model-based analysis of wheel speed vibrations for road friction classification … ·...
TRANSCRIPT
MODEL-BASED ANALYSIS OF
WHEEL SPEED VIBRATIONS FOR
ROAD FRICTION CLASSIFICATION
USING MF-SWIFTAntoine Schmeitz, Mohsen Alirezaei
CONTENTS
Introduction
Road friction classification from wheel speed vibrations
Tyre modelling
Simulation environment
Evaluation of the estimation method
Concluding remarks
21 April 2015MODEL-BASED ANALYSIS OF WHEEL SPEED VIBRATIONS FOR ROAD FRICTION CLASSIFICATION USING MF-SWIFT
INTRODUCTION
Consider an Autonomous Emergency Braking (AEB) test:
21 April 2015MODEL-BASED ANALYSIS OF WHEEL SPEED VIBRATIONS FOR ROAD FRICTION CLASSIFICATION USING MF-SWIFT
INTRODUCTION
Now imagine that it is winter
and the road is icy?
What would happen with the cyclist
if the controller does not know the
friction coefficient between
tyre and road?
Probably this:
21 April 2015MODEL-BASED ANALYSIS OF WHEEL SPEED VIBRATIONS FOR ROAD FRICTION CLASSIFICATION USING MF-SWIFT
INTRODUCTION
Basic problem: identifying friction potential during normal driving conditions
Solution becomes more important with higher levels of vehicle automation
when task are taken over from the human driver
Interesting possible solutions was presented by Toyota (Umeno):
Wheel speed vibrations are used for friction estimation
T. Umeno: Estimation of Tire-Road Friction by Tire Rotational Vibration Model,
Review of Toyota CRDL, vol. 37 No. 3, 2002.
21 April 2015MODEL-BASED ANALYSIS OF WHEEL SPEED VIBRATIONS FOR ROAD FRICTION CLASSIFICATION USING MF-SWIFT
http://www.tytlabs.com/english/review/rev373epdf/e373_053umeno.pdf
INTRODUCTION
In 2006 this method was evaluated by Pavković et al. together with the slip-
slope method (SAE paper 2006-01-0556)
Able to distinguish different road friction conditions from both methods
Identified slip stiffnesses from both methods did however not agree
Contribution of this paper:
Review method and tyre modelling
Developing a simulation environment to study the estimation method in
terms of accuracy and robustness
Simulation environment to develop Integrated Friction Estimator using
multiple methods
21 April 2015MODEL-BASED ANALYSIS OF WHEEL SPEED VIBRATIONS FOR ROAD FRICTION CLASSIFICATION USING MF-SWIFT
ROAD FRICTION CLASSIFICATION
FROM WHEEL SPEED VIBRATIONS
Estimation method:
identify parameters of a second order linear time-invariant (LTI) system
use discrete time autoregressive (AR) model in combination with
instrumental variable (IV) method21 April 2015MODEL-BASED ANALYSIS OF WHEEL SPEED VIBRATIONS FOR ROAD FRICTION CLASSIFICATION USING MF-SWIFT
ABS wheel speed sensor
Signal conditioning
Bandpass filter
Estimation method
Tyre model• Slip stiffness• Friction
TYRE MODELLING
Basic idea:
Slip stiffness is measure
for friction
Slip stiffness drops at
low friction surfaces
Magic Formula model:
Slip stiffness is changed
by scaling factor LKX
21 April 2015MODEL-BASED ANALYSIS OF WHEEL SPEED VIBRATIONS FOR ROAD FRICTION CLASSIFICATION USING MF-SWIFT
TYRE MODELLING
Simple tyre model of Umeno
2nd order model for describing the 1st tyre in-plane resonance
Natural frequency affected by sidewall stiffness (e.g. inflation pressure)
Damping ratio increases with decreasing slip stiffness (i.e. friction)
21 April 2015MODEL-BASED ANALYSIS OF WHEEL SPEED VIBRATIONS FOR ROAD FRICTION CLASSIFICATION USING MF-SWIFT
( )22
1
2
1
RC
JJ
J
C
J
Cf
F
ba
a
an
κ
θ
θ
ζ
π
+=
=
TYRE MODELLING
MF-Swift
rigid ring
Magic Formula slip model
contact patch model
enveloping model
Well-validated
Accurately represents
primary tyre modes
(rigid ring modes)
21 April 2015MODEL-BASED ANALYSIS OF WHEEL SPEED VIBRATIONS FOR ROAD FRICTION CLASSIFICATION USING MF-SWIFT
SIMULATION ENVIRONMENT
Quarter vehicle model
rigid car body, wheel carrier and rim bodies
suspension stiffness and damping
linear elements, uncoupled in X, Z
TNO’s MF-Swift tyre model
Measured road profiles
21 April 2015MODEL-BASED ANALYSIS OF WHEEL SPEED VIBRATIONS FOR ROAD FRICTION CLASSIFICATION USING MF-SWIFT
SIMULATION ENVIRONMENT
Power spectral densities of wheel speed for different slip stiffness (i.e. friction)
Tyre resonance ‘disappears’ for low friction surfaces (i.e. small LKX)
21 April 2015MODEL-BASED ANALYSIS OF WHEEL SPEED VIBRATIONS FOR ROAD FRICTION CLASSIFICATION USING MF-SWIFT
SIMULATION ENVIRONMENT
Natural frequencies and damping ratios from eigenvalue analysis
Conclusions:
Damping increases with decreasing slip stiffness (i.e. friction)
(similar conclusion as Umeno et al.)
However: natural and peak frequencies (fn, fp) also decrease21 April 2015MODEL-BASED ANALYSIS OF WHEEL SPEED VIBRATIONS FOR ROAD FRICTION CLASSIFICATION USING MF-SWIFT
EVALUATION OF ESTIMATION METHOD
21 April 2015MODEL-BASED ANALYSIS OF WHEEL SPEED VIBRATIONS FOR ROAD FRICTION CLASSIFICATION USING MF-SWIFT
Algorithms have
been implemented
Results can be
compared
LKX = 0.5
Model Estimation
fp [Hz] 32 34
ζζζζ [-] 0.24 0.20
EVALUATION OF ESTIMATION METHOD
similar trends as found by others
generally resonance frequency is
rather accurately estimated
LKX = 0.2 (~ice): no excitation
inaccurate results
increased damping is still good
measure for low friction
Inaccurate simple tyre model:21 April 2015MODEL-BASED ANALYSIS OF WHEEL SPEED VIBRATIONS FOR ROAD FRICTION CLASSIFICATION USING MF-SWIFT
LKX = 1 LKX = 0.5 LKX = 0.2
Model Estimation Model Estimation Model Estimation
fp [Hz] 35 35 32 34 23 30
ζζζζ [-] 0.17 0.14 0.24 0.20 0.44 0.32
ζζζζref / ζζζζ [-] 1 1 0.71 0.70 0.39 0.44
( ) ( )( ) 22* LKX2
1
2
1,
2
1
RC
JJ
J
C
RC
JJ
J
C
J
Cf
F
ba
aF
ba
aan
κ
θ
κ
θθ ζπ ⋅
+=
+==
EVALUATION OF ESTIMATION METHOD
Influence of road roughness
as expected influence of road
roughness
deviations are relatively small
estimation method is quite
robust
21 April 2015MODEL-BASED ANALYSIS OF WHEEL SPEED VIBRATIONS FOR ROAD FRICTION CLASSIFICATION USING MF-SWIFT
Model Estimation
Smooth road Rough road 1 Rough road 2 Rough road 3
fp [Hz] 35 35 39 37 37
ζζζζ [-] 0.17 0.14 0.15 0.13 0.15
ζζζζref / ζζζζ [-] 1 0.93 1.1 0.93
CONCLUDING REMARKS
First tyre resonance as indicator for monitoring tyre-road friction is studied
Simulation environment to study this application has been developed:
results qualitatively agree well with results found in literature
reference for algorithm design and testing, i.e.: ‘real’ values are known
frequencies and damping ratios can be estimated with reasonable accuracy
trends observed with changing road friction are consistent
Review of method:
Simple tyre model (Umeno) is insufficiently accurate to identify slip stiffness
MF-Swift model can be used to translate identified frequencies and damping
ratios to correct slip stiffness values21 April 2015MODEL-BASED ANALYSIS OF WHEEL SPEED VIBRATIONS FOR ROAD FRICTION CLASSIFICATION USING MF-SWIFT
CONCLUDING REMARKS
Next steps:
Test the method using real car data
identify operating range, accuracy and limitations
Implement method in an Integrated Friction Estimator
sensor / estimator fusion
basic idea: several algorithms should give the same tyre characteristic
values (i.e. comparing apples with apples)
one source (tyre model) to relate tyre characteristic values to friction
estimate
21 April 2015MODEL-BASED ANALYSIS OF WHEEL SPEED VIBRATIONS FOR ROAD FRICTION CLASSIFICATION USING MF-SWIFT
THANK YOU FOR YOUR ATTENTION