SAE TECHNICAL

PAPER SERIES 2002-01-0319

The Influence of Damper Properties on

Vehicle Dynamic Behaviour
Adrian Simms and David Crolla
The University of Leeds, School of Mechanical Engineering

Reprinted From: Steering and Suspension Technology Symposium 2002

(SP–1654)

SAE 2002 World Congress

Detroit, Michigan
March 4-7, 2002

400 Commonwealth Drive, Warrendale, PA 15096-0001 U.S.A. Tel: (724) 776-4841 Fax: (724) 776-5760
The appearance of this ISSN code at the bottom of this page indicates SAE’s consent that copies of the
paper may be made for personal or internal use of specific clients. This consent is given on the condition,
however, that the copier pay a per article copy fee through the Copyright Clearance Center, Inc. Operations
Center, 222 Rosewood Drive, Danvers, MA 01923 for copying beyond that permitted by Sections 107 or
108 of the U.S. Copyright Law. This consent does not extend to other kinds of copying such as copying for
general distribution, for advertising or promotional purposes, for creating new collective works, or for
resale.

Quantity reprint rates can be obtained from the Customer Sales and Satisfaction Department.

To request permission to reprint a technical paper or permission to use copyrighted SAE publications in
other works, contact the SAE Publications Group.

All SAE papers, standards, and selected

books are abstracted and indexed in the
Global Mobility Database

No part of this publication may be reproduced in any form, in an electronic retrieval system or otherwise, without the prior written
permission of the publisher.

ISSN 0148-7191
Copyright © 2002 Society of Automotive Engineers, Inc.

Positions and opinions advanced in this paper are those of the author(s) and not necessarily those of SAE. The author is solely
responsible for the content of the paper. A process is available by which discussions will be printed with the paper if it is published in
SAE Transactions. For permission to publish this paper in full or in part, contact the SAE Publications Group.

Persons wishing to submit papers to be considered for presentation or publication through SAE should send the manuscript or a 300
word abstract of a proposed manuscript to: Secretary, Engineering Meetings Board, SAE.

Printed in USA
 2002-01-0319

The Influence of Damper Properties on Vehicle

Dynamic Behaviour
Adrian Simms and David Crolla
The University of Leeds, School of Mechanical Engineering

Copyright © 2002 Society of Automotive Engineers, Inc.

ABSTRACT out prior to the manufacture of any vehicle prototypes,

and this technique would be far less subjective in nature.
This paper details a non-linear hysteretic physical shock Such an approach would necessitate an improved
absorber model, and the processes utilised to identify method of characterising the damper, such that the
the constituent parameters. In the current paper the important dynamic features are represented, and
model parameters are extracted from experimental data comprehension of the links between more subtle
for the ‘sport’ setting of a prototype front shock features of the damper response. The current paper
absorber for a vehicle in the luxury class. The model is attempts to address the aforementioned requirement for
validated by comparing simulated results to experimental improved damper characterisation in the context of the
data for a test damper, for three discrete frequencies of ‘sport’ setting of a triple rate prototype adaptive shock
sinusoidal excitation of 1,3 and 12 Hz. Finally the shock absorber.
absorber model is included in a quarter car vehicle ride
model and output characteristics are compared to those SHOCK ABSORBER MODELLING APPROACH
obtained with classical damper representations.
In order to select the optimum damper modelling
INTRODUCTION strategy for a ‘virtual damper tuning environment’, the
suitability of the differing approaches found within
The detailed dynamic properties of dampers are known published literature were determined with respect to the
to influence substantially some of the subtle, and yet following criterion:
nevertheless hugely important, refinement aspects of
vehicle ride and handling. However, damper properties · Ability to capture damper non-linearity and dynamic
are typically characterised by quasi-steady properties for behaviour.
vehicle simulation purposes. The classic 14 speed test · Flexibility to model different shock absorber types.
[1], for example, involves subjecting a damper to 14 · Ease of model generation (Experiment/Parameter
differing frequency levels of fixed amplitude sinusoidal identification).
excitation, and then plotting the peak force values · Suitability for use in vehicle simulations.
obtained versus the relevant test velocity. Such a · Usefulness as a predictive tool.
representation of shock absorber behaviour is clearly
deficient for the purpose of vehicle simulations as only a Clearly black box methods such as the Restoring Force
snap shot view of the damper’s behaviour is utilised and Mapping method [2-5] and neural networks [6], do not
much information is discarded. As a direct consequence satisfy the need for a predictive tool as they are
the process of damper valve tuning is still carried out to inherently non-tuneable. The same applies to
a great extent via ride work. This consists of ride elementary models constructed with spring and ideal
engineers subjectively rating the performance of the viscous damping elements [7,8]. The required need for
prototype vehicle(s) over a series of test tracks/ride tuneable elements for this damper model application
routes. The damper valve is then possibly adjusted to lends itself to an explicit physical model, where the
improve the vehicle’s ride/handling or to obtain damper control force is related to physical parameters
characteristics expected for that class of vehicle. This that govern the dampers internal flows and pressures.
process is clearly open to subjective evaluation which
may vary driver to driver, and even from one day to the The high detail model developed by Lang [9,10], for a
next. twin tube shock absorber, was the first physical model
that actually aimed to predict damper behaviour over a
A more scientific approach to the issue of damper wide range of operating conditions. This 87 parameter
tuning, via vehicle simulations, would offer a number of model gave good correlation with experiment, but both
significant benefits. This is simply due to the fact that the simulation and parameter identification processes were
bulk of the damper selection process could be carried highly iterative. It was also very specific to the shock
absorber under investigation. It did however identify the higher velocities the blow-off valve progressively opens
contribution of internal compliance’s to the hysteretic thus reducing the rate of increase of damper force with
nature of dampers. respect to velocity.

More recently models by Duym [11-13], Lang and Piston valve Compression
Sonnenburg [14] and Herr et al [16] attempted to Rebound assembly chamber, Pcom
generate more readily identifiable physical damper Chamber
models. Each of these models are similar in overall Preb Gas
structure in that they consist of a pressure model(s) and phase,
valve pressure/flow characterisations. The pressure
Pcom,gas
models are a selection of first order non-linear
differential equations, which are utilised to determine the
various internal chamber pressures, and are derived x, x&
from pressure dependent oil compressibility models. For x0
the pressure/flow model Duym [11-13] identified valve Lpt
parameters from several simple dynamometer tests and
an additional ‘incompressible’ model (see definitions);
these parameters were then used to analytically P1
determine valve flows for given pressure drops in the
main model. In contrast Lang and Sonnenburg [15] used Port Restriction
experimentally obtained pressure/flow data, and Herr et
al evaluated the pressure flow characteristics of the
differing valve components using Computational Fluid Leak
Dynamics (CFD). In the absence of a large database of Restriction
pressure/flow data for differing valve formulations, as
was available to Lang and Sonnenburg, this non Blow-off valve
parametric form of pressure model does not satisfy the
required criteria of tune-ability. The CFD approach is a P2
time consuming process requiring detailed modelling
even for varying damper valves of the same basic Flow into Rebound chamber +ve.
architecture. As a result of the aforementioned
considerations the damper modelling approach to be Flow into Compression chamber -ve.
taken is one of a physical model similar to that of Duym Control force in Rebound +ve.
[11-13].

SHOCK ABSORBER ARCHITECTURE, AXES Figure 1. Mono-tube layout / axis system (top), and flow path
SYSTEMS AND SIGN CONVENTIONS architecture (bottom).

The prototype shock absorber, in the ‘sport’ setting, falls MONO-TUBE DAMPER MODEL
into the mono-tube category; the only flow paths for the
internal hydraulic fluid are thus through the piston’s The twin-tube damper model presented in [11-13] is a
valves (see Figure 1). The additional ‘softer’ damper compressible model which is used solely for the purpose
settings, which are achieved by the solenoid activation of of simulations. To yield a readily identifiable damper
additional flow paths, will not be considered in this model the parameters of the compressible model of [11-
paper. The mono-tube shock absorber consists of a 13] were identified by fitting an additional incompressible
rebound chamber and a compression chamber which model to experimental data from which the hysteresis
are both oil filled. A high pressure nitrogen gas volume, had been removed as described in [16,17]. As a result of
typically 20-30 Bar, is present at the end of the pressure the differing model architectures the pressure/flow
tube, separated from the damper fluid by way of a model, required for the parameter identification
floating piston. Expansion and compression of this gas processes, was different to that of the full compressible
volume compensates for the differing volumetric model. In this paper a single pressure/flow model will be
changes encountered in bump and rebound strokes as a derived for a mono-tube shock absorber for the
result of the piston rod’s presence in the rebound parameter identification process, and the valve
chamber. Several valve assemblies are present on the pressure/flow characteristics will then be supplied to the
piston, one for compression and one for extension compressible damper model in the form of a simple
strokes. look-up table. This modelling approach should yield
more rapid simulations and additional robustness without
The flow path architecture of these valves are outlined in compromising model tune-ability. As will be discussed
Figure 1. For low damper velocities the preloaded blow- later this can also yield improvements in model accuracy
off valve remains closed and the net pressure/flow and flexibility; for non-standard valve architectures it will
characteristic can be described by the series now only be necessary to derive a single new
combination of the port channel and the fixed bleed. For pressure/flow model.
PRESSURE MODEL -The contribution of the damper & = x& ( A - A )
Q tot ,reb® com = - Vreb pt rod
oil’s compressibility to the hysteretic nature of shock (7)
absorbers dynamic behaviour is modelled with the Q tot ,com® reb = -(Q reb® com ) = - x& ( A pt - A rod )
assumption that the relative change in oil volume is
proportional to pressure:
VALVE PRESSURE/FLOW CHARACTERISTICS -The
individual valve characteristics of the port and leak
V0 - V restrictions can be characterised by simple power law
= ap (1)
V0 expressions. An exponent of 1.75 was proposed by
Reybrouck [18] and shown to be accurate for a range of
damper valves [11-13,15-17].
The mono-tube pressure model consists of a first order
non linear differential equation, (2), derived from (1) for 7/4
the rebound chamber and a simple relationship from the DPport = K port n 1 / 4 Q port (8)
adiabatic gas law which relates the compression
chamber pressure to piston position (3). 7/4
DPleak = K leak n1 / 4 Q leak (9)
dp reb ( x& ( A pt - A rod ) - Q pv )(1 - ap reb )
= (2) The blow-off valve pressure/flow characteristic is
dt ( L pt - x 0 - x )( A pt - A rod )a modelled as a function of two independent parameters,
DP0, the pressure required to overcome the valve pre-
g load and the subsequent valve stiffness, Kspring. This
æ Vgas,static ö expression is a simplified version of that utilised by Lang
Pcom = Pgas,static ç ÷ (3)
çV + A x ÷ [9,10] proposed by Duym et al [11,17].
è gas,static rod ø

Assuming a simple Coulomb friction model the damper

K spring Q blow - off = ( Dp blow - off - Dp o ) Dp blow - off (10)
force is thus given by:
The total valve characteristics are represented as a
Fdamper = (A pt - A rod )Preb - A pt Pcom + Ffriction sgnx& (4) combination of the individual valve characteristics in
series and in parallel. For two valves in series the
pressure drop over the total valve is the sum of the
For a full derivation of (2) the reader is referred to [13]. individual pressure drops and the flow through each is
equal. For a parallel combination the total flow is the
INCOMPRESSIBLE MONO-TUBE DAMPER MODEL -In sum of the flow through each component and the
this section an incompressible mono-tube damper model pressure drops are equal:
will be derived that evaluates damper control force as a
function of velocity, utilising the axis system and sign
conventions detailed in Figure 1. The model structure is
Dp tot = Dp leak + Dp port (11)
very similar to that presented in [11-13] for a twin-tube
shock absorber. Substitution of the port’s pressure/flow characteristic, (8),
yields:
CALCULATION OF THE VALVE FLOW RATES -
Since the damper hydraulic fluid is assumed to be Dp tot = K port n1 / 4 Q port
7/4
+ Dp leak (12)
incompressible the rebound chamber volume may be
directly determined at any time increment from the
damper’s geometry. The port flow can then be described as the sum of the
leak and blow-off flows:
Vreb = ( L pt - x 0 - x )( A pt - A rod ) (5)
Dp tot = K port n1 / 4 (Q leak + Q blow off )7 / 4 + Dp leak (13)
Differentiation of this expression yields an expression for
the rate of change of the chamber volume with respect While the blow-off valve remains closed the leak flow
to time: equals the total flow. Inserting this into (9) yields:

& = - x& ( A - A )
V (6) DPleak = K leak n1 / 4 Q tot
7/4
(14)
reb pt rod closed closed

From (6) it is thus possible to define the flow rates Substituting this expression into (13), and for zero blow-
through the individual damper valves to the defined sign off valve flow thus yields an explicit relationship for the
convention: pressure drop across a closed valve as function of flow
rate:
Dp tot closed
= (K port + K leak ) n1 / 4 Q tot ( closed
)7/4
(15)
A third order Taylor series expansion yields a cubic
equation which can then be solved utilising Cardano’s
rule [19]. The Cardano coefficients are presented in
For 2 individual valves in parallel: Appendix 2. The resulting explicit representation for the
leak flow is as follows:
Q tot = Q leak + Q blow - off (16)
B2
Q leak open
= Q leak , blow - off - +
Rearranging (16) to gain an expression for Qblow-off and 3B 3
substituting into (10) yields: (22)
D D
3 - 2 + D3 + 3 - 2 - D3
K spring (Q tot open
- Q leak open
)= 2 2
(17)
( Dp blow - off - Dp o ) Dp blow - off Now at the blow-off point DPleak is equal to DP0.
Rearranging (9) and inserting this condition yields:
For 2 individual valves in parallel the pressure drop is
4/7
the same across each element, therefore for pressures æ Dp 0 ö
greater than the blow-off pressure, DPleak is equal to Q leak ,blow off = çç 1/ 4
÷÷ (23)
DPblow-off. Substituting this into (9) leads to the following è K leak n ø
expression:
From (9) the pressure drop across a leak restriction for
7/4
DPblow - off = K leak n1 / 4 Q leak open
(18) open blow-off valve flow is given by:

7/4
Combination of (17) and (18) yields an explicit DPleak open
= K leak n1 / 4 Q leak open
(24)
relationship for the total open valve flow in terms of the
leak flow: The pressure drop over the port restriction is then
calculated using (8). Noting that the flow through the port
Q tot open
= Q leak open
+ restriction is equal to the total flow:

7/4 7/4 7/4

( K leak n1 / 4 Q leak open
- Dp o ) K leak n1 / 4 Q leak open
DPport = K port n1 / 4 Q tot open
(25)
open

K spring
The total pressure drop over the valve assembly is then
obtained by adding the pressure drops of the series
combination of individual valves:
(19)

AN EXPLICIT OPEN VALVE CHARACTERISTIC - In Dp tot open

= Dp leak open
+ Dp port open
(26)
order to avoid an iterative numerical solution of (19), to
determine the leak flow rate for a given total flow rate,
the right hand side of (19) is expanded into a Taylor Substituting (24) and (25) into (26) yields the final
series about the blow off point, i.e. Qleak= Qleak,blow-off : expression for the total pressure drop over an open
valve assembly:

( )
3 i

Q tot = å B i (Q leak
7/4 7/4
- Q leak , blow - off ) Dp tot open
= n 1 / 4 K leak Q leak open
+ K port Q tot open
open
i =0 (20)
for Q leak > Q leak , blow - off (27)

with: A SMOOTHED BLOW-OFF CHARACTERISTIC - The

outlined model results in a sharp transition between
‘closed valve’ and ‘open valve’ operating scenarios when

Bi =
1 ¶i
i! ¶ Q leak open i
Q tot open
(Q leak open ) viewed on a pressure/flow diagram. In reality this
transition between regimes is rounded off possibly due
Q leak = Q leak , blow - off
to leakage through the blow-off valve prior to the blow-off
condition being met. The following empirical formula is
thus utilised to provide a smoothing effect:
(21)

The derived Taylor series coefficients are given in

Appendix 1.
 Dp tot ´ Dp tot routines were implemented in Matlab (v5.3) using the
open closed sub routine, Lsqnonlin, which is provided within the
Dp smooth = (28)
G G optimisation toolbox. At each time increment, ti, for a set
G Dp tot open
+ Dp tot closed of parameters, P, and for N samples of experimental
data the squared difference between the modelled
output of interest, Umodelled, and the corresponding
For a closed blow-off valve the Taylor series
experimental input, Umeasured, was evaluated with respect
approximation cannot be utilised to represent the
to a set of experimental inputs, Imeasured viz:
pressure drop over an open valve hence the explicit
characteristic given by (22) and (27) is linearly 2
extrapolated at the blow-off point as follows: N
é U mod elled ( t i , I measured ( t i ), P) ù
C( P) = å ê ú (33)
i =1 ë - ( U measured ( t i , I measured ( t i )) û
Dp tot open
= H 0 + H 1 (Q tot - B 0 ) (29)
The model parameters, P, were thus optimised to
Where H0 and H1 are the extrapolation coefficients minimise the function, C.
defined in appendix 3.
To facilitate an efficient optimisation process the
For the compressible damper model of Duym [11-13], Jacobian of the modelled signal was derived and
where the valve flows are calculated as a function of implemented into the optimisation routines. If the
pressure drop, it is necessary to smooth the calculated experimental data, to which the model is to be fitted, is a
flow, as opposed to pressure, and with an alternative matrix of m components and n is the number of
empirical formula to that used in the incompressible parameters then the Jacobian, J, is an m-by-n matrix
model(28). As such this is a possible source of where J( i, j ) is the partial derivative of Umodelled(i) with
inaccuracy when switching between the models. respect to P(j). Explicitly defining the Jacobian makes
the optimisation process more rapid simply due to the
THE DAMPER FORCE - As a result of the differing fact that if the Jacobian is not supplied it has to be
areas in the rebound and compression chambers a static evaluated numerically which is more costly in terms of
force exists at rest which is given by the following time.
equation:
A simple measure of model fit quality is given by the
Fstatic = - A rod Pgas,static (30) RMS of the residual signal viz.

In order to facilitate the parameter identification process C

RMS of the residual = CRMS = (34)
the compression chamber pressure is assumed to N
remain constant at the static gas pressure, Pgas,static.
Therefore: A relative measure of the fit quality can also be obtained
by comparing the residual RMS to the RMS value of the
Preb = Dp + Pgas,static (31) measured signal:

Substitution of (31) with (30) into (4) yields the desired C RMS
expressions for the damper force: Relative RMS of Residual = (35)
U Measured - RMS
Frebound = Fstatic + Dp( A pt - A rod ) + Ffriction sgn x&
(32) QUASI-STATIC TEST - For the selected test velocity
-1
Fcompression = Fstatic - Dp( A pt - A rod ) + Ffriction sgn x& of 4mms the generated damper force was assumed to
consist of little viscous damping and could thus be
attributed to the compression/expansion of the gas
PARAMETER IDENTIFICATION volume contained within the shock-absorber and due to
friction. To ensure that this assumption was valid the
For the current paper the same parameter identification quasi static test was also performed for the ‘comfort’ and
approach was taken to that of Duym et al [12,13,17]. ‘normal’ settings of the prototype shock absorber. A
This approach will be briefly introduced here. comparison of these softer damper settings to that of the
Experimental data was collected from two dynamometer ‘sport’ setting showed little variation in results between
tests. The first being a quasi-static test, where the shock the three. Assuming the pressures are equal in each of
absorber was extended and compressed at a constant the three chambers the following applies:
low velocity. For the second test the shock absorber was
excited by a sine chirp signal.
F = Pcom ( A pt - A rod ) - Pcom A pt + sign( x& ) Ffriction (36)
The parameter values were then determined through the
process of least squares minimisation. Optimisation
Simplifying and applying the reserve chamber isentropic Hz signal which was 0.005m. These simulated results
law based function (3) yields: were then compared to those obtained from experiment
for identical excitation signals.
g
æ Vgas,static ö
F = - A rod Pgas,static ç ÷
çV + A x ÷ (37)
1
è gas,static rod ø
0.8
data
hysteresis filtered
+ sign( x& ) Ffriction 0.6

0.4
The static gas pressure and volume and friction value
0.2
were thus identified from (37) and experimental data

velocity, m/s
using the aforementioned technique of least squares 0
minimisation. Figure 2 shows the predicted and -0.2
measured force/displacement characteristics. The
-0.4
calculated value for the RMS of the residual was 6.7 N,
indicating a close match has been obtained. -0.6

-0.8

-320 -1

-340 0 1 2 3 4 5 6
time, s
-360
Figure 3. Velocity/time plot for chirp signal, and corresponding plot with
-380 simulated hysteresis removed.
force, N

experimental
-400
5000
-420
Data
4000 Modelled
-440
3000
-460
2000
-480
force, N

-0.05 0 0.05
1000
displacement , m
0
Figure 2 . Quasi-static test/simulation results (sport)
-1000
SINE CHIRP EXCITATION TEST - The parameters for
the incompressible model were optimised using the -2000
previously defined techniques using data obtained from
a ‘modified’ sine chirp excitation signal. The chirp signal -3000
-1.5 -1 -0.5 0 0.5 1
shown in Figure 3 was utilised, as in [11-13,17], so that vel m/s
the required frequency range would be swept through
exponentially preventing prolonged excitation at elevated Figure 4. Experimental force/velocity identification data with hysteresis
velocities. Hysteresis was consequently removed from removed overlaid with incompressible damper model prediction.
the experimental data, prior to the parameter
identification process, by only retaining data where Looking at Figure 5 it can be seen that the simulated
velocity and acceleration had different signs. For an in curves match the experimental data very well for each of
depth description of the hysteresis ‘filtering’ and data the three frequencies of excitation that were evaluated.
pre-processing steps taken the reader is referred to [17].
Figure 4 shows a plot of the resulting force/velocity data All of the parameter values listed in Figure 6 are either
superimposed with that predicted by the incompressible identified from the two experimental procedures or
damper model. As can be seen the incompressible gathered from manufacturers data such as the geometry
model matches the experimental data closely. based parameters. At present the only parameter that is
not identified experimentally or taken directly from data
VALIDATION OF MONO-TUBE DAMPER MODEL sheets is the one for compressibility, a. In order to
account for compliances other than that of the oil itself it
In order to validate the described damper model the is preferable to subtly adjust the value of oil
output characteristics of interest were simulated for compressibility which can be obtained from data sheets.
-9 -1
sinusoidal excitations of 1, 3 and 12 Hz. All of the sine For this study a value was taken of 1.5e Pa .
wave amplitudes were 0.05m with exception of the 12
 3000 3000

2500 2500

2000 2000

1500 1500
measured 1Hz
1000 1000 simulated 1Hz
force N

force N
500 500

0 0

-500 measured 1Hz -500

simulated 1Hz
-1000 -1000

-1500 -1500
-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 -0.05 0 0.05
vel m/s displacement m

6000 4000

3000
4000
2000

2000
1000
force N

force N measured 3Hz

0 0
simulated 3Hz

-1000
-2000 measured 3Hz
simulated 3Hz -2000

-4000 -3000
-1 -0.5 0 0.5 1 -0.05 0 0.05
vel m/s displacement m

3000
3000

2500
2000
2000

1000 1500
force N

1000
force N

0
500

0
-1000
measured 12Hz -500 measured 12Hz
simulated 12Hz simulated 12Hz
-2000 -1000
-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4
vel m/s -1500

Figure 5. Experimental and simulated force/velocity and -2000

force/displacement plots for differing sinusoidal excitation frequencies. -6 -4 -2 0 2 4 6
displacement m -3
x 10
 improvements that the proposed damper model offer
when compared to conventional representations by
Geometric/ Magnitude Units comparisons to vehicle objective data. In order to ensure
data-sheet that the comparative study focuses on variations
Parameters between the overall damper modelling approaches, the
14 speed test approach was also simulated for an
Lpt 0.453 m elevated number of test frequencies to eliminate
X0 0.1355 m inaccuracies/variations due to exact frequency selection
and as a result of a 28 point fit to a force/velocity curve .
frod 13e-3 m
fpiston 36e-3 m
Mb z2
g 1.4 N/A
2
n 21e-6 m /s
Cs Ks
Friction 43.7 N
Quasi-static

parameters

3
Vgas,static 1.2299e-4 m
identified

Mw
z1
Pgas,static 2.9995e+6 Pa
Kt
x0
Compression Rebound Units
parameters
Identified

Figure 7. Quarter car ride model.

Valve

The equations of motion for the quarter car model shown

27/4 in Figure 7 are:
Kleak 6.54e+014 7.63e+014 kg/m

27/4
M w &z&1 = K t (x 0 - z 1 ) - K s (z 1 - z 2 ) - C s (z& 1 - z& 2 )
Kport 2.02e+012 1.20e+011 kg/m (38)
M b &z& 2 = K s (z 1 - z 2 ) + C s (z& 1 - z& 2 )
3/2 2 9/2
Kspring 1.56e+012 6.50e+012 Kg /s m
The input parameters for this model are outlined in
6.82e+005 2.74e+006 Pa Figure 8.
Dp 0
Two types of road surface input were applied to the
G 1.7 1.5 N/A quarter car model, one representing a length of
simulated road, and one of a standard feature, namely a
pothole. Time domain results were recorded and then
Figure 6. Non-linear mono-tube damper model parameters. converted to the frequency domain in the case of the
‘random road’ in order to evaluate weighted RMS
QUARTER CAR RIDE MODELLING vertical body accelerations according to ISO 2631 [21].

This section describes a simple investigation to quantify Parameter Magnitude Units

any improvements in accuracy that a hysteretic damper
model might have on simulated vehicle behaviour when Mw 55.5 kg
compared to simpler damper representations. The
Mb 451.8 kg
proposed damper model was incorporated into a
classical vehicle quarter car ride model, Figure 7, and Kt 250 kN/m
both time and frequency domain related characteristics
were compared to those obtained with conventional 14 Ks (Wheel rate) 29.7 kN/m
speed test and linear damper representations. To ensure Cs (Damper rate) 2622 Ns/m
a fair comparison for this section the 14 speed test was
simulated from the non-linear damper model, and not Damper Motion 1.5 N/A
performed experimentally. In this way the investigation Ratio
may be carried out in isolation of the effects of slight
inaccuracy of the hysteretic damper model compared to
actual characteristics. The purpose of this section is to Figure 8. Parameters for quarter car ride model (Front)
identify the need for a hysteretic damper model but,
clearly there is also a requirement to validate the exact
RANDOM ROAD INPUT MODELLING - For this study damper will provide higher forces and hence ACC values
a time dependant road profile was derived from a and vice versa for lower vehicle speeds.
spectral density road description, defined in terms of
frequency, in cycle/s, as described by Crolla et al [20].
The spectral density is given by the following simple 0.06
formula, where G is a roughness coefficient, V is vehicle
speed in m/s and p is equal to the gradient of the log-log
spectral density curve: 0.04

GV p -1
S(f ) = (39) 0.02
fp

SWS m
0
For the first section of this study a vehicle speed of
45mph (20m/s) was selected along with typical values
-6
for a minor road of roughness coefficient of 5x10 , and -0.02
an index p value of 2.5.
linear
The linear damper rate, given in Figure 8, was carefully -0.04 14 speed
selected to yield an equivalent compromise between model
RMS Dynamic Tyre Load (DTL) variation and RMS
weighted vertical body acceleration (ACC), to that of the -0.06
non-linear hysteretic damper model, for the defined 0 0.5 1 1.5 2
time s
vehicle speed of 20m/s. The damper rate selected is
slightly softer than that required to minimise DTL for this
condition. Figure 9. Simulated SWS versus time for first 2 seconds of 500m of
typical minor road being traversed at 20m/s, for three differing damper
representations.
Figure 9 shows the predicted time histories for
Suspension Working Space (SWS), which is given by z1-
z2, for each damper representation, for the initial 2 Units Linear 14 Non-
seconds of simulation. An interesting feature is that the damper speed linear
linear damper predictions are generally vertically offset damper damper
relative to each of the other two damper representations. SWS (RMS) m 0.026 0.028 0.028
This general trend is maintained for the full length of 2
simulated road of 500m. This offset yields a lower peak Weighted m/s 2.26 2.29 2.33
value of SWS in damper compression and a higher Body accel
value in extension compared to the other damper (RMS)
representations, see Figure 10. This is clearly a
DTL (RMS) 1.63 1.69 1.59
manifestation of the effect of asymmetric damping for the
14 speed and modelled dampers, with rebound being Peak SWS m 0.078 0.091 0.091
higher than bump.
-0.091 -0.079 -0.079
2
A further interesting result is that for similar predicted Peak body m/s 8.52 7.92 8.02
weighted vertical body accelerations and RMS SWS accelerations
values the modelled non-linear hysteretic damper model -10.50 -9.72 -9.67
predicts a RMS DTL variation of 6% less than that for Peak DTL kN 5.18 5.86 5.33
the 14 speed test representation with peak DTL’s some
10% lower. -5.68 -5.85 -5.47

Figure 11 shows the variation of the discomfort

Figure 10. Simulated RMS and Peak output values for 20m/s
parameter, ACC, for each of the damper models, for a simulation over 500m of typical minor road for three differing damper
range of vehicle speeds over the defined typical minor representations.
road. The linear damper curve intersects the modelled
damper curve in the region of 20 m/s. This is to be When comparing the outputs obtained for the 14 speed
expected due to the method chosen to select the linear test damper to the non-linear hysteretic damper model
damper rate. It can be seen that the discrepancy several overall trends are displayed. For low vehicle
between the predicted ACC values for the linear damper speeds the predicted discomfort parameter is similar,
and the other two representations are proportional to the however for higher speeds there is more variation.
percentage speed increase or decrease relative to this Generally for all speeds the damper model predicts
point. Again this trend is to be expected since for higher higher ACC values than the 14 speed damper.
vehicle speeds, and hence damper velocities, the linear
 levels of damping relative to the other two damper
4 representations.

3.5 Units Linear 14 Non-

damper speed linear
weighted rms body accel (m/s 2)

3 damper damper
SWS (RMS) m 0.0325 0.04201 0.0430
2.5
2
Body accel m/s 8.21 6.62 6.52
2
(RMS)
1.5 DTL (RMS) kN 5.51 6.66 6.55
linear damper
14 speed damper Peak SWS m 0.0416 0.0670 0.0683
1 damper model
-0.0624 -0.0715 -0.0733
0.5
5 10 15 20 25 30 35 40 2
Peak body m/s 17.85 11.37 10.52
vehicle speed (m/s) acceleration
-14.41 -11.30 -11.094
Figure 11. Discomfort parameter versus vehicle speed for three
differing damper models traversing a typical minor road.
Peak DTL kN 15.31 14.87 14.03
STANDARD FEATURE - A pothole matching the -11.184 -11.255 -10.977
geometry of that given in Figure 12 was applied as the
road input to the quarter car ride model. This is an
example of a standard pothole that is used by Jaguar for
test work at MIRA [20]. Figure 13. RMS and Peak output values for 10m/s simulation over a
typical pothole for three differing damper representations.
15mm

155mm 300mm 155mm

0.1
60mm

z1-z2 m

0
linear
14 speed
model
Figure 12. Standard pothole geometry. -0.1
0 0.05 0.1 0.15 0.2

For this investigation a vehicle speed of 10 m/s was 20

selected.
baccel m/s2

The simulated peak and RMS values for suspension 0

working space, body acceleration, and tyre load variation
are given in Figure 13, and time histories for each of
these outputs in Figure 14. Interestingly from Figure 13 it -20
can be seen that the non-linear damper predicts peak 0 0.05 0.1 0.15 0.2
values for both DTL and body acceleration of 20
approximately 10% and 3% smaller for tyre compression
Kt(x0-z1) kN

and extension respectively than that of the 14 speed

damper. This is for an increase in predicted RMS SWS 0
of the order of 2% and peak value of 1%. The output
characteristics displayed for the linear model damper
model, in terms of RMS and peak values, differ -20
significantly from those of the other damper 0 0.05 0.1 0.15 0.2
representations, and indicate higher levels of overall time s
damping. This finding correlates well with the results of
the investigation of the effects of differing road speeds
on discomfort parameter for a minor road, see Figure 11; Figure 14. Simulated time histories for 10m/s simulation over a pothole
at higher velocities a linear damper provides higher for three differing damper representations.
CONCLUSIONS 6. Fash, J. Modeling of Shock Absorber Behaviour
using Artificial Neural Networks. SAE paper 940248,
The parameters of a relatively complex non linear 1994.
damper model have been successfully extracted, in an 7. Besinger, F. H., Cebon, D. and Cole, D. J. Damper
automated fashion, from data obtained from two simple Models for Heavy Vehicle Ride Dynamics. Vehicle
experimental procedures. The model has been System Dynamics, 24, pp. 35-64, 1995.
successfully validated to a high degree of accuracy by 8. Willumeit, H.P. and Tong, Z. Investigation of the
comparing simulated results to experimental data for 3 high frequency disturbance response of vehicle
discrete frequencies of sinusoidal excitation of 1,3, and suspension. FISITA Congress, IMechE, London, pp.
12Hz. 319-327, 1992.
9. Lang, H.H. A Study of the Characteristics of
The differences between simulated vehicle ride output Automotive Hydraulic Dampers at High Stroking
characteristics for the non-linear damper model and Frequencies, PhD Dissertation, The University of
several simpler damper representations have been Michigan, 1977.
investigated for a simple 2 d.o.f quarter car ride model. 10. Segel, L. and Lang, H.H. The Mechanics of
This work has indicated that gains to vehicle modelling Automotive Hydraulic Dampers at High Stroking
accuracy may be achieved by utilising the derived Frequencies. Vehicle System Dynamics 10, pp. 82-
damper model in preference to a 14 speed 85, 1981.
representation and has also identified the inadequacy of 11. Duym, S. Simulation Tools, Modelling and
a single linear damper rate to accurately predict vehicle Identification, for an Automotive Shock Absorber in
outputs over a broad range of operating conditions. the Context of Vehicle Dynamics, MDI conference,
France, 1998.
Future work will investigate the effects of improved 12. Duym, S. and Reybrouck, K. Physical
friction representations within the damper model in the Characterisation of Nonlinear Shock Absorber
context of vehicle dynamics. The damper model will Dynamics. European Journal of Mechanical
also be incorporated into vehicle models of elevated Engineering, Vol. 43, No4, pp. 181-188, 1998.
complexity and a case study will be performed of the 13. Reybrouck, K. and Duym, S. A Physical and
adaptive damping strategy for the vehicle introduced Parametric Model for Nonlinear Dynamic and
within this paper. Further work is also needed to validate Temperature–Dependant Behaviour of Automotive
the findings of the vehicle ride modelling work carried out Shock Absorbers. Proceedings of the 11th ADAMS
within this paper. Users Conference, Frankfurt.
14. Lang, R, and Sonnenburg, R. A Detailled Shock
ACKNOWLEDGMENTS absorber Model for Full Vehicle Simulation.,
presented at the 10th European ADAMS Users’
This research is sponsored by RICARDO VEHICLE Conference, Frankfurt, Germany, November 14-15,
ENGINEERING as part of an ongoing project to deliver 1995.
an improved systematic methodology for the early 15. Herr, F., Mallin, T. and Roth, S. A Shock Absorber
development of damper properties to influence the Model Using CFD Analysis and Easy5. SAE
ultimate ride and handling properties of vehicles. Steering and Suspension Technology Symposium
1999, p.p. 267-281, 1999.
16. Duym, S. Stiens, R. and Reybrouck, K. Fast
REFERENCES Parametric and Nonparametric Identification of
Shock Absorbers. Proc. 21st INT. Seminar Modal
1. Richardson, C. Ride Manual. Monroe Europe. Analysis Noise and Vibration Engineering
2. Belingardi, G. and Campanile, P. Improvement of Conference. Leuven, pp. 1157-69, 18-20 Sep 1996.
the Shock Absorber Dynamic Simulation by the 17. Duym, S., Stiens, R. and Reybrouck, K. Evaluation
Restoring Force Mapping Method. Proceedings of of Shock Absorber Models. Vehicle system
the 15th International Modal Analysis Conference, dynamics, 27, pp. 109-127, 1997.
Leuven, Belgium, pp. 441-454, Sep 1990. 18. Reybrouck, K. A Non Linear Parametric Model of
3. Audenino, A. and Belingardi, G. An Application of an Automotive Shock Absorber, SAE International
the Restoring Force Mapping Method for the Congress, Detroit, Michigan, February 28- March 3,
Diagnostic of Vehicular Shock Absorbers Dynamic pp. 79-86, 1994.
Behaviour. Proc 2nd. Int. Machinery Monitoring and 19. Cardano, G. Ars Magna, Nuremberg, 1545.
diagnostics conference, LA, California, pp.560-566, 20. Crolla, D.A., Firth, G., Horton, D. An Introduction to
1990. Vehicle Dynamics. Internal University of Leeds
4. Duym, S., Schoukens, J. and Guillaume, P. A Local document pp. 12.
Restoring Force Surface Method. Proceedings of the 21. International Standard ISO 2631-1:1997, Mechanical
13th International Modal Analysis Conference , Vibration and Shock – Evaluation of Human
Nashville, Tennessee, pp. 1392-1399, 1995. Exposure to Whole Body Vibration – Part 1: General
5. Cafferty, S., Worden , K. and Tomlinson, G. R. Requirements.
Characterisation of Automotive Shock Absorbers
Using Random Excitation. Proc Instn Mech Engrs,
Vol 209, Part D, pp. 239-248, 1995.
CONTACT

Adrian Simms. [email protected] APPENDIX 2.

DEFINITIONS The Cardano Coefficients:

Compressible damper model: A damper model which æ 3B B - B 2 ö

models effects of damper internal compliances to yield D1 = çç 1 3 2 2 ÷÷ (44)
hysteretic dynamic behaviour. è 3B 3 ø
Incompressible damper model: A damper model 3 2
which assumes damper internal compliances are 2 B 2 - 9 B1 B 2 B 3 + 27( B 0 - Q tot ) B 3
negligible, and that control force is related solely to D2 = 3
(45)
damper velocity; hence non-hysteretic behaviour is
27 B 3
predicted.
2 3
APPENDIX 1. D D
D3 = 2 + 1 (46)
The Taylor Coefficients:
4 27

APPENDIX 3.
B 0 = Q leak , blow -off (40)
Linear extrapolation coefficients:
3
7 Dp 0 7/4
B1 = 1 + (41) H 0 = B0 n1 / 4 ( K leak + K port ) (47)
4 K spring Q leak ,blow - off

3
Dp tot open 7n1 / 4 B0
3/ 4

35 Dp 0 H1 = = ´
B2 = 2
(42) dQ tot 4
16K spring Q leak , blow - off blow - off

æ ææ D2 ö
-2 / 3
ö ö
ç çç- + D3 ÷ (2 - D 2 ) + ÷ ÷
217 Dp 0
3
ç K leak çè 2 ø ÷ ÷
B3 = (43) ç ç -2 / 3 ÷ + K port ÷
ç 12 B3 D 3
3
512 K spring Q leak ,blow - off çæ- D2 ö ÷ ÷
ç çç - D3 ÷ (2 + D 2 ) ÷ ÷
è èè 2 ø ø ø

(48)