Vehicle Road Interaction - STP1225

Copyright by ASTM Int'l (all rights reserved); Sat Dec 24 01:23:58 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
STP 1225
Vehicle-Road Interaction
Bohdan T. Kulakowski, Editor
ASTM Publication Code Number (PCN):

04-012250-08
ASTM
1916 Race Street
Philadelphia, PA 19103
Printed in the U.S.A.
Library of Congress Cataloging in Publication Data
Vehicle-road interaction/Bohdan T. Kulakowski, editor.
p. cm.--(STP; 1225)
"Papers presented at the Vehicle-Road Interaction II Conference
held in Santa Barbara, California on 17-22 May 1992"--Foreword.
Includes bibliographical references and index.
ISBN 0-8031-1893-7
1. Pavements--Live loads--Congresses. 2. Trucks--Tires--
Congresses. I. Kulakowski, Bohdan T. II. Vehicle-Road
Interaction II Conference (1992: Santa Barbara, Calif.)
III. Series: ASTM special technical publication; 1225.
TE250.V34 1994
625.7--dc20 94-2400
CIP
Copyright 9 1994 AMERICAN SOCIETY FOR TESTING AND MATERIALS, Philadelphia, PA.
All rights reserved. This material may not be reproduced or copied, in whole or in part, in any
printed, mechanical, electronic, film, or other distribution and storage media, without the written
consent of the publisher.
Photocopy Rights
Authorization to photocopy items for internal or personal use, or the internal or personal
use of specific clients, is granted by the AMERICAN SOCIETY FOR TESTING AND MATERI-
ALS for users registered with the Copyright Clearance Center (CCC) Transactional
Reporting Service, provided that the base fee of $2.50 per copy, plus $0.50 per page is
paid directly to CCC, 222 Rosewood Dr., Danvers, MA 01923, Phone: (508) 750.8400; Fax:
(508) 750-4744. For those organizations that have been granted a photocopy license by
CCC, a separate system of payment has been arranged. The fee code for users of the
Transactional Reporting Service is 0-8031-1893-7/94 $2.50 + .50.
Peer Review Policy
Each paper published in this volume was evaluated by three peer reviewers. The authors
addressed all of the reviewers' comments to the satisfaction of both the technical editor(s) and
the ASTM Committee on Publications.
The quality of the papers in this publication reflects not only the obvious efforts of the authors
and the technical editor(s), but also the work of these peer reviewers. The ASTM Committee
on Publications acknowledges with appreciation their dedication and contribution to time and
effort on behalf of ASTM.
Printed in Philadelphia
March 1994
Foreword
This publication, Vehicle-Road Interaction, contains papers presented at the Vehicle-Road

Interaction II conference held in Santa Barbara, California on 17-22 May 1992. The conference
was sponsored by the Engineering Foundation (related ASTM Committees are
D-4 on Road and Paving Materials and E-17 on Pavement Management Technologies). Bohdan
T. Kulakowski of Penn State University presided as symposium chairman and was editor of
this publication.
Contents
Overview
MODELING AND SIMULATION OF VEHICLE DYNAMICS AND VEHICLE-ROAD

DYNAMIC INTERACTION
Simulation and Analysis of Trucks Using the Modeling and Simulation

Program B A M M S ~ c H R I S H. VERHEUL, G. BATSTRA, AND S. T. H. JANSEN
Predicting Vertical Dynamic Tire Forces of Heavy Trucks--DAVID J. COLE

AND DAVID CEBON 27
Computer Simulation of Road-Vehicle Dynamic Interaction Forces of Three-

and Four-Axle T r u c k S - - M A T T I HUHTALA, MAURI VESIMAKI,
AND PEKKA HALONEN 36
Heavy Truck Properties Significant to Pavement Damage--THOMAS D. GILLESPIE

AND STEVEN M. KARAMIHAS 52
LABORATORY AND FIELD TESTS OF VEHICLE-INDUCED PAVEMENT LOADING
Some Factors Affecting the Design and Use of the Texas Mobile Load
S i m u l a t o r - - F R E D E R I C K HUGO 67
Measured and Calculated Pavement Response in the Danish Road Testing

Machine--JORGEN KRARUP 89
TIRE CHARACTERISTICS
Influence of Tread Polymer on Traction, Rolling Resistance, and Wear

Properties of TireS--IAN R. GELLING 107
Traction Tests on an Expressway Treated with a Pavement Rejuvenator--

FRANCIS NAVIN 119
vi Contents
RIDE QUALITY AND ROAD ROUGHNESS
Development and Evaluation of the Ride Number Concept--

ELSON B. SPANGLER AND WILLIAM J. KELLY 135
A Geometry-Based Contact/No-Contact Simulation Model Applied to Ride

QualitymBRIAN GILMORE 150
ADVANCES iN VEHICLE SUSPENSION DESIGN AND CONTROL
Steering and Suspension Development of Road VehiclesmCHRIS J. ALSTEAD AND

JOHN P. WHITEHEAD 173
NOISE EMISSION DUE TO VEHICLE-TIRE-ROAD INTERACTION
Vehicle Noise Testing Problems Due to Vehicle/Tire]Road Interactions--

PETER WAGNER 185
Road Noise Simulation Using Road Surface Weighting Functions--

MARK P. VOUTYRAS AND KEVIN R. THOMSON 204
FUEL EFFICIENCY AND ROLLING RESISTANCE
The Effect of Vehicle-Road Interaction on Fuel Consumption--LASZLO GYENES

AND C.G.B. MITCHELL 225
The Influence of Pavement Evenness and Macrotextureon Fuel Consumption--

DELANNE 240
Rolling Resistance Character~tics of New Zealand Road Surfaces--

PETER D. CENEK 248
Index 265
STP1225-EB/Mar. 1994
Overview
Most of the problems associated with the safety, economy, and overall quality of road
transportation are affected by the characteristics of both roads and vehicles and by the manner
in which these two dynamic systems interact. In spite of the complex interaction between
roads and vehicles, there has been little interaction and a rather limited flow of information
between road and vehicle researchers. At most professional meetings and conferences, the
emphasis is usually placed either on roads or on vehicles, but very rarely on both. Moreover,
in those conferences that do include both road and vehicle topics in their programs, the road
and vehicle presentations take place in separate and often concurrent sessions. In 1990, the first
Vehicle-Tire-Pavement Interface Conference was conducted by the Engineering Foundation. In
this conference, equal emphasis was placed on the vehicle and the roadway. In 1992, the
second Engineering Foundation conference on Vehicle-Road Interaction was held at Santa
Barbara, California. This special technical publication has been published as a result of the
Vehicle-Road Interaction II Conference in an effort to communicate the current state of the
art and future research needs to those involved in studies of vehicles and roads.
The 16 papers published in this volume can be grouped into seven subject areas. These
areas are: modeling and simulation of vehicle dynamics and vehicle-road dynamic interaction,
laboratory and field tests of vehicle-induced pavement loading, tire characteristics, ride quality
and road roughness, advances in vehicle suspension design and control, noise emission due
to vehicle-tire-road interaction, and fuel efficiency and rolling resistance.
Modeling and Simulation of Vehicle Dynamics and Vehicle-Road Dynamic Interaction

There are four papers in this section. The first paper, by Verheul et al., presents a general-
purpose computer simulation program, called BAMMS, developed from the bond graph model
of multibody systems. The features that set BAMMS apart from many other vehicle simulation
programs are its flexibility and expandability to a great variety of dynamic systems that can
be modeled using bond graphs. Two applications of the program are described; one is the
analysis of performance and optimization of design of a racing truck, and the other is the
investigation of stability of tractor-trailer combination.
Two other papers in this section deal with computer simulation of the dynamic tire forces
of heavy trucks, which constitute a very important element of vehicle-road interaction. The
dynamic tire forces of heavy vehicles are the primary cause of pavement damage and have
become the subject of many research studies in recent years. However, only a few computer
models, of many that have been developed in those studies, have been validated using experi-
mental data. The paper by Cole and Cebon describes a nonlinear three-dimensional model of
a heavy vehicle and the results of an experimental validation study. In general, very good
agreement was found between measured and calculated tire forces. The three-dimensional
model was then compared with a two-dimensional model. The authors concluded that the two-
dimensional model is satisfactory for predicting tire forces of typical leaf-sprung, articulated
vehicles operating under typical conditions of speed and road roughness. Huhtala et al. used
a commercial multibody system analysis software, called ADAMS, to simulate tire forces
generated by three-axle and four-axle trucks. The paper provides a thorough analysis of the
results of computer simulation. Experimental validation of the truck models is planned for the
near future when data from tire force measurements become available.
The fourth paper, authored by Gillespie and Karamihas, uses analytical relationships between
truck properties and pavement damage to determine which truck characteristics have the
1
Copyright* 1994 by ASTM International www.astm.org
2 VEHICLE-ROAD INTERACTION
strongest effect on pavement damage. The truck characteristics studied included truck type,
axle loads, number of axles, spacing between axles, suspension type, and tire parameters.
Static axle loads were found to have the strongest effect on pavement damage.
Laboratory and Field Tests of Vehicle-Induced Pavement Loading

The two papers published in this section deal with laboratory devices used for accelerated
testing of pavement response to vehicle loading. The first paper, by Hugo, focuses on the
design of the Texas mobile load simulator, whereas in the other paper, Krarup reports results
from tests conducted on the Danish road testing machine. The expected dynamic performance
of the Texas simulator was investigated using computer simulation. Two 1:10 scale models
of the simulator were then built and tested to augment the design of the full-scale machine.
Krarup's paper can serve primarily as a source of relevant technical data including pavement
stress and strain distributions measured under two types of loads, rolling truck tire and falling
weight deflectometer. The interpretation and conclusions from the test results are left up to
the reader to consider.
Tire Characteristics
This section includes two papers. In the first paper, Gelling describes the tire tread properties
and their effect on traction, rolling resistance, and wear. Tire traction, rolling resistance, and
wear are known to depend on tire construction and road surface texture characteristics. Gelling
shows that these tire properties can also be related, to some extent, to viscoelastic properties
of the tread material. In addition, the paper raises an issue of environmental effects of tire
properties. Tire recycling and vehicle/tire contribution to the greenhouse effect are two major
areas of concern to the tire and automotive industries.
The paper by Navin presents a skid resistance case study conducted on a surface treated
with a pavement rejuvenator using a 1991 Lincoln Town Car equipped with an antilock
brake system (ABS). The main difficulty in traction tests involving ABS-equipped vehicles
is determining when and where the braking actually starts and stops. The testing procedure
described by the author should be of interest because it offers an economic alternative to an
expensive skid trailer.
Ride Quality and Road Roughness

Ride quality is an important aspect of vehicle-road interaction, not only as a matter of
comfort but, more importantly, as a factor affecting driver fatigue and thus traffic safety. Ride
quality is a subjective quantity based on a level of comfort perceived by persons traveling
over a selected section of road. Since obtaining subjective ratings of ride quality is a rather
involved and costly process, highway engineers developed objective measures calculated from
road profile and/or vehicle response data to estimate the ride quality. Spangler and Kelly
compared five different measures of ride quality that are in common use today, and found the
Ride Number to be superior to the other measures in this group. The Ride Number is derived
from the road profile measurements, then processed with a digital filter that emphasizes the
range of wavelengths to which human subjects have been found to be most sensitive while
traveling in highway vehicles.
The paper by Gilmore, while concentrating also on objective evaluation of ride quality,
introduces a ride quality measure derived from the dynamics of interaction between a rider
and a seat. A new computer program is presented that can be used to simulate the rider-seat
interaction. The program can also be employed as a tool in seat and vehicle design. A new
feature of the program is that it includes nonlinearities due to geometry or changing kinematic
OVERVIEW 3
constraints in the rider-seat-vehicle system. These nonlinearities become increasingly important

for ride quality estimation when the road roughness increases and when the vehicle suspension
is stiff, which may lead to situations where contact between the seat and the rider is lost or
the seat bottoms out.
Advances in Vehicle Suspension Design and Control

The part of a vehicle that has the strongest impact on how the vehicle interacts with the
road is the suspension. Vehicle suspension characteristics are crucial for ride quality, handling,
and dynamic tire forces. Traditionally, suspension designers have relied on experience, intuition,
and subjective measures of a vehicle's ride and handling performance. This was primarily
because of the complexity of the vehicle design and due to the lack of adequate analytical or
numerical models of vehicle dynamics. In their paper, Alstead and Whitehead outline some
of the objective methods developed and used by the Motor Industry Research Association in
England in evaluating vehicle ride and handling performance. The authors suggest, however,
that in spite of an increase in the use of objective techniques, subjective techniques will
continue to play an important part in the development of new suspension designs.
Noise Emission Due to Vehicle-Tire-Road Interaction

The level of noise generated by a vehicle must be considered in two categories----exterior
and interior to the vehicle. Concerns about the environmental effects of exterior vehicle noise
have been growing rapidly, especially in densely populated areas. These concerns lead to
legislation reducing the acceptable level of exterior noise or creating tax advantages to the
manufacturers of low-noise vehicles. The implementation of the various legislative actions
requires accurate and reliable noise testing procedures, which is the topic of the paper by
Wagner. This paper presents the results of several studies conducted by Volkswagen in an
effort to develop vehicle noise testing methods. The effects of several test variables such as
road surface, type of tire, tire inflation pressure, and temperature are examined. If these
variables are not controlled sufficiently, the measurements of exterior noise may vary by 3
dB or more, which constitutes a very significant error. Another very important result reported
by Wagner is that, contrary to widespread opinions, tires with good traction characteristics
don't necessarily have to be noisy.
The second paper in this section, authored by Voutyras and Thomson, focuses on vehicle
interior noise. The primary objectives of the research reported in the paper were to investigate
the effects of different road surfaces on vehicle interior noise and to develop a method that
would allow for predicting the level of noise generated on different road surfaces based on
vehicle noise measured on a single surface. The method, called the road surface weighting
functions method, produces results that have been found to be in very good agreement with
experimental data.
Fuel Efficiency and Rolling Resistance

Rolling resistance is generally considered to be a characteristic of a tire. However, from a
broader perspective, rolling resistance and vehicle fuel consumption depend on characteristics
of both vehicle tires and the road surface. In their paper, Gyenes and Mitchell review the
results of experimental measurements of the rolling resistance and fuel consumption of cars
and trucks using different types of tires and traveling on different road surfaces. Road surface
macrotexture and roughness can each increase the fuel consumption of cars by 5% and trucks
by 10%. On unpaved roads, this effect is further magnified by 15 to 20%. It is also stated
that rolling resistance and thus fuel consumption of trucks can be reduced by the use of higher
4 VEHICLE-ROADINTERACTION
inflation pressure and wide single tires in place of dual tires. However, increased tire inflation
pressure as well as replacing dual tires with wide single tires causes greater pavement wear.
In the two other papers in this section, the authors discuss the effects of the road surface
characteristics, roughness and texture, on the rolling resistance of typical passenger car/tire
combinations. In his paper, Delanne presents the results of a very comprehensive study con-
ducted at the Laboratoire Central des Ponts et Chauss6es in France involving tests performed
in the laboratory, on a test track, and on in-service roads. The results are in good agreement
with the results reported by Gyenes and Mitchell earlier in this section. The important question
raised in this paper is how the low rolling resistance requirements relate to other vehicle and
road characteristics such as traction and noise. The author states that road surfaces that are
beneficial for fuel consumption can be detrimental to other road qualities. The paper by Cenek
reports on a similar study conducted in New Zealand. It describes the development of the test
method for the on-road determination of roiling resistance and the results of application of
this method on a variety of roads having widely differing surface texture characteristics.
The papers published in this volume should provide the reader with the latest information
and the direction for future research in the area of vehicle-road interaction. I would like to
acknowledge sincerely the efforts of the authors, reviewers, ASTM personnel, and the confer-
ence organizing committee. Claude Lamure of INRETS, France, served as the conference co-
chairman and the following served as members of the organizing committee: Alexander A.
Alexandridis, R. B. J. Hoogvelt, Bfirckhard Horn, Byron N. Lord, Mark A. Poelman, Ulf
Sandberg, Elson B. Spangler, Margaret M. Sullivan, and A. Roger Williams. In addition, J.
J. Henry, Charles F. Scheffey, Frank W. Schmidt, and James C. Wambold were the "ex
officio" members of the committee. The support of the conference by the Federal Highway
Administration is also gratefully acknowledged.
Bohdan T. Kulakowski
The Pennsylvania State University,
Pennsylvania Transportation Institute,
University Park, PA 16802.
Conference Chairman and Editor
Modeling and Simulation of Vehicle
Dynamics and Vehicle-Road Dynamic
Interaction
Chris H. VerheuL 1 G. Batstra, 2 a n d S. T. H. J a n s e n 1
Simulation and Analysis of Trucks Using

the Modeling and Simulation Program
BAMMS
REFERENCE" Verheul, C. H., Batstra, G., and Jansen, S. T. H., "Simulation and Analysis
of Trucks Using the Modeling and Simulation Program BAMMS," Vehicle-Road Interaction,
ASTM STP 1225, B. T. Kulakowski, Ed., American Society for Testing and Materials, Philadel-
phia, 1994, pp. 7-26.
ABSTRACT: A description is given of a general purpose, open-structured simulation tool

based on the bond graph method. The method and capacities of the program are discussed.
Research performed with the help of the program are discussed. Models are made of a race
truck in order to study and optimize the road-handling behavior of the vehicle. The final part of
the paper discusses models made of truck trailer combinations. The purpose of these simulations is
to study parameter variations with respect to the stability of these vehicles. The general intention
of the paper is to give an impression of the modeling and analysis approach used when working
with BAMMS.
KEY WORDS: dynamic simulations, bond graphs, vehicle dynamics, handling, and stability
This paper discusses methods and features of a simulation program developed in a cooperative
effort at TNO in Delft and at the Department of Mechanical Engineering of the Delft University
of Technology. Two examples of research illustrate use of the program. The two topics are:
the optimization process of a race truck and parameter variations in truck trailer models to
study their impact on vehicle stability.
In this paper, presentation of research results is a secondary target. The main purpose of
the paper is to demonstrate the advantages of a simulation program with an open-structured
user interface and the possibility for on-the-fly creation of modeling and analysis tools dedicated
to a certain research.
The BAMMS Simulation Program
The acronym BAMMS stands for Bondgraph-based Algorithm for Modeling Multibody
Systems. The purpose of the program is to offer a flexible, adjustable, and user-extendable
software tool for development and analysis of continuous simulation models. The BAMMS
program can be used to model discrete systems also, but was originally intended to model
and analyze physical systems whose behavior can be described by differential and algebraic
equations.
1Research engineer, TNO Road-Vehicles Research Institute in Delft, Schoemakerstraat 97, PO Box
6033, The Netherlands.
z Graduate student, Department of Mechanical Engineering, Delft University of Technology.
7
Copyright* 1994 by ASTM lntcrnational www.astm.org
History of BAMMS' Development

The development of the BAMMS program started in 1985 as part of diploma work at the
Delft University of Technology. The goal of the research was to obtain an advanced three-
dimensional model of a passenger car. The model should be detailed enough to study the
"torque steer" phenomenon. Much experience was available in the department on using the
bond graph method in modeling and simulation studies [1,2]. At this point, it was decided to
develop a program that introduces the advantages of the bond graph method in a multibody
simulation program. Successive versions of BAMMS were used in a number of related vehicle
studies and applications. The program started as a model generator for the simulation programs
TUTSIM and ACSL. In 1987, in a Ph.D. study by the first author, the simulation environment
of BAMMS was developed to include the possibility of on-line wire frame animations. In
1991, the Delft University of Technology and the Institute of Road-Vehicles of TNO in Delft
decided to cooperate in the development of the BAMMS program. In past years, a number
of full vehicle models made with BAMMS have been successfully validated by measurement
data both in steady-state and transient situations.
Using Bond Graphs

The BAMMS software is separated in the modeling program and in simulation programs.
In the modeling program, users create models by defining and connecting equations from
predefined submodeis. These submodels (macros) are stored as ASCII data in user-definable
macro files. In the present version of the program, models and analysis runs are defined in a
command-driven dialogue. In future versions, direct graphical input of bond graph data will
be available.
Equations in the submodels are derived from bond graphs. Basically, the bond graph method
defines the relations in a system by means of a graphical representation of the power flow in
the system. The physical relations of the system are represented as components connected by
power bonds. A limited number of standard components define power transformations and
storage or dissipation of energy. Two types of summation blocks are defined in the bond graph
syntax; the "one" (1) junction represents a summation of effort variables. In the mechanical
domain these are forces and moments. At the same time, only one flow is defined at such a
one junction. The bond graph "zero" (0) junction is a summation point of flow variables:
linear and angular velocities. Similarly, at a zero junction, an effort variable is transmitted to
all bonds connected to the junction. The bond graph relations in BAMMS macros are already
made causal. An integrative calculation sequence has been maintained as much as possible to
prevent numerical problems.
Some advantages of using bond graphs are:
1. The bond graph syntax is not limited to one physical domain; it can be used to define
models in several domains. In the mechanical domain, the effort variables are represented
by the forces and torques in the system. The flow variables are represented by the
velocities and angular velocities in the system.
2. The bond graph method presents a logical graphical schematization of the system
modeled. It offers insight in the relations existing in the system and can be adjusted
and expanded very easily.
3. The method can be used both to model scalar relations and vector relations. Multibonds
are defined in the syntax to model vector bonds with a variable dimension.
Although it is helpful, knowledge of the bond graph method is not required when using
BAMMS.
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized
VERHEUL ET AL. ON BAMMS 9
The BAMMS programs consist of a fixed number of commands and a variable number of
macros. Commands are available t o perform a number of elementary actions in the model
composition and in the analysis phase of the program. Macros are essentially user defined
and are stored as ASCII data in separate macro files. An on-line macro interpreter is included
in the BAMMS programs, which gives users the opportunity to develop and test new macros
while running the program. Once a new macro has been tested and stored in one of the macro
libraries, its user interface is fully identical to that of existing "standard" macros or commands.
Program Features
A number of features are implemented in BAMMS:
User-Definable Block Operators--All equations in a model are stored as blocks in an overall

block diagram table. Block operators define what action must be performed to produce output
variables. Output variables can be scalars, vectors, or matrices. All operators are user defined
and can connect (commercially available) Fortran subroutine libraries as new program elements.
Syntax checks will be performed automatically when new block operators are used in
model equations.
Interactive and Batch Mode--All programs can be used both interactively and in batch. A
graphical representation of the model geometry is directly generated. This offers a frame for
rapid model prototyping and fast model debugging. In both program phases, all commands
and macros are echoed to the model description file.
Minimized Pre- and Post Processing--The BAMMS program is an integrated set of Fortran
77 subroutines with a minimum of pre- and post processors required. Implementations of the
program have been made on Unix Workstations, on MSDOS machines, and on VAX and Cray
hardware platforms.
The Modeling Phase

In the modeling phase of the program, all model equations are generated. The program flow
of both the modeling and the simulation phase is visualized in Fig. 1. The figure demonstrates
the main program, an extendable amount of macros, and possible program input and output.
In the final stage of the modeling phase, model equations are checked, compressed, optimized,
and written out as a Fortran 77 file. This file is compiled and linked to the BAMMS program
libraries to produce a stand-alone simulation program.
Mechanical Macros Available in BAMMS

BAMMS contains a library of macros defining equations for models in the mechanical
domain. The models are defined in a Cartesian space. Each (rigid) body introduces six degrees
of freedom. The three components of the angular velocity are observed in the local body fixed
frame. The three components of the linear velocity of the center of gravity of the body are
considered with respect to the global inertial frame. Experience has shown that this is the
most effective choice of motion variables.
Orientation transformations are performed by using direction cosines. For definition of the
cosines of frames, the user can select Tayt-Bryant angles or Euler parameters.
Kinematical constraints can be set up by using Lagrangian multipliers [3,4]. The equations
connected with constraint elements contain the necessary information for a symbolical genera-
tion of the constraint Jacobean matrix. In the simulation program, the required constraint forces
10 VEHICLE-ROADINTEF:iACTION
I E777~
wire-
t........... frame
[~a*ros I
I I
[~.d~ I
1
I=m! 'e/' I" 1
~UNCOM ~ ; ; ~ model L
l~acros ~ - - ~ "exe F
~messaoe
'f~ ei-e~l~-
~pline plots
K h'ea.transfer
[P~
I ~
[
simuI,data
vslarlabl data [ wireframe
~odemovles [~
rl '
I Other
Programs
FIG. 1--Flowchart of a BAMMS modeling and simulation run.
are represented by Lagrangian multipliers and are solved by using Wittenburg's method.
Possible numerical errors are corrected by using the Baumgarte stabilization method. The
advantage of the method is that all bodies are uncoupled, meaning that the rank of the biggest
inertia matrix is three. Thus, the dimension of the largest matrix to be solved in the system
is equal to the number of constraints.
Mechanical macros are separated in:
Flow-producing Macros--This group contains body macros, a gyrostat macro, and the
external flow source.
Transformation Macros--Defining the algebraic relationships between flows and efforts of

different points in a body or mechanism and the transformation from one frame of axes
to another.
Effort-producing Macros--These macros define massless connections between flow-produc-

ing macros. Macros are available to model point, line, and beam connections. To these macros,
the actual effort-producing elements are attached, such as spring-dampers, relaxation spring-
dampers, external effort sources, and kinematic constraints. Some other examples of effort-
producing macros are the wheel macro based on the so-called "magic formulae" equations
and a macro of an engine crankshaft with differential.
Simulation Program Features

Figure 1 also demonstrates that communication to external programs is possible by means
of ASCII input and output data files. The format in which data is written to these files can
be adjusted while running the program. This feature can be used to export BAMMS simulation
data to (graphical) post processing programs. Importing numerical data from measurements
and/or other software programs is also supported by this feature. Once compiled and linked,
a BAMMS simulation model can be used in a flexible way. All parameters and initial values
of integration blocks can be adjusted while running the simulation program.
In the simulation program, a range of commands and macros is available to perform both
time domain analysis and frequency domain analysis. Time domain simulations can be per-
formed with a selection of integration methods [5,6]. A number of available simulation options
will be discussed in the case studies treated in the remaining part of the paper.
Analysis and Optimization of a Race Truck
Introduction
A study has been carried out for a Dutch trucking company to optimize the road-handling
behavior of a Scania 143m. This truck has been used in races for a number of years and has
undergone a number of modifications to improve its road holding and handling behavior.
Together with students of the Delft University of Technology, the relevant parameters of the
truck have been assessed, and.a number of two-dimensional models have been established to
analyze the behavior of the vehicle. The purpose of this research was to obtain optimum values
of the roll stiffnesses and damping of the truck wheel suspensions with respect to the handling
of the vehicle.
Next, more complex models have been established with BAMMS in order to obtain more
accurate models of a high-powered road vehicle with a relatively flexible chassis. Models of
different levels of complexity are described and compared below. A number of different
methods are discussed that were used to obtain information about the influence of certain
parameter variations on the vehicle's handling and stability.
Simulations were performed with two different models, a complex model and a simplified
model. A short description of both models follows.
The Complex Model--The complex truck model consists of a relatively light flexible chassis
frame that carries a rigid body. The body mass includes the engine and the cabin of the truck.
A number of joints are defined between chassis and body. The rear of the body is connected
to the chassis frame by means of a massless beam element. The chassis frame consists of two
rigid bodies in the longitudinal direction of the vehicle. By means of massless flexible beams,
the bodies are connected at the front and rear of the truck. The bending stiffness values of
the beams are derived from static measurements at the Delft University of Technology.
The wheel axle suspensions of the truck are simply modeled as point connections. The
suspensions introduce no relevant geometric effects. The connections are considered rigid in
longitudinal direction of the vehicle. They provide a compliance in vertical direction equal to
that of the parabolic springs of the actual truck. Lateral movement of the axles is not constrained
by the point connections. For this purpose some additional constraints have been defined. A
schematic overview of the chassis and suspensions is shown in Fig. 2. The number of degrees
of freedom of the complex model is 18.
The Simple Truck M o d e l - - - A simplified truck model was made as a fast prototype for the
complex model. With the simple model, test simulations (e.g., for determination of control
parameters) were speeded up significantly.
The simplified model consists of a rigid vehicle body to which two gyrostats are mounted
to model the front and rear axle. Gyrostat macros are rotational inertias with one degree of
freedom mounted to a carrier frame. In this application they represent a rotation about the
longitudinal axis of the vehicle.
The torsional frame elasticity and the roll stiffnesses of the wheel axle suspensions are
modeled by rotational stiffnesses between the vehicle roll and the gyrostat rotations. The
number of degrees of freedom of this model is 14.
Special M o d e l Features Both models have double tires at the rear and single tires at the
front axle, which are represented by magic formula steady-state tire models (according to SAE
Paper 890087 [7]) capable of simulating combined slip conditions. The engine and drive line
are modeled as a torque acting on the rear wheels through a differential that can be locked.
PI controllers are defined to control the steering angular velocity and the throttle position.
Handling C u r v e s - - G e o m e t r i c linearization of the system equations (i.e., for a small path

curvature) of a two-wheeled vehicle model results in the following equations
l
= ~ - (a2 - al)
mV z
K = may = R
where
= steer angle of front wheel,

eq = slip angle of the front wheel,
etz = slip angle of the rear wheel,
1= wheel base of the vehicle,
R = radius of the turn,
K = centripetal force applied to the vehicle,
m = vehicle mass,
ay = lateral acceleration of vehicle, and
V= vehicle forward velocity.
These equations form the basis for the construction of handling curves as defined in Ref 8.
The latter equation can be rewritten to yield the equation for lines of constant velocity
Chassis and body support
(topview)
11
! ....
i2
!
mainchassis
Suspension beam
(1-1 and 2-2) l /
(frontview) /
chassisbeam torsional/lateral
joint~ front./rear
L,,
= rotations constrained I = displacement constrained
= rotations free I = displacement free
= torsional spring t = linear spring
I = massless flexible element ~ = mass (inertia)
FIG. 2--Chassis components of the complex truck model.
K V2 l
mg gl R
In the handling curves in this paper, the dimensionless vehicle lateral acceleration, a~lg, is
plotted against the difference in axle slip angle, cr - oq. In this representation, negative x-
14 VEHICL~ROADINTERACTION
values indicate understeer, and positive x-values indicate oversteer. Vehicle-handling effects
in the simulation runs performed will be explained from handling curves.
Simulation Runs Performed

Both models have been employed in a number of simulation runs. All three simulation
types are included in an analysis macro. In general, during research performed in BAMMS,
most time is used in developing a set of dedicated automated analysis macros. Ideally, these
macros will allow the complete analysis in one day. This allows the researcher to put maximum
focus on possible model bugs or parameter discrepancies before performing actual "produc-
tion runs."
Speed Sweep While Running Straight Ahead--A number of runs combined in a "while loop"
where the lateral acceleration is set to zero and the forward velocity of the vehicle is linearly
increased from 10 to 50 m/s in 100 s. The end speed approximately corresponds to the
maximum speed of the race truck. The simulation is interrupted at 40 intervals to calculate
the eigenvalues of the model. Because of the acceleration, the models are not in a steady-
state situation but the deviations are considered small. This set of simulation runs is performed
to determine root loci quantifying stability of a straight running vehicle as a function of the
forward speed.
Spiral Curve at Fixed Speed--This is a variation of the steer angle where the lateral
acceleration of the vehicle is linearly increased from zero to 0.6 g in 100 s. The controller
for the throttle position is activated to obtain a constant forward velocity, and the steering
wheel controller is used to follow the prescribed desired lateral acceleration. The resulting
slip angles of the front and rear axles and the lateral acceleration are determined and stored.
These variables are used to generate handling curves for the vehicle. Again, the model will
not exactly be in a steady-state situation, but the deviations are considered small.
Speed Sweep at Constant Path Curvature--In this simulation, a lateral acceleration sweep
up to 0.6 g is performed by increasing the forward speed by 0.1 rn/s 2 while negotiating a
curve with a constant radius. Again, the resulting slip angles of the front and rear axles and
the lateral acceleration are determined and stored to generate the handling curves.
Parameter Variations
Simulations have been run with a number of variations of the model parameters to study
some aspects of the handling and stability of the truck at hand. The parameters that have been
varied in the simulation runs are:
Chassis Stiffness--A change in stiffness of the chassis will change the roll stiffness distribu-
tion of the vehicle and therefore influence the handling characteristics.
Stiffness of the Roll Stabilizers--A relative increase of the roll stiffness at the front/rear
will enforce understeer/oversteer as a consequence of the increased load transfer at the front/
rear axle.
Damping of the Differential--This parameter determines whether the differential is locked

or unlocked. When negotiating a curve, a locked differential will keep the rotational speeds
of the rear wheels equal, which will generate a change in longitudinal slip characteristics. This
will introduce a dependence of the handling curve on the path curvature. The double rear tires
will have a similar effect on the handling curve.
Simulation Results
Comparison of Simple and Complex Models--In Fig. 3, handling curves are shown for a
spiral run at 25 rn/s for the simple and the complex model. The figure shows that the handling
behavior at small lateral accelerations is identical for both models. At higher lateral accelera-
tions, the simple model tends more to the oversteer direction. This can be explained by the
more accurate method of modeling the steering system in the complex model. Also, the
geometric effects caused by the deformation of the chassis which are considered in the complex
model will cause a difference in handling behavior.
(1) simple model and (2) complex model

a,/g, lateral acceleration
0.7
0.525
I
I
0.35 -q
I
I
I
I
I
0.175 --t-
I I
I I
I I
I ~J
I
O.
' I
-.4 -.2 O~ 0.2 0.4
2-o, (~
FIG. 3--Handling curves for the simple and the complex model for a 25-m/s steer sweep.
Increased Frame Stiffness--In Fig. 4, the handling curve for the simple model with normal
frame stiffness is compared to the model with a doubled frame torsional stiffness. The curves
are calculated during a spiral run at a speed of 10 m/s. As the roll stiffness of the front
suspension is considerably larger than that of the rear suspension, an increase in frame stiffness
will mainly lead to an effective increase of the roll stiffness of the rear suspension. This has
no noticeable influence on the handling at smaller lateral accelerations. At larger lateral
acceleration values, the vehicle shows less understeer due to the increase in frame stiffness.
Roll Stabilizer Stiffness--The stiffness of the rear roll stabilizer of the complex model has
been multiplied by 1.3. The resulting handling curve shows no noticeable difference with the
standard vehicle. This was identical to the results of previous research on the handling of this
vehicle at the Delft University.
(1) Normal and (2) double frame stiffness

a~/g, l a t e r a l a c c e l e r a t i o n
0.7
1 2
0.525
0.35
I I
I I
I I
t I
0.175 5- ----t
.
-2. -1. 0, 1~ 2.
FIG. 4--Handling curves for the simple model with normal and doubled frame stiffness.
Speed Sweep Versus Spiral Run--The influence of path curvature on the handling curve
was investigated by comparing the methods described by the complex model and the simple
truck model. Simulations are performed with the simple model with single rear tires; slip
stiffness values of each single tire are equal to that of two double tires.
Speed Sweep at Constant Path Curvature

When the differential is locked, the handling curves show a horizontal offset which increases
with increasing path curvature. Because of the locked differential, an extra steer angle is
needed above the Ackermann steer angle (the steer angle following from kinematic motion
of the vehicle at low speeds) to reach the desired path curvature. This causes handling curves
to show a horizontal offset equal to the extra steer angle (the values on the abscissa of the
handling curve represent the steer angle minus the Ackermann steer angle) (see Fig. 5). In
this case, a locked differential combined with extra driving forces due to the forward acceleration
mainly generates oversteer due to the decreased lateral slip stiffness of the rear tires.
Effect of differential lock and curve radius

a~/g, l a t e r a l acceleration
0.7
unlocked
0.525
I
I
I
I
I
0.35
7" - - -
I
I
I
I ~ locked, R = 2 0 0 m
I ~ locked., R = 1 0 0 m
0.175
I locked, R = 5 0 m
I
"'~..
I
I
.
-2. -1. O. 1. 2.
.2-m( ~
FIG. 5--Handling curves for the simple model with single rear tires at constant curve radius.
The fourth curve in Fig. 5 shows the result for the vehicle with unlocked differential.
Spiral Run at Constant Speed

In this case, the simulation starts with the vehicle running straight ahead. Therefore, the
handling curve starts at zero lateral acceleration and at zero slip angle difference between
front and rear axle. Smaller drive forces at rear wheels are required at zero longitudinal
acceleration. Thus, the difference between drive forces left and right will increase, causing an
increase in the moment about a vertical axis. This moment increases at increasing path curvature
(i.e., at lower speeds) which is shown in the right picture of Fig. 6. Handling curves for an
unlocked differential are shown in the left picture of Fig. 6.
In another set of runs, an unlocked differential is combined with double tires at the rear
axle. At higher speed values, the effect of double tires is similar (though smaller) to that of
a locked differential, as shown in Fig. 7.
Mode Analysis and Root Loci

The EIGENFREQ analysis command calculates the eigenfrequencies of the modeled system.
Numerical perturbation is performed to generate the system matrix at any given model position.
Sequences of eigenfrequency calls can be used to generate root loci with every possible model
parameter as the ordinate. The root loci of the simple model will be discussed in detail. All
loci in Fig. 9 contain 40 points, each representing a certain speed value.
U n l o c k e d differential L o c k e d differential
%lg, lateral a c c e l e r a t i o n ay/g, lateral a c c e l e r a t i o n
n.Kg~ . . . . . . - - -~k-~'t~J . . . . . . . I _ _
10 s ~ i
! 20 m/s lO m h
! ~ / ~ _~_ 30 m ) s tl 20 m / s
0.35 . . . . . . . . . . . . . . ]" -- 30 m / s
I llll I
iI
I
I
i
I
l
I
'
I
I
I
'
t
I
I
0.176 . . . . . . . . . . . . "I" . . . . . . . l-
I I
I I I
I I I
I I I
I
0. I
-2. -1. 0. -2. -1. 0. 1.
~2-~, (o) c,2-0~1 (o)
FIG. ~-Handling curves for the simple model with single rear tires at constant forward speed.
Handling curves with double rear tyres

%/g, l a t e r a l a c c e l e r a t i o n
.7 _
I I
f I
0.525 I _ ~ 10 m/s
m/s
m/s
I
0.35 "3" "l

I
I
{ I
I I
I I
0.175 -+ -I
I I
I
I I
I
o. I
-2. -1. O. 1. 2.
a2-c~1 ( ~)
FIG. 7--Handling curves for the simple model with double rear tires at constant forward speed.
The numeric values of the first six vehicle modes are listed in Table 1. A selection of the
modes is visualized in Fig. 8, The effect of removing differential lock on the vehicle stability
is visualized in the right portion Fig. 9 and shows that yaw mode (6) of the model becomes
oscillatory at higher speeds. A relatively small effect of longitudinal velocity on mode damping
TABLE l--Relevant modes of the simple race truck model at 30 mls.

Mode Frequency, Hz Damping ( - ) Explanation
1 10.0 0.35 Rear axle roll

2 8,25 0.28 Front axle roll
3 8,2 0.30 Pitch about center of gravitational
4 3.64 0.13 Pitch about rear axle
5 1.46 0.03 Chassis roll
6 (-34,12) (Real mode) Yaw motion
FIG. 8--A selection of the mode shapes of the simple truck model
can be seen in the loci of Modes 2 and 3. In general, modes related to tire lateral slip will
demonstrate a decreasing damping at increasing speed. Root loci for the complex model are
considered beyond the scope of this paper.
Conclusions
The following conclusions are drawn in the race truck analysis:
1. At larger lateral accelerations the handling behavior of the simple and the complex
model shows a difference. Both models show initial understeer changing to oversteer
at higher lateral accelerations.
2. In the case of a vehicle with a locked differential and double rear tires, the handling
curves depend strongly on the path curvature and the presence of driving forces. There-
v~
ir I
9 ,ol
0!
I. . . . . . . . . . . . T . . . . .
I
!
I
I
!
. . . . . 0
E
. . . . . . 't
-I . . . . . .
~ N I I
I !
I t~ ~ I
E
0
": I
I
II
I- I t
0 e ~ Q
. . . . . . !. . . . . . r . . . . . T . . . . .
I
I I I
I I I
I I I
I I I
I I I
o
. . . . . . I. . . . . . I- . . . . . + . . . . . I
! I I
I I I
I N I I
I ~ I I
"l ~ f , l
Im-i~
, I I i
fore, several different maneuvers (preferably with small driving forces) will have to be
studied to get a good impression of the handling of such a vehicle.
3. The effect of double rear tires on the handling is similar to that of a locked differential;
however, the effect of a locked differential is larger.
4. Doubling the frame stiffness will only have a noticeable influence on the handling at
larger lateral acceleration. It will decrease the amount of understeer.
5. The damping of the differential has influence on the stability of the vehicle when
running straight ahead. Damping of the yaw mode of the vehicle is decreased when
the differential lock is removed.
Stability Criteria for Truck Trailer Combinations

Introduction
This investigation deals with criteria for the determination of stability of truck trailer
combinations. The research has been performed to improve knowledge on stability of trucks
with a variable number of trailers. The root loci of truck models at different driving speeds
have been produced and analyzed. The following aspects have been investigated:
1. The need for three-dimensional modeling for determining straight-ahead stability. For
each configuration, both a two-dimensional and a three-dimensional model have been
made and root loci curves are compared. The need for three-dimensional modeling is
demonstrated by the influence of the height of the center of gravity on the stability of
the vehicle.
2. Special attention has been paid to the effect of steering the single axle of a trailer on
vehicle stability. Different values for the constant steering ratio of the mechanical
coupling between the relative angle of two vehicle bodies and the axle steering angle
have been applied.
Models Used
Three different vehicle combinations have been modeled. The most relevant eigenmodes
of the models are shown in Fig. 10. All vehicle combinations are loaded to their gross vehicle
weight as set by the Dutch legislation. In each model a user-developed BAMMS macro was
used to model the axles. Each axle introduces two degrees of freedom: vertical axle displacement
and axle roll. All main vehicle bodies are modeled as rigid bodies.
The following vehicles are modeled:
Configuration 1--A 19-ton truck pulling a 27-ton three-axle full trailer. Figure 10a shows
the five relevant modes for the lateral motion of the combination. The modes range from
0.4 to 1.4 Hz. Each mode shows a unique combination of truck and trailer roll, yaw, and
lateral motion.
Configuration 2 - - A 7-ton tractor with a 36-ton three-axle semi-trailer. Figure 10b shows
the three relevant modes for the lateral stability of the combination. The frequency of Modes
I and II is approximately 0.4 Hz, and the frequency of Mode II! is approximately 1.15 Hz.
Configuration 3 - - A 7-ton tractor with a 22-ton single-axle (city) trailer. Figure 10c shows
the relevant modes for this combination. The modes are fairly comparable to those of Configura-
VERHEUL El" AL. ON BAMMS 23
b) T R U C K + c) T R U C K +
SEMI-TRAILER CITY-TRAILER
FIG. lO---Mode shapes of the relevant lateral modes of the three used truck models.
tion 2. Due to the difference in vehicle parameters, the loci of Modes I and III of the two
models have swapped places.
The kingpin connection between the main bodies of Configurations 2 and 3 is modeled as
a rigid joint allowing only rotations about y- and z-axis. The full trailer of Configuration 1
consists of a main vehicle body, a dolly for the front axle, and a triangular connection to the
motor vehicle. The dolly has been modeled as a rotor with only one degree of freedom (the
yaw angle of the dolly). The triangular arm can rotate around the lateral axis at the dolly and
is connected to the motor vehicle with a rigid point connection.
Parameter Variations and Simulation Results

For Configurations 2 and 3, the height of the trailer center of gravity (cg) is set at 2.02,
2.42, and at 2.82 m. For Configuration 1, the height of the motor vehicle cg is set at 1.52,
1.72, and at 1,92 m. Simultaneously, the height of the trailer cg is set at 1.78, 2.18, and 2.58 m.
At a speed increasing from 1 to 30 m/s, the eigenvalues of the combinations are calculated
in a straight running steady-state position. For each set of parameters, relevant root loci are
plotted in Fig. 11. The stability of the combinations can be determined from the distance of
each locus to the imaginary axis.
The results of each combination are listed columnwise with the top pictures representing
the lowest positions of cg. Arrows in the loci represent the direction of increasing vehicle speed.
Figure 11 leads to the following conclusions:
1. For most modes, the damping decreases at increasing speed.

T R U C K § FULL ~ TRUCK + S E M I T R A I L E R T R U C K § CITY T R A g . ~

10.
i I I I
C.O.O, Position
| I ' ; I I 3
7.5 ............ ~:.... .i_/'__3_ .... ! ..... i- .... -!-- - Low
"~ .... i .... ! ..... ",
;
'L_
l
l
5. ._4 .... ! ..... ---.
I I ...... ' ..... '] .... T ..... i i i
It Ii It :L.., i i
2.3
""-I .... I"- Tf-~"- t~ 4 ,.t- . . . . . : ._L~...
,, ,. /./,
0.
10. i i t
~
t t i
I i I
i i i
I I i i ;
7.5
..... "; .....
i
i .....
i
,--'~-"
|
. . . . . ii . . . . . ";, . . . . . . i;......
; ; ; ...... ! ..... T .... V II iI I
I~ If3
i i i
5. i i , i :
..... " ..... ~ ...... r'<=
i i i
I I I
1 ; I
2.5 i i i ....~ . . !
.... 4 ..... ; ; l
0.
10.
i
i
l i i I
i
i
'
i I I
7.5 ..... k..... J ...... i. . . . .
..... ~~. . . . J 1. .7. . . ~
...... ...... i ..... i ...... i......
5. ..... :_.. ;
i
i
, i
i
, ~ i
i
: ,
..... r ..... i ..... r
2.5 i 1
i
i
I
i
I
i
P
l
I
,/'!
I I
,,'~-~'~
0. IJ ~ i l
i
i
I
i
i
I
VI, ..... i,
I 9 1 . I
. . . .
, I -
2
-10. -'/.5 -5. -2.5 0. -tO. -7.5 -5. -2,5 O. -1o. :1.5 -5. -2.5 0.
R e a l part o f e i g e n v a l u e R e a l part o f e i g e n v a l u e Real pan of eigenvalee
FIG. 11--Root loci for variation of the position of the center of gravity for all three models.
2. The real part of the eigenvalues seems to indicate that most curves reach an asymptotic
value. At higher velocities the damping of these modes will change less drastically.
This becomes more apparent when the real parts of the eigenvalues are plotted against
the vehicle speed.
3. The changes in the root loci, while varying the height of cg, indicate that oscillations
in the systems depend very strongly on this parameter. This leads to the conclusion
that use of two-dimensional models is not allowed when accurate determination of the
stability is desired.
The loci in Fig. 12 are obtained from a simulation where the steering ratio of the trailer
axle in the city trailer model was varied. The ratio, K, denotes the steering angle of the trailer
axle relative to the articulation angle between truck and trailer body. It was varied from a
value of zero to a value of two. The figure shows that an increasing value of the steering ratio
will increase the damping of Mode I and will decrease the damping of Mode III. The overall
conclusion is that steering will decrease the stability of the vehicle.
TRUCK + CITY 33~AILE~

Variation of the steering ratio K (0,..,2)
I0.
7.5
5.
"1"
I i 3
I I
I I
I I
2.5 -t- -4
I I I
I I I I 2
I I t
I I I
I I I
O.
-10. -7.5 -5. -2.5 0.
Real part of eigenvalue

FIG. 12--Root loci for variation of the steering ratio for the rear trailer axle.
Conclusions
1. Use of two-dimensional models will not give results accurate enough for a detailed
study of the stability of the vehicle combinations. The reason for this is that the roll of the
vehicle is strongly present in the relevant modes.
2. A first indication for the use of mechanically steered trailers is that an increasing
amount of steering will destabilize the vehicle combination.
References
[1] Pacejka,H. B., "Modeling Complex Vehicle Systems using Bond Graphs," Journal of the Franklin
Institute, Vol. 319, No. 1/2, 1985, pp. 67-83.
[2] Margolis,D. L., "Bond Graphs, Normal Modes and Vehicular Structures," Vehicle System Dynamics,
Vol. 7, 1978, pp. 49-63.
[3] Nikravesh, P. E., Computer-Aided Ancdysis of Mechanical Systems, Prentice-Hall International,
Englewood Cliffs, NJ.
[4] Baumgarte, J., "Stabilization of Constraints and Integrals of Motion in Dynamical Systems," Compu-
tational Methods in Applied Mechanics, Vol. 1, pp. 1-16.
[5] Shampine, L. F. and Gordon, M. K., Computer Solutions of Initial Value Problems, Freeman, 1975.
[6] Fuehrer, C., "Numerical Integration Methods in Vehicle Dynamics Simulation," Proceedings, 3rd
ICTS Seminar on Advanced Vehicle System Dynamics, Amalfi, May 1986, pp. 329-345.
[7] Bakker, E., Pacejka, H. B., and Lidner, L., "A New Tire Model with an Application in Vehicle
Dynamics Studies," SAE Paper 890087, Society of Automotive Engineers, Warrendale, PA, Janu-
ary 1989.
[8] Pacejka, H. B., "Simplified Analysis of Steady-State Turning Behavior of Motor Vehicles," Vehicle
System Dynamics, Vol. 2, 1973, pp. 185-204.
D a v i d J. C o l e 1 a n d D a v i d C e b o n 1
Predicting Vertical Dynamic Tire Forces of

Heavy Trucks
REFERENCE: Cole, D. J. and Cebon, D., "Predicting Vertical Dynamic Tire Forces of
Heavy Trucks," Vehicle-Road Interaction, ASTM STP 1225, B. T. Kulakowski, Ed., American
Society for Testing and Materials, Philadelphia, 1994, pp. 27-35.
ABSTRACT: Two models of a four-axle articulated vehicle are described; one is two-dimen-
sional, and the other is three-dimensional. Attention is given to modeling the tandem-axle, leaf-
sprung trailer suspension. Vertical dynamic tire forces measured from tests on the full-size vehicle
are compared with predicted forces from the three-dimensional model, and good agreement is
observed. The predictions of the two- and three-dimensional models are compared, and the
conditions under which a two-dimensional model can satisfactorily simulate vehicle behavior
are examined.
KEY WORDS: vehicle simulation, dynamic tire forces, articulated vehicle, leaf spring suspen-
sions, heavy trucks
Previous work on heavy vehicle vibration simulation was extensively reviewed by EIMadany
et al. [1] in 1979 and by Cebon [2] in 1985. Cebon noted that very few heavy vehicle
simulations had included experimental validation. Cebon [3] developed and validated two-
and three-dimensional nonlinear models of a three-axle rigid fuel tanker, but the test conditions
(speed and road roughness) were not representative of typical freight operating conditions.
Several simulation studies have been reported since the review by Cebon [4-10], but only
two of these have included comparison with measured responses. Hu [8] described a simulation
model used for the study of articulated vehicle ride dynamics. The model was two-dimensional
and incorporated nonlinear spring characteristics and a first order model of frame bending. A
limited comparison of simulated responses with measurements indicated good agreement.
Sakuma et al. [10] created two-dimensional nonlinear models of rigid vehicles using a finite
element program, and flexible frame modes were included. Agreement between measured and
simulated vertical acceleration of the cab was good.
In summary, many heavy vehicle simulations have been developed, but only a small number
have been compared with experimental measurements, and no articulated vehicle simulations
have been convincingly validated. Most of the simulations have considered the pitch and
bounce motions of rigid-framed vehicles only. Consequently, there is no definite opinion on
whether roll motions and frame flexibility are significant in contributing to dynamic tire forces.
This paper describes a heavy vehicle simulation and the results of a validation study. The
test vehicle was a typical UK articulated lorry, with four axles and a maximum gross mass
of 32.5 metric tons. The tractor unit had two axles, each fitted with multi-leaf springs and
hydraulic dampers. The trailer was fitted with a wide-spread, tandem-axle group, with mono-
Research associate and lecturer, respectively, Cambridge University Engineering Dept., Trumpington
St., Cambridge, CB2 1 PZ, United Kingdom.
27
leaf springs and a load-equalizing mechanism, but no hydraulic dampers. The axles were
instrumented to measure vertical dynamic tire forces by means of strain gages and accelerome-
ters on the axle [16]. The vehicle and test facilities were provided by the Transport and Road
Research Laboratory (TRRL). The vehicle was tested under realistic conditions of speed and
road surface roughness.
Simulation
Three-Dimensional Mathematical Model
The three-dimensional model is shown in Fig. 1. The model has 21 degrees of freedom and
includes the important sources of nonlinearity. The hysteresis of the leaf spring elements is
modeled using the method of Fancher et al. [11,12]. The dampers on the tractor axles are
modeled with different rates in bump and rebound. The tires have the ability to lose contact
with the road, although the input conditions of the test did not cause this to occur. The
envelopment of short wavelength irregularities by the tire is accounted for by simple tire
contact patch averaging [3,13]. The sprung masses of the tractor and trailer are considered
rigid. The two-dimensional model is similar, but includes only the eleven pitch and bounce
degrees of freedom.
Inertia properties of the sprung and unsprung masses were estimated using the measured
static tire forces and dimensions of the test vehicle. Tire stiffness values were derived from
the manufacturer's force-deflection data. Tractor suspension properties were provided by the
manufacturer. The leaf spring data consisted only of values for linear (large deflection) stiffness.
The hysteresis characteristics were estimated from results of tests on springs of similar type
[11,12,14,15]. Particular attention was given to modeling the trailer suspension, which is
described in the next section. The parameter values for the three-dimensional model can be
found in Ref 16.
Four-Spring Trailer Suspension Model
The four-spring trailer suspension consisted of two mono-leaf springs attached to each axle.
The springs had slipper mounts at each end, and longitudinal location of each axle was provided
by two radius arms inclined at an angle to the horizontal (Fig. 2). The force-deflection
characteristic of a single trailer leaf spring tested on a laboratory rig (with slipper end mounts,
but without radius arms) was provided by the suspension manufacturer and used to determine
suitable values for the parameters of the leaf spring model. The manufacturer's data and the
fitted characteristic are shown in Fig. 3.
The radius arms were found to have a significant effect on the behavior of the suspension.
When an axle displaces upwards, the inclined radius arm causes the axle to move rearwards.
This movement is resisted by friction forces at the slipper ends of the leaf spring. The friction
forces are reacted by force in the radius arm, which, because it is inclined to the horizontal,
influences the vertical force on the axle. To include this effect in the simulation, additional
force elements were included in the model (for clarity these are not shown in Fig. 1). For
details see Ref 17.
Published measurements from tests performed by TRRL on the same test trailer [18] were
used to determine values for the parameters of the tandem suspension model. In the TRRL
tests, the laden trailer was supported at the fifth wheel by a crane, and weighpads were placed
under each wheel to measure the tire forces. The fifth wheel was raised and lowered through
several cycles to give a trailer pitch angle amplitude of 0.8 ~ The force amplitude at the tires
was measured to be 10.3 kN for the leading axle and 7.6 kN for the trailing axle. The maximum
COLE AND CEBON ON VERTICAL DYNAMIC TIRE FORCES 29
a3
al~ a2 bl b2
i 0z2
/-' I i
| 0 x |
Oyl
2ca m
section A - A '
FIG. l--Three-dimensional whole vehicle model For clarity, the force elements necessary for
simulating the trailer suspension radius arms are not shown. A double-headed arrow indicates
angular motion or torque.
tire force on each axle occurred when the corresponding axle-to-trailer frame distance was
a minimum.
Figure 4 compares the TRRL measurements with the results of the tandem suspension
simulation. The agreement is good, and the simulation displays most of the important features
of the experimental results.
Starting at the point of minimum tire force on the leading axle (Point A on Fig. 4), the tire
force increases at a rate controlled mainly by the "wind-up" stiffnesses of both leading axle
chassis f r a m ~ slipper ends
radius arm levelling beam monc~leaf spring

FIG. 2--Schematic view of one side of the trailer suspension.
70 " ' ' 1 . . . . I . . . . I . . . . I . . . . I . . . . I . . . .
manufacturer's data
60 fitted characteristic .'
50
Z
~_r
2O
10
.... -t'...-':", I .... J .... I .... I .... i..,

0
0 5 10 15 20 25 30 35
D I S P L A C E M E N T / ram
FIG. 3--Measured and simulated behavior of a single trailer leaf spring.
and trailing axle springs. At Point B, sliding begins at the slipper ends of the now lesser-laden
trailing axle spring. The leading axle tire force then increases at a lower rate because the
trailing axle radius arm cannot exert any further force. On reversing the tire displacements at
Point C, sliding at the ends of the trailing axle spring stops, and the tire forces again change
at a rate controlled by the wind-up stiffnesses of both springs. At Point D, sliding begins at
the ends of the leading axle spring.
The corresponding measured characteristic in Fig. 4 does not show the distinct changes in
stiffness predicted by the simulation; this is probably because in practice the transition from
COLE AND CEBON ON VERTICAL DYNAMICTIRE FORCES 31
45~ .... I .... i .... I ....
measured leading
axle tyre f o r c e
0 0 0
measured trailing
axle tyre f o r c e
simulated t y r e force
" -,o . . . . o . . . . ,o . . . .
D I S P L A C E M E N T / mrn
FIG. 4--Measured and simulated trailer pitch test. The measurements are from tests performed
by TRRL. The horizontal axis is the road displacement seen by the leading axle in the simulation.
sticking to sliding occurs gradually. The difference in force amplitudes for the leading and
trailing axles measured in the tests could not be simulated.
Validation
Three-Dimensional Model
Response time histories for the three-dimensional model were calculated by numerical
integration using the measured test lane wheelpath profiles as the input. The IRI values of the
nearside and offside wheelpaths were 4.25 and 4.35 m/kin. Figures 5a and 5b show measured
and simulated vertical tire force histories (at the tire/road interface) for two of the nearside
wheels of the fully laden vehicle traveling at 22 m/s. In these graphs, the horizontal axes show
the position of each axle along the test section. Slight variations in test vehicle speed mean
that the measured and simulated force histories are not always exactly synchronized: the
maximum discrepancy in longitudinal position apparent from these graphs is about -+0.5 m.
Figures 5c and 5d show measured and simulated tire force spectral densities for the same
two wheels. Overall the agreement between measurement and simulation is very good. T h e r e
are some small discrepancies in the region of the sprung mass modes (1 to 4 Hz). In the region
of the unsprung mass modes (10 to 15 Hz), the drive axle tire forces agree quite closely, but
the agreement is not so good for the trailer axle tire force. This result suggests that further
refinement of the trailer suspension model is needed to simulate its complex behavior accurately.
Comparison of measured and simulated responses for other conditions of speed and payload
showed similar agreement (see Ref 16).
Some of the discrepancies may be partly due to the assumption of rigid sprung masses. The
tractor sprung mass was treated as one rigid mass. In practice the cab and engine are flexibly
mounted to the chassis and have modes of vibration in the frequency range of interest. The
trailer sprung mass was also considered to be rigid, but the measured trailer bending acceleration
showed a lightly damped bending resonance at 7 Hz. Note that the measured tire force spectral
7O I I I
6O
-~ 50
40
~3o !.j
~2o
lO
, I , , , I , , , I , , , ,
(~6'o 170 180 190 200
DISTANCE / m
(a) Drive axle ns tyre force history.
70 I " ' I I
60
~ 5o
~ 40
~ 30
0 20
1(1
I I i i ,
"Ion6" . . . . I
170 . . . . I
180 ..,
190 200
DISTANCE / m
(b) 'l~Ml!ng trailer axle ns tyre force history.
20.0 25.0 .... , . . . . , .... , . . . . , ....
' ~ . 15.0
20.0 !
15.0
I~ 10,0
r~ 1o.o
0 0 ~.0 ~
0.0' < . . . . . .
5 10 15 20 25 0"00" 5 10 15 20 25
FREQUENCY / Hz F ~ q E E N C Y / Hz
(c) Drive axle as tyre force (d) Trailing trailer axle ns tyre f o r c e
spectral density. spectral density.
FIG. 5--Measured and simulated nearside tire force responses--fully laden, 22 m/s: _ _
measured, . . . . . . . simulated.
densities do not show significant response at 7 Hz. This is thought to be because the nodes
of the bending mode are close to the suspension mounting points.
An additional source of error may be the tractor suspension parameters, which were derived
entirely from the manufacturer's data and published measurements on similar components.
The simple bilinear characteristic used for the damper simulation may also be inaccurate.
Detailed studies of automotive damper behavior [19,20] have revealed highly nonlinear and
hysteretic behavior.
Experimental error and data processing may contribute to some of the discrepancy between
measurement and simulation. It is estimated that the vehicle speed varied by up to -+5% during
each test run. The error in the tire forces measured by the lorry instrumentation was estimated
to be 1.5% rms [16].
It is difficult to determine how each of the possible sources of error contributes to the
overall discrepancy. However, none of the experimental or numerical errors are thought to be
excessive, and therefore the results of this study indicate the likely agreement that can be
obtained when using a nonlinear rigid body model with accurate road profile input and good
estimates of the vehicle properties. Overall, the agreement is considerably better than that
achieved in previous heavy vehicle simulation validation studies [1-3,8,10].
Comparison of Two- and Three-Dimensional Models
Response time histories for the two-dimensional model were calculated using the nearside
wheel track of the measured lane profile as input. Figure.6 compares nearside tire force
histories and spectral densities of the two- and three-dimensional models (fully laden, 22 m/
s). Similar results were obtained for the offside wheel track.
Agreement between the two models is good in the region of the sprung mass modes. Detailed
analysis of the measured left- and right-hand wheel track profiles of the test lane shows that,
at the test speed of 22 m/s, the long wavelength undulations that excite the sprung mass modes
(1 to 4 Hz) are in phase (left to right) [17]. This means that the sprung mass roll modes are
not excited significantly, and therefore the two models are expected to give similar results.
Analysis of the measured tire forces confirms that the response of sprung mass roll modes is
small [17].
There is significant difference between the tire force responses of the two models in the
region of the unsprung mass modes (10 to 20 Hz), particularly for the tractor suspension. The
shorter wavelength undulations that excite unsprung mass modes are uncorrelated left to right;
therefore, the bounce and roll modes are both excited [17]. If the unsprung mass roll modes
contribute significantly to the response, then there is likely to be a difference between tire
forces predicted by the two- and three-dimensional models in this frequency range (Fig. 6).
In summary, a two-dimensional model may be satisfactory for predicting the tire forces of
a heavy vehicle if: (1) the vehicle speed is high enough to prevent excitation of sprung mass
roll modes, and (2) the contribution of the unsprung mass roll modes to the tire forces is small.
Conclusions
1. The measured tire force responses of the vehicle were compared in time and frequency
domains with predictions from a three-dimensional nonlinear model. Agreement was generally
good and better than that achieved in previous studies.
2. The main sources of error in the simulation were thought to be the trailer suspension
model properties and the assumption of rigid sprung masses. The measured trailer bending
resonance did not significantly affect the measured tire forces.
3. A two-dimensional model should be satisfactory for predicting the tire forces of typical
leaf-sprung articulated vehicles with well-damped unsprung mass modes operating under
typical conditions of speed and road roughness.
70 ; . . . . i " " ' ' I . . . . I . . . .
~ 40
lo
i i i i ~ i i , I , , , , I
~6'0 170 180 190 200
DISTANCE / m
(a) Drive axle ns tyre force history.
70 I I !
60
~5o
~4o
~3o
o20
lO
i , i I
~6' . . . . I . . . . , ,
0 170 180 190 200
DISTANCE / m
(b) Trailing trailer axle ns tyre force history.
15.0 30.0
•:•25.0
e~
.,.~ 10.0 .~20,0
I ~ 15.0
t~
5.0 ~ I0.0
Q ~ 5.0
0.0
5 10 15 20 25 0"0'0 5 10 15 20 25
FREQUENCY / Hz FREQUENCY [ tIz
(e) Drive axle ns tyre force (d) Trailing trailer axle ns tyre force
spectral density. spectral density.
FIG. 6---Tire force responses from two- and three-dimensional models--fully laden, 22 m/s; _ _
three-dimensional, two-dimensional.
Acknowledgment
The authors are grateful to the Director of the Transport and Road Research Laboratory
and members of the Vehicles and Environment Division for assistance with the vehicle testing
program described in this paper. Any views expressed are those of the authors and not
necessarily those of the Department of Transport or the Transport and Road Research Labora-
tory.
References
[1] E1Madany, M. M., Dokainish, M. A., and Allan, A. B., "Ride Dynamics of Articulated Vehicles--A
Literature Survey," Vehicle System Dynamics, Vol. 8, No. 4, 1979, pp. 287-316.
[2] Cebon, D., "An Investigation of the Dynamic Interaction Between Wheeled Vehicles and Road
Surfaces," Ph.D. dissertation, University of Cambridge, 1985.
[3] Cebon, D., "Heavy Vehicle Vibration--A Case Study," Proceedings, 9th IAVSD Symposium on
the Dynamics of Vehicles on Roads and on Tracks, Linkoping, Sweden, Swets and Zeitlinger,
Lisse; The Netherlands, 1985, pp. 30--45.
[4] ElMadany, M. M., "Nonlinear Ride Analysis of Heavy Trucks," Computers and Structures, Vol.
25, No. 1, 1987, pp. 69-82.
[5] E1Madany, M. M., "An Analytical Investigation of Isolation Systems for Cab Ride," Computers
and Structures, Vol. 27, No. 5, 1987, pp. 679-688.
[6] Gadala, M. E., E1Madany, M. M., and Gadala, M. S., "Finite Element and Analytical Modeling
of a Tractor-Semitrailer Vehicle," Computers and Structures, Vol. 23, No. 6, 1986, pp. 831-836.
[7] Hedrick, K. and Yi, K., "The Effect of Alternative Heavy Truck Suspension on Flexible Pavement
Response," Proceedings, 2nd International Symposium on Heavy Vehicle Weights and Dimensions,
Kelowna, British Columbia, June 1989.
[8] Hu, G., "Simulation of Heavy Truck Ride Using a Desktop Computer," SAE 871557, Society of
Automotive Engineers, Warrendale, PA, 1987.
[9] O'Connell, S., Abbo, E., and Hedrick, K., "Analyses of Moving Dynamic Loads on Highway
Pavements, Part I: Vehicle Response," Proceedings, International Symposium on Heavy Vehicle
Weights and Dimensions, Kelowna, British Columbia, 1986, pp. 363-380.
[10] Sakuma, N., Kadono, I., Dohi, M., and Nakai, H., "Heavy Duty Truck Ride Comfort Analysis by
Computer Simulation," International Journal of Vehicle Design, 1986, pp. 279-290.
[11] Cebon, D., "Simulation of the Response of Leaf Springs to Broad Band Random Excitation,"
Vehicle System Dynamics, Vol. 15, 1986, pp. 375-392.
[12] Fancher, P. S., Ervin, R. D., MacAdam, C. C., and Winkler, C. B., "Measurement and Representation
of the Mechanical Properties of Truck Leaf Springs," SAE 800905, Society of Automotive Engineers,
Warrendale, PA, 1980.
[13] Captain, K. M., Boghani, A. B., and Wormley, D. N., "Analytical Tire Models for Dynamic Vehicle
Simulation," Vehicle System Dynamics, Vol. 8, No. 1, 1979, pp. 1-32.
[14] Winkler, C. B., "Measurement of Inertial Properties and Suspension Parameters of Heavy Highway
Vehicles," SAE 730182, Society of Automotive Engineers, Warrendale, PA, 1973, pp. 720-732.
[15] Winkler, C. B. and Hagan, M., "A Test Facility for the Measurement of Heavy Vehicle Suspension
Parameters," SAE 800906, Society of Automotive Engineers, Warrendale, PA, 1980, pp. 2754-2782.
[16] Cole, D. J., "Measurement and Analysis of Dynamic Tyre Forces Generated by Lorries," Ph.D.
dissertation, University of Cambridge, 1990.
[17] Cole, D. J. and Cebon, D., "Validation of a Vehicle Vibration Simulation," VehicleSystem Dynamics,
Vol. 21, No. 4, 1992, pp. 197-223.
[18] Simmons, I. C. P. and Wood, J. G. B., 'q'he Equalization of Multiaxle Bogies Fitted to Commercial
Vehicles," TRRL Research Report 277, 1990.
[19] Hall, B. B. and Gill, K. E, "Performance of a Telescopic Dual Tube Automotive Damper and the
Implications for Vehicle Ride Prediction," Proceedings, Institution of Mechanical Engineers, Vol.
200, No. D2, 1986.
[20] Besinger, E H., "The Performance of Passive and Semi-active Suspensions for Heavy Lorries,"
Ph.D. dissertation, University of Cambridge, 1992.
M a t t i Huhtala, l M a u r i Vesimakg 2 a n d P e k k a H a l o n e n I
Computer Simulation of Road-Vehicle

Dynamic Interaction Forces of Three-
and Four-Axle Trucks
REFERENCE: Huhtala, M., Vesimaki, M., and Halonen, E, "Computer Simulation of Road-
Vehicle Dynamic Interaction Forces of Three- and Four-Axle Trucks," Vehicle-Road Interac-
tion, ASTM STP 1225, B. T. Kulakowski, Ed., American Society for Testing and Materials,
Philadelphia, 1994, pp. 36-51.
ABSTRACT: The road-vehicle interaction forces were studied with two truck models, one
with four axles (twelve tires) and the other with three axles (ten tires). The truck models were
created using the multibody system analyzing program (ADAMS). The input data for the models
are based on real values derived from the truck manufacturer. Two modifications of suspensions
were made, one with three dampers. The suspensions were multileaf, parabolic, and air suspen-
sions.
The road profiles were measured by VTr. The model computes instantaneous locations,
speeds, and road contact points of tires. Then it computes vertical and longitudinal forces acting
on tire-road contact surfaces as a function of tire stiffness, damping, and longitudinal slips.
These forces are then used to study repeatability of dynamic axle loads along the road profile.
KEY WORDS: simulation, axle loads, dynamic behavior, vehicle pavement interaction
The effect of dynamic loads on selected test pavement sections were studied using computer
simulations. An original truck model was created in cooperation with the manufacturer, Oy
Sisu-Auto Ab, which specializes in heavy transport equipment. The truck model was first used
in ride comfort analysis. This presentation concentrates on dynamic axle loads, especially by
varying the number of axles, spring types, damping, and speed of the truck. The repeatability
of road-vehicle interaction forces on different road profiles was also examined.
The truck models were created using the ADAMS (Automatic Dynamic Analysis of Mechani-
cal Systems) simulation program. In this paper, a brief introduction of ADAMS is given, the
truck model is described, and the road profiles used in this study are explained. The simulation
and the results are also described.
ADAMS
ADAMS is a commercially available multibody system analysis software. It is the most

widely used software package of its kind. ADAMS is used in a large variety of industrial
applications throughout the world. Some examples of its use in vehicle dynamics are presented
in Refs 1 and 2.
Technical Research Center of Finland (VTT), Road, Traffic and Geotechnical Laboratory, Espoo, Fin-
land.
2MBS-Models, Inc., Espoo, Finland.
36
Copyright 9 1994 by ASTM lntcrnational www.astm.org
HUHTALA ET AL. ON COMPUTER SIMULATION 37
ADAMS offers an extendable library of modeling elements. The elements are very general,
so they do not restrict the type of mechanism being modeled. In fact, they are the basic building
blocks of dynamics, i.e. forces, parts, joints. By selecting and combining these elements, the
user can build a dynamic model to the required level of detail.
Once the model is defined, the user can request static, quasi-static, linear, kinematic, dynamic,
or inverse dynamic analysis of the system. ADAMS automatically forms the equations of
motion and solves them in the time domain. The solution method employed in ADAMS has
proved to be general. It can even handle severe nonlinearities and impacts quite well.
With the aid of ADAMS, even complex mechanisms can be modeled straightforwardly. The
use of recognizable physical input values makes the models less prone to errors and easy to
modify. In fact, the modeling itself is easy, but finding out all the input data needed can be
a lot of work. A mathematical model is always a simplified representation of the actual system
being studied. The amount of input data is strongly related to the level of detail in the model.
One has to create an idealized model of the system and judge which parameters are important
and which can be neglected. For that reason, the models should always be easy to modify.
ADAMS solves the equations of motion using variable order and time step predictor-
corrector integration technique. The user sets the upper limit for time steps by defining the
number of output steps of the simulation. ADAMS may use smaller steps if it is necessary
for achieving required accuracy or in case of sudden change in the equations. ADAMS predicts
the solution one step forward in time and then corrects it with iteration. If the step is successful,
the next step is taken; otherwise, the step is rejected and a smaller time step is tried.
The Truck Model
Originally, a four-axle truck was studied because of its complex dynamic behavior (there
are also three-axle trucks available similar to the one chosen). Four-axle trucks are common
in Finland, especially in lumber transportation. The equipment on a truck varies greatly,
depending on its intended use. For the truck model, the equipment needed to carry logs was
chosen in order to represent typical use of such a truck in Finland.
The first two axles steer the truck, and there are single tires on them. There are dual tires
on the two axles of the bogie, so altogether the truck has twelve tires. The first axle is suspended
by parabolic leaf springs, the second by air springs, and, on the bogie, there are two multileaf
springs. The two rear axles are kinematically coupled by swing arms so that one multileaf
spring on each side of the bogie can suspend both axles. The second and fourth axles can be
lifted off the ground to save fuel and tires when the truck is driven without load.
The original four-axle truck model was created in cooperation with the manufacturer to be
used in studying ride comfort. In Fig. l, a side view of the model is shown. It consists of 35
inertial masses connected to each other with kinematical joints and nonlinear force elements.
Each joint reduces the amount of active degrees of freedom in the model. After adding the
kinematic couplings, there are still 32 left. Altogether the model forms a set of 1166 differential-
algebraic equations, which is solved iteratively by ADAMS. Different versions of the original
four-axle model were also created for use in this study. These models can be created by varying
chassis characteristics such as number of axles, spring types, and dampers.
The structure of the model is modular. The parts forming logical groups in the actual vehicle
are first put together as submodels, and then the submodels are connected to each other. This
way the main model is easy to modify and to expand. All the submodels also work independently,
so their correctness can be checked and their dynamic behavior can be studied.
In the following, the input data and the submodels used are discussed in more detail.
II II
FIG. 1--Side view of the truck model.
The Input Data

The geometry and kinematics of different parts were derived from drawings. The mass and
inertial properties needed were measured as far as possible. For complex structures such as
the frame, the inertial data were calculated by dividing them in smaller pieces. The properties
of each piece were found by using a separate computer program, after which they were put
together to represent the whole structure.
The characteristics of dampers, rubber mountings and bump stops, tires, and springs are
based on the data derived from the manufacturer in question. Most of those dynamic properties,
which were not directly available, were measured during the project. Some data that were
difficult to achieve through measurements had to be approximated. Every approximated value
was first carefully considered to have a minor effect on the dynamic behavior of the whole truck.
Modeling of Leaf Springs

Multileaf springs can support an axle in every direction. At the same time they act as springs
and as connecting arms between the frame and an axle. Under deformation, a multileaf spring
moves the axle both vertically and longitudinally. This kinematic effect was taken into account
in the model by replacing each multileaf spring with four links (Fig. 2). The first link acts
between the frame and the axle, which itself forms the second link. One end of the third link
is connected to the axle and the other to the shackle, which is the fourth link. Finally, the
shackle is connected to the frame.
The first and the third links are connected to the axle by fictitious revolute joints. The
location of the joints and the lengths of those links are chosen so that there is very little
difference between the kinematical movements of the model and of the multileaf spring itself.
The spring force is modeled with equivalent torsional springs generating moments on the
revolute joints. This way the model can represent the behavior of a multileaf spring with
HUHTALAETAL. ON COMPUTERSIMULATION 39
A= Bushing A, i \ link
/ ~" \ ~ " "~ "A (shackle)
B= Revolute joint
M= Torsional spring link 1-/ "\..._ \ // ~-Iink 3
~-- link 2 (axle}
FIG. 2--Model of the multileafspring.
reasonable accuracy also when it is under torsion, for example, when there is a driving or a
braking force acting on a tire. Significant hysteresis of the multileaf springs is taken into
account through the definition of the equivalent torsional springs.
There are parabolic leaf springs on the front axle and, in modified versions of the model,
also on the bogie. Modeling them was analogous to the one described above. The main difference
was hysteresis, which is much smaller for parabolic leaf springs than for multileaf springs.
Modeling of Air Springs
The behavior of the air springs, used on the second axle, is highly nonlinear. A special
subroutine was added to the model to calculate the actual forces acting on the air springs.
First, the subroutine uses built-in information to define the volume of the air spring as a
function of instantaneous length of the spring. Then, instantaneous pressure inside the air
spring is calculated using the following equation
p ( I ) = p o " V ~ 4 / V ( l ) ~4
where
p(/) -- instantaneous pressure,

1= length of the air spring,
P0 = initial pressure,
V0 = initial volume, and
V(/) = instantaneous volume.
An adiabatic process is assumed to take place there. The initial pressure depends on the
load of the left multileaf spring of the bogie, i.e., the load of the truck. The actual force
generated by the air spring is the pressure multiplied by the effective area of the spring. The
latter is again defined through a built-in nonlinear function of length by the subroutine.
Modeling of Dampers
Nonlinear and asymmetrical dampers are commonly used in vehicles. The force they generate
depends nonlinearly on the relative velocity between their ends (Fig. 3). Asymmetrical means
that they each have different properties for push and pull. The properties of the dampers used
in the model were measured and then used as they were, in numerical form. A spline-function
in ADAMS was used for creating smooth curves going through the measured points.
Modeling of Bump Stops and Rubber Mountings

Rubber has two important material properties, hysteresis and nonlinear stiffness (Fig. 4).
With small deformations the stiffness can be assumed linear. Also, the hysteresis of rubber
can be modeled using equivalent velocity-dependent damping if the frequency spectrum of
the vibration is narrow. These simplifications were used in modeling the rubber mountings of
the cabin suspension.
There are rubber bump stops on the front axle and on the bogie. Normally they have no
effect on the dynamic behavior of the truck. They are used only to prevent plastic deformations
of the multileaf springs in extreme dynamic load conditions. Their nonlinear stiffness was
included in the model, but for hysteresis, an equivalent damping coefficient was used.
The Frame of the Truck

The frame acts as a connecting part for all the other submodels. It was assumed to be rigid,
which was considered to be a reasonable assumption for this particular truck. This assumption
is known to be one of the major simplifications in the model. Further studies are needed to
define the input data for modeling the flexibility of different types of frames with different loads.
Another simplification is that the engine and the gearbox are connected rigidly to the frame.
In reality they are connected with rubber mountings to each other. This simplification can
easily be eliminated, if necessary.
SHOCK ABSORBER
FORCE
"<
pull
FIG. 3--Nonlinear model of the damper.
RUBBER HYSTERESIS
DEFORMATION
FIG. 4--Nonlinear model of rubber bump stops.
The Front Axle

The front axle is connected to the frame through parabolic leaf springs and dampers (Fig.
5). In the model, both springs are replaced with four links and two torsional springs, as
described earlier. Both dampers are replaced by a force acting between the actual attachment
points. The global direction of the force changes from time to time during the simulation
because the axle is moving with respect to the frame. The magnitudes of the forces depend
on the instantaneous velocity between the attachment points. The velocity dependence is both
nonlinear and asymmetric.
The front axle is equipped with rubber bump stops, which can generate forces between the
springs and the frame. They are active only when the parabolic leaf springs have almost
reached their maximum deformation allowed.
The tires are connected to the axle through kinematical joints that allow them to rotate only.
The steering was not included in this model, but it is possible to add this into the model if
_shock o b s o r b e r
\ !
/
M = Torsional spring \ /
\ /
~... ../ ,,
FIG. 5--Model of the front axle.
i p o i r spring /
, - / - R t,
"<~..~ t,___i)/7-----'I---, shock
~'~-~-~--t-- s ~Y ~ absorber
/
\ /
\ /
FIG. 6--Model of the second axle.
handling maneuvers are to be analyzed. The properties of the tires are discussed later in
this paper.
The Second Axle

The second axle is attached to the frame with two longitudinal struts and one lateral bar
that reduces vibrations of the axle (Fig. 6). For this axle, two air springs and two dampers
are used. The dampers and the tires were modeled similar to those of the front axle.
The Bogie
The drive axle of the bogie is suspended with two multileaf springs, both of which have
rear eyes kinematically coupled with the other pair of dual tires, respectively (Fig. 7). The
damping is taken care of solely by the hysteresis of the multileaf spring, and no dampers are
needed. In between the frame and the drive axle, there are rubber bump stops. They are
modeled similarly to those of the front axle.
The four-link approximation was used for modeling the multileaf springs. The actual stiffness
and hysteresis characteristics of such a multileaf spring were first measured by the manufacturer
and then included in the model through a separate subroutine.
Dual tires are used on both axles of the bogie. In the model, a pair of tires act together as
one mass, but the tire-road interaction forces are calculated separately for both. In order to
/-bump stop
I L)_
/ j \
M = Torsional spring
\ / /
Me = Driving moment / ~ /
FIG. 7--Model of the bogie.
maintain constant speed of the truck, the drive torque was added. It is modeled using a simple
control function that adjusts the rolling speed of the driving tires by changing the torque.
The Cabin
The original project was focused on the cabin suspension and on ride comfort. Therefore,
the cabin suspension and even the driver's seat were modeled in detail. They have a minor
effect, however, on loads on pavements. When only vehicle-road interaction forces are of
interest, they can be eliminated without significant loss in accuracy in order to save CPU time.
The Tires
A special tire model was used to calculate the tire-road interaction forces separately for
each of the twelve tires. The radial and lateral stiffness, damping, static and dynamic friction,
and the rolling resistance of the tire are given as parameters to this subroutine. The longitudinal
slip, loaded radius, and the instantaneous speeds of the tire are defined by the subroutine itself
at each time step of the simulation. The tire is divided in small sectors (1 to 2 ~ in order to
calculate the longitudinal impact on uneven road. The road profile is given as a group of
points for each tire. The subroutine then connects these points with straight lines.
Because all the parameters are defined independently for each tire, one can change the types
of tires or vary the tire pressure on different axles. Separate road profiles for each tire can
also be used. In this study the same road profile was used for each tire, but the pressure of
tires varied from one axle to another. The radius of the front axle tires was also slightly bigger
than that of the others. The tires used in the model are listed in Table 1.
Different Versions from the Original Truck Configuration
In order to examine the repeatability of road-vehicle interaction forces on different road

profiles, the number of axles, spring types, damping, and speed of the truck were varied. Due
to the number of different possible combinations, not all of them were simulated.
First of all, a three-axle version was created by eliminating the second air-suspended axle.
The payload had to be lowered to achieve legal axle loads and total weight. The inertial
properties of the payload were naturally updated at the same time.
As different spring types, the original multileaf spring and a parabolic leaf spring were used
on the bogie. A parabolic leaf spring has a very low hysteresis, so additional dampers must
be used. In order to study the effect of damping further, the strength of the dampers were
varied as well as the number of dampers on the bogie. Variations are presented in Table 2.
TABLE 1--The truck tires.

Unloaded Radius,
Axle Number of Tires Pressure, bar Type of Tires mm
Front 2 8.5 315/80 R22.5 538

Second 2 8.0 315/80 R22.5 538
Drive 4 7.5 295/80 R22.5 525
Fourth 4 6.0 295/80 R22.5 525
TABLE 2--Simulated combinations: 4L = four-axle leaf spring, 4P = four axle parabolic spring,
3P = three-axle parabolic spring.
Suspension Type
4L, Speed, km/h 4E Speed, km/h 3E Speed, km/h
Road 70 80 90 70 80 90 70 80 90
6 * * X~ * X"
7 * * *
8 ~ * *
"With three dampers.
The Road Profiles

The three road profiles studied were measured on national highways using the road surface
monitoring vehicle fabricated by VTT. It measures the profiles by accelerometers and laser,
and thus, the wavelength is limited approximately to 30 m.
The first, called "Road 6" (length I00 m), represents a rather good highway with IRI
(International Roughness Index) value of 1.52. Its profile consists of long waves, and there
is little local unevenness. "Road 7" (100 m) has an IRI value of 2.40, but includes more short
radical changes in its profile, especially in the end part. The third profile used, "Road 8" (200
m), has an overall IRI value of 1.89; the first 100 m is, however, much worse (IRI value of
2.73). Unfortunately, even Road 8 is not an extreme example of existing roads in Finland,
which are somewhat rougher than samples taken from the national highway system.
Simulations
The dynamic properties of only one truck were available. It was realistic, however, to take
off one axle, and the second type was available. The parabolic springs could be replaced by
realistic muitileaf springs. Because the number of trucks was very limited, three different
speeds (70, 80, and 90 kin/h) were used. The legal speed limit for trucks, 80 km/h, was chosen
to be the most frequently used speed. In this study the other speeds were used only with one
model version on two different road profiles. The combinations are shown in Table 2.
The simulations were run on a Sun 4/390 workstation. CPU time used for a 100-m simulation
with 200 output steps per second was 1 to 1.5 h, depending on the speed and the model
version. The total number of time steps during the simulation varied between 1000 and 1500,
and the number of cumulative corrector iterations needed were roughly 4000.
The values used in the simulation are based on the measured properties of the vehicle
components. The manufacturer makes the vehicles according to the order of the buyer, and
thus only a few vehicles used in this simulation are made.
The Road and Traffic Laboratory of the Technical Research Center (VTT) has the technology
to measure the dynamic axle loads in a vehicle, but because all the vehicles of that type were
in use, no calibration of the model could be made at this time.
Results
Frequency Effects
There are usually two specific frequencies: a lower one due to the pitch and the bounce of
the truck body, and a higher one due to the bounce of the axle.
Some results of different suspensions are shown in Fig. 8 (the driving axle in the bogie)
as a function of time. It can be seen that the greatest dynamic axle loads are much greater if
multileaf springs are used compared to parabolic springs with dampers. If the dynamic force
is great, the stronger dampers have relatively little effect on the greatest dynamic axle loads
(for instance, at 1.5 and 2.5 in Fig. 8). The stronger damper extinguishes the oscillations faster,
and the dynamic forces are generally smaller.
Figure 9 shows dynamic axle loads of the front axle as a function of the road station at
three speeds. The longitudinal profile of Road 6 (note, wavelengths over 30 m are neglected)
is also shown. The low frequency due to the pitch of the truck body dominates, and the higher
frequency from the axle bounce is much smaller.
The dominating frequency is higher in Fig. 10 (the driving bogie axle on the same Road
6) because it is due to the bounce of the axle. The pitch mode is nearly nonexistent.
Figure 11 presents the front axle on Road 8, which is rougher (IRI 2.73) than Road 6, and
consequently, dynamic axle loads are greater. It is possible to distinguish both lower frequencies
due to truck body pitch and higher frequencies from axle bounce. Higher frequencies are
greater than on Road 6 (Fig. 9), but still much less important than low frequencies.
Figure 12 presents the dynamic axle loads of the driving axle on the rougher Road 8.
Dynamic axle loads are much greater than on Road 6 (Fig. 10). The higher frequencies due
to the axle bounce dominate.
Axles 2 (the second axle) and 4 (the carrying axle in the bogie) behaved basically like Axle
3 (the driving axle in the bogie).
120
....... MULTILEAF SPRING
- - pARABOLIC SPRING WIOAMPINO
.... PARABOLIC 9PRING W/STRONGER

DAMPING
i.
110
Z
a
<
o. J 100
UJ !! .'
.J
90
I I I I
1 2 3 4
TIME (s)
FIG. 8--Dynamic axle loads of the driving axle on Road 6.
200
120
150
100
~" 80 100 E
I/J
r, ..i
Ik
o
re
"J 50 13.
l,U
.J
50 "
....... ?okn~
40 - - 80kn~
.... 90 kmth
20
I I I I -50
20 40 60 80 150
STATION (m)
FIG. 9--Dynamic axle loads of the front axle on Road 6.
200
120
150
~" 80 100
..I
i: rr
o, 0
E
Q.
iJU
,=J 50 o
x ....... 70 knVh
- - e0 kra/h
40
.... !)0 kmth
20
I I I I -50
20 40 60 80 100
STATION (m)
FIG. lO---Dynamic axle loads of the driving axle on Road 6.
HUHTALA ET AL, ON COMPUTER SIMULATION 47
25O
120
20O
E
8O 150 e
UJ
O .J
u.
O O
n-
tl,I O.
~,e ....... 70 k r r ~ loo
~ I~km/h O:
40 . . . . gQ k m / h
SO
20
I I I I
20 40 60 80 100
STATION (m)
FIG. 1l--Dynamic axle loads of the front axle on Road 8.
250
It
120 It
t I :" ~l
I
200
11111
150 E
Ul
a .wJ
tl
U
0 o
n-
n
w ....... 70 V,m/ll
.J
x 100 ~
~ 80lml/h 0Ilc
. . . . gO k m / h
PROFILE
50
20
I" I I I
20 40 60 80 100
STATION (m)
FIG. 12--Dynamic axle loads of the driving axle on Road 8.
FIG. 13--Dynamic loading coefficients on Road 6.
Effect of Roughness and Speed

The effect of road roughness and speed can be seen in Figs. 13 (Road 6) and 14 (Road 8).
The dynamic load coefficients (DLC = standard deviation of dynamic axle loads divided by
the mean dynamic axle load) vary between 0.03 and 0.11.
The DLC of the front axle on Road 6 is relatively great because of the pitch. The center
of gravity of the truck is close to the bogie; thus, the pitch has very little effect on the dynamic
FIG. 14--Dynamic loading coefficients on Road 8.
loading of the bogie, and the DLCs of the bogie axles are much smaller than that of the
driving axle.
The DLCs of the front axle are smaller on the rough road. The pitch of the truck is smaller
on the bad road, and the axle bounce is greater. Their combined effect makes the DLC smaller.
The DLCs of the bogie axles are smaller than that of the front axle on Road 6. The air
suspension of Axle 2 and the parabolic suspension of Axle 3 behave similarly, and Axle 4 is
slightly worse.
The DLCs of the bogie axles are clearly greater on Road 8 than on Road 6. The air suspension
is the best, the parabolic suspension on the driving bogie axle (Axle 3) is somewhat worse,
and the parabolic suspension of Axle 4 is the worst. The suspensions on bogie Axles 3 and
4 are not, of course, independent from each other; neither are they similar. The axle loads and
the masses of the axles are different, too.
As the speed increases, the DLC increases nearly linearly in all cases. The three-axle version
of the vehicle behaved almost like the four-axle vehicle, and thus it is not discussed here.
Spatial Repeatability
It was found that if a truck passes the same road at the same speed the dynamic loadings
occur at exactly the same points or the repeatability or spatial concentration of dynamic
pavement loads is perfect [3]. Very little is known if the speeds and vehicles are different;
that is to say, what is the spatial concentration of the dynamic axle loads of the whole
vehicle fleet.
The much-used formula by Professor Eisenmann since 1975 [4] assumes that the dynamic
loadings are normally distributed (Gaussian) or evenly distributed along the road.
qrp = E [P(t) 4] = (1 + 65 2 + 3 T 4) P4STAT
where
P(O = axle load at time t,

PSTAT = static axle load,
s = DLC, and
E[] = expectation operator.
The spatial concentration of dynamic axle loads of the whole vehicle fleet is not easy to
show experimentally, although important work has been done in the United Kingdom and in
the United States [3,5,6]. Because only one truck with modifications was available, the effect
of the whole vehicle fleet was simulated in this research project by driving the same vehicle
at different speeds. Each axle has its own specific frequency. If the speeds of a vehicle are
different, the maximum axle loads should occur at different points. The results were shown
as a function of the station of the road, not as a function of time, as is usual.
The spatial repeatability of the front axle on Road 6 is presented in Fig. 9. All the dynamic
axle loads are in the same phase in the beginning because before the starting point of the
simulation the road is perfectly smooth, and it takes some time before the truck body begins
to oscillate. From about 40 m, the dynamic axle loads begin to separate as a function of space
but not of time (the frequency is the same). The pitch mode dominates, and no spatial
repeatability can be found on this good road.
The dominating frequency in Fig. 10 (the driving bogie axle on the same Road 6) is higher
because it is due to the bounce of the axle and not to the pitch of the truck body. The dynamic
axle loads separate, but in some cases a single unevenness gets them into the same phase.
However, frequent small unevenness, for instance between 60 and 75 m, seems to "kill" the
spatial repeatability.
Figure 11 presents the front axle on Road 8. The dynamic axle loads are greater. Higher
frequencies are more important than in Fig. 9. The dynamic axle loads at different speeds
begin to separate, but a sharp and short unevenness gets them in the same phase, for instance,
at 40 m. The spatial repeatability is perfect between 40 and 60 m, but then the dynamic axle
loads begin to separate. At 90 m, the axle loads get once again in the same phase due to a
greater unevenness at 87 m.
The dynamic axle loads of the driving axle in the bogie (Fig. 12) show similar behavior,
but the repeatability is clearly stronger than that of the front axle, for instance, between 40
and 60 m.
The results showed that, if the road is very smooth, there is no spatial concentration of
dynamic loadings. However, only a very small unevenness in the road puts the dynamic
loadings in phase, and the spatial concentration of dynamic loadings is very clear. The same
phenomenon was found at all the bogie axles and in both roads, but not clearly at the front axle.
This study showed very clearly the assumption that the dynamic axle loads are not normally
distributed along the road. Thus the equations based on the assumption are not valid. This
limited study could not give any answer as to how the dynamic loads are distributed, but
showed the importance of the spatial concentration of the dynamic axle loads.
Concluding Remarks
The road-vehicle interaction forces were studied with two truck models, one with four axles
(twelve tires) and the other with three axles (ten tires). The truck models were created using
a multibody system analyzing program (ADAMS). The input data for the model are based on
real values derived from the truck manufacturer.
Two longitudinal road profiles were used. The behavior of the model seemed to be realistic,
but there was no possibility to verify the model.
The simulation model was used to study dynamic axle loads on two measured longitudinal
road profiles. The test truck was equipped with different suspensions and dampers. The
spatial repeatability of the dynamic axle loads was simulated with different speeds of the
simulated vehicle.
The simulation showed very clearly the basic behavior of the vehicle on two roads of
different roughness. The results indicated that if the road is very smooth there is no spatial
concentration of dynamic loadings; but, only a very small unevenness in the road put the
dynamic loadings in phase, and the spatial concentration of dynamic loadings is very clear.
However, frequent small unevenness seems to "kill," in certain cases, the spatial repeatability.
Only a few profiles have been analyzed. The approach has not been statistical but determinis-
tic. The basic idea was to find what kind of profiles cause spatial repeatability in which cases.
Simulation provides good possibilities because different profiles can be used. Usually only
one profile is available for the experimental measurements concerning spatial repeatability.
This research project showed that spatial repeatability is influenced by both the road profiles
and the vehicles. More profiles are available now, and dynamic axle loads have been measured
with an instrumented vehicle. The results are not yet available. The next goal is to develop a
simple dynamic truck index which can be used in the road monitoring vehicle which measures
IRI and the rut depth.
References
[1] Antoun, R. J., Hackert, P. B., O'Leary, M. C., and Stichin, A., "Vehicle Dynamic Handling Computer
Simulation--Model Development, Correlation and Application Using ADAMS," SAE Technical
Paper 860574, Society of Automotive Engineers, Warrendale, PA, 1986.
[2] lchikawa, A. and Susuki, Y., "Truck Controllability and Stability Analysis by using ADAMS,"
Mitsubishi Motors Corporation, 1991.
[3] Gyenes, L. and Mitchell, C. G. B., "The Spatial Repeatability of Dynamic Pavement Loads Caused
by Heavy Goods Vehicles,Heavy Vehicles and Roads," Proceedings, Third International Symposium
on Heavy Vehicle Weights and Dimensions, 28 June-2 July 1992, Cambridge, Thomas Telford
Services, Ltd., London, pp. 95-101.
[4] Eisenmann, J., "Dynamische Radlastschwankungen--Strassenbeanspruchung," Strasse und Auto-
bahn, No. 4, 1975, pp. 127-128.
[5] Cebon, D., "Vehicle-GeneratedRoad Damage: A Review," Vehicle System Dynamics, Vol. 18, 1989,
pp. 107-150.
[6] Cebon,D. and Winkler, C. B., "A Study of Road Damage Due to Dynamic Wheel Loads Using a Load
Measuring Mat," Strategic Highway Research Program, National Research Council, Washington,DC,
1991.
Thomas D. Gillespie 1 and Steven M. Karamihas 2
Heavy Truck Properties Significant to

Pavement Damage
REFERENCE: Gillespie, T. D. and Karamihas, S. M., "Heavy Truck Properties Significant
to Pavement Damage," Vehicle-Road Interaction, ASTM STP 1225, B. T. Kulakowski, Ed.,
American Society for Testing and Materials, Philadelphia, 1994, pp. 52-63.
ABSTRACT: Heavy trucks are a primary source of road damage due to the stresses imposed
by heavy axle loads. The mechanics of truck-pavement interaction are examined to determine
which truck characteristics have the most direct and strong relationship to road damage.
Static axle loads are found to be one of the most important factors affecting both rigid and
flexible pavement fatigue damage because of the power relationship of load to damage. As a
result, the load-sharing capabilities of tandem suspensions are also important. Tandem axle
spacing is only significant to performance on rigid pavements and is always beneficial, Dynamic
loading effects dependent on road roughness, travel speed, and truck dynamic properties are
also quantified. Finally, the relevant tire properties are examined. Tire size and inflation pressure
have strong influences on fatigue damage of flexible pavements. Tire construction type (radial
and bias) appear to have influences on tracking that may explain the recent problems with
accelerated rutting.
KEY WORDS: heavy trucks, pavement stress, pavement damage, cracking, rutting
Heavy trucks have long been recognized as a source of pavement damage due to the stresses
imposed by heavy axle loads. The mechanistic relationship between truck properties and
pavement damage has been studied both empirically (e.g., the AASHO Road Test of 1958
[1]) and using analytical models of pavements to calculate generalized responses [2-4].
Recently, new insights on these relationships have been obtained analytically in the NCHRP
Project 1-25(1), "Effects of Heavy Vehicle Characteristics on Pavement Response and Perfor-
mance" [5]. The paper summarizes the truck properties found in that work to be most important
to pavement damage.
Analytical models of trucks and pavement structures were developed to allow systematic
study of the pavement responses to the moving, dynamic loads of various truck configurations.
The truck characteristics studied included: truck type (straight trucks, tractor-semitrailers, and
multiple-trailer configurations); axle loads; number of axles; spacing between axles; suspension
types (single axles with leaf and air springs and tandem axles with leaf springs, air springs,
and walking beams); and tire parameters (single/dual configurations, radial/bias construction,
and inflation pressure). The response was determined in both rigid and flexible pavements of
various designs and properties, with variations in road roughness and vehicle speed. Pavement
1Research scientist, The University of Michigan Transportation Research Institute, 2901 Baxter Rd.,
Ann Arbor, MI 48109.
Research Assistant II, The University of Michigan Transportation Research Institute, 2901 Baxter
Rd., Ann Arbor, MI 48109.
52
GILLESPIE AND KARAMIHAS ON HEAVY TRUCK PROPERTIES 53
responses (stresses/strains/deformations) were evaluated throughout the pavement, and the

damage from the passage of the truck was estimated using accepted pavement damage models.
The discussion that follows summarizes key findings with regard to the truck characteristics
that are most significant as causes of pavement damage.
Analytical Approach
The generalized approach to modeling truck-pavement interaction is illustrated in Fig. 1.

Pitch-plane models of 33 trucks of interest were developed. The trucks were "run" on roads
with generalized roughness properties designed to represent the characteristic spectral content
of each type of road [6]. In the case of rigid pavements, the random roughness was augmented
with periodic components representing typical failure mechanisms of faulting and slab-flit
(correlated spatially with the pavement being modeled). The dynamic loads along the pavement
were calculated for each axle of the combination.
The pavements were analyzed for their response to truck wheel loads using pavement
structural models that compute influence functions for the response at each point to a unit
load imposed anywhere else along the pavement. The rigid pavement model was ILLI-SLAB,
which is a finite element slab on an elastic base with provisions for different joint properties
[3]. Nineteen rigid pavement designs were analyzed. Thirteen flexible pavements were analyzed
using a model derived from VESYSDYN, which is a multilayer elastic model [7]. Examples
of the influence functions are shown in Fig. 2. The elastic models are configured for calculating
stress and strain response affecting fatigue. Rutting behavior was approximated by modifying
the visco-elastic model to a visco-plastic form. This was accomplished by replacing elastic
material properties with their viscous equivalents [5].
By combining the dynamic loads from the truck simulations with the influence functions,
the time-varying stresses and strains could be calculated at points along the wheel path due
to the combined loads of all wheels of the truck as it passed by. The time history of stress
was evaluated at each point using the most commonly accepted models to estimate the
cumulative damage from passage of the truck. For estimating fatigue damage (which results
in cracking), a power law relationship was used, with failure limits appropriate to the material
properties of the road structure. For rutting damage, permanent deformation in each layer was
computed using the visco-plastic model.
Computations were made over road sections of sufficient length to allow a statistical analysis
of the responses at incremental distances along the roadway. The statistics were compiled to
yield the average damage from each truck, along with that representative of the worst 5% of
the pavement.
Summary of Results
The various truck-pavement combinations were analyzed to determine systematic relation-
ships between truck properties and pavement damage and the interactions that occur. The
primary observations are discussed here.
Static Axle Load
The static axle load applied to the pavement is the single vehicle factor that has the greatest
effect on fatigue damage. Typical loaded truck axles vary over the range of about 4.6 ton (10
kip) for front axles to 10 ton (22 kip) for heavily loaded rear axles. Because fatigue is assumed
related to the fourth power of the load on both rigid and flexible pavements, the damage varies
by a factor of more than 20:1 over the range of static axle loads. On the other hand, the
FIG. 1--Analytical approach to damage evaluation.
pavement structural models used in this study predicted that rutting damage is directly dependent
on axle load, as explained in Ref 11. Thus, rutting varies over a range of 2.2:1 for the same loads.
Vehicle Gross Weight

The vehicle gross weight only has a direct influence on rutting because rutting is linearly
related to weight. Although fatigue of both rigid and flexible pavements varies significantly
E ~ Rigidpavement
O
e't Flexiblepavement
._=
Tension ~_._../i
Compressio~
I I I I I
-g -6 -3 0 3 6 9
AxlePositionAlongtheWheelPath(m)
FIG. 2--Influence functions for rigid and flexible pavements.
among the various types of vehicles using the road system, it is not systematically related to
gross weight, but varies in accordance with the maximum axle loads on each vehicle combina-
tion. Heavier trucks do not necessarily cause more fatigue.
Axle Spacing
Axle spacing has a moderate effect on rigid pavement damage, particularly the spacing of
the axles within axle groups. The mechanics become evident when examining the influence
function for rigid pavements, as shown in Fig. 2.
On either side of the tension peak directly under the wheel, a region of compressive response
occurs. An additional axle placed in this location benefits from the compressive influence of
the adjacent axle. Thus, the relative fatigue damage, in equivalent 8.2-ton (18-kip) single axle
loads (ESALs), for tandem axles varies with spacing, as shown in Fig. 3. For rigid pavements,
the optimal spacing for tandem axles falls between 2 and 2.75 m for weak and strong pavements,
respectively. On weak rigid pavements the damage caused by the common closely spaced (1.3
m) tandem axle can be reduced by 25% if the tandem spacing is increased to 1.8 m. On strong
concrete pavements such as those that include a bonded concrete base layer, a spacing of up
to 2.7 m will best reduce damage. However, pavements of this type are not expected to fail
due to truck loading by virtue of their strength. The benefit of closely spaced axles on rigid
pavements would permit much higher loads to be carded with no additional road damage.
Figure 4 shows the damage (in ESALs) for single, tandem, and tridem axle groups.
~_. 4, iiii
~, 16.4t (36-klp) Tandem

~ -. 25.4-cm (10-inch) Thick Pavement
\ ---- 20.3-cm (8-inch) Thick Pavement
\ Bonded to a 10.2-cm (4-inch) Thick
E \ Base Layer
._~
it.
.>=_ 1
n-
......... ; ........ ; .......... I ......... I......... I .........
0 1.0 2.0 3.0 4.0 5.0 6.0
Tandem Axle Spacing (m)
FIG. 3--Influence of tandem axle spacing on rigid pavement fatigue.
10
9
8
/
uJ
E 6
7
n.
r
14.
r 5
4 I / /
=> 3
2 I /
n- 1
0
0 10
20 30 40
Axle Group Load (t)
FIG. 4--Relative fatigue of rigid pavements for closely spaced, multiple-axle configurations.
Axle spacing has little effect on flexible pavement fatigue under the asphalt-bound layer
because the compressive range extends little more than 1 m from the tire, and minimum axle
spacing is 1.3 m for the common truck tire sizes. These findings apply to pavements with
asphalt-bound layers from 5 to 17 cm thick. Thus, these flexible pavements see multiple axles
simply as a series of separate and independent loads. Flexible pavements are the primary
barrier to carrying heavier loads on multiple-axle groups, although these results do not indicate
a need for limiting individual loads on tandem axles differently than for single axles.
Surface rutting is unaffected by axle spacing.
Static Load Sharing
Static load sharing within a multiple-axle group influences fatigue of rigid and flexible
pavements moderately as a result of the higher load on one axle when sharing is not equal.
Increasing the load on one axle of a tandem set disproportionately increases the fatigue from
that axle because of the power-law relationship between load and fatigue. The reduced load
on the other axle reduces its contribution to fatigue, but does not fully offset the increase from
the heavy axle. Figure 5 shows the relative fatigue damage on flexible pavements from
variations in load sharing coefficient. The load sharing coefficient is defined as the load on
the heaviest axle divided by the average load.
If the individual loads of a multiple-axle group are held to within 5% of the mean load for
the group, very little additional fatigue will result. Load disparities have been observed as
high as 25% [8], which increase fatigue damage as much as 60%. Static load sharing has no
influence on rutting by virtue of the linear relationship between rutting and axle load.
The significance of loads and load-distribution factors discussed above is not directly linked
to the dynamic behavior of trucks. Thus, they have been evaluated under static load conditions.
The dynamic component of axle loads can elevate the damage experienced by a pavement
above that induced by static axle loads. The dynamic effects are directly evident in the damage
influences of roughness, speed, and suspension type.
Road Roughness
Road roughness excites dynamic behavior of trucks such that instantaneous axle loads can
be much higher than static axle loads9 However, the fatigue damage that results varies only
modestly over the typical range of roughness on existing roads9 Over most roads, roughness
varies from 1.25 to 3.75 m/km on the International Roughness Index (IRI) scale [9]. A smooth
road at 1.25 m/km IRI represents a pavement serviceability index (PSI) level of approximately
3.0
Wear Course Thickness
--a--a- 5 cm
-o--o- 10 cm
2.5 16.25 cm
c)
==
2.0
~>
1.5 I = - - , - - - ,
1.00 1.05 1.10 1.15 1.20 1.25
Load Sharing Coefficient
FIG. 5--Influence of load sharing coefficient on flexible pavement fatigue damage.
4.25, and a rough road at 3.75 rn/km represents a PSI level of approximately 2.5. Figure 6
shows flexible pavement fatigue (in ESALs) caused by various drive axle suspensions as a
function of roughness. Since fatigue damage varies along the pavement section when the loads
are dynamic, the damage is evaluated at the 95% level, which characterizes the damage on
the 5% of the pavement length exposed to the most severe loadings. The 95% level is a more
sensitive indicator of damage than is an average over the section and highlights the localized
damage possible with dynamic loads that may render a road unserviceable.
The level of damage caused by the suspension running over a perfectly smooth road (i.e.,
at the static load) is indicated by a horizontal line. At even the lowest levels of roughness,
damage increases by approximately 50% above that of the static axle loads. On the roughest
roads, damage may increase by 200 to 400%, depending on the type of road and truck
suspension properties.
Speed
Vehicle speed influences pavement fatigue damage by road roughness excitation of reso-
nances in the truck and its suspensions, which increases the peak dynamic wheel loads. The
"tuning" is evident in the dynamic load coefficient (DLC) [8] of the drive and trailer axles
of the three-axle tractor-semitrailer illustrated in Fig. 7. Due to the tuning, the damage may
increase nonlinearly with speed as shown for various tandem axle types in Fig. 8. Compared
to the static case (equivalent to zero speed), the 95% damage is at least 50% and up to 300%
greater on the moderately rough road (2.5 m/km IRI) selected for these calculations. Yet, over
the typical operating speed range of 70 kph (45 mph) to 105 kph (65 mph), the damage varies
by only 20 to 30%.
Vehicle speed affects the primary response of flexible pavements through the dynamic
loading and the duration of loading. While dynamic loads (and their propensity to damage
the pavement) generally increase with speed, the time duration of the load application dimin-
5 "l
| Tandem Drive Axle Suspension Type
4 ~ 4-Spring Flat-Leaf
| "~ 4-Spring Taper-Leaf re,r
~" 4 "1 = Air-Spring
| ~ Walking-Beam j l ~
.] 88 kph
l 15.5 t (34-kip) Suspension
o) 3
o) 2
~ N ~u
1
3 4
Roughness (rrv~m)
FIG. 6---Influence of road roughness on rigid pavement fatigue damage.
0.09
A
q
t3
v
0.08,
O
O
o,
"1~0.07.
t,--
IRI = 2.5 m/kin

Drive Axle (Flat-Leaf)
Trailer Axle (Flat-Leaf)
0.06 9 i . i 9 i 9 i . i 9 i . i . i 9 i 9 I 9 i i
30 50 70 90 110 130 150

Speed (kph)
FIG. 7--Example of a vehicle "tuning" to a road.
~" 3.5 15.5 t (34-kip) Tandem Suspension Type

---o--- 4-Spring Flat-Leaf
4-Spring Taper-Leaf
9 Air-Spring
~E 3"0 "'~ Walking'Be~ -~'=
2.5 IR .
.~1.5
~ 1.0 !
40 60 80 100 120 140

Speed (kph)
FIG. 8--Interactions of speed, suspension type, and rigid pavement fatigue.
ishes. On a visco-elastic material this partially compensates to reduce the damage. This is
demonstrated in Fig. 9, which shows the 95% damage in ESALs caused by a four-spring
suspension at various levels of roughness. On a perfectly smooth road (0 IRI), where the loads
remain equal to their static values, damage decreases with speed due to the shorter load
application time. With low levels of roughness, the two effects largely cancel each other, such
that flexible pavement damage remains fairy constant with speed. With high levels of roughness
the dynamics prevail and damage increases with speed.
3.5
"l
Or)
LU 3.0
tD
E
(1I
t:3 2.5
::3
2.0 R O ~
O~
145 = 3.75 m/km .-..-..,+

"d
* 2.5 m/km
tr 9 1.25 m/km 4-Spring Flat-Leaf
I 0 rn/km Loaded to 15.5 t (34 kips)
1.0 ! I I I
40 60 80 1O0 120 140

Speed(kph)
FIG. 9--Interactions of speed, road roughness, and flexible pavement fatigue.
Rutting is decreased by the shorter load application times at high speed. Because it is linear,
the average rut depth is not affected by the truck dynamics, although the localized damage
may contribute to a buildup of roughness.
Single Axle Suspension Type

Single axle suspension types (air- and leaf-spring) have only a moderate effect on rigid or
flexible pavement damage. Although the suspension plays a primary role in dynamic behavior
and the increased damage that results, the range of variation in stiffness properties believed
typical of single-axle suspensions is small enough that the suspension type has only second-
order influence on damage.
Tandem Dynamics
The dynamic loads from tandem axles and suspensions have a much greater influence on
fatigue of rigid and flexible pavements. Fatigue damage of rigid and flexible pavements may
vary by 25 to 50% between the best (air-spring) and worst (walking-beam) suspensions on
moderately rough roads and nearly 100% on very rough roads. Suspension type has little
influence on flexible pavement rutting.
Tire Configuration
Variations in the size of the tire-pavement contact patch are responsible for the wide variation
in the pavement damaging potential of single, dual, and wide-base tires. Flexible pavement
fatigue is highly sensitive to variations in size of the tire contact area over which the load is
distributed. Single tires are so damaging relative to duals that an axle loaded to 5.5 ton (12
kip) with single tires (typical of a steer axle) is often more damaging than an axle with dual
tires loaded to 9.0 ton (20 kip).
Rigid pavement fatigue is not as sensitive to tire contact conditions. Thus, axles with single
tires are no more damaging than those with duals when operated within the rated loads of
the tires.
Rutting is dependent on load and contact area. For a given load, rut depth is higher when
it is carded on single tires, although the rut volume differs little between single and dual tires.
Tire Inflation Pressure
Variations in tire inflation pressure affect pavement damage by changing the size of the
contact patch and the tire vertical stiffness. Inflation pressure has only a moderate impact on
rigid pavement fatigue because it is not very sensitive to the decrease in contact area at high
inflation pressures. On the other hand, flexible pavement fatigue is strongly affected by these
changes, as shown in Fig. 10. In the single tire case, damage may increase by more than 100%
with a 70 kPa (10 psi) increase in pressure. Rutting also moderately increases with inflation
pressure. Changes in tire vertical stiffness with inflation pressure have little impact on damage.
Overall, changes in tire dimensions and inflation pressures that will reduce contact pressures
can reduce road damage. This translates into use of tires with the widest available tread and
largest diameters. The competitive pressure for truck operators to adopt small, low-aspect tires
to their fleets carries with it the potential for even more road damage from trucks in the future.
6 6
Wear Course Wear Course
Thickness Thickness
m
g, 5 | 5
- 7.6 cm / 7.6 cm
oJ 1 2 . 7 cm
c~ 4
z 1
0
z
0
1 __f 1
75 90 1'05 120 (psi) 75 120 (psi)
520 620 720 830 (kPa) 520 620 720 830 (kPa)
Inflation Pressure Inflation Pressure
(a) 15R22.5 at 3640 kg (b) 11 R22.5 duals at 4550 kg
FIG. lO---Flexible pavement fatigue damage as a function of inflation pressure for wide-base
single and dual tires.
Tire Construction
Tire construction type (radial versus bias-ply) has little direct effect on fatigue of rigid or
flexible pavements. The differences in vertical stiffness and contact patch size are second
order in magnitude and may not be systematic between tire types.
However, the different camber and cornering properties of radial and bias-ply tires may
affect wheel tracking behavior and consequently the rutting damage. Trucks with radial-ply
tires will tend to track more precisely, and the low camber stiffness makes it easier for tires
to track in existing pavement ruts. Figure 11 shows the forces acting on a truck tire operating
in a rut. Tracking behavior depends on the balance between the gravitational force component
trying to push the tire to the bottom of the rut and the camber thrust opposing it. The critical
level of camber coefficient determining the direction of tire movement is 0.0175 kgf/kgf/
degree. From the sparse amount of data available on these properties of radial and bias-ply
tires (see Fig. 12) [10], it appears that radial fires lack camber thrust to run out of the rut,
while bias-ply tires do. Thus, trucks with bias-ply tires will tend to climb out of ruts, thereby
distributing rutting damage over the width of the wheeltrack. Those with radial tires will run in
the rut once one has formed, thereby accelerating fatigue and rutting damage and concentrating it
in a narrow wheeltrack. This factor may be one of the primary causes for the frequent incidence
of dual-wheel ruts occurring on flexible pavement roadways [11] (sometimes known as "sudden
early rutting") that have appeared concurrently with the industry transition to radial-ply tires.
Conclusions
In examining truck characteristics that affect the amount of road damage they may cause,
it is clear that there are many ways in which their damaging potential can be reduced. The
most direct impact is obtained by reducing axle loads. While maximum loads are defined by
road use laws, even for trucks below the legal load limits, more uniform distribution of loads
between axles can do much to reduce damage. This translates into loading tractor-semitrailers
to obtain equal loading between tractor drive axles and trailer axles and using tandem suspen-
sions with demonstrated ability to provide equal loading between tandem axles. Concurrently,
the industry should strive to use the largest tires and lowest inflation pressures practical in
each application to minimize fatigue damage to flexible pavements.
q~ Camber
r ~ T Thrust
~1 due
LateralForce
to Load
FIG. l 1--Forces acting on a tire on a cross-slope surface.
I
Tracks in rut ~ Climbs out of rut
20- I
0 , I
0.01 t.02 0.03
Camber Coefficient, Cy/Fz (kgf/kgf/deg)
FIG. 12--Frequency distribution of camber coefficientfor radial and bias-ply passenger car tires.
References
[1] American Association of State Highway and Transportation Officials, "The AASHO Road Test.
Report 7, Summary Report," Highway Research Board Report 61, Highway Research Board,
Washington, DC, 1962.
[2] Westergaard,H. M., Stresses in Concrete Pavements Computed by Theoretical Analysis, John Wiley
and Sons, Inc., New York, No. 2, April 1926, pp. 25-35.
[3] Tabatabaie, A. M., Barenberg, E. J., and Smith, R. E., "Longitudinal Joint Systems in Slip-Formed
Rigid Pavements. Volume II--Analysisof Load Transfer Systems for Concrete Pavements," Report
Number FAA-RD-79-4, Federal Aviation Administration, Washington, DC, 1979.
[4] O'Connell, S., et al., "Analyses of Moving Dynamic Loads on Highway Pavements, Part I; Vehicle
Response," Proceedings, International Symposium on Heavy Vehicle Weights and Dimensions,
Kelowna, British Columbia, 1986.
[5] Gillespie, T. D., et al., "Effects of Heavy Vehicle Characteristics on Pavement Response and
Performance," Preliminary Draft Final Report, The University of Michigan Transportation Research
Institute, Report No. UMTRI 92-2, Ann Arbor, MI, December 1991.
[6] Sayers,M. W., "Dynamic Terrain Inputs to Predict Structural Integrity of Ground Vehicles," Wright
Patterson Air Force Base, University of Michigan Transportation Research Institute, Report No.
UMTRI 88-16, Ann Arbor, MI, 1988.
[7] Kenis, W. J., et al., "Verification and Application of the VESYS Structural Subsystem," Proceedings,
5th International Conference on the Structural Design of Asphalt Pavements, 1982, pp. 333-345.
[8] Sweatman,P. E, "A Study of the Dynamic Wheel Forces in Axle Group Suspensions of Heavy
Vehicles," Australian Road Research Board, Vermont South, Australia, Special Report 27, 1983.
[9] Sayers,M. W., Gillespie, T. D., and Paterson, W. D. O., "Guidelines for Conducting and Calibrating
Road Roughness Measurements," World Bank Technical Paper, ISSN 0253-7494, No. 46, World
Bank, Washington, DC, 1986.
[10] Ervin, R. D., "The State of the Art of Knowledge Relating Tire Design to Those Traction Properties
which May Influence VehicleSafety," The University of Michigan TransportationResearch Institute,
Report Number UM-HSRI-78-31, Ann Arbor, MI, 1978.
[11] Smith, H. A., "Critical Assessment of Tire Pressure Research," National Cooperative Highway
Research Program Project 20-7, Task 36 Final Report, Highway Research Board, Washington,
DC, 1991.
Laboratory and Field Tests of
Vehicle-Induced Pavement Loading
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorize
Frederick Hugo t
Some Factors Affecting the Design and

Use of the Texas Mobile Load Simulator
REFERENCE: Hugo, E, "Some Factors Affecting the Design and Use of the Texas Mobile
Load Simulator," Vehicle-Road Interaction, ASTM STP 1225, B. T. Kulakowski, Ed., American
ABSTRACT: The Texas Department of Transportation (TxDOT) is implementing a strategy

for full-scale accelerated pavement testing (APT). The Texas mobile load simulator (TxMLS)
is being designed and manufactured as part of this strategy. Two 1:10 scale MLS models are
also being used to conduct laboratory tests to enhance the design, understanding, and use of
the machine.
The expected dynamic behavior of the MLS was investigated using computer simulation.
This greatly enhanced the understanding of the vehicle pavement interaction as well as the
structural response of the MLS. The feasibility of the structural system was confirmed, and the
need for a continuous ramp and transition zones at the junction of the circular and straight
sections was demonstrated. The simulation was also used as a means of reducing unwanted
dynamic effects and finding the optimal solution for the electric drive. Variable frequency drive
(VFD)-induction motors were selected.
The full-scale prototype is expected to be completed by the summer of 1993. Tests with the
modei-MLS on pavements demonstrated that it could be used to conduct selective studies of
wheel/pavement interaction to enhance the future use of the full-scale MLS. The first tests have
focused on rutting of asphaltic concrete. The effects of accelerated aging of asphaltic concrete
by heat is expected to be important in the future use of APT.
KEY WORDS: accelerated pavement testing, vehicle simulation, mobile load simulator (MLS),
model pavements, rutting, creep test
At the annual meeting of the Transportation Research Board (TRB) in 1991, a presentation
was given on the development of a strategy for the implementation of full-scale accelerated
testing for the Texas Department of Transportation (TxDOT) [1]. Attention was given to the
role of accelerated pavement testing in pavement engineering, and the development of the
mobile load simulator (MLS) by T x D O T was discussed. Details were given of the proposed
configuration of the machine as well as the structure and composition, using off-the-shelf
items as far as possible in order to ensure greater accuracy of simulation of real traffic. It was
also pointed out that two 1:10 scale models had been constructed that would be used to
augment the design and construction of the prototype, as well as to conduct laboratory tests
under controlled conditions on various pavement materials in order to further the understanding
of the use of the machine.
A schematic drawing of the model-MLS is presented in Figs. la, and l b shows a photographic
side view of it during a test. It has six double bogies, linked together to form an endless chain
which moves around a set of looped rails mounted in the vertical plane on a fixed frame.
1Research fellow, Center for Transportation Research, University of Texas at Austin, also professor at
the University of Stellenbosch, Republic of South Africa.
67
P1..ATFnRN F'OR 700kg DEADVE'IGHT

f
1
190
-- LINK 130
553
/J~,.. A,~-! L
,F~ ~--'~_~o~ p : I
F I ~176I l J I ~ ~ I. L J !lf~
, i
.d/,.,/
~W~I,~CC~.~I POVERRAILS-vI SLIDINGCONTACTS-~-'I~- , ~~
" -'~~/
1734
NOTE: All m e a s u r e m e n t s are in millimeters

FIG. la--Schematic drawing of the model-MLS.
FIG. lb--Photographic sideview of the model-MLS during a test.
HUGO ON THE TEXAS MOBILE LOAD SIMULATOR 69
Each double bogie contains two axles, each with suspension springs and two sets of dual tires.
One axle is electrically driven. The double bogies are similar to the rear axles of a heavy
truck. The solid rubber tires have a resilience comparable to that of full-scale truck fires.
Lateral distribution of the wheel tracking is achieved by different axle lengths.
When a bogie moves along the bottom part of the rails, it is loaded by the fixed frame
while the fires are in contact with the underlying pavement. During this phase the bogies are
powered via sliding contacts on a power rail. A typical tire contact pressure of 560 kPa is
maintained. Loading on the wheels can be adjusted by raising or lowering the fixed frame on
its four adjustable pods. With its twelve recirculating axles, trafficking is applied at a rate of
more than 10 000 equivalent real axle loads per hour.
The loading of the wheels is monitored by two displacement transducers on the two suspen-
sion springs of one of the axles. The two signals are transmitted by an infrared link from the
moving bogie to a receiver on the fixed frame, from where it can be recorded on an oscilloscope
or data logger. The number of applied axle loads is recorded by a mechanical counter.
The full-scale prototype of the MLS machine is presently under construction in Victoria,
Texas. It is ten times larger than the model and utilizes conventional truck components. More
details are given in Ref 1. This development of the full-scale prototype is being done under
the auspices of the Center for Transportation Research at The University of Texas at Austin,
and commissioning is expected by the summer of 1993. This paper presents some of the
findings during the design phase and details of the first test with the model-MLS on model
pavement structures.
Design of the MLS
Initially it was considered best to develop the MLS prototype by means of a "design and
construct" process. This would have allowed the possibility of exploring certain options by
trial and error. After consideration of this procedure and review of the methods that had been
adopted for the development of both the South African heavy vehicle stimulator (HVS) and
the Australian loading facility (ALF), it was decided that it would be better to conduct a
preliminary design in order to establish the feasibility of the concept. Once this had been
successfully concluded, it would be more feasible to complete the detailed design and then
have the machine constructed according to comprehensive drawings and manufacturing specifi-
cations. It was anticipated that this would minimize unnecessary expenditure and at the same
time ensure a cost-effective manufacturing process since it was considered likely that once
the machine was constructed, there could be interest in the production of further models.
It was apparent that the machine would require a systems approach involving various
disciplines of engineering such as electrical, mechanical, structural, and electronics. The dynam-
ics were expected to be complex, including a consideration of pavement/mechanical and
mechanical/structural interaction. Detailed specific operational requirements (SORs) were
developed prior to embarking on the design of the MLS [2].
The Center for Transportation Research (CTR), which is responsible for the conceptual
design, design review, and management of the project, used Stress Engineering Services (SES)
in Houston as a subcontractor for the preliminary design.
During the developmental phase, a number of manufacturers of trucks, as well as manufactur-
ers of items that were being considered as possible elements of the MLS, were visited. Research
institutions studying vehicle/pavement interaction were also visited. This allowed a very
comprehensive database of information to be provided for the use of the design team as well
as the management group, who were responsible for the setting of the specific operational
requirements (SORs) and the resulting design specifications. With the aid of these, it was
possible to manage the project as the design progressed and exposed difficulties.
The draft report on Phase 1 of the design [3] was completed in July 1991, and, in view of
the findings on the feasibility, recommendations were made for the continuation of the prelimi-
nary design in Phase 2. Phase 2 was completed in January 1992. Presently the project is in
Phase 3, the final design and manufacture phase which began in April 1992 and which deals
with the detailed design and manufacture, with Victoria Machine Works (VMW) of Victoria,
TX, as the contractor. In the next section some of the findings from the feasibility study and
the subsequent design phases will be discussed.
Overview of Findings from the Feasibility Design Phase
The feasibility design phase focused on the dynamics of the system as well as electromechani-
cal and electronic elements of the system. Each of these will be briefly considered by high-
lighting some of the salient points from the report [3].
Challenges During the Feasibility Study
During the feasibility study it soon became apparent that the following were to be major
challenges facing the design team:
1. Finding a solution to enable the MLS to simulate travel on a gradient as well as making
it able to simulate braking forces.
2. Catering for differential speed between the chain linking the bogies and the bogie tires
during the circular phase of travel.
3. Determining the impact forces owing to the geometry of the system, including the effect
of road surface characteristics. In particular, catering for the jerk developed during a
sudden change in acceleration of the bogies exiting or entering the circular travel phase.
4. Solving nonstandard mechanical features of the system in a cost-effective manner.
Computer Simulation of Dynamic Forces Generated by the MLS
In order to gain a better understanding of the transient dynamic system of the rotating
bogies, SES developed a microcomputer-based simulation of the MLS [3]. This was done by
considering all masses in the system and determining their instantaneous acceleration through
numerical integration of computed velocities. This "time domain simulation" was repeated
iteratively. With this it was possible to gain a better understanding of the dynamic response
of the machine because of pavement/structure interaction. Figure 2a shows the schematic
outline of the MLS with four phases of travel of the bogie, and Fig. 2b shows the bogie
dynamic model schematically.
The analysis included the original design concept, where the wheels were driven while in
contact with the pavement surface while traveling along Phase I; but once they had emerged
under the machine, the drive force was switched off during Phases II, III, and IV, and power
was switched on again only once the wheels were about to re-enter under the machine. Other
systems involved the possible options of maintaining contact between the wheels and the
rolling surface for the duration of the full rotational cycle using a steel ramp, and also the
possibility of having the rolling wheels in contact with the ramp surface for only a portion of
the full cycle once it exits from under the structures. Furthermore, the dynamic response was
determined by considering the response of a standard truck bogie.
PHASE III
,) o [o) (o] o [o) PHASE II
@ 0 colD(,
PHASE IV !) I o) {o) (o] 9 Io3 {'~') (oi o ((
NOTE: SOLUTION STARTS @ Xg=O
FIG. 2a--Phases of travel of MLS bogies.
YI
' k~.V~, - AI
COULOMB DAMPER - - " I~ _/ I I"~, " . / TIRE DAMPER
ONLY LOCAL Y MO~IONS ARE ADMITTED
FIG. 2b--Schematic outline of bogie dynamic model.
Cases analyzed thus included inter alia:
l. Variation in the response of the suspension system from full freedom of movement to
limited movement using:
a. A strap to shackle the suspension when the load is reduced.

b. Applying a preload.
c. Provision of a structural ramp for bogies to run on during Phases II, III, and IV of
their movement.
2. Alternative track geometry for the circular Sections II and IV using spiral transitions.
3. Speed variation of the bogies.
4. Variation of the surface profile of the pavement.
5. Variation of the axle load in the different travel phases.
The results are graphically illustrated in Figs. 3 through 8 and Figs. 11 and 12.
Discussion of Computer Analyses

Findings with the computer analyses can be summarized as follows:
1. The dynamic forces with the simulation of the initial proposal were as anticipated except
for the jerk at the entry and exit of Phases II and IV, respectively. This is demonstrated
in Fig. 3. Applicable parameters are indicated in the figure. It is apparent that the forces
vary with time as far as size and frequency are concerned. It was also apparent that
changes in acceleration would cause dynamic impulses to the structure and to the
underlying pavement.
1 - Front axle tire 2 - Front axle suspension spring

3 - Rear axle tire 4 - Rear axle suspension spring
IO.
I
I
O,
EL
I
1
v
I
L,I I
O -10.
FY
C)
LL I I
I I
--20.
I ! I
~HASE I i, PHASE ~ ,II J PHASE I~ II, PHASEI ' I, PHASEI - - - ~

-30.
NOTES: Stropped axles, Preload = 1 0 0 0 0 Ibs

No ramp
V = 2 0 mph
I I I I
4-.8 5.8 6.8 7.8
TIME (sec)
FIG. 3--Dynamics of the MLS--System A [3].

3 - Rear axle tire 4 - R e a r axle suspension spring
tO.
I
I
I
I
I
O_ 1 4,
d . 1 3 1 1 13 I ~ i 13 !
EL
i I I
i
t
.3(
I
t..d
(.J. - 1 0 . I
i
I
FF I I
I I
O I I
I-L I I
-20.
'HASE 1, j ,
PHASE t PHASE III , I, PHASE "I~ ,L
I PHASE I ,!
-30.
NOTES: Stropped axles, Prelood = 5 0 0 0 Ibs

No romp
V = 2 0 mph
I I
4.8 5.8 6.8 7.8
TIME (sec)
FiG. 4---Dynamics of the MLS System B [3].
2. Figure 4 shows how a change in the strapping force (reduced from 44.6 to 22.3 kN)
influences the response.
3. The effect of leaving the suspension unrestrained is demonstrated in Fig. 5, while Fig.
6 shows how this is changed with a reduction in the speed.
4. Figure 7 shows the benefit of having a continuous ramp with the suspension stressed
under a force of 89.2 kN on the top section.
5. The advantage of building in a spiral transition to the curves can be seen in Fig. 8, and
the benefit achieved by structural configuration using transition curve is apparent if
one compares Figs. 3, 7, and 8. It can be seen that the dynamic load in Phase I is very
small and the effect on the structure is also reduced.
6. In order to evaluate the effect of pavement unevenness, data from a typical pavement
surface were modeled into the system. Figures 9 and 10 show the surfaces of the
pavement on a macro and micro scale, and the results of the dynamic analyses are
shown in Figs. 11 and 12. It is apparent that the dynamics owing to the absence of the
transition are drowned within the dynamics due to the pavement surface. Furthermore,
the extent of the dynamics that are likely to be transmitted to the pavement structure
is very similar to those measured by Sweatman in his Australian study cited by Mam-
louk [4].
From the analysis it was thus clear that vehicle dynamics can be controlled sufficiently to
render the MLS an acceptable simulator of truck traffic.
I -- F r o n t axle tire 2 - Front axle suspension sl~ing

3 - R e a r axle t i r e 4- -- R e a r a x l e s u s p e n s i o n s p r i n g
10.
O.
v
LLJ
C) -10.
n-"
O
LL
-20,
v-- " I I
! I
I I
I I
P H A S E 11 '
-30. i
PHASE i , PHASE : PHA~ i [ -JJ-II'HHHH~
, i
NOTES: No strop, No r o m p
-40. V = 20 mph
3
I I I I
4.8 5.8 8.8 7.8
TIME (sec)
HG. 5--Dynamics of the MLS--System C [3].
Structural System
The dynamic analyses showed that no insurmountable problems were to be expected.
However, it was apparent that the deflection of structural elements had to be minimized.
Furthermore, the structure had to be tested for fatigue. More than one structural system was
found to be feasible for making mobility of the MLS possible. The provision of counter-
weight through the installation of water tanks with baffle plates was found feasible, but other
factors may ultimately determine which material is the most appropriate. It would also depend
upon the final design mass of the structure.
Mechanical System
The use of standard bogies was found to be feasible. However, it was decided to use a
single drive system in the first MLS model to minimize mechanical complexities. Some of
the other findings were as follows:
1. It was found that the simulation of braking forces while maintaining motion of the
bogies would be very difficult because of the size of motor required and the spacial
1 -- Front axle tire 2 - F r o n t axle suspension spring

3 -- Rear axle tire 4 - Rear axle suspension spring
10.,
I 2
IA
O.
C~_
1,1 2~ ' 4'

0 -10.
fY
0
I,
-20.
PHASE I Z
I PHASEB ~ P H A S E B t PHASE PHASE I "I
-30.
NOTES: No strop, N o r o m p
V = 15 m p h
I I 1 I
6 7 8 9 10 11
TIME (sec)
FIG. 6---Dynamics of the MLS--System O [3].
(restrictions) limitations of the mechanical/structural system. Accordingly, it was decided

to concentrate initially only on simulation of the road gradient.
2. As was shown by the computer analysis, the dynamic forces were reduced to an
acceptable size and extent for both the continuous ramp and the strapped suspension
system. However, it was considered prudent to use the continuous rolling surface in
order to:
a. Enhance safety and mechanical operation.

b. Improve the power drive system.
c. Reduce noise.
d. Prevent spinout damage to axles.
Traction resistance to simulate gradient was found to be feasible by having the bogies
generate power in the "off-pavement" Phases II, III, and IV. However, the prototype will not
have this feature. The internal friction of the system is expected to simulate a 3% gradient.
A particularly difficult mechanical problem was the transition of the bogie from the structure
onto/and off the pavement surface, the reason being the changing nature of the pavement
surface under traffic. Several options are being explored during the final design phase.
1 - Front oxle tire 2 - Front axle suspension spring

3 - R e o r oxle tire 4 - Reor oxle suspension spring
10.
O.
LO
(D -10.
rY
0
It-
-20. 2
-30.
PHASEII.
I
PHASE n ,1.
I
PHASE rn
'i
.~,
I
PHASE I =~-
I
PHASE I .i
I
NOTES: Ramp, 2 0 0 0 0 Ibs on t o p section
No strop
V = 20 rnph
I I t |
4.8 5.8 6.8 7.8
TIME (sec)
FIG. 7--Dynamics of the MLS--System E [3].
Electrical Configuration
The drive system was studied in detail in order to establish the most beneficial method of
using electric power. It was found that:
l. The electrical system of the MLS could be built primarily from standard components.
2. 150-kW VFD-induction motors were needed for the drive bogies, with only one of the
two axles being driven in order to simulate a 3% gradient.
The power would be provided by using pick-up shoes, and variable frequency motor devices
will be used to control the functioning of each motor. The control system was dependent upon
the final design configuration.
Current Status of the Project
Overall, the project has progressed according to schedule, and it is anticipated that the
design drawings and manufacturing specifications will be completed by the end of 1992. This
should enable the manufacturing of the prototype to be completed by the summer of 1993,
when commissioning could commence.
Several presentations on the project have been made to the Trucking Industry's Maintenance
Council, who have expressed a keen interest and willingness to participate in the project.
H U G O O N T H E TEXAS MOBILE LOAD SIMULATOR 77

3 - Rear axle tire 4 - Rear axle suspension spring
10.
NOTES: Ramp, 2 0 0 0 0 Ibs on top section

Transition spiral length = 3 ft
No strop
V = 20 mph
0,
C~
I I
LLI I I
I I
0 -10o
13s
0
LL
1!47.. I,
I
I,
I
,2
I I I
: .... :
-20. 1 3
2 ; I
I- PHASE E" 'I PHASE ':' PHASE ]E 1 1 PHASEI

-30. : !
I PAVEMENT SECTION--~
I I
I I I
5.2 6.2 7.2 8.2
TIME ( s e c )
FIG. 8--Dynamics of the MLS--System F [3].
2,000
~r ". 1,500 - - - Left Side

,m ', ,RightSide i~
o
o 1,000
o
,ll : , I
--- 500
C
O
= 0
e
m -500
-1,000
0 100 200 300 400 500 600
Distance (ft)
FIG. 9--Macro pavement surface profile.
2000
.... Left S i d e
":.
.E 1 5 0 0 -- Right Side
o
~
o
1000
500
g
.m
>
0
_e
UJ
-500 '" .... " "-J'
-1000 i i I I I I I
120 125 130 135 140 145 150 155 160
D i s t a n c e (ft)
FIG. lO--Selected micro pavement surface profile.
1 - Front axle tire 2 -- F r o n t o x l e s u s p e n s i o n s p r i n g

3 - Rear axle tire 4 - Rear a x l e s u s p e n s i o n s p r i n g
10..
NOTES: Romp, 20000 Ibs on top section
No transition spiral
No strop
V - 20 mph
Pavement section 120-150
0.
CL
v I
I
i i I
LJ I
i I I I
~.~ -10.
Q:: I I I I
0 I I
LL r I 1A i 2 i i
-20.
PHASE In 'l' PHASE I ~ PHASE I " I

-30.
I
PAVEMENTSECllON---J
I t t
4,8 5.8 6.8 7.8
TIME (sec)
FIG. 11--Dynamics of the MLS--System O [3].
1 - Front oxle tire 2 - Front oxle suspension spring

3 - Reor oxle tire 4 - Reor oxle suspension spring
10.
NOTES: Romp, 2 0 0 0 0 Ibs on top section
Tronsition splrol length = .3 ft
No s t r o p
V = 2 0 mph
P o v e m e n t section 1 2 0 - 1 5 0
O.
A
s
I,I
r.) -10.
l"v"
0 I !
Ix_ 1
I
' !
I
: I
I I
I
-20.
a2 2
I I
,
I
i!
1" PHASE " ~ P H A S E m LI, PHASE ]~ i %h PHASE I
-30.
/ i
I
i'-- PAVEMENT SECTION - -
I I I I
5.2 6.2 7.2 B.2
TIME (sec)
FIG. 12--Dynamics of the MLS--System H [3].
Concurrently with the development of the prototype MLS, the model-MLS [1] has been
used to conduct tests on pavement structures. These will be reviewed in the next section.
Pilot Rut Tests Using the ModeI-MLS

As a first step in the implementation of the model-MLS test program, a pilot study was
made of the response of an existing asphalt pavement surface. An attempt was also made to
relate the result of model rut test to static and dynamic creep tests. These tests were carried
out at the Institute for Transport Technology in Stellenbosch, South Africa, using a later version
of the model-MLS. Subsequent testing with model-MLSs is focusing on high- and low-
temperature behavior, as well as fatigue and accelerated aging of the asphalt.
Methodology
An asphaltic concrete surfacing of a parking lot was selected for the test. A temperature
chamber was constructed on the selected test site, and the room temperature was maintained
between 35 and 45~ depending on external weather conditions. Thermocouples were installed
into the pavement at surface level and at a depth of 15 nun below the surface. With the aid
of conventional heater fans, the pavement temperature was maintained within 4-1 of the 40~
test temperature.
The model-MLS was set up and the mean of the dynamic wheel load adjusted to 800 N at
a wheel pressure of 560 kPa. Prior to testing, a transverse surface profile was measured at
two positions using a digital straight edge capable of measuring to 0.1 of a millimetre. Levels
were measured at transverse intervals of 15 mm. During the course of the test, levels were
initially measured at half-hour intervals and subsequently at 1-and 2-h intervals or between
20 000 and 35 000 load applications to ensure that a reliable rut pattern was recorded. Figure
13 shows the maximum average rutting trend in the two wheel paths for the duration of the
test. The results of the transverse measurements of the pavement surface are shown in Fig.
14. The results are the mean of two adjacent transverse profiles. Static and dynamic creep
tests were done on untrafficked core samples taken adjacent to the test site in conjunction
with other conventional analytical laboratory tests.
The static creep test [5] was conducted by applying a static axial load to the cylindrical
asphaltic concrete (AC) samples under the following test conditions:
1. Test temperature 40~

2. Stress--200 kPa.
3. Test t/me--100 mins.
4. Preconditioning--1 min.
The static creep modulus was calculated as
Ec-,t~t = (o'J~,) = (o-JAl) 9 l
where
Uc = compressive stress, kPa,

A= deformation in millimetres,
Ec-,~t = static creep modulus,
r = compressive strain, and
l= sample length in millimetre.
The dynamic creep test [6] was conducted by applying a cyclic load to the cylindrical AC
samples under the following test conditions:
1. Test temperature--40~
2. Stress--lO0 kPa.
3. Load mode----4).5 Hz square wave (1 s load, 1 s rest).
4. Preconditioning--30 cycles at 5% of test load after 2 h preheating.
5. Duration--2 h (3600 cycles).
The dynamic creep modulus was calculated as
Ec-dy. = (o'J~.)= (o'JAl) 9 l
It should be noted that the above test methods differ slightly from those used by Van de
Loo [7] and Cooper and Brown [8].
The results of the tests are shown in Tables 1 and 2. The average gradation of the asphaltic
concrete is shown in Fig. 15.
Discussion of Test Results

As can be seen from Figs. 13 and 14, the plastic deformation of the asphaltic concrete was
primarily due to consolidation of the asphalt. This was in accordance with the open structure
.I
| §
,II
t~
P
| 4-
~11 I I I I II I 9 I I I I I l 4"
" c; ~ .
[ram]ua!l~wJo~O l . ~ u ~ a
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions aut
I
1,000,000 Axles
1.5 130,000 Ax!~
...... 60,000 Axles
0.5
=E
==
o
.,t.., i t#L... 9
g
s -0.5
"J i J
-1
-1.5 I
0 50 100 150 200 250 300 350
Horizontal M e a s u r e m e n t [ram]
FIG. 14---Transverse profile of rutting trends.
of the asphalt, which had a void content of 12%. The low penetration value of the binder,
viz., 20, was evidence of the age of the asphaltic concrete (17 years). Because of the location,
it had been exposed only to car parking during this period of time. It was interesting to find
that it was still so pliable at 40~ despite the age and penetration of 20 (0.1-mm units) of the
recovered binders.
The total rut depth after one million axles was approximately 1.4 mm. This yields a plastic
strain rate of 4.5 x 10 -2 mm/mrrdl06 cycles, which is considerably higher than the results of
HVS tests found previously by the author [9]. In that case, the rutting under an 80-kN wheel
load using the HVS gave a plastic strain of 5.1 X 10 -3 mrn/mm/106 cycles at 25~ However,
it should be noted that the MLS test was at 40~ but at a higher speed than the HVS.
Furthermore, the stress bowl under the 80-kN/axle load was deeper, while the tests were done
at 25~ using an 80-kN wheel load. These facts will have to be taken into account in order
to make a proper comparison of the relative plastic strains. It was anticipated that conventional
shear deformation would start and cause shoving of the asphalt transversely once the void
content had been reduced to some critical level. This was indeed what happened, as can be
seen in Fig. 14. Figure 16 shows a pictorial view of the rutting.
The results of the static and dynamic creep tests were very low in comparison to standard
specifications. This was expected from the MLS rutting response.
In order to relate the rutting to the laboratory creep tests, a "field creep modulus" (Ec-r,ota)
relative to the rutting was calculated as follows
(E~-~ld) = (Average Compression x h)lAh

r~
"a,
o
~
r +++~
~ +
t~
~t
i
~J
,r,,,i
=-
Q
t~
U.S.SjeveSize
200 100 50 30 16 8 4
100 0
90
/'
80
70
e-.
60 40
=3
O.
50 50
/
._~
ell
40 60
E
e-i
30 70
20 J 80
f
10 f
J
J 90
0 100
75 150 300 600 1.18 2.36 4.75 9.5
IJJTI ~ ~ mm v
Sieve Size
FIG. 15---Average particle-size distribution of asphaltic concrete.
where
h = total asphalt thickness, and

Ah = permanent change in layer thickness.
Therefore,
(Ec-nold) =(0.312 X 31.5)/1.4

=7.0 MPa
This compares favorably with the static creep shown in Table 1 and augurs well for future
application of the machine for testing and predicting material behavior.
Operational Observations
The operational effectivity of the MLS model during the ten days of testing is presented
in pie chart form in Fig. 17. This reflects the results of the first extended test of this nature.
FIG. 16--Photograph showing rut profiles under the MLS.
MEASUREMENTS
Qo/^
BREAKDOWI
9% "ICKING
5%
INOPERATIVE
(STAFF UNAVAILABI
3O%
INOPERATIVE
(SERVICING)
17%
PERIOD 10 DAYS
24 h Workday
FIG. 17--MLS pilot Study operational efficiency.
It can be seen that the machine was operational 35% of the time. This increases to 50% of the
time if the time where staff was unavailable is discounted. This was considered very acceptable.
When horizontal traction forces due to wind resistance, braking forces, or gradients are to
be simulated while maintaining constant speed with the MLS, a counterforce has to be provided
somewhere in the system. In the prototype this will be achieved by using the induction
motor in regenerative mode. With the model this can be provided by reaction torque and
system friction.
An analysis of the horizontal drive force of each drive axle of the model-MLS was made
ignoring the electrical losses. It was found to vary between 101 and 136 N/axle, depending
on whether three drive motors or one drive motor was in action. Taking this on average to
be 118 N/axle, this equates to a rolling resistance which simulates a grade of 1.9%. To increase
this, the model would require the wheels to be in contact with a rolling surface at the top
similar to the prototype.
Because of the straddled alignment of the wheels [2], the number of wheel coverages varied
over the width of the wheel track. This is reflected in the transverse profile shown in Fig. 14.
Conclusions Regarding Model-MLS Testing 2

From the pilot study with the model-MLS, it was apparent that it would serve very well
for conducting exploratory tests prior to the application of the full-scale testing machine.
Furthermore, it is anticipated that it will serve as a valuable, inexpensive tool for investigating
appropriate materials and pavement structures in conjunction with the MLS. The machine was
commissioned on August 26, 1993. It will, of course, require further in-service evaluation and
development of technology to enhance the understanding of the behavior of model test pave-
ments. This is already in progress and in fact results were reported at the Seventh International
Conference on Asphalt Pavements in Nottingham, United Kingdom in August 1992. A particular
aspect that proved to be of great importance was the testing of asphaltic concrete after it had
been subjected to accelerated artificial aging by heat [10]. The author considers this to be a
very necessary integral part of accelerated pavement testing.
Acknowledgments
This paper is published with the permission of the Texas Department of Transportation and
the Institute for Transport Technology at the University of Stellenbosch. The author gratefully
acknowledges the support of both Institutions and in particular of the associate executive
director, Roger Welsch, and technical co-ordinators AI Luedecke and Bob Briggs. The dedicated
support and input of the following MLS team members is also recognized: Frank McCullough,
Bill Ward, Mike McNerney, Mike Murphy, Jeff Jackson, Bob Sieberg, Johan MUller, Nick
Reck, Kruger du Plessis, Johan Burger, Riaan de Waal, Anton Cordier, and the project staff
of Stress Engineering Services Incorporated (SES) and VMW Industries Incorporated. A special
word of thanks and recognition is due to Heping Zhang from SES, who was responsible for
the structuring and writing of the software program for the dynamic modeling of the MLS,
and Christo Venter, a student assistant from Stellenbosch, who was responsible for the model-
MLS testing.
2 Since the presentation of the paper, further analyses of the dynamic behavior of accelerated pavement
testing devices have been completed. See report by Michael T. McNerney, Frederick Hugo, and B. Frank
McCullough, "Technical and Economic Analyses of an Accelerated Pavement Test Facility for the Florida
Department of Transportation," Report 997-2, Center for Transportation Research, University of Texas
at Austin, July 1993.
Disclaimer
The contents of this report reflect the views of the author, who is responsible for the facts
and the accuracy of the data presented herein. The contents do not necessarily reflect the
official views or policies of the Federal Highway Administration or the Texas Department
of Transportation.
References
[1] Hugo,F., McCullough, B. F., and van der Walt, B., "Full-Scale Accelerated Pavement Testing for
the Texas State Department of Highways and Public Transportation [now TxDOT]," Pavement
Design, Management, and Performance, Transportation Research Record 1293, Transportation
Research Board, National Research Council, Washington, DC, 1991, pp. 52--60.
[2] Hugo, E, McCullough, B. E, and van der Walt, B., "The Development of a Strategy for the
Implementation of Full-Scale Accelerated Pavement Testing for TxDOT," Report 1246-2F, Center
for Transportation Research, University of Texas at Austin, November 1990.
[3] Young,R., Miller, J., Albert, J., and Zhang, H., "Design of Texas Mobile Load Simulator," report
presented to the Center for Transportation Research, The University of Texas at Austin, by Stress
Engineering Services, Inc., Houston, June 1991.
[4] Mamlouk,M. S., "A Rational Look at Truck Axle Weight," Proceedings, Annual Meeting of the
Transportation Research Board, Washington, DC, 1991.
[5] Committeeof State Road Authorities, "Draft Method for Determination of the Creep Characteristics
of Asphalt Mixtures," Method C6T, Standard Test Method for Road Building Materials, TMH1,
Pretoria, 1990.
[6] Grobler,J. E., "Development of Procedures for Large Stone Asphalt Mix-design," Masters thesis,
University of Stellenbosch, Republic of South Africa, 1990.
[7] Van de Loo, E J., "The Creep Test: A Key Tool in Asphalt Mix Design and in the Prediction of
Pavement Rutting," Proceedings of the Association of Asphalt Paving Technologists, Vol. 47, 1978.
[8] Cooper, K. E. and Brown, S. E, "Development of Simple Apparatus for the Measurement of the
Mechanical Properties of Asphalt Mixes," Proceedings, Eurohitume Symposium, Madrid, 1989.
[9] Hugo,E, "Surface Cracking of Asphalt Mixtures in Southern Africa," Ph.D. dissertation, University
of Texas at Austin, 1984.
[10] Van der Merwe, C. J., Theyse, H. L., Horak, E., Hugo, E, and Du Plessis, J., "Evaluation of the
Rehabilitation Design of a BTB Pavement and the Effects of Artificial Aging Using Accelerated
Wheel Load Testing," Proceedings, 7th International Conference on Asphalt Pavements, Notting-
ham, UK, August 1992.
JCrgen Kraruf
Measured and Calculated Pavement

Response in the Danish Road Testing
Machine
REFERENCE: Krarup, J., "Measured and Calculated Pavement Response in the Danish
Road Testing Machine," Vehicle-Road Interaction, ASTM STP 1225, B. T. Kulakowski, Ed.,
ABSTRACT: This paper reports results from a research program assessing pavement response
from heavy vehicles. The Technical University of Denmark and The Danish Road Institute
cooperated in the research program. To estimate the validity of theoretical values of stress and
strain in pavements, stress and strain distributions were measured under two types of loads.
One was the rolling truck tires of the Road Testing Machine, and the other was falling weight
deflectometer (FWD) loadings. Distributions of stress and strain are explored and mapped in
three-dimensional graphs. Some steps are taken in an attempt to test measured response distribu-
tions against calculated distributions. The measured responses are: (1) vertical contact stress in
the tire/pavement interface, (2) longitudinal and transversal strain at the bottom of the asphalt
concrete layer, (3) vertical soil stress, and (4) vertical soil strain.
As temperatures and layer thicknesses in the Road Testing Machine are well known, only
Poisson's ratio is estimated for the calculations. The layer's E-values were backcalculated
from the FWD load and measured surface deflections. Using the elastic layer theory with
backcalculated E-values and the chosen values of Poisson's ratio, values of stress and strain
are calculated at the positions of the instruments. FWD load response values are considered
the means for in situ calibration of stress cells.
KEY WORDS: pavement, test, instrumentation, measurements, stress, stress analysis, strain,
Denmark, research project
Pavement design is in the process of changing from an empirical craft into an engineering
science. A lot of success has already been achieved, but persistent problems still exist. The
above-mentioned process involves a solid need for measurements verifying the reliability of
mathematical models used for calculating pavement response (stress, strains, and deflections).
When models become valid for future materials and load conditions, they will provide substan-
tial benefits by design of optimal pavements.
In the field of pavement engineering, measurements and experimental designs are costly
and long lasting. The significant parameters are material properties, traffic load distributions,
and environmental variations, all of a complex nature. In the Danish Road Testing Machine
(RTM), it is possible to control several parameter levels. As an example, the full-scale pavement
model can be kept in a regime of constant temperature and moisture content while pavement
response measurements are conducted at different levels of load from a well-defined load
configuration (tire or F W D loading plate). Basically the difference between measured and
Senior research engineer, Danish Road Institute, National Road Laboratory, Elisagaardsvej 5, Roskilde,
Denmark DK-4000.
89
calculated values becomes the issue for evaluation and discussion of the validity and reliability
of models and measurements.
The Danish Road Testing Machine (RTM)

The RTM is an indoor installation for monitoring test pavement response and performance.
It is located at the Technical University of Denmark and is owned by the Danish Road Institute.
It is a linear track facility applying loads in both directions. The present instrumentation for
response measurements records contact stress, horizontal asphalt strain, vertical stress, and
vertical strain in the unbound layers. Pore water pressure, water content, and temperature is
also measured. This paper reports measured stress and strain response data from contact stress
cells, asphalt strain gages, and soil deformation transducers. Loads are applied by rolling truck
wheel configurations or FWD loading plate.
To vehicle engineers, the FWD can be described as a device simulating load events of truck
tires on a pavement by dropping a heavy load on a load-distributing plate located on the
pavement. The pavement deformations responding to the load create a basin in the plane
pavement. The basin is measured by seven accelerometers (geophones) positioned along a
radial line from the center of the plate to 1.50 m from the center. As the thickness of each
pavement material layer is known, layer elastic theory is applied in an iterative procedure
(backcalculation)--when the calculated basin fits a measured basin, the stiffness (E-value) of
each layer is found. Strong pavements with long residual lifetime possess high stiffness,
whereas deteriorated pavements in need of rehabilitation possess lower E-values.
The test pavement was built in a concrete pit 27 m long, 2 m deep, and 2.5 m wide. It has
a 9-m-long test section where the instruments are located and where the load carriage, simulating
one half of a single axle, travels at a constant speed. The maximum speed is 25 km/h, and
the maximum hydraulically controlled axle load is 130 kN. In spring 1992 a new wide-base
single tire assembly was manufactured and tested; until then, the dual wheel was the only
moving load applied. The dual tires are 2 • 12 R 22.5, and the wide base tire is 445/65 R
22.5 with axle load capacities of 116 and 112 kN, respectively.
The present full-scale test pavement is a thin flexible pavement (local road) made of the
following layers: 60 mm asphalt concrete, 130 mm granular base, 390 mm granular subbase,
and 1000 mm nonplastic untreated sandy till.
The response instrumentation layout comprises four identical groups of gages (one is shown
in Fig. 1). The type of gages and their signals are presented in the following section.
Measuring Strains at the RTM

Asphalt Strain
Measurements of horizontal strain response at the bottom of the bituminous layer are made
by means of asphalt strain gages (ASG). In the RTM, twelve ASGs are installed, eight
longitudinal (in the travel direction) and four transversal. The ASGs were produced in the
RTM workshop and are the H-shaped type (Fig. 2). Because the acrylic bedding of common
asphalt gages, according to their manufacturers, is not intended for temperatures normal during
asphalt paving, a gage embedded in epoxy is used [1]. The strain gage is glued to a flexible
strip of polyester film possessing a very low modulus of elongation compared to the strain
gage itself. In both ends the strip is fixed to stainless steel anchors 7.5 cm long. The initial
distance between the anchors is approximately 10 cm. The 120-fI strain gage itself is a precision
device having a "gage factor" equal to 2.0 from delivery with accuracy better than what
could be obtained by a workshop calibration. Consequently, no efforts were made to reassess
the instrument.
KRARUP ON CALCULATED PAVEMENT RESPONSE 91
Ir'lll.//lI/l./lI, E///L/,~ AsphaltConcrete

(Pavement Surface)
"~' ]~ ~c,, "=" _ ' ~ ' "=' Base
ASG- Asph. Strain Gauge
SOT14 ~ SPCIO
SPC - Soil Pressure Cell
="'* Subbase
SDT - Soil Deformation $
Transducer
$OTI~
Subgrade
Station (cm)
- 240 -lO0 -120
C o n c r e t e Test
Pit B o t t o m
FIG. 1--One of four identical groups of gages (cross section in the travel direction).
FIG. 2--Asphalt strain gage.
Figure 3 shows three-dimensional plots generated by software made by the author. The
plots show distributions of strain in the driving direction recorded under dual tires at 100 kN
axle load and 700 kPa inflation pressure, and under a single wide-base tire at 100 kN and
800 kPa. Values for the first half of the distributions are recorded and then mirrored to give
the full picture of symmetrical distributions.
The transversal asphalt strain from load series (shown only for one of the dual tires) can
be seen in Fig. 4. Note the compressive strain between the dual tires and the remaining
transversal tensile strain behind the zone with tensile strain.
In its standard setup, the data acquisition system reads each gage 208 times per second,
corresponding to one reading every time the wheels travel 27 mm at 20 km/h. The wheel
FIG. 3--Distributions of longitudinal strain from dual wheel and wide-base single tires (Gage
ID: ASG5).
FIG. 4---Distributions of transversal strain from dual wheel and wide-base single tires (Gage
ID: ASG6).
kPa 500
Legend:
* Contact stress
4o0 from FWD load (in kPa)
13)
i
IZI
o Strain in asphalt
m, 300
o)
concrete in microstraln
+,.,, (i.e. 10 .6 m/m)
C
"..,s
o Asphalt layer thickness
'~
t,-
20o at the ASG position
0
(in mm)
t,-
M
r
~,. tO0
X
O)
0 I i i i i i i i i
3 4 5 6 7 8 9 10 11 12 ASG number
FIG. 5--Strain under FWD load. Response from twelve asphalt strain gages.
travel line can be moved sideways in 22-mm steps. All measurements reported in this paper
were conducted this way, and when nothing else is specified, the graphs are single gage signals
without curve fitting or the averaging of signals from more instruments.
As the stress condition and variations from moving wheel loads on the pavement surface
are rather uncertain, the FWD load was applied as precision load to check the measured peak
value from each ASG. Figure 5 shows the values from the twelve ASGs.
Soil Strain
Vertical strain response in the unbound layers is recorded by soil deformation transducers
(SDT) using common LVDTs shown in Fig. 6. In the unbound layers, vertical strain response
FIG. 6---Soil deformation transducer.
distributions have been recorded for both tire configurations. Distributions recorded in the
unbound base are given in Fig. 7. At the end of the 3 s recording time, the residual "permanent"
strain remaining after the resilient response is 10 9 10 -6 m/m or 1% of the peak value of 939
9 10 -6 m/m for the dual wheel tires. The "permanent" strain is expected to be superposed by
contributions from the axle load series of the individual vehicles. The indoor pavement facility
is not capable of loadings from real trucks, but the Danish Road Institute has plans for
instrumentation of an in-service highway pavement section, which might give answers about
response phenomena like accumulating response from true traffic loads.
Measuring Stress at the RTM
Contact Stress (Between Wheel and Pavement Surface)
Three contact stress cells (CSC) were installed in the RTM pavement. A CSC is a double
diaphragm pressure cell for registration of vertical stress at the fire/pavement interface; this
principle is illustrated in Fig. 8. The material of the cell is the inactive metal titanium. Due
to the ratio of the size of the two membranes, only very little deformation of the surface
membrane is needed for measurement. The cell itself, therefore, acts like a stone in the
material where stress values are measured. The space between the membranes is filled with
incompressible hydraulic liquid. A strain gage rosette produced for this particular purpose is
glued to the back of the smaller membrane. The thickness of the surface membrane is 2 mm,
and its diameter is 50 mm to fit the tread of the 12 R 22.5. (Additional CSCs for the wide-
base tire tread are manufactured but not yet installed.)
It must be emphasized that the cell is calibrated statically, having the total area of the 50-
mm-wide surface membrane in full contact with the stressing material. In this state the
deformation of the membrane is axisymmetrical. If the load does not cover the full membrane
area, the signal from the CSC will indicate lower stress than the actual contact stress. As the
tread of the tire has a pattern, problems arise when full contact is not gained. The three 50-
mm-wide bands of the pattern will give good readings, but values read between full contact
zones will be rather uncertain. Fortunately, peak values of contact stress are assumed to be
located in the middle of the full contact bands of the tread. Therefore, the peak value readings
will be the most reliable values. The gage reading frequency for CSC measurements is 526
readings per second per cell, so the signal is digitized and stored every time the wheels have
traveled 11 mm. CSC measurements are taken for wheel path sideway steps of 11 mm also.
Figure 9 shows two examples of the signals from only one CSC, describing stress distributions
under one of the dual tires.
One of the CSCs was tested at various proper stress levels using a Dynatest 8000 falling
weight deflectometer. Between the plate and the CSC a thin, water-filled rubber bag in a steel
frame was attached to the FWD foot plate to ensure uniform stress distribution. In Fig. 10,
the FWD load is kept close to 330 kPa, and the F W D foot plate is moved in 5-cm steps across
the CSC in two perpendicular directions, giving twelve positions where the falling weight was
activated. It is satisfactory that the signal mean from the CSC was 5.5% above the FWD
stress signal.
Soil Pressure Cells
Vertical soil stress is measured by soil pressure cells (SPC). The SPC design is similar to
the CSC, though the dimensions of the membranes are designed to meet the expected stress
ranges. The individual SPC is calibrated in the laboratory under static stress levels in the
material of the layer in which it will be installed. The static calibration stress values are read
FIG. 7--Vertical strain distributions in the unbound base. Response from dual wheel and wide-
base single tires. (Gage ID: SDT16--because of electronic signal noise, the curve values are
"moving average" of five.)
~ S u f f a c e Membrane
IncompressibleU q u i d
__ _ . . ~ / / ~ l ~ t 4ncrete~ . . . . ~ ~ ~ LittleMembrane
~,, . . . . . . . . . . . . . _.ff,_d ,.~......-- RosetteStrain Gauge
: - 7 / Z A -
R"~L~\\ \ \ \ \ \,L'\ \ \\\ \ \ \ \~,-r ~Jaue

FIG. 8--Cross section of contact stress cell (CSC).
on a precision manometer. Stress distributions recorded in the unbound granular base can be
seen in Fig. 11.
The maximum stress values from the wide base tire are approximately 30% higher than
those from the dual wheel tire. Ratios of single/dual maximum values for a number of relevant
responses are listed later in this paper.
Measured and Calculated Response

The shown response distributions do not possess axial symmetry, and from comparisons of
values from different gages of identical type and from the same depth, peak values are observed
to vary by a factor of two. The variation may be caused by the dynamics of the traveling
wheels and structure nonhomogeneity. To get response data suitable for comparison to theoreti-
cal values, the FWD loading was applied as a precision load. The load cell of the FWD is
specified to read the FWD load within an uncertainty of 5% of the peak value. Again the
water-filled bag was attached to the underside of the 300-mm-diameter FWD foot plate to
ensure uniform stress distribution.
FWD Loadings and SPC Readings

The FWD weight was adjusted to deliver a load, P~VD, in the range from 35 to 40 kN. A
series of response measurements were then carried out during F W D drops at every 15 cm of
the test pavement. The stress data from eight SPCs (from the unbound base and the subgrade)
were integrated numerically assuming axial symmetry of the stress distribution, giving the
values of the vertical reaction, Psrv, and balancing the FWD load (Table 1). The ratio, PFWD/
PsPc, is considered an in situ calibration coefficient needed to compensate for nonhomogeneous
material surrounding the cells. The eight coefficients have a mean deviation from unity of 0.13.
Response Calculation
As input to the elastic layer theory program BISAR [2], the thickness values of the individual
layer are known from construction phase rod and level readings taken in steps of 60 cm along
the line of instruments and positions of SPCs from leveling during installation of gages. The
E-values were assessed by FWD measurements in steps of 30 cm and load levels of 35 to 40
kN. Backcalculations were carried out with both ELMOD [3] and MODCOMP3 [4] with
Poisson's ratio fixed at 0.35 for all layers. In pavement analysis, 0.35 is the traditional default
value for Poisson's ratio. ELMOD calculations were carried out with equivalent depth to
bedrock (bottom of concrete pit) fixed at 2200 mm, where MODCOMP3 by itself computed
a higher E-value for the bottom layer than for the layer above the bottom layer. Table 2 is
FIG. 9--Measured contact stress distributions from radial tire 12 R 22.5:100 and 80 kN axle
loads and 700 and 900 kPa tire pressures, respectively.
kPa 500
400
_= 3OO
8
10
-i 200
100
0 ! 1 ! ! i ! i ! i i
1 2 3 4 5 6 7 8 9 10 11 12 Position number
aFWD .CSC
FIG. lO---Measured contact stress compared to applied F W D stress.
not intended for comparison of ELMOD against MODCOMP3; it simply shows the different
moduli obtained with two different settings of theoretical boundary conditions.
As can be seen from Table 2, the standard deviations are large, but the same variation of
E-moduli along the instrumentation line is seen in some 30 measurements over a two-year
period. The variation must be caused by surface and structure nonhomogeneity (in evenness,
thickness, and density). The variation supports the general warning against FWD measurements
on thin pavements for pavement-managing purposes.
In this project the FWD E-values were supported by laboratory material testing. The asphalt
layer is composed by 4-cm asphalt concrete and 2-cm wearing course. Material sample beams
compacted in the laboratory were tested in a three-point bending machine leading to master
curves. At 10~ and a load cycle time matching a wheel speed of 40 km/h, the E-values of
the AC material was 11.000 MPa and for the wearing course 5.500 MPa. Finding FWD-
measured E-values for the total thickness of the asphalt layer at 7000 MPa when back-calculated
with ELMOD and 5200 when back-calculated with MODCOMP3 is considered acceptable in
comparison with the E-values found from bending machine testing.
With the in situ calibration coefficients (PFwD/PsPc), the reaction from one of the tires of the
dual wheel was calculated to 32.4 kN, 30% more than the static load of the wheel, and the
single tire reaction was computed to 38% more than the static wheel load.
The ratios in Table 3 show obviously that the choice of boundary settings for backcalculation
is less significant than the general deviation between applied load and the reaction calculated
by BISAR and numerical integration of stress distributions. The FWD load data from the
series of 15-cm step loadings were input loads to BISAR.
Better solutions for computing load/reaction ratios will be studied in the future. Maybe
smaller steps in the FWD loading series could provide a better response. Figure 12 shows
measured (raw data) and calculated response from two SPCs in the unbound base.
FIG. 11--Vertical stress distributions in the unbound granular base. Dual wheel and wide-base
single tires. (Gage ID: SPC14).
TABLE 1 Load and reactionforces.
PFWDLoad, Pspc Measured

SPCs kN Reactions, kN PFwD/PsPc
Unbound base layer:

SPC13 38.7 58.8 0.66
SPC14 36.7 34.8 1.05
SPC15 35.8 32.7 1.10
SPC 16 34.9 43.4 0.81
Subgrade layer:
SPC1 38.3 39.3 0.97
SPC2 37.3 45.3 0.82
SPC3 35.8 38.8 0.92
SPC4 35.0 37.6 0.93
TABLE 2--Backcalculated E-values, in MPa.
ELMOD MODCOMP3
Std.
Standard Std. Dev./ Standard Dev./
Layer Average Deviation Average Average Deviation Average
Asphalt 7001 2160 0.31 5178 2482 0.48

Unbound base 307 22 0.07 478 98 0.20
Subbase 189 14 0.07 102 9 0.08
Subgrade 131 8 0.06 297 30 0.10
TABLE 3--FWD load and B1SAR reactionforces."
Positions for PCE Calculated PcM Calculated

BISAR PFWDLoad, kN Reaction, kN, Reaction, kN PcE/PFwD PcM/PFwD
Base layer:
SPC13 38.7 39.8 40.9 1.03 1.06
SPC14 36.7 38.4 39.6 1.05 1.08
SPC 15 35.8 37.4 38.6 1.04 1.08
SPC 16 34.9 36.0 37.7 1.03 1.08
Top of subgrade:
SPC1 38.3 35.8 36.3 0.93 0.95
SPC2 37.3 34.5 35.2 0.92 0.94
SPC3 35.8 33.7 34.5 0.94 0.96
SPC4 35.0 32.6 33.5 0.93 0.96
~Index C for Calculation, E for ELMOND, and M for MODCOMP 3.
Response Equivalency Factors

Response measurements also supply pavement engineers with information about response
equivalency factors. Equivalency factors may be considered a damage index ranking different
tire configurations with respect to deteriorating effects and response parameter. The equivalency
factors are important to the road authorities in decision making about consequences of changes
in limits for legal weight and dimensions for vehicles. Table 4 lists peak values from the same
kPa
700- SPC 14
D Measured
600- . BISAR (E)
o BISAR (M)
500-
400-
300-
200-
100-
0, I ~" I ~ ! ~ I 11 I .,IPm ! i.
-300 -280 -260 -240 -220 -200 -180 -160 -140 -120 -100
FWD positions, cm
kPa
700- SPC 15
600-
500-
400-
300-
200 -
100-
O~ ,=, ~. ~-, , , , , , , .~ ,~ ~ ,r r ,=
0 20 40 60 80 100 120 140 160 180 200
FWD positions, cm
FIG. 12--Vertical stress, measured and calculated at positions of the pressure cells SPCI4 and
SPC15 in the unbound base.
~c-,I
.~
Ou")
'No~ u"~ t ~ ,,_;
o [...~ ~
N<~
o,No
"1~ ~
,,.-.~ 0 0
"~.~
,-1
~t--I
~'~.~
wheel load series of dual and wide-base single tires for Figs. 3, 4, 7, and 11. Values of
traditional critical parameters are listed.
Comment
It must be emphasized that the results come from a physical model, a full-scale pavement
model of one half of a local road pavement with only 60 mm of asphalt concrete. The test
pavement location in a concrete pit disturbs the stress and strain domain. There are many
sources introducing uncertainty to the complex sets of measurements. May the presented results
be received and treated with this in mind.
Conclusions
For pavement engineers analyzing consequences of changes in vehicle legal weight and
dimensions, this paper delivers measured response equivalency factors comparing different
response parameter peak values from dual tires and wide-base tires. The observations reveal
equivalency factors in the range of 0.98 to 1.64.
With respect to the wide range of equivalency factor values, future research on critical
response parameters for pavement design should emphasize the study of strain energy calcula-
tions locating the centers of deteriorating energy under different load distributions.
Graphical presentations of different response distributions have been produced in three-
dimensional plots generated from measured time series by software made by the author.
By the means of the FWD device, precision loads were produced for the purpose of calibrating
in situ soil pressure cells.
References
[1] OECD Scientific Expert Group 12, "Strain Measurements in Bituminous Layers," OECD Road and
Transport, Paris, France, 1985.
[2] Shell, "BISAR-PC" manual, Shell International Petroleum Company Limited, England, Septem-
ber 1987.
[3] Dynatest Engineering A/S, "ELMOD/ELCON User's Manual (ELMOD 3.2)," Dynatest Engineering
A/S, Vedbaek, Denmark, 1989.
[4] Irwin, L. H., "User's Guide to MODCOMP2," Cornell Local Roads Program Report No. 83-8,
Cornell University, Ithaca, NY, November 1983.
Tire Characteristics
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproduc
lan R. Gelling 1
Influence of Tread Polymer on Traction,

Rolling Resistance, and Wear
Properties of Tires
REFERENCE: Gelling, I. R., "Influence of Tread Polymer on Traction, Rolling Resistance,
and Wear Properties of Tires," Vehicle-Road Interaction, ASTM STP 1225, B. T. Kulakowsld,
Ed., American Society for Testing and Materials, Philadelphia, 1994, pp. 107-118.
ABSTRACT: Tire construction is a major factor in determining performance, but the polymeric
composition of the tire tread also influences performance. Traction, rolling resistance, and to
some extent wear properties can be related to the viscoelastic properties of the tread material.
Rolling losses and skiding phenomena occur at different temperatures and frequencies, and
hence it is the hysteresis properties of the tread compound under these conditions that are
important. General purpose polymers such as SBR, NR, and BR do not fulfill these requirements,
though blends are a compromise. Ideal hysteresis requirements for a tire tread are described.
High-vinyl butadiene and styrene butadiene rubbers and epoxidized natural rubber are evaluated,
and it is shown that these rubbers succeed to some extent in meeting the requirements of the
ideal polymer.
KEY WORDS: tire tread rolling resistance, traction and wear properties, ideal polymer hystere-
sis properties, vinyl butadienes, vinyl styrene butadiene, epoxidized natural rubber
Tires are the interface between vehicle and highway, and property requirements can be
placed in three categories: (1) safety, tire integrity, handling/traction; (2) economy, rolling
resistance, and wear; (3) comfort, and acoustic and mechanical damping. The relative impor-
tance of these factors has varied with geographical location and over time.
The major factors that influence these properties are tire construction and highway surface.
The greatest influence on tire properties in the last few decades was brought about by the
introduction of the radial tire, which resulted in higher mileage, lower rolling resistance, and
better handling properties due to smaller slip angles for a given side force. Within a particular
property category, the polymer composition of the tread compound can be a minor determining
factor. This is illustrated in Fig. 1 for wet traction properties measured by braking characteristics
[1]. The tread pattern influences stopping distances by a factor of eight on going from a "slick"
to a standard tread pattern, and the composition of the highway can have a five-field effect.
In contrast, the polymer composition has only a maximum contribution of 1.5.
For a medium-sized car at a constant speed of 80 kin/h, about 30% of the available mechanical
energy is dissipated by the tires. For a radial tire, 61% of the energy loss is in the tread region
[2] (Fig. 2). The composition of the tread compound is a factor in determining tire rolling
resistance and hence the fuel economy of the vehicle [3,4]. Observed fuel savings of freight
and passenger vehicles are shown in Fig. 3.
Head of Dry Rubber Applications Technology, Malaysian Rubber Producers' Research Association,
Brickendonbury, Hertford, SGI3 8NL, United Kingdom.
107
Copyright 9 1994 by ASTM lntcrnational Www.astIII.OI'g
108 VEHICLE-ROAD
INTERACTION
Level of effect
Tread pattern 8 :1
Road surfacecomposition 5 :1
Speed 10 :1
Brakingsystem 3 :1
Tread material 1.5:1
FIG. 1--The major factors that influence the wet breaking traction of tires at high speeds.
-- ~ o 7 Treadband61%
~ i / o 4,,o/ Sidewall
~._~ 14% B:gidn14%

FIG. 2--Distribution of energy losses in radial tires.
Although many factors affect tire performance within one tire type, the hysteresis properties
of the tread compound dominate. The carbon black type and loading, the oil level, and the
cross-link density all contribute to the hysteresis profile [5], but the characteristics of the tread
polymer are the basic controlling factor. Glass transition temperatures (Tg), even though they
are a static measurement, are an indicator of properties. Tire tread wear is a complex phenome-
non, and two basic mechanisms have been identified: tensile [6] and fatigue [7] modes. The
dominant mechanism depends on the road surface and driving conditions. Under low-severity
wear conditions, the fatigue mechanism dominates, but at high severities a tearing-type mecha-
nism operates and can be related to the Tg of the polymer (Fig. 4). The lower the Tg of the
tread polymer, the higher the abrasion resistance.
Tires operate under dynamic conditions, and thus the dynamic glass transition, i.e., the
position of the maximum of the loss peak and the overall shape of the hysteresis temperature
profile, is a more appropriate parameter. Requirements for high wet traction are opposed to
GELLING ON INFLUENCEOF TREAD POLYMER 109
Fuel
consumption
g/km
0.6 --
0.5 --
0.4-
0.3-
0.2
Car (a)
0.1
I I I
0.5 1 1.5 2
Rolling loss, KN
FIG. 3--Relationship between rolling tire resistance and fuel consumption: (a) for a medium-
passenger vehicle [3], four tires, and (b) a heavy truck [4], eighteen tires.
those of wear, as the former requires high hysteretic properties. The relationship [8] between
wet traction and peak loss angle (tan ~5)is illustrated ia Fig. 5, i,e,, the wet traction of elastomers
increase with increasing tan ~ value. Thus, those rubbers which exhibit high wet traction
properties will have poor wear resistance and vice versa (Fig. 6). For low rolling resistance,
the tread polymer should be a low hysteresis material. Rolling resistance increases with the
hysteresis of the tread polymer. This is illustrated in Fig. 7 where peak tan ~ values of various
tread compounds, differing only in the type of polymer, are plotted against the rolling resistance
of tires treaded with these compounds.
Rolling resistance and wet traction occur at different frequencies and temperatures: rolling
resistance at low frequency and moderate temperatures, and wet traction at high frequencies
and high temperatures. Laboratory hysteresis measurements carried out at frequencies of 1 to
10 Hz can be related to the various tread properties, as illustrated in Fig. 8.
At temperatures above 30~ the loss angle should be low to minimize rolling resistance
and heat buildup, whereas between 0 and 30~ the requirement is for high hysteresis to
maximize wet traction. The peak in the curve should be at a low temperature for abrasion
resistance. The general purpose rubbers-natural rubber (NR), styrene butadiene (SBR), and
butadiene (BR) rubber-that have traditionally been used in tread compounds do not fulfill all
these requirements. NR exhibits the desired low rolling resistance and low heat buildup
Wear mm/1000 miles
0.6
0.5
0.4
0.3
0.2
I I I I I I 9
-110 -100 -90 -80 -70 -60 -50
Tg of polymer, ~
FIG. 4---The influence of polymer glass transition temperature on tread wear
properties but poor wet traction. The reverse is the case with SBR. The hysteresis profile of
the ideal polymer has been described (Fig. 9) that would exhibit the desired tread properties.
This paper describes the tread properties of a number of vinyl butadiene and styrene butadiene
rubbers and epoxidized natural rubber.
Experimental
Materials
The vinyl butadiene and styrene vinyl butadiene rubbers were obtained from the lapan
Synthetic Rubber Co. Ltd. (Rubbers 1 through 4, Table 1) or Hills (Rubbers 5 through 8, Table
1). The SBR was a standard commercial SBR 1500 and the NR an SMR 10. The epoxidized
natural rubbers (epoxyprenes) were produced by Kumpulan Guthrie Berhad in Malaysia. A
standard formulation was employed to evaluate the polymers: ISAF N220 carbon black, 45
phr; Dutrex 729 process oil, 5 pphr; zinc oxide, 5 pphr; stearic acid, 2 pphr; Santoflex 13, 2
phr; sulphur, 1.4 pphr; N-oxydiethylenebenzothiazole-2-sulphenamide. A standard mix cycle
was employed and the compounds were vulcanized to optimum cross-link density at 150~
for laboratory specimens. New 165 by 13 steel radial tires, whose tread had been buffed off,
were retreaded with the test compounds by conventional bead-to-bead remolding at 150~
Testing
Rolling Resistance--Rolling resistance was measured from the torque of the test tires under
load on a Heenan Fraude test rig less the torque with the same tire in skimming contact with
GELLING ON INFLUENCE OF TREAD POLYMER 1 11
Wet skid
resistance
65
T 55
45 9
o••e
OSBR
VBR (50%)
VBR (72%)
35
25
15- BR I I I J
- .5 -1.0 -0.5 0 0.5
log tan8
FIG. 5--Relationship between the dynamic loss angle of polymers and their wet skid properties.
Loss angles, tan ~, were measured at O~ and 110 Hz. Wet traction was measured on a pendulum
tester (Stanley, London, UK).
W e a r rating
160 -
140 -
120
100
I I I i
85 90 95 100
Skid rating
FIG. 6--Relationship between wear properties and skid resistance for tire tread compounds.
Rolling
Resistance
Rating
110
100
90
I I I
0.1 0.2 0.25
tank
FIG. 7--The influence of polymer loss angle (25~ and 10 Hz) on tire rolling resistance in
otherwise identical tread compounds.
the wheel (1.71 m diameter). Initially, all tires were conditioned for 60 min at 80 km/h and
80% of rated load. Inflation pressures were set at 248 kPa (cold). The tires were run to
equilibrium temperature (45 min), and three torque measurements were recorded at 15-min
intervals for each of two speeds, 48 and 80 km/h. The rolling resistance was calculated from
the average of the three torque readings.
Wet Traction Testing--The 165 by 13 steel radial test tires were fitted to a two-wheel
Schallamach trailer with the wheels at zero slip angle. The trailer was towed at 40 km/h over
a wet (minimum water depth 2 mm) standard asphalt surface and the trailer braked to bring
the whole vehicle to a halt. The braking distance and forces were automatically recorded.
A set of control tires were run between each test sequence to eliminate any variation in
track conditions.
Wear Testing--Accelerated wear testing was carded out employing a Schallamach trailer
(weight 1000 kg) with the wheels set at a slip angle of 1.5~ For each test run, a test and
control tire were fitted to the trailer and wear determined from the weight loss of the tires
after deflation. The same route was used in all tests with the position of the tires reversed for
the return leg.
Laboratory physical testing was carried out employing the following standard procedures:
Hardness to ISO 480 [ASTM Test Method for Rubber Property--International Hard-
ness (D 1415)]
GELLING ON INFLUENCE OF TREAD POLYMER 1 13
tans
1.0 I-
I
I rg
I
101 I-
I
10-2 -
/
8
=
"9
"
E~E
| 1 7 4 0}
.o
t~ oc~ " e-
lO-3 t___J__ b_J___

- 100 -50 0 50 100
Temperature~
FIG. 8--The temperature hysteresis profile of a SBR1500 tire tread (1 Hz) and its relationship
to tread properties.
Tensile strength to ISO 37.

Resilience, Dunlop tripsometer to BS 903: Part A8.
Abrasion DIN to ISO 4649.
Laboratory wet traction data were obtained with a laboratory pendulum skid tester (Stanley &
Co., London) employing a wet, smooth concrete surface.
Results and Discussion
The epoxidation of NR is achieved by reaction with peroxycarboxylic acids [9], and a range
of materials can be readily prepared covering a wide range of epoxide contents (Fig. 10). Each
one mole percent increase in epoxidation raises the Tg by approximately one degree celsius
from that of NR [10].
The epoxidation reaction is stereospecific, and the all cis 1,4-structure of NR is retained.
ENR, like NR, thus undergoes strain crystallization and hence exhibits high-strength properties
[10]. The epoxide groups are randomly dispersed along the NR backbone [11].
The increase in Tg on epoxidation would be expected to increase the wet traction properties
of treads built with these materials. At 25 mol% epoxidation, ENR-25, the Tg is -47~ and
the expected increase in wet traction is observed. Figure 11 compares data obtained for radial
passenger tires on a wet asphalt service. Traction increases with the extent of epoxidation,
and at 25 tool% epoxidation and beyond these treads are superior to those based on oil-
extended SBR.
tan~
1.0
I I I I I I
,~162tg~~162 X'**J IflepIpolympr I

10-1 i i i
7; #/l"J-~" 'L-"kL~ I I
10-2
i
I
I
I
I
10-3
, h I , ,
-100 -50 0 50 100
Temperature *C
FIG. 9--The hysteresis profile (1 Hz) of an ideal tire tread polymer [121.
TABLE 1--Structures of vinyl SBRs and BRs.

Rubber 1 2 3 4 5 6 7 8
Styrene, % 24 24 28 15 0 20 15 25
Butadiene
cis, % 20 20 20 16 . . . . . . . . . . . .
trans, % 40 40 40 27
vinyl, % 40 40 40 57 5()" ~" 45" 55"
NR Epoxidlzed NR
FIG. lO--The epoxidation of natural rubber employing a peroxycarboxylic acid.
GELLING ON INFLUENCE OF TREAD POLYMER 115
126 124
121 123
FIG. 11--Wet traction properties of epoxidized natural rubbers compared to NR and oil-extended
SBR. Testing was carried out employing a Schallamach trailer as described in the experimental sec-
tion.
Rolling resistance measurements were carried out on the same series of tires, and the results
are recorded in Fig. 12. The rolling resistances of the ENR treaded tires were much lower
than expected from the Tg values. The reason for this combination of properties can be
ascertained from the hysteresis temperature profile (Fig. 13). Compared to general purpose
119
114
OESBR 103 102 100
ENR-30 I
93
ENR-25 ENR-20 NR II
ENR-25
silica
FIG. 12--Rolling resistance of radial passenger tires retreaded with various polymers. Details
of the experimental procedure are recorded in the experimental section.
Tan b
0.25 ENR-25
NR OESBR
0.15
0.05 I I I I I
20 40 60 80 100
Temperature ,~
FIG. 13--Hysteresis temperature profile of ENR-25 compared to those of NR and oil-extended SBR.
rubbers, the hysteresis of the ENR-25 compound has a much greater temperature dependence
over the critical region.
At low ambient test temperatures, the wear properties of ENR-25 treaded fires are inferior
to those of NR (Fig. 14), but as the test temperature moves further away from the Tg, wear
resistance increases.
High-vinyl butadiene rubbers and copolymers with styrene have also been examined as tire
tread compounds [12,13]. The materials examined are recorded in Table 1, and some laboratory
properties are listed in Table 2. Although the vinyl styrene butadiene rubbers with 40% vinyl
groups and 24 and 28% styrene (Rubbers 3 and 4) do not have as good a combination of wet
Log wear rate

0.3 Wear rate = Wt loss test tyres
Wt loss NR control tyres
0.2 (50phr N220)
0.1
0 -- -- -- 4"-e--
-0.1 8,
-0.2
-0.3
I I I I I I I
0 10 20 30
Ambient test temperature, *C
FIG. 14---The effect of temperature on the wear properties of ENR-25 (35 pphr N220115 pphr
silica) retreaded radial tires.
GELLING ON INFLUENCE OF TREAD POLYMER 117
TABLE 2--Laboratory properties of ENR-25 and some vinyl tyrene butadiene rubbers.
Rubber ENR-25 3 4 5 6 7
Hardness (1RHD) 65 65 63 65 65 66
Modulus, 100%, MPa 2.6 2.5 2.3 2.2 2.1 2.2
Tensile strength, MPa 27.8 25.5 21.6 15.7 19.6 21.2
Tear strength, N/ram 21.2 8.6 5.9 6.7 7.8 8.6
Wet grip 115 100 75 90 108 104
Abrasion, DIN Index 88 121 121 96 111 98
Dunlop Resilience, %
23~ 47 55 64 59 50 56
35~ 54 61 68 63 56 60
50~ 58 63 70 65 59 62
70~ 66 70 75 68 62 64
100~ 72 75 78 72 67 69
traction and rolling resistance properties as ENR-25, their abrasion resistance is superior. The
hysteresis temperature profiles in the regions relating to wet traction and rolling resistance are
recorded in Fig. 15.
The hysteresis profiles of rubbers can be altered to improve tire tread properties by chemical
modification or by modifying the microstructure during polymerization. Solution-polymerized
SBRs with modified microstructures are currently being employed in certain tire tread applica-
tions. However, none of these materials meet the criteria of the ideal tread polymer.
In the future, "green" issues will probably play an increasing role in the tire industry and
will go beyond improving rolling resistance and wear properties. Recycling in the automotive
industry is a current issue in Europe, and there is likely to be governmental pressure to increase
the retreading of passenger car tires. In the longer term, the overall energy balance of tires
and contribution to the "greenhouse effect" may become an important issue. It should be noted
Tanb
0.25
3"~ ~" ~ ".o

0.15 4 , ~ "~ ~ .0. ~.
0.05 : , , : '
20 40 60 80 100
Temperature ,~
FIG. 15--The hysteresis temperature profile of ENR-25 compared to two vinyl styrene butadiene
rubbers: Rubber 3, 24% styrene, 40% vinyl butadiene; Rubber 4, 28% styrene, 40% vinyl butadiene.
that it takes 3 metric tons of oil to produce one metric ton of synthetic rubber, whereas N R
is essentially a renewable resource material [14].
References
[1] Albert, B. J. and Walker, J. C., "Tire Wet Road Friction at High Speed," Proceedings of the
Institution of Mechanical Engineering, Vol. 180, Part 2A, No. 4, pp. 1965-1966.
[2] Hill, S. W. and Moore, R. C., "Tire Rolling Resistance, Its Effect on Vehicle Fuel Economy and
Other Performance Areas," Proceedings, Akron Rubber Group Technical Seminar, 27 Oct. 1978.
[3] Evans, R. D., "Factors Affecting the Power Consumption of Pneumatic Tires," Proceedings, 2nd
Rubber Technology Conference, W. Heffer & Sons, Cambridge, England, 1948.
[4] Campbell, K. L., "Tire Rolling Lossess and Fuel Economy," SAE Conference Proceedings, P74,
Vol. 13, Society of Automotive Engineers, Warrendale, PA, October 1977.
[5] Hess, W. M. and Klamp, W. K., "The Effect of Carbon Black and Other Compounding Variables
on Tire Rolling Resistance and Traction," Rubber Chemistry and Technology, Vol. 56, 1983, p. 390.
[6] Schallamach, A., "Friction and Abrasion of Rubber," Rubber Chemistry and Technology, Vol. 3 l,
1958, p. 982.
[7] Kraguski, L. V. and Nepomnyashchii, E. E, "Fatigue Wear under Elastic Contact Conditions,"
Wear, Vol. 8, 1965, p. 303.
[8] Rahalkar, R. R., "Dependence of Wet Skid Resistance upon the Entanglement Density and Chain
Mobility According to the Rouse Theory of Viscoelasticity," Rubber Chemistry and Technology,
Vol. 62, 1989, p. 246.
[9] Gelling, I. R., British Patent 2,113,692, 1982.
[10] Davies, C. K. L., Gelling, I. R., Thomas, A. G., and Wolfe, S. V., "Strain Crystallization in Random
Copolymers Produced by Epoxidation of cis, 1,4-polyisopreue," Polymer, Vol. 24, 1983, p. 107.
[11] Davies, J. E. and Loadman, M. J. R., "A Chemical Demonstration of the Randomness of Epoxidation
of Natural Rubber," British Polymer Journal, J, Vol. 16, 1984, p. 134.
[12] Nordsick, K. H., "Model Studies for the Development of an Ideal Tire Tread Rubber," Paper No.
48, Proceedings, American Chemical Society, Rubber Division, Indianapolis, Spring 1984.
[13] Day, G. and Futamura, S., "A Comparison of Styrene and Vinyl Butadienc in Tire Tread Com-
pounds," Paper No. 22, Proceedings, American Chemical Society, Rubber Division, New York,
Spring 1986.
[14] Allan, P. W., "Accounting: Natural versus Synthetic Rubber," Rubber Developments, Vol. 32, 1979,
p. 104.
Francis Navin 1
Traction Tests on an Expressway Treated

with a Pavement Rejuvenator
REFERENCE: Navin, E, "Traction Tests on an Expressway Treated with a Pavement
Rejuvenator," Vehicle-Road Interaction, ASTM STP 1225, B. T. Kulakowski, Ed., American
ABSTRACT: A rash of accidents on a major freeway immediately after it was treated with a
pavement rejuvenator prompted the British Columbia Ministry of Transportation and Highways
(BCMoTH) to commission a set of traction tests by the author. A series of 31 maximum braking
tests were run at speeds of 50, 70, 90, and 120 km/h on wet normal and treated pavements,
and two additional tests were run on dry pavements at 50 km/h.
A 1991 ABS-equipped Lincoln Town Car was outfitted with a Lamar~'-5th wheel, g-ANA-
LYST TM,and two bumper guns and was driven by a police-trained driver. The tests were divided
more or less equally between maximum ABS and slide-to-stop braking.
The untreated surface had an average coefficient of friction at 50 km/h of 0.71 for peak
(ABS) braking and at 75 km/h of 0.61 for maximum locked wheel braking. The standard
deviation in both cases was 5% of the mean value. The similar conditions for the treated surface
were 0.68 for the peak ABS at 50 km/h and 0.53 for the locked wheel at 75 km/h.
An interesting outcome of this testing is in the experience with the equipment. It was difficult
to determine when and where the onset of braking actually started and stopped. The devices
failed to function at a rate of 6 to 12 in 100. It was found that all measuring devices were
needed to provide sufficient data to confidently estimate the average coefficient of friction.
KEY WORDS: traction tests, coefficient of friction, skid testing, wet pavements
Purpose of Vehicle Skid Testing
The purpose of the vehicle skid testing on Highway 1, a four-lane 100 km/h freeway just
east of Vancouver, BC, was to compare the relative average coefficient of friction at operational
speeds on a wet section that had a pavement rejuvenator treatment with another that had not
been treated. Pavement rejuvenators are petroleum-based materials applied to a pavement
surface that shows premature distress, in this case, extensive microcracking. The section under
study was thought to have had more skidding accidents than normal, so a comparative study
was requested by the British Columbia Ministry of Transportation and Highways (BCMoTH).
The purpose of this paper is to outline how the data were obtained with off-the-shelf
equipment, explain the results, and suggest a probabilistic model to determine braking distance.
The B C M o T H had undertaken British Pendulum tests at 77 sites to determine if a skidding
problem existed within a few days of pavement rejuvenator treatment. The results of the tests
on the normal pavement were, in units o f British Pendulum Number: the wheel paths ranged
from 68 to 75 BPN with an average of 72 BPN, and the shoulders ranged from 77 to 83 BPN
Professor and co-ordinator of accident investigation team, Dept. of Civil Engineering, University of
British Columbia, 2324 Main Mall, Vancouver, BC, V6T 1Z4, Canada.
119
Copyright* 1994 by ASTM International www.astm.org
with an average of 80 BPN. On the treated pavement, the results were: the wheel paths ranged
from 64 to 69 BPN with an average of 65 BPN, and the lane center line ranged from 64 to
74 BPN with an average of 70 BPN. The manufacturer suggests the following minimum values:
1. Difficult sites such as curves, grades over 5%, and approaches to traffic signals----65 BPN.
2. Highways with traffic greater than 2000 vehicles per d a y - - 5 5 BPN.
3. Other locations---45 BPN.
These tests indicated that the road's friction properties were acceptable.
The Royal Canadian Mounted Police (RCMP) also performed a few all-locked wheel skid
tests using a bumper gun to mark the start of braking and radar to estimate speed. The tests
were run a week after the surface was treated. The results were: an average coefficient of
friction of 0.45 on the wet treated surface, 0.71 on the dry treated surface, and 0.71 on the
dry normal pavement.
The BCMoTH decided to go ahead with full-scale vehicle testing in part because of the
media attention that this particular issue was attracting. The Accident Research Team from
Civil Engineering at the University of British Columbia was requested to undertake full-scale
traction tests to compare the normal highway pavement to that treated with the pavement rejuve-
nation.
The comparison was accomplished by first estimating a relation between the ABS peak and
four-wheel locked average coefficients of friction during maximum braking and the vehicle
speed at the start of braking on wet pavements for both the treated and untreated pavement
surfaces. The emergency braking conditions were maximum ABS (independent anti-lock
braking system on all four wheels) and four-wheel locked skidding. The need for this relation-
ship arose because the average operating speed was about 120 krn/h, well beyond the safe
speed for any four-wheel locked skidding tests.
The average coefficient of friction is generally dependent on the driver, the vehicle, the
tires, the pavement surface, and the environment. The testing program used the same RCMP-
trained driver and a new car (1991 Lincoln Town Car) with good tires as given in the Appendix.
The tire wear was not measured, nor were the tires changed during the tests. A wet pavement,
both treated and untreated, was the test surface.
The instruments used to record or estimate the average coefficient of friction were a Lamar TM
5th wheel, a g-ANALYST TM, and two bumper guns supported by police radar. The general
specifications of these instruments are given in the Appendix.
Test Program
The basic idea underlying the tests was to establish the relationship between maximum all-
wheel ABS braking decelerations and those experienced during four-wheel locked wheel slide
to stop. The test program simply ran a series of controlled experiments to measure the
average coefficient of friction on a treated surface and another that was not treated. Additional
information from the testing would be a relationship between the instantaneous braking coeffi-
cient and the average value.
The testing was conducted on a suburban section of the Trans Canada Highway, a four-
lane freeway that runs east from Vancouver, British Columbia. The highway was closed from
0100 to 0630 h for two evenings on the 16th and 17th of May 1991. This was 49 days after
the rejuvenation treatment, which was applied on 26 and 27 March 1991.
The actual testing plan called for an ABS test on dry pavement to confirm that the equipment
functioned properly. The highway was then wetted, and the Society of Automotive Engineers
Inc. procedure SAE J345a [1] for car braking traction tests was followed as closely as possible.
NAVIN ON TRACTION TESTS 121
The speeds tested were 50, 75, and 100 km/h for ABS and 50 and 75 km/h for locked wheel
braking. There was one additional ABS test at 120 km/h.
Because of the importance that the MoTH put on these tests, four independent measuring
devices were used, and the data were retrieved from each device at the completion of every
test. Each test followed the SAE J345a procedure as closely as possible: the test car was
positioned, the two bumper guns armed, and the test controller informed. The water truck wet
the highway high-speed travel lane. The test car was cleared to start and had prior instructions
to stabilize speed and then to start maximum braking between two cones in an area that was
flood-lit. Once the car had stopped, measurements were taken from the bumper gun shot marks
to the test car, the radar speed was obtained from the RCMP officer, the g-ANALYST TM
memory was stored on disk, and the Lamar TM 5th wheel printout was retrieved.
The direct measure from each instrument is given in Table 1. The estimate of the duration
of braking was most easily estimated from the Lamar TM 5th wheel data. The g-ANALYST TM
was used as a check on the other braking time and as the actual estimate when the 5th wheel
failed to operate.
The estimated time to complete a test was originally planned to be 20 mins, a time found
during previous nighttime practice tests. The actual in-field experience was almost half an
hour for the first few tests. This was eventually reduced to just over 12 min at the end of the
second night of testing.
The additional tests were, a high-speed 120 km/h with maximum ABS braking, a cluster
of tests at 50 km/h with maximum ABS, and another cluster at 75 km/h with locked wheels,
all on the untreated section. The clustered testing was used to judge the consistency of repeated
tests with the off-the-shelf measurement instruments.
Findings
The average coefficient of friction from an initial speed of braking on the rejuvenation
treated surface and normal surface are shown in Figs. 1 and 2, respectively. The data points
represent the average of the three methods used to estimate the average coefficient of friction
experienced over the entire stopping process (see Tables 2 and 3). The agreement between
the average coefficient of friction estimated by the various devices is surprisingly good. The
values that appear to be somewhat erratic are those for maximum ABS braking on the untreated
pavement, particularly the results at 100 km/h. The ABS tests at 100 km/h were repeated three
times, and all were a little greater than expected, so Tests 28 and 29 were conducted to check
on any systematic error. Test 30 was another attempt to see if the high coefficient of friction
persisted at high speed. The general conclusion was that the variation was probably due to
the logic in the ABS system, so Tests 31, 32, and 33 were performed. These latter tests
clustered around the previous results, so the suspicion of a systematic error in the procedure
was dropped. The test value from the bumper guns was usually the average value between
TABLE 1--Data measurement devices.

Device
Measurement 5th Wheel g-ANALYST
TM Bumper Gun Radar
Distance Direct m e a s u r e Calculated Direct measure

Time Direct measure Used if needed
Velocity Calculated Calculated Direct measure
Deceleration Calculated Direct measure Calculated
1.0
I
O DRY,MAXIMUMABS
0.8
WET, MAXIMUMABS
Z
O
I
2 99
I--
O - e e - 9 7 12 tl
% 3 4 __li
r,,
IJ. % 6
11. 0.6 WET, LOCKEDWHEEL
~" , J + . ,s 9
O
l-
Z 9 10
LU
I
O
I
u.
u_ 0.4
ILl
O
O
LU
O
LU 0.2
<>
9 MAXIMUMABS
9 LOCKEDWHEEL
TEST CAR, 11~1 UNCOLN TOWN CAR
0.0 I I I I I I
0 20 40 60 80 100 120
I N I T I A L S P E E D (kin/h)
FIG. 1--Average coefficient o f friction versus initial speed on normal pavement, Highway l, 2
km east of 232 St., eastbound fast lane.
the L a m a : M 5th wheel and g-ANALYST TM on the treated surface, but was more erratic on
the untreated surface.
The reliability of the equipment is also shown in Tables 2 and 3. The Lamar TM 5th wheel
failed in four tests, the g-ANALYST TM failed in two tests, both bumper guns failed on three
tests, and the marks could not be found in one test. Finally, the remarks column indicates the
media attention that these tests received. In the early morning, an investigative reporting TV
crew arrived uninvited and unannounced rather than waiting for the tests at 0530 h that were
to be conducted for both the press and the experiment. Much of this testing was conducted
under the glare of public television.
The actual relationship between the ABS peak and four-locked wheel, wet road, average
coefficient of friction values is a reduction of about 20% at 100 km/h. At 100 km/h the results
for the rejuvenated pavement 49 days after application and highway use was a peak average
value of 0.64 versus 0.73 for the normal untreated section. These coefficients are about the
best to be expected given the high quality of the vehicle, good tires, and an experienced driver.
1.0
15, 16
O DRY,MAXIMUM ABS
WET. MAXIMUM ABS 29 25

0.8 %. ~ 27
17 ~ ~
Z
O
I- WET, LOCI~EOWHEEL "~ 2 Q 9
19 24~,'~ ~ ~ ,~. ,..

31
u. 0.6
33
LI.
o
I-
Z
ILl
O
U. 0 . 4
LI.
ILl
O
LU
LU 02.
<>
9 MAXIMUM ABS
9 LOCKEDWHEEL
TEST CAR, t991 LINCOLN TOWN CAR
0.0 I I I I I I
0 20 40 60 80 100 120
INITIAL SPEED (km/h)
FIG. 2--Average coefficient of friction versus initial speed on treated section, Highway 1, 2 Ion
west of 264 St., eastbound fast lane.
The relationship between the peak and slide average and instantaneous coefficient of friction
are shown in Figs. 3 and 4. The actual shape of the instantaneous curves follows that found
by other authors [2-5].
The instantaneous values obtained from the g-ANALYST TM and the Lamar T M 5th wheel had
approximately the same shape. The g-ANALYST T M curves were smoother and slightly below
the 5th wheel curves, (Fig. 5). The instantaneous peak coefficient of friction at 100 km/h was
0.59 on the treated surface and 0.70 on the untreated surface, and the locked-wheel values
were 0.43 and 0.54, respectively.
Finally, an analysis of variance was performed with ABS Tests 17, 18, 28, and 29 and for
locked-wheel Tests 23, 24, 31, 32, and 33. In both situations the total variance was about 5%
of the mean value of the average coefficient of friction. The between ABS test variance was
roughly two thirds of the total variance. The remaining one third was accounted for by the
instrumentation. For the locked-wheel test, the devices accounted for slightly more than half
of the variance.
TABLE 2--Vehicle traction results on a rejuvenator-treated section.
Test Average Coefficient of Friction

Number Brake Mode Speed, km]h |a 2b 3c Average Rema~s
1 ABS 52.9 0.88 0.81 0.84 0.84 Dry pavement

2 ABS 52.9 0.63 0.70 0.70 0.68 Remaining wet
3 ABS 50.3 0.66 0.66 0.71 0.68
4 ABS 52.6 0.66 0.70 0.68 0.68 Test slow lane
5 L-W 53.4 0.60 0.59 0,63 0.57 Car pulls left
6 L-W 53.4 0.64 0.60 0.68 0.64
7 ABS 77.6 0.68 0.68 0,68 0.68
8 ABS 76.9 0.72 0.71 0.67 0.70
9 L-W 75.8 0.53 ND d 0.52 0.53 Disrupted U TV
10 L-W 78.0 0.57 0.46 0.53 0.53 Car pulls left
I1 ABS 102.6 0.70 0.61 0.66 0.66
12 ABS 101.0 0.69 0.64 0.64 0.66
13 ABS 102.9 0.65 0.52 0.57 0.58 Media present
14 L-W 49.9 ND d 0.60 0.61 0.58 Media, car pulls left
"Lamar TM 5th wheel.
bg-ANALYST TM .
CBumper gun.
"No data.
L-W = locked wheel. ABS = ABS engaged.
TABLE 3--Vehicle traction results on a normal pavement surface.
Test Average Coefficient of Friction

Number Brake Mode Speed, km/h 1a 26 3~ Average Remarks
15 ABS 51.3 0.88 0.87 0.90 0.88

16 ABS 51.0 0.88 0.87 0.90 0.88
17 ABS 50.0 ND d 0.77 0.68 0.73
18 ABS 49.6 0.71 0.72 ND d 0.72 Both misfire
19 L-W 52.0 0.66 0.68 0.60 0.65 BC TV arrive
20 L-W 52.0 0.66 0.70 0.65 0.67
21 ABS 75.6 0.73 NI~ 0.75 0.74
22 ABS 76.4 0.75 0.81 0.74 0.75
23 L-W 76.1 0.67 0.66 0.62 0.65 Car pulls left
24 L-W 76.8 0.65 0.66 0.60 0.64 Car pulls left
25 ABS 102.8 0.77 0.82 ND a 0.80 Both misfire
26 ABS 103.3 0.76 0.81 ND a 0.79 Mark not found
27 ABS 103.1 0.77 0.8l ND d 0.79 Both misfire
28 ABS 52.l 0.73 0.70 0.76 0.73 Cluster test
29 ABS 51.7 0.76 0.79 0.78 0.78 Cluster test
30 ABS 122.6 0.73 0.65 0.72 0.70 Brakes heated
31 L-W 76.1 0.65 0.66 0.60 0.64 Cluster test
32 L-W 77.0 ND d 0.64 0.66 0.63 Cluster test
33 L-W 77.0 ND a 0.66 0.58 0.60 Cluster test
a5th wheel.
~g-ANALVST".
CBumper gun.
'No data.
L-W = locked wheel. ABS = ABS engaged.
1.0
3 0.8
,••IMUM
ABS
U.I 0.6
E
*'~ .,,. ,,~ LOCKED WHEEL
0.4
Z
0.2
00
Z
TEST CAR, 1991 UNCOLNTOWN C A R
0.0 I I I I I I
20 40 60 80 100 120
SPEED (km/h)
FIG. 3--Instantaneous coefficient of friction versus speed on treated section, Highway 1, 2 km
east of 232 St., eastbound fast lane.
Expected Value Braking Model
Using expected values as outlined by Ang and Tang [6], the expected braking distance and
variance may be estimated as
_ E(V) 2 + 1 E(f)3 (1)

(ao) 2e-------~
g 2 | _ E (_v) [Cov (V,?) + Cov (), v)]
k E (f)2
E (V)4 E (V)2
- - [Cov (V,f) + Cov ~ V)] (2)
L2-~J
Copyright by ASTM Int'l (all rights reserved); Sat Dec 24 01:23:58 4 (.~)4Gr~
ESTE2011 2E (~)z
1.0
~ = . = . MAXIMUM
ABS
LU O.6
imm
LOCKED
WHEEL
m
E
0
0.4
o
Z
Z
0.2
Or)
Z
TESTCAR,1901LONCOLN
TOWNCAR
0.0 I I I I I I
0 20 40 60 80 100 120
SPEED (kin/hi
FIG. 4 Instantaneous coefficient of friction versus speed on normal pavement, Highway 1, 2
km west of 264 St., eastbound fast lane.
where
BD = braking distance, m,
V= driving speed, m/s,
.f = average coefficient of friction,
g = 9.81 rn/s 2,
0 .2 = v a r i a l l c e ,
Coy = covariance, and
E = expected value.
Assume that speed is independent of the average coefficient of friction, Cov(V, J) = 0, that
the variance of speed is zero, and that the variance of the average sliding coefficient of friction
is that found in the cluster tests of this experiment. For these conditions, the expected braking
distance and standard deviations at 75 km/h are 36 and 5 m, respectively, on the untreated road.
The relationship between l a n d V is a fairly well-established relationsh!p for wet pavements.
The relationship between a person's speed, V, and the apparent f or slipperiness of the road
1.0
Z
O
i
I- 0.8
_o
IZ
u.
LL
O
I-
Z 0.6
uJ
m
_o
IJ.
LL
UJ
0
0
r~ 0.4
0
IJJ
Z
Z
0.2
Z
O TESTNUMBER
5th WHEEL
g-ANALYST
0.0 I I I I I I
0 20 40 60 80 100 120
SPEED (km/h)
FIG. 5--1nstantaneous coefficient of friction versus speed from Lamar TM 5th wheel and g-ANA-
LYST TM on untreated section, Highway 1.
has not been reported in the literature. Observation by Olsen et al. [7] indicates that drivers
do not change their speed on dry or wet pavement and that the speed distribution is normal
with a standard deviation of about 15%. Casual speed observations by police and highway
officials at the test sections did not indicate any difference in driver behavior. It seems reasonable
to assume that drivers alter their speed only in exceptional circumstances.
The formulation of a model that allows both speed and braking to be random may be
estimated by a Monte Carlo simulation. The model for the results presented here used a normal
distribution of speed that had a standard deviation of 15% of the mean. The mean speed for
this example was 75 km/h. The average deceleration rates were taken from the observed data
of Figs. 1 and 2. The standard deviation was set at 5% of the mean deceleration, and the
distribution was assumed to be uniform since there was not enough data to develop a reli-
able distribution.
The resulting braking distance from the simulation is shown in Fig. 6 and is numerically
compared in Table 4. The probability that the demanded braking distance at 75 km/h exceeds
that recommended by AASHTO, given the need for maximum braking, is 1.1 in 100 for the
untreated road and 5 in 100 for the treated road. Ervin and Winkler [8] presented an interesting
100
umR~o
80 ~.. ~ " - TREATEO
/
60
i I
i=
I
I AASHTO
2O I
i
~~ESIGN)
i
i
i
0 t ~ J/Ill I I I I I
25 ,SO 75 lOO
BRAKING DISTANCE, m
FIG. 6---Braking distance distribution for 75 kmlh average speed (speed, coefficient of friction
are random).
TABLE 4--Braking distances at 75 kmlh; speed and

deceleration are random.
Treated, Untreated,
m m
Mean 42.9 39.0

Standard deviation 13.4 12.3
P(BD > AASHTO design) 5:100 1.1:100
NOTE: P = probability; BD = breaking distance.
variation on the demand for and supply of coefficient of friction throughout a road system.
They found that for a vehicle such as that tested, the chance of wheel lockup is about 1 in
7000 on a wet road. If the brakes were poorly balanced with an efficiency of about 60%, then
the chance of lockup is about 1 in 65.
Conclusions
The g-ANALYST TM and Lamar TM 5th wheel produced similar results for the instantaneous
coefficient of friction at the speed tested. The average coefficient of peak and sliding friction
values estimated with the g-ANALYST, TM Lamar TM 5th wheel, and bumper guns were in
reasonable agreement. The repeated tests found a variance in the measured average coefficient
of friction of about 5% of the mean value. The machine failure rate was 6 in 100 for the g-
ANALYST T M , 12 in 100 for the Lamar TM 5th wheel, and slightly greater than 12 in 100 for
the bumper guns. All three measurement devices were found to be needed for the experimental
work. First, the bumper gun results provided a quick check that the experiment was actually
measuring the coefficient of friction of the vehicle tire on a wetted road. The g-ANALYST TM
also provided an easily available summary value of peak and average values that could be
compared to the bumper gun results. The actual time of braking was best determined from
the Lamar TM 5th wheel results. The consensus was that all three measuring devices were needed
to provide the redundancy and cross checking needed to have confidence in the results.
The relationship between the all-wheel ABS peak and the four locked-wheel average coeffi-
cient of friction is a reduction of about 20% at 100 km/h. The shape of the curve of the all-
locked-wheel instantaneous coefficient of friction versus speed follows that reported by Lamm
[4]. The ABS peak instantaneous coefficient curves showed the influence of the logic within
the ABS. The resulting curves were adequate for the purposes of this research, which was to
estimate the all-locked wheel average coefficient of friction at operational speeds of about
120 km/h.
The influence of the pavement rejuvenator treatment, 49 days after application, was to
reduce the average coefficient of friction by 0.10 for the Lincoln Town Car. Casual observations
by the highway authorities found that there was no change in driver speed through the treated
section. This observation is in keeping with those of Olsen et al. [7] in which they found that
driver speed did not change between dry and wet pavements.
The experience of this experiment suggests that it may be possible to adequately measure
the instantaneous coefficient of friction using an automobile equipped with ABS. The logic
behind the ABS should be known so that its influence may be factored out. Such a system
might provide an economic alternate to an expensive skid trailer. Also, the experiment showed
that "off-the-shelf" deceleration measurement devices can, when carefully used, provide accu-
rate estimates of a vehicle's instantaneous deceleration and the average all-wheel locked
coefficient of friction.
Acknowledgements
The data for this project were collected by Innovative Vehicle Testing Ltd., assisted by Mr.
N. Cumming, P.Eng., and members of E Division of the Royal Canadian Mounted Police. Mr.
R. Thomson did much of the data reduction.
APPENDIX
Vehicle Make and Description
1991 Lincoln Town Car

VIN: 1LNCM81WOMY65
Date of Manufacture: 11/90
Test Weight: 2060 kg (includes driver, observer, all instrumentation, and 3/4 tank of fuel)
Vehicle Braking System
The vehicle is equipped with vacuum-assisted, four-wheel disk brakes with anti-lock (ABS).
The ABS controls each brake separately through a hydraulic control unit which modulates the
line pressure to the individual brake caliper pistons. Magnetic reluctance wheel speed sensors
are located on each wheel and send signals to an electronic control unit, which determines
which wheel(s) could be close to lock-up by monitoring wheel deceleration values.
Vehicle Tires
Tire Manufacturer: Michelin
Model and Size: XW4, P215/70 R15, M + S
Tire Pressure: at operating pressure
LF: 220.6 KPa RF: 213.7 KPa
LR: 213.7 KPa RR: 213.7 KPa
Tread Depth: average of four points across face of the tread
LF: 6.35 mm RF: 6.03 mm
LR" 6.43 mm RR: 6.27 mm
Lamar 5th Wheel TM
The Lamar 5th W h e e l " is manufactured by Lamar Instruments of California. It is a micropro-

cessor-based road test device using a small-diameter wheel with an inner-mounted steel ring
with 508 slots cut out of the ring. An optical reader picks up the intermittent pulses, and speed
may be determined to the nearest 0.2 km/h. The wheel is calibrated by traveling over a known
distance and comparing that to the measured results.
The wheel may be mounted directly to the vehicle frame or bumper. The data are stored
on an internal microprocessor. The data, time, distance, and speed are printed out in 100-
ms intervals.
The g-ANALYST TM
The g-ANALYST TM is manufactured by Valentine Research Inc. It uses a three-axis servo-
accelerometer that measures lateral, longitudinal, and vertical acceleration. The transducers
are sampled at 10 Hz and then the signals go through an anti-aliasing filter. The accelerometers
have a 60 V, d-c supply and are activated with a reference voltage of 2.5 V, d-c. The signals
are corrected for roll and pitch in software so that accelerations are measured in the plane of
the road.
Bumper Guns
Bumper guns are manufactured by Pacific Institute of Traffic Safety Inc. A switch is mounted
in the brake light circuit, and a 38-caliber chalk-filled shell is fired when the brake light goes
on. The system uses the car's 12 V, d-c battery as the power source and a solenoid to fire the gun.
Radar Gun
The radar gun is manufactured by Muniquip and has an accuracy of ---0.5 km/h.
References
[1] Society of Automotive Engineers Inc., "Wet or Dry Pavement Passenger Car Tire Peak and Locked
Wheel Braking Traction," SAE J345a, 4:30.09, Society of Automotive Engineers, Warrendale, PA.
[2] Henry,J., "Comparison of the Friction Performance of a Passenger Car Tire and the ASTM Standard
Test Tires, Frictional Interaction of Tire and Pavement," Frictional Interaction of Tire and Pavement,
ASTM STP 793, W. Meyer and J. Walter, Eds., American Society for Testing and Materials, Philadel-
phia, 1983, pp. 219-231.
[3] Yandell,W., Taneerananon, R, and Zankin, V., "Prediction of Tire-Road Friction from Surface
Texture and Tread Rubber Properties," Frictional Interaction of Tire and Pavement, ASTM STP 793,
W. Meyer and J. Walter, Eds., American Society for Testing and Materials, Philadelphia, 1983,
pp. 304-322.
NAVlN ON TRACTION TESTS 131
[4] Dijks, A., "Influence of Tread Depth on Wet Skid Resistance of Tires," Transportation Research
Record 621, Transportation Research Board, Washington, DC, 1977, pp. 136-147.
[5] Lamm, R., "Driving Dynamic Considerations: A Comparison of German and American Friction
Coefficients for Highway Design," Transportation Research Record 960, Transportation Research
Board, Washington, DC, 1984, pp. 13-20.
[6] Ang, A. and Tang, W., Probability Concepts in Engineering Planning and Design, John Wiley and
Sons, Toronto, 1984.
[7] Olsen, P. et al., "Parameters Affecting Stopping Sight Distance," NCHRP Report 270, Transportation
Research Board, Washington, DC, 1984.
[8] Ervin, R. and Winkler, C. D., "The Influence of Braking Efficiency on the Probability of Wheel
Lockup," No. 870334 Society of Automotive Engineers, Warrendale, PA, February 1987.
Ride Quality and Road Roughness
Elson B. Spangler I and William J. Kelly I
Development and Evaluation of the Ride

Number Concept
REFERENCE: Spangler, E. B. and Kelly, W. J., "Development and Evaluation of the
Ride Number Concept," Vehicle-Road Interaction, ASTM STP 1225, B. T. Kulakowski, Ed.,
ABSTRACT: The statistical analysis of pavement ride quality was introduced in the United
States during the 1950s at the AASHO Road Test. It was at the Road Test that Carey and Irick
developed the statistical relationship between pavement surface profile and measured subjective
ride quality. Interest in pavement ride quality continued in the 1960s with the development of
the road meter and the inertial profilometer, which for the first time provided a fast and accurate
method for measuring and recording pavement surface profile in a computer-readable format.
During the 1970s, ride quality transforms relating pavement surface profile measurements
made with the inertial profilometer to measured subjective ride quality were developed at the
University of Texas and at the Michigan Department of Transportation. During the 1980s, the
ride number concept for estimating pavement ride quality from surface profile measurements
was developed by Janoff et al. in a National Cooperative Highway research project. This paper
discusses the further development of the ride number transform, its evaluation in the NCHRP
and Ohio DOT research projects, and the comparison of its performance to other candidate
measures of pavement ride quality including the Texas Serviceability Index, the Michigan Ride
Quality Index, the Mays Road Meter Index, and the International Roughness Index.
KEY WORDS: ride number, pavement ride quality, pavement roughness, inertial profilometer,
International Roughness Index, Mays Road Meter Index, Texas Serviceability Index, Michigan
Ride Quality Index
As a result of research performed by Carey and Irick [1] at the AASHO Road Test in the
1950s, an interest has developed in the United States in the objective measurement of pavement
ride quality. In the AASHO Road Test Program, those investigators developed a method for
estimating the subjective ride quality of a pavement test section from measurements made on
the pavement's surface profile. This estimated pavement ride quality was called the Present
Serviceability Index (PSI) and was used extensively in the United States in subsequent years
to describe the present ride quality of pavements being evaluated.
Since the direct measurement of pavement surface profile was not possible at that time, the
AASHO Present Serviceability Index implementation used surface profile data obtained from
the geometric response of a slope variance measuring device to the surface profile of the
pavement being evaluated. Although the Present Serviceability Index concept has had continu-
ing support since the AASHO Road Test Program, the usefulness of this concept suffered
from the limitations of the existing slope variance measuring equipment to accurately produce
the required slope variance data. However, one of the lasting heritages of that program was
the 0 to 5 pavement ride quality rating scale which has continued to be used in the United
l Surface Dynamics, Inc., Suite 102, 3883 Telegraph Rd., Bloomfield Hills, MI 48302.
135
States to describe pavement ride quality. A modern variation of the AASHO rating scale is
shown in Fig. 1.
Shortly after the AASHO Road Test, two important events occurred in the United States
in the area of pavement ride quality evaluation:
1. Introduction of the road meter.

2. Introduction of the inertial profilometer.
In the early 1960s, two automotive vehicle-based road meters were introduced in the United
States: (1) the Portland Cement Association (PCA) road meter [2], and (2) the Mays road
meter (MRM) [3]. The operation of both of these road meters was based on the measurement
and accumulation of the relative motion of the sprung and unsprung masses of automotive
vehicle suspension systems. The two road meters differed in how the measured displacement
was weighted in the accumulation process. The PCA road meter weighted the measured
displacement as a function of the amplitude of the displacement measurement, whereas the
Mays road meter applied the same weighting regardless of the measured amplitude. Due to
the weighting method used in the PCA road meter, its output was found to be very sensitive
to the changes in vehicle running height caused by normal vehicle load changes. Although
both of these road meter types had their advocates and were widely used during the 1960s,
the Mays road meter has prevailed and is still used today by many state highway departments.
PERFECT I 15
VERY
GOOD I IRIDE QU~ITY
DOES NOT NEED
INPROVEMENT
GOOD
FAIR
RIDZ~UALITY
POOR NEEDS
IMPROVEMENT
VERY
POOR
IMPASSABLE -0
FIG. 1--Ride quality rating form.
SPANGLER AND KELLY ON THE RIDE NUMBER CONCEPT 137
It was also during the 1960s that the inertial profilometer was developed by Spangler and
Kelly [4] at the General Motors research laboratories. Although this device was developed to
provide a road profile input to computer simulations of vehicle suspension systems, it was
immediately obvious that the ability to measure a road surface profile would also be very
useful to the highway community in the evaluation of pavement ride quality. Through a
licensing arrangement, the General Motors Corporation made the new inertial profilometer
technology available to the highway community through the company of K. J. Law Engi-
neers, Inc.
Two of the more aggressive users of the new inertial profilometer technology were the
University of Texas, Center for Highway Research [5], and the Michigan Department of State
Highways [6]. Both of these centers of activity have used subjective ride panel rating of
pavement roughness to develop transforms between measured pavement surface profile and
estimates of pavement ride quality. Hudson [7] at the University of Texas developed a surface
profile-based serviceability index (SI) transform presently being used by the Texas Department
of Highways and Public Transportation as a measure of pavement ride quality. At about the
same time, the Michigan Department of Transportation was developing a ride quality transform
[8] they have named the Ride Quality Index (RQI). Both of these ride quality transforms will
be discussed and evaluated in this report.
Starting in the 1980s, the National Cooperative Highway Research Program (NCHRP)
sponsored two research projects which addressed the use of pavement surface profile measure-
ments to estimate pavement ride quality. In these projects, Janoff [9,10] used statistical correla-
tion techniques to develop a relationship between measured pavement surface profile and
measured subjective ride quality. In the first of these studies [9], Janoff isolated the spatial
wave length content of the pavement surface profile which appeared to be the biggest contributor
to the measured variable of subjective ride quality. Janoff then went on to develop a ride
quality transform based on this spatial wave length content. In his work, Janoff found that the
ride quality estimates produced by his transform were independent of pavement type, road
class, vehicle size, vehicle speed, and other variables that might affect subjective rating
panel measurements.
In the second NCHRP project [10], Janoff extended the study to four different regions in
the United States to evaluate the regionality variable. In this study, Janoff found no significant
difference in how subjective panel rating members rated pavement ride quality over roads in
the seven states included in the study. It was also in this study that Janoff proposed the name
"ride number" for the estimated subjective ride quality value computed from measured pavement
surface profile.
Concurrent with the two NCHRP projects, the Ohio Department of Transportation (Ohio
DOT) contracted with Surface Dynamics, Inc. [11] to assist that agency in the integration of
a newly acquired K. J. Law Engineers, Inc. inertial profilometer into the Ohio DOT pavement
management system. One of the tasks in that project was to develop and evaluate a pavement
surface profile-based ride quality transform that could be used by the Ohio DOT to rate the
ride quality of its pavement inventory and to use as a predictor of the need for nonroutine
maintenance. It was soon realized that the NCHRP and the Ohio DOT ride quality projects
had much in common, and that the ride quality data collected in the state of Ohio could be
shared by the two projects. In this data collection task, Michael Janoff was responsible for
the collection of the subjective ride panel data used in the Ohio study. This 1983 ride quality
data was the basis for development by Janoff and Surface Dynamics, Inc. of two different
ride quality transforms that produced basically the same results.
As a continuation of the Ohio DOT ride quality research project, the Ohio Department of
Transportation contracted with Surface Dynamics, Inc. to conduct a repeat of the 1983 ride
quality study five years later in 1988. The objective of the new study [12] was to evaluate
the time stability of the subjective ride panel rating data over the five-year time period. A
second objective of the new study was to evaluate the performance a simplified ride number
transform developed in the project and any other transforms that might be considered as good
candidates for pavement ride quality evaluation.
The purpose of this report is to discuss the development of the ride number transform and
to present the results of the evaluation of five different candidate ride quality transforms.
Research Objectives
The ride number transform development and evaluation to be discussed in this paper is
basically a continuation of the research efforts of Janoff and Spangler [9-11] in the area of
pavement ride quality. The objective of the new research effort was to evaluate the long-term
time stability of the subjective ride panel rating data over the five-year period from 1983 to
1988. A second objective was to use the additional ride quality data produced in the new
project to simplify and fine tune the ride number transform developed in the earlier studies.
The final objective of the new project was to evaluate the performance of five candidate ride
quality transforms in their ability to estimate measured subjective ride quality from pavement
surface profile measurements.
Measured Subjective Ride Quality Data

In the new Ohio Dot ride quality study, the collection of the subjective ride quality data
was performed exactly as in the two earlier NCHRP and 1983 Ohio DOT ride quality studies.
To ensure the replication of those studies, Michael Janoff, the principal investigator on the
NCHRP projects and collector of the subjective ride quality data on the ftrst Ohio DOT project,
was given the responsibility of collecting the subjective ride quality data on the new Ohio
DOT project.
As in the 1983 Ohio DOT ride quality study, the Ohio DOT provided the required support
for the subjective ride quality data collection including the rating panel members, the rating
panel vehicles, the vehicle drivers, the rating panel instruction site, and the 89 pavement test
sections included in the rating process.
As in the 1983 Ohio DOT ride quality study, the Ohio subjective ride quality rating panel
was made up of 36 rating panel members. The 36 rating panel members were divided into
three groups of twelve for the rating panel activities. Daily instructions to the rating panel
members were provided by Michael Janoff. The instructions in this study were identical to
those given on the earlier NCHRP and Ohio DOT studies. Each rating panel member rated
the ride quality of all of the pavement test sections using the ride quality rating form shown
in Fig. 1. Upon the completion of the ride quality rating session, the mean panel rating (MPR)
for each pavement test section was computed from the rating forms from the 36 rating
panel members.
Surface Profile Measurements

As in the 1983 Ohio DOT ride quality study, the surface profiles of the 89 pavement test
sections were measured and recorded on floppy disk by the Ohio Department of Transportation
staff using a K. J. Law Engineers, Inc. inertial profilometer. Although a concerted effort was
made to obtain surface profiles over the full spectrum of profile roughnesses, there was the
practical difficulty of obtaining the desired number of surface profiles for the very rough
pavements since these pavements are constantly being resurfaced. The fewer very rough
pavements might have the potential to influence the test results at the rough end of the pavement
roughness spectrum.
TABLE 1--Pavement test section distribution in 1983 study.
Functional Pavement Type

Road Bituminous Portland Cement
Classification Concrete Concrete Composite Total
Interstate 0 14 5 19
Principal arterial 1 2 2 5
Minor arterial 0 1 2 3
Major collector 2 0 6 8
Minor collector 3 0 2 5
Local 12 0 0 12
Total 18 17 17 52
The distribution of the pavement test sections by pavement type and by the Ohio DOT
functional road classifications are shown in Table 1 for the 1983 study and in Table 2 for the
1988 study.
Ride Quality Transform Evaluation
Rather than immediately evaluating the time stability of the subjective ride panel rating
data, the first research effort was to use the combined 1983 and 1988 ride quality datasets to
develop a new simplified fide number transform that would be used in the remainder of the
project. The form of the new fide number (RN) transform is shown in Eq 1.
RN=5"[ e-2~ +] e-:~

2
(1)
where PIL and PIR are the profile indices for the left and fight wheel paths, respectively,
computed from the following equations
TABLE 2--Pavement test section distribution in 1988 study.
Functional Pavement Type

Road Bituminous Portland Cement
Classification Concrete Concrete Composite Total
Interstate 0 15 17 32
Principal arterial 1 9 7 17
Minor arterial 5 0 4 9
Major collector 7 0 3 10
Minor collector 9 0 0 9
Local 12 0 0 12
Total 34 24 31 89
PIL = RMS S 2 + 2~toS + to~ (PL)
PIR = RMS
52 )]
S 2 + 2~toS + to2 (PR)
where
RMS = root mean square of the filtered profile,

PL = the measured elevation profile of the left wheel path, in.,
PR = the measured elevation profile of the fight wheel path, in.,
S = complex operator, lift,
to = spatial natural frequency, cycles/ft, and
= damping ratio, 0.7.
The effect of Eq 1 is to compute the root mean square (RMS) of the measured elevation
profiles of the individual left and right wheel paths (PL and PR) where the displacement
amplitude of the measured pavement profiles have been attenuated by a second-order high
pass filter with a spatial natural frequency (to) of 0.125 cycles/ft or a wavelength cutoff of 8
ft (2.4 m). The combined 1983 and 1988 Ohio DOT ride quality datasets were used to determine
the coefficient - 2 0 in Eq 1 using the best-fit straight line (regression equation) to the measured
mean panel rating (MPR) and the estimated mean panel rating (RN) computed from measured
surface profile data.
This best-fit straight line is represented by the line of equality in Fig. 2 along with the
coefficient of correlation (0.91) and the standard error of estimate (0.29 RN). The correlations
shown in Fig. 2 for the fide number transform (Eq 1) are for all the pavement types (BC,
PCC, and composite), functional road classifications, pavement test section locales, and lane
configurations evaluated in the combined 1983 and 1988 ride quality datasets. The effects of
these four variables on the performance of the fide number can be evaluated by treating the
pavement test section in each of these categories as a subset of the combined 1983 and
1988 datasets.
Pavement Type as a Variable
In the combined 1983 and 1988 ride quality datasets, there were a total of 141 pavement
test sections in the following categories:
Bituminous concrete 52
PCC 41
Composite 48
All pavements 141
To determine the effect of the four different pavement types on the performance of the ride
number transform, a best-fit straight line regression to the measured mean panel rating and
the computed ride number was computed individually for the data pairs in each of these three
pavement types, as shown in Fig. 3.
In Fig. 3, the regression lines for the PCC and composite pavements are identical and fall
on top of each other. Also from Fig. 3, it can be observed that the difference between the
1983 § 14 56 R i d e Q~ s i l t y D a t a .
mmt
9 o
,J m 9
el m el m 9
~2 a
C oFrelaUon 0.91
B st. E r r o r = 0.29
O 9 9 I * 9 | 9 9 9 9 9 9
0 I 2 3 4 5
Estimated MPR (RN)
FIG. 2--Correlation between measured M P R and ride number.
measured MPRs and the estimated MPR for the three pavement types is less than the accuracy
of the ride quality rating panel experiment of 0.2 to 0.3 MPR units. The computed correlation
coefficients and standard errors of estimate for the individual pavement types are shown in
Table 3.
Functional Road Classification
In the analysis of the effect of the Ohio DOT functional road classifications on the panel
ratings, a best-fit straight line regression equation to the measured mean panel rating (MPR)
and the estimated mean panel rating were computed individually for the six Ohio DOT
functional road classifications. When plotted (Fig. 4), the regression lines for the six Ohio
DOT functional road classifications fell on top of each other, and the difference between their
mean panel ratings is less than the accuracy of the ride quality rating panel experiment of 0.2
to 0.3 MPR units. Correlation coefficients and standard errors of estimate for the regression
lines for the six Ohio DOT functional road classifications are shown in Table 4.
19,$~ + 191 8 Ride (tultllty D ktit
4
/r F~
-'- BC
-~" P C C
Comp
I
/
m-
9 i 9 * 9 9 i I * ~ | 9 * 9 i I 9 9 J |
0 @
0 1 2 :3 4
Estlmlted MPI~ (RH)
FIG. 3--Effect of pavement type on ride number performance.
TABLE 3--Ride number transform performance by

pavement type.
Correlation Coefficient, Standard Error of

Pavement Type r Estimate MPR
Bituminous concrete 0.95 0.27

PCC 0.92 0.20
Composite 0.88 0.31
All pavements 0.91 0.29
Urban Versus R u r a l Roads
It was thought that the ride quality rating panel members might rate urban roads differently
than rural roads. To test this hypothesis, best-fit straight line regression equations to the
measured mean panel rating and the computed ride number were computed individually for
the urban and rural roads in the interstate and principal arterial Ohio DOT functional road
classifications, as shown in Fig. 5.
The regression lines for the urban and rural ride quality data are nearly the same, and the
difference between the two is less than the accuracy of the ride quality rating panel experiment
of 0.2 to 0.3 MPR units. This comparison was of interest to the Ohio Department of Transporta-
tion since there is a public perception that the State of Ohio urban interstate and principal
arterial roads are in worse shape than the rural interstate and principal arterial roads. Correlation
15
4,
'- ---(l)
~3
/, --- ( s )
-"-(4)
/l- 9 §
.1"
ot.
0
| r
:1.
9 9
2
* 9 9 9
~
9 9 9 9
4.
9 9 9 9
15
~mtimmtecl MPR (RN)
FIG. 4~Effect of Ohio DOT functional road classifications.
TABLE 4--Ride number performance by road classification.
Correlation Standard Error

Functional R o a d Coefficient, of Estimate,
Classification r MPR
(1) Interstate 0.88 0.28

(2) Principal arterial 0.87 0.23
(3) Minor arterial 0.85 0.30
(4) Major collector 0.91 0.29
(5) Minor collector 0.97 0.18
(6) Local 0.90 0.32
coefficients and standard errors of estimate for the regression lines for the urban and rural
ride quality data are shown in Table 5.
Lane Configuration
To evaluate the effect on the ride panel ratings of the number of lanes in the road being
rated, best-fit straight line regression equations to the measured mean panel ratings and the
computed ride number were computed individually for the two-lane and multi-lane configura-
tions, as shown in Fig. 6.
19113 + 19a ! R i d e C a a l t t . y DI ~ta

/-
~3 Orba:
//
\ Rural
0 , i , .
o 1 2
Estimated ldPR (RN)
FIG. 5--Effect of urban versus rural on ride panel ratings.
TABLE 5--Ride number performance by locale.

Locale r Estimate, MPR
Urban 0.93 0.18

Rural 0.85 0.30
The regression lines for the two-lane and multi-lane road ride quality data are nearly the
same, and again the difference between the two is less than the accuracy of the ride quality
rating panel experiment of 0.2 to 0.3 MPR units. This small difference seems to show that there
is no difference in the public perception of pavement ride quality based on lane configuration.
Correlation coefficients and standard errors of estimate for the regression lines for the two-
lane and multi-lane ride quality data are shown in Table 6.
Time Stability
The evaluation of the ride number performance reported here was based on the use of a
combined ride quality dataset composed of the ride quality data collected in 1983 and 1988
(Tables 1 and 2 and Fig. 2). To evaluate the time stability of ride quality ratings made by a
ride quality rating panel, best-fit straight line regression equations to the measured mean panel
ratings and the computed ride numbers were computed individually for the 1983 and 1988
ride quality datasets, as shown in Fig. 7.
A comparison of the 1983 and 1988 ride quality datasets shows little difference in pavement
ride quality ratings by the 1983 and 1988 ride quality rating panels in the roughness areas of
5
\
, - Two ~ne
1
r a I I I t P |
0
0 1 2 3 4 5
Zstixnated MPI~ (RN)
FIG. 6--Effect of lane configuration on ride panel ratings.
TABLE 6---Ride number performance by lane configuration.

Number of Lanes r Estimate, MPR
Two-lane 0.94 0.26

Multi lane 0.81 0.29
most interest to the user. The biggest deviations are in the rougher end of the pavement
roughness spectrum, where the fewer number of pavement ride quality data points might
unduly influence the ride quality study results. It again will be noted that the difference in
the measured MPR for the same estimated MPR over most of the range of interest is less than
the accuracy of the ride quality rating panel experiment of 0.2 to 0.3 MPR units.
From this comparison, it has been concluded that there was no significant difference between
the 1983 and 1988 Ohio ride quality rating panels, and that the estimated MPRs were time
stable for the five-year period between 1983 and 1988.
Although the estimated MPRs were time stable for the last five-year period, no conclusion
can be made for their time stability from 1988 forward. It is therefore suggested that the time
stability of the ride quality rating panel and thus the ride number transform be periodically evalu-
ated.
5 ''
~ 3 f
,,'o
~2 ,. , f
J
1 X 1903 1988
Cor: 'elation -~ 0.93 0.91

Eat. Error = 0.31 0,27
I | l l l * | 9 i | . I t
O 9 . | ! i | , .
O 1 2 3 4 5
Estimated MPR (RN)
FIG. 7--Time stability of ride quality rating panels.
Ride Quality Transform Evaluation

A second objective of the 1988 Ohio Department of Transportation pavement ride quality
study was to compare the performance of the newly developed ride number transform to the
performance of other transforms used by the highway community to evaluate pavement ride
quality. As discussed in the introduction, there are four other transforms currently in use in
the United States to evaluate pavement ride quality. These four transforms are:
1. Texas Serviceability Index (SI).

2. Michigan Ride Quality Index (RQI).
3. Mays Road Meter Index (MRM).
4. International Roughness Index (IRI).
As part of the evaluation process, computer algorithms were implemented for each of the
four transforms. The inputs to the computer algorithms would be the 1983 and 1988 surface
profile datasets used to develop and evaluate the ride number transform discussed in the
previous sections. The outputs of the algorithms would be the computed indices for each of
the 141 surface profile datasets. To assist in the implementation of the transform algorithms,
the Texas Department of Highways and Public Transportation and the Michigan Department
of Transportation provided computer programs and sample datasets to ensure the correctness
of the implementation of their ride quality transforms.
The Mays road meter and International Roughness Index computer algorithms were based
on ASTM and World Bank parameters presented in their various publications [13-15].
The Texas Serviceability Index used the same 0 to 5 ride quality rating scale used in the
ride number transform. Rating scale transforms were developed for the other three ride quality
transforms to convert their outputs to that same 0 to 5 rating scale. Details on the computer
algorithm implementations for these four transforms are included in the final report on the
1988 Ohio Department of Transportation pavement ride quality [12].
Ride Quality Transform Comparison

For comparison purposes, the performance of the four other candidate ride quality transforms
were compared to the performance of the newly developed ride number transform. Using these
four transforms, an estimated mean panel rating (MPR) was computed from the measured
surface profile for each of transforms for the 141 pavement test sections included in the 1983
and 1988 Ohio Dot ride quality studies. As with the ride number analysis, a best-fit straight
line (regression equation) to the measured mean panel ratings (MPR) and the estimated mean
panel ratings was used to compute correlation coefficients and standard errors of estimate for
the four transforms.
The correlation coefficients and standard errors of estimates for the ride number transform
and the four other candidate ride quality transforms were then compared, as shown in Table
7. The other ride quality transforms are arrayed below the ride number transform with the
International Roughness Index having the lowest correlation and highest standard error of esti-
mate.
This comparison can be extended further by comparing the correlations and standard errors
of estimates of the five transforms by the three individual pavement types included in the
1983 and 1988 Ohio DOT ride quality studies.
The comparison of correlations coefficients for the five transforms by individual pavement
types is shown in Table 8. This comparison shows that the ride number transform has the
highest and most consistent correlations for the three individual pavement types with the
correlations for the other transforms being generally lower and less consistent.
The standard error of estimate comparison, Table 9, shows that the ride number transform
has the lowest and most consistent standard errors of estimate for the three individual pavement
types, with the standard errors of estimate for the other transform being generally higher and
less consistent.
From this comparison, it can be seen that the performance of the ride number transform in
estimating the subjective ride quality of highway pavement is superior to the other candidate
ride quality transforms for each of the three individual pavement types included in the Ohio
DOT ride quality study. This superior performance also suggests that the newly developed
TABLE 7--Comparison of ride number and other transforms.
Standard Error
Transform Correlation Coefficient of Estimate
Ride number 0.91 0.29

Michigan
RQI 0.81 0.42
Texas SI 0.79 0.44
MRM Index 0.79 0.44
IRI 0.77 0.46
TABLE 8--Correlation coefficients by pavement types.
Pavement Type
Bituminous Portland
Transform Concrete Concrete Composite All
Ride number 0.95 0.92 0.88 0.91

Michigan
RQI 0.80 0.82 0.85 0.81
Texas SI 0.92 0.84 0.67 0.79
MRM Index 0.92 0.78 0.70 0.79
IRI 0.91 0.78 0.70 0.77
TABLE 9--Standard errors of estimate by pavement types.
Pavement Type
Bituminous Portland
Transform Concrete Concrete Composite All
Ride number 0.27 0.20 0.31 0.29

Michigan
RQI 0.53 0.32 0.35 0.42
Texas SI 0.41 0.30 0.55 0.44
MRM Index 0.37 0.37 0.56 0.44
IRI 0.40 0.39 0.56 0.46
ride number transform should be considered a candidate for the preferred method for measuring
and reporting pavement ride quality in the highway community. To this end, an ASTM standard
practice for computing the ride number from longitudinal road profile measurements has been
proposed and is presently being balloted within ASTM Committee E-17.
To assist possible future users of the ride number transform, a computer listing and a sample
surface profile dataset were included in the final report on the Ohio DOT ride quality study
[12], which is available to the user.
Conclusions
This paper has discussed the development of the ride number transform as a method of
estimating subjective quality of highway pavements from surface profile measurements made
with the Ohio DOT inertial profilometer. As discussed in the paper, the development of the
ride number transform is the result of four NCHRP and Ohio DOT ride quality studies which
are reported in more detail in Refs 9 through 12. Included in these studies are an evaluation
of the performance of the newly developed ride number transform over an array of highway
pavement attributes and a comparison of the performance of the ride number transform to the
performance of four other candidate ride quality transforms including Michigan RQI, Texas
SI, MRM Index, and the International Roughness Index.
In this comparison, the performance of the ride number transform was found to be superior
to the performance of the other candidate ride quality transforms, producing estimates of
pavement ride quality with the highest correlation to the measured subjective ride quality and
with the lowest standard error of the pavement ride quality estimate.
The superior performance of the ride number transform suggests that the ride number
transform should be considered as a candidate for the preferred method for measuring and
reporting pavement ride quality in the highway community.
References
[1] Carey, W. N. and Irick, E E., "The Pavement Serviceability Performance Concept," Bulletin 250,
Highway Research Board, Washington, DC, 1960.
[2] Brokaw, M. P. "'Development of the Road Meter," Special Report 133, Highway Research Board,
[3] Walker, R. S. and Hudson, W. R., "Method for Measuring Serviceability Index with the Mays
Road Meter," Special Report 133, Highway Research Board, Washington, DC, 1973.
[4] Spangler, E. B. and Kelly, W. J., "GMR Road Profilometer--A Method for Measuring Road
Profile," Record 121, Highway Research Board, Washington, DC, 1966.
[5] Hudson, W. R., "High-Speed Road Profile Equipment Evaluation," Record 189, Highway Research
Board, Washington, DC, 1967.
[6] Darlington, J. R. and Milliman, P., "A Progress Report on the Evaluation and Application Study
of the General Motors Rapid Travel Road Profilometer," Record 214, Highway Research Board,
[7] Nair, S. K. and Hudson, W. R., "ServiceabilityPrediction from User-Based Evaluation of Pavement
Ride Quality," Record 1084, Transportation Research Board, Washington, DC, 1986.
[8] Holbrook, L. E, "Prediction of Subjective Response to Road Roughness by Use of the Rapid
Travel Profilometer," Record 291, Highway Research Board, Washington, DC, 1969, p. 212.
[9] Janoff, M. S., Nick, J. B., and Davit, P. S., "Pavement Roughness and Rideability," Report 275,
National Cooperative Highway Research Program, Washington, DC, 1985.
[10] Janoff, M. S., "Pavement Roughness and Rideability Field Evaluation," Report 308, National
Cooperative Highway Research Program, Washington, DC, 1988.
[11] Spangler, E. B. and Kelly, W. J., "Use of the Inertial Profilometer in the Ohio DOT Pavement
Management System," Report FHWA/OH-87t005, Ohio Department of Transportation, Columbus,
OH, 1987.
[12] Spangler, E. B., Kelly, W. J., Janoff, M. S., and Hayhoe, G. F., "Long-Term Time Stability of
Pavement Ride Quality Data," Report FHWA/OH-91/001, Ohio Department of Transportation,
Columbus, OH, 1991.
[13] ASTM E 1170-87, "Standard Practice for Simulating Vehicular Response to Longitudinal Profiles
of a Vehicular Traveled Surface," Annual Book of ASTM Standards, VoL 04.03, American Society
for Testing and Materials, Philadelphia.
[14] Sayers, M. W. et al., "The International Road Roughness Experiment," Technical Paper No. 45,
World Bank, Washington, DC, 1986.
[15] Sayers, M. W. et al., "Guidelines for Conducting and Calibrating Road Roughness Measurements,"
Technical Paper No. 46, World Bank, Washington, DC, 1986.
Brian Gilmore I
A Geometry-Based Contact/No-Contact
Simulation Model Applied to Ride
Quality
REFERENCE: Gilmore, B., "A Geometry-Based Contact/No-Contact Simulation Model
Applied to Ride Quality," Vehicle-Road Interaction, ASTM STP 1225, B. T. Kulakowski, Ed.,
ABSTRACT: In this work, ride quality is considered a function of the dynamic interaction
between a rider and the vehicle/seat. The seat is coupled to the human, which influences its
dynamics. This coupling may introduce situations where contact between the seat and the human
is lost, the seat bottoms out, etc., resulting in discomfort and/or personal injury.
This paper presents a design tool in the form of a computer simulation program for the above
systems. The tool uses a mechanical model of a human to evaluate ride quality and therefore
removes subjectivity from the analysis. The program predicts contact/no-contact situations and
automatically reformulates the equations of motion to represent the actual system topology.
Thus, the program can predict the bottoming out of a seat or a human leaving (losing contact
with) the seat. Since the program uses the geometric model of the system to make decisions
about contact/no-contact,the geometry of the seat and its effect on ride comfort can be predicted.
KEY WORDS: ride quality, mechanical system simulation, vehicle dynamics
Ride comfort is strongly affected by the dynamic characteristics of the vehicles tires and
suspension. A smooth ride requires a "soft" suspension that allows the wheels to move easily
over a bumpy road without transmitting significant forces to the car body. From the point of
view of vehicle safety and handling, "stiff" suspensions are required for good performance in
braking and cornering maneuvers. Most modern suspension designers try to compromise
between these two strategies to provide acceptable ride comfort without severely sacrificing
control and safety in cornering and emergency maneuvers [1]. In many instance the compromise
is difficult to achieve. For example, trucks and other vehicles with stiff suspensions for
supporting heavy loads subject their riders to a harsh ride. Emergency vehicles, ambulances
in particular, require stiff suspensions for high-speed safety but also need a smooth ride for
the patient's well-being. Much of the work on "ride quality" involves static test and human
juries. Thus, they neglect the system (including the seat and human) dynamics and are subjective
as opposed to objective evaluations. This paper presents a computational analysis and design
tool to consider the dynamic interaction between the rider and the seat.
Several early studies, Reiher and Meister [2], Jackline and Liddel [3], Janeway [4] and
TenCate [5], studied subjective response to vibrations. Most of the resulting tolerance curves
have similar shapes. Due to the definition of ride level comfort by each investigator, differences
among the comfort criteria exist.
Associate professor, Dept. of Mechanical Engineering, The Pennsylvania Transportation Institute, The
Pennsylvania State University, University Park, PA 16802.
150
GILMORE ON CONTACT/NO-CONTACT SIMULATION MODEL 151
Other efforts have attempted to quantify ride comfort through measurement of human
response. Van Deusen [6] measured acceleration at the seat and head of the passenger in three
mutually perpendicular directions. He found that the mean vertical acceleration of the body
correlated with jury ratings. Griffin presents methods of quantifying vibration with respect to
human response [7]. He suggests six measures: comfort ride value, overall ride values, vibration
spectra, vibration dose values, activity disturbances, and motion sickness dose values. However,
Griffin collects his data from human subjects, even though subjects' variability can affect
results [8,9]. Sato and Johansen applied the concept of pressure distribution to design car seats
in order to improve ride comfort [10]. They used a plate-deflection-photographic method to
determine the pressure distribution. In a related effort, Reed et al. [11] considered driver
discomfort and seat design. They measured driver movement, seat pressure distribution, and
driver perceptions using a static driving simulator. Therefore, their work did not consider the
dynamic interaction of the driver and the seat.
Several researchers have considered the dynamic interaction of the rider and seat. Donoghue
[12] used "in-the-vehicle" instrumentation and concluded that the interposition of the seat and
back cushion between the vehicle and human body drastically changes the nature of the
vibrations transmitted to the seated person. For example, at low frequencies, vibrations may
be amplified while high frequencies are attenuated. Therefore, Donoghue measured this cushion
effect by simultaneously recording support vibration and human vibration. Pietsch [13] devel-
oped an automatic leveling air suspension seat which was independent of the driver's weight
and therefore provides a full range of motion. This work is based on the power spectral density
(PSD). He states that his study does not include events where the seat bottoms or tops out.
Scarzalski and Loverich [14] conducted a forensic study of a spinal injury due to a vehicle
seat bottoming out while transversing a berm.
In an effort to develop a design and analysis method, Pradko and Lee [15-17] introduced
the concept of "absorbed power." Absorbed power is an energy flow rate dependent on the
anatomical properties of the human body and relatable to passenger subjective responses.
International Standards Organization (ISO) standards provide performance guidelines. Three
degrees of disturbance are postulated in the ISO standard: reduced comfort, fatigue-decreased
proficiency (FDP), and an exposure limit above which an acute physical hazard due to the
vibration is considered to exist [18]. The standard suggests methods of dealing with broad
frequency band linear vibration in a given direction, but does not deal adequately with combined
effects of different directional vibration components or effects of rotational vibration. Griffin
[71 points out the importance of pitch and roll motions.
In order to remove the subjectiveness of human jurors from the evaluation and consider
the dynamic interaction, Wambold and Park [19-21] developed a mechanical simulator to
evaluate ride quality. This simulator eliminates subjectiveness due to day-to-day variations
inherent in humans, such as mood, as well as differences between humans, such as sex,
physique, etc. The mechanical simulator is based on a lumped parameter model. This simulator,
composed of a passive system of springs, dampers, etc., can be made to give very nearly the
same response as a human body [22,23]. As indicated in Refs 19-21, this model when
used with the absorbed power concept accurately and objectively predicts ride quality. Other
advantages of a mechanical simulator include:
1. Identification of specific frequencies and amplitudes that cause poor ride quality.
2. Ability to obtain data from a prototype vehicle which has not been certified for
human use.
3. Capability of comparison of rides through use of identical mechanical simulator models
(an objective standard).
4. Reduced need for large numbers for subject for routine ratings.
5. Mechanical simulator is invariant with respect to season of year, sex, prejudice, day-
to-day feelings, etc. (i.e., a simulator is a constant transducer).
6. Applicable to computer simulation.
This paper presents a mechanical system simulation approach to model the dynamic interac-
tion of the driver with the vehicle including situations where the driver may separate from
the seat. Many works, such as those in Refs 24-29 and 31 and 30, contributed to the development
of simulation algorithms for dynamic mechanical systems. These algorithms have event handlers
for changing kinematic constraints; however, the constraint changes and resulting kinematic
configuration must be anticipated by the user. This can be a formidable task when nonlinear
dynamic mechanical systems are considered.
Even though some of the above programs like ADAMS and DADS can handle anthropomor-
phic figures, another set of simulation programs have evolved in the automobile industry
(i.e., MADYMO and CAL-3D) which have been developed to consider the interaction of an
anthropomorphic figure with a vehicle [31-34]. These programs are typically used to study
vehicle/human interaction during a collision.
Gilmore and Cipra presented an algorithm that automatically predicts and detects the changes
in the constraints and then reformulates the dynamic equations of motion for planar dynamic
mechanical systems with changing topologies [35]. They characterized the kinematic constraint
changes using the concepts of point-to-line contact kinematic constraints, force closure, and
ray firing together with the information provided by the rigid bodies' boundary representation
(b-reps), state variables, and reaction forces. No a priori data pertaining to the constraint changes
or the intermediate system topologies is required by the algorithm, nor is any assumption made
on the system behavior. Later work by Han and Gilmore and Han, Gilmore, and Ogot extended
this algorithm to handle stick/slip friction, impact with friction, and bodies whose boundary
representations include arcs as well as lines [36,37].
Simulation Approach
In this work each physical body is modeled as a separate geometric entity and dynamic
subsystem described by its boundary representation and stated variables, respectively. Physi-
cally, the contact, friction, and impact between the bodies takes place at the boundaries of
each body. Therefore, the geometry of the boundaries is considered by the algorithm. The
approach employs the concepts of geometric modeling, dynamic mechanical systems, and kine-
matics.
This problem and its solution has many components. The time at which the constraints
change must be determined, and the proper constraint equation must be generated to couple
the bodies or deleted to uncouple the bodies. Friction physically plays an important role and
therefore must be considered along with its two regimes, static and kinetic friction. The solution
to the impact must be determined, and the differential dynamic equations of motion must be
reformulated and solved. Even though these components may appear diverse, they must be
considered together and are combined into a viable package.
Even though the major portion of this work will focus on the rules to change the kinematic
constraints, the equations of motion formulation must be considered such that the equations
are easily manipulated to reflect the constraint changes. Therefore, the following discussion
is broken into two major portions: the formulation of the equations of motion and the develop-
ment of the rules to determine the kinematic constraint changes.
The approach has several sections. First, a brief overview of the concept is discussed. Then
the geometric model for the bodies is presented. The general point contact constraint equations
are derived along with the dynamic and impact equations of motion. The rules to change the
kinematic constraint equations is then presented.
Force Closure and Point Contact
The type of kinematic constraint which exists between two physical bodies is a force-closed
kinematic joint [38]. This means that a force is required to keep two bodies in contact and
implies that two bodies in contact through a force-closed joint may separate when the closing
force becomes insufficient to maintain the joint. Separation can also be caused by an impulsive
force due to an impact or when one body slides off another. Conversely, force-closed joints
are established when two bodies collide. The boundaries of the bodies prevent them from
penetrating each other.
As mentioned earlier, the geometry of the boundaries is considered by the algorithm, and
therefore the geometric model of the bodies must be established. The geometric boundaries
of the bodies are assumed to be composed of lines and circular arcs. It is felt that this is not
a major restriction since the geometric boundaries of most manufactured parts can be represented
by a combination of lines and arcs. The bodies can then be reduced to a set of vertex points
(i.e., the endpoints of the boundary lines and circular arcs) and center points of the arcs. The
centers of the arcs are considered since they are a reference for all of the points on the circular
boundary. In this way, a body is reduced to a representation of points in order to carry out
the prediction and detection of the constraint changes. Also, the constraint equations represent-
ing the physical contact can be written using these points.
In many situations it may not be clear which points will provide an independent set of
constraints since an infinity of points may be in contact. This problem is overcome by
constraining those points which provide an equivalent set of "point contacts." In planar
kinematics, three basic types of joints exist and each one can be modeled by an equivalent
set of point contacts [39]. Each independent point contact removes one degree of freedom.
This reduces the problem to writing a constraint equation for each independent point contact.
However, this requires point contact kinematic constraint equations for all possible combinations
of geometry which can result from bodies defined by arcs and lines.
Point Contact Kinematic Constraint Equation and Formulation of the Dynamic Equations
of Motion--The dynamics of the bodies are coupled together through point contact constraint
equations. Shown in Fig. 1 are two bodies in point contact. The (X,Y) coordinate system is
the inertial coordinate system, while the (x,y) coordinate system is fixed to a boundary of
Body 1 (i.e., the contacting boundary). Point a2 on Body 2 is instantaneously in contact with
Point a~ on Body I. Point a2 can slide along the boundary of Body 1, and therefore Body 2
may rotate or slide with respect to Body 1. In this case there is one point contact and two
degrees of relative freedom. Since the bodies are rigid and cannot penetrate each other, Eq 1
is true while the bodies are in contact.
y.2(X2. Y2,02, r.2) = y.,(Xb YI,OI, l'al ) (1)
Note that Y,2 and y,~ are implicit functions of the geometric model. The vectors r,2 and r,, are
defined from the center of gravity to the points on the boundaries which are constrained. Since
the dynamic equations of motion are in terms of accelerations, Eq 1 must be differentiated
twice to yield the required acceleration constraint equations.
02
X
r176
~'- 0,1
01
/
/
Contact Point a
X
VIlIIiIA
FIG. 1--Point-to-line contact.
yo2(R2,~'2,O2,O2,ra2) = ya,(Xl,?~,~l,Ol,r,,) = f(position, velocity) (2)
where the right-hand side is a function of position and velocity. Equation 2 is the general
form of the contact constraint equations. Reference 35 derived this relation for a vertex
point in contact with a line and is shown by Eq 3, the acceleration point-to-line constraint
equation [35].
"~l sindO. - R2 sindO. - i~1 cosdO. + ~2 cosdOo - r~,,O, + r.~02 = r.,yO~ + ro~O2 (3)
+ 2{(X2 - X,) cosdO. + (f'2 - I~) sindO. - r.vO2)}Ot
Similar relationships are derived and applied to the other types of contact, such as arc-line,
convex arc-concave arc, etc.
The point contact acceleration constraint equations are placed into a sparse coefficient matrix
formulation of the system equations of motion. Since each body is considered as a separate
subsystem, the equations of motion of each subsystem appears in the system equations and
are coupled together through the constraint equations. Equation 4 is a representation of the
sparse coefficient matrix formulation, where [M] is the system mass matrix, {X} is the
acceleration vector, {h } is the lagrange multiplier vector which represents the reaction forces,
f(~,x,t) is the applied force vector, [C] is the coefficient matrix for the left-hand side of Eq 2,
and {f(position, velocity)} is the vector corresponding to the right-hand side of Eq 2.
= } (4)
[C] [dO] J [ h J [f(pos,vel)J
For each independent point contact in the system, a row appears in [C] and {f(position,
velocity)}, representing that constraint.
GILMORE ON CONTACT/NO-CONTACTSIMULATION MODEL 155
Knowing how to write the kinematic constraint equation and reformulate the equations of
motion is straightforward once the system topology is known. Therefore, the algorithm predicts
and detects the changes in the constraint equations to keep track of the system topology, which
allows it to reformulate the equations of motion. This necessary component to the algorithm
is presented later.
Static~Kinetic Friction--A very essential component to the dynamics is the friction between
the bodies. However, friction is a nonlinear effect since it has two regimes, static and kinetic.
Therefore, the simulation model must switch between these two regimes.
Previous methods to simulate systems with static/kinetic friction apply a kinematic constraint
to the system during the static friction phase [26]. This can produce two difficulties. First, the
kinematic constraint must be applied and released accordingly, which requires a time step
interpolation or halving technique. Second, for many systems, the additional kinematic con-
straint due to the static friction can be a redundant constraint, which would cause the coefficient
matrix of the dynamic equations of motion to become singular. This would preclude the
simulation of a system switching between static friction and kinetic friction.
The model presented here considers the friction as an applied force, regardless of its regime.
The product of the static coefficient of friction and reaction force is only the maximum possible
static friction force, and therefore the static friction force must be determined [40]. During
the static friction phase, the relative velocities at the points of contact are zero. Thus, the
magnitude of the static friction force is the value which balances the forces such that acceleration
and therefore the velocity at the points of contact remain zero [41]. The static friction forces
can then be applied at the points of contact. This is possible only because the boundary
representations of the bodies are included in the dynamic analysis. In situations where the
static friction forces are indeterminate, an equivalent force-couple system is applied as discussed
in Ref 37.
The kinetic friction phase occurs when the body is moving or the applied forces are greater
than the maximum possible static friction force. Similar to the static friction force, the kinetic
friction force is also applied at the points of contact. Both the kinetic and static friction forces
are placed into the right-hand side of Eq 4. The logic to determine the friction phases and the
corresponding forces is summarized and presented in Fig. 2 [37].
The general approach presented here offers several advantages over previous models when
static/kinetic friction is considered. The normal reaction forces are the lagrange multipliers
from the equations of motion and are the normal reaction forces. This is very helpful since
the dynamic friction force is considered to be a function of the normal reaction force. Also,
since the boundary representation of the bodies is included into the dynamic analysis, the
friction forces can be applied directly at the points of contact. Finally, the conversion between
static friction and kinetic friction is inherent in the dynamic equations of motion formulation
and do not require a time interpolating technique as previous methods.
Algorithmic Prediction of the Kinematic Constraint Changes

The system equations of motion change at particular points in time while the simulation is
discrete in time. Therefore, the time at which the kinematic constraints change must be
determined and synchronized with the numerical integration procedure. To accomplish this
task, an interval halving technique is employed to determine the point in time at which the
kinematic constraints change.
The following discussion identifies and characterizes each of the four ways by which
kinematic constraints change. The characterization provides the algorithm with the rules to
determine how and when the constraints change, permitting it to reformulate the equations of
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authori
~- Begin ]
+
Monitor the relative tangential velocity on each contacl[
point and adjust time step according to the sliding /
velocities ]
t
Classify contact points L
V
Formulate system equation
no Is there any redundant

contact point?
yes
Estimate the independent static friction forces
Solve system equation
110
iterations small?
any siletion point, is smile frie~ yes

larger than max. static friction?
no
Re~ 1
FIG. 2--Sticklslip friction algorithm [37].
motion. Also, these rules provide a fundamental understanding of how kinematic con-
straints change.
Essentially, the rules determine when in time and the points which become coincident when
contact between the bodies is established or separate when the bodies break contact. Therefore,
the rules determine when to change the dynamic equations of motion and the points on the
boundaries which are to be constrained or unconstrained. These rules are broken into two
groups: the prediction rules which determine the possibility of a constraint change and the
time until the event occurs and the detection rules which state that a change is occurring at
that instant in simulation time. The rules of an algorithmic approach to determine the boundary
collision are discussed first.
Constraint Formation Due to Boundary Collision
In order to write the constraint equations for the impact and dynamic equations of motion,
the points of contact and time of impact between the bodies must be determined. However,
the general dynamic motion of bodies is typically difficult for a person to predict. Therefore,
the algorithm must pick up the burden. The constraint is due to a boundary of a body establishing
contact with a boundary of another body. Thus, the problem is reduced to determining when
one or more points on a body's boundary come into contact with a boundary of another body.
According to the geometric model proposed above, collisions between the bodies can occur
in three ways: between two circular arcs, a line and an arc, and a line and a line (which can
be reduced to a vertex point and a line [35]). As an example, the prediction and detection
rules for a vertex point and line are presented.
Physically, if a point is going to collide with a boundary, it must have an instantaneous
velocity towards that line. As shown in Fig. 3, this implies that the dot product o f the relative
velocity of a point and the exterior normal of a boundary line must be negative for a collision
to occur. This rule is stated analytically by Eq 5 and is used to determine from the entire set
of points and lines, those pairs which have a potential to collide.
V(j,n) 9 n(i,m) < 0 (5)
Equation 5 is applied to each combination of the bodies. However, Eq 5 is only a statement

of relative orientation, and therefore a test to determine relative position is also required.
Since a line boundary is finite in length, relative position must be considered to determine
if a point moving along a path defined by its relative velocity vector will collide with a body.
A ray r~.,) is extended from the point along its instantaneous relative velocity vector. If the
ray intersects the boundary line, then the point has a potential to collide with the line. This
technique is similar to the ray firing technique used in computer graphics [42], This condition
is analytically stated by Eq 6.
V(j+l,n)
~ , n )
n(i-l~
,/r v(j,n)
FIG. 3--Prediction of collision.
ro,.) 9 VO.n)
r0,. ~ n boundary i of body m I = 1 (6)
Iro..~ 9 vo,.)l
The portion of the ray which extends from the point to the boundary line defines the relative
distance between their respective bodies and is symbolically represented by Ir0..)h Equations
5 and 6 define the prediction rule. The relative distance, as defined above, and its time
derivatives are used to predict the time of impact and with an interval halving technique is
employed to determine the time of impact. The predicted time is used to estimate a time step
such that the colliding bodies will not be allowed to pass through each other. They are
repetitively applied until either they are no longer true or Eq 7 becomes satisfied.
Ir0.,)l = 0 (7)
Equation 7 is the detection rule for this event. When Eq 7 is true, the boundaries of the bodies
have interfered with their motion. Thus, the bodies have collided. Note the prediction rules
are necessary but not sufficient conditions for boundary collision.
The above rules reduce the problem of determining how and when boundaries collide to
that of considering particular points of the boundaries' geometry. These points, along with the
bodies' state variables, are used to predict and detect boundary collisions.
When the algorithm detects that the bodies have collided, it performs an impact analysis.
The impact analysis determines the system response. The new contact is represented by an
additional constraint equation in the impact analysis. The system impact response is then
determined. The algorithm then determines if contact is maintained or broken due to the
restitution impulse.
Constraint Removal--It is also necessary to develop prediction and detection rules for the
event of constraints breaking such that the algorithm can determine how and when kinematic
constraints break in order to remove the corresponding point contact constraint equation from
the dynamic equations of motion. A kinematic constraint may break due to three causes:
restitution impulse of an impact, an insufficient closing force, or a point sliding off a boundary.
Constraints Break Due to the Restitution Impulse o f an I m p a c t - - T h e effect of an impact

is to impart an instantaneous velocity change to the system. If a constraint breaks due to an
impact, then the bodies have enough relative velocity to separate at the contact and its point
contact velocity constraint is violated.
Since the change in velocities occurs within an infinitesimal time period after the boundary
collision, the time at which the constraints break is known. Therefore, the characterization of
constraints breaking due to an impact requires only the detection rule and is stated by Eq 8
in terms of Fig. 1.
9 + 9 +
yat(timpaet) -- y,2(timpact)< 0 (8)
where the velocity terms are the post impact velocities. Equation 8 is applied to each point
contact pair to determine if the constraint has been broken by the restitution impulse. If Eq
8 is true, then the constraint has broken due to the impact, and the constraint equation is
deleted from the dynamic equations of motion.
Constraints Breaking Due to an Insufficient Closing F o r c e - - A constraint breaking due to

an insufficient closing force is characterized by considering the constraint's reaction force.
The kinematic constraint equations are formulated such that the corresponding normal reaction
force has a negative magnitude when it is compressive. As the closing force become insufficient
to maintain the constraint, the reaction force approaches zero from the negative side. Therefore,
when the reaction force is equal to zero, the constraint is broken and its equation is removed
from the dynamic equations of motion. The detection rule of this event is stated analytically
by Eq 9.
F~ = 0, i = 1 . . . number of constraints (9)
Also, Eq 9 and the secant root finding method are combined into an interval halving method
to form the prediction rule for this event.
Constraints Break Due to Relative Sliding--A third way for kinematic constraints to break
is for a point constrained to a boundary to slide off. Two situations can result from this type
of change in the kinematic constraints; as an example, consider the situation shown by Fig.
4. Initially Block 2 is sliding relative to Block 1. Points A and B are on Block 2 and are
constrained to Line CD, and Points C and D are on Block 1. Point B is the endpoint of the
line and arc on Block 2. As Block 2 slides from Position 1 to Position 2, Point B slides off
Line CD, and Point D becomes constrained to Line AB. Effectively, the constraint has reversed
its incidence. In this case, the previous constraint is removed from the equations of motion
while the latter constraint is added. As the motion of the system progresses from Position 3
to Position 4, Point D relative to Block 2 slides off Line AB on Block 2. When this constraint
breaks, the system gains an additional degree of freedom due to relative sliding.
Even though relative sliding between the bodies may have two different results, when a
point slides off a boundary line, the corresponding kinematic constraint equation must be
removed from the equations of motion. Again, to develop the prediction and detection rules,
the event is characterized. The characterization of this event is similar to the boundary collision
characterization. The type and location of contact are known from the topological data. This
identifies the possible set of constraints which may break. Each line has two endpoints, while
the constrained point can slide off only one end. To determine the end which the point is
going to slide off, the dot product of the point's relative velocity vector with the exterior
normal of the boundary on either side of the line is determined. The dot product indicates
which end the point is moving towards as shown by Fig. 5. The condition for a point to slide
off a particular end of a line is analytically stated by Eq 10.
v(j.,) 9 n (i_*i. mj > 0 (10)
For example, the Point j on Body n in Fig. 5 will slide off endpoint (i) since v0,,) 9 n(H.,) >
0. Note, if the adjacent boundary is an arc, then the normal of the line from the vertex point
to the arc center is considered. The distance from the point to the end of line is defined by
that length of the boundary line from the point of contact to the designated end, as shown by
Fig. 5 and symbolically represented as r0,, ). Hence, using the topological data, Eq 10, and the
point's relative acceleration, velocity, and position, the time for the constraint to break can be
algorithmically predicted. When the magnitude of r0.,) is equal to zero, the point has slid off
the line, and the corresponding kinematic constraint equation is removed from the equations
of motion. This condition is analytically stated by Eq 11 and is the detection rule for this event.
{(rr = 0 (11)
Body 2
A I:;:'
"":::'
~;::~::;:~.
::~'~::':"":":;:':::'~BJ.~.-
.......
~ I =~ ~
IN\\\\\\\\\\\\\\\\N21
Position 1
Body 2
~ '
I~'~\\\\\\\\\\\\\\\\~,~1
~
Position 2
Body 2 \ 9
Position 3
Body 2
Position 4
FIG. 4--Changes in kinematic constraints during sliding.
Impact Analysis
When a multibody system collides with a single body or with another multibody system,
impact dynamics with friction should be considered. Han and Gilmore [36] present a general
computer-oriented analysis of impact dynamics incorporating friction. The presence of friction
between sliding contacts during the impact makes the problem difficult since the events such
as reverse sliding or sticking, which may occur at different times throughout the impact, must
be determined. The boundary representations of the bodies are used to solve for the velocities
at the points of contact. Using this information and a classification of the modes of impact,
the frictional impact with sliding contact problem is solved. It was shown by theory and
,, Iy
+1
O
Body n J"
Vd,n)
n (i- 1 , m )
, i kldi /
i-1 -]ZTql+
(~ Body m ]~n(~+l,m)
i+2
FIG. 5--Prediction of constraint breaking due to relative sliding.
experiment that several modes of impact such as sticking, reverse sliding, and forward sliding
exist. These modes of impact are characterized and formulated into a general computational
strategy. Many difficulties of previous approaches have been obviated since each physical
rigid body is considered as a separate geometric entity (i.e., described by its boundary representa-
tion) and dynamic subsystem.
Examples
Two examples are presented. One example considers the rolling motion of the vehicle, while
the other example considers the pitching motion. The vehicle for both examples is based on
that presented in Ref 43 and is considered to be traveling at 1.3 m/s. The simulated rider in
both cases has a mass of 140 kg. It is based on the mechanical simulator developed by Ref
21, but is a reduced model which considers only the predominant vibratory mode at 4.7 Hz.
Both examples start the simulation at steady-state initial conditions.
Example 1--Vehicle Roll

Trunk sway and stability of the rider is often of concern, especially for geriatric riders [44].
This example demonstrates the effect which seat curvature (i.e., geometry) has on rider comfort.
The rider model has a two-point contact with the seat, as discussed in Ref 44. The vehicle
transverses a 0.09-m-deep by 0.1-m-long pothole. Figure 6 shows the system model. Simulation
results are shown in Figs. 7 through 11, where the solid lines denote the response for a seat
radius of 0.4 m and the dot-dashed lines indicate the response for a seat radius of 1.0 m.
Figure 7 shows the normal reaction force between the rider and seat multiplied by the vertical
velocity with respect to the ground of the rider's lower mass. This is important since absorbed
power can be defined as the integral of this quantity. The values for the two cases are very
close, and therefore the curves appear to lie over each other for this example. Figures 8 and
9 show the reaction forces between the seat and the rider. As defined by the above equation
of motion formulation, negative reaction forces are compressive. Figure 10 shows the difference
of trunk sway between the two cases, while Fig. 11 compares the time derivative of this value.
The short dashed curves show the vehicle angular velocity for both cases. Note that the angular
velocity and position of the rider are not explicitly stated in the equations to develop the data
~/ Seat Radius
i~i uman Simulator
I':
~/ ~/~
j
i ~ Roll
F1 F2
f Excitation Excitation f
FIG. 6---Example 1, simulation model.
100 I
-I00"
-200
0.0 0.5 1.0 1.5 2.0
Time (Sec)
FIG. 7--Force multiplied by velocity for both cases.
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions a
- 25 A
< Y tA[- -x
~.
_,ooI,,0.0 0.5
,,,
1,0
Time (Sec)
1.5
,,,
2.0
FIG. 8--Reaction forces (F2) between seat and rider.
for Fig. 7. Thus, this example shows that, for two different seat curvatures, the absorbed power
can be similar, while trunk sway can be different. If absorbed power was used as the only
measure of performance for this example, the difference between the two cases would not
become apparent. Therefore, two situations can have similar values of absorbed power, but
due to different geometry the situations can have different motions, in this case increased
sway. The differences in the response can be explained by considering the seat/rider as an
inverted pendulum. The larger radius seat causes the instant center between the seat and the
rider to be longer than that for the smaller radius seat.
Example 2--Vehicle Pitch

This example uses the same model for the rider as above; however, the vehicle is a side
half-car. The vehicle is again moving at 1.3 m/s and transverses a berm. The berm is a ramp
which rises 0.23 m in 0.15 m and then has a constant amplitude of 0.10 m. The purpose of
this example is to demonstrate that the simulation algorithm can handle contact/no-contact
situations. The model used in this example is shown by Fig. 12. Figure 13 shows the reaction
force between the seat and rider multiplied by the vertical velocity with respect to the ground
of the rider's lower mass. Figures 14 and 15 show the reaction forces between the rider and
the seat. Again, negative reaction forces are compressive forces. These figures show that the
rider loses contact with the seat shortly before 0.3 s. This occurs when the reaction forces
become equal to zero. Contact is re-established shortly after 0.4 s, as evidenced by the
non-zero reaction forces. Figures 14 and 15 show another discontinuity or disturbance at
approximately 1.45 s. At this point the rider and seat have again collided, where the rider
-200T
-2so-.1- v
- 3oo--"~'~..~ ~- " / ' ~ , ; .,...: 7,

:~.j;
=" ,% , [ " 4 ~
" t :V .V-
,ooJ I,i
f
0.0
, , , , I , ,0 .,5 , I , , , , I1.0, , ,,,,
Time(,Sec)
1.5
FIG. 9--Reaction forces (Fz) between seat and rider.

I
2.0
moves towards and hits the back of the seat. Figure 16 shows the angular position of the rider
and the vehicle. Figure 17 shows the angular velocity of the rider and the vehicle. After 0.4
s when contact is re-established, the rider vibrates in the angular direction with a high-frequency
component. This is due to the fact that contact is not re-established evenly on the seat and
the fact that the model is a discrete system. In this simulation model, the contact points did
not consider local deformation and therefore there was a very tight coupling of the dynamics
of individual bodies. In actuality, local deformation (i.e., deflection of the seat cushion) at the
contact points (or areas of contact) would exist. Thus, some of the high vibration shown in
Figs. 16 and 17 would be absorbed by the seat cushion.
Summary and Conclusions

Many previous studies of ride quality involved human subjects or are restricted to vibration
analysis of linear systems. The subjective approaches are not well suited for computer imple-
mentation and are more of an evaluation tool as opposed to a design and analysis tool. The
linear vibration studies removed much of the subjectivity. However, they typically have not
been concerned with nonlinearities due to geometry or changing kinematic constraints between
the rider, seat, and vehicle.
This paper considers the dynamic interaction between the rider and seat. The computer
program predicts contact/no-contact situations and automatically reformulates the equations
of motion to represent the actual system topology. Thus, the program can predict the bottoming
out of a seat or a human leaving (losing contact with) the seat. These are important occurrences
since they can harm the rider. Since the program uses the geometric model of the system to
make decisions about contact/no-contact, the geometry of the seat and its effect on ride comfort
.003
.002-
~ .O01-
v
i .000=
i' 1 / ,.
~-/
~ ".l I
.', %1
9
.001=
8'
,(
-b ',i I 9
-'
.002 'UI
.003
0
,,,,I,,,,I,,,,I,,,,I 0.5 1.0 1.5 2.0
Time(Sac)
FIG. lO---Angular positions of rider.
0.04-- --
II I /%
o.o~ ?~. i~ i~
9~. I I,.. :, ;:.i ;
1 1~1 ' .P %. II .'t"~1 I o
I 2 I. 9 '; . u. I . I io r 9 ,
0.00-.I '' ~4% .~" -,, " : .- , ,.- "
i'r:" "! !? ":,~? ; 9

o.o~. -~,".., ;J !? ,~::
0.04 /
O0
I,,,,I,,,,i,,,,I
0.5 1.0 1.5 2.0
Time (Sac)
FIG. 11--Angular velocities of rider and vehicle.
can be predicted. Therefore, the program can serve as a simulation tool to consider geometry
effects which the rider, seat, and vehicle apply to each other. It should be noted that the approach
presented in this paper is based on a discretized rigid body model of the system. Thus, the values
of the discrete parameters must be determined, and the contact points do not account for local
deformation or deflection in the seat cushion. In actual practice, care must be used to ensure that
the vibration modes correlate to the physical continuous systems.
Human Simulator
FIG. 12 Example 2, simulation model.
500-
i~
-500
I000 m
0 O0 0.25 0.50 0.75 1.00 1.25 1.50
Time (Soc)
FIG. 13--Force multiplied by velocity for both cases.
9 Og .25 g .5~ 9.75 1.00 1.25 1.50

Time (See)
FIG. 14---Reaction force (Ft) between seat and rider.
O.OO 0.25 e.so 0.75 1.00 1.25 1.50

(See)
FIG. 15----Reaction force (F2) between seat and rider.
.004
.OOZ
.OOe-
!
.OOZ-
9'" .... I,. ,I .... I .... I .... I .... I

o.ee o.25 o.se o.75 1.co 1.z5 1.5o
Time (Sec)
FIG. 16---Angular position of rider and vehicle.
1 . O..-l--
~,fRid&
o.5-:--
O. O,. ,-,
>
~' -o. ~-:--
- I . ~.4=-
-2..
O.O0 O.Z5 O.SO 0.75 1.00 1.25 1.50
(See)
Time
FIG. 17--Angular velocity of rider and vehicle.
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions
Current work is applying similar concepts as those discussed above to determine appropriate
parameters for the seat suspension [45]. Then the suspension can be kinematically synthesized to
yield the appropriate parameters. In mass transit systems where the layout is constrained, the above
concepts can be used to help design seats and their supports.
References
[1] Kulakowski, B. T., "Effect of Seat Design on Ride Comfort," unpublished notes, November 1987.
[2] Reiher,H. and Meister, F. J., "Die Emfindlichkeit des Menschengegan Erschultengen," Forschung
aufdem lngenierswesens, Vol. 22, 1931, pp. 381-386.
[3] Jacklin, M. H. and Liddel, G. J., "Riding Comfort Analysis," Engineering Bulletin, Research Series
No. 44, Purdue University, Lafayette, IN., May 1933.
[4] Janeway, R. N., "Ride and Vibration Data," Special Publication SP-6, Society of Automotive
Engineers, Warrendale, PA, 1950.
[5] Ten Cate, W., translated by R. D. Murray, "Vibration Nuisance," Royal Aircraft Establishment,
Great Britain, October 1957.
[61 Van Deusen, B. D., "Analytical Techniques for Designing Riding Quality," SAE 670021, Society
of Automotive Engineers, Warrendale, PA, 1967.
[7] Griffin, M. J., "Evaluation of Vibration with Respect to Human Response," SAE 860047, Society
of Automotive Engineers, Warrendale, PA, 1986.
[8] Wambold,J. C. and Park, W. H., "Objective Measures of Ride Quality for Railroads," Report PTI
7801, The Pennsylvania Transportation Institute, The Pennsylvania State University, University
Park, PA, January 1978.
[9] Fairley, T. E. and Griffin, M. J., "The Apparent Mass of The Seated Human Body: Vertical
Vibration," Journal ofBiomechanics, Vol. 22, No. 2, 1989, pp. 81-94.
[10] Sato, T. and Johansen, D. W., "Comfort Design Using Molded Seat Foam Techniques," SAE
890846, Society of Automotive Engineers, Warrendale, PA, 1989.
[11] Reed, M. P., Saito, M., Kakishima, Y., Lee, N. S., and Schneider, L. W., "An Investigation of
Driver Discomfort and Related Seat Design Factors in Extended-Duration Driving," SAE 910117,
Society of Automotive Engineers, Warrendale, PA, 1989.
[12] Donoghue, W. C., "Instrumentation for Truck Ride Data," SAE 660136, Society of Automotive
[13] Piestsch,H., "The Automatic Leveling Air Suspension Seat for Industrial Vehicle," SAE 840507,
[14] Scarzalski, K. J. and Loverich, E. B., "Vibration Analysis Methods Applied to Forensic Engineering
Problems," Structural Vibration and Acoustics, DE-Vol. 34, American Society of Mechanical
Engineers, New York, 1991, pp. 197-206.
[15] Pradko,F. and Lee, R. A., "Vibration Comfort Criteria," SAE Publication No. 660139, Society of
Automotive Engineers, Warreudale, PAL,1966.
[16] Pradko, F. and Lee, R. A., "Analytical Analysis of Human Vibration," SAE 680091, Society of
[17] Pradko,F., Lee, R. A., and Kauluz, V., "Theory of Human Vibration Response," ASME Publication
No. 66-WAIBHF-15, American Society of Mechanical Engineers, New York, 1966.
[18] "Guide for the Evaluation of Human Exposure to Whole-Body Vibrations," ISO 2631, 2nd ed.
Np. 2631-1978 (E), International Standards Organization, Paris, 1978.
[19] Park, W. H. and Wambold,J. C., "Objective Measures of Ride Quality," PTI Report 7425, Pennsylva-
nia Transportation Institute, The Pennsylvania State University, University Park, PA, December
1974.
[20] Park, W. H. and Wambold, J. C., "Objective Ride Quality Measurement," SAE paper 760360,
[21] Wambold,J. C., "Vehicle Ride Quality--Measurement and Analysis," SAE 861113, Society of
[22] Coermann, R. R. et al., "The Passive Dynamic Mechanical Properties of the Human Body System,"
Aerospace Medicine, Vol. 31, No. 6, 1960, p. 443.
[23] Suggs, C. W., Abrams, C. F., and Strikeleather, L. F., "Application of a Damped Spring Mass
Human Vibration Simulator in Vibration Testing of Vehicle Seats," Ergonomics, Vol. 12, 1969,
pp. 79-90.
[24] Smith,D. A., "Reaction Forces and Impact in Generalized Two-DimensionalMechanical Systems,"
Ph.D. thesis, University of Michigan, Ann Arbor, 1971.
[25] Haug, E. J., Computer Aided Kinematics and Dynamics of Mechanical Systems Volume 1: Basic
Methods, Allyn and Bacon, Needham Heights, MA, 1989.
[26] Hang,E. J., Wu, S. C., and Yang, S. M., "Dynamics of Mechanical Systems with Coulomb Friction,
Stiction, Impact and Constraint Addition-Deletion-I: Theory,-II: Planar Systems," Mechanisms and
Machine Theory, Vol. 21, No. 5, 1986, pp. 401-416.
[27] Lotstedt, P., "Numerical Simulation of Time-Dependent Contact and Friction Problems in Rigid
Body Mechanics," SlAM Journal of Scientific and Statistical Computing, Vol. 5, No. 2, 1984,
pp. 370-393.
[28] Lotstedt, P., "Mechanical Systems of Rigid Bodies Subject To Unilateral Constraints," SIAM Journal
of Applied Mathematics, Vol. 42, No. 2, 1982, pp. 281-296.
[29] Nikravesh, P. E., Computer-Aided Analysis of Mechanical Systems, Prentice-Hall, Inc., Englewood
Cliffs, NJ, 1988.
[30] Orlandea, N., Chace, M. A., and Calahan, D. A., "A Sparsity-Oriented Approach to the Dynamic
Analysis and Design of Mechanical Systems--Part I," ASME Transactions, Journal of Engineering
for Industry, American Society of Mechanical Engineers, New York, 1977, pp. 773-779.
[31] Huijbers, J. J. W. and Janssen, E. G., "Experimental and Mathematical Car-Bicycle Collision
Simulators," SAE 881726, Society of Automotive Engineers, Warrendale, PA, 1988.
[32] Nieboer,J. J. and Wismans, J., "Status of MADYMO 2D Airbag Model," SAE 881729, Society
of Automotive Engineers, Warrendale, PA.
[33] Deng, Y.-C., "The Importance of the Test Method in Determining the Effect of Door Padding in
Side Impact," SAE 892429, Society of Automotive Engineers, Warrendale, PA, 1989.
[34] Deng, Y.-C., "Analytical Study of the Interaction Between the Seat Belt and A Hybrid III Dummy
in Sled Tests," SAE 880648, Society of Automotive Engineers, Warrendale, PA, 1988.
[35] Gilmore, B. J. and Cipra, R. J., "Simulation of Planar Mechanical Systems with Changing Topolog-
ies, Part 1: Characterization and Prediction of the Kinematic Constraint Changes, Part 2: Implementa-
tion Strategy and Simulation Results for Example Dynamic Systems," ASME Transactions, Journal
of Mechanical Design, Vol. 113, American Society of Mechanical Engineers, New York, 1991,
pp. 70-83.
[36] Han, I. and Gilmore, B. J., "Multi-Body Impact Motion with Friction--Analysis, Simulation and
Experimental Validation," ASME Transactions, Journal of Mechanical Design, Vol. 115, American
Society of Mechanical Engineers, New York, 1993, pp. 412-422.
[37] Han, I., Gilmore, B. J., and Ogot, M. M., "The Incorporation of Arc Boundaries and Stick/Slip
Friction in a Rule-Based Simulation Algorithm for Dynamic Mechanical Systems with Changing
Topologies," ASME Transactions, Journal of Mechanical Design, Vol. 115, American Society of
Mechanical Engineers, New York, 1993, pp. 423-434.
[38] Reuleaux, E, The Kinematics of Machinery, MacMillan and Co., London, 1876.
[39] Soni, A. H., Mechanism Synthesis and Analysis, McGraw-Hill Book Co., New York, 1974.
[40] Greenwood,D. T., Principles of Dynamics, 2rid ed., Prentice-Hall, Englewood Cliffs, NJ, 1988.
[41] Karnopp, D., "Computer Simulation of Stick-Slip Friction in Mechanical Dynamic Systems,"
Transactions, ASME Journal of Dynamic Systems, Measurement, and Control, Vol. 107, March
1985, American Society of Mechanical Engineers, New York, pp. 100-103.
[42] Newman, W. M. and Sproul, R. E, Principles of Interactive Computer Graphics, McGraw-Hill
Book Co., New York, 1973.
[43] Chalasani, R. M., "Ride Performance of Active Suspension Systems--Part 2: Comprehensive
Analysis Based on a Full-Car Model," ASME AMD Vol. 80, DSC, Vol. 2, L. Segel, Ed., American
Society of Mechanical Engineers, New York, 1987, pp. 205-234.
[44] Ferguson-Pell, M. W., "Seat Cushion Selection," Journal of Rehabilitation Research and Develop-
ment--Clinical Supplement No. 2, March 1990, pp. 49-73.
[45] Berger,E. and Gilmore, B. J., "Seat Dynamic Parameters for Ride Quality," SAE 930115, Society
for Automotive Engineers, Warrendale,PA, 1993, also accepted for publication in SAE Transactions.
Advances in Vehicle Suspension
Design and Control
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authori
Chris J. A l s t e a d 1 a n d John P. W h i t e h e a d 1
Steering and Suspension Development of

Road Vehicles
REFERENCE: Alstead, C. J. and Whitehead, J. P., "Steering and Suspension Development
of Road Vehicles," Vehicle-Roadlnteraction, ASTMSTP 1225, B. T. Kulakowski, Ed., American
ABSTRACT: The evaluation and development of a vehicle's ride and handling characteristics
are still largely carried out by subjective means within the automotive industry. However, the
need to reduce product development times and to "design in" acceptable handling characteristics
at the prototype stage requires a more formalized approach to handling development. The use
of objective measurement methods, alongside subjective techniques, in a vehicle development
program is described, and it is argued that these methods allow a better understanding of vehicle
characteristics and progress development more rapidly than using subjective techniques alone.
Prediction of the ride and handling characteristics of a vehicle at the design stage is only
possible in very general terms, and this leads to lengthy development work being carried out
on the preproduction prototype. Continuing and future research programs are expected to lead
to closer correlation of subjective and objective results and the verification and refinement of
a mathematical model for use as a tool for the prediction of handling behavior at the design stage.
KEY WORDS: vehicle behavior, ride, handling, objective testing, subjective testing
When the design of a prototype vehicle is laid down, the designers cannot predict with any
confidence the vehicle's final dynamic behavior. Constraints are imposed on the suspension
designer by such considerations as drivetrain packaging, body styling, and chassis or floorpan
details and tend to dictate the type of suspension system finally chosen. The prototype's
suspension design is based, at best, on past experience of chassis systems that have been
found, after development, to give acceptable ride and handling behavior. The suspension
designer therefore faces a herculean task, and the prototype inevitably requires a considerable
amount of development work to achieve ride and handling standards acceptable in the market-
place.
Chassis development is largely achieved by subjective means and requires the development
engineers to have not only a sound understanding of vehicle dynamics, but also highly developed
driving skills. Although subjective techniques will never be eliminated from the development
process, they provide little or no formal feedback to the design team. Economic requirements
to reduce development times and market pressures to improve vehicle behavior have brought the
need to objectively define vehicle ride and handling characteristics. Such objective information
provides a link between a vehicle's subjective behavior and design changes and ultimately
will aid the prototype designer.
For a number of years the Motor Industry Research Association (MIRA) has been integrating
subjective and objective test techniques in its development programs, together with laboratory
Consultant and department manager, respectively, Vehicle Dynamics, Motor Industry Research Associ-
ation, Watling St., Nuneaton, Warwickshire, CVI0 0TU, United Kingdom.
173
measurement of suspension kinematic and compliance characteristics, and damper behavior.

This has led to a greater understanding of steering and suspension design requirements and
more rapid and successful development of prototype vehicles.
Current and future research programs in computer modeling of suspension and whole vehicle
systems and the correlation of subjective opinion with objective test results will further the
understanding of chassis dynamics and ease the task of prototype design.
Ride
Work carried out by MIRA in the 1950s and 1960s on the correlation of subjective and
objective ride measurements led to the (now obvious) conclusion that the acceleration levels
to which vehicle occupants were exposed were the most important measured parameters. By
filtering and weighting the acceleration signals, it was found that good subjective/objective
correlations could be obtained. These studies led to the building of a single-channel Ridemeter
which grew in sophistication over the years and evolved into the present six-channel machine.
This machine was designed in conjunction with M. Griffin of ISVR (the Institute of Sound
and Vibration Research at Southampton University) around the requirements of the Ford
Motor Company. It is in use by vehicle manufacturers around the world and is employed in
development programs and in objectively ranking competitor vehicles.
The philosophy behind the Ridemeter is to reduce the complex acceleration signal from
each channel to a single number by taking its averaged root mean square (rms) value over a
fixed time period. For comfort evaluation, accelerometers are positioned between the occupant
and the seat cushion and seat back and adjacent to the feet. To provide data that correlate well
with subjective ratings, the acceleration signals can be filtered and weighted according to
several criteria, including BS 6841 [I] and ISO 2631 [2].
By being quick to install and by providing easily assimilated results at the time of test, the
Ridemeter lends itself to use in subjective development programs, especially in spring and
damper tuning exercises, where changes to the vehicle are made rapidly. The Ridemeter
augments the engineer's subjective judgment and allows an objective record to be kept of the
development process. Very consistent and repeatable data can be produced if the vehicle is
run at an accurate and constant speed over a surface devoid of any significant discrete features,
and by careful choice of test surfaces, different aspects of a vehicle's ride behavior may be
investigated separately.
In a ride development program, it is normal practice to include laboratory measurements
of component and system characteristics in the initial and in subsequent modified conditions.
Indeed, both the laboratory and road test measurements will lead the direction in which the
modification process proceeds. Typically, laboratory examinations would include, for example,
measurements of suspension vertical wheel deflection versus load characteristics, suspension
friction and hysteresis levels, bump stop characteristics, bushing characteristics (and consequent
suspension lateral and longitudinal compliances), and damper (or shock absorber) characteris-
tics.
If a specific problem is being investigated, the Ridemeter approach is often too broad, and
time history records are necessary from accelerometers and displacement transducers in loca-
tions other than in the seating area. Spectral analysis can be used to identify possible resonant
conditions, and examination of the time histories of a vehicle's response to discrete events is
often useful.
An example is given in Figs. 1 and 2 of the results of a damper development exercise using
the MIRA Ridemeter set to the ISO 2631 comfort criteria. The vehicle was an air-suspended
doubledecker bus, and Ridemeter measurements were taken on three road surfaces representing
a good quality major road or motorway, a secondary road characterized by long wave undula-
ALSTEAD A N D WHITEHEAD O N STEERING 175
E
I ,d-
E
oI
0)
r~ ~
o t-
o iZ
rr 0 0
/
A
o
o
ii., i i 0
co o
punoq~s d~n8
q I
d
L_
ot
c5"~ t-
. D
E E
o9
r~ >,
~
o
o
~
r
0
LL o
A
Z ~
punoq~N dtuns"
{,,"3
ALSTEAD AND WHITEHEAD ON STEERING 177
tions, and a poor quality minor road with surface breakup. Measurements were made of the
seat cushion vertical and seat back longitudinal averaged rms accelerations from the extreme
front and rear seating positions. Damper settings were changed numerous times during the
development program, and the effect made on ride quality was monitored using the Ridemeter
results and the test engineers' subjective judgment. Figure 1 shows the damper force against
velocity characteristics at the start and the end of the development program, and Fig. 2 shows
the changes made to the seat cushion vertical ride quality by the altered damper settings.
As can be seen, the changes made to the rear damper settings made virtually no difference
to the force against velocity characteristics, while at the front only the rebound characteristic
was changed to any significant degree. The results on the front seat vertical acceleration levels
were reductions of 6% on the smooth surface and 10% on the other surfaces. At the rear seat
position, the reductions were 12% on the smooth surface, 5% on the secondary road surface,
and 3% on the poor quality surface. The reason for the generally higher levels of acceleration
on the secondary road surface compared to the poor quality surface at the rear seat position
is that the secondary road surface contains a long wave undulation, which induces pitching
motion most evident at the rear of a bus.
Handling
The term "handling" embraces many different aspects of a vehicle's lateral dynamics, and
at least two objective test procedures are required to quantify a vehicle's basic handling
characteristics. The correlation between subjective "feel" and objective results is less well
established for handling than for ride, and interpretation of objective test data requires much
experience. For these reasons, the evaluation and development of a vehicle's handling character-
istics is still largely carried out by subjective means within the automotive industry. However,
the need to reduce product development times and to "design in" acceptable handling character-
istics at the prototype stage requires a more formalized approach to handling development pro-
grams.
Objective handling test techniques are used by MIRA alongside subjective methods in both
development and evaluation programs and have proved to be especially useful in problem-
solving exercises. Original work on the objective evaluation of handling performance was
carried out by MIRA in the 1960s and 1970s and led to a significant contribution to the
formulation of the ISO handling test procedures. These test procedures concentrate on "open
loop" maneuvers, where the driver makes a specific control input irrespective of the vehicle's
response. Tests involving "closed loop" maneuvers such as a lane change or slalom around
obstacles are unsuitable for discrimination of all but gross vehicle characteristics because the
driver control inputs obscure the vehicle behavior. This effect is more pronounced with a
highly skilled driver who unconsciously uses complex control strategies.
It has been found that two open loop tests give a good general description of vehicle handling
behavior. These are the steady state circular test, and the pseudo random steer input test.
The steady state circular test consists essentially of driving the vehicle on a constant radius
at a series of incremented constant speeds. At each speed, the vehicle is held steady on the
radius with steady handwheel and throttle positions. Data are collected at each steady-state
condition using a PC-based acquisition system running dedicated analysis software that allows
results to be computed and checked before leaving the test area. Parameters normally measured
are: lateral acceleration, handwheel steer angle, handwheel steer torque, roadwheel steer angles,
body slip angle, and body roll angle.
The principal information yielded by the steady state circular test is the steady-state handling
balance (understeering or oversteering behavior). Other useful information to be gained from
the test are the body roll angle gradient with respect to lateral acceleration, handwheel steer
torque characteristics (a very strong feedback cue to the driver), and steering system compliance
characteristics (from the difference between handwheel steer angle and roadwheel steer angles).
Results are usually plotted against lateral acceleration, and the handling balance is most
simply described by the relationship between the handwheel angle and lateral acceleration. A
more sensitive measure of understeer is the parameter dHW/dSLIP, which is the rate of change
of the handwheel angle with respect to the rate of change of rear wheel slip angles. This
parameter can be used to distinguish the effects of small modifications to the vehicle and
allows the comparison of different vehicles when normalized by steering ratio and wheelbase.
It can also be used to calculate the position of the neutral steer point [3], which is an important
parameter in predicting a vehicle's crosswind stability.
The pseudo random steer input test is designed to measure the transient handling behavior
in the linear operating regime, which is typically up to about 0.4 g lateral acceleration for a
medium-sized European sedan. The vehicle is driven at a constant speed in a nominally straight
line, and the driver inputs a. swept sine wave to the handwheel from the lowest possible
frequency allowed by the track width to the highest attainable frequency of about 4 Hz. The
handwheel input amplitude is kept nominally constant and is of such magnitude as to keep within
the vehicle's linear handling regime. Parameters normally measured are: lateral acceleration,
handwheel steer angle, handwheel steer torque, roadwheel steer angles, body yaw rate, body
roll rate, and body roll angle. As with the steady-state circular test, data are collected using
a PC-based acquisition system, and frequency response functions of vehicle and steering system
with respect to handwheel steer input are computed.
These two tests, steady-state and pseudo random input, have been used as the bases for
ISO Standard Procedures 4138 and 7401 [4,5]. Together they are able to quantify the behavior
of a vehicle over its normal operating range in everyday use on the public road. With MIRA's
current analysis methods, the effects of suspension modifications to a vehicle are readily
apparent.
Other procedures are available to extend the vehicle testing regime to higher lateral accelera-
tions, even up to limit conditions (steering step input test [6]) or to "real life driving maneuvers"
such as braking in a turn [7], power change in a turn, and response to a wind gust (ISO
procedures in draft).
Suspension and steering development to improve handling, as for ride improvement, will
involve vehicle modifications. Laboratory measurements will be necessary to check that system
behavior matches design intent, where applicable, and to monitor the modification process.
Measurements of suspension kinematic and compliance characteristics and of steering system
friction, hysteresis, and power assistance characteristics would be normal here.
To illustrate the sensitivity of the objective track measurements to vehicle modifications,
the effect of a change of tire fitment has been chosen. The vehicle concerned was a 3-L, front
engine, rear-wheel-driven sedan. The two tire fitments, both steel-braced radials, were of the
same brand, of identical size, and set to identical inflation pressures, but one was S speed
rated and the other was H rated.
Figure 3 shows the lateral force versus vertical load characteristics of the two tires for
various slip angles at the appropriate inflation pressure for the vehicle of 1.85 bar (26 lb/in.2).
The wheel loads of the vehicle in test condition were approximately 40 kN (400 kgf) front
and 37 kN (370 kgf) rear. The tires are labelled S and H appropriate to their speed rating.
The results of the steady-state test are shown in Fig. 4 as a plot of the parameter dHW/
dSLIP, normalized for steer ratio and wheelbase, plotted against lateral acceleration. Values
above the horizontal axis, i.e., positive, indicate understeer, the greater the value the greater
the understeer. Negative values would indicate oversteer. As can be seen, the vehicle is
understeering with both tire fitments, but the understeer decreases with increasing lateral
Lateral Force (kN)
50 10 deg
Slip Angle
Tyre Size 175-14
Inflation Pressure ~ ~
40 7 deg
Slip Angle
30
y / .,"
4 deg
Slip Angle
20
2 deg
10 Slip Angle
0 I I I I I I b
0 10 20 30 40 50 60 70
Vertical Load (kN)
--H Speed Rated S Speed Rated

. . . . . .
FIG. 3--T/re lateral force versus vertical load.
acceleration. At low levels of lateral acceleration, tire H provides more understeer, but then
levels reduce more sharply than those provided by tire S.
In the context of this paper, the most interesting result from the pseudo random steer input
test is the plot against frequency of the transfer function of yaw rate with respect to handwheel
steering input. Figure 5 presents the yaw rate transfer functions for the vehicle with both tire
fitments at a road speed of 80 km/h (50 mile/h). The desirable shape of the yaw rate response
curve is flat to within +-- 3 dB (gain ratio of 0.7 to 1.4) to high frequencies with a peak at the
yaw natural frequency. Modem passenger cars typically produce a response "flat" up to 2.5
or even 3 Hz with a peak around 1 Hz. The peak is a result of understeer, and an oversteering
vehicle or a slightly understeering, heavy, and slow-responding vehicle such as a typical laden
17-ton truck produces no peak at all.
As can be seen, Tire H produces a lower gain level at low frequencies (0.2 Hz) but a larger
rise at around 1 Hz. Both these features are a result of the greater amount of understeer at
low lateral acceleration conferred by Tire H and is shown in Fig. 5.
Figure 6 shows the time delay of the yaw rate response relative to the handwheel input.
Tire H produces the shortest time delay, and this is consistent with its more understeering
behavior. Objectively, the difference between the behaviors of the car with the two tire fitments
were small, and this was also the case subjectively, although, nevertheless, drivers showed a
clear preference for Tire H.
(dHW/dSLIP)/(Steer Ratio x W h e e l b a s e ) deg/m/deg

0.4
0,3
0.2
0.1
Anti-clockwise Clockwise
I ..... I I _ I l I I
0
-0.8 -0.6 -0.4 -0.2 0.2 0.4 0.6 0.8
Lateral Acceleration (g)
--H Speed Rated ...... S Speed Rated

FIG. 4---Gradient of normalized handwheel steer angle with respect to rear slip angle for rear
drive sedan.
Gain Ratio (Normalised to Response of H Tyre at 0.2 Hz)

1.2
1.1
0.9
0.8
I .... 1 t I I . I l t I
0.7
0.1 1
F r e q u e n c y (Hz)
-- H Speed Rated ....... S Speed Rated

FIG. 5--Frequency response function gain of yaw rate with respect to handwheel angle input
for rear drive sedan.
Time Delay (ms)
-60
-80
-100
-120 I I J I I I I I I I I I
0.1 1
Frequency (Hz)
-- H Speed Rated ........ S S p e e d Rated

FIG. 6---Time delay of yaw rate with respect to handwheel angle input for rear drive sedan
Subjective Techniques
Subjective evaluation of vehicles will always be important and will never be completely
replaced by objective tests. Quite often a subjective evaluation is a quick way to discover the
existence of a problem that can be more thoroughly investigated by objective testing. It is
also impossible to evaluate ride and handling together by objective methods, whereas these
can be dealt with together subjectively. This is important because a modification made to
change a vehicle's handling characteristics is quite likely to change the ride characteristics
and vice versa.
Subjective techniques are only useful if carried out with considerable discipline on the part
of the driver. As with objective testing, "open loop" maneuvers are used as much as possible,
and the vehicle is tested on particular road track surfaces and discrete features at consistent
speeds. Only by these methods is the driver able to discriminate details of the vehicle's ride
and handling characteristics. This leads to a better understanding of the effect of suspension
and steering changes on the behavior of a vehicle and to a better correlation of subjective
judgment against objective measurement.
Future D e v e l o p m e n t s
A research program to evaluate "on center" handling, or steering "feel" and vehicle directional
response during predominantly straight line driving, has recently been completed. The aim
was to correlate subjective feel by expert test drivers with objective measurements and to
define an objective test schedule for evaluation and development purposes. The work is
expected to lead to more efficient trouble shooting on problem vehicles.
A doctoral research project on active suspension systems applied to heavy goods vehicles
is nearing completion and has yielded a patented system with possible applications to both
commercial and passenger vehicles. A second doctoral research program is due to start in the
near future. It will address the problem of correlating subjective feel with objective test
measurements and verification and refinement of a mathematical model to be used as a tool

for the prediction of handling behavior at the design stage.
A new twelve-channel digital Ridemeter is being designed to be PC compatible and user
programmable. Thus, while including ISO and BS specified filters and analysis, it would also
enable users to specify their own filter shapes and analysis routines and to record short time
histories. The latter would be useful for examining vehicle responses to discrete (as opposed
to continuous) events.
Plans are in formulation to build a facility which would provide full kinematic and compliance
data on a car within one or two days. Consideration is being given to augmenting this with
an MTS Flat Trac II tire tester. These facilities would provide a more rapid development
capability and a full data-gathering service for computer modeling.
Conclusions
The paper has argued that the need to reduce development times and improve the quality
of vehicle ride and handling characteristics necessitates the increased use of objective test
techniques to quantify vehicle behavior. It has been noted, however, that subjective testing
will continue to play an important part in prototype development programs.
Objective measurement methods and improved subjective/objective correlation are used to
quantify vehicle behavior. The paper has outlined the methods in use and under development
at MIRA, methods which comprise laboratory measurements and dynamic track and road
testing of vehicles. For the latter, the paper stresses the need for "open loop" rather than
"closed loop" test schedules.
References
[1] BS 6841:1987, "Measurement and Evaluation of Human Exposure to Whole-Body Mechanical
Vibration and Repeated Shock," British Standards Institution, London.
[2] ISO 2631:1974, "Guide to the Evaluation of Human Exposure to Whole-Body Vibration," Interna-
tional Organization for Standardization, Geneva.
[3] Barter, N. F., "Measurement of Vehicle Handling by Tethered Testing," Proceedings of the Institute
of Mechanical Engineers, London, Vol. 184, Part 2A, No. 11, 1969-1970.
[4] ISO 4138:1982, "Road Vehicles--Steady-State Circular Test Procedure," International Organization
for Standardization, Geneva.
[5] ISO 7401:1988, "Road Vehicles--LateralTransient Response Test Methods," International Organiza-
tion for Standardization, Geneva.
[6] ISO TR 8726:1988, "Road Vehicles--Transient Open-Loop Response Test Method with Pseudo-
Random Steering Input," International Organization for Standardization, Geneva.
[7] ISO 7975:1985, "Road Vehicles--Braking in a Turn---Open Loop Test Procedure," International
Organization for Standardization, Geneva.
Noise Emission Due to
Vehicle-Tire-Road Interaction
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproduction
Peter Wagner ~
Vehicle Noise Testing Problems Due to

Vehicle/Tire/Road Interactions
REFERENCE: Wagner, R, "Vehicle Noise Testing Problems Due to Vehicle/Tire/Road
Interactions," Vehicle-Road Interaction, ASTM STP 1225, B. T. Kulakowski, Ed., American
ABSTRACT: European Community legislation will decrease the limits of drive-by noise
according to ISO 362 from 77 to 74 dB(A) in 1995. In Germany, the government will also
enforce a new emission-related car tax, probably in 1993.
These planned restrictions make it necessary to look more closely at the influence of tire/
road noise in relation to the drive-by noise level during vehicle type approval. Volkswagen has
built four test tracks according to ISO/DIS 10844 and compared the influence of tire/road noise
on these and other tracks in relation to drive-by noise.
The main differences of noise level were measured on different road surfaces. Nearly compara-
ble is the effect of different tires. It must be considered that tire/road noise may be much higher
under accelerating conditions than under coast-by conditions. These noise levels generally do
not correlate. The effect of temperature has a large influence on drive-by noise, especially when
measuring in the third gear. Aging of the test track should be recognized the first year after
construction. The car itself influences the sound propagation through the design of the wheelhous-
ing and the floor of the car body. Tire inflation pressure and vehicle weight may influence the
noise level of low-noise cars.
If these effects are not taken into account, drive-by noise may spread in the range of 3 dB(A)
even on an "ISO" test track.
KEY WORDS: tire/road noise, road surface influence, tire influence, temperature effects, drive
torque influence, heated "ISO" track, aging of "ISO" tracks
European Legislation
The growing traffic density, particularly in large cities and surrounding areas, leads to the
inhabitants of the central and outlying areas being increasingly disturbed by noise. This gives
the legislator reason to tighten existing legal regulations by reducing the existing limits or by
creating tax advantages for low-noise vehicles. Currently, the most critical legislation with
regard to the limit value and its introduction date is planned in Europe. Figure 1 shows
the scenario.
There will be a modification of Regulation 70/157/EWG (EEC) in Version 84/424/EWG
by reducing the current limits of 77 to 74 dB(A) (draft of 06/25/91). The application is planned
for approval of each new type from October 1994 and for all first-time registrations from
October 1995, with the measuring method as per ISO 362 (revised in N o v e m b e r 1992: October
1995 and October 1996).
l Development engineer, Volkswagon AG, Research and Development, Dept. 1731, PKW-Versuch-
Aknstik, Postfach, 38436 Wolfsburg, Germany.
185
Copyright 9 1994 by ASTM International www.astm.org
k;
,%
r~
WAGNER ON VEHICLE NOISE TESTING PROBLEMS 187
The Nordic Council (DK, IS, N, S, SF) has been in favor of tax support for "low-noise"
vehicles [<73 dB(A)] since 1992. This applies to first-time registrations (measuring method
as per ISO 362).
The EC plans an additional reduction of the external noise limit in this decade to 7l dB(A),
which allows tax support before this time.
In Germany, an emission-dependent motor vehicle tax beginning in 1993 is being discussed.
In this case, the minimum noise-dependent tax that has been suggested is achieved with 71
dB(A) or less.
Other national legislations (e.g., Switzerland and Austria) will automatically go into effect
when the above regulations become effective.
Due to these modifications to the regulations expected in the near future, considerable
pressure results to develop vehicles with reduced external noise levels.
In order to avoid competitive distortion under these basic conditions, it is necessary to
standardize the conditions under which type approval takes place. Influences which may lead to
the increase of the unavoidable variation ranges must be recognized and, if necessary, corrected.
For this reason, Volkswagen began early to conduct studies on the influencing variables,
particularly with regard to tire/road noise. Four test tracks were built in accordance with ISO
recommendations in order to gather knowledge and to include the results in the work of the
group ISO/TC43/SC1/WG27.
Some of the most important results are presented below. All measurements were made with
cars and car tires. The results do not correlate with the tire/road interaction of heavy vehicles.
Measurements were made under two drive-by conditions:
1. Accelerated drive-by as per ISO 362.

2. Coasting or coast-by, which means the car is driven at constant speed with the sound
sources of engine, intake, and exhaust virtually eliminated. The maximum noise level
in dB(A) is measured at the microphone position as per ISO 362.
Noise Sources of Passenger Cars With Diesel and Spark Ignition Engines
In order to determine the contribution of tire/road noise in the total noise of passenger cars,
a partial sound source analysis was used. This means that each single sound source like engine,
intake, exhaust, and tire/road noise has been identified by eliminating the other sound sources.
Based on three typical vehicles taken from the Volkswagen model line, it can be shown
what part of the overall external noise level the tire/road noise makes up. The vehicles
chosen were:
Vehicle 1--2.0 L/100 kW Passat, five-speed manual gearbox, Conti Supercontact 195/
60 VRI4.
Vehicle 2 - - 1 . 9 L155 kW Diesel Passat, five-speed manual gearbox, Conti Supercontact 195/
60 R14.
Vehicle 3 - - 2 . 8 L/128 kW VR6 Passat, automatic, Conti Supercontact 195/55 RI5.
The results in Fig. 2 show that the tire/road noise proportion for the vehicles with a manual
gearbox has a main influence on the level in third gear. In second gear, engine noise is
dominant, and tire/road noise contribution directly follows the engine noise share. The intake
noises and the exhaust system noises occur at a lower noise level. The tire/road noise is
dominant for the vehicle with an automatic gearbox.
=t
,&
Influence of Road Surface and Tires
The first systematic studies by Volkswagen with the University of Stuttgart from 1987 show
that the road surface influence leads to greater noise level differences when compared to the
tire influence within the tire/road noise. The noise and traction of twelve tire types were tested
on five different road surfaces [1]. The tire sizes were 175/70 R13, 185/60 HR14, and 185/
60 VR14, and the roads were made of asphalt concrete with various grain sizes and cement
concrete. The measurements were taken with the coasting method.
Differences of 2.3 dB(A) (Tire 10) to 4.3 dB(A) (Tire 3) per tire resulted at 50 km/h due
to the road surface influence. This is shown in Fig. 3. The differences per road surface between
the tires ranged from a minimum of 2.3 dB(A) (Surface 3) to a maximum of 4.1 dB(A)
(Surface 1). At higher speeds, the variation ranges increased. This study did not confirm the
fear that tires with good traction must also be noisy.
However, measurements conducted at a later time on twelve different tire types (185/60
R14, 175/70 R13S, and 175/70 R13H) showed that, in general, tires with good traction
characteristics tend to be noisier. In this case, lateral hydroplaning, longitudinal hydroplaning,
wet handling, wet braking, and circular driving were studied. An example of wet handling is
shown in Fig. 4 (best value means the subjectively identified best handling quality). Neverthe-
less, it can be stated that quiet tires with good traction can also be produced.
Similar studies were conducted and published in recent years by various institutions and
automobile manufacturers. Some of the results published in Germany should be mentioned here.
Betzl [2] conducted measurements with one tire type on 17 different road surfaces (cement
concrete, asphalt concrete, "ISO" asphalt concrete in accordance with ISO/DIS 10844, and
drainage asphalt). At 50 km/h with the coasting method, differences up to 9 dB(A) resulted.
The effect of the different road surfaces on drive-by noise according to 84/424/EWG was 5.3
dB(A) in second gear and 6.4 dB(A) in third gear. The average value differs by 5.8 dB(A)
(Fig. 5).
Steven [3] conducted measurements with nine tire types on nine different road surfaces.
The influence of the road surfaces is given as nearly 9 dB(A) here (Fig. 6), while the influence
of the studied tires on one road surface resulted in noise level differences of up to 4 dB(A)
(coasting method).
It is recognized that the road surface influence on the tire/road noise is dominant and may
reach up to 9 riB(A). The road surface is not defined in 84/424/EWG, and a restriction only
exists in Germany by the Federal Ministry of Transportation: "drainage asphalt surfaces are
not allowed." This means that there is an urgent need to define and legally establish a "standard
road surface" for type approval. The proposal on this is currently being discussed in the form
of ISO/DIS 10844.
Influence of Tires on "ISO" Road Surfaces
All tires to be approved for handling quality by Volkswagen are studied with regard to their
influence on type approval. Measurements of tire/road noise were taken with a vehicle on
which the other sound sources have been virtually eliminated and which provides a high drive
torque in accordance to the test procedure described in ISO 362. It has been shown that the
coasting test is not suitable for describing the effect of tire/road noise on an accelerated drive-
by as the rank ordering according to noise level of tires measured during coasting does not
always correlate with the rank ordering under torque. The noise level increase as compared
to coast-by level was measured on the "ISO" road surface with a maximum of 5 dB(A). In
particular, wide tires and tires with a low aspect ratio, which have a high noise level even
while coasting, show clear noise level increases under torque. This also applies to winter tires.
=t
k;
;=
.3
FIG. 4---Tire~road noise and wet-handling. (Source: Volkswagen.)
Figure 7 shows the variation range of the tire/road noise and the accelerated drive-by noise
for a Volkswagen vehicle with 17 different tire types (195/50 R15, 185/55 R15, 195/60 R14,
and 205/50 R15) on the "ISO" road surface. From this it can be seen that the tire influence
is 3.6 dB(A) even in second gear. This means that under torque a tire/road noise more than
78 dB(A) results. The influence of the tested tires in third gear is about 3 dB(A). This means
that the tire/road noise reaches 72 dB(A). During coast-by a variation range of 5.7 dB(A) was
measured (including the slicks).
Steven [3] indicates a noise level increase under torque of 3 dB(A) for summer tires and
even up to 5 dB(A) for winter tires. The influence of tire width is illustrated in Fig. 8.
Hieronimus [4] reports a variation range of 3 dB(A) for production tires (coast-by) and a
noise level increase of up to 4 dB(A) under traction.
Enz and Steven [5] describe a comparison of ten "ISO" road surfaces carried out with six
vehicles (and tire types). Averaged for all vehicles, a road surface influence with the maximum
variation of 2 dB(A) results for the measurement of accelerated drive-by noise as per ISO 362.
It is shown, with regard to type approval, that in addition to standardization of the test
track, there is a further need to standardize the tire type used during the test procedure.
Temperature Influence on Noise Value as per ISO 362
The sources of external vehicle noise are temperature-dependent in various ways. In the
least favorable case these contributions add up and lead, for example, to a clearly less favorable
type approval value in winter than is achieved in summer.
Figure 9 shows the results of the analysis of a vehicle at different ambient temperatures on
the "ISO" road surface. The temperature dependency of the tire/road noise is dominant and
particularly influences the drive-by noise as per 84/424/EWG in third gear. The mechanical
noise of the engine is constant in the first approximation, as the type approval is only carried
out with the engine at operating temperature. However, an increasing air temperature results
in a power output drop of the engine (ISO 1585, SAE J 1349, EWG 80/1269, DIN 70020)
and thus in altered inlet and exhaust noises. If a possible temperature range from - 1 0 to
+30~ (February and May) is used for type approval periods in Germany, this means that the
type approval values may fluctuate by approximately 2 to 3 dB(A).
k;
J~
L~
=:
k;
=::
('~
t~
r~
I
FIG. 8--Coast-by noise level as a function of tire width. (Source: Steven.)
For this reason it continues to be necessary that, in addition to the standardization of the
road surface and the used tires, a temperature correction of the measured value also be stipulated.
Only then can it be avoided that, for example, vehicle manufacturers from the Mediterranean
countries have a systematic competitive advantage over the Scandinavian manufacturers. A
reduction of drive-by noise by 2 dB(A) can mean an increase in production costs of several
hundred Deutsche marks because of further reduction of the noise sources and the measures
against the temperature increases usually resulting from this. This can be a measure of just
how important a legally prescribed temperature correction is against systematic competitive dis-
tortions.
Analyses of the temperature influence are described in Ref 3. Steven suggests the formula
AL = 0.1 9 ( T - 20~
for a correction in the range T > 0~

In Fig. 10 the temperature dependency of the tire/road noise for two tires is indicated [6].
Volkswagen recommends a correction as follows: standard temperature T = 20~ _ 2.5~
correction of the noise increase for a temperature drop in intervals of 5~ by 0.5 dB(A). This
gives an average temperature correction for car tires used on cars (third gear, ISO 362). For
trucks and truck tires, different values may occur. Temperature effects and tire selection will
be the new work item of ISO/TC43/SC1/WG27.
Experiences With a Heated "ISO" Road Surface

Since 1990 a heated "ISO" track has been in use at Volkswagen. There was a need for this
as weather conditions in Wolfsburg considerably impair testing due to the frequent high
humidity and the resulting damp road surfaces. A test track as per Fig. 11 was constructed.
The entire test stretch is 477 m long. Turning loops are located at both ends. The heated area
is 60 m long. The actual measuring area (20 m) and the bay for the microphone on one side
=t
k;
L~
=:
t~
<3
L)
I
O',
LE
k;
=t
t.
FIG. lib--Heated "ISO" track with optical speed detector and all-weather microphone.
(Source: Volkswagen.)
are located in the middle. Here the road surface is of asphalt concrete as per ISO/DIS 10844.
Drainage asphalt topping with a length of 40 m adjoin the heated area on each side. This
serves to reduce possible wetness on the tires before driving into the measuring stretch. There
are three moisture sensors and one temperature sensor on the road. When surface moisture is
present, the surface is heated to a maximum of 30~ Snow is melted and dried off around
the sensors. The control time constant has been selected to be sufficiently long to achieve
stable conditions.
A stationary measuring system is in operation around the clock. One large signal display
each stands 170 m to the left and right of the microphone position. Here the data of the
entrance speed of the vehicle and the dB(A) level are indicated to the test engineer (Fig. 12).
A list of the measuring results is printed out in the measuring station together with the
environmental influences (temperature, air pressure, and humidity). A mobile measuring system
is available for more in-depth frequency analyses.
By using this track, it was possible to increase the usable test time by more than 100%.
The track is used by the Passenger Car and Commercial Vehicle Development Division to
optimize prototypes, and by Quality Assurance to check conformity of production.
The following characteristics have become apparent after two years of operation.
In heated operation, sections at a higher level dry faster and thus become warmer. This
means that the temperature distribution at this point in time is inhomogeneous. If the entire
road surface is dried, the air temperature is decisive for the surface temperature.
Individual heat loops are meanwhile defective. Their repair is not possible without damaging
the "ISO" surface.
FIG. 12--Display for drive-in speed and noise level. (Source: Volkswagen.)
The tire rub from new tires during production conformity testing leads to a reduction of
the surface texture depth and thus to an increase of the tire/road noise. It is currently being
investigated whether this process can be reversed by cleaning the road surface.
Increase in Tire/Road Noise Due to Aging of " I S O " Tracks

Through measurements on the "ISO" road surface built most recently (1990) in comparison
to the older road surfaces, it should be clarified what influence aging has on the tire/road
noise and thus on the accelerated drive-by noise. In the course of the first operating year of
the heated test track, the aging effect was checked with a reference vehicle under consideration
of the temperature dependency. It was shown that a noise increase of approximately 1 dB(A)
(third gear) in the first year must be expected.
Tonhauser [6] indicates a value of > 2 dB(A) (coast-by) for the first 13-month period (Fig.
13). Here the initially different degree of compression of the road toppings may be a cause.
The aging effect should be checked during the first year after completion of the road surface
with a reference vehicle and reference tires. It is to be expected that this runs strongly digressive
with regard of time. More detailed investigations appear to be required here.
In our opinion, changes in the surface texture and the voids content due to settling or
compression during driving must not be feared. The oldest "ISO" tracks (construction year
1989) are in use in three-shift operation on the Ehra-Lessien proving grounds. The voids
content and the surface texture were checked after two years. Changes could not be proven.
Optically, as well, no difference can be seen between the wheel tracks and the adjoining area.
#,
.2
Additional Influences on Tire]Road Noise

In addition to the dominant influences on the tire/road noise given above, there are other
influencing variables which, however, only begin to play a greater role in the case of "quiet"
passenger cars [<73 dB(A)].
The vehicle influence affects the sound propagation through the design of the vehicle
underbody and in particular of the wheel housings. The same tires on different vehicles (of
the same weight) generally also cause different noise levels. A typical example is the vehicles
Rallye Golf and Golf GTI G 60. The different wheel housing design here causes a 2 dB(A)
lower tire/road noise for the Rallye Golf (Fig. 14). The wheelhousing forms a kind of reverberant
room. The volume of it and the size of the openings influence the noise level to be emitted.
Different reflection conditions result in a different noise level.
Steveu [3] found for one tire set on five different cars an influence of 2.5 dB(A). These
effects have to be studied more closely.
On passenger cars the influence of weight on the tire/road noise is usually negligible if the
car is noisier than 73 dB(A). The noise level always increases with increasing wheel load.
Within the scope of permissible loading, the effects on noise are mainly determined by the
altered ability of the vehicle to accelerate.
The same applies to passenger cars for the influence of the tire inflation pressure on tire/
road noise. The noise produced increases with increasing tire inflation pressure. Within the
scope of the permissible inflation pressure fluctuations, however, the effects are <1 dB(A).
Steven [3] confirms these findings.
Conclusion
The legislation to be expected in Europe makes it necessary to consider the growing
proportion of tire/road noise in type approval more closely.
The greatest noise differences are measured on different road surfacing. Here the need exists
to define a "standard road surface" for the regulation ISO 362. A proposal has been made
with ISO/D1S 10844.
The tire influence is of the second greatest importance. The goal conflict between traction
and noise must be solved by the tire manufacturers. In this case the noise level increase under
drive torque must be considered. Here also exists a need to standardize the tire type used for
vehicle type approval.
FIG. 14---Different styling of wheelhousings causes different tire~road noise.
The temperature dependency of tire~road noise must be taken into consideration during type
approval with a corrective formula.
The aging of the test track should be checked during the first year after the completion
of construction.
Vehicle influences, weight, and tire inflation pressure also have an effect on tire/road noise.
The influence on accelerated drive-by is, however, generally negligible.
If the effects named are not taken into account, even on a standardized "ISO" track the
drive-by noise value may show a spread up to 3 dB(A).
References
[1] Wolf,A., "Vergleichende Rollger~iuschuntersuchungenan PKW-Reifen auf unterschiedlichen Fahr-
bahnen," (Comparative Drive-by Noise Studies on Passenger Car Tires on Different Road Surfaces),
Report 16/87, Forschungsinstitut ftir Kraftfahrwesen und Fahrzeugmotoren FKFS, Stuttgart 1987/
Volkswagen, 1987.
[2] Betzl,W., "Einflul3 der Fahrbahnoberfl~icheauf das Ergebnis von Gerauschmessungen an Nutzfahr-
zeugen," (The Influence of the Road Surface on the Results of Noise Measurements on Commercial
Vehicles), Research Report No. FP 1.9008, TUV Bayern/Bundesanstalt fiir Stral3enwesen, March
1991.
[3] Steven, H., "Entwicklung eines Mel3verfahrens fur das Reifen-Fahrbahn-Gerausch yon PKW,"
(Development of a Measuring Method for Car Tire/Road Noise), Research Report No. 105 05 211,
FIGE, Aachen/Umweltbundesamt, July 1990.
[4] Hieronimus, K., "Die Absenkung der Aul3enger~iuschemissionenyon PKW---eine Herausforderung
fUrden Akustik-Entwicklungsingenieur,"(The Reduction of the External Noise Emissions of Passen-
ger Cars--a Challenge for the Developmental Acoustics Engineer), T-30-312-056-2, Haus der
Technik, Essen, March 1992.
[5] Enz, W. and Steven, H., "Developing Scientific and Technical Principles for EC Guidelines in the
Field of Noise Abatement," Research Report No. 105 05 993/01, FIGE, Aachen/Umweltbundesamt,
May 1991.
[6] Tonhauser, J., "Normfahrbahn ftir Aul~engerituschmessungen--AnwenderorientierteErkenntnisse,"
(Standard Road Surface for External Noise Measurements--User-Oriented Knowledge), VDI Reports
No. 916, October 1991.
Mark P. Voutyras 1 a n d Kevin R. Thomson 1
Road Noise Simulation Using Road

Surface Weighting Functions
REFERENCE: Voutyras, M. P. and Thomson, K. R., "Road Noise Simulation Using Road
Surface Weighting Functions," Vehicle-Road Interaction, ASTM STP 1225, B. 1", Kulakowski,
Ed., American Society for Testing and Materials, Philadelphia, 1994, pp. 204-222.
ABSTRACT: The purposes of this study were to investigate the effects of different road surfaces
on vehicle interior noise and to investigate a method of predicting road noise from multiple
surfaces based on noise measurements of a single surface. The primary focus is on coarse road
surfaces. However, both smooth asphalt and rough road surfaces are included in the study.
A road surface survey was conducted to determine the effect of different road surfaces on
vehicle interior noise. Weighting functions were developed for the purpose of modifying the
noise response from one road to match the response of other roads. The weighting functions
were experimentally derived from road noise measurements of a number of vehicles driven
over different surfaces. A second group of vehicles were then tested to determine if road
noise due to different road surfaces could be predicted accurately based on a single surface
noise measurement.
It is concluded that road surface weighting functions (RSWFs) are a useful tool for estimating
the vehicle interior noise due to different surfaces and investigating differences in road surfaces
and vehicle nonlinearity. The potential applications of the RSWFs include both laboratory
chassis rolls and computer model simulations. The implications to computer simulations are
that similar RSWFs may be used to predict the noise response from a control road profile.
KEY WORDS: road noise, laboratory simulation, weighting functions, coarse road surface,
dynamometer rolls surface
Through contact with test equipment suppliers and competitive automotive companies, it
is known that road noise simulations exist in many forms in the automotive industry today,
including simulations in the laboratory, via computer modeling or combinations of both.
Although the intent of this project is to benefit the laboratory road simulation, the findings
are applicable to computer simulations as well.
Improved repeatability, reduced time per test, and perhaps most importantly, the avoidance
of changing weather conditions are some of the benefits of the laboratory simulation. The
simulation also provides a more precise tool to diagnose problems, to assess development
changes, and to gain insight about vehicle system dynamics.
Most road noise laboratory simulations today use a chassis rolls or other forms of artificial
excitation, e.g., shakers. In any of the simulations, there are always compromises to be made.
Shakers typically compromise tire dynamics, airborne paths, input frequency content, and
directionality. The chassis rolls simulation accounts for tire dynamics and tire patch input
characteristics, but introduces other considerations such as roll harmonics and rolls surface
dependency.
Supervisor and product engineer, respectively, Body/Chassis Dynamics, Chrysler Corporation, 800
Chrysler Dr., Bldg. S1007, CIMS 481-47-10, Aurburn Hills, MI 48326-2757.
204
VOUTYRAS AND THOMSON ON ROAD NOISE SIMULATION 205
Rolls surfaces can range from generic inputs such as cleats, wires, or other unrealistic inputs
to surfaces cast from actual road surfaces. The generic-type inputs are often justified as a
diagnostic tool, but are almost always questioned with regard to creating noise representative
of actual road surfaces. The rolls surfaces cast from actual road surfaces are more representative,
but can still only represent the specific surface they are modeled after. However, evaluations
have shown that subjective road noise impressions of one surface versus another can be
quite different.
At the time of this work, there was no published information found on subjective impressions
of interior noise due to varying road surfaces. This study was initiated to investigate the effects
of different road surfaces on vehicle interior noise, and a method to minimize rolls surface
dependency by predicting road noise due to other surfaces. The primary focus was on coarse
road surfaces. However, both smooth asphalt and rough road surfaces were included in the study.
A road surface survey was conducted to determine the effect of different road surfaces on
vehicle interior noise. Weighting functions were developed for the purpose of modifying the
noise response from one road to match the response of other roads. The weighting functions
were experimentally derived from road noise measurements of a number of vehicles driven
over different surfaces. A second group of vehicles was then tested to determine if road
noise due to different road surfaces could be predicted accurately based on a single surface
noise measurement.
Objective
The objectives of this study were to investigate the effects of different road surfaces on
vehicle interior noise and to investigate a method of predicting road noise from multiple
surfaces based on vehicle noise measurements of a single surface.
Road Surface Survey
Road noise may be described in very different terms depending on the road surface. Evaluators
often use terms like "boomy," "roar," "hashy," and "pure tones" when describing road noise.
An understanding of what creates these different impressions was necessary before road
simulation development. For this purpose, a survey was conducted to investigate the effect of
different road surfaces on vehicle interior noise.
Survey Procedure
Seven different road surfaces in the Detroit area were chosen to represent a broad range of
road noise characteristics (Table 1). Surfaces 1 through 5 were classified as coarse road surfaces
creating varying degrees of "roar," "boom," and "pure tones." The remaining two roads were
a smooth asphalt surface creating what was perceived as more high-frequency noise and a
rough road generating mostly "boom" noise from impacts. Eight different vehicles (Table 2)
were driven over each surface at 45 mph while interior noise measurements were recorded.
Microphones positioned at the driver's left and right ear positions were recorded using a Sony
PCM digital audio recorder. Recordings were made when no traffic was present and with wind
speeds below 10 mph. The recordings were analyzed using a Hewlett Packard 3562 spectrum
analyzer. The measurements were A-weighted and analyzed in a 20 to 1000 Hz narrow band
format. The sum of the driver's left and right ear responses were used for the study.
a,-
.e,
~a o
,.m
~ ga ,.-1 Z
~a
0a
e~
e~
<
0
,...I
r~
~8
~z
~ ~ ~ ~
uJ
..1
~.~ .~o~.~ -~
~ ~ ~ ~-~ ~ ~ ,~ ~ ~
.~ ..~ .~ .~. ~ .~ .~ ~ .~ .~_
Survey Results
A sample of road noise spectra measured in Vehicles 1 and 3 over four of the surfaces are
provided (Fig. 1). Road Surface 7, a rough road, excited frequencies below 100 Hz the most,
while the roads described as generating "pure tones" had the highest response levels in the
100 to 300 Hz range. Surface 5, a concrete road with large worn stones, caused the highest
noise levels between 600 to 1000 Hz.
Each of the four different road surfaces appear to excite the same vehicle resonances. Very
little frequency shifting is observed. Only the relative magnitudes of the resonances are changed.
This indicates that each road surface provides enough broad band random input energy to
excite the primary vehicle resonances, and that the change in energy distribution is the basis
of different subjective impressions.
Road Survey Conclusions

The results lead to the following conclusions about rolls surface design and potential road
noise predictions:
1. Generic-type roll surfaces, e.g., random-spaced cleats or impact wires, may provide
adequate input to excite the primary vehicle resonances. However, because the input
energy distribution of the generic surface does not match any actual surface, the resulting
noise will not be representative. Therefore, the noise simulation may be adequate
for diagnosis of problem resonances and some comparative studies, but actual road
measurements will be necessary to determine the importance of specific resonances.
2. A roll surface modeled after an actual surface has the advantage of representing at least
one surface. In many cases, it may also be adequate to determine the importance of
specific resonances for diagnostic and development work. However, the question of
"How will it sound on a different road?" still cannot be answered.
3. The results indicate a potential answer to this question. The primary difference between
roads is the input energy distribution. In the case of the rough road, more energy was
input below 100 Hz, and so on. Although one roils surface cannot match all of the
different energy distributions of interest, it may be possible to mathematically adjust
one surface response to predict that of another.
Road Surface Simulation

Weighting functions were proposed as a means of estimating the vehicle road noise responses
to varying road surfaces. The basis and development of weighting functions are described below.
Theory
A transfer function is defined as the ratio of output or response of a system divided by the
input to the system. It is a linear system if that transfer function remains constant regardless
of changes in the input or other conditions. If we assume the vehicle's road noise transfer
functions are linear, the road noise due to one surface can be predicted from measurements
of another provided that the changes in input are known. Weighting functions may be used
to estimate the effects of changes in input. The road surface weighting function (RSWF) is
derived as follows.
Let the transfer functions describing one vehicle's road noise transmissibility characteristics
to Road Surfaces A and B be defined as
66
64
62
60
58
6 56
, :l/,i
54 L
52 i i u.
I ' 1 IVY/!
I~:~.~.. _=
50
[]
~8
r~! !
r .f
i
44
m
,u r ifi!:
~0
i
38
V " " ~';~% d..\
,,v / ,,,,
36
,/
3~
ZOO ~00 600 80( 1000
Frequency CKz)
--Rd #2 ..... #4 ..... 15 ----- #?
66
64
62
60 i A.
58
d 56
5t i\i r'l u
.~ sz
t ~1 tl i:
~ 50
ii
, bq,
N 4s
\ A"~
~.;~.
ii .~
Ptlv
m
'IJ '~2
4O I!' 'Co ~,. \ /

38
36 llt
34 !11
0 200 400 600 800 1000
Frequency (Hz)
--Rd #2 ..... #4 ..... #5 ----- #7
FIG. 1--Road surface survey (sample of four roads), driver's ear microphone narrow band
responses. Top plot: Vehicle 1; bottom plot: Vehicle 3.
OUT(w)A OUT(w)B
TF(W)A = ~ TF(w)B - - -
IN(w)A IN(w)a
where
TF(w) = transfer function,

OUT(w) = interior sound level response (measured by test),
IN(w) = input displacements from road surface, and
w = frequency.
If the vehicle dynamic characteristics are linear, then the transfer functions are equal and
a road surface weighting function, RSWFBA, may be defined to describe the relationship
between the Surfaces A and B,
IN(w)s OUT(w)B
RSVv'FBA(W) - -
IN(w)A OUT(W)A
This weighting function may now be used to predict another vehicle's response to road Surface
B from the vehicle's response measurement of Surface A. The resulting predicted response
of the second vehicle, designated by ('), is defined as
OUT(w)'B = RSWFBA(W) 9 OUT(w)'A
Road Surface Weighting Function (RSWF) Development

In theory, the RSWF may be developed to predict the response of any vehicle over any
surface. Initial measurements would be required to determine each RSWF as a function of a
control surface. One measurement over each surface would be required. Once the RSWF is
determined, only measurements over the control surface would be required to predict any new
vehicle's response over other road surfaces.
In practice, however, a vehicle's road noise transfer function may not be linear or may be
linear over a limited range of input. Furthermore, time variant noise sources, such as tire
impacts, and tire/surface sources, such as tread whine noise, may also be a basis for concern.
To explore this further experimentally, a survey was conducted to evaluate the limitations of
the RSWF.
Weighting functions were developed for six different surfaces (Surfaces 2 through 7, Table
1) using six different vehicles (Vehicles 1 through 6, Table 2). The road surfaces selected
ranged from smooth asphalt to multiple irregularly spaced tar strips. Vehicles with significantly
different suspensions, structures, and acoustic treatments, ranging from a convertible to a
minivan, were selected. Road Surfaces 2 through 7 were compared to Control Surface 1.
The RSWFs were calculated for each of the vehicles, 1 through 6 (Table 2). The resulting
RSWFs are shown (Figs. 2a through 2c). Each figure shows RSWFs for four different vehicles
and the average RSWF of all six vehicles. In theory, each curve should overlay. The differences
between the curves may be attributed to unknowns, such as tire contact differences, vehicle
nonlinearity, weather variation, measurement, and operator error.
A composite of the RSWF averages for each road surface is shown (Fig. 3). This comparison
illustrates the differences between roads very well. It shows that Surface 5 provides the most
high-frequency content, while Surface 7 produces the most "boom" content. The abrupt changes
VOUTYRAS AND T H O M S O N ON ROAD NOISE SIMULATION 211
3.5
3
u
tn
2.5
v
~. all.
11 I.I,
o_
t.5 !t !.!,
n i i
1 9V ~. ,-~ ~-~r - - ' I / ....
0.5
0
200 r 600 800 1000
Frequency (Hz)
--Veh tl ..... #2 ..... ~3 ----- 14 ........A v g ( 1 - 6 )
3.5
m 3
m 2.5
,z 1,5
II I
ii i
0.5
o Zoo 4o0 6oo 800 i000
Frequency (Hz)
- - V e h ~1 ...... #2 ..... 13 ----- #4 ........ A v 9 C 1 - 6 )
FIG.2a--RSWFs (microphone response of target surface)l(microphone response of base surface),
Vehicles I through 4, and average of all six vehicles. Top plot: Target Surface 2; bottom plot: Target
Surface 3.
4
3.5
t~
2.5
g~
-.-4
e-4
2
1.5
..A IV
1
~ *..2,~../~ , ~ I ~ t~ . ~-~:.-~
0.5
0
200 400 600 800 1000
Frequency (Hz)
- - V e h #t ..... #2 ..... #3 ----- #4 ........ Avg(l-6)
3.5
3
r"l
fJ
Ul
1.4
2.5
OI
2 n ,,~I,
1.5 , i'.=, ~ "~.

IL. ,0 a l ~
!
1 W/w JV"
0.5 I
0
0 ZOO 400 600 800 1000
Fzequency (llz)
.... Veh ~ml ..... #9. ..... e3 .__ #m~ ........ Av9"C1-6}
FIG. 2b---RSWFs (microphone response of target surface)/(microphone response of base surface),

Surface 5.
V O U T Y R A S A N D T H O M S O N ON R O A D NOISE SIMULATION 213
3.5
3
r-i
m
(.i
2.5
oJ
r-.i
1.5
r.,,-,
t~ 1
,.cl II l
0.5
0
0 200 ~OO 600 801 100o
Frequency (Hz)
Veh #1 ..... #2 ..... s3 ----- #4 ........ A v g ( l - 6 )
3.5
I"
I1".
oJ 3 lit.
r4 .111:
ro
~J
p%
2.5 i ul
I~,1
Q&
~4
2
t
,i
1.5
D~
0.5 v ~ v. to
0
0 200 400 600 800 1000
Frequency (Ez)
Veh #1 ..... #2 ..... #3 ----- #4 ........ A v g ( 1 - 6 ]
F I G . 2c--RSWFs (microphone response of target surface)/(microphone response of base surface),

Surface 7.
.4
3.5
9 3
u
m
2.5
L
@
t-"
2
A b.J~.
f-
-I 1.5 :_ ~/ V
U-
,, ~.,~ ./'~,
ki
l"'d ''~ ./ V
& 1
r-
.;
0.5
0
0 200 400 ;00 800 1000
Frequency CHz)
--Rd#2 ...... #3 ..... #4 .... #5
3.5
9 3
U
m
2.5
I_
0
C
2
:: :,:
U
r-
1.5
,,=
& 1 t'"
0.5
0
0 200 400 600 800 1000
Frequency CHz:)
RcJ#6 .... #7
FIG. 3---Averaged RSWFs of Vehicles I through 6. Top plot: Coarse Road Surfaces 2-5; bottom
plot: Smooth and Impact Surfaces 6 and 7.
VOU'I'YRAS AND THOMSON ON ROAD NOISE SIMULATION 215
in slope of the weighting function in the "boom" frequency range indicate the likelihood of
vehicle nonlinearity.
Predicted versus Actual Road Noise Levels
The average value RSWFs (Fig. 3) were used to predict the road surface noise responses
of Surfaces 2 through 7 for Vehicles 7 and 8 using only responses measured on Surface 1.
The two vehicles were not included in the average RSWF calculation. The resulting vehicle
response, predicted versus actual noise for each of the six surfaces, is shown (Figs. 4a through
43'). A very good correlation between predicted and actual road measurements exists for the
four coarse road noise surfaces (Figs. 4a through 4d). Predictions for smooth Surface 6 (Fig.
4e) and rough impact Surface 7 (Fig. 4~ are very good below 100 Hz, but less accurate above
100 Hz. On the smooth Surface 6 we suspect that the greater margin of error is due to the
tire tread contact becoming a more significant contributor to the interior noise. Error will
increase as the road surface becomes smoother. Tread noise becomes more dominant because
the input excitation no longer remains constant, but instead varies with each different tread
pattern/tire.
Road Surface 7 is an impact-type surface which produces much higher suspension forces
than any other surface. Each vehicle's nonlinearity due to load-dependent suspension bushing
stiffness may be the cause of error in predictions of Surface 7 responses. If the transmissibility
of the vehicle varies depending on the surface (load-dependent nonlinearity), error will be
introduced into the RSWE
Conclusions
This investigation has revealed the following interesting findings about the effects of different
road surfaces on interior road noise:
1. Most surfaces excite the primary resonances in each vehicle. This indicates that even
crude generic chassis rolls surfaces which excite broad band noise will be adequate for
diagnostic purposes, but actual road measurements would be required to determine
which resonances are most important.
2. Differences in the input energy distribution are the primary causes for different subjective
impressions of road noise over various surfaces, with the exception of very smooth
surfaces where tire tread noise may dominate.
This second observation led to the proposition that RSWFs could be used to predict road noise
due to various road surfaces based on a measurement over a single surface. The investigation into
RSWFs showed:
. The correlation between predicted and actual road noise was excellent for the different
coarse road surfaces (Surfaces 2 through 5).
2. Extremely different surfaces, such as smooth asphalt or impact "boom" surfaces, can
be predicted with limited accuracy. The resulting predictions for the impact "boom"
Surface 7 were excellent for the primary boom frequency range from 20 to 100 Hz. A
second control surface providing a more severe input may be necessary to improve
accuracy further.
It is concluded that the RSWFs are a useful tool for estimating the vehicle interior noise
due to different surfaces and investigating differences in road surfaces and vehicle nonlinearity.
66
64
62
50
58
6
E ss i ;I
54
L ~2 I All
.= ~o { 11 II~l 1~ ^:"
48 I illlU
I 44
42
40
It11
]V
38
u "'I
36 i: "' '.w.'"
l,,
34
0 200 400 601 800 1000
Frequency CHz~)
Pred I cted ..... Actua I
66
64
62
60
58
6 56 A
E 54
L
52
5O :V'
L 48
E~
O 46
I
44
<c
m 42 I A
~7 -~,R.
40
/ "U :',
38
36
34
0 200 4 00 600 800 1000
Frequency CHz~)
Predicted ..... Actual
FIG. 4a--Surface 2, predicted (base surface microphone response x RSWF of target surface)
versus target surface actual driver's ear microphone response. Top plot: Vehicle 7; bottom plot:
Vehicle 8.
66
64
62
60
58
56
E
54
L
52
U) 50
48
6 46 l!/ :t!:
!
44 ,,"
"0
42
40
38
36
34
200 400 60{ 800 1000
F r e q u e n c y (I-IZZ)
- - Pred|cted ..... Actual
06
64
62
60
58
U 56
54
L
9
o 52
o~ 50 -I' r"
L
r~ 48 V. I,/k
46
' 44 -v
oo
ID 42
4O
.~.--.u
38 v~
36
34
200 400 600 800 1000
Frequency CHz)
-- Predlcted ..... A c t u a l
FIG. 4b--Surface 3, predicted (base surface microphone response • RSWF of target surface)
Vehicle 8.
66
64
62
60
58
5 56
54
L
O)
52
.(. 50
48
o 46 !Ill ':
!
44
m
"O 42
40
38 A
36
34
0
l 200 400
Vl 600 800 1000
Frequency CHz)
Pred| ctecl ..... Actua I
66
64
62
{50
58
U
56
54 AI
L
13
@ 52 Itd ^
fl) 50 I~1
48
46
I
44
"0 42
40
38
36
34
0 200 400 600 800 1000
Frequency (Hz~)
- - Predicted ..... Actua I
FIG. 4c--Surface 4, predicted (base surface microphone response • RSWF of target surface)
Vehicle 8.
66
64
62
60
58
6 56
E
54
L
@ 52
m 50
48
46
!
44
m 42
"U
40
38
36
34
0 200 400 600 800 1000
F r e q u e n c y (l-lz)
Predlcted ..... Actuel
66
64
62
60
58
56 ; k
E 54
L
e 52
.~ 50
~5 48
~ ,46 A
' 44
VI . . . .
.I . .
m
"lJ 42 Lh "'"
vk
40
38
36
34
0 200 400 600 800 '1000
Frequency (FIz)
,,, P r e d ic t e d ..... A c t u a I
F I G , 4d--Surface 5, predicted (base surface microphone response • RSWF of target surface)

Vehicle 8.
66
6,4
62
6O
58
u 56
54
L
c 52
e
50
48
0 46
!
44 It, ~t
113 42 tin
40 t tfhf
Ilia/ m~
38
3G
fywv Vv tl \l~'"k,~.- +'- ^
34 f '++"~-,+i "V h+'..,k ~ .~v'~ I \
200 400 600 800 1000
Frequency CHzZ)
- - Predicted ..... Actual
66
64
62
60
58
6 56
54
L
52
50
48 ,i
O 46 JIJ-t ,
I
44 9 I~V.-.!I ~ I;\_
<
tl3
lO 4;2
40
38
36
34
~:\ t, v ~ v , V ....
0 200 400 600 800 1000
Fr equer~cy ~l-lz~)
FIG. 4e--Surface 6, predicted (base surface microphone response • RSWF of target surface)
Vehicle 8.
66
64
62
60
6
58
56 A
E
54 h
52
311 50 !, I ~, ",,
48 J'~d '
6 415
I
44 Lr~ ;,:'.~.'.
42 V___ y
9 vVl" "
40
38
:36
34
200 400 600 800 1000
F r e q u e n c y (I-Iz 3
-- I~edlcted ..... Actual
66
64
62
60
58
6 56
E 54
L
r 52
r; 'V.;
.tn
~5
~
50
48
45
7
V
' 44 v•
m
"o 42
40
3R
v~ ,-'~,~..41
36
r
3,4
0 200 400 600 800 1000
F r e q u e n c y (Hz:)
FIG. 4f---Surface 7, predicted (base surface microphone response X RSWF of target surface)
Vehicle 8,
Based on these findings, RSWFs may be established to predict vehicle interior noise over a
variety of surfaces using measurements from one surface, such as a chassis rolls surface. It
may also be possible to use the RSWFs to modify computer simulation responses to estimate
interior noise due to different road surfaces.
Acknowledgments
The authors wish to thank the following people for their support: Nils Omholt, for continual
project support, Allen Zwierzchowski for initial road surface investigation, and Ken Bazydlo
for technical support.
Fuel Efficiency and Rolling Resistance
Laszlo Gyenes 1 and C. G. B. Mitchell 1
The Effect of Vehicle-Road Interaction on

Fuel Consumption
REFERENCE: Gyenes, L. and Mitchell, C. G. B., "The Effect of Vehicle-Road Interaction
on Fuel Consumption," Vehicle-Road Interaction, ASTM STP 1225, B. T. Kulakowski, Ed.,
ABSTRACT: This is a review paper of experimental measurements of the effects of road and
tire characteristics on the rolling resistance and fuel consumption of cars, goods vehicles, and
buses. The rolling resistance of tires has been reduced by improved design and can be reduced
in use by higher inflation pressures and, for goods vehicles, by the use of wide single tires in
place of duals. For heavy vehicles, higher inflation pressures and wide single tires increase
road wear.
On paved roads, rolling resistance and fuel consumption is increased separately by the surface
macrotexture and the profile unevenness. These effects are additive and over the range of
conditions found on European roads can each increase car fuel consumption by about 5%.
Unevenness can increase goods vehicle fuel consumption by up to 10%. On unpaved roads in
developing countries, fuel consumption of goods vehicles and buses is increased by 15 to 20%
over that on paved roads. In addition, loose material on the road surface further increases
fuel consumption.
KEY WORDS: vehicles, roads, vehicle-road interaction, fuel consumption
Fuel is used by road vehicles to accelerate the vehicle, to raise its center of gravity when
driving uphill; to overcome rolling and aerodynamic resistance to motion at steady speed, to
overcome energy losses in the power plant and drivetrain; to operate auxiliary equipment such
as compressors, heaters, and air conditioners; and to overcome energy losses in suspensions.
Some of these effects, particularly the rolling resistance and the energy dissipation in the
suspension, are influenced by the interaction between the vehicle and the road. This interaction
depends in turn on the characteristics of the tires and the road surface, and possibly on the
type of suspension.
This paper reviews the current state of knowledge on the effects of vehicle characteristics,
road surface characteristics, and vehicle-road interaction on vehicle fuel consumption. The
authors' experience covers studies of vehicle fuel consumption but does not include a detailed
knowledge of tire technology. In the review, the opinions of various authors are quoted with
appropriate acknowledgments.
Vehicle Effects
Importance of Rolling Resistance for Cars and Trucks

Cars--Of the energy used by a car, about 72% is lost through thermodynamic heat rejection
by the power plant, 2% by mechanical losses in the transmission, and 8% to power auxiliaries.
Transport Research Laboratory, Crowthorne, Berkshire, United Kingdom.

225
Of the remaining 18% of the total energy that overcomes forces resisting motion, about 22%
(4% of total energy) is used to overcome aerodynamic drag, 33% (6% of total) to overcome
rolling resistance, and 44% (8% of total) for acceleration and braking [1]. Waters and Laker
[2] estimate that, for a car, a 20% reduction in tire rolling resistance would reduce fuel
consumption by about 3% for both urban and rural driving.
Trucks--For a 32-ton goods vehicle, rolling resistance contributes about 70% of total drag
at 50 km/h and about 37% at 100 km/h, the remaining drag being aerodynamic. On a representa-
tive route, Waters [3] has calculated the contributions to energy use for a 36-ton articulated
goods vehicle as
Engine losses 62%
Accessories 3%
Transmission losses 4%
Rolling resistance 15%

Aerodynamic drag 12%
Acceleration and braking 4%
Schuring and Redfield [4] measured the effect of rolling resistance on the fuel consumption
of a five-axle tractor-trailer. The results show a linear relationship between fuel consumption
and the total rolling resistance of all tires. To achieve a 1% reduction in fuel consumption for
the fully loaded truck, the rolling loss had to be reduced by 2.7% for crossply tires or 3.7%
for radials. For the empty truck, the corresponding figures were 5.0 and 6.6%. Knight [5]
obtained similar results from tests on two tractor-trailer combinations and showed good correla-
tion between values of tire rolling resistance obtained from road tests of fuel economy and
from laboratory tests. Thus vehicle/road interaction is responsible for a higher proportion of
the energy used by trucks than cars, and the scope for improving fuel consumption through
improvements to tires or road surfaces is correspondingly greater.
Cars--Effects of Tire Type, Size, and Pressure

The rolling resistance, or energy use, of tires has been reviewed by Walter and Conant [6],
Williams [7], Clark et al. [8], and Schuring and Futamura [9]. The development of tires with
lower rolling resistance, better wet grip, and longer life is continuous, and modern tires are
more efficient than those quoted in the first three references. Schuring and Futamura [9] have
reviewed this development, and Fig. 1 shows the reduction in rolling resistance coefficient
for car tires between 1975 and 1989. Not only has the mean resistance reduced, but the
variation between tires has become much smaller.
Williams [7] showed that radial car tires have a rolling resistance that at speeds of up to
about 100 km/h is about 25% less than crossply tires (Fig. 2). There have been many estimates
of the effect of inflation pressure. Figure 3 shows the results of a number of these. It is clear
that rolling resistance reduces as inflation pressure increases. Schuring and Futamura [9] have
reviewed the evidence of the effect on rolling resistance of the tire aspect ratio (ratio of cross-
section height to cross-section width). The results are complicated, and they conclude that,
because aspect ratio can be changed in several different ways, it is not possible to derive a
trend with aspect ratio alone.
Lippman et al. [10] showed, for a P185/80R-15 tire at 8 km/h, a linear increase in rolling
resistance with increased inflation pressure and load at constant deflection. Clark et al. [8]
quote a large number of studies that show a small increase in rolling resistance with increasing
speed up to a speed of about 100 kin/h, and a much larger increase with further increases in speed.
GYENES AND MITCHELL ON FUEL CONSUMPTION 227
1,0
1989
0.8
~ 0.6
E
.0
0.4
e 1981
0,.
/s~ X 9149 1975
0.2
..~ "..
e9 \ ;
\ sQ
f
~,. e9
! I I
0.005 0.010 0.015 0.020
Rollln 9 resistsnce coelficlent
FIG. 1--Rolling loss coefficients of three passenger car tire samples taken around 1975, 1981,
and 1989. The sample size was less than 200 tires. All tires were tested at 80% maximum load and
maximum pressure on a 67-in. (107-cm) roadwheel [9].
0.03
"8 I 0.02 m
Cross ply
Bias belted
0--
Radial
C
= 0.01
"6
n,-
o I I I
0 40 80 120 160
Speed (kin/hi
FIG. 2--Effect of tire construction on rolling resistance coefficient (car tires) [7].
0 .o4
0.02
.w
u
E
o=
r- 0.03
==
e-
o
\
rr Lorry
48km/h
0.01 I
0 I I I I I I
0 100 200 300 400 500 600 70(
Inflation pressure (kN/m 2)

FIG. 3--Range of results of the effect of inflation pressure on rolling resistance coefficient, speed
80 km/h (50 mile~h) [7].
Overall, Williams [7] concluded that changes in rolling resistance coefficient with load are
small. Tire construction can reduce the rolling resistance coefficient. Tire materials, tread,
carcass, and cord will change the rolling resistance. Purely to minimize rolling resistance
irrespective of other issues, a tire is required to operate at high inflation pressure, with radial
ply construction, low aspect ratio, natural rubber, minimum tread depth, zero slip angle,
and a relatively high working temperature. Some of these features would not be acceptable
in practice.
Trucks--Effects of Tire Type, Size, and Pressure

In addition to the references on rolling resistance given above, Ramshaw and Williams [11]
have made a major study of the rolling resistance of commercial vehicle tires. This considered
tires ranging in size from 7.00 by 16 to 18.00 by 19.5 and both crossply and radial ply
construction. The rolling resistance coefficient tended to reduce as the tire size increased (Fig.
4) and, like the car tires, was lower for radial ply than for crossply. It reduced with increasing
inflation pressure and, for crossply tires, increased with increasing load at constant pressure
(Fig. 4). It appeared that the rolling resistance coefficient depended on the static deflection
i Radial ply . . . . . Cross ply J
Inflation p r e s s u r e
(a) 7.00 x 16 tyres ( k N / m 2)
C 0.015 _ . - .400
IJ I
f ...-.- 5 0 0 ~
8
u f
f t
. . - - - 6 0 0 --
C
r
0.010
.400
I 500~
600~
O
tr
0.005 t I I
6 10 14 kN
Inflation pressure
(b) 10.00 x 20 tyres
E (kNIm 2)
0.015 w
~ ~ .--- .__...--.- --.:i00~ ~ ~

u
if_,
......,-'-" --- .._.,, ....-- 7 0 0 . ~ ~
0.010
,....-- . 8 3 0 ~ ---"
9 400
-- 500
0 60O
700
B30
0.005 I I I I I I I
16 20 24 28 kN
0.006 Inflation pressure

r
(C) 1 8 . 0 0 x 19.5 lyre (kN/m2l
550--
O
(J 650
cal
,- 750
850
cn
r
O
n"
0.005 l I
20 30 40 kN
Tvra load
FIG. 4--Variation of rolling resistance coefficient with load on various sizes of radial ply and
cross ply tires [11].
of the tire. If the inflation pressure and load are adjusted to keep the tire deflection constant,
then the rolling resistance coefficient is constant [11].
Ramshaw and Williams [11] quote tests of the fuel consumption of a five-axle, 32.5-ton,
articulated goods vehicle on a track at 48 km/h and on a motorway at 80 km/h. The two axles
of the trailer were fitted in turn with twin crossply tires, twin radial ply tires, and wide single
radial ply tires. On the track the dual radials gave a 2% and the wide singles a 6% fuel
saving over the dual crossply. On the motorway the corresponding savings were 4 and 10%.
Papagiannakis and Haas [12] quote anecdotal evidence that radial and wide single tires lead
to noticeably lower fuel consumption.
Figure 5 shows that there is a slight tendency for rolling resistance to increase with increasing
speed, but that this is a small effect [11].
The energy dissipated in the suspension of a vehicle manifests as an increase in rolling
resistance and, hence, fuel consumption [13]. If the energy dissipated in different types of
suspension is different, then on rough roads the apparent rolling resistance should vary with
the type of suspension. The authors are not aware of measurements of this effect having
been reported.
From a fuel economy point of view, there is pressure on the operators of heavy vehicles to
use wide single tires, to use high inflation pressures, and to use radial ply tires. The first two
factors will increase road wear [14], while the third will probably reduce it [12].
Road Effects
The rolling resistance of tires, and hence the fuel consumption of vehicles, is affected by
the characteristics of the road surface. The characteristics of the road that may be significant
are: whether it is paved, cobbled, or unpaved; its surface texture; its roughness or unevenness;
its softness or the depth of loose material if it is unpaved; and perhaps the material from
which the road is constructed. In addition, the overall vertical and horizontal alignment of the
road will affect the fuel consumption of vehicles using it (see, for example, Everall [15]).
Reducing gradients and curvature and allowing vehicles to operate at steady speeds will reduce
fuel consumption, but this effect is not considered in detail in this paper. Everall [15] has
shown that if averages are taken of fuel consumption over both up and down gradients, then
gradients of up to 3% do not significantly affect the fuel consumption of cars. However, even
small gradients affect the fuel consumption of commercial vehicles, and the effect increases
steadily with both gradient and load.
Paved Roads
A review by Hiersche [16] showed that, under free-flow traffic conditions, increasing the
unevenness of the road increases fuel consumption. There is considerable variation in the
magnitude of the effect. For a typical range of unevenness on highways in the United States,
Ross [17] found an increase in fuel consumption of 1.7% for three cars at 80 km/h on a rough
road relative to the consumption on a smooth road. In France, Bonnot and Boulet [18] reported
an effect of about 5%.
Laganier and Lucas [19] found, for a car on paved roads of the French highway system,
extra fuel consumption:
1. Ranging from 0 to 0.4 L/100 km (about 0 to 5%) for evenness rating from excellent
to poor.
2. Ranging from 0 to 0.4 L/100 km (about 0 to 5%) for macrotextures ranging from fine
to exceptionally coarse.
0 (
~
eb xil
,! ~o <~ ,.%
E ~
"D
\
I
I
i,r) O r5
o
o o o.
0 O o
luii!~!itlo:J ilOUlll!,li 6U!llOl:l
These effects are additive, and for neither effect does extra fuel consumption increase linearly
with unevenness or macrotexture. In practice, evenness and macrotexture are equally important,
but neither has as much effect on fuel consumption as road alignment, traffic control, and
driving habits.
Sandberg [20] found that, for a car on paved roads at 50 to 70 km/h, the fuel consumption
could be influenced by up to 12% by the ranges of macrotexture and unevenness found on
roads in Sweden. Macrotexture could cause 7 or 8% influence, but because macrotexture and
short wavelength unevenness are closely correlated, the effect of pure macrotexture is probably
about 5%.
Ramshaw and Williams [11] consider the effect of surface texture on the rolling resistance
of a 7.00 by 20 radial ply commercial vehicle tire (Fig. 6). This shows a difference of about
7% in the rolling resistance of a goods vehicle tire over lightly brushed concrete and BS 594
motorway type asphalt. This would imply an effect on fuel consumption of about 1 to 2%.
Walter and Conant [6] suggest a difference of up to 20% in the rolling resistance of a car tire
on concrete and asphalt and give the following rolling resistance coefficients for several
common surfaces:
Concrete 0.010-0.020
Asphalt 0.012-0.022
Dirt 0.025-0.037
Sand 0.060--0.150
Knoroz and Shelukhin [21] showed that there were large increases in rolling resistance on
a cobbled road compared to an asphaltic concrete over the whole speed range tested. An
increase in rolling resistance coefficient from 0.024 to 0.054 was measured at 40 km/h.
Visser and Curtayne [22] measured the separate effects of surface texture and roughness
on the rolling resistance of a passenger car. Schuring and Futamura's [9] presentation of their
results is shown in Fig. 7, where it can be seen that the rolling loss varies as a function of
roughness at two texture levels. The graph suggests that both texture and roughness have
equally important influences on rolling loss. Du Plessis et al. [23] extended this work to
medium trucks and buses at 80 km/h on paved and unpaved roads. They obtained the results
listed in Table 1.
Young [24] reports a number of experiments to measure the influence of road unevenness
on fuel consumption for cars and goods vehicles. Tests of a goods vehicle on the TRL track
at speeds of 6 to 13 krn/h used small wood humps to artificially increase the road unevenness.
Tests with an instrumented car at 65 km/h used rubber mats to increase unevenness. The
results are given in Table 2.
Two further series of measurements were made on public roads. In the first, two cars (Cars
2 and 3) were driven side by side along adjacent lanes which had different unevenness. In
the second, Car 2 was driven at steady speed along five lengths of road with different roughness.
The results of both of these series of measurements are shown in Fig. 8. Young combined all
these results in Fig. 9 to show the increase in fuel consumption of a goods vehicle and of the
cars due to road unevenness.
Unpaved Roads
The results already quoted from Walter and Conant [6] show the increase in rolling resistance
on dirt and sand roads as compared to metalled roads. INRETS [25] quotes fuel consumptions
for three classes of goods vehicle on paved and unpaved roads as given in Table 3. The fuel
consumption on unpaved roads is 15 to 20% greater than on paved roads.
oE~
t~
cE _o.
c5
L,,
o
"ION~
xm~ q~
~;O'm
g
-i ~5
, I
0 I
O
o. 8. ~5
0 O O
lu.~!:)!JJoo:~ ~ u u l s ! ~ J 6U!llO U
40
Te x t u r e ~
3O
_~ 2o
0
.J
10
I I I I I
0 20 40 60 80
Road roughness QI (Quarter-car Index) countslkm

FIG. 7--Relative rolling-loss increase over base loss (smooth road) as afunction of road roughness
(QI scale) and texture depth for an intermediate station wagon. Adapted from Visser and Curtayne
at 45~ tire temperature [9].
TABLE 1--The effect of road roughness on rolling resistance and fuel consumption for a 12-ton
truck and a 12-ton bus [23].
Roughness Quarter-Car Index QI, counts/km ~
20 (Good Paved) 80 (Poor Paved) 200 (Poor Unpaved)
Tire pressure, 540 kPa

Rolling resistance 0.0110 0.0129 0.0168
Truck fuel consumption,
L/100 km 29.7 31.6 35.3
Bus fuel consumption,
L/100 km 30.9 32.8 36.5
Tire pressure, 640 kPa

Rolling resistance 0.0092 0.0112 0.0150
Truck fuel consumption
L/100 km 2Z9 29.8 33.6
Bus fuel consumption
L/100 km 29.1 31.0 34.8
The quarter-car index is a measure of road roughness made with a single-wheeled trailer with a
suspension representative of a passenger car. The vertical acceleration of the trailer is measured and a
count made of the number of acceleration peaks per km greater than a specified threshold.
Abaynayaka et al. [26] give an estimate of the increase in fuel consumption of various
classes of vehicle in developing countries on gravel and earth roads relative to paved roads.
This lists the increase in terms of the road roughness and the depth of loose material on the
surface. Hide et al. [27] give similar information, shown in Figs. 10 and 11. Du Plessis et al.
[23] provide similar information from a number of developing countries.
TABLE 2--The effect of artificial roughness on the TRL track

on the fuel consumption of a goods vehicle and a car [24].
Average Fuel Average Fuel

Unevenness--3-m Consumption of Consumption of
Moving Average Profile the Goods Vehicle, the Car,
Variance, mm2 L/IO0 km IJlO0 km
0.21 21.7 ...

2.18 23.1 ...
3.74 24.0
2.21 . . .
8.10 ... 8.05
8
0
5.4 I80 percentconfidence 6.6
== limitsonmean
=
t-
O
5.2
(Z
E
:3
r
5.0 ~ roadsI 6.2
o
(b
:3
LL 6.0
t I ., I I I !
0 2 4 6 8
Unevenness varience (ram2|
FIG. 8--Public roads: examples of the effect of unevenness of fuel consumption [24].
Concluding Discussion
Tire-rolling resistance accounts for typically 30% of the energy used for acceleration, rolling
resistance, and air drag on a car, and 50% for a goods vehicle. To reduce the fuel consumption
of a car by 1%, the tire-rolling resistance needs to be reduced 6 to 7%. For a laden goods
vehicle, the corresponding figure is 2.5 to 3.5%, and, for an empty goods vehicle, 5 to 6%.
The rolling resistance of tires has been steadily reduced through improved design. Tire
materials, tread thickness, carcass, and cord all affect the rolling resistance. Purely to minimize
rolling resistance irrespective of other issues, a tire requires to be of radial ply construction
using natural rubber, to have minimum tread depth, and to operate at a high inflation pressure
and a relatively high working temperature. Some of these features may not be acceptable for
other reasons.
For trucks, the use of wide single tires instead of duals gives measurable fuel saving.
However, for heavy vehicles, the use of high inflation pressures and wide single tires will
increase road wear by a factor of two or more [12].
The rolling resistance and fuel consumption of a vehicle on paved roads are affected
separately by the macrotexture of the surface and the unevenness of the profile. These effects
appear to be additive. Over the range of conditions on European roads, macrotexture can
increase the fuel consumption of a car by about 5% or 0.4 L/100 km. The unevenness can
increase the fuel consumption of a car by up to about 5% or 0.4 L/100 krn and that of a goods
vehicle by up to about 10%.
FIG. 9--Averagedfuel-unevennessrelationships[24].
Relative to paved roads, rolling resistance and fuel consumption are increased on cobbled
and unpaved roads. Several measurements show that for buses and goods vehicles the fuel
consumption on unpaved roads in developing countries is increased over that on paved roads
by 15 to 20%. In addition, loose material on the road surface increases fuel consumption.
Twelve millimetres of loose material has been found to increase the consumption of a car by
1 L/100 km and of light and medium goods vehicles by about 2 L/100 Ion.
TABLE 3--Fuel consumption of goods vehicles in West

Africa [25].
Fuel Consumption L/lO0 km

Class of Goods Vehicle Paved Roads Unpaved Roads
Class I: capacity > 1 0 ton

Full 48 55
Empty 35 40
Class II: capacity 5-10 ton
Full 32 38
Empty 24 29
Class III: capacity < 5 ton
Full 24 29
Empty 18 22
35
30
25
(Bedford)
Lorry (all loads)
g,
O
O
Private car V
20
E
C 15
0
u
10
51~ (LanRover)
d
0
0 2 4 6 8 10 12 14 16 18
Roughness (ram/kin x 1000)
FIG. lO--Extra fuel consumed due to roughness on gravel roads [26].
35
Lorry (all loads)
(Bedford)
30
O
O
o 25
~
Light goods
( Land Rover)
C
.~ 2o
E
g t5
x
Ul
10
Private Car
(Cortlna)
or1. .f i I I I I I I I |
0 2 4 6 8 10 12 14 I6 18
Deplh of loose material (mini

FIG. 11--Extra fuel consumed due to loose material on gravel roads [26].
Acknowledgments
The work described in this paper is part of the program of the Transport Research Laboratory,
and the paper is published by permission of the Chief Executive. The research customer for
the work is Vehicle Standards and Engineering Division, Department of Transport.
The views expressed in this paper are not necessarily those of the Department of Transport.
Extracts from the text may be reproduced, except for commercial purposes, provided the
source is acknowledged. (Crown copyright 1992.)
References
[1] Martin, D. J. and Shock, R. A. W., "Energy Use and Energy Efficiency in UK Transport up to the
Year 2010," Energy Efficiency Series 10, Energy Efficiency Office, Department of Energy, HMSO,
London, 1989.
[2] Waters, M. H. L. and Laker, I. B., "Research on Fuel Conservation for Cars," TRRL Laboratory
Report 921, Transport and Road Research Laboratory, Crowthome, UK, 1980.
[3] Waters, M. H. L., "Research on Energy Conservation for Cars and Goods Vehicles," TRRL
Supplementary Report 591, Transport and Road Research Laboratory, Crowthorne, UK, 1980.
[4] Schuring, D. J. and Redfield, J. S., "Effect of Tire Rolling Loss on Fuel Consumption of Trucks,"
SAE Paper 821267, Society of Automotive Engineers, Warrendale, PA, 1982.
[5] Knight, R. E., "Correlation of Truck Tire Rolling Resistance as Derived from Fuel Economy and
Laboratory Tests," SAE Paper 821266, Society of Automotive Engineers, Warrendale, PA, 1982.
[6] Walter, J. D. and Conant, E S., "Energy Losses in Tires," tire Science and Technology, TSTCA,
Vol. 2, No. 4, 1974, pp. 235-260.
[7] Williams, T., "Power Consumption of Tires," TRRL Supplementary Report 192 UC, Transport and
Road Research Laboratory, Crowthome, UK, 1975.
[8] Clark, S. K., Dodge, R. N., Ganter, R. J., and Luchini, J. R., "Rolling Resistance of Pneumatic
Tires," U.S. Department of Transportation Report DOT-TSC-OST-74-33, PB 242 985, Office of
Systems Development and Technology, U.S. Department of Transportation, Washington, DC, 1975.
[9] Schuring, D. J. and Futamura, S., "Rolling Loss in Pneumatic Highway Tires in the Eighties,"
Rubber Chemistry and Technology, Vol. 63, 1990, pp. 315-367.
[10] Lippmann, S. A., Oblizajek, K. L., and Metters, J. J., SAE Paper 780258, Society of Automotive
[11] Ramshaw, J. and Williams, T., "The Rolling Resistance of Commercial Vehicle Tires," TRRL
Supplementary Report 701, Transport and Road Research Laboratory, Crowthorne, UK, 1981.
[12] Papagiannakis, A. T. and Haas, R. C. G., "Wide-Base Truck Tires: Industry Trends and State of
Knowledge of Their Impact on Pavements," Department of Civil Engineering, University of
Waterloo, Ontario, Canada, 1986.
[13] Lu, X. E and Segel, L., "Vehicle Energy Losses Associated with the Traversal of an Uneven Road,"
Proceedings, 9th IASVD Symposium on the Dynamics of Vehicles on Road Sand Tracks, Linkrping,
June 1985.
[14] Addis, R. R., "The Effect of Wheel Loads on Road Pavements," Proceedings, Seminar on Road
Wear: the Interaction Between Vehicle Suspensions and the Road, April 1991, London, Institution
of Mechanical Engineers, London, 1991.
[15] Everall, P. E, "The Effect of Road and Traffic Conditions on Fuel Consumption," TRRL Laboratory
Report 226, Transport and Road Research Laboratory, Crowthome, UK, 1968.
[16] Hiersche, P., "Summary of Results of a Literature Review on the Relationships between Pavement
Condition, Energy, and Time Requirement," Federal Highway Research Institute, Bergisch Glad-
bach, Germany, 1985.
[17] Ross, F. R. "Effect of Pavement Roughness on Vehicle Fuel Consumption," Transport Research
Record 846, Transportation Research Board, Washington, DC, 1983.
[18] Bonnot, J. and Boulet, M., "View of LCPC on the Results of the IRRE Concerning the APL
Longitudinal Profile Analyser," Proceedings, TRB 65th Annual Meeting, January 1986, Washington,
Transportation Research Board, Washington, DC, 1986.
[19] Laganier, R. and Lucas, J., "The Influence of Pavement Evenness and Macrotexture on Fuel
Consumption," Surface Characteristics of Roadways: International Research and Technologies,
ASTM STP 1031, W. E. Meyer and J. Reichert, Eds., American Society for Testing and Materials,
Philadelphia, 1990.
[20] Sandberg, U. S. I., "Road Macro- and Megatexture Influence on Fuel Consumption," Surface
Characteristics of Roadways: International Research and Technologies, ASTM STP 1031, W. E.
Meyer and J. Reichert, Eds., American Society for Testing and Materials, Philadelphia, 1990.
[21] Knoroz, V. I. and Shelukhin, A. S., "Test Data on the Rolling Resistance of Vehicle Tires on Roads
with a Hard Surface," Moscow Automotive Prom., Vehicle Research Institute, No. 2, 1964, pp. 16-21.
[22] Visser, A. T. and Curtayne, E C., "A Pilot Study on the Effect of Road Surface Properties on the
Fuel Consumption of a Passenger Car," NITRR Contract Report C/PAD/46.5, CSIR, National
Institute for Transport and Road Research, Pretoria, 1985.
[23] Du Plessis, H. W., Visser, A. T., and Curtayne, E C., "Fuel Consumption of Vehicles as Affected
by Road-Surface Characteristics," Surface Characteristics of Roadways: International Research
and Technologies, ASTM STP 1031, W. E. Meyer and J. Reichert, Eds., American Society for
Testing and Materials, Philadelphia, 1990.
[24] Young, J. C., "The Influence of Road Unevenness on Vehicle Fuel Consumption," Proceedings,
PTRC Transport and Planning Summer Annual Meeting, September 1988, Bath, PTRC Education
and Research Services Ltd., London, 1988.
[25] Inrets, "Politiques de rrduction des co/)ts du camionnage en afrique subsaharienne (Cameroun,
crte d'Ivoire, Mali)," Rapports de Recherche INRETS/LET, Institut National de Recherche sur les
Transports et leur Srcurit6, Arcueil, 1989.
[26] Abaynayaka, S. W., Hide, H., Morosiuk, G., and Robinson, R., ''Tables for Estimating Vehicle
Operating Cost on Rural Roads in Developing Countries," TRRL Laboratory Report 723, Transport
and Road Research Laboratory, Crowthorne, UK, 1976.
[27] Hide, H., Abaynayaka, S. W., Sayer, I., and Wyatt, R. J., "The Kenya Road Transport Cost Study:
Research on Vehicle Operating Costs," TRRL Laboratory Report 672, Transport and Road Research
Laboratory, Crowthorue, UK, 1975.
Y. D e l a n n e I
The Influence of Pavement Evenness and

Macrotexture on Fuel Consumption
REFERENCE: Delanne, Y., "The Influence of Pavement Evenness and Macrotexture on
Fuel Consumption," Vehicle-Road Interaction, ASTM STP 1225, B. T. Kulakowski, Ed., Ameri-
can Society for Testing and Materials, Philadelphia, 1994, pp. 240-247.
ABSTRACT: This paper is concerned with light vehicle fuel consumption on paved roads. It
starts with a recall of different factors affecting car consumption and then focuses on the effect
of road irregularities. Recall of results of practical tests carried out at the LCPC, on the one
hand, and more recently in collaboration with a tire manufacturer (Michelin-France), on the
other hand, are presented. Then, when considering effects of road irregularities on road qualities
such as: riding quality, braking efficiency, inside and outside noise (tire/road interaction), and
component vibrations, it appears that required specifications for low fuel Consumptionsurfaces
can be detrimental to other road qualities. Nevertheless, some new surfaces offer a very
good tradeoff.
KEY WORDS: fuel consumption of vehicles, influence of roughness and texture, road sur-
face characteristics
It is well known that the most efficient actions to reduce fuel consumption are the following:
1. Improvement of traffic flow and traffic congestion.

2. Preventing vehicles from entering city centers.
3. Changing driving habits.
The stake of these actions is at least 30% improved efficiency. But enforcing such a policy
seems difficult. The first action implies improvement of the road network (junction, crossing,
dimension). The second implies an important improvement of the public transport network.
These two actions must be considered as long-term actions. Regarding the third action, it is
very difficult to convince drivers that a change in their mode of driving would be cost efficient.
Research on rolling resistance and fuel consumption were carded out a few years ago at
the Laboratoire Central des Ponts et Chauss6es (LCPC). The work aimed at finding out what
could be gained by developing low fuel consumption roads. Regarding roads, it is well known
that the following factors affect fuel consumption:
1. Road alignment.
2. Road unevenness.
3. Road megatexture.
4. Road texture.
aHead, Roads/Vehicles Interaction and Road Accoustics Group, Laboratoire Central des Ponts et
Chauss6es, BP 19, 44340 Bouguenais, France.
24O
DELANNE ON PAVEMENT EVENNESS AND MACROTEXTURE 241
Effects on road alignment were not investigated in our research. Our experiments were
concerned mainly with the effects of short wavelength unevenness (0.7 < h < 2.8 m) and
texture (0.5 < k < 50 mm).
Experimental Means
Experimental Principles
The originality of the French approach, which is otherwise quite pragmatic, results from
the agreement of three lines of experimentation:
1. Laboratory simulation of roughness on a vibration bench. Such equipment can produce

vertical signals of any shape or simulate a road profile. However, a test of this kind
underestimates the effects of roughness by eliminating the changes in the drag coefficient
that occurs with changes of trim on an actual moving vehicle.
2. Road test track comparisons of surfaces having extreme macrotexture but the same
(good) degree of evenness.
3. Analysis of direct fuel consumption measurements on a sampling of actual roads having
various evenness and macrotexture levels.
Measurements of Unevenness and Texture
Unevenness--Unevenness was characterized by the weighted average of marks computed

from waveband bioctave filtered digitized profiles as explained in Fig. 1. The profile was
measured with our Road Profile Analyzer (APL) [1].
Macrotexture--The British Minitexture Meter was used to obtain the "standard mean texture
depth" (SMTD or Rq value), then converted into sand patch texture depth.
Conduct of Tests [2]
Hydraulic bench tests. The hydraulic bench (two hydraulic actuators) has a ram holding a
passenger-car wheel. The shock absorber temperature is recorded by an infrared camera. The
power dissipated (in watts) for a vertical loading of known frequency and amplitude is calculated
by analysis of the temperature-versus-time curve. The bench was used to reproduce both
sinusoidal loadings related to the conventional roughness rating and actual road profiles
recorded by the A P E Using a few simple assumptions, these energy losses are converted into
"computed excess consumption values."
Test-track measurements of rolling resistance. A dynamometer hub was installed on a
(nondriving) front wheel of a passenger car. This hub has four three-direction quartz transducers,
allowing simultaneous measurement of the forces and couples in all three axes. The device
has a very large measurement span and no significant drift in tests of short duration (less than
10 rain). The force measurement results (daN) are converted into rolling resistance (daN~ton),
then into a "computed excess consumption value." The reference surface strips of the Nantes
Test Track were used.
Fuel consumption measurements on roads. For the choice of road sections and experimental
precautions, the methodology used was similar to that used by automakers. All tests include
four valid sequences (the experimental constraints are such that less than 20% of sequences
~1 o0
l+
i ,
-=
c,I
~
! i o
II
!
L~
~a
, __+t
\
| i "o
.1.1
I,.I
=1
"---.tJ
+J+l+m!ll.tm t~
El 'F-,
9 w,,,I 0
DELANNE ON PAVEMENT EVENNESS AND MACROTEXTURE 243
are valid) on about ten different sections in addition to the test carried out at the Nantes Test
Track with a direct fuel-consumption measurement (in addition to the rolling-resistance tests).
Results
Extra fuel consumptionltexture. The extra fuel consumption/texture relationship was esti-
mated from computed extra vehicle fuel consumption and direct extra fuel consumption
measurement on the test track and very good paved road regarding unevenness. In this case
it was assumed that vehicle fuel consumption can be expressed in the form
C m = C n + S(r)
where
Cm -- measured consumption,
Cn = nominal fuel consumption of the vehicle, and
S(r) = extra fuel consumption induced by macrotexture.
Cn was estimated on a very flat and smooth road section on our test track.
Extra fuel consumption versus SMTD or sand patch texture depth is given in Fig. 2. The
higher the texture, the higher the extra fuel consumption. Extra fuel consumption due to
macrotexture can reach 5%. 2
Extra consumption versus unevenness. To estimate extra consumption versus unevenness it
was assumed that vehicle fuel consumption can be expressed in the form
Cm : Cn + S(r) + S(u)
Knowing Cn and S(r), it was possible to get an evaluation of S(u) from Cm measured on
various paved roads. These results are given in Fig. 3. Around 6% extra fuel consumption
can be explained by unevenness defects. Hydraulic bench experiments confirmed these results,
although it was not possible from our study to investigate the effects of different wave-
length bands.
This study was conducted in 1985. A new experiment was carded out in collaboration with
Michelin in 1986 on five trial strips used to establish the German-French procedure to measure
type/road noise.
The five trial strips were:
1. Surface dressing.
2. Bituminous concrete (0/14 mm).
3. Drainage asphalt (0/10 mm).
4. Sand bituminous concrete (0/2 mm).
5. Cement concrete.
The measurements were completed by measurements carried out on the Michelin test track.
Results given in Fig. 4 are consistent with results given in Fig. 2 and results obtained by other
researchers [4].
z Experiments were carried out with cars consuming 71/100 km/h.

4%
3%
2e
1%
l
Macrotexture
S.M.T.D value (mm)
0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2
Sand patch
I I I I value (ram)
0.5 1.0 1.5 2.0
Bituminous ooncrete ,,,J Surface dressing >
[1 '
Textured cement concrete
FIG. 2--Texture and extra fuel consumption.
6%
im 4%
e-
8 3%
9 ", -
1 2 3 4 5 6 7 8 9
~ - Very bad Excellent -ib
Unevenness: weighted averaged waveband marl=
FIG. 3--Unevenness and extra fuel consumption.
Conclusions Regarding our Results
From our experiment we can say that unevenness can influence fuel consumption up to
6%, and macrotexture can influence fuel consumption up to 5%.
We can conclude that minimizing extra fuel consumption required a smooth and even road
surface. The question is: Are the features compatible with other required road surface qualities?
D E L A N N E O N PAVEMENT E V E N N E S S A N D M A C R O T E X ' I U R E 245
E
I=
>
J=
=3
m
0
N
@
0
@
Oo
@ o~
I
0
0
0 0 0 0
v In
UOlldUJnsuoo IonJ ~JIX~l
.o
.Q
Z
,r-,,
E
0
o ~ ~ ~
eeSwluoo,oed O^liWinuJn ~
~E
l r--
.Q
0
J
1
,, ,'f
Ill ~ L~- ! s_
9 U
w,eol~omm| | | | ~
DELANNE ON PAVEMENT EVENNESS AND MACROTEX'TURE 247
Demand on Road Surface Characteristics

Table 1 gives the relationship between road irregularities and road user qualities. Considering
comfort, safety, noise, and user cost, it appears that the main problem is relative to macrotexture,
which is required for high-speed adherence on compact wearing courses and is responsible
for extra fuel consumption, inside and outside low frequency noise, and inside vibrations.
TABLE 1--Effects of road irregularities.
Road Irregularities Dynamic Comfort Safety Noise/Vibration Consumption
Long wavelength, Bad Bad Not directly Bad

octave 16-32 m related
Medium wavelength, Bad Bad Not directly Bad
octave 4-8 m related
Short wavelength, Very bad Very bad Bad Very bad
octave 1-2 m
Megatexture, octave Very bad Very bad Very bad Bad
0.625-0.5 m
Texture, octave Not related Good for Very bad Bad
0.0625 m high speed
Microtexture Not related Good for all Not directly Not directly
speeds related related
In France, two regulations enacted by the Road Directorate are concerned with unevenness
and texture. For unevenness, specifications are mainly defined in terms of statistical distribution
of an index more or less related to decametric wavelengths (Fig. 5).
To improve road qualities regarding unevenness, a working group composed of Road
Direction Technical Service servants and Road Private Companies representatives is currently
studying the introduction of short wavelengths specifications.
As for texture, specifications are defined regarding the kind of roads and demand for skid
resistance. A 0.8-mm sand patch texture depth is actually the minimum required, and the
average level is practically more than 1 mm.
Only open-graded bituminous eonerete (0/10 ram) and drainage asphalt friction eourses
meet this requirement in offering good quality as far as rolling resistance and tire road
noise are concerned [3].
Conclusion
Our study led to results in agreement with results obtained in other countries [6], but we
believe we still lack knowledge with regard to the influence of different unevenness wavelengths
(short, medium, long) on extra fuel consumption. This information would be very useful in
defining unevenness specifications for different wavelength bands for new roads and mainte-
nance works.
References
[1] Delanne, Y., "L'analyseur de profil en long num6rique," Revue G~ndrale des Routes et Adrodromes,
July/August 1992.
[2] Laganier,R. and Lucas, J., "The Influence of Pavement Evenness and Macrotexture on Fuel Consump-
tion," Surface Characteristics of Roadways: International Research and Technologies, ASTM STP
1031, American Society for Testing and Materials, Philadelphia, 1990, pp. 454-459.
[3] Delanne, Y., "Les enrob~s drainants: analyse de leurs proprirt~s vis ~t vis des bruits de roulement
et de l'adhrrence des pneumatiques des vrhicules de tourisme," Bulletin de liaison des Laboratoires
des Ponts et Chauss~es, No. 162, July/August 1989.
[4] Gyenes, L. and Mitchell, C. G. B., "The Effect of Vehicle/Road Interaction on Fuel Consumption,"
this publication.
P e t e r D. C e n e k I
Rolling Resistance Characteristics of New

Zealand Road Surfaces
REFERENCE: Cenek, E D., "Rolling Resistance Characteristics of New Zealand Road
Surfaces," Vehicle-Road Interaction, ASTM STP 1225, B. T. Kulakowski, Ed., American Society
for Testing and Materials, Philadelphia, 1994, pp. 248-262.
ABSTRACT: New Zealand's road network utilizes pavement types of widely differing proper-
ties, ranging from very smooth asphaltic concrete to coarse chipseal surfaces having macro-
texture levels not often found elsewhere in the world. An experimental program, therefore, was
undertaken to investigate the effect of surface texture on the rolling resistance of a typical tire/
car combination. A comparison of the coastdown and steady-state torque methods established
that the latter was more suitable for measuring vehicle drag forces on public roads. Accordingly,
it was used to obtain static and dynamic rolling resistance coefficients for twelve different road
surfaces having approximately the same roughness, but with macrotexture depths which varied
from 0.14 to 2 mm. The coarsest-textured surface investigated had a static component of rolling
resistance that was 55% greater than that of the smoothest. Regression analyses established that
the static component of rolling resistance is strongly correlated to surface profile wavelengths
between 2 and 50 mm. In addition, a limited investigation of rolling resistance changes due to
aging/traffic loading effects was conducted for a porous friction course surface.
KEY WORDS: rolling resistance, road surface texture, vehicle operating costs, road/tire interac-
tion, chipseal, coastdown, steady-state torque method
Nomenclature
A Vehicle frontal area, m 2

A, Transmission loss constant, N
A. Undriven wheel loss constant, N
Bt Transmission loss linear coefficient, N s / m
B. Undriven wheel loss linear coefficient, Ns/m
Co Static coefficient of rolling resistance
c. Speed-dependent coefficient of rolling resistance, where n = 1 is the linear coefficient,
s/m, and n = 2 is the quadratic coefficient, s2/m2
c~0 Aerodynamic drag coefficient at zero yaw
AC~ Yaw angle aerodynamic drag variation coefficient
C, Transmission loss squared term, Ns2/m 2
cx~) Coefficient of drag against yaw angle characteristic
CL(~,) Coefficient of lift against yaw angle characteristic
D Distance, m
Section Head, Aerodynamics, Works Consultancy Services Ltd., Central Laboratories, EO. Box 30-
845, Lower Huu, New Zealand.
248
Copyright 9 1994 by ASTM International www.astm.org
CENEK ON ROLLING RESISTANCE 249
F, Aerodynamic drag force, N

Fo(V) Total drag force function, N
FM Mechanical drag force, N
FR Rolling resistance force, N
Fr Tractive effort, N
g Gravitational acceleration, m/s 2
IRlqc International Roughness Index, quarter car simulation, m/km
K~ Temperature correction coefficient, K -1
K~ Coefficient of V2 in tractive effort, Ns2/m2
M Vehicle mass, kg
M~ Vehicle effective mass, kg
ri Residual error at point i, rrds
Tmc Sand circle texture depth, mm
TDMTM TRRL minitexture meter texture depth, mm
TMo Macrotexture, mm
To Observed temperature, K
T~ Standard ambient temperature, K
t Time, s
V Vehicle speed, rn/s
v, Total relative air speed, m/s
Road gradient, degrees
P Air density, kg/m 3
Yaw angle, degrees
The choice of road surface to be used in any given situation involves the consideration of
several factors: maintenance, environmental, economic, and road safety. Economic considera-
tions have chiefly focused on costs of laying and maintaining a surface and on the life span
of the surface. The cost of surface usage, i.e., the energy consumed by a vehicle in overcoming
the resistance of the road (the rolling resistance), is a further consideration that is frequently
overlooked. Previous studies have often concluded that the difference in fuel consumption over
the range of surfaces available is insignificant compared with overall fuel consumption [1,2].
New Zealand's road network utilizes pavement types of widely differing properties, ranging
from very smooth asphaltic concrete to coarse chipseal surfaces with macrotexture levels not
often found elsewhere in the world. Macrotexture has been identified as the most important
surface characteristic affecting rolling resistance [3]. It might, therefore, be expected that
pavement type may have a greater influence on fuel consumption in New Zealand than it
does elsewhere.
A research program was designed and undertaken to address the question of surface depen-
dence of fuel consumption for the surfaces in use on New Zealand roads. As the energy used
in overcoming rolling resistance comprises only a fraction of total energy usage, there will
be a smaller percentage change in fuel consumption for a given percentage change in rolling
resistance. The ratio between the change in roiling resistance and the resulting change in fuel
consumption is typically taken to be about 5:1 [3]. Consequently, rolling resistance is a far
more sensitive function of pavement type than is fuel consumption. Therefore, the research
program was concerned with the measurement of the former rather than the latter.
This paper describes the development of a test method for the on-road determination of
both static and dynamic components of rolling resistance and its application to a variety of
New Zealand road surfaces to establish the effects of road surfacing type and condition on
the rolling resistance of a representative tire/car combination.
Determination of Rolling Resistance Forces from Tractive Effort Measurements
Theoretical Background
The following derivation of a mathematical model for describing the total drag force which
acts on a vehicle has been taken from Ref 4 and modified in the context of instrumenting a
vehicle for use on public roads to gather rolling resistance data.
Tractive effort is taken to be the total forces resisting the straight line motion. Mathematically,
it can be expressed as
dV
Fr = Fo(V) + M, --~ + Mg sin a (1)
where
F r = tractive effort,
Fo(V) = drag force,
MaY
e dt = inertial force, and
Mg sin ot = gravitational force.
The main difficulty is associated with establishing the characteristics of the drag force
function. It consists of mechanical and aerodynamic terms, and it is assumed that their contribu-
tions can be considered separately as
Fo(V) = Fu + Fa (2)
Mechanical drag force includes tire rolling resistance, driveline losses, undriven wheel
losses, and suspension losses. Generally these effects are combined for ease of modeling.
Rolling resistance is a function of many operational variables, among them: tire type, tire
inflation pressure, tire cord temperature, applied torque, ambient air temperature, road surface,
and velocity [3]. Only the last four of these are of interest, as the other functions can be
held constant.
A common representation of rolling resistance for tractive effort modeling is
FR = Mg (Co + C.V") (3)
The speed-dependent contribution is generally assumed to be linear, n = 1 [4], or quadratic,

n = 2 [1,5].
Ambient temperature is an external effect imposed on the tires which causes a direct change
to the tire characteristic. Rolling resistance reduces as the ambient temperature increases. A
simple correction is therefore applied to give
FR = Mg {[Co + C~V~] [1 + Kr(T, - To)]} (4)
The temperature correction factor, Kr, takes a value of 0.013/~ [4].

Tire drag is a function of the normal load (Mg), which is reduced by the presence of
aerodynamic lift of the vehicle. Including this effect into the model, the normal load (Mg) in
Eq 4 is modified as
normal load = Mg - 1/2 O A Cd~) V 2, (5)
The lift characteristic CL(~) is determined from either model or full-scale wind tunnel tests.
Driveline losses are incurred in transmitting power from the engine to the axle shafts. These
losses are a function of vehicle speed and dfiveline oil temperature. They are modeled as
either a linear (Eq 6) or a quadratic (Eq 7) function of vehicle speed, although the quadratic
form is preferred [4]
For (V) = At + Bt V (6)
For (V) = A, + Bt V + Ct V 2 (7)
Undriven wheel losses (F) depend on bearing friction and brake drag. A linear variation
with speed is adequate, resulting in
F. (V) = A. + B. V (8)
Suspension damping losses are related to the absorption of the relative movements of the
wheels and body by the shock absorbers. This movement is primarily due to defects in the
longitudinal evenness of the road surface. To include suspension losses is difficult and leads
to increased complexity of the model [6]. Accordingly, a low roughness level (IRlqc < 2.5)
was a requirement of the test sections so that the contribution from suspension losses would
be minimized, roughness in this case being defined as surface profile wavelengths greater than
0.5 m.
Aerodynamic drag force for a road vehicle is dominated by the wake that develops behind
the vehicle [4] and is therefore primarily normal pressure drag. In practice, the magnitude of
the aerodynamic drag force is a function of the relative attitude of the vehicle to the airstream,
in particular yaw angle. The aerodynamic drag force is commonly written as
Fa = 1/~ p A C_,(~) V~ (9)
The characteristic of Cd(t~) is a minimum at zero degrees yaw, increasing to a maximum

somewhere between 25 and 40 ~ yaw. Like CL(~), it can be determined from either model or
full-scale wind tunnel tests.
Summing the mechanical and aerodynamic forces gives the mathematical model of the total
drag force on a vehicle as
Fo(V) = (M g - V2 p A CL(t~) VZ~)([Co + C.V"] [1 + Kr (To - T~)])

+ [A, + B,V + C,V 2] (lo)
+ [A. + B.Vl
+ 1/2 o A c~N,) v~
The unknowns in Eq 10 are the coefficients Co, C,, and Cd(~). The third term is modeled
as [Ca0 + ACa(O)] to indicate the variation of aerodynamic drag with yaw angle. If driveline
and undriven losses are not removed from the data, they will be incorporated in the derived
values of Co, C,, and Cd0. However, provided the same warmup procedure is adopted, these
losses will be constant from test to test and so are unimportant in road surface comparisons.
A minimum warmup period of 20 min is suggested for tires and the vehicle to reach normal
operating conditions [4].
From above, prerequisites for accurate tractive effort measurement include:
1. On-vehicle anemometry to measure aerodynamic yaw angle and total relative air speed.
2. Accurate aerodynamic drag and lift characteristics of the test vehicle.
3. Detailed grade information for the test sites.
4. A conditioning period of at least 20 min to allow the tires, bearings, etc, to reach
equilibrium conditions.
5. Accurate vehicle speed and wheel torque measurement.
Test Methods
Two test methods are widely used for determining the tractive effort coefficients Co, C,,
and Cd0: t h e velocity-time coastdown and the steady-state torque method. The coastdown
method is to allow the vehicle to coast down to a slower speed by disengaging the motor
from the wheels and to measure the subsequent position-time trajectory from which deceleration
and thus resistance is computed. The resistive force causing the deceleration includes driveline
losses. The steady-state torque method involves the vehicle being driven at steady speeds
between 20 and 100 km/h. At each steady speed, wheel torque along with total relative air
speed and aerodynamic yaw angle are continuously measured. The wheel torque is divided
by the dynamic tire radius to obtain the force, which is corrected for the ambient wind.
During the steady-state test, the wheel torque meter records the total resistance excluding the
driveline losses.
Prior to investigating the effect of New Zealand pavement types on tire-rolling resistance,
both test methods were used on two different test sections, a smooth-textured asphaltic concrete
surface (Site A) and a coarse-textured Grade 4 chipseal (Site B). The results were compared
in an attempt to establish the relative accuracies of the two methods.
Experimental Methodology
The instrumentation, test procedures, and data analysis are described in detail in Ref 7, so
only the principal features are presented here.
Instrumentation
The test vehicle was a 1985 Nissan Pulsar fitted with 175/70, SR13 steel-belted radial tires.
It was equipped with a data acquisition system which simultaneously measured torque at the
left-hand driven wheel, body and wheel strut accelerations, vehicle velocity, and relative wind
speed and direction.
Data Acquisition
All measured variables, with the exception of wheel strut accelerations, were low pass
filtered at 1 Hz to suppress high frequency noise and sampled at a frequency of at least 4 Hz
via an AXVIIC twelve-bit analogue/digital converter attached to a PDP 11/23 + microcomputer.
The wheel strut accelerations were sampled at 200 Hz and bandpass filtered between 0.2 and
100 Hz, enabling power spectra of the wheel response to be generated.
General Test Method
The general test practices required for tractive effort measurement applied to both steady
state and coastdown methods and involved:
1. Measurement of ambient temperature and pressure at the site.

2. The vehicle being weighed before and after every test session and the mass during
testing taken as a linear function of distance traveled.
3. Close monitoring of tire pressure, which was set before testing when the tires were
cold and maintained constant during testing.
4. A minimum warmup period of 20 min prior to testing.
5. The acquiring of a calibration file at intervals during testing to check the zeros of
the transducers.
Data Analysis
Coastdown
As outlined earlier, the equation of motion for a vehicle during coastdown may be written as
M"--d-[
dV = - M g (Co + C,W) - 1/2 p A(Cao + ACd(d~)) V~ -- M g sinct (11)
A nonlinear least squares technique was used to fit velocity data to Eq 11. The algorithm,
which is presented in detail in Ref 8, minimizes the sum of squares of the residual function
F (C)
F(C)= 2 [r,(C)12=~ [V~-V(t~,C)]2 (12)

i=1 i=l
where C is the vector of parameters, i.e., C = (Co, C,, C~0, V0).

Here V~is the measured velocity at time tl, and V(ti, C) is the model velocity at ti obtained
from integration of Eq 11.
Prior to the commencement of the experimental work, preliminary studies using rounded
data from the analytic solution of Eq 11, with n = 1, were carded out. These indicated that,
with a maximum velocity error of 0.1 ms -~, errors in Co and Cd0 would not exceed 5%. The
corresponding error in C~ was as large as 20%, indicating that Cl would be less well determined
than the other two parameters.
Steady-State Torque Method
Separating the tractive effort equation into known and unknown terms results in
Fr - M g sin a - M d V _ V2 p A AC~(~) V~ (14)

dt
The rolling resistance term, FR, takes the general form of Eq 3. To obtain average values,
Eq 14 is integrated with respect to distance using the transformation dx = V dt and divided
by the total distance traveled, D. The distance averaged form of Eq 14 therefore becomes
Y = Mg Co + Cdo X + Mg C, VN (15)
where
Y = 1/D [
o[ Fr - Mg sinet - M,--~ - l~ p A ACd(~) V~
] V dt (16)
X= 1/a p A 1/D f V dt (17)

0
D
VN = lID
f 0
V "+l dt (18)
All quantities in Y can be readily determined, as can X and VN. Therefore, regressing Y
against X and VN yields the required drag coefficients Co, C,, and Cd0.
Summary of Preliminary Test Findings
Coastdown
A total of twelve coastdowns were performed, six at Site A and six at Site B. The experimental
data were analyzed:
C a s e / - - B y allowing all three coefficients (Co, (7,, Cd0) to vary (the three-parameter model).
Case 2 - - B y allowing Co and Cdoonly to vary, keeping C, constant (the two-parameter model).
Case 1 lead to unrealistic values of the parameters. Computed values of the variance of
error for each of the two cases were similar, but the confidence limits on the parameters Co
and Cd0 in the three-parameter model were ten times larger than those in the two-parameter
model. All datasets were therefore analyzed with the dynamic roiling resistance coefficient
arbitrarily set to C~ = 0.0002 s/m for the linear case and C2 ~--- 0.000 018 sZ/m2 for the quadratic
case. For the vehicle used in the tests, a +50% variation in CL led to a consistent variation
in each of Co and Cd0 of - 4 . 5 % . A corresponding variation in C2 led to a variation in Cd0 of
- 9 % with Co unaffected. It is expected that an estimate of C~ or C2 with an uncertainty of
less than 50% can be obtained from a laboratory test on a single tire/drum combination.
TABLE 1--Mean parameter values with 95% confidence,

two-parameter coastdown model
n = 1, Cl = 0.0002 s/m n = 2, (72 = 0.000 018 s2/rn~
Site A Co = 0.0136 +--0.0009 Co = 0.0158 +--0.0007

C~ = 0.485 _+ 0.019 Cd0 = 0.424 + 0.020
Site B Co = 0.0184 --- 0,0008 Co = 0.0204 -+ 0.0008
C~ = 0.522 + 0.010 C~ = 0A37 + 0.018
With reference to Table 1, which summarizes the results of the analysis, both linear and
quadratic rolling resistance models show a significant difference in the rolling resistance
between the two surfaces, with the smooth-textured Site A having a static rolling resistance
coefficient that was about 24% less than the coarse-textured Site B.
Steady-State Torque Method
The instrumented vehicle was driven at near constant speed along the test road sections.
Measurements were made at five speeds equally spaced between 20 and 100 km/h and were
made in both directions. Both the forward and the return runs were made in the same lane
for Site A. Runs were repeated for at least five sets of speeds in either direction, so at least
ten measurements at each speed were made at each site. In practice, rather than determining
the gradient term M g sina in Y, a dummy variable taking the value of 1 for forward runs and
- 1 for reverse runs was added to the regression model. This simply and adequately compensated
for gradient effects.
The results presented in Table 2 show that the coefficient of the static component of rolling
resistance is well determined with standard deviations at least half those obtained with the
coastdown method. The driving torque data also provided evidence for a quadratic rolling
resistance model. The consequence of taking n = 2 is that the two independent variables in
Eq 15 are highly correlated. If the actual wind speed relative to the ground is zero, then X
and V2 are identical, and the regression of Eq 15 will be unable to distinguish between the
coefficients C~ and (72. Therefore, to ensure that the regression was yielding consistent results
unaffected by colinearity of the independent variables, final coefficient values were calculated
from regressions on the variables, X - V2 and X + V2. Identical results were obtained to
those derived from regressions directly on X and V2.
From Table 2, it is clear that a successful distinction can be drawn between C2 and Cd0.
Nevertheless, their standard deviations are relatively large when compared with Co, and hence
the significance of the values obtained is reduced. Furthermore, the lesser the ambient wind
speed, the greater the resulting colinearity between X and V2, and so the larger the standard
deviations of the regression-derived coefficients Cz and Cd0.
A more useful measure for the purpose of comparison between road surfaces might be the
combined parameter K~ defined by Eq 19. This represents the variation with speed of the
driving force required to maintain the vehicle at a constant speed
TABLE 2--Mean parameter values derived from steady-state wheel torque data.
Site Co C2 (sZ/m2) Cdo Kv (Ns2/m2) R2
A 0.0124 -+ 0.0003 0,000 014 _+ 0.000 004 0.59 -+ 0.04 0.83 -+ 0.09 0.993
B 0.0151 _+0.0005 0.000019 --- 0.000004 0.51 -+ 0.04 0.76 -+ 0.08 0.995
Kv = 1/2p A Cdo + MgC2 (19)
Because Cd0is invariant, Kv should be a function of road surface only. The use of Kv to compare
surfaces was found to be especially valuable when the steady-state torque measurements were
made in near to calm conditions, so that the colinearity of V2 and X made independent
evaluation of Cd0 and C2 difficult.
Comparison of Results
The steady-state torque method resulted in lower coefficients of the static component of
rolling resistance than the coastdown method. This difference is attributed primarily to the
characteristics of the test methods. During a coastdown, the drag forces acting on the test
vehicle, and implicitly contributing to Co, include brake drag, wheel-beating drag, and driveline
drag. Of these, only the first two contribute to the roiling resistance term calculated by the
driving torque method, which should therefore be lower. Reference 4 shows that comparable
coefficients can be obtained if driveline losses are eliminated from the coastdown force/velocity
data. This is achieved by acquiring wheel torque data during the coastdown. It is, however,
reassuring that the observed trends between road surface types, the main object of this investiga-
tion, are consistent irrespective of method. Furthermore, the derived Cd0 values also show
good agreement.
Measurement of Rolling Resistance Characteristics of New Zealand Roads

The steady-state torque method was used to determine the static and dynamic rolling
resistance coefficients for twelve test sites covering the range of New Zealand sealed road
surfaces. The steady-state torque method was selected in preference to the coastdown
method because:
1. The statistical accuracy of the static coefficient of rolling resistance was higher for a
similar number of test runs.
2. It provided reliable information on the speed variance of rolling resistance.
3. The required test speed range of 20 to 100 km/h fell within legal speed limits for
public roads.
A survey of available literature [9] indicated that macrotexture and roughness of the road
surface exert a considerable effect on the roiling resistance of vehicles. Measurements were
therefore carried out in the wheelpaths to quantify the test sites in terms of these two parameters.
Macrotexture of the road refers to the individual pieces of aggregate in the surface. For this
exercise, macrotexture measurements were made using the sand circle method and two laser-
based devices: the TRRL mini texture meter and the stationary laser profilometer developed
by the Swedish Road and Traffic Institute (VTI). The relationships between the various
measures are as follows
Tosc = 0.34 + 2.16 TMa (20)
Tosc = - 0 . 2 0 + 2.27 ToMrM (21)
Because it is art internationally recognized measure, all macrotexture measurements were

converted to sand circle texture depth using Eqs 20 and 21.
Road roughness of the test surfaces were measured using the NAASRA roughness meter,
which is a response-type mechanical device fitted to a passenger vehicle to register cumulative
relative vertical displacement of the vehicle body in one direction [10]. N A A S R A roughness
measures can be converted to the quarter car International Roughness Index
NAASRA (counts/km) = 26.2 IRIqc (rn]km) (22)
The characteristics of the test sites and the associated steady-state torque results are summa-
rized in Tables 3 and 4, respectively. With reference to Table 3, it will be noted that for Site
4, two different surfaces were measured, as during the test period the surface was relaid. The
opportunity was also taken to remeasure the rolling resistance characteristics of the newly laid
surface after being exposed for 30 months to an average daily traffic count of approximately
12 600 vehicles.
The steady-state torque measurements were made over two separate periods: during February
and March 1988 (reported in Ref 7) and May 1991 (reported in Ref 11). The latter measurements
concentrated on smooth-textured roads.
D i s c u s s i o n o f Results
Tl~e Static Coefficient of Rolling Resistance, Co

Regression modeling was used to analyze the data obtained from the experimental program.
The model providing the most satisfactory fit in terms of simplicity and accuracy is
Co = 0.0122 + 8.24 • 10 -4 T2o,c (23)
This equation has an R-squared value of 70.7% and a standard error of 0.001. Inclusion of
road roughness improved the fit slightly, although for this data its effect is minimal as can be
seen by
Co = 0.0102 + 8.35 • 10 -4 T2,c + 1.05 • 10 -3 IRlqc (24)

R 2 = 72.5% SE = 0.001
With regard to the different wavelength measures of texture obtained with the VTI stationary
laser profilometer, Co appears least correlated to fine texture (2 to 6-mm wavelengths) corres-
ponding to very small chippings and the sharpness of the chippings.
The driving torque data obtained by Central Laboratories was compared with the results of
an experiment by VTI in which a Volvo 242 car, instrumented to accurately measure fuel
consumption, was run over 20 road surfaces with various textures [12]. The following procedure
was used to convert our static rolling resistance coefficient values to equivalent fuel consumption
at 70 km/h
1. The asphaltic concrete surface of Site 11 was selected as the reference surface, and the
excess rolling resistance for the other test surfaces were calculated on a percentage basis.
2. The excess rolling resistance was divided by five to obtain fuel consumption.
3. Quite arbitrarily, a value of 7.35 l/lO0 km fuel consumption was assigned to Site 11.
Fuel consumption of the other sites was calculated by applying the excess values derived
in Step 2 above.
g==~ggggggggg~
0 oo ~
r162162
ga OOE.~. 0000000000 ~ 0 0 ~
,.4
0 0 0 0 00~
o
~ ~ 0 0 0 8 8 0 0o o
o o o o.o o o o o o o
ii
,..-1
g
0
0 ~ -o . . . . . . . d
. . . . .~ ~
. . . . . d d ~ o d d d -
" " " d . . . . . . . d
9 "O . . . . . . . O
,g
r,.)
,...1
J ~1~
d
Z
+1 § 2 4 7
§247247247247247247247247
+l+l+l§247247247247247
.~. x
l..r.l
F
0
+1+1+1+1+1§247
x
. . ~ . . . . . . ~
-a "~''8 8-''8 o88
~ ~ ' ~ ~ ~ ~'~"
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions auth
The "macrotexture" values for the New Zealand test sites were converted to logarithmic
values by using the relation
Tua (dB) = 20 log10{0.001 [(Tosc - 0.34)/2.16]} (24)
As can be seen from Fig. 1, the New Zealand values fell well within the range of VTI
measurements, the linear regression lines demonstrating virtually complete coincidence. This
substantiates the finding that tire rolling resistance, and hence vehicle fuel consumption,
increases with increasing macrotexture, as both sets of results were obtained independently.
The Dynamic Coefficient of Rolling Resistance, C2
The standard errors were such that it was not possible to identify significant differences
between either the C2 or K~ values obtained for each surface. The accuracy of the measurements,
however, would have been improved if either more runs were performed at each test site or
if the tests were conducted in windier conditions. Regression analyses showed lack of correlation
between C2 and sand circle texture depth, To= (R 2 = 9.7%), and also IRIroughness (Rz = 15.2%).
Age/Traffic Dependency of Rolling Resistance
A difference of 7% was found for the friction course surface of Site 4 between the measure-
ments performed in 1988 (Co = 0.0152) and 1991 (Co = 0.0141). The standard error in
determination of both Co values was about 3%, and so the difference is significant, suggesting
a traffic loading dependency on roiling resistance.
8.50
o o
8.00 4 9 .......
I
I o o
--o o
~ -- 0 0
m
o
c.)
.~ "/.00 oV'n
9 Central Laboratories
6.50 i t , i | l t , t I t i t i | i t , , I s , ~ ,
45 50 55 60 65 70
Macrotexture (dB)
FIG. 1--Linear regression o/fuel consumption on road surface macrotexture, Swedish and New
Zealand data.
Conclusions
1. An on-road test procedure, utilizing wheel torque measurement and on-board anemometry,
has been developed that can reliably determine the rolling resistance characteristics of road sur-
faces.
2. The static component of rolling resistance was found to be significantly affected by
surface texture depth as determined from the sand circle method, with a measured increase
of 55%, resulting from a 2-mm increase in texture depth.
3. Experimental data from this program indicated that, over a 20 to 100 km/h speed range,
rolling resistance is weakly dependent on vehicle velocity squared. However, more definitive
measurements are required to establish whether this dynamic component of rolling resistance
is independent of both surface texture and roughness.
4. Preliminary data have been collected to demonstrate age/traffic loading dependency on
rolling resistance. However, additional investigations are required to establish the relationship
between tire rolling resistance and traffic loading for particular road surfacing types.
Acknowledgment
The author is grateful to Transit New Zealand, the funders of the research program, for
permission to quote results of the work undertaken, and to the ASTM referees for their
helpful comments.
References
[1] Bester, C. J., "Effect of Pavement Type and Condition on the Fuel Consumption of Vehicles,"
Transportation Research Record 1000, Transportation Research Board, Washington, DC, 1984.
[2] du Plessis, H. W. and York-Hart, M. A., "The Effects of Road Surface Properties on the Fuel
Consumption of Trucks and Buses," Research Report 487; National Institute for Transport and
Road Research, CSIR, Pretoria, South Africa.
[3] Lees, G., "Cost and Energy Considerations in Relation to Pavement Surface Characteristics,"
PIARC Technical Committee Report on Surface Characteristics, Proceedings, 17th World Road
Congress, Sydney, Australia, Permanent International Association of Road Congresses, Paris,
France, 1983.
[4] Passmore,M. A. and Jenkins, E. G., "A Comparison of the Coastdown and Steady State Torque
Methods of Estimating VehicleDrag Forces," SAE Paper 880475, Society of Automotive Engineers,
Warrendale, PA, 1988.
[5] Ljubic, D. A., "Analysis of Productivity and Cost of Forestry Transportation," Technical Report
No. TR53, Forest Engineering Research Institute of Canada, Vancouver, Canada, 1982.
[6] Korst,H. H. and Funfsinn, M. A., "Determination of Effective Rolling Resistance by Coastdown
Experiments on Smooth and Rough Roads," Tire Rolling Losses and Fuel Economy, SAE Conference
Proceedings P-74, R. H. Snyder, Ed., Society of Automotive Engineers, Warrendale, PA, 1977,
pp. 133-141.
[7] Cenek, P. D. and Shaw, P. F., "Investigation of New Zealand Tyre/Road Interactions," RRU Bulletin
85, Transit New Zealand, Wellington, New Zealand, 1990.
[8] Swift,A., "Calculation of Vehicle Aerodynamic Drag Coefficients from Velocity Fitting of Coast-
down Data," Journal ofWindEngineering andlndustrialAerodynamics, Vol. 37, 1991, pp. 167-185.
[9] Shaw,P. E and Cenek, P. D., "Review of Vehicle Drag Losses," Report 9-86/1, Central Laboratories,
Ministry of Works and Development, Lower Hutt, New Zealand, 1986.
[10] Prem, H., "NAASRA Roughness Meter Calibration Via the Road-Profile-Based International
Roughness Index (IRI)," Australian Road Research Board, Research Report ARR 164, Vermont
South, Victoria, Australia, 1989.
[11] Cenek, P. D., "Rolling Resistance Measurements for Smooth Textured Road Surfaces," Report 91-
29306, Central Laboratories, Works Consultancy Services Ltd., Lower Hutt, New Zealand, 1991.
[12] Sandberg, U. S. I., "Road Macro and Megatexture Influence on Fuel Consumption," Surface
Characteristics of Roadways: International Research and Technologies, ASTM STP 1031, W. E.
Meyer and J. Reichert, Eds., American Society for Testing and Materials, Philadelphia, 1990,
pp. 460--479.
Author Index
A K
Alstead, C. J., 173 Karamihas, S. M., 52

Kelly, W. J., 135
B Krarup, J., 89
Kulakowski, B. T., Editor
Batstra, G., 7
M
C
Mitchell, C. G. B., 225
Cebon, D., 27
Cenek, P. D., 248 N
Cole, D. J., 27
Navin, F., 119
D
S
Delanne, Y., 240
Spangler, E. B., 135
G
T
Gelling, I. R., 107
Gillespie, T. D., 52
Gilmore, B., 150 Thomson, K.R., 204
Gyenes, L., 225
V
H
Verheul, C. H., 7
Halonen, P., 36 Voutyras, M. P., 204
Hugo, E, 67 Vesimaki, M., 36
Huhtala, M., 36
W
J
Wagner, P., 185
Jansen, S. T. H., 7 Whitehead, J. P., 173
263
Copyright* 1994 by ASTM lntcrnational Www.astIII.org
Subject Index
A Development and analysis of simulation
tools, 7
AAASHO. American Association of State Development programs, road vehicles. 173
Highway and Transportation Drag forces, measurement, 248
Officials. Drive-by noise, 185
AASHO present serviceability index, 135 Drive torque influence, 185
AASHO road test, 52, 135 Driving habits, effect on fuel economy, 240
Accelerated pavement testing, 67 Dynamic behavior
Aging of ISO tracks, 185 heavy trucks, 36
Analysis of simulation tools, 7 mobile load simulator, 67
Articulated vehicle, 27 rider/seat interaction, 150
Axle loads, 36 steering and suspension, 173
Dynamic loads, computer simulation of road-
B vehicle interaction, 36
Dynamic simulations--analysis of trucks, 7
BAMMS. See Bondgraph-based algorithm for Dynamic tire forces, 27
modeling multibody systems. Dynamometer rolls surface, noise
Bond graphs, 8 simulation, 204
Bondgraph-based algorithm for modeling
multibodysystems
(BAMMS) trucks, 7
vehicle simulation program, 1 (overview)
Economy, road transportation, l(overview)
Bond graphs, 7
Epoxidized natural rubber, 107
Buses, rolling resistance, 225
Evenness, pavement, influence on fuel
consumption, 240
C
Cars, fuel consumption, 240, 248

Cars, rolling resistance, 225, 248
Chassis development, 173 Friction, pavement rejuvenator, 119
Chipseal road surfaces, 248 Fuel economy, 107, 225, 240, 248
Coarse road surface, noise simulation, 204 Fuel tanker, simulation validation, 27
Coastdown, 248
Coefficient of friction, 119 G
Computer simulation programs
BAMMS, 1 (overview)
Goods vehicles, rolling resistance, 225
mobile load simulator, 67
road noise test equipment, 204
road-vehicle interaction, 36 H
Contact/no-contact simulation model, ride
quality, 150 Handling, modeling and simulation program
Cracking, 52 road testing machine, 89
Creep test, 67 steering and suspension, 173
trucks, 7
D Heated ISO track, 185
Heavy trucks
Danish road testing machine, 89 tire forces, 27
Denmark, research project, road testing pavement damage, 52
machine, 89 Heavy vehicle simulation, 27
Design, mobile load simulator, 67 Highway surface, vehicle/road interaction,
Design, prototype vehicle, 173 107
265
tires, 107
Prediction, noise, 204
Ideal polymer hysteresis properties, 107
Inertial profilometer, 135 Q
Instrumentation, Danish road testing
machine, 89 Quality, road transportation, 1 (overview)
Interior noise, effect of road surfaces, 204
International roughness index, 135 R
Research project, Denmark, road testing

machine, 89
Laboratory simulation, road noise, 185, 204 Ride comfort, 150
Leaf spring suspension, 27 Ride number, 135
Light vehicle fuel consumption, 240 Ride quality, 2 (overview), 150, 173
Ride quality index, 135
M Rider/seat interaction, 2 (overview), 150
Road characteristics
Macrotexture, road surface, effect on fuel fuel economy, 240
consumption, 240, 248 noise testing, 206(table)
Mays road meter index, 135 Road damage, heavy trucks, 52
Measurements, Danish road testing machine, Road handling behavior of race truck, 7
89 Road meter index, 135
Mechanical system simulation, 150 Road noise testing, 185, 204, 206-207(tables)
Michigan ride quality index, 135 Road roughness, 2 (overview)
Mobile load simulator (MLS), 67, 68(figs) Road surface influence, 185, 204, 240
Modeling and simulation program Road surface irregularities, effect on fuel
BAMMS demonstration, 7 consumption, 240
Models Road testing machine, Danish, 89
BAMMS demonstration, 7 Road/tire interaction, 248
contact/no-contact ride quality, 150 Road transportation, characteristics affected
pavements, 67 by roads and vehicles, 1
truck-pavement interaction, 52 (overview), 248
Road-vehicle interaction
N fuel consumption, 225
heavy trucks, computer simulation, 36
New Zealand road surfaces, rolling steering and suspension, 173
resistance, 248 Rolling resistance
Noise testing, vehicle tire/road interactions, effect on fuel consumption, 225, 240, 248
185, 204, 206-207(tables) influence of tread polymers, 107
New Zealand road surfaces, 248
O Roughness, 135, 240
Rutting, 52, 67
Objective testing, 173
Safety, road transportation, 1 (overview)

Pavement loading, 2 (overview) Safety, tires, 107
Pavement damage, heavy truck properties, 52 Seating. See Rider/seat interaction.
Pavement design and engineering, 89 Serviceability index, 13
Pavement rejuvenator, 119 Simulation programs
Pavement ride quality, 135, 150 BAMMS demonstration, 7
Pavement roughness and texture,135,240,248 contact/no-contact model, 150
Pavement stress, 52 mobile load simulator, 67
Pavement testing, 67, 89 ride quality, 150
Polymer hysteresis properties, influence on road noise, 204
SUBJECT INDEX 267
road testing machine, 89 influence of tread polymer on wear

three and four-axle trucks, 36 properties, 107
tools for, development and analysis, 7 integrity, 107
Simulation--continued Tools for modeling and simulation program,
vehicle dynamics, 150 development and analysis, 7
Skid testing, 119 Traction and wear properties, 107
Software, BAMMS, 8 Traction tests, 119
Stability, simulation and analysis of trucks, 7 Tractor trailor stability
Steady-state torque method, 248 BAMMS analysis, 1 (overview)
Steering/suspension development of road Traffic regulation, effect on fuel economy, 240
vehicles, 173 Tread polymer, influence on traction, 107
Strain, 89 Truck-pavement interaction, 52
Stress, 89 Truck, racing
Stress analysis, 89 BAMMS analysis of performance
Subjective testing, 173 Truck tires, 52'
Surface and texture, New Zealand roads, 248 Truck trailor model--stability, 7
Suspension Trucks, three- and four-axles, road-vehicle
development of road vehicles, 173 interaction, 36
fuel economy, 225
heavy truck pavement damage, 52 V
ride quality, 150
Vehicle
behavior dynamics, 7, 150, 173
Vehicle
noise interaction, 185
Vehicle
operating costs, 248
Vehicle
pavement interaction, 36
Tandem axle spacing, 52 Vehicle
pavement interaction, influence of tire
Temperature effects, vehicle noise testing, 185 materials, 107
Test equipment, road noise simulations, 204 Vehicle ride and handling, 173
Test vehicles for road noise, 207(table) Vehicle-road interaction
Testing overview, 1
drive-by noise, 185 noise, 185
fuel economy, 225, 240, 248 fuel economy, 225
road noise simulation, 204, 206-207(tables) Vehicle simulation, 67
road testing machine, 89 Vehicle-simulation
rolling resistance, 248 BAMMS, 1 (overview)
steering and suspension, 173 heavy trucks, 27
temperature, influence on noise, 185 Vehicle skid testing, 119
vehicle/road/tire noise, 185 Vehicles, description for noise tests, 207(table)
Texas Department of Transportation, 67 Vibration simulation, heavy trucks, 27
Texas mobile load simulator, 67 Vinyl butadienes, 107
Texas serviceability index, 135 Vinyl styrene butadiene, 107
Tire/car rolling resistance, 248
Tire characteristics, 2 (overview) W
Tire construction, 107
Tire properties, heavy trucks, 52 Wear properties of tires, influence of tread
Tire/road influence, 185 polymer, 107
Tire/road noise, 185 Weighting functions, road noise simulation,
Tire tread rolling resistance, 107 204
Tires Wet pavements, 119
computer simulation of dynamic forces, 1 Wheel/pavement interaction, mobile load
(overview) simulator, 67
Z~SN O-aO3~-lSq3-?
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductio

Vehicle Road Interaction - STP1225

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Vehicle Road Interaction - STP1225

Uploaded by

Copyright:

Available Formats

Copyright by ASTM Int'l (all rights reserved); Sat Dec 24 01:23:58 EST 2011

Bohdan T. Kulakowski, Editor

ASTM Publication Code Number (PCN):

Peer Review Policy

This publication, Vehicle-Road Interaction, contains papers presented at the Vehicle-Road

MODELING AND SIMULATION OF VEHICLE DYNAMICS AND VEHICLE-ROAD

Simulation and Analysis of Trucks Using the Modeling and Simulation

Predicting Vertical Dynamic Tire Forces of Heavy Trucks--DAVID J. COLE

Computer Simulation of Road-Vehicle Dynamic Interaction Forces of Three-

Heavy Truck Properties Significant to Pavement Damage--THOMAS D. GILLESPIE

LABORATORY AND FIELD TESTS OF VEHICLE-INDUCED PAVEMENT LOADING

Measured and Calculated Pavement Response in the Danish Road Testing

Influence of Tread Polymer on Traction, Rolling Resistance, and Wear

Traction Tests on an Expressway Treated with a Pavement Rejuvenator--

RIDE QUALITY AND ROAD ROUGHNESS

Development and Evaluation of the Ride Number Concept--

A Geometry-Based Contact/No-Contact Simulation Model Applied to Ride

ADVANCES iN VEHICLE SUSPENSION DESIGN AND CONTROL

Steering and Suspension Development of Road VehiclesmCHRIS J. ALSTEAD AND

NOISE EMISSION DUE TO VEHICLE-TIRE-ROAD INTERACTION

Vehicle Noise Testing Problems Due to Vehicle/Tire]Road Interactions--

Road Noise Simulation Using Road Surface Weighting Functions--

FUEL EFFICIENCY AND ROLLING RESISTANCE

The Effect of Vehicle-Road Interaction on Fuel Consumption--LASZLO GYENES

The Influence of Pavement Evenness and Macrotextureon Fuel Consumption--

Rolling Resistance Character~tics of New Zealand Road Surfaces--

Modeling and Simulation of Vehicle Dynamics and Vehicle-Road Dynamic Interaction

Laboratory and Field Tests of Vehicle-Induced Pavement Loading

Ride Quality and Road Roughness

constraints in the rider-seat-vehicle system. These nonlinearities become increasingly important

Advances in Vehicle Suspension Design and Control

Noise Emission Due to Vehicle-Tire-Road Interaction

Fuel Efficiency and Rolling Resistance

Simulation and Analysis of Trucks Using

ABSTRACT: A description is given of a general purpose, open-structured simulation tool

The BAMMS Simulation Program

History of BAMMS' Development

Using Bond Graphs

User-Definable Block Operators--All equations in a model are stored as blocks in an overall

The Modeling Phase

Mechanical Macros Available in BAMMS

Transformation Macros--Defining the algebraic relationships between flows and efforts of

Effort-producing Macros--These macros define massless connections between flow-produc-

Simulation Program Features

Analysis and Optimization of a Race Truck

Handling C u r v e s - - G e o m e t r i c linearization of the system equations (i.e., for a small path

= steer angle of front wheel,

= rotations constrained I = displacement constrained

= rotations free I = displacement free

= torsional spring t = linear spring

I = massless flexible element ~ = mass (inertia)

FIG. 2--Chassis components of the complex truck model.

Simulation Runs Performed

Damping of the Differential--This parameter determines whether the differential is locked

(1) simple model and (2) complex model

(1) Normal and (2) double frame stiffness

Speed Sweep at Constant Path Curvature

Effect of differential lock and curve radius

Spiral Run at Constant Speed

Mode Analysis and Root Loci

Handling curves with double rear tyres

0.35 "3" "l

TABLE l--Relevant modes of the simple race truck model at 30 mls.