industries as the true damping characteristics of rubber materials are very
complex and difficult to define, especially on three-dimensional evaluations. In this impact
investigation, the rubber hysteresis is expressed in the form of the Rayleigh damping equations
instead of a usual viscoelastic form. In addition, the rubber anti-vibration component modelled
is a solid three-dimensional spring, not simplified as either one-dimensional spring (and/or
dashpot) elements or multi-layer linked simpler elements.
In this article, we first present static and impact experiments on a rubber anti-vibration
component. Then, we introduce a concept for Rayleigh damping and a hyperelastic model,
concluding with experimental validation following simulation of the component.

2. Experiment
An industrial anti-vibration spring, MDS (Multi-Directional Snubbing) mounting shown in Fig.
1(a), was selected for experiment. The product was made from natural rubber with shear
modulus of 0.7 MPa. The main dimensions of the spring are demonstrated in Fig. 1(b). They
are usually used in engine installations. This type of mountings is easy to install based on a
two-part single bolt installation. There is no requirement for radius or chamfered installation
hole and a steel flange prevents rubber wear at the bracket interface. The bonded steel snubbing
cup limits vertical movements and prevents excessive strain in rubber. The cup is encapsulated
in rubber to prevent corrosion. This design is a solution to a complex problem, achieving a
higher level of sophistication in isolation, noise attenuation and motion control, especially for
vehicles. In anti-vibration system design, quasi-static load-deflection curves are normally
required. Hence, a quasi-static experiment was first performed at loading rate 10 mm per
minute. The mounts were installed in a pair similar to a real engine installation with 5 mm
preload at both ends in vertical direction. A load-deflection curve up to approximately 5 kN is
presented in Fig. 2 (a). The shape of the load-deflection curve (S form) is a typical loading
response from rubber material. This quasi-static experiment was also served as a benchmark
2
for material properties used for further impact simulation analysis as the elastic property of the
rubber material is directly linked to the spring stiffness expressed in a stiffness matrix for both
static and dynamic simulations (see simulation section).

(a) MDS mount

(b) Main dimensions of the MDS mount

Figure 1 MDS mount and its main dimensions

In a real service environment for vehicles, geometric irregularities from the road surface
generate various dynamic events on suspensions and engine installations. The structural
components most affected by the dynamic loads are the rubber anti-vibration components. To
evaluate these effects correctly, an impact experiment was performed on the MDS mounts that
were bolted together and fixed in a test rig. A 250 kg mass was dropped on to the top washer
of the springs at a velocity of approximately 440 mm/s in a vertical direction. A deflection
response from these mounts was collected using a digital record system. A sample of the
deflection histories in one second frame is plotted in Fig. 2(b). The maximum impact deflection
is approximately 7 mm which occurred at first peak. The main response frequency is
approximately 10 Hz with some higher frequency deflections in the vicinity of the first three
rebound peaks.

3
 (a) Static load-deflection curve of the MDS mounts

(b) Impact response of the MDS mounts

Figure 2 Experimental results for the MDS mounts (static and impact)

3. Constitutive models and simulations

The general dynamic equations can be written as
̇ + [𝐾]{𝛿} = {𝐹}
[𝑀]{𝛿̈ } + [𝐶]{𝛿} (1)

Where [M] is the system’s mass, [C] is the system’s damping, [K] is the system’s stiffness,
{𝛿} is the deflection vector, {𝛿} ̇ is the system’s velocity vector, {𝛿̈ } is the system’s
acceleration vector, and {𝐹} is the external force vector.
For quasi-static analysis, the system’s mass matrix [M] and the system’s damping matrix [C]
can be ignored. The system’s stiffness matrix [K] is formed from the elastic part of the rubber
material that is usually modelled using a hyper-elastic expression of strain energy density. The
finite element software Abaqus [30] is used to perform the analysis.

3.1 Rubber material model and quasi-static simulation

In order to predict the quasi-static response of the rubber component, an appropriate material
model should be used. There are several hyper-elastic models are used to describe rubber

4
material based on the strain energy density [28-29]. These hyperelastic models can be
expressed in a general form:

𝑊 = 𝑊𝐼 (𝐼 )̅ + 𝑊𝐽 (𝐽𝑒𝑙 ) (2)

Where 𝑊𝐼 (𝐼 )̅ is the deviatoric part of the strain energy density of the primary material response
and 𝑊𝐽 (𝐽𝑒𝑙 ) is the volumetric part of the strain energy density.

𝐼 ̅ can be further expanded into 𝐼1̅ and 𝐼2̅ , which are the first and second deviatoric strain
invariants defined as

I1  1   2   3 and
2 2 2
(3)
2 2 2
I 2  1   2   3 , (4)

Where  i are the deviatoric stretches.

J el is the elastic volume ratio.

For this quasi-static analysis on the MDS mounts, a polynomial form of strain energy density
is used:

𝑊 = 𝐶10 (𝐼1̅ − 3) + 𝐶01 (𝐼2̅ − 3) + 𝐶20 (𝐼1̅ − 3)2 + 𝐶11 (𝐼1̅ − 3)(𝐼2̅ − 3) + 𝐶02 (𝐼2̅ − 3)2
+ 𝐶30 (𝐼1̅ − 3)3 + 𝐶21 (𝐼1̅ − 3)2 (𝐼2̅ − 3) + 𝐶12 (𝐼1̅ − 3)(𝐼2̅ − 3)2 + 𝐶03 (𝐼2̅
1 1 1
− 3)3 + (𝐽𝑒𝑙 − 1)2 + (𝐽𝑒𝑙 − 1)4 + (𝐽𝑒𝑙 − 1)6
𝐷1 𝐷2 𝐷3
(5)
Where Cij and Di are material parameters.
The material data for equation (5), which are typically either gathered by a given company or
externally contracted, were obtained from the manufacturer’s database, whose accuracy and
suitability for our engineering design and applications have been verified.
The finite element model of the rubber spring in pairs is shown in Fig. 3. Two materials have
been included in the component, one is mild steel (modelled as a linear material) and the other
is rubber. Three-dimensional finite elements with full integration and constant pressure
(C3D8H) were used for rubber and standard brick elements (C3D8) were used for mild steel.
The total degrees of freedom for the model is approximately 0.4 million.
A quasi-static simulation provides insights into the elastic behaviour of the rubber structure.
The response of a system subjected to a quasi-static load can be expressed by the following
equation
[𝐾]{𝛿} = {𝐹} (6)
Quasi-static simulation result and comparison with experimental data is shown in Fig. 4.

5
The two results are agreed well, which has confirmed the suitability of the rubber elastic model
and the stiffness matrix [K]. Hence, further impact analysis can be carried out based on the
verification of this quasi-static simulation.

Figure 3 Finite element model of the MDS mounts (a pair)

Figure 4 Static simulation and comparison with experimental data

3.2 Damping consideration and impact simulation

For impact simulation, both elastic characteristics and damping effect of the rubber material
are considered. Rayleigh method is employed to include the damping characteristics.
The damping [C] can be expressed as a function of mass [M] and stiffness [K] based on
Rayleigh assumption:
[𝐶] = 𝛼[𝑀] + 𝛽[𝐾] (7)

Where α and β are constants related to two free vibration modes of the rubber springs. Let 𝜔1
and 𝜔2 be the two specific natural frequencies of the modes, 𝜉1 and 𝜉2 are the damping ratio
associated with 𝜔1 and 𝜔2 . The relations between them can be expressed in the following
equations:
𝛼 𝜔 𝜔 𝜔2 −𝜔1 𝜉1
{𝛽 } = 2 𝜔21−𝜔22 [ −1 ]{ } (8)
2 1 −𝜔2 𝜔1−1 𝜉2

In real practice it can be assumed that the same damping ratio is applied to both 𝜔1and 𝜔2 . The
equations (8) can then be simplified to

6
 𝛼 2𝜉 𝜔1 𝜔2
{𝛽 } = { } (9)
𝜔1 + 𝜔2 1
The damping ratio 𝜉 can be obtained in several ways in an experiment [29], e.g. free-vibration
decay method, resonant amplification method and half-power method et al. 𝜔1 can be chosen
as the lowest frequency and 𝜔2 as the highest frequency over the range of interest frequencies.
Free vibration analysis should be performed before an impact analysis. The main objective is
to obtain free vibration modes, which are not dependent on external loading, and the
corresponding frequencies for determination of the damping coefficients.
The equations of motion for the free-vibration of a system without damping can be written as

[𝑀]{𝛿̈ } + [𝐾]{𝛿} = {0}

(10)

where 0 is a zero vector.

The preloading condition is the same as the quasi-static experimental settings with an additional
250 kg impact mass. To determine the conditions under which equation (4) allows motions to
occur, the Lanczos method was used.
Fig. 5 shows the first five vibration modes. The first mode (Fig. 5 (a)) is a vertical vibrating
shape in Y direction with frequency of 10.21 Hz, which is the same mode as the impact loading.
The second one (Fig. 5 (b)) is a shear shape in X direction and the third one (Fig. 5 (c)) is also
a shear mode in –X direction. Their frequencies are the same value (54.78 Hz), which indicates
that they are conjugate mode shapes. For the same reason, the fourth and the fifth mode shapes
(Fig. 5 (d) and (e)) are also conjugate mode shapes. The impact on the MDS mounts is along
the vertical direction and the response should be dominated in this direction. Hence, the first
mode is selected as a response base and the second one is selected for possible additional higher
frequency response. For the rubber material of the MDS mounts, damping ratio ξ was measured
in the laboratory using free-vibration decay method and the value was 0.04. Therefore, based
on equation (9) and using mode frequencies 𝜔1 = 10.26 HZ and 𝜔2 = 54.78 Hz , we can
obtain α = 0.69 and β =0.0012.

(a) The first mode shape with frequency =10.21 Hz

7
(b) The second mode shape with frequency =54.78 Hz

(c) The third mode shape with frequency =54.78 Hz

(d) The forth mode shape with frequency =55.60 Hz

8
 (e) The fifth mode shape with frequency =55.64 Hz

Figure 5 First five mode shapes

Figure 6 Impact histories of the simulation and the experiment using point mass method

For an impact analysis, the finite element model, used for free vibration evaluation has been
modified to an axisymmetric model for cost-savings as a large number of time steps are
required. Fig. 6 shows the simulation result and the comparison with the experimental data
using the axisymmetric model with the point impact mass. The first peak response from the
impact is correctly predicted in both magnitude and time scale. The magnitudes of the predicted
curve are also agreed well with the experimental ones. However, the response curve simulated
is so smooth without having any higher frequency interferences in the vicinity of first three
rebound peaks. It is not clear whether these interferences were generated by the real impact or
from the environmental noise. In addition, there are some differences on time scales between
the predicted response and the tested one from the first rebound peak. Further investigation was
conducted in order to ascertain the cause of these differences. The impact object is a deformed
body made from steel and has both dimensions and elasticity. Hence, a new finite element
model, including the impact block with real dimensions and elasticity, the link rod and the
supporting washer, is created. The impact block is a circular plate with a ring; the total thickness
is 100 mm with maximum 700 mm in diameter. The new finite element model is shown in Fig.
7. The comparison between the simulated result and the experimental data is shown in Fig. 8.
The predicted deflection is nearly identical to those observed from the experiment except a
small amplitude difference in the first rebound peak region and slight time delay after the 7th

9
cycle response. This comparison explains the phenomena of the higher frequency disturbance
in the vicinity of the rebound peaks and the time difference between the predicted result and
the experimental data using the simplified point mass method for impact simulation. It is
evident that the effect of the flexibility of the impact body needs to be taken into consideration
to obtain the accurate response. However, the simplified point mass method may be used for a
quick impact assessment as the calculated result from this method is good enough to provide a
reference for design engineers.

Figure 7 Finite element model with real impact body

Figure 8 Impact histories of the simulation and the experiment using real mass model

In dynamic analysis, stability and accuracy are the key control points to obtain reliable results.
To exam the accuracy of the simulation results from the view of the numerical integration
aspect, different allowable time increments are evaluated. Fig. 9 demonstrates this effect using
three maximum allowable time increments, i.e. 0.1, 0.01 and 0.001 second. All three
simulations are stable with different accuracies. The smaller time increments tend to give
shorter response time than the larger increments. In the first rebound peak region, some
predicted responses from the higher frequency impacts are missed using the larger time
increments. An impact response is also obtained using a maximum time increment 0.0001
second. The result obtained from 0.0001 second is the same as that from 0.001 second. Hence,
maximum 0.001 second is used to obtain the impact response.

10
Figure 9 Simulation results with different maximum time increment using real impact body

4. Conclusions
An integrated quasi-static and impact analysis with validation on the solid anti-vibration spring
has been presented. This engineering approach is different from usual viscoelastic based
approach. Rayleigh method for dynamic analysis is widely used in linear materials but rarely
used for non-linear materials with high damping like rubber, especially for impact applications.
This approach is simple and effective without complicated assumptions and parameters. The
key points to use this approach are:
A. A proper quasi-static analysis is needed to make sure that the elastic properties of the
material are correctly defined.
B. Proper damping coefficients should be carefully selected based on the free vibration
evaluations on the rubber spring and the external dynamic loadings.
C. A quick impact assessment could be performed using a simplified point mass for the impact
objects.
D. To obtain detailed evaluation accurately, the real geometry of the impact body with its
elastic properties needs to be included even though it would take much longer time.

This proposed approach may be suitable for solid rubber anti-vibration systems. The detailed
deformation and strain (stress) profiles of the components can be obtained and optimised at an
appropriate design stage. It is suggested that this methodology could be used for dynamic
design of rubber components in industrial applications. More engineering cases may also be
needed to further verify this approach.

Acknowledgements
The author would like to thank Mr. Andrea and Mr. Simpson in Trelleborg IAVS for their
technical discussions.

11
 REFERENCES

1. R. K. Luo, W. X. Wu & W. J. Mortel. Plastics, Rubber & Composites, 2000, 2(29),88-91

2. R. K. Luo, W. X. Wu & W. J. Mortel. Journal of Materials: Design & Applications, Vol. 217, No.4, 2003,
287-290
3. R. K. Luo, W. J. Mortel, P. W. Cook and J. Lake. Plastics, Rubber & Composites, 2011, 4(40), 155-160
4. R. K. Luo, L.M. Peng, X.P. Wu and W. J. Mortel. Journal of Rail and Rapid Transit, Volume 226 Issue 2,
March 2012 pp. 228 - 232.
5. R. K. Luo, W. X. Wu, P. W. Cook, & W. J. Mortel. Journal of Rail and Rapid Transit, 218, No.2, 2004,173-
177
6. R. K. Luo and W. X. Wu. Engineering Failure Analysis, Vol. 13, No. 1, 2006,110-116
7. R. K. Luo, W. J. Mortel & X. P. Wu. Engineering Failure Analysis, Vol. 16, No.5, 2009,1366-1378
8. S. Sharma. In Constitutive Models for Rubber III (Eds J.J.C. Busfield, A.H. Muhr), 2003, pp. 161-167 (A.A.
Balkema, Rotterdam, The Netherlands).
9. E. Verron, G. Chagnon and J.-B. Le Cam. In Constitutive Models for Rubber VI (Eds G. Heinrich et al), 2010,
pp. 279-284 (A.A. Balkema, Rotterdam, The Netherlands).
10. R. Lillbacka, E. Johnson and M. Bellander. In Constitutive Models for Rubber VI (Eds G. Heinrich et al),
2010, pp. 241-248 (A.A. Balkema, Rotterdam, The Netherlands).
11. Magalini A. et al. In Constitutive Models for Rubber III (Eds J.J.C. Busfield, A.H. Muhr), 2003, pp. 365-372
(A.A. Balkema, Rotterdam, The Netherlands).
12. S. L. Grassie. Proc Insln Mech Engrs, 2007 203, 25-32
13. D. J. Thompson, C. J. C. Jones, T. X. Wu and G. D. France. Proc Instn Mech Engrs, 1999, 213, 233-241
14. R. K. Luo, B. L. Gabbitas & B. V. Brickle. Vehicle System Dynamics, 25(s1), 1996, 438-449
15. R. K. Luo, B. L. Gabbitas & B. V. Brickle. Engineering Failure Analysis, 1(3), 1996, 53-64
16. R. K. Luo, B. L. Gabbitas & B. V. Brickle. Journal of Rail and Rapid Transit,208, 1994, 123-132
17. R. K. Luo, B. L. Gabbitas & B. V. Brickle. Vehicle System Dynamics,23, 1994,334-345
18. K.L. Knothe and S.L. Grassie. Vehicle System Dynamics, Vol. 22, 1993, 209–262.
19. M. Enelund and P. Olsson. International Journal of Solids and Structures, 1999, Vol. 36, 939–970.
20. N. Gil-Negrete, J. Vinolas and L. Kari. In Constitutive Models for Rubber IV (Eds P. –E. Austrell, L. Kari),
2005, pp. 479-485 (A.A. Balkema, Rotterdam, The Netherlands).
21. Tarrago M.J., Gil-Negrete N. Garcia and J. Vinolas. Int. J. Vehicle Design, 2009, 49(4), 303-317.
22. J.J.C. Busfield, C. Deeprasertkul & A.G. Thomas. In Constitutive Models for Rubber (Eds Al Dorfmann and
A. Muhr), 1999, pp. 87-93 (A.A. Balkema, Rotterdam, The Netherlands).
23. P.-E. Austrell, A.K. Olsson and M. Johnson. In Constitutive Models for Rubber II (Eds D. Besdo et al), 2001,
pp. 231-236 (A.A. Balkema, Rotterdam, The Netherlands).
24. R.K. Luo, E.R. Green, C.J. Morrison. Journal of composite part B: Engineering, Vol. 32, No.1 2001, 513-520
25. R. K. Luo. Composites Sciences and Technology, 2000, 60, No.1, 49-58
26. R. K. Luo, E. R. Green & C. J. Morrison. International Journal of Impact Engineering, 1999, 4(22), 435-448
27. R. K. Luo, X.P. Wu and W. J. Mortel. J Rail and Rapid Transit 227(1),2013, 103–112
28. R. W. Ogden. 1984 (Ellis Horwood, Chichester).
29. Ray W. Clough and Joseph Penzien. 1993, McGraw-Hill, Inc.
30. Dassult Systems. 2014 (Dassult Systems, USA).

12

View publication stats