You are on page 1of 10

783

Identification of steering system parameters by


experimental measurements processing
S Data2, M Pesce2 and L Reccia1*
1Dipartimento di Ingegneria Meccanica, Tor Vergata University, Rome, Italy
2CRF (Fiat Research Centre), Orbassano, Italy

Abstract: The object of this investigation is to propose a procedure for the parameter identification
of a steering system, processing experimental measurements obtained on a test bench by means of
a developed software. This allows the isolation of a set of parameters and the performance ratings
of a steering system with an objective approach. The adopted approach overcomes problems of
repeatability of the test, influence of driver style, and feelings associated with the subjective drivers
evaluation method.
In this paper a two degrees of freedom (dof ) model dependent by synthesis parameters is used.
The model has been developed in Simulink environment, both for hydraulic and electrical power
steering systems. The experimental data, obtained by an ad hoc collected set of measurements, are
used to identify the parameters in two steps. A set of parameters is first computed by means of a
simple signal processing procedure. Then, the model is controlled by an optimization tool, which
varies the parameters values in order to reach the best fitting between experimental data and model
output. Identification leads to a complete characterization of the steering system. In particular, the
following parameters are identified: steering ratio, torsional stiness, power-steering characteristic
curve, friction forces, and damping.
Keywords: identification, steering system, hydraulic power steering, electrical power steering (EPS)

NOTATION
C
a
C
b
C
SW
F
a
F
r
F
s
I
SW
K
t
M
r
x
d
SW
t
q

friction torque on the column


torsion bar/torque sensor torque
steering wheel torque (feedback torque)
friction force on the rack
wheel force on the rack
f (C )=power steering force
b
moment of inertia of the steering wheel about
rotation axis
torsion bar/torque sensor torsional stiness
rack and front wheels reduced inertia
pinion radius
d r=rack displacement
SW
steering wheel rotation angle
d
/d =steering ratio
Wheel SW
C /K =torsion bar/torque sensor torsion angle
b t

The MS was received on 9 April 2003 and was accepted after revision
for publication on 7 May 2004.
* Corresponding author: Dipartimento di Ingegneria Meccanica, Tor
Vergata University, Via del Politecnico 1, 00133 Rome, Italy. email:
reccia@ing.uniroma2.it
D05403 IMechE 2004

1 INTRODUCTION
The steering system is the main way for a driver to feel
the handling behaviour of a car. During the cornering
manoeuvre the driver applies a torque to the steering
wheel and simultaneously feels on his or her hands the
tyre forces. Perceived feedback gives the confidence
needed to drive in safety. However, even in less favourable conditions, such as parking, modern vehicles require
very little eort to turn the steering wheel. The current
state of the art records three solutions adopted by automotive car makers for vehicle steering: (a) steer-bywire; (b) hydraulic power steer; and (c) electric power
steer. The last two have the same mechanical lay-out
(depicted in Fig. 1) but dier by the elements of the
power assistance unit.

1.1 Steer-by-wire system


The most advanced (and still not industrialized) system
is the steer-by-wire. No mechanical links are present.
The required steering-wheel position is translated into
an electric signal and an electronic control module
Proc. Instn Mech. Engrs Vol. 218 Part D: J. Automobile Engineering

784

S DATA, M PESCE AND L RECCIA

1.3 Electric power system (EPS)


The working principle is similar to the hydraulic system.
A torque sensor has the same role as the torsion bar,
but provides an electrical signal instead of a mechanical
eect. Collected signal is managed by a control unit,
which provides the control current to the electrical
motor. The motor can be placed on the steering column,
on the rack, or joined to the pinion. EPS can also include
an active damping system (managed by the control unit),
which works supplying a torque to the steering column
in order to damp the oscillations proportional to the
rotation speed. The eects of this system can be felt by
the driver during high vehicle-speed manoeuvres.
Fig. 1 HPS/EPS steering system lay-out

1.4 Comparison between EPS and HPS


manages the system and steers acting on the front wheels.
Even though experimental prototypes have already been
evaluated and a lot of research has been done (see e.g.
[1, 2]), the path to obtain a satisfactory steer-by-wire
still seems long. The main problems are the road forcefeedback reconstruction and the legal limitation to a
by-wire system. In fact, current homologation rules
require a mechanical link to connect the steering wheel
to the wheels.
1.2 Hydraulic power steer (HPS)
HPS is widely tested and has been used for a very
long time. Assistance power is supplied by means of
an hydraulic circuit. Oil is maintained in pressure by a
pump and the force on the rack is modulated by a spring
element, the torsion bar, moving inside a distributor [3].
Supplied force is proportional to the torsional displacement, a consequence of the externally applied torque.
Figure 2 shows a typical plot of torsion bar torque versus
rack force characteristic curve.

Fig. 2 Steering-wheel torque versus rack-force characteristic


(qualitative)
Proc. Instn Mech. Engrs Vol. 218 Part D: J. Automobile Engineering

EPS has several advantages [4, 5]. The EPS system


requires power only when it is really needed, thus reducing
fuel consumption. In a hydraulic system a pump always
works to maintain the pressure in the circuit. Moreover,
EPS requires a wiring harness instead of tubes. This
allows easy management of the lay-out of the mechanical
parts under the bonnet. Furthermore, there are no
problems of maintenance and of disposal of exhausted
oil. Finally, the electronic control permits the system to
work dierently depending on the situation [6 ]. For
example several cars equipped with EPS can switch
between normal and city control programs, in order
to supply dierent levels of assistance during normal
drive and parking (or low-speed driving) respectively.
Nevertheless, HPS, especially in high-velocity manoeuvres, has better performance than EPS. Furthermore,
the reliability of EPS is not at the same level as HPS.
For these reasons HPS is still an object of research and
development [7], even though car manufacturers have
already started to install EPS in several cars.
In any case, the quality of the steering system, as well
as the whole vehicle handling performance, has to be
correctly evaluated to support the designer in improving
car performance. In this matter, two dierent approaches
are generally employed: subjective evaluations, based
on driver ratings, and objective measurements [8, 9].
Pursuing the objective approach, the identification of the
main characteristics, in terms of frictions, power steering
supplied, stiness, and dynamic damping, leads to a
representative evaluation of performance, quality, and
defects of a steering system.
In this work a model, based on a set of experimental
bench tests and a numerical identification procedure, has
been developed to synthesize the characteristics of the
steering system for both EPS and HPS.
The parameter identification procedure is suitable for
dierent purposes, such as:
1. Tuning of the steering control systems during the
design phase (especially for EPS).
D05403 IMechE 2004

IDENTIFICATION OF STEERING SYSTEM PARAMETERS

2. Reverse engineering. The benchmarking of dierent


steering system models provides the designer with new
information.
3. Development of diagnostic applications.
4. Improvement of handling models knowledge.

2 MODELS DESCRIPTION
In this work two similar software models have been
developed respectively for the HPS and EPS systems.
Both models depend on a limited set of lumped parameters. Several reasons justify the level of detail employed.
As observed in [10], the guideline to choose the order
of the model is to tune the number of masses and stiffness depending on the frequency domain of interest. The
frequencies of the cyclic bench manoeuvres used in this
work involve parameters that influence almost-static
response (0.01 Hz test) and dynamic response (1 Hz test)
separately. Furthermore, as will be explained later, the
optimization tool is based on the search for the minimum error in terms of dierence between measured data
and model output. The MatLab code devoted to this
task works in a satisfactory manner with few uncoupled
parameters. In the presence of a large set of unknowns
the use of a dierent identification technique is needed.
The steering system can be divided into two parts: the
first one includes the steering wheel and the steering
column; the second one the rack-pinion steering gear,
the lever arms, the tie rods, and the wheels. The two
parts are joined by the torsion bar (or by the torque
sensor, for an EPS). Therefore, the system is composed
of rigid bodies, connected by an elastic element, and
the theoretical model has two dof. The stiness characteristic of the elastic element is represented by a bi-linear
curve, to take into account the limit stop. The eect of
the cardan joints and the dierence between left- and
right-wheel steer angle, under the hypothesis of small
angles, are neglected. Thus, the steer ratio is considered
constant. In this work a vehicle equipped with a powersteer motor on the column was considered. Since the
torque sensor is placed between the steering wheel and
the motor, equations are formally identical for EPS and
HPS.
The equilibrium equations take the form
I d =C C C
SW SW
SW
b
a
C
Mx = b +F F F
s
a
r
r

(1)
(2)

Friction appears in both expressions and can be


distinguished in three types as follows:
1. Dry friction. This causes the typical hysteretic shape
to the steering-wheel angle versus feedback torque
diagram (see e.g. Fig. 11). Such friction is represented
by terms C and F .
o
o
D05403 IMechE 2004

785

2. Viscous damping due to lubricated elements such as


cardan joint, bearings, etc. The typical formulation
for a generic damping force has the form
(3)

F=mx

where m is the coecient of viscous friction and x the


relative velocity. This expression does not give a satisfactory fitting of experimental measurements. Thus
a modified formulation, dependent on the velocity
at the a (or b)-th power, has been used instead.
Consequently the coecient of xa does not have the
dimension of a viscous coecient and is indicated
with m .
3. Load-dependent friction, named Coulombian terms.
This term appears in the rack equilibrium equation.
In fact, experimental observation shows a linear
increase of resistant forces with the load.
Summarizing, frictions in equations (1) and (2) are in
the following forms
C =sgn(d )C +m d a
(4)
a
SW
o
v SW
|C |
F =sgn(x)F +m xb+K sgn(x) b
(5)
a
o
c
c
r
C , F , K , m , m , a and b are coecients to be identified.
o o c v c
Since friction always acts against the relative motion
of the mating part, friction force and velocity vectors are
opposed. Hence the sgn function is needed. Nevertheless,
the numerical integration subroutine employed does not
easily deal with discontinuity due to sign function. This
problem has been solved using the Dahl friction model
[11] based on the equation

2
dF
F
(6)
=kx 1 sgn(x)
dt
F
o
where F is the friction force (or torque), F is the
o
hysteretic value of the friction force (or torque), x is
the velocity of the rack (or the angular velocity of the
column), and k is a coecient that relates the velocity
to the friction.
Both HPS and EPS software models have been
developed using the block representation available in
Simulink environment [12] (see Figs 3 and 4). The
models are mainly based on equations (1) and (2). Time
histories of steering-wheel angle and wheels torque are
the input data. The feedback torque and the wheel
angles are then computed. All the blocks depend on the
parameters to be identified. Every coecient can be set
by a MatLab script or directly in a Simulink model.
In order to increase the readability of the model, subsystems have been masked in rectangular block. For
instance Rack block in Fig. 3 contains the blocks needed
to integrate equation (2), as can be seen in Fig. 5.
During experimental tests the steering wheel is removed
to connect an hydraulic motor to the steering column.
The motor drives the system imposing a constant speed
motion. Thus the inertia of the steering wheel is neglected.
Proc. Instn Mech. Engrs Vol. 218 Part D: J. Automobile Engineering

786

S DATA, M PESCE AND L RECCIA

Fig. 3 HPS Simulink model

Fig. 4 HPS Simulink model

In the EPS model the active damping, which introduces


a torque linearly dependent by the steer angular speed
(and also by the velocity of the car), has been merged
into the block named EPS, as shown in Fig. 4.

3 EXPERIMENTAL MEASUREMENTS
The handling division of CRF has developed a methodology to evaluate steering characteristics by means of
bench tests. This kind of test can be executed without
disassembly of mechanical parts and oers several
advantages. First of all it is possible to evaluate the performance of a subsystem of the vehicle, in this instance
the steering system, uncoupling it from other parts such
as tyres, suspensions, brakes, etc. Moreover, it is possible
to perform the test with the engine turned on or o, to
investigate the influence of power steering. Finally, a test
bench has better repeatability features than a road test,
which is aected by environment variables and driver
personal style.
The experimental lay-out is shown in Fig. 6. The car
is clamped in the frontal part of the chassis; the front
wheels lay on appropriate plates that permit them to
Proc. Instn Mech. Engrs Vol. 218 Part D: J. Automobile Engineering

turn. The bench is equipped to gather the following


signals: steering-wheel angle, steering-wheel torque, and
left and right front-wheel angle and torque. Figures 7, 8
show the typical plots of collected data.
On the left wheel a system of springs act as a torsional
stiness (for small angles) while the right wheel is free.
Hence the elastic displacements due to loads can be
evaluated from the dierence of measured angles. The
manoeuvre consists of a cycle of almost 5060 rotation
of the steering column actuated by an hydraulic motor
assembled in place of the steering wheel. The motor
provides a constant rate rotation based on a feedback
control. The frequencies set for the steering cycles are
0.01 Hz, 0.5 Hz, and 1 Hz. The 0.01 Hz test represents
the quasi-static behaviour. The 0.5 Hz and 1 Hz tests
involve viscous friction and inertia eects. Consequently,
the amplitude in terms of steering wheel torque of the
steering wheel angle versus feedback torque cycle,
increases (see dotted line diagram in Fig. 7). Furthermore, tests are conducted with the car engine on and
o, in order to appreciate the influence of the power
steering. Moreover, both free left wheel and loaded
left wheel cases have been considered. Testing vehicles
equipped with EPS, a fictitious signal is given to the
D05403 IMechE 2004

IDENTIFICATION OF STEERING SYSTEM PARAMETERS

787

Fig. 5 Look under mask of Rack block

Fig. 6 Test bench equipment

electronic control processor to simulate dierent values


of vehicle velocity. In this way the dependence of the
power steering characteristic on the velocity of the car
is investigated.

4 PARAMETERS IDENTIFICATION
As discussed earlier, the parameters identification is split
into two phases. The first one is based on a numerical
approach to the experimental signals. The second one
D05403 IMechE 2004

Fig. 7 Steering-wheel angle versus feedback torque diagram


obtained by test bench (engine on, loaded wheels; cycle
frequencies: 0.01 Hz, 1 Hz)

works by comparing experimental measurements and


models output. The parameter values are varied by an
optimization software until best fitting is reached. This
technique, widely used in various branches of engineering (and with dierent mathematical detail degrees) is a
standard at Fiat Research Centre (see e.g. [13]). In Fig. 9
a logical blocks diagram describes a generic identification
procedure.
Proc. Instn Mech. Engrs Vol. 218 Part D: J. Automobile Engineering

788

S DATA, M PESCE AND L RECCIA

Fig. 8 Steering-wheel angle versus left/right wheel-angle


diagram obtained by test bench

Fig. 9 Identification logical blocks scheme

Figure 10 depicts the typical shape of the steering


wheel angle versus wheels angle diagram. In the AB
stroke no steering is obtained while rotating the column.
This is due to the angular deformation of the torsion
bar. Then the torsion angle of the bar reaches the limit
stop and the wheels start steering following the steering-

Fig. 10 Steering-wheel angle versus wheels angle diagram


(qualitative)
Proc. Instn Mech. Engrs Vol. 218 Part D: J. Automobile Engineering

wheel angle. The torsion bar deformation remains constant. Hence FC and DE strokes are characterized by a
constant slope. The value of this slope, assessed interpolating experimental data, equals the steering ratio.
The torsional stiness of the HPS torsion bar (or
of the EPS torque sensor), is obtained computing the
dierence of slope of FC (and DE) strokes (see Fig. 10)
for the measurements gathered in loaded wheels test and
free wheels test. The details of the procedure are outlined
in Appendix 1.
The half width of the steering wheel angle versus feedback torque cycle (in Fig. 11 the diagram relative to
engine o and free wheels test is depicted) assessed by
the AB stroke length, gives information about the whole
friction in the steering system. Since C and F represent
o
o
the constant (dry) friction term, respectively on steering
column and rack, assuming the pinion radius is equal
to r, AB length is equal to AB=C +F r. Calculating
o
o
this quantity in both engine on and engine o tests it is
possible to define the upper and the lower limits for the
whole friction AB. In fact, with the engine on, the width
of the friction hysteresis is reduced by the eect of the
power steer, which works against the friction downstream of the pinion. Furthermore, the friction measured
is lower because of the eect of the vibrations (even
though the car is stationary).
Hence, the values of AB quantity in the two cases
discussed permit the definition of a boundary suitable
to constraint friction parameters in the identification
procedure.
Identification procedure is coded in a MatLab script
in the following five steps:
1. Setting of the parameters deduced by means of signal
processing.
2. Definition of a trial value for the parameters to
identify.

Fig. 11 Steering-wheel angle versus feedback torque diagram


(qualitative)
D05403 IMechE 2004

IDENTIFICATION OF STEERING SYSTEM PARAMETERS

789

Table 1 Identification procedure


Parameters to identify
t
K
b
AB=C +F r
o
o
C , F , K , F = f (C )
o o c s
b
m , m , a, b (active damping
v c
C=c C only for EPS)
ad SW

Parameters identified in previous


steps

Experimental measure used

t, K , AB
b
t, K , C , F , K , F = f (C )
b o o c s
b

d
, d from 0.01 Hz, engine o, free wheels test
Wheel SW
d
, d from 0.01 Hz, engine o, free wheels and loaded wheels test
Wheel SW
C , d from 0.01 Hz, engine o, and engine on, free wheels test
SW SW
C , d from 0.01 Hz, engine on, loaded wheels test
SW SW
C , d from 0.5/1 Hz, engine on, loaded wheels test
SW SW

3. Run of the Simulink model. The model is driven


by the experimental time-histories of steering-wheel
angle and wheels load and outputs wheel angles and
steering wheel torque.
4. Evaluation of the error function in terms of mean
square dierence between measured feedback torque
time-history and steering-wheel torque output from
the model.
5. Optimization of the parameters output taking into
account the error function until convergence (i.e. error
lower than a threshold) is reached.
Table 1 shows the sequence of steps in identification
procedure. For each step are indicated the experimental measurement used, the parameters to identify,
and those fixed because already identified in previous
steps.
Several interpolating functions are apt to represent
power-steer characteristics F = f (C ). The dierence of
s
b
behaviour of HPS and EPS required the use of dierent
formulations for the power-steer characteristics. The
mathematical expressions used are reported in Table 2.
Since EPS has two levels of assistance (city and normal),
two dierent curves have been measured for the EPS.

Figures 12 to 16 show the comparison between experimental measurements and the numerical output obtained
from the Simulink models once all the parameters have
been identified.

Fig. 12 Steering-wheel angle versus torque feedback diagram


(experimental data: 0.01 Hz, engine on, loaded wheels
test; model output: HPS with identified parameters)

Table 2 Summary of identified parameters


HPS

EPS

t=1/16.01=0.0625
K =1.073 Nm/
b
C =0.0300 Nm
o
F =184.7 N
o
K =0.0449
c

t=1/16.85=0.0593
K =0.6879 Nm/
b
C =0.05 Nm
o
F =285 N
o
K =0.02
c

F=csgn(C )[e|Cb|1] c=6.34


b

F=c C +c Cc3
1 b
2 b

m =0.0136
v
a=2.14
m =2661.8
c
b=1.95

c =0.29
1
c =1.22
2
c =1.68
3

Normal

c =0.21
1
c =0.34
2
c =1.84
3

Speed

60 km/h
100 km/h
120 km/h

0.05
0.015
0.09

m =0.002
v
a=0.5
m =10
c
b=0.5
Active damping
C=c C
ad SW

D05403 IMechE 2004

City

ad

Proc. Instn Mech. Engrs Vol. 218 Part D: J. Automobile Engineering

790

S DATA, M PESCE AND L RECCIA

Fig. 13 Steering-wheel angle versus torque feedback diagram


(experimental data: 1 Hz, engine on, loaded wheels
test; model output: HPS with identified parameters)

Fig. 14 Steering-wheel angle versus torque feedback diagram


(experimental data: 0.01 Hz, engine on, loaded wheels
test; model output: EPS with identified parameters,
normal mode)

Figures 12 and 13 are concerned with HPS, for


0.01 Hz engine on, loaded wheels and 1 Hz engine on,
loaded wheels test respectively. Figures 14 and 15 are concerned with EPS for 0.01 Hz engine on, loaded wheels
for normal and city behaviour respectively. Figure 16
is concerned with the 0.5 Hz test, engine on, loaded
wheels.
A satisfactory matching in 0.01 Hz tests can be
observed. Dynamic behaviour of EPS has been studied
using 0.5 Hz test (Fig. 16) instead of 1 Hz test because
of oscillations observed in 1 Hz measures, probably caused
by the diculty of EPS control unit to manage highfrequency dynamics. For the same reason even 0.5 Hz
fitting does not have the same accuracy of HPS analysis.
Proc. Instn Mech. Engrs Vol. 218 Part D: J. Automobile Engineering

Fig. 15 Steering-wheel angle versus torque feedback diagram


(experimental data: 0.01 Hz, engine on, loaded wheels
test; model output: EPS with identified parameters,
city mode)

Fig. 16 Steering-wheel angle versus torque feedback diagram


(experimental data: 0.5 Hz, engine on, loaded wheels
test; model output: EPS with identified parameters,
normal mode)

However, qualitative values for viscous friction and


active damping have been obtained. All the parameters
identified are summarized in Table 2.

5 CONCLUSIONS
In this investigation the problem of the characterization
of the automobiles steering system has been addressed.
An experimental-analytical procedure to identify the
parameters of a model, suitable both for HPS and EPS,
has been developed and implemented in a software tool.
In particular, the two dof model takes into account static
D05403 IMechE 2004

IDENTIFICATION OF STEERING SYSTEM PARAMETERS

and dynamic frictions, torsion bar/torque sensor stiness, power-steer characteristic, and also active damping
and operation program (city, normal) selected for
EPS.
The identification tool requires, first, experimental
data collected with a test bench. Then a numerical procedure simulates the static and dynamic behaviour of
the steering system. The numerical results are in good
agreement with experimental data for quasi-static tests
and in satisfactory matching for dynamic test of HPS.
It can be concluded that the proposed models are suitable for investigation steering system behaviour, for
reverse engineering applications, and also for tuning,
benchmarking, or diagnostic of actual steering systems.
REFERENCES
1 Park, Y. and Jung, I. Semi-active steering wheel for steerby-wire system. In Proceedings of the ATTCE Congress,
volume 6, Barcelona, 2001.
2 Oh Se Wooh, Park Tong-Jin and Han Chang Soo. The
design of a controller for the steer-by-wire system. FISITA
Congress, Helsinki, 2002.
3 Adams, F. J. Power steering Road feel, SAE Technical
Paper No. 830998, June 1983.
4 Nakayama, T. and Suda, E. The present and future of
electric power steering, JSAE Rev., October 1990, 11(4).
5 Amberkar, S., Kushion, M., Eschtruth, K. and Bolourchi, F.
Diagnostic development for an electric power steering system.
SAE Technical Paper No. 2000-01-0819, January 2000.
6 Okamoto, K., Chikuma, I., Saito, N. and Miyazaki, H.
Improvement of drivers feel of electric power steering, SAE
Technical Paper No. 890079, January 1989.
7 Herkommer, R. Steering the right waySteering the
economic way: energy saving concepts in hydraulic steering
systems. FISITA Congress, Helsinki, June 2002.
8 Ugo, A. and Data, S. Objective evaluation of steering
system quality. 26th FISITA Congress, Prague, 1996.
9 Data, S. and Frigerio, F. Objective evaluation of handling
quality, J. Automob. Eng., 2000, 216.
10 Badawy, A., Zuraski, J., Bolourchi, F. and Chandy, A.
Modelling and analysis of an electric power steering system.
SAE Technical Paper No. 982878, January 1999.
11 Dahl, P. A solid friction model. Technical Report
TOR-0158(3107-18), The Aerospace Corp., El Segundo,
California, 1968.
12 The MATHWORKS Inc. SimulinkDynamic System
Simulation for MATLABUsing Simulink Version 4, The
MATHWORKS, November 2000.
13 De Bernardi, L., Turco, P., Cheli, F. and Resta, M.
Identification of nonlinear cornering forces from road tests,
with extended Kalman filter technique, for vehicle handling
analysis. EAEC, Barcellona, 1999.

APPENDIX 1

D05403 IMechE 2004

According to the notation, q torsion angle can be


defined as q=d d , where d and d are the angles
SW
b
SW
b
of the side steering column end and side pinion end
respectively. Equation (7) can be rewritten as
dC dd
dC
K = b= b r
b
dq
dd dq
r

(8)

where d is the front-wheel steering angle. Under the


r
hypothesis of quasi-static test (0.01 Hz) frictions can
be considered constant. Moreover, there is a cinematic
dependence between the wheels angle and the pinionend angle, because other elasticity can be neglected with
respect to torsion-bar stiness, defined by the steering
ratio
t=

dC
b
dC
r

where C is the loaded wheel torque. By substituting into


r
equation (8) the expression (9) is obtained
dC dd
K =t r r
b
dd dq
r

(9)

Considering steering-wheel, angle-front wheels angle diagram for 0.01 Hzengine-o test. For the free-wheels
test the slope of DE and FC stroke (see Fig. 10) is

dd
dd
SW $ b
dd
dd
r l
r

(10)

(l subscript means free-wheels test) because torsion


angle q can be considered constant outside the CD and
EF strokes. Instead, for the loaded-wheels cycle

dd
dd +dq
SW = b
dd
dd
r r
r

(11)

(r subscript means loaded-wheels test) because the load


varies with steering angle and consequently the torsionbar angle. Subtracting equation (10) from equation (11)
the following is obtained

dd
dd
dq
SW SW =
dd
dd
dd
r r
r l
r

(12)

that, substituted in equation (9), gives

The torsion bar stiness.


The torsion bar stiness is defined as
dC
K = b
b
dq

791

(7)

1
dC
K =t r
b
dd dd
dd
r
SW SW
dd
dd
r r
r l

(13)

Proc. Instn Mech. Engrs Vol. 218 Part D: J. Automobile Engineering

792

S DATA, M PESCE AND L RECCIA

The terms

dC dd
r , SW
dd dd
r
r r
and

dd
SW
dd
r l
obtained interpolating the wheels angle-wheels torque
and steering-wheel, angle-wheels angle diagrams (steering

Proc. Instn Mech. Engrs Vol. 218 Part D: J. Automobile Engineering

ratio t is calculated in the previous step of identification


procedure) permit the calculation of the torsional
stiness K .
b
It can be observed that slopes determined with
equations (10) and (11) are obtained by measures
relative to right (free) wheel both for loaded and free
left-wheel test. In fact, under the spring system force,
the suspension elements bear a displacement. This
quantity must be separated in order to evaluate only the
steering system elastic displacement.

D05403 IMechE 2004

You might also like