Vehicle System Dynamics

ISSN: 0042-3114 (Print) 1744-5159 (Online) Journal homepage: http://www.tandfonline.com/loi/nvsd20

High Performance System Dynamics Simulation of

the Entire System Tire-Suspension-Steering-Vehicle

DIETER AMMON , MICHAEL GIPSER , JOCHEN RAUH & JÜRGEN WIMMER

To cite this article: DIETER AMMON , MICHAEL GIPSER , JOCHEN RAUH & JÜRGEN WIMMER
(1997) High Performance System Dynamics Simulation of the Entire System Tire-Suspension-
Steering-Vehicle, Vehicle System Dynamics, 27:5-6, 435-455, DOI: 10.1080/00423119708969341

To link to this article: http://dx.doi.org/10.1080/00423119708969341

Published online: 27 Jul 2007.

Submit your article to this journal

Article views: 172

View related articles

Citing articles: 3 View citing articles

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=nvsd20

Download by: [University of California Santa Barbara] Date: 23 November 2015, At: 19:21
 Vehicle System Dynamics, 27 (19971, pp. 435-455 0042-3 1 14/97/2705-435$12.00
O Swets & Zeitlinger

High Performance System Dynamics Simulation of

the Entire System Tire-Suspension-Steering-
Vehicle
Downloaded by [University of California Santa Barbara] at 19:21 23 November 2015

DIETER AMMON ' , MICHAEL GIPSER * ,JOCHEN RAUH * and

JURGEN WIMMER *

SUMMARY

A problem-specific multi-body-system approach. the dynamic suspension models, was developed to

obtain an almost complete description of the nonlinear dynamic properties of modem suspensions. The
concept includes the spatial elasticities and damping effects of the bushings. forces and moments of
inertia of the links, and related friction and clearance effects. In combination with the sophisticated
physical tire simulation model BRIT [4] and comparable models for the hydraulics of the power
steering system, we obtain an effective simulation tool for the analysis of the entire suspension system
dynamics as well as for the entire vehicle system dynamics. This approach enables steady-state
vibration phenomena as well as transient response problems of the suspension system such as shimmy,
steering roughness, or excitations due to misbalanced wheels to be investigated efficiently and with
respect to realistic excitation conditions. The results are compared with driving test data.

1 . MOTIVATION

The dynamic properties of modern car wheel suspension systems are determined
both by the kinematics of the wheel location as well as by the elastic linkages of
the suspension to the car body. This makes it possible to better adapt the steering
properties to different driving states [6].This vehicle dynamics benefit parallels the
problem of manifesting more severe non-linear elastokinematic interactions in the
wheel suspension, which, in combination with the vertical and wheel location
force dynamics of the tires and the dynamics of the power steering, produce
complex problems in harmonizing the entire system. To be in a position to fully
understand the dynamic interactions of the suspension components and to optimize

Daimler-Benz AG, Fahrzeugdynamik (FlM/SD), 70546 Stungart, Germany. Tel: Dr.-Ing. Dieter
Ammon, 07 1 1 /1721537, Dr.-Ing. Jochen Rauh, 071 1/1721538, Dipl.-Ing. Jurgen Wimmer,
071 1/1720871.
* *
Fachhochschule Esslingen, Hochschule f ~ Technik,
r Fachbereich Fahrzeugtechnik, 73728 Esslin-
gen, Germany. Tel: 071 1 /3973341.
 436 DIETER AMMON ET AL.

these, it is therefore necessary to develop, interlink and, on the basis of measure-

ments, verify appropriate detailed simulation models.

2. INTEGRATION OF SIMULATION METHODS AND MODELS

Because of the need to combine numerous modelling disciplines, simulation of the

system dynamics of a vehicle places high demands on simulation technology.
Although the market for simulation systems is in itself large, to the extent of being
unclear, with the most varied range of orientations, the situation remains the same
Downloaded by [University of California Santa Barbara] at 19:21 23 November 2015

as before in that no satisfactory solution exists for typical tasks from this field.
The situation is rather that the limits of commercially available simulation tools
are revealed when it comes to systems of the complexity of vehicles. It is not
possible and also not practical to simulate all the subsystems in a self-contained
simulation environment. The only solution is to make use of model export and
import subsystems from various sources and to integrate these to form an entire
system, comparable to the assembly of a vehicle from individual components.
The thinking in respect of this approach is presented more closely in the
sections which'follow and then elucidated in the remaining chapters by taking a
concrete example.

2.1 Modelling Concepts

When it comes to modelling it is possible to distinguish various description layers,
see Fig. 1:
The physical description is defined in system-oriented quantities, such as mass
and dimensions in the case of mechanical systems, resistance and capacitance
in the case of electrical systems, etc.

Preprocessor 1 10
0-
Physical model description

Model compiler
/In
16-
Mathematical description

Linker+lntegrator
Scheduler *Inn
Behavioral description

Postprocessor
0-
Fig. 1. Model layers.
 HIGH PERFORMANCE SYSTEM DYNAMICS 437

The mathematical description is the image of the model in an abstract

representation independent of the system, such as differential equations.
The behavioral description is the result of simulations or other analyses of the
mathematical model. A link to the physical model is also required in the
majority of cases to obtain a practical evaluation.
It is not intended to look any more closely in this paper at physical substitute
representations, e.g. the representation of mechanical systems by means of electri-
cal networks as they are losing significance with the availability of suitable model
compilers for the various technical disciplines.

2.2. Mathematical Description Forms

Downloaded by [University of California Santa Barbara] at 19:21 23 November 2015

Following modelling, the physical model representation obtained is transferred to

the mathematical represention. It is possible to distinguish various types in this
connection, such as
algebraic models which describe simple, non-dynamic interrelations by means
of characteristic curves, families of characteristics, approximation approaches
or functional relations,
ordinary diferential equations (ODE) in explicit or implicit representation
which can be used to describe the majority of components which occur in the
system dynamics of the vehicle,
d~ferential-algebraicequations ( D A E ) which may occur, for example, in
modejling mechanical systems with self-contained kinematic loops,
linear dlferential equations which can be obtained from a linearization of the
aforementioned differential equations or with the aid of direct linear modelling,
and finally
time- (and possibly also state) discrete differential equations which occur in
every software-in-the-loop or hardware-in-the-loop application as well as in
the model approaches described in detail in the sections which follow.

2.3 Model Import and Export

Export of the mathematical model representation is either not supported at all by
the majority of simulation tools or only to an inadequate extent as a direct link to
the solution method often exists within the system. On the other hand, the
available simulation tools usually do not permit any, or only a very inefficient
import of typical subsystems.
All in all, the prerequisites for the division of labor are rather poor when it
comes to be in a position to merge model components as a mathematical function
description of the vehicle components from different sources in the simulation to
achieve an efficient entire system. One example of nevertheless achieving this on
the basis of available simulation tools, has been studied more closely as part of the
AIT pilot phase [9].
The Working Group "Simulation of System Dynamics of the Vehicle in the
German Automobile Industry" has been working on the questions of a suitable
open simulation landscape for about 4 years now [7 and I I]. In addition to work
 43 8 DETER AMMON ET AL.

into standardization of the physical model description, one of the achievements of

this Working Group has been the interactions between simulation tools for
mechanical and hydraulic components.

2.4 Interactions of Subsystems with Local Integration

The first unwelcome side effect which already occurs in the interaction of ODES
and DAEs with a central integrator, is that the most demanding system compo-
nents influence the integration increment for the entire system. A typical example
of this are hydraulic subsystems which generally have a "stiff' behavior. Even if
this subsystem is interfaced only "softly" to the rest of the system, the entire
Downloaded by [University of California Santa Barbara] at 19:21 23 November 2015

system is integrated with a small increment. It would therefore be desirable to

achieve a separation of integration methods for both system parts.
A pragmatic approach to separation is local integration of the individual system
parts with the methods and increments suitable for each part. Interfacing to form
an entire system then corresponds to the linking of time-discrete system blocks. It
is essential, in this connection, to devote considerable care to the choice of the
communication increment. Basically, linking the time-discrete blocks to form
fixed communication increments introduces a "block" Euler method, which
influences the accuracy and stability of the entire system. In doing this, it is not
the dynamic properties within the subsystems which are considered, but primarily
the summary properties of the modules, as they are rendered visible by the
interface parameters.
As no automatic methods exist for this task, the choice of the communication
increment has to be made from the knowledge of the properties of the subsystems
and from the desired scope of the simulation of the entire system. Experience
gained with the CASCaDE simulation system for studying vehicle dynamics,
shows that it is possible to efficiently achieve good results if the interfaces
between the subsystems are properly selected, with a communication frequency
approximately ten times above the limit frequency in question for the entire
system [8]. The section which follows describes a detailed problem-specific
modelling of the entire tire-suspension-steering-vehicle system dynamics on the
basis of this tried-and-tested approach.

3. BRIT - A PHYSICAL TIRE MODEL

The acronym BRIT stands for Brush and Ring Tire Model and is the name given
to the dynamic non-linear tire model used within the suspension model presented
here. Modelling is conducted based as closely as possible to the physics of the tire.
The following requirements had to be met in this connection:
- highly accurate steady-state tire characteristics while taking into account all
the essential influencing parameters,
correct simulation of the steady-state response in the frequency range up to
about 100 Hz, also in all critical ranges such as: wheel lock-up, spinning, riding
 HIGH PERFORMANCE SYSTEM DYNAMICS 439

over obstacles, stopping and moving off on an inclined plane, brake applica-
tions with ABS control loops, etc.,
computing effort no more than the level of the effort for the remaining model,
available or easily identifiable model parameters.
From the vehicle dynamic simulation aspect, the tire model represents a general
force and moment element between the vehicle, or more precisely the wheel, and
the road surface. This linking element is strongly non-linear in almost all operating
conditions of relevance for vehicle dynamics and, as an elastic structure, has
complicated transportation and friction processes in the contact zone, characterized
by momentum with a very large number of internal state variables.
Downloaded by [University of California Santa Barbara] at 19:21 23 November 2015

The input parameters for BRIT are the total of 12 translatory and rotary
position and velocity quantities of the hub, as well as the height profile and
friction coefficient distribution of the road surface. BRIT supplies as the output
parameters the 6 components of the force/moment vector acting on the hub in
the center of the wheel.
It is also possible for BRIT to adopt the integration of the intrinsic wheel
rotation for numerical reasons, also because of the severe mass coupling between
wheel and tire. In this case, the driving and the (unsigned maximum) braking
moment are added to the input parameters. The output parameters are supple-
mented accordingly by wheel rotation angle and wheel velocity, these values being
required for the elastic linking of the wheel to the drive train. In addition, we
obtain from the wheel speed the gyrostatic moments acting on the vehicle (or more
precisely on the wheel hub). The actually effective braking moment when the
wheel is locked up is, according to sign and amount, just large enough so as to
compensate for the other moments acting on the wheel in the direction of the axis

Rotary hub position

Rigid body state variables, radial ply, cleat deflections, ...

Fig. 2. BRIT interface and state Tm-abTes.
 440 DIETER AMMON ET AL.

of rotation. The bralung moment, corrected in this way, is likewise added to the
output parameters.
In addition to the input and output parameters, BRIT manages mechanical and
thermodynamic state variables which describe the momentum of the tire, such as
position and velocity of the steel radial ply as a 'rigid body', deflections of tread
cleats as well as the internal pressure and temperatures.
In BRIT, the steel radial ply is modeled as an elastic rigid body connected to
the wheel. This approach is supplemented by additional quasi-static elastic disloca-
tions, twisting and deformations of the contact area relative to the position of the
radial ply as a function of the active contact forces and moments, cf. Fig. 3. The
parameters required for this description can be easily derived from measurements
Downloaded by [University of California Santa Barbara] at 19:21 23 November 2015

of 'integral' tire springing stiffnesses, as well as from intrinsic frequencies and

damping of the freely oscillating tire. Another possibility is to use static or
dynamic condensation of FE models to conduct an approximate calculation.
The next step, after calculating the position of the radial ply in the current time
increment of the simulation, is to calculate the contact area, as well as the input for
the contact pressure distribution in this area. Both are a function of the position of
the radial ply. The penetration of the radial ply (modeled geometrically as a radial
shell) with the tangential plane of the road surface is specified as the boundary of
the contact area in this case.
The contact pressure is calculated only in discrete zero-mass cleats which are
distributed equidistantly over a fixed number of strips, in a similar way to a
'brush', over the tread strip. The input of the contact pressure distribution in BRIT
is done by using simply parameterized statement functions.
In this aspect, BRIT differs significantly from elastomechanical FE models
such as DNS Tire [2,3], where the contact pressure distribution itself is part of the
solution and no 'statement' is required. The lesser depth of this mechanical
modelling is reflected in an enormous reduction in computer time as the numeri-
cally extremely rigid equations of motion of the steel radial ply-shell structure are
eliminated. Even if the numerical integration of these equations of motion are
optimized in DNS Tire, the computer time in BRIT is reduced by the order of 3.5
to 4.
The rolling resistance is allowed for in the statement functions for the contact
pressure distribution as a peaking toward the lead-in side.
A prerequisite which still has to be respected at present when using BRIT is
that there are no extremely high frequency portions in the road height profile.
A simple statement for the relation between such road unevennesses and the
resulting contact pressure does not exist. Exact calculation of this relation would
require the very involved solution of a mathematically complicated variation
inequation, allowing for the radial ply bending stiffnesses for each roll of cleats
(cf. Fig. 4).
If the contact pressure distribution is known, it is possible to integrate for each
cleat the longitudinal and lateral deflection (from the equilibrium of forces
resulting from the shear deformations, the sliding friction on the road surface and
Downloaded by [University of California Santa Barbara] at 19:21 23 November 2015

HIGH PERFORMANCE SYSTEM DYNAMICS

44 1
 442 D E E R AMWON ET AL.
Downloaded by [University of California Santa Barbara] at 19:21 23 November 2015

Fp 4 R o d uncvcnneqs and conrxt pmscurc dinnbutlon: Conraa plane spl~rtingdue lo h~phfrquent

rond profiles.

the inner damping). The corresponding differential equation, in a simplified form

for pure longitudinal or transverse dell ection. reads

In [his a designales the cleat deflection, v the velocity of the cleat base fixed in the
radial ply. c the cleat shear stiffness. d the cleaa shear damping. p the contact
pressure and p, the friction coeii'icien~,dependent on the slidinp velocity (and
additionally on the contact pressure).
By using material and non-stationary coordinates in the latch. i t is possible to
define a hinary adhesion/sliding state for each cleat. This indicates the 'bough'
of the branching on which the clcat is located. This hranchinp, in other words the
ambiguity in the relation between friction coefficient and sliding velocity. occurs
because of the non-monotonic curve of the friction coefficient and the resultant
ambiguous solution of the differential equation.
The adhesion/slidinp slate is also managed in the case of the implicit numeri-
cal integration of the differential equations for the cleat deflections. As the state in
question here is a binary, in other words switching state. it is not possible to use
conventional integra~orswhich recognize on1y continunus states. The adapted
numerical integration of the rigid body degrees of frecdom and the cleat deflec-
tions is, for this reason, an integral element of BRIT.
As a distinction is made individually on the hasis of adhesion and slip Tor each
cleat. BRIT is also able to correctly simulatc the passage of several adhesion/slip
limits through the contact area. Such phenomena occur in the case of very low
rnllinp velocities or in the case of very rapid changes in slip. as are present. in
particular. during ABS brake applications [I].
It is possible. hy numerical squaring. to calculate from the cleat shearing forces
resulting from the deflections, lhe contact forces and moments which act on the
 HIGH PERFORMANCE SYSTEM DYNAMICS
Downloaded by [University of California Santa Barbara] at 19:21 23 November 2015

IQg(v6dc
Fig. 5. Frication coefficient related to the sliding velocity.

radial ply as impressed forces. Filtered by the mass moments of inertia of the
radial ply, they continue to act on the wheel as a result of the elastic mounting of
the radial ply.
Precautions have been taken in numerical integration to avoid adhesion/slip
transitions of individual cleats causing discontinuities in the resulting contact
forces. Moreover, provision is made for correction terms for the numerical error
which arises from the fact that start and end of the contact area do not coincide
with an interpolation point of the square.
BRIT is able on request, as mentioned, to also adopt the (implicit) integration
of the intrinsic rotation of the wheel. In this case, the numerically stiff dynamic
interaction of the intrinsic rotation of the wheel is regarded with the rotational
movement of the radial ply relative to the rim.
All the BRIT parameters are of a 'physical' nature, in other words can
theoretically be directly measured. Nevertheless, a number of these parameters can
only be determined approximately at justifiable cost. That is why a program for
parameter identification has been developed parallel to BRIT. This program
contains inter alia a stationary model BRST (BRIT Slarionary) which is data-
compatible with BRIT and which in the steady-state case supplies exactly the
same circumferential and cornering force characteristics as BRIT.

4. MULTI-BODYSYSTEM MODEL OF WHEEL SUSPENSION AND OF

MECHANICAL PART OF STEERJNG

Wheel location and suspension as well as the mechanical part of the steering are
described with the in-house developed multi-body system software DA ( ~ y n a m i c
Suspension ~ o d e l r )DA,
. like the sub-models for the tire and the hydraulics of the
powersteering, is also integrated in the entire vehicle model CASCaDE [8].
 444 DIETER AMMON ET AL.

Fig. 6. DA suspension model: animation single frame of running over an obstacle.

Downloaded by [University of California Santa Barbara] at 19:21 23 November 2015

DA is not a multipurpose MBS system but has been specifically and solely
developed for efficient simulation of the dynamics of wheel suspension systems.
This is clearly portrayed in the relatively small elements catalog in which
complex, assembly-oriented 'macro' elements predominate (cf. Table 1).
Like all other sub-models in CASCaDE, DA also integrates the state variables
locally with simple partially-implicit integration methods. The numerical accuracy
which this produces is not as great as in the case of complicated increment- and
order-controlling multi-purpose integrators, but is always still significantly better
than the lack of definition resulting from not exactly known parameter values and
simplifications or negligences in modelling. The advantage of local integration is
the high flexibility when it comes to enlargements and the specific localization and
special treatment of rigid system parts, for example the treatment of dry friction in
the disc brake and ball bearing model and the allowance made for 'almost'
kinematic linkages.
The section which follows elucidates more closely the calculation of zero-mass
links in DA as representative for a large number of other numerical aspects (all
the links can, of course, also be modeled with mass as the elements catalog offers
all the elements required for this).
The lower wishbone arm of a front suspension with subframe is connected to
the hub by a ball joint and to the subframe by two bushings (the comfort and the
location bushing) which have different degrees of stiffness and are not rotation-
ally-symmetrical. The coil spring and an elastically mounted damper are posi-
tioned between wishbone and body. All of these elements give rise to forces and
moments acting on the participating rigid bodies, which depend on the translatory
and rotary position and velocity of all of these rigid bodies and on other state
variables, such as the axial deflection of the damper mounting.
The position and velocity variables x, v are, in the case of the positive mass
subframe and the body, state variables and are thus known and fixed at the
commencement of the next time step. The position and velocity of the link, in
other words a total of 12 variables, now have to be calculated so that (a) the total
of all the forces, and (b) the total of all the moments acting on the link, vanish.
 HIGH PERFORMANCE SYSTEM DYNAMICS 445

Table 1. Mechanical elements of the dynamic suspension model toolbox.

Element interface parameters Intrinsic dynamics
on off (state variables)

Rigid body 6 12 12
Forces and moments Position, velocity Position, velocity
Bushing 24 12
with stops and damping, Position, velocity of Forces, moments on
off-diagonal stiffnesses, both connected rigid the two connected
opt. general characteris- bodies rigid bodies
tic (splines)
Beam 24 12
Position, velocity of Forces, momemts on
Downloaded by [University of California Santa Barbara] at 19:21 23 November 2015

both connected rigid the two connected

bodies rigid bodies
Ball joint 24 12
rigidly-elastic with dry Position, velocity of Forces, moments on
friction both connected rigid the two connected
bodies rigid bodies
Spring 12 12
with preload, length un- Position of the two Forces, moments on
tensioned and character- connected rigid bod- the two connected
istic ies rigid bodies
Damper 24 12
with characteristic, dry Position, velocity of Forces, moments on
friction, gas force, and both connected rigid the two connected
elastic stops for bounce bodies rigid bodies
and rebound with char-
acteristic
Elastically mounted 24 12 1
damper Position, velocity of Forces, moments on Axial deflection of
with characteristic, fric- both connected rigid the two connected damper bushing
tion, gas force and stops bodies rigid bodies
in bushing
Damper strut 24 12 1
optionally including Position and velocity Forces and moments Axial deflection of
body spring, with bush- of wheel carrier and on wheel carrier and damper bushing
ing (above) attached to body body
body
Zero-mass link 24 12
elastic or rigidly elastic, Position and velocity Forces and moments
with rotationally sym- of the two pivoted on the two pivoted
metrical bushings with rigid bodies rigid bodies
stops and damper
Zero-mass link a. 36 18 1
zero-mass wishbone Position and velocity Forces and moments Axial deflection of
arm with non-rotation- of wheel carrier, sub- on wheel carrier, damper bushing
ally symmetrical bush- frame and body subframe and body 12 pseudo states
ings with stops and Position and velocity of
damper (above) option- link as initial value for
ally including body the Newton-Raphson it-
spring and elastically eration for calculating
mounted dam p e r state of equilibrium
(above)
(Table continues)
 446 DIETER AMMON ET AL.

Table 1. (continued).
Element Interface parameters Intrinsic dynamics
on off (state variables)

Torsion anti-roll bar 36 18

Position and velocity Forces and moments
of body and of both on body and on both
wheel carriers wheel carriers
Rack-and-pinion steer- 39 19 5
ing with zero-mass Position and velocity Forces and moments Steering wheel angle
steering gear of body and of both on body and on both and steering wheel an-
including track rods, wheel carriers, im- wheel caniers, result- gle velocity (in the case
elastically mounted pressed steering ing steering wheel of steering moment in-
Downloaded by [University of California Santa Barbara] at 19:21 23 November 2015

zero-mass steering hous- wheel angle or im- moment or resulting put), relative opening of
ing with non-linear pressed steering steering wheel angle Cway valve, pressure in
b u s h i n g , s t e e r i n g wheel moment, en- rod part, pressure in pis-
d a m p e d s ) , variable gine speed and vehi- ton part
steering ratio with stop cle speed (with 14 pseudo states
and play, torsionally speed-sensitive servo Dislocation of rack rela-
flexible torsion bar with assistance) tive to steering housing
stop, torsionally flexible and position and veloc-
steering tube with fric- ity of steering housing
tion and universal joint, as initial value for the
steering wheel with mo- Newton-Raphson itera-
ment of inertia and in- tion for calculating state
ternal retrieval of hy- of equilibrium, supply
draulic interface pressure of power steer-
ing
Rack-and-pinion steer- 39 19 20
ing with positive mass as above as above as above, additionally:
steering gear position and velocity of
additionally positive steering housing, dislo-
mass steering housing cation and rate of dislo-
and positive mass rack cation of rack relative to
steering houding
1 pseudo state
Supply pressure of
power steering
Wheel bearing 24 12
with friction, axial play Position and velocity Forces and moments
and cardanic stiffnesses of wheel canier and on wheel carrier and
wheel wheel
Disc brake 25 13 I
is able to block, temper- Position and velocity Forces and moments Temperature
ature model prepared, of wheel carrier and on wheel carrier and
uneven disc surface wheel, brake pres- wheel, ABS wheel
sure sensor signal
Tire interface 12 or 14 6 or 9 +
22 3 No. of cleats
BRIT retrieval Position and velocity Forces and moments BRIT state variables
Interface to road is re- of wheel canier. on wheel carrier.
trieved internally by With integration of With integration of
BRIT wheel intrinsic rota- wheel intrinsic rota-
tion in BRIT: driving tion in BRIT: wheel
moment, unsigned rotation angle, wheel
maximum braking speed, effective
moment signed braking mo-
ment
(Table continues)
 HIGH PERFORMANCE SYSTEM DYNAMICS 447

Table 1 . (continued).
Element lnterface parameters Intrinsic dynamics
(state variables)
on off
Hydraulic interface 5 2 3
Open-loop and closed- Torsion bar torsion, Scalar force on rack Relative opening of 4-
loop control algorithms dislocation and rate in direction of dislo- way valve, pressure in
for variable power assis- of dislocation of rack cation, moment on rod part, pressure in pis-
tance are retrieved inter- relative to steering torsion bar ton part
nally by the hydraulic housing, engine 1 pseudo state
model speed, vehicle speed Supply pressure of
power steering
Downloaded by [University of California Santa Barbara] at 19:21 23 November 2015

What applies to the position and velocity of the wishbone is a kinematic

relation of the type

or, after fully implicit discretization,

The velocity parameters can be eliminated in the force/moment balance with the
aid of this equation so that we obtain a non-linear algebraic equation system with
now only 6 equations (force and moment total = 0) and 6 unknown quantities
(translatory position and angular position of the wishbone).
This system is resolved in DA for each zero-mass link separately with the
Newton-Raphson method. The method converges very rapidly; provided the
starting approximations are sufficiently good, only one to three steps are required
as a rule in order to get below an accuracy tolerance of cab, = lo-'.
The Newton-Raphson method requires the partial derivatives of all the equa-
tions on the basis of all the unknown quantities. These parameters are closely
related to the linearized stiffness, damping and mass matrices and are re-procured
in each time step by numerical differentiation.
The solutions from the last time step are taken as the starting values for the
Newton-Raphson method and, for this reason, are designated in the elements
catalog as 'pseudo state variabIes'. In this case, the position variables are first of
all extrapolated to the new time step with the aid of the velocity variables.
Despite the apparently involved calculation required, this method is usually
faster than the positive mass modelling of the links. Because of the relatively small
mass and the extreme stiffness of the ball joint, the equations of motion are
numerically very stiff in a number of components and would thus, with explicit
integration, demand very small increments. The method described moreover
results as a boundary case of implicit BDF integration, with vanishing masses and
moments of inertia.
 448 DETER AMMON ET AL.

The principal aspects of the further development of DA are:

the completion of the elements catalog (further steering configurations,
multi-link systems, hydraulic bushings, .. .),
allowance for link elasticities, described in the zero-mass case by the non-lin-
ear two-dimensional series connection of three stiffnesses: bushing 1, link
stiffness matrix (from the preceding FE calculation and static condensation),
bushing 2,
the physically more accurate description of the damper property of elastomers
by modelling the hysteresis in the time range (which is characterized by a
state-discrete momentum),
Downloaded by [University of California Santa Barbara] at 19:21 23 November 2015

implementation of interfaces to detailed models for semi-active and active

suspension systems, and
the implementation of an interface to the structure-mechanical tire model
DNS-Tire([2,3]).

5. DYNAMIC DESCRIPTION OF THE POWER STEERING

Despite the manifold design differences, it is possible to explain the majority of

hydraulic power steering systems by the function diagram presented in Fig. 7. In

. --. - - --.--

Fig. 7. Kinematics and kinetics of the hydraulic and mechanical components of the steering system.
 HIGH PERFORMANCE SYSTEM DYNAMICS 449

this case, a force measurement signal A6 dependent on the static steering wheel
moment, is recorded in order to actuate a bridge circuit of hydraulic servo valves,
the flow equilibrium of which controls the pressures p, and p, in the servo
actuator so that the physical forces do not exceed certain limits. In the case of
unitized steering systems, the actuator is generally integrated in the steering gear;
the force is transmitted to the track rods through further reduction ratios but is
nevertheless comparable - taking into account the complicated kinematics.
The static moment gain characteristic curves are generally severely non-linear.
They are configured by the type of "steering moment measurement", in other
words the elasticities c, and frictions M R and also, as the case may be, by
Downloaded by [University of California Santa Barbara] at 19:21 23 November 2015

additionally impressed servo pressure feedbacks p,(pL 1, p,(p,) together with the
control valve properties so that the driver is provided with adequate feedback from
the steering wheel during small steering movements, but is otherwise relieved of
major physical efforts, for example when moving into a parking place. This driver
adaptation of the steering, characterized primarily by static or steady-state consid-
erations, has to be examined in respect of its consequences for the dynamics of the
tire-suspension-steering system, in particular in respect of its disturbance transmis-
sion properties.
In a similar way to the component-oriented steering analysis (cf. [5]), it is also
necessary in the case of the entire system analysis, to model the principal
hydraulic and mechanical components. What this means for the mechanical part of
the model is correctly describing the moments of inertia, stiffnesses, damping and
frictions of the actuator as well as its linkages to the vehicle structure and to the
steering column, with due allowance for the partially dominant non-linearities. As'
the hydraulics are an explicit part of the steering elasticity, it is necessary to
simulate both the compressib~lityP of the hydraulic fluid as well as the compli-
ance of the pressure hoses:

The pressures p, and p, in the servo actuator are thus determined by the volume
flows sL, sR of the control valves and of the actuator position Ay and its velocity.
From this results the hydraulic part of the actuator force F,,,,, = A(pL - p,),
whereby A describes the effective area. B,,, and V ,, represent the change in
compressibility and the volumes as a consequence of the feed pipes.
This leads to a dynamic model of the flows through the control valves. As the
latter display properties which are both laminar-like (in the base state) as well as
properties typical of orifice flows (in the case of higher stresses) because of the
design boundary conditions, the description generally becomes very complicated
as dependencies on the existing Reynold's numbers and on the viscosity of the
 450 DIETER AMMON ET AL.

medium occur. Such problems can be avoided if use is made of a summary

approach for the flow resistances of the control valves:

In this case, the resulting volume flow s with small pressure differences is
primarily determined by the laminar resistance c,,,. [bar/(l/min)], in the case of
greater pressure differences or high flow velocities it is the effective orifice
Downloaded by [University of California Santa Barbara] at 19:21 23 November 2015

resistance cD/B2 [bar/(l/min)*] which dominates the situation. Both parameters

can be derived from the characteristic data of the control valves, whereby due
allowance requires to be given to the temperature dependence of the viscosity for
determining c,,,: The normalized valve actuation a may adopt values between
[ - 1,1]. a = 0 describes the force-free center position; in this case the orifice
resistances adopt their nominal value c D (B = 1).
The pressure supply in regular operation supplies a constant volume flow s,
whereby a certain maximum servo pressure p y x cannot be exceeded. This simple
model has to be enlarged by more accurate statements for pressure generation,
volume control and pressure limiting for failure studies and for investigations in
the idle speed range, in other words it is necessary to integrate the steady-state
pump characteristics and suitably operated and parameterized control valves. The
same also applies for any hydraulic circuits which are provided for the servo
pressure feedback, in respect of the steering moment measurement, which, in the
case of a so-called speed-sensitive power steering, may additionally depend on the
vehicle speed and further control variables. The resulting forces are finally added
to the mechanical balance equations for the steering.
The parameterization of the model can be conducted largely on the basis of and
static or steady-state measurements of the input-output behavior of the steering as
well as steady-state valve characteristics. A particularly important aspect in this
connection is accurately determining the mechanical friction in the actuator and in
the hydraulic feedback; this may require to be formulated as a function of the
pressure because of the sealing influences. For reasons of efficiency, what is
recommended for the numerical integration of the hydraulic components is a local,
quasi-implicit statement of the type:

The time increment is designated with h; the gradients in respect of s are obtained
analytically from the volume models, while the changes in volume flows are
determined either numerically or analytically with the aid of the participating
 HIGH PERFORMANCE SYSTEM DYNAMICS 45 1

valve components. All in all, what is obtained in this way is a system description
for the steering dynamics which is balanced in terms of model depth and scope
and with which it is also possible to integrally investigate higher frequency
suspension vibration phenomena such as brake judder, shimmy, steering imbalance
or excitations resulting from misbalanced wheels.

6. ENTIRE SYSTEM STRUCTURE

Downloaded by [University of California Santa Barbara] at 19:21 23 November 2015

Fig. 8 clearly reveals that it is in fact necessary to model the entire system and that
it is not sufficient to regard each individual component for itself and separately
from the others. This image which shows the occupancy structure of the system
matrix following linearization in a specific operating condition, reveals the pro-
nounced mutual influencing of the sub-components of the 'front suspension'
system. Both the severely non-linear elastokinematics of the suspension as well as
the dynamic force generation of tire and steering exert a considerable influence on
the stability and ruggedness of the entire system. Considering the off-diagonal
elements, the coupling of the mechanical parts of the steering system, the
hydraulic components, and the wheel hubs is evident.
In general, analysing the frequency response helps to interpret the system
behavior due to affecting disturbances. As an example, Fig. 9 shows the transfer
characteristics from the road height at the left wheel to the steering wheel torque,
which results for a fixed steering wheel. In the diagram, responses obtained by
driving straight ahead are shown for two different tire types.
The deviation can be explained by the fact, that with a tire of larger dimension,
the self aligning torque increases, if the same slip angle and wheel load are taken
as a basis. Consequently, changes in the wheel load have a more distinctive effect
on the steering wheel torque. In both cases, the largest amplification can be
observed at a frequency around 17 Hz, which corresponds to the natural frequency
of the wheel as well as the steering system. However, the effect discussed may not
be solely led back to the different tire dimensions. Further factors of influence are,
for example, the tire stiffness, the mass of the tire, the contact patch shape, and the
pressure distribution at the contact patch.
The effect of the road excitations becomes even clearer while driving in a
curve. In this case, the wheel load at the outer wheel is increased, which in turn
increases the system sensitivity. In Fig. 10, the frequency responses are drawn for
a tire of dimension 195/65 R15.
Although the frequency responses allow to draw conclusions w.r.t. the system
behavior, one has to keep in mind, that these are only valid at specific operating
conditions. Furthermore, as a supposition for linearization, a11 sources for disconti-
nuities in the right-hand sides of the equations of motion have to be eliminated;
Such phenomena, as friction or play, belong to this category. As a consequence,
the required simplifications lead to an incomplete representation of the system
Downloaded by [University of California Santa Barbara] at 19:21 23 November 2015

DIETER AMMON ET AL.

 HIGH PERFORMANCE SYSTEM DYNAMICS

Gain [Nmlmm]

... ... ... ... ... ... ... ... . . . ....

. . .. . .. . . . ... . .. . . . .
m . . . . . .
a . . . . . .

6
.. .. .. .. .. .. .. .. 8 a 8 6 ,

. . a
. 6
.
. . . . a
.
*
. a
. .
,

.. .. .. .. .. .. .. .. ... ... ... ... ... ... ...

... ... ... ... ... ... ... ... .
.. .
.. m
.. a
.. .. .. ... . .

... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
. . . . . . . .
... ..... ..... ..... ..... ..... .....
. . . . a s . . . . a * . . .

.....,.......(.. .,.......,. .
... ... .. .. .. .. ... ... .. ... . . . .
.,. . ., ., ..
a , .
.. . . .
8 . a .

a
. . , . . * a

. . . . . a ,
Downloaded by [University of California Santa Barbara] at 19:21 23 November 2015

Frequency [Hz]

Fig. 9. Frequency response road height fl -,steering wheel torque for tires of dimension
( X ) and 225/60 R 15 (0).respectively.

characteristics. Hence, for a detailed analysis, a complete nonlinear description, as

is done in the DA suspension model, is indispensable.
An important development objective of the simulation package presented was
the shortest possible computing times. The entire system (4 X BRIT, DA suspen-
sion models for front and rear suspension, steering hydraulics, entire vehicle) can
at present be stably and adequately accurately integrated with increments varying
from 0.2 to 0.5 ms. It is only necessary to analyse the right-hand sides once for

Gain [Nmlmm]

........
.. .. .. .. .. .. .. ..
... ... ... ... ... ... ... ... .. .. . . . . . . . .. .. ..
. . . . . . . .
.. .. .. .. .. .. .. ..
.. ... ... ... ... ... ... ...
..... .. .. .. .. .. ..
... ... ... ... ... ... ... ...
. . ......

Frequency [Hz]

Fig. 10. Frequency response road height fl + steering wheel torque while driving straight ahead ( X )
and in a curve (0).
 454 DIETER AMMON ET AL.

each integration step. The computing time on typical workstations is about 15

times greater than real-time. As a result of being integrated in CASCaDE, DA is
downward-compatible to simpler static suspension models with which computing
times far below realtime are achieved.

7. VERIFICATION ON THE BASIS OF DRIVING TESTS

Driving test results, which were kindly made available to us by MBAG, were used
for conducting a qualitative and quantitative verification of the simulations. The
results in question were obtained from a driving maneuver which highly dynami-
Downloaded by [University of California Santa Barbara] at 19:21 23 November 2015

cally excites the complete transmission line between road surface and steering so
as to reveal the dynamic interactions of the components involved.
The test maneuver consists of driving over a series of ledges at a constant speed
(about 100 km/h) and with a high lateral force stress (a, = 7.5 m/s2). The ledges
are short, about 1 cm high and distributed on the basis of a random pattern over a
distance of about 25 m. They produce short force shocks at the tires which are
converted in the wheel suspension system primarily to vertical and longitudinal
oscillations. Because of the high lateral acceleration all of the wheels are mas-
sively stressed; the suspension bushings are located in different segments of the
respective dynamic characteristic curves. Nevertheless, it is possible to also
simulate the high frequency wheel oscillations, as Fig. 11 shows. The low
damping is attributable to a structural oscillation of the tire which is severely
excited by the shape of the ledge.
The rotary accelerations of the steering wheel calculated at the end of the
transmission chain, depend very greatly on the frictions and the internal dynamics
of the servo assistance. In addition, the test signals are always superposed by
structural oscillations of the steering column as a result of which additional high

Time s
Fig. 1 I . Venical hub acceleration due to curving on a ledge, measurement (broken line) and simulation
(full line).
Downloaded by [University of California Santa Barbara] at 19:21 23 November 2015