Você está na página 1de 76

AP-T97/08

AUSTROADS TECHNICAL REPORT

Development of HDM-4 Road Deterioration


(RD) Model Calibrations for
Sealed Granular and Asphalt Roads
Development of HDM-4 Road Deterioration (RD) Model
Calibrations for Sealed Granular and Asphalt Roads
Development of HDM-4 Road Deterioration (RD) Model Calibrations for
Sealed Granular and Asphalt Roads

First Published May 2008

© Austroads Inc. 2008

This work is copyright. Apart from any use as permitted under the Copyright Act 1968,
no part may be reproduced by any process without the prior written permission of Austroads.

Development of HDM-4 Road Deterioration (RD) Model Calibrations for


Sealed Granular and Asphalt Roads

ISBN 978-1-921329-56-2

Austroads Project No. AT1064

Austroads Publication No. AP–T97/08

Project Manager
Ian Hickson

Prepared by
Zahidul Hoque, Dr Tim Martin and Lith Choummanivong

Published by Austroads Incorporated


Level 9, Robell House
287 Elizabeth Street
Sydney NSW 2000 Australia
Phone: +61 2 9264 7088
Fax: +61 2 9264 1657
Email: austroads@austroads.com.au
www.austroads.com.au

Austroads believes this publication to be correct at the time of printing and does not accept
responsibility for any consequences arising from the use of information herein. Readers should
rely on their own skill and judgement to apply information to particular issues.
Development of HDM-4 Road Deterioration (RD) Model
Calibrations for Sealed Granular and Asphalt Roads

Sydney 2008
Austroads profile
Austroads’ purpose is to contribute to improved Australian and New Zealand transport outcomes
by:
ƒ providing expert advice to SCOT and ATC on road and road transport issues
ƒ facilitating collaboration between road agencies
ƒ promoting harmonisation, consistency and uniformity in road and related operations
ƒ undertaking strategic research on behalf of road agencies and communicating outcomes
ƒ promoting improved and consistent practice by road agencies.

Austroads membership
Austroads membership comprises the six state and two territory road transport and traffic
authorities and the Australian Department of Transport and Regional Services in Australia, the
Australian Local Government Association and Transit New Zealand. It is governed by a council
consisting of the chief executive officer (or an alternative senior executive officer) of each of its
eleven member organisations:

ƒ Roads and Traffic Authority New South Wales


ƒ Roads Corporation Victoria
ƒ Department of Main Roads Queensland
ƒ Main Roads Western Australia
ƒ Department for Transport, Energy and Infrastructure South Australia
ƒ Department of Infrastructure, Energy and Resources Tasmania
ƒ Department of Planning and Infrastructure Northern Territory
ƒ Department of Territory and Municipal Services Australian Capital Territory
ƒ Department of Infrastructure, Transport, Regional Development and Local Government
ƒ Australian Local Government Association
ƒ Transit New Zealand

The success of Austroads is derived from the collaboration of member organisations and others in
the road industry. It aims to be the Australasian leader in providing high quality information, advice
and fostering research in the road sector.
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

SUMMARY

Background and Scope


Pavement deterioration data for sealed granular and asphalt roads in the LTPP database are
currently insufficient, in terms of the variation in the values of the variables, to develop reliable
calibrations for HDM-4 road deterioration (RD) models in Australasia. However, wide ranging
historical deterioration data from the state and New Zealand road authority (SRA) networks are
available, although the quality of this SRA data is not as good as the LTPP data. Consequently,
SRA data from some states were used to calibrate HDM-4 RD models which should improve the
reliability of deterioration predictions for the State and New Zealand road networks in which the RD
model calibrations were made. These calibrated HDM-4 RD models can be either refined or
simplified when new data from the existing and the additional LTPP sites become available.

This report documents the calibration of HDM-4 RD models for sealed granular and asphalt
pavements based on SRA historical deterioration data. SRAs supplied historical roughness and
rutting deterioration data and some supplied cracking deterioration data and maintenance history.
HDM-4 RD models were calibrated to suit conditions in Victoria, Tasmania, South Australia,
Queensland and New Zealand. Rutting and roughness RD models were calibrated for all these
SRAs, except Queensland. Cracking RD models were calibrated for South Australia due to the
reasonable quality of its cracking data. Cracking data from Victoria, Tasmania and New Zealand
were either of inadequate quality or too insignificant, in terms of extent of cracking, to be
considered in the analysis, hence cracking models were not calibrated for the data from these
authorities.

The development of generic RD models for roughness, rutting, cracking and deflection is expected
to commence during 2007-08 to enable wider application of these models.

Findings
An objective approach to estimating the underlying rate of deterioration from a time series of
deterioration data was used to represent the observed deterioration which was then matched with
the RD model predicted rate of deterioration to calibrate the RD models. The objective approach
to estimate the underlying rate of pavement deterioration was developed by another Austroads
project using a set of decision rules to eliminate data noise and the effect of maintenance or
rehabilitation treatments on deterioration. A detailed description of the calibration procedure is
presented in this report.

Victorian road network


A total of 55 asphalt and sprayed seal pavement sections were used for RD model calibration in
Victoria based on their observed deterioration using the objective estimation approach outlined
above. These sections were grouped according to road type, pavement type, traffic loading and
climate conditions. The results from the current RD model calibration were compared with the
values obtained from a previous calibration study. Mean values of the calibrated coefficients (Krst
and Kgm) were estimated for each group of pavements. The mean values of the calibration
coefficients derived from the current and previous studies were on average comparable, although
the current study generally estimated lower mean values compared to those from the previous
study.

Austroads 2008

— i—
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

When comparing the calibration coefficient (K) values with the estimated rate of pavement
deterioration, the results from the current calibration study showed an improved correlation
compared to the results from the previous calibration study. The correlation coefficients (r2) from
the current study between the estimated rates of deterioration and the K values were greater than
0.66, while those from the previous study were only 0.39.

Tasmanian road network


A total of 58 pavement sections from the Department of Infrastructure, Energy and Resources
(DIER), Tasmania, road network were calibrated by this study. Similar to the Victorian network, the
selected sections were grouped according to road type, pavement type, traffic loading and climate
condition in order to estimate and compare mean calibration coefficient (K) values from the current
calibration study with those from a previous calibration study. A comparison of the mean K values
for each group of pavements indicated that the current study generally yielded lower K values
compared to the K values from the previous study.

The calibration results from the current study were good because the correlation between the
calibration K values and the estimated rates of deterioration was high. The current study estimated
lower K values (Krst and Kgm) because the observed rates of deterioration were estimated to be
lower. The correlation coefficients (r2) between the calibration values and the estimated rates of
deterioration for the revised method were greater than 0.82, while the correlation coefficients (r2)
for the previous study were less than 0.13.

South Australian road network


Of the 129,000 pavement sections with deterioration data available, only 14,765 sections were
estimated to be experiencing deterioration concurrently with cracking, rutting and roughness and
these were used in the current study. Because of limited computational capacity, the calibration of
all the 14,765 sections was not undertaken. Instead, pavement sections were grouped in terms of
road type, pavement type, traffic loading and climatic conditions prior to calibration. The mean
values of observed and predicted deterioration for each group of pavements were estimated or
assumed and calibration was based on these mean values.

The calibration coefficients for cracking, rutting and roughness (Kgm, Krst and Kgm) for each group of
South Australian pavements are presented in this report. It should be noted that these K values are
applicable to pavement sections whose characteristics, in terms of road and pavement type, traffic
loading and climatic conditions, are similar to the mean values used to group the pavement
sections. The correlation between the calibration coefficient K values and the estimated rate of
deterioration was high for rutting (r2 = 0.89), practically nil for cracking, and reasonable for
roughness (r2 = 0.45). With future improvements in computational capacity and an efficient
deterioration algorithm, it may be possible to estimate the deterioration of all individual pavement
sections in a large road network so that the mean calibration coefficient K values for various
groups of roads could be estimated more accurately.

Queensland road network


Roughness deterioration, traffic and inventory data for over 28,600 pavement sections, each 1 km
in length, were made available by the Queensland Department of Main Roads (QDMR) for the
study. Of these, more than 5300 sections had satisfied the selection criteria which included the
following: (1) traffic with AADT less than 5,000; and, (2) deterioration identified and estimated by
the principles and decision rules of the underlying rate of deterioration approach outlined above.

Austroads 2008

— ii —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

Due to the large number of sections representing the network, the number of sections for the
calibration study was reduced by grouping the sections based on the usual criteria of road type,
pavement type, traffic loading and climate conditions. Group mean values of observed and
predicted deterioration, not the values of individual sections, were used as input data for
calibration.

The roughness calibration of 57 section groups, based on the mean values of estimated and
predicted deterioration, produced a range of calibration coefficient, K values, generally less than
the HDM-4 default value of 1.0. The correlation between group means of the estimated rate of
deterioration and the K values was reasonable (r2 = 0.5) given that there was some reduction in the
quality of input data due to using mean values of deterioration.

New Zealand State Highway network


A large data set from 63 pavement sections, each 300 m in length and monitored on a yearly basis
to date, was provided by Transit New Zealand (TNZ). Performance data from 2001 to 2005
including roughness, rutting, cracking and deflection were collected either from each lane of a two-
lane two-way road or from one of the lanes on a divided carriageway. As each lane was treated
independently, the total number of lane sections available for the analysis increased to 124.

As for the Australian states, the selection of pavement sections for calibration was based on the
estimation of underlying rate of deterioration to represent the observed deterioration. It was found
that 54 of the 124 sections showed concurrent increases in rutting and roughness whereas only a
small number of sections were experiencing cracking deterioration. Consequently, RD models
were calibrated only for rutting and roughness using input data from the individual sections.

The calibration coefficients for rutting and roughness (Krst and Kgm) were plotted against and
correlated with their corresponding estimated rates of deterioration. The correlation between the
calibration coefficients and the estimated rates of deterioration was strong (r2 = 0.70 and 0.85 for
rutting and roughness respectively), indicating that the calibration approach produced reliable
estimates.

Findings
The calibration of a range of HDM-4 RD models was completed using all the data made available
by the SRAs in 2005/06 and 2006/07. The following outcomes were achieved:
ƒ estimation of calibration coefficients for roughness, rutting and, where possible, cracking for
road networks in Victoria, Tasmania, South Australia and Queensland
ƒ estimation of calibration coefficients for roughness and rutting for the State Highway network
in New Zealand.

Conclusions
The objective approach used to estimate the observed deterioration from SRA data allows the
calibration of RD models based on the underlying pavement deterioration. The correlation between
the calibration coefficients and the estimated rates of observed deterioration was usually strong,
indicating that the calibration approach produced reliable estimates.

With future improvements in computational capacity and an efficient deterioration algorithm, it may
be possible to calibrate all individual pavement sections in a large road network so that the mean
calibration coefficient K values for various groups of roads can be estimated more accurately.

Austroads 2008

— iii —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

CONTENTS
1 INTRODUCTION ............................................................................................................ 1
1.1 Background..................................................................................................................... 1
1.2 Scope of this Report ....................................................................................................... 1
1.3 Report Structure ............................................................................................................. 1
2 HDM-4 ROAD DETERIORATION (RD) MODELS ......................................................... 2
2.1 Background of HDM-4 RD Models ................................................................................. 2
2.2 Distress Modes ............................................................................................................... 2
2.3 Calibration Factors.......................................................................................................... 3
2.4 RD Models ...................................................................................................................... 4
2.4.1 Background....................................................................................................... 4
2.4.2 Cracking Models and K Factors........................................................................ 5
2.4.3 Rutting Models and K Factors .......................................................................... 6
2.4.4 Roughness Models and K Factors.................................................................... 6
3 METHOD OF RD MODEL CALIBRATION .................................................................... 8
3.1 Previous Calibration........................................................................................................ 8
3.2 Calibration Approach ...................................................................................................... 8
3.2.1 Assumptions ..................................................................................................... 9
3.2.2 Calibration Procedure ....................................................................................... 9
4 CALIBRATION OF RD MODELS ................................................................................ 11
4.1 Scope of Calibration ..................................................................................................... 11
4.2 Victorian Road Network ................................................................................................ 11
4.2.1 Road Sections Considered for RD Model Calibration..................................... 11
4.2.2 Performance History ....................................................................................... 12
4.2.3 Maintenance Intervention ............................................................................... 13
4.2.4 Calibration Results.......................................................................................... 13
4.3 Tasmanian Road Network ............................................................................................ 14
4.3.1 Road Sections Considered for RD Model Calibration..................................... 14
4.3.2 Performance History ....................................................................................... 15
4.3.3 Maintenance Intervention ............................................................................... 15
4.3.4 Calibration Results.......................................................................................... 16
4.4 South Australian Road Network.................................................................................... 19
4.4.1 Road Sections for RD Model Calibration ........................................................ 19
4.4.2 Performance History ....................................................................................... 22
4.4.3 Maintenance Intervention ............................................................................... 22
4.4.4 Calibration Results.......................................................................................... 22
4.5 Queensland Road Network........................................................................................... 27
4.5.1 Road Sections for RD Model Calibration ........................................................ 27
4.5.2 Performance History ....................................................................................... 27
4.5.3 Maintenance Intervention ............................................................................... 27
4.5.4 Calibration Results.......................................................................................... 30
4.6 New Zealand State Highway Network .......................................................................... 35
4.6.1 Road Sections for RD Model Calibration ........................................................ 35
4.6.2 Performance History ....................................................................................... 36
4.6.3 Maintenance Intervention ............................................................................... 36
4.6.4 Calibration Results.......................................................................................... 36

Austroads 2008

— iv —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

4.7 Summary of Calibration Factors ................................................................................... 41


4.7.1 Victorian Road Network .................................................................................. 41
4.7.2 Tasmanian Road Network .............................................................................. 41
4.7.3 South Australian Road Network...................................................................... 41
4.7.4 Queensland Road Network............................................................................. 42
4.7.5 New Zealand State Highway Network ............................................................ 42
4.7.6 Comparison of Calibration Factors ................................................................. 43
5 WORK SUMMARY....................................................................................................... 44
6 CONCLUSIONS AND RECOMMENDATION .............................................................. 45
6.1 Conclusions .................................................................................................................. 45
6.2 Recommendation.......................................................................................................... 45
REFERENCES ...................................................................................................................... 46
APPENDIX A HDM-4 PAVEMENT CODES, CLASSIFICATIONS AND
DETERIORATION MODELS............................................................ 47
APPENDIX B ESTIMATION OF UNDERLYING RATE OF DETERIORATION...... 59

Austroads 2008

— v—
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

TABLES
Table 2.1: Types of distress and independent variables ................................................... 2
Table 2.2: Effect of pavement classification on deterioration models ................................ 3
Table 2.3: Calibration factors used in the deterioration models......................................... 4
Table 4.1: Characteristics of Victorian pavement sections considered for analysis ........ 12
Table 4.2: Calibration results for pavement sections in Victoria ...................................... 13
Table 4.3: Characteristics of the Tasmanian pavement sections considered for
analysis........................................................................................................... 16
Table 4.4: Calibration results for the pavement sections in Tasmania ............................ 17
Table 4.5: Characteristics of the South Australian pavement sections considered
for analysis...................................................................................................... 20
Table 4.6: Calibration results for the pavement groups in South Australia ...................... 24
Table 4.7: Characteristics of the Queensland pavement sections considered for
analysis........................................................................................................... 28
Table 4.8: Calibration results of the pavement groups in Queensland ............................ 31
Table 4.9: Characteristics of the New Zealand pavement section groups for
analysis........................................................................................................... 38
Table 4.10: Calibration results for the pavement groups for the New Zealand road
network ........................................................................................................... 39
Table 5.1: Summary of HDM-4 RD calibration results ..................................................... 44

FIGURES
Figure 2.1: Deterioration phases of sealed granular pavements ........................................ 5
Figure 4.1: Relationship between rate of deterioration and RD model coefficients for
the Victorian road network .............................................................................. 14
Figure 4.2: Relationship between rates of deterioration and RD model coefficients
for the Tasmanian road network ..................................................................... 18
Figure 4.3: Relationship between rate of deterioration and RD model coefficients for
South Australian road network........................................................................ 26
Figure 4.4: Relationship between rate of deterioration and RD model coefficients for
Queensland road network............................................................................... 35
Figure 4.5: Relationship between rate of deterioration and RD model coefficients for
New Zealand State Highway network ............................................................. 40

Austroads 2008

— vi —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

1 INTRODUCTION

1.1 Background
There is a need for improved road deterioration (RD) models for the sealed and asphalt road
network to reliably predict future pavement conditions due to changes in traffic loading and
maintenance strategies. Pavement deterioration data for sealed granular and asphalt roads in the
LTPP database are currently insufficient, in terms of the variation in the values of the variables, to
develop reliable calibrations for HDM-4 RD models in Australasia. However, wide ranging historical
deterioration data from the state and New Zealand road authority (SRA) networks are available,
although the quality of this SRA data is not as good as the LTPP data. SRA data from some states
may therefore be used to calibrate HDM-4 RD models which should improve the reliability of
deterioration predictions for the state and New Zealand road networks in which the RD model
calibrations can be made. The calibrated HDM-4 RD models can be either refined or simplified
when new data from the existing and the additional LTPP sites becomes available.

In 2004/05, the interim calibration of HDM-4 RD models was undertaken for sealed granular
pavements in the Victorian road network (Hoque and Martin 2005). In 2005/06 calibration of
HDM-4 RD models was undertaken for Victorian, South Australian and Tasmanian road networks
based on deterioration data provided by each of these SRAs (Hoque and Martin 2006b). These
calibrations were based on an approach (Martin and Hoque 2006) that aimed at assessing the
underlying rate of deterioration from the time series of SRA supplied deterioration data to represent
the observed deterioration experienced on specific pavement segments. These rates of
deterioration were then matched with those predicted by the RD models on specific pavement
segments to calibrate the RD models.

1.2 Scope of this Report


This report documents the calibration of the HDM-4 RD models for sealed granular and asphalt
pavements based on historical deterioration data supplied by SRAs. SRAs supplied roughness and
rutting data and some supplied cracking and maintenance history. The time series deterioration
data on sealed granular and asphalt pavement segments in the road networks for Victoria,
Tasmania, South Australia, Queensland and New Zealand was made available to ARRB via either
related projects or supplied directly.

1.3 Report Structure


The report contains the following major parts:
ƒ Introduction – provides background, summary of the previous interim RD calibration report,
objectives and scope
ƒ HDM-4 Road Deterioration (RD) Models – provides an overview of HDM-4 RD models
ƒ Calibration method – provides details of the improved calibration procedure
ƒ Calibration results – provides calibration results and analysis
ƒ Conclusions – provides general conclusions of the investigation and recommends
methodology to use in the future.

Austroads 2008

— 1—
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

2 HDM-4 ROAD DETERIORATION (RD) MODELS


HDM-4 RD models are considered to be robust and are increasingly being adopted by many
SRAs. Some SRAs have engaged ARRB to conduct separate studies to calibrate these RD
models to suit their local conditions (Martin et al. 2006). An overview of these models is presented
in the following sub-sections.

2.1 Background of HDM-4 RD Models


An international collaborative study known as ‘The International Study of Highway Development
and Management (ISOHDM)’ was initiated in 1993 to extend the scope of the HDM-III models. The
HDM-III models were statistically estimated from data collected during a multi-year empirical study
carried out in Brazil (GEIPOT, 1982). The statistical relationships were validated and extended
using data from several other deterioration studies such as those from Kenya, the Caribbean,
India, Texas, etc. The revised and improved models are known as HDM-4 RD models and are
described in detail in:
ƒ HDM-4 Volume 4. Analytical Framework and Model Descriptions (Odoki and Kerali 2000)
ƒ HDM-4 Volume 6. Modelling Road Deterioration and Works Effect (Morosiuk, Riley and
Odoki 2001).

2.2 Distress Modes


The types of deterioration of a sealed granular pavement can be categorised into cracking, surface
disintegration, permanent deformation, longitudinal profile and friction. The development of these
modes of deterioration can be dependent on a number of factors which are broadly classed as
pavement strength, material properties, traffic loading and environment. Table 2.1 shows the
distress types which are modelled and the independent variables which are used in the
deterioration models.

Table 2.1: Types of distress and independent variables


Distress Distress Pavement Materials Traffic Environment
mode type strength properties loading
Structural √ √ √ √
Cracking Reflection √ √
Transverse thermal √ √
Ravelling √ √ √
Potholing √ √ √ √
Disintegration
Rutting – surface wear √ √
Edge break √ √ √
Rutting – structural √ √ √ √
Deformation
Rutting – plastic flow √ √ √
Profile Roughness √ √ √ √
Texture depth √ √
Friction
Skid resistance √ √

Austroads 2008

— 2—
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

HDM-4 Volume 4 (Odoki and Kerali 2000) introduced the system of pavement classification used in
HDM-4 (Appendix A). The structure of a model used to predict the initiation or progression of a
certain distress may be governed by surface type, base type or a combination of both (pavement
type). In other cases the model structure is the same for all types of surfacing and base, but the
default model coefficients are dependent on surfacing or base type. In other cases the model
structure and default coefficients are independent of both surfacing and base types. Table 2.2
summarises these relationships.

Table 2.2: Effect of pavement classification on deterioration models


Distress mode Distress type Surfacing type Base type
Structural S S
Cracking Reflection C C
Transverse thermal C
Ravelling C
Potholing
Disintegration
Rutting – surface wear
Edge break C C
Rutting – structural C
Deformation
Rutting – plastic flow C
Profile Roughness
Texture depth C
Friction
Skid resistance C
S – structure of model may change by pavement type
C – coefficients of model may change by pavement type

2.3 Calibration Factors


The RD models include a number of calibration factors to facilitate calibration of the RD models to
local conditions. The calibration factors are denoted by the letter ‘K’ together with identifying
subscripts given in Table 2.3.

These factors are multiplicative and are used to change the scale of a particular distress. The
default value for all the ‘K’ factors is 1.0. For example, Kcia is the calibration factor for the initiation
of all structural cracking in bituminous pavements. By increasing the value of Kcia to 2.0, the time to
the initiation of all structural cracking is doubled, implying that the pavement will last longer before
cracks appear than that predicted by a default calibration of HDM-4. Similarly, increasing the
calibration factor for the progression of all structural cracking, Kcpa, to 2.0 implies that the pavement
will deteriorate, in terms of the rate of crack progression, twice as fast as that predicted by the
default calibration of HDM-4.

In addition to an increased number of calibration factors, another important addition to the RD


models in HDM-4 is the use of adjustable model coefficient values, referred to as the ‘ai’ values. In
HDM-4, the ai values for the variables in each relationship are not hard coded into the software.
Instead a default value has been assigned to each of these model coefficients, which the user can
alter. Calibration is discussed in more detail in Volume 5 of the HDM-4 series – A Guide to
Calibration and Adaptation (Bennett and Paterson 2000).

Austroads 2008

— 3—
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

Table 2.3: Calibration factors used in the deterioration models


Deterioration model Calibration factor
Wet/dry season SNP ratio Kf
Drainage factor Kddf
All structural cracking – initiation Kcia
Wide structural cracking – initiation Kciw
All structural cracking – progression Kcpa
Wide structural cracking – progression Kcpw
Transverse thermal cracking – initiation Kcit
Transverse thermal cracking – progression Kcpt
Rutting – initial densification Krid
Rutting – structural deterioration Krst
Rutting – plastic deformation Krpd
Rutting – surface wear Krsw
Ravelling – initiation Kvi
Ravelling – progression Kvp
Pothole – initiation Kpi
Pothole – progression Kpp
Edge break Keb
Roughness – environmental coefficient Kgm
Roughness – SNPK Ksnpk
Roughness – progression Kgp
Texture depth – progression Ktd
Skid resistance Ksfc
Skid resistance – speed effects Ksfcs
Source: Morosiuk et al. (2001)

2.4 RD Models
2.4.1 Background
Accelerated load testing on sealed granular pavements in South Africa (Freeme 1983) indicated
three distinct phases of deterioration as illustrated in Figure 2.1. These deterioration phases, i.e.
the initial densification phase, the gradual deterioration phase, and the rapid deterioration phase,
were also identified in HDM-4 (Morosiuk et al. 2001) and are explained below:
ƒ The initial densification phase is the short period between just after opening to traffic and
initial settlement. The rate of deterioration is comparatively higher due to the initial
settlement.
ƒ In the gradual deterioration phase pavements are assumed to be in stable condition and thus
the rate of deterioration is expected to be gradual and linear.
ƒ The rapid deterioration phase starts at the end of the gradual phase when pavement
condition deteriorates at a faster rate and leads to the onset of failure.

Austroads 2008

— 4—
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

The observed data available for the last 10 years indicated that the pavement sections generally
remained either in an unchanged condition or experienced a gradual rate of deterioration. Some
sections appeared to have experienced improved conditions possibly due to maintenance
treatments. A few sections also saw a very high rate of deterioration. Therefore the sections
considered for calibration were assumed to represent a wide spread of stages and conditions in
their performance lives.
Deformation or distress

Initial densification phase


Rapid deterioration phase

Gradual deterioration phase

0
0 Traffic load or time

Source: Martin and Hoque (2005)Figure 2.1: Deterioration phases of sealed granular pavements

A specific calibration sequence is required for RD calibration because of the interactive nature of
the HDM-4 RD models. This means that the cracking model is calibrated before the rutting model
calibration because the predicted rutting is influenced by the predicted cracking. Similarly, the
roughness models are calibrated after cracking and rutting calibration.

2.4.2 Cracking Models and K Factors


HDM-4 includes RD models for both structural and thermal cracking. Thermal cracking is caused
by binder stiffening and temperature variations. The RD thermal cracking model was not calibrated
nor is it relevant in Australia.

In terms of structural cracking, RD models exist for ‘all’ cracking and ‘wide’ cracking (>3 mm). As
the time to initiation of wide cracking is dependent on the time to initiation of ‘all’ cracking, only the
‘all’ cracking RD model was calibrated and the default calibration factors for ‘wide’ cracking were
adopted (Kciw=1, Kcpw=1). Details of the cracking models are provided in Appendix A.

The ‘all’ cracking model has two calibration factors; one for adjusting the time to initiation of
cracking (Kcia) and the other for adjusting the rate of crack progression (Kcpa). The data were not
sufficient to determine the all cracking initiation calibration factor (Kcia) independently of the
progression calibration factor (Kcpa). Consequently, the calibration was undertaken assuming the
following relationship between the two factors (Bennett and Paterson 2000, Section 7.2.3):

Kcpa =1/Kcia 1

Austroads 2008

— 5—
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

2.4.3 Rutting Models and K Factors


The details of rutting models are given in Appendix A. The rutting model is based on four
components of rutting, the rut depth at any time being the sum of the four components. These four
components are as follows:
ƒ initial densification (rutting in the first year after new construction or reconstruction that
includes a new base layer)
ƒ structural deformation (structural rutting following initial densification)
ƒ plastic deformation (shoving/deformation of asphalt layers)
ƒ surface wear from studded tyres (rutting from studded tyres used on snow covered roads -
applicable only to cold climates).

The model for initial densification is used to predict the deformation of unbound granular materials
and subgrade for new pavements in the first year of trafficking. In a previous study (Toole et al.
2004a, Toole et al. 2004b) the K factors for initial densification (Krid) were selected based on the
experience on similar roads and on local engineering experience. The same Krid values were used
in this analysis of the Tasmanian and Victorian road networks. For South Australia a set of Krid
factors was assumed.

The model for surface wear is not relevant in Australia.

The model for plastic deformation predicts the plastic flow (shoving) of asphalt layers or long-term
plastic deformation (creep) of thick asphalt pavements. This model is not relevant because the
pavement sections considered for calibration are sealed granular and thin asphalt pavements.

The structural component of the rutting model was considered in this analysis and its coefficient
(Krst) was calibrated.

2.4.4 Roughness Models and K Factors


Roughness progression is predicted in HDM-4 as the sum of five components; structural, cracking,
rutting, potholing and the environment. Roughness is calculated at the end of each year, taking into
account the change in condition for each mode of distress sequentially for each year of an analysis
period.

The total annual incremental change in roughness is the sum of the various components described
above and shown in Eq. (2). The details of roughness models are given in Appendix A.

ΔRI = Kgp [ΔRIs + ΔRIc + ΔRIr + ΔRIt ] + ΔRIe 2

where;
Kgp = calibration factor for roughness progression
ΔRI = total incremental change in roughness during analysis year, in m/km IRI
ΔRIs = incremental change in roughness due to structural deterioration during analysis year,
in m/km IRI
ΔRIc = incremental change in roughness due to cracking during analysis year, in m/km IRI
ΔRIr = incremental change in roughness due to rutting during analysis year, in m/km IRI
ΔRIt = incremental change in roughness due to potholing, in m/km IRI
ΔRIe = incremental change in roughness due to environment during analysis year, in m/km
IRI
RIa = roughness at the start of the analysis year, in m/km IRI

Austroads 2008

— 6—
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

m = environmental coefficient
Kgm = calibration factor for the environmental component.

As seen from the above, there are two calibration factors which can be calibrated:
ƒ environmental component (Kgm)
ƒ roughness progression (Kgp).

Bennett and Paterson (2000) advise that usually only the Kgm factor is used to adjust the HDM-4
roughness predictions to local conditions. However, during the calibration it was found that the Kgp
factors needed adjustment to match the predicted and measured roughness. In some cases the
difference between the predicted and measured roughness was large when the Kgp was set to the
default value of 1.0. When Kgp was set to equal Kgm, the gap between the predicted and measured
roughness was reduced. Therefore Kgp was always set equal to Kgm for calibration.

Austroads 2008

— 7—
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

3 METHOD OF RD MODEL CALIBRATION

3.1 Previous Calibration


As noted earlier, calibrations of the coefficients for the HDM-4 RD models for roughness, rutting
and cracking were undertaken in a separate study for the Victorian and Tasmanian sealed and
asphalt road network (Toole et al. 2004a, Toole et al. 2004b). In this calibration, pavement
conditions were predicted by assuming the K factor(s) in question. Using a trial and error method,
K factors were chosen so that the predicted conditions, using HDM-4 software, matched as close
as possible the observed conditions or expected trend. A number of assumptions and adjustments
were made so that the predicted condition would follow an expected trend for a pavement section
under a particular loading and environmental condition. Where expected trends were used they
were estimated from local experience of pavements under similar loading and environmental
conditions.

Site specific initial pavement conditions (just after pavement construction, rehabilitation or in some
cases maintenance treatments) in terms of roughness, rutting and cracking were assumed based
on the experience of similar road types under similar climate and traffic loading conditions in
conjunction with the local knowledge of the responsible engineers.

The following values were assumed for all sections:


ƒ routine maintenance - annual drainage maintenance and potholes patched within two weeks
of occurrence
ƒ construction quality defect factor for bituminous surfacing (CDS) was 1.0 and base (CDB)
equal to 0 (no defects). Relative compaction was varied
ƒ drainage life calibration factor was 1.0
ƒ drainage factor equal to 1.0
ƒ crack retardation factor equal to 1.0
ƒ zero percentage thermal cracking
ƒ ravelling retardation factor equal to 1.0
ƒ climate conditions were chosen based on the site location and data from the Bureau of
Meteorology.

3.2 Calibration Approach


The principle used for calibration was to calibrate HDM-4 RD models based on continuous
deterioration data without including the immediate effect of maintenance treatments.

As noted in Section 1, a robust method of estimating pavement underlying rate of deterioration was
developed under another Austroads project. This is an automated process with flexibility that
considers data scattering (or noise) as well as the effect of maintenance or rehabilitation. The
effect of maintenance or rehabilitation was identified by a pre-defined change in pavement
condition. This pre-defined value is assumed and is adjustable subject to local experience and
conditions. Details of this method can be found in Martin and Hoque (2006) and in Appendix B.

Austroads 2008

— 8—
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

The method for estimating the underlying rate of deterioration is an objective approach that is not
influenced by judgement and operational speed. In this approach, a large number of pavement
segments can be analysed with the same set of rules in a short time. This facilitates the calibration
of a larger number of individual road network segments than previously possible. The approach
considers only those pavement segments that experience positive deterioration (increased distress
with time) and discards those that experience negative deterioration (less distress with time).

The approach only calibrates coefficients of RD models for pavement segments that have at least
three or more ‘valid performance data’ points. These data points are defined in Appendix B.

3.2.1 Assumptions
The following assumptions were made:
ƒ The available performance data were assumed to represent pavement conditions some
years after construction as the pavement performance history data did not contain
performance data for the whole life of the pavement.
ƒ The performance data were assumed to be relevant from the last construction/reconstruction
or rehabilitation. The effect of any previous (if any) maintenance, rehabilitation or
reconstruction was not considered.
ƒ Due to the above, the RD model coefficients were only calibrated for the period (or pavement
life) where pavement segments were experiencing positive deterioration.
ƒ It was assumed that only those pavement segments which experienced deterioration for all
distresses could have their relevant RD models calibrated. In other words, pavement
segments that showed negative deterioration were excluded from calibration.
ƒ The initial condition and all other K factors in previous calibrations (Toole et al. 2004a, Toole
et al. 2004b), that were not calibrated by this study, were assumed to be suitable for the
calibrations undertaken by this study.

3.2.2 Calibration Procedure


Calibration of the RD models involved the following steps:
1. Estimate the underlying rate of deterioration for roughness, rutting and cracking (if reliable
cracking data are available) on pavement segments, based on the objective approach
(Appendix B), using the observed time series pavement distress data.
2. Exclude the pavement segments that experience negative deterioration in any of the
distresses considered in this analysis. i.e. roughness, rutting and cracking.
3. For each segment considered for calibration, identify the period (may be defined as
‘deterioration period’) for which the rate of deterioration was estimated as well as identify the
pavement condition at the start of the deterioration period.
4. Estimate the input parameters to run the HDM-4 models. Key parameters include pavement
type, structural number, environment classifications, environmental data, i.e. rainfall,
temperature, traffic, start of the analysis year, pavement initial condition, etc.
5. Run HDM-4 models to predict pavement condition with the input parameters. Calculate the
predicted rate of deterioration for the ‘deterioration period’ and compare it with the estimated
underlying rate of deterioration.
6. Adjust the calibration coefficients for the RD models and repeat step five. Continue this
process until the predicted rate of deterioration matches closely with the underlying rate of
deterioration.

Austroads 2008

— 9—
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

7. Use the calibration coefficients that produce the closest match. In some cases, the
coefficients may be restricted to a minimum or maximum value even if the predicted and
observed underlying rates of deterioration did not match well.

Step two above was implemented because the RD models are for the deterioration phase only,
which excludes the separate influence of works effects (Section 3.2) that give an apparent
reduction in distress conditions which is assessed as negative deterioration. Because of the
structure of the HDM-4 RD models, the calibration sequence was cracking (when cracking
calibration was considered), rutting and roughness, as noted in Section 2.4.1.

Austroads 2008

— 10 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

4 CALIBRATION OF RD MODELS

4.1 Scope of Calibration


Pavement deterioration manifests in a variety of distresses, however, cracking, roughness and
rutting models are only considered for calibration because they are common distresses and their
performance data were readily accessible.

The historical deterioration data supplied by most SRAs contained sets of roughness and rutting
data, but only a few SRAs had cracking data. Consequently, roughness and rutting deterioration
were considered for RD model calibration. Cracking data from the South Australian road network
appeared to be reliable so the cracking RD model was calibrated using the South Australian data
set.

4.2 Victorian Road Network


4.2.1 Road Sections Considered for RD Model Calibration
The same pavement sections that were calibrated for the HDM-4 RD models for roughness, rutting
and cracking in the previous study (Toole et al. 2004b) were re-calibrated using the approach
outlined in Section 3.2.

The original data set comprised 82 sections covering a wide range of climate and loading
conditions. However, not all these sections were considered in the re-calibration. Estimation of the
underlying rate of deterioration showed that a number of sections had improved conditions (i.e.
exhibited negative deterioration of either roughness or rutting). As the aim of the re-calibration
exercise was to calibrate RD models for sections experiencing deterioration over the last several
years, those sections that had negative deterioration were excluded from the analysis. Therefore,
the analysis considered only 55 sections out of the 82 sections originally considered.

Pavement sections considered for the model development were grouped according to road type so
that RD models were calibrated to suit road types M, A, B, C and ‘Other’. Road type ‘M’ represents
freeways and motorways; ‘A’ roads represent major arterials; ‘B’ roads are state highways
connecting major cities; and ‘C’ roads are rural roads connecting smaller towns. ‘Other’ roads are
defined as roads that do not fall in the above four categories.

The pavement sections were further grouped according to pavement type, traffic level and climate
conditions. In order to maintain consistency, the same pavement classifications used in HDM-4
was also used in this exercise. Three groups of traffic levels and three groups of climate conditions
were used. The traffic groups were defined as: low (AADT < 5000), medium (5000 < AADT <
15000) and heavy (AADT > 15000). The climate groups were defined in terms of Thornthwaite
Index (TI): dry (TI < 0), temperate (0 < TI < 50) and wet (TI > 50). Table 4.1 lists characteristics of
pavement sections considered in the analysis.

Austroads 2008

— 11 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

Table 4.1: Characteristics of Victorian pavement sections considered for analysis


Road Pavement Traffic Climate No. of Pavement SNP Roughness Rutting
type1 type2 group3 group4 sections age5 (average)7 progression progression
(average (average)7 (average)7
years)7 (NRM/year)6 (mm/year)6
Dry 5 12 ~ 49 3.96 ~ 4.95 0.64 ~ 4.0 0.24 ~ 0.65
Low (33.75) (4.42) (1.81) (0.42)
A SRGB Wet 1 21 4.75 1.50 0.31
Temperate 5 20 ~ 37 5.7 ~ 5.95 0.64 ~ 1.59 0.03 ~ 0.35
Medium
(30) (5.8) (1.2) (0.16)
Dry 1 21 3.91 8.26 0.38
Low Temperate 10 15 ~ 41 3.58 ~ 5.78 .72 ~ 7.12 0.04 ~ 0.99
(32.6) (4.87) (2.47) (0.44)
B STGB
Dry 2 35 ~ 36 6.09 ~ 6.14 1.15 ~ 3.46 0.24 ~ 0.87
Medium (35.5) (6.12) (2.31) (0.55)
Temperate 1 12 5.95 0.51 0.22
Temperate 12 10 ~ 25 2.6 ~ 8.23 0.59 ~ 4.03 0.25 ~ 1.56
(19.33) (4.0) (1.89) (0.57)
C STGB Low
Wet 2 24 2.69 1.93 ~ 3.16 0.56 ~ 0.90
(2.54) (0.73)
AMAP Heavy Wet 1 21 7.0 1.29 0.15
AMGB Heavy Wet 1 26 6.68 6.82 0.05
AMSB Heavy Wet 1 26 7.0 3.52 0.10
Dry 6 22 6.11 0.12 ~ 6.82 0.05 ~ 0.64
M Low
(1.7) (0.27)
Temperate 2 13 6.29 0.42 ~ 0.46 0.08 ~ 0.10
STGB Medium
(0.44) (0.09)
Temperate 3 17 ~ 20 5.12 ~ 5.54 0.60 ~ 1.09 0.05 ~ 0.47
Heavy
(18.33) (5.35) (0.81) (0.22)
Temperate 2 25 4.4 ~ 4.5 4.35 ~ 10.15 0.79 ~ 1.26
Other AMGB Heavy
(4.45) (7.25) (1.03)
Notes:
1. Road types are defined as per VicRoads classifications.
2. Pavement types are defined as per HDM-4 classifications (Appendix A
3. Traffic group: low (AADT < 5000), medium (5000 < AADT < 15000), heavy (AADT > 15000)
4. Climate group: dry (Thornthwaite Index, TI, < 0), temperate (0 < TI < 50), wet (TI > 50)
5. Pavement age refers to the time since last construction or rehabilitation or reconstruction to year 200
6. Roughness and rutting progressions were the underlying rate of progression
7. Values in the brackets represent the average values.

4.2.2 Performance History


The continuity of the time series data for roughness, rutting and cracking for the Victorian road
sections varied from three to ten years. The approach and equipment used to measure
performance was not documented by the SRA supplying the data. From the data supplied it was
not possible to identify how the performance data was reported. For example, it was not known
whether a 1.2 m or a 2.0 m straightedge simulation was used to calculate rut depths. However, a
laser profiler was typically used in collecting reliable roughness and rutting data, while cracking
data were usually collected manually and their reliability was less than desirable because of the
subjective nature of this type of survey. This was confirmed by VicRoads, consequently, cracking
RD models were not calibrated for Victorian road sections.

Austroads 2008

— 12 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

The above time series performance data were used to estimate the underlying rate of pavement
deterioration for roughness and rutting. The rate of deterioration for roughness (NRM/year) and
rutting (mm/year) for each group of pavement sections are shown in Table 4.1.

4.2.3 Maintenance Intervention


Data sets from the Victorian road network had a good record of maintenance history. However,
maintenance history data were not used in the analysis because the estimation of the underlying
rate of deterioration (Appendix A) excluded data that were considered to be influenced by the
immediate effect of maintenance and rehabilitation treatments. The maintenance data were used
to determine other parameters such as pavement and surface age.

4.2.4 Calibration Results


Calibration coefficients for rutting structural deterioration (Krst) and roughness progression (Kgm)
were obtained using the approach explained in Section 3.2. The resulting calibration coefficients
are shown in Table 4.2 as mean values for each group of pavements. Included in Table 4.2 are
also the results from the previous calibration (Toole et al. 2004b) for comparison.

Table 4.2: Calibration results for pavement sections in Victoria

Road Pavement Traffic Climate Rutting coefficients (Krst) Roughness coefficients (Kgm)
type1 type2 group3 group4 Previous result5 Revised result6 Previous result5 Revised result6
Dry 2.63 2.13 1.25 0.53
Low
A SRGB Wet 3.00 2.13 0.30 0.26
Medium Temperate 1.38 0.66 0.30 0.28
Dry 0.75 1.40 2.25 1.34
Low
Temperate 1.53 1.80 0.41 0.44
B STGB
Dry 3.35 3.58 0.95 0.84
Medium
Temperate 3.00 1.58 0.30 0.19
Temperate 2.37 2.82 0.54 0.33
C STGB Low
Wet 3.35 1.83 0.55 0.33
AMAP Heavy Wet 1.00 1.10 0.25 0.46
AMGB Heavy Wet 0.50 0.28 0.75 2.00
AMSB Heavy Wet 0.50 0.49 1.00 0.86
M
Low Dry 0.50 0.48 0.35 0.15
STGB Medium Temperate 2.67 1.17 0.25 0.36
Heavy Temperate 2.17 1.72 0.42 0.39
Other AMGB Heavy Temperate 4.50 4.92 1.00 0.96
Notes:
1. Road types are defined as per VicRoads classifications.
2. Pavement types are defined as per HDM-4 classifications (Appendix A)
3. Traffic group: low (AADT < 5000), medium (5000 < AADT < 15000), heavy (AADT > 15000).
4. Climate group: dry (Thornthwaite Index, TI, < 0), temperate (0 < TI < 50), wet (TI > 50).
5. Previous calibration results obtained from Toole et al. (2004 b)
6. Revised calibration results obtained using the calibration approach described in Section 3.2.

Austroads 2008

— 13 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

As shown in Table 4.2, the pavement group mean calibration coefficient values from the revised
method and the previous method are on average comparable, although some 38% of these
comparisons have differences of at least 40% between their respective calibration coefficient
values. The current approach usually yielded lower coefficient values compared to the previous
method. When group specific results were compared, roughness coefficients from both analyses
appeared to be better matched.

However, the plots of RD model coefficients and rate of deterioration progressions for individual
pavement sections revealed a different outcome (Figure 4.1). Theoretically, roughness and rutting
K calibration coefficients (Krst and Kgm) should be related to the rate of deterioration. As shown in
Figure 4.1, the revised calibration produced a better relationship between the rate of deterioration
and the K factors compared to the previous calibration. Also, better correlations (r2 of 0.67 for
rutting and 0.75 for roughness) were observed in the revised results compared to the results (r2 of
0.39 for rutting and 0.39 for roughness) from the previous study.

8
Revised calibration
7 Previous calibration
Trend line (revised calibration) R2 = 0.669
6 Trendline (previous calibration)

R2 = 0.3881
5
Krst values

0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8

Rutting progression (mm/year)

Figure 4.1: Relationship between rate of deterioration and RD model coefficients for the Victorian road network

4.3 Tasmanian Road Network


4.3.1 Road Sections Considered for RD Model Calibration
The inventory and performance history of 80 pavement sections from the Department of
Infrastructure, Energy and Resources (DIER), Tasmania, road network were available to ARRB
from a previous RD model calibration project (Toole et al. 2004a) for Tasmania. These pavement
sections covered both asphalt and spray seal surfaced pavements in the wet and temperate zones.
Traffic on these roads varied between 76 to 15000 AADT.

Austroads 2008

— 14 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

The same set of pavement sections that were calibrated for the HDM-4 RD models for roughness,
rutting and cracking in the previous study (Toole et al. 2004a) were considered for re-calibration
using the calibration approach outlined in Section 3.2. Estimation of the underlying rate of
deterioration showed that a number of sections experienced improved conditions (i.e. negative
deterioration of either roughness or rutting conditions). Of the 80 sections, 58 showed positive
deterioration and these 58 sections were considered for the re-calibration exercise.

The selected pavement sections were grouped according to road type, pavement type, traffic
group and climate zone. As per DIER practice, the sections were grouped into five road types, i.e.
type one to five, while HDM-4 classifications were used for pavement types. The same traffic and
climate criteria as used for the Victorian data set were also used for the Tasmanian data set.

As shown in Table 4.3, the selected pavement sections fell into two traffic groups, i.e. low and
medium, and two climate groups, i.e. wet and temperate. Also, most of the test sections were
surfaced with spray seal (STGB), while a couple of the sections were asphalt surfaced (AMGB).

4.3.2 Performance History


Roughness, rutting and cracking time series deterioration data were available for the Tasmanian
road network. This data included five to twelve measurements of roughness and four to nine
measurements of rutting. Also included were two to nine measurements of cracking data, which
showed that there was little cracking over the last ten years. Therefore, as for the Victorian data
set, only roughness and rutting data were used for the re-calibration.

4.3.3 Maintenance Intervention


Data sets from the Tasmanian road network had one to five records of maintenance treatments.
However, for the same reasons as for the Victorian analysis, maintenance data were not used in
the analysis.

Austroads 2008

— 15 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

Table 4.3: Characteristics of the Tasmanian pavement sections considered for analysis
Road Pavement Traffic Climate No. of Pavement SNP Roughness Rutting
type1 type2 group3 group4 sections age5 (average)7 progression progression
(average (average)7 (average)7
years)7 (NRM/year)6 (mm/year)6
Temperate 2 15 ~ 25 (20) 3.18 ~ 3.39 0.80 ~ 1.07 0.20 ~ 0.77
AMGB Medium
(3.29) (0.93) (0.49)
Temperate 3 20 ~ 26 2.46 ~ 3.55 0.06 ~ 1.86 0.04 ~ 0.22
Low (22.33) (2.94) (1.21) (0.13)
1 Wet 1 21 3.55 0.63 0.18
STGB Temperate 2 27 ~ 38 (32.5) 3.76 ~ 4.77 0.03 ~ 0.42 0.08 ~ 0.15
(4.27) (0.22) (0.12)
Medium
Wet 3 12 ~ 16 (14) 2.13 ~ 2.48 0.13 ~ 0.47 0.00 ~ 0.07
(2.25) (0.26) (0.05)
AMGB Low Temperate 1 23 3.46 3.26 0.08
Temperate 3 19 ~ 28 3.23 ~ 3.74 0.34 ~ 1.22 0.01 ~ 0.86
(23.33) (3.48) (0.74) (0.29)
Low
2 Wet 5 11 ~ 37 (21.6) 2.75 ~ 4.60 0.02 ~ 1.64 0.13 ~ 1.48
STGB
(3.51) (0.42) (0.91)
Temperate 2 32 ~ 43 (37.5) 3.20 ~ 3.45 1.02 ~ 1.51 0.07 ~ 0.52
Medium
(3.33) (1.26) (0.30)
Temperate 4 13 ~ 26 (17.5) 2.46 ~ 3.54 1.11 ~ 10.10 0.03 ~ 0.37
(3.07) (4.42) (0.23)
Low
3 STGB Wet 7 10 ~ 36 (19) 2.10 ~ 2.68 0.13 ~ 1.46 0.00 ~ 0.53
(2.47) (0.63) (0.25)
Medium Wet 1 29 3.95 0.62 0.06
Temperate 4 19 ~ 23 (21) 2.97 ~ 4.40 0.26 ~ 1.46 0.05 ~ 0.73
(3.52) (0.68) (0.28)
4 STGB Low
Wet 8 10 ~ 42 2.48 ~ 2.72 0.19 ~ 1.46 0.01 ~ 0.77
(19.25) (2.51) (0.78) (0.33)
AMGB Low Temperate 1 10 2.48 1.21 0.18
Temperate 3 21 ~ 24 2.39 ~ 2.50 0.25 ~ 0.75 0.05 ~ 0.36
5 (29.33) (2.46) (0.45) (0.18)
STGB Low
Wet 8 13 ~ 39 2.48 ~ 5.37 0.14 ~ 0.89 0.15 ~ 0.75
(21.38) (2.91) (0.52) (0.31)
Notes:
1. Road types are defined as per DIER classifications.
2. Pavement types are defined as per HDM-4 classifications (Appendix A).
3. Traffic group: low (AADT < 5000), medium (5000 < AADT < 15000), heavy (AADT > 15000).
4. Climate group: dry (Thornthwaite Index, TI, < 0), temperate (0 < TI < 50), wet (TI > 50).
5. Pavement age refers to the time since last construction or rehabilitation or reconstruction to year 2005.
6. Roughness and rutting progressions were the underlying rate of progressions.
7. Values in the brackets represent the average values.

4.3.4 Calibration Results


Calibration coefficients for rutting structural deterioration (Krst) and roughness progression (Kgm)
were obtained using the approach explained in Section 3.2. The calibration coefficients are shown
in Table 4.4 as mean values for each group of pavements. Included in Table 4.4 are also the
results from the previous calibration (Toole et al. 2004a) for comparison.

Austroads 2008

— 16 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

The group mean calibration coefficients from the revised method are generally lower compared
with those of the previous method. It should be noted that, unlike the previous method, the
calibration results from the revised method were calculated so that the rate of predicted
deterioration more closely matched the observed rate of deterioration.

Table 4.4: Calibration results for the pavement sections in Tasmania


Road Pavement Traffic Climate Rutting coefficients (Krst) Roughness coefficients (Kgm)
type1 type2 group3 group4 Previous result5 Revised result6 Previous result5 Revised result6

AMGB Medium Temperate 0.75 1.86 0.40 0.20


Temperate 1.00 0.38 0.43 0.24
Low
1 Wet 1.00 0.48 0.30 0.14
STGB
Temperate 1.25 0.50 0.20 0.15
Medium
Wet 0.67 0.17 0.13 0.13
AMGB Low Temperate 1.00 0.34 0.40 0.82
Temperate 0.90 1.22 0.63 0.17
2 Low
STGB Wet 0.84 3.03 0.18 0.18
Medium Temperate 1.50 0.90 0.40 0.23
Temperate 0.85 0.83 0.45 0.71
Low
3 STGB Wet 0.93 0.91 0.21 0.22
Medium Wet 0.50 0.20 0.30 0.21
Temperate 1.80 1.39 0.43 0.17
4 STGB Low
Wet 0.88 1.32 0.26 0.26
AMGB Low Temperate 1.00 0.60 0.30 0.27
5 Temperate 1.17 0.60 0.13 0.17
STGB Low
Wet 1.00 1.55 0.18 0.17
Notes:
1. Road types are defined as per DIER classifications.
2. Pavement types are defined as per HDM-4 classifications (Appendix A).
3. Traffic group: low (AADT < 5000), medium (5000 < AADT < 15000), heavy (AADT > 15000).
4. Climate group: dry (Thornthwaite Index, TI, < 0), temperate (0 < TI < 50), wet (TI > 50).
5. Previous calibration results obtained from Toole et al. (2004 b).
6. Revised calibration results obtained using the calibration approach described in Section 3.2.

Figure 4.2 shows the relationship between roughness and rutting progressions and Krst and Kgm
calibration coefficients. As expected, the results from the revised method indicate that pavement
sections with a higher rate of deterioration have higher roughness and rutting calibration
coefficients (Krst and Kgm), and vice versa. The revised method produced K values that were highly
correlated (r2 > 0.82) with the estimated deterioration.

Austroads 2008

— 17 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

7
Revised calibration

6 Previous calibration
Trend line (revised calibration)
R2 = 0.8284
5 Trendline (previous calibration)
Krst values

1 R2 = 0.0007

0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6

Rutting progression (mm/year)

1.6
Revised calibration
2
1.4 Previous calibration R = 0.8498
Trend line (revised calibration)
1.2 Trendline (previous calibration)

1
Kgm values

0.8
2
R = 0.1263
0.6

0.4

0.2

0
0.0 2.0 4.0 6.0 8.0 10.0 12.0

Roughness progression (NRM/year)

Figure 4.2: Relationship between rates of deterioration and RD model coefficients for the Tasmanian road network

Austroads 2008

— 18 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

4.4 South Australian Road Network


4.4.1 Road Sections for RD Model Calibration
The inventory and time series deterioration data of a large number of pavement sections from the
Department for Transport, Energy and Infrastructure (DTEI), South Australia, were available to
ARRB from another project (Martin et al. 2006). The available data contained more than 129000
pavement sections, each 100 m in length. The pavement sections covered both asphalt and spray
seal surfaced pavements in the dry and temperate zones.

All sections were initially considered for the calibration of HDM-4 RD models for South Australian
conditions. Estimation of the underlying rate of deterioration on each section showed that a large
number of sections showed improved conditions (i.e. negative deterioration of either roughness or
rutting or cracking). All sections with negative deterioration were excluded from the analysis,
resulting in 14765 sections with positive deterioration concurrently for cracking, rutting and
roughness progression being considered for calibration.

Calibration of the HDM-4 RD models was iterative and time consuming. On average, it took 20 to
30 minutes to calibrate one pavement section (on a Pentium 4 desk top computer) for the cracking,
rutting and roughness models. It was therefore too time consuming to analyse all the 14765
sections in this study. However, with improved computing capacity and an efficient programming
algorithm, it will be possible to analyse a large number of pavement sections.

Typically, candidate pavement sections were calibrated using individual pavement section
parameters with the results placed into groups based on a set of criteria. The results were reported
as a mean of each group of pavements. The 14765 pavement sections were grouped on the basis
of road type, pavement type, traffic loading and climate conditions prior to calibration.

The sections were grouped into four road types, i.e. national highways (NH), urban arterial (UA),
urban local (UL) and rural arterial (RA). HDM-4 classifications were used for pavement types and
the same traffic and climate criteria as used for Victorian and Tasmanian data sets. This resulted in
51 groups of pavements with the characteristics of these pavement sections shown in Table 4.5.
As seen in Table 4.5, the pavement sections fall into two climate conditions, i.e. dry and temperate,
while the majority were in the dry condition. In terms of traffic loading, the majority of the sections
carry low traffic (< 5000 AADT), while some sections also carry medium to heavy traffic.

Mean values for HDM-4 RD model coefficients were estimated or assumed for each group of
pavements. These mean values were used for the calibration of the RD models. It should be noted
that the results of this analysis are applicable to the pavements whose characteristics are similar to
the mean values used in this analysis.

Austroads 2008

— 19 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

Table 4.5: Characteristics of the South Australian pavement sections considered for analysis
Road Pavement Traffic Climate No. of Pavement SNP (avg)# Cracking Rutting Roughness
type1 type2 group3 group4 sections age5 (average prog. (avg)# prog. prog. (avg)#
years)# (% area /yr)6 (avg)# (NRM/yr)6
(mm/yr)6
Dry 2 10 ~ 12 (11) 3.77 ~ 3.90 1.23 ~ 3.21 0.09 ~ 0.60 2.12 ~ 9.26
AMAP Heavy
(3.84) (2.22) (0.35) (5.69)
Dry 766 1 ~ 49 (20) 2.80 ~ 3.84 0.0 ~ 10.51 0.02 ~ 2.55 0.01 ~ 16.87
Low
(3.64) (0.96) (0.52) (2.37)
Dry 1254 1 ~ 47 (25.5) 2.98 ~ 3.90 0 ~ 10.18 0.02 ~ 2.55 0 ~ 19.57
(3.69) (0.73) (0.54) (2.12)
Medium
Temperate 7 4 ~ 6 (4.5) 3.76 ~ 3.77 0.46 ~ 1.80 0.13 ~ 1.35 1.96 ~ 10.06
AMGB
(3.77) (1.12) (0.78) (5.98)
Dry 408 2 ~ 66 (33) 3.16 ~ 4.0 0 ~ 9.18 (1.59) 0.01 ~ 3.75 0 ~ 18.12
(3.62) (0.60) (3.52)
Heavy
Temperate 107 1 ~ 33 (31) 3.66 ~ 3.91 0 ~ 3.11 (0.08) 0.01 ~ 1.34 0.04 ~ 10.85
(3.68) (0.41) (1.33)
Dry 19 4 ~ 6 (5.5) 2.99 ~ 3.77 0 ~ 3.86 (0.59) 0.10 ~ 2.15 0.33 ~ 8.58
Low
(3.36) (0.98) (4.16)
Dry 33 3 ~ 11 (7.8) 3.61 ~ 3.82 0.52 ~ 8.68 0.06 ~ 1.32 0.33 ~ 12.68
Medium
(3.77) (2.14) (0.65) (4.22)
AMSB
Dry 52 9 ~ 12 (10.5) 3.69 ~ 3.95 0 ~ 6.45 (2.24) 0.06 ~ 1.48 0.40 ~ 10.53
NH
(3.84) (0.49) (2.56)
Heavy
Temperate 6 (5) 3.75 ~ 3.79 0.05 ~ 5.45 0.07 ~ 1.39 0.08 ~ 3.24
(3.77) (1.45) (0.95) (1.30)
Dry 4 (1) (3.68) 0.53 ~ 1.40 0.55 ~ 2.23 1.69 ~ 3.95
STAB Low
(1.0) (1.60) (3.03)
Dry 5014 1 ~ 47 (28) 2.61 ~ 3.84 0 ~ 12.55 0 ~ 4.19 0 ~ 23.61
Low
(3.42) (0.53) (0.65) (2.14)
Dry 396 1 ~ 47 (26) 2.70 ~ 3.77 0 ~ 8.44 (0.91) 0.01 ~ 2.75 0.07 ~ 18.91
STGB Medium
(3.42) (0.67) (2.57)
Dry 14 3 ~ 47 (32) 3.49 ~ 3.71 0.23 ~ 7.55 0.24 ~ 1.65 0.19 ~ 13.79
Heavy
(3.59) (2.52) (0.69) (4.89)
Dry 140 3 ~ 10 (8.2) 2.94 ~ 3.74 0 ~ 6.63 (0.67) 0.01 ~ 3.35 0.25 ~ 10.43
Low
(3.59) (0.57) (3.08)
Dry 3 (3) (3.77) 0.03 ~ 2.66 0.77 ~ 0.91 1.51 ~ 5.76
STSB Medium
(1.10) (0.83) (3.79)
Dry 73 (10) (3.83) 0.24 ~ 6.38 0.05 ~ 1.02 0.19 ~ 8.01
Heavy
(1.73) (0.37) (1.51)

Austroads 2008

— 20 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

Road Pavement Traffic Climate No. of Pavement SNP (avg)# Cracking Rutting Roughness
type1 type2 group3 group4 sections age5 (average prog. (avg)# prog. prog. (avg)#
years)# (% area /yr)6 (avg)# (NRM/yr)6
(mm/yr)6
Dry 5 25 ~ 59 (32) 3.58 ~ 3.77 1.26 ~ 4.46 1.05 ~ 2.35 0.99 ~ 17.95
Medium
(3.72) (3.19) (1.67) (6.54)
AMAB
Temperate 1 11 3.49 1.07 2.50 3.51
Heavy Dry 1 33 3.71 4.14 1.55 11.66
Dry 5 8 ~ 10 (9.2) 3.72 ~ 3.85 1.38 ~ 3.56 0.18 ~ 1.95 0.55 ~ 3.39
AMAP Heavy
(3.80) (2.73) (0.93) (1.87)
Dry 212 5 ~ 73 (33.3) 2.57 ~ 3.83 0 ~ 8.51 (1.7) 0 ~ 3.80 0.02 ~ 15.67
Low
(3.09) (1.39) (3.4)
Temperate 54 1 ~ 70 (37) 2.19 ~ 3.26 0.06 ~ 13.31 0 ~ 3 (0.77) 0.30 ~ 12.30
(2.69) (4.51) (3.15)
Dry 1966 1 ~ 80 (28) 2.48 ~ 3.89 0 ~ 13.44 0 ~ 4.90 0 ~ 20.79
Medium
(3.32) (2.23) (1.39) (3.61)
AMGB
Temperate 496 1 ~ 70 (29) 2.90 ~ 3.74 0 ~ 10.74 0 ~ 4.20 0.07 ~ 18.87
(3.31) (2.10) (1.33) (3.62)
Dry 1266 1 ~ 80 (29.2) 3.01 ~ 4.03 0 ~ 13.09 0 ~ 3.75 0 ~ 21.83
Heavy
(3.51) (2.80) (1.27) (4.08)
Temperate 349 2 ~ 77 (28.6) 2.40 ~ 3.82 0 ~ 9.53 (2.30) 0 ~ 3.95 0 ~ 19.24
(3.40) (1.38) (3.82)
Dry 60 5 ~ 11 (7.37) 3.38 ~ 3.78 0 ~ 6.84 (1.78) 0.06 ~ 3.85 0.01 ~ 13.89
Medium
(3.61) (1.51) (2.91)
UA Temperate 6 4 ~ 11 (5.3) 3.12 ~ 3.74 0.32 ~ 3.39 0.62 ~ 3.75 0.13 ~ 3.77
(3.50) (1.82) (2.15) (1.56)
AMSB
Dry 23 7 ~ 12 (8.6) 3.51 ~ 3.83 0 ~ 3.86 (1.21) 0.13 ~ 4.60 0.13 ~ 6.35
Heavy
(3.72) (1.62) (1.90)
Temperate 56 4 ~ 11 (8.3) 2.72 ~ 3.78 0 ~ 5.03 (0.61) 0.03 ~ 2.90 0.07 ~ 14.74
(3.31) (1.74) (3.34)
Dry 56 4 ~ 55 (33.3) 2.76 ~ 3.28 0 ~ 4.98 (0.74) 0 ~ 1.3 0.42 ~ 16.03
Low
(2.99) (0.30) (2.71)
Temperate 49 37 ~ 72 (39) 2.38 ~ 3.11 0 ~ 8.66 (3.4) 0 ~ 1.72 0.50 ~ 7.28
(2.64) (0.41) (2.88)
Dry 108 1 ~ 70 (22.7) 3.03 ~ 3.98 0 ~ 9.76 (1.48) 0 ~ 3.15 0.07 ~ 16.07
Medium
(3.47) (1.16) (3.97)
STGB
Temperate 114 2 ~ 38 (31) 2.47 ~ 3.42 0 ~ 5.96 (0.49) 0 ~ 3.55 0 ~ 10.95
(3.14) (0.70) (2.71)
Dry 233 1 ~ 67 (22) 3.19 ~ 3.93 0 ~ 11.59 0 ~ 3.80 0 ~ 19.57
Heavy
(3.58) (1.59) (1.09) (3.78)
Temperate 3 5 ~ 31 (13.67) 2.74 ~ 3.53 0 ~ 3.60 (1.57) 0.92 ~ 3.20 0.27 ~ 8.93
(3.27) (1.96) (4.59)
Dry 11 4 ~ 10 (8.9) 3.47 ~ 3.66 1.52 ~ 5.87 0.26 ~ 2.50 0.11 ~ 9.60
Medium
(3.58) (3.06) (1.34) (3.63)
STSB
Dry 11 (10) 3.86 ~ 3.92 0.12 ~ 3.10 0.53 ~ 2.45 1.17 ~ 14.55
Heavy
(3.88) (1.12) (1.70) (5.14)

Austroads 2008

— 21 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

Road Pavement Traffic Climate No. of Pavement SNP (avg)# Cracking Rutting Roughness
type1 type2 group3 group4 sections age5 (average prog. (avg)# prog. prog. (avg)#
years)# (% area /yr)6 (avg)# (NRM/yr)6
(mm/yr)6
Low Dry 1 72 2.87 1.98 0.04 0.56
AMGB Dry 6 55 ~ 72 (69) 2.86 ~ 2.96 1 ~ 7.23 (2.86) 0.05 ~ 1.85 1.05 ~ 6.25
Medium
UL (2.88) (0.40) (2.65)
Dry 11 (35) (2.27) 0 ~ 1.41 (0.16) 0.77 ~ 2.05 1.42 ~ 7.97
STGB Low
(1.33) (5.06)
Dry 189 6 ~ 75 (38.4) 2.16 ~ 3.40 0 ~ 11.12 0 ~ 2.35 0 ~ 8.34
Low
(2.88) (1.47) (0.42) (2.42)
Temperate 15 22 ~ 24 (22.4) 2.74 ~ 2.75 0.37 ~ 9.41 0.08 ~ 1.25 0.38 ~ 4.17
AMGB (2.75) (3.41) (0.49) (2.06)
Dry 33 5 ~ 47 (26.2) 2.37 ~ 3.52 0 ~ 1.67 (0.37) 0.01 ~ 1.36 0.15 ~ 5.87
Medium
(3.09) (0.42) (1.79)
Temperate 1 46 3.14 3.79 1.85 1.86
Dry 890 4 ~ 75 (39.2) 2.16 ~ 3.46 0 ~ 10.15 0 ~ 2.05 0 ~ 10.69
Low
RA (2.95) (0.47) (0.43) (2.19)
Temperate 107 24 ~ 50 (41) 2.49 ~ 3.10 0 ~ 9.36 (2.27) 0.01 ~ 2.35 0.03 ~ 8.57
(2.75) (0.38) (2.2)
STGB
Dry 60 12 ~ 47 (40) 2.37 ~ 3.27 0.53 ~ 1.60 0.01 ~ 1.27 0.35 ~ 9.07
Medium
(3.03) (1.11) (0.47) (2.52)
Temperate 61 2 ~ 46 (30.8) 3.14 ~ 3.50 1.67 ~ 3.96 0.01 ~ 1.67 0.17 ~ 9.04
(3.27) (2.57) (0.58) (2.45)
Temperate 7 (5) (3.48) 2.13 ~ 2.23 0.16 ~ 0.52 0.22 ~ 4.83
STSB Moderate
(2.18) (0.36) (2.21)
Notes:
1. Road types are defined as per DTEI classifications.
2. Pavement types are defined as per HDM-4 classifications (Appendix A).
3. Traffic group: low (AADT < 5,000), medium (5,000 < AADT < 15,000), heavy (AADT > 15,000).
4. Climate group: dry (Thornthwaite Index, TI, < 0), temperate (0 < TI < 50), wet (TI > 50).
5. Pavement age refers to the time since last construction or rehabilitation or reconstruction to year 2005.
6. Roughness and rutting progressions were the underlying rate of progressions.
7. Values in the brackets represent the average values.

4.4.2 Performance History


Roughness, rutting and cracking time series deterioration data were available for the DTEI road
network. This data included cracking data from 1999 to 2004 and roughness and rutting data from
1994 to 2004. DTEI advised that the quality of the cracking data was reasonable. Therefore, in
addition to roughness and rutting model calibrations, cracking models were also calibrated for the
DTEI road network.

4.4.3 Maintenance Intervention


Maintenance treatment data were not available. However, as explained before, maintenance
historical data were not necessary for this analysis. The method of estimating the underlying rate of
deterioration (Appendix A) adopted in this analysis excluded condition data that were affected by
maintenance intervention.

4.4.4 Calibration Results


Calibration results for cracking progression (Kcpa), rutting structural deterioration coefficients (Krst)
and roughness progression coefficients (Kgm) were estimated using the approach explained in
Section 3.2. The results are recorded in Table 4.6.

Austroads 2008

— 22 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

As seen in Table 4.6, the analysis yielded smaller K factors (Kcpa and Kgm) for the cracking and
roughness models (less than the default 1.0 values). The rutting K factors (Krst), on the other hand,
were generally higher (greater than 1.0). As noted previously, these K factors were not estimated
using the individual pavement section parameters, they were based on the calculated mean values
for the 51 groups of pavements.

Therefore, the K values shown in Table 4.6 are applicable only to the pavement sections whose
characteristics are similar to the mean values shown in Table 4.5. Because the mean values were
used for the calibration, it was expected that the relationship between the K factors and the rates of
deterioration would be low. However, as shown in Figure 4.3, the regression coefficient for rutting
(r2 = 0.89) was very high, very low for cracking (r2 = 0.0) and reasonable for roughness (r2 = 0.45).

Again with greater computing capacity and a sophisticated algorithm, a large number of individual
pavement sections could be calibrated. The results from these individual pavement sections can
then be grouped to find mean K values, similar to those for the Victorian and Tasmanian road
networks.

Austroads 2008

— 23 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

Table 4.6: Calibration results for the pavement groups in South Australia
RD model
Roughness
Road type1 Pavement type2 Traffic group3 Climate group4 coefficients5 Rutting (Krst)
(Kgm)
Cracking (Kcpa)
AMAP Heavy Dry 0.44 1.38 0.94
Low Dry 0.14
Medium Dry 0.14
AMGB Temperate 0.44
Heavy Dry 0.18
Temperate 0.14
Dry 0.14
Dry 0.22
Dry 0.20
Temperate 0.24
Dry 3.27
Dry 0.15
Dry 0.15
Dry 0.15
Dry 0.24
Dry 0.54
Dry 0.16
Dry 0.26
Temperate 0.32
Dry 0.38
Dry 1.30
Dry 0.18
Temperate 0.34
Dry 0.20
Temperate 0.22
Dry 0.24
Temperate 0.22
Dry 0.28
Temperate 0.44
Dry 0.17
Temperate 0.24
Dry 0.69
Temperate 0.68
Dry 0.68
Temperate 0.69
Dry 0.68
Temperate 0.14
Dry 0.32
Dry 0.14
Dry 0.16
Dry 0.24

Austroads 2008

— 24 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

RD model
Roughness
Road type1 Pavement type2 Traffic group3 Climate group4 coefficients5 Rutting (Krst)
(Kgm)
Cracking (Kcpa)
Dry 0.14
Dry 0.14
Temperate 0.31
Dry 0.14
Temperate 0.28
Dry 0.68
Temperate 0.14
Dry 0.14
Temperate 0.69
Temperate 0.44
Notes:
1. Road types are defined as per DTEI classifications.
2. Pavement types are defined as per HDM-4 classifications (Appendix A).
3. Traffic group: low (AADT < 5000), medium (5000 < AADT < 15000), heavy (AADT > 15000).
4. Climate group: dry (Thornthwaite Index, TI, < 0), temperate (0 < TI < 50), wet (TI > 50).
5. Previous calibration results obtained from Toole et al. (2004 b).
6. Revised calibration results obtained using the calibration approach described in Section 3.2.

12
Revised calibration
Trend line (revised calibration)
10
R 2 = 0.888

8
K rst values

0
0.0 1.0 2.0 3.0

Rutting progression (mm/year)

Austroads 2008

— 25 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

4.0
Revised calibration
Trend line (revised calibration)

3.0
K cpa values

2.0

1.0
R 2 = 0.0005

0.0
0.0 1.0 2.0 3.0 4.0 5.0

Cracking progression (% area/year)

1.8
Revis ed calibration R 2 = 0.452
1.6 Outlier
Trend line (revis ed calibration)

1.4

1.2
K gm values

0.8

0.6

0.4

0.2

0
0.0 0.1 0.2 0.3 0.4 0.5

Roughness progression (NRM/year)

Figure 4.3: Relationship between rate of deterioration and RD model coefficients for South Australian road network

Austroads 2008

— 26 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

4.5 Queensland Road Network


4.5.1 Road Sections for RD Model Calibration
The inventory and time series roughness condition data collected from a large network in
Queensland were made available for the study by the Queensland Department of Main Roads
(QDMR). The data set held more than 28,600 road sections of 1 km in length with roughness,
pavement characteristics, traffic, climate and inventory data up to 2005.

As a previous analysis of the Queensland road network was limited to rural roads with AADT less
than 5000 vehicles per day (Martin and Hoque 2006), the original number of records was reduced
to just over 7300 for this calibration. After the final selection process based on the estimated
underlying rate of deterioration, the number of sections with positive deterioration available for the
RD model calibration was 5327. As for the South Australian data, it was not possible to calibrate
each section individually because of processing time; calibration of a group of sections was
undertaken with an expected potential reduction in the accuracy of the results.

Four categories of road type were identified in the Queensland road network. These were National
Highway (NH), State Strategic (SS), Regional Roads (RR) and District Roads (DR). Pavement
types were those classified by HDM-4, namely AMGB, STGB, STSB (as defined in Table A 2 of
Appendix A), and sub-grade soil was designated as either reactive or non-reactive. The last
parameter used for the grouping was climatic conditions. This was based on the ranges of
Thornthwaite moisture indices which determined whether the site climate was dry, temperate or
hot. Soil reactivity was used instead of traffic for grouping the road sections as there was only one
group of traffic considered for the network analysis.

For each group of sections the mean value, range (minimum and maximum) and sample size for
the required parameters were calculated and presented in a table as summary information. Only
mean values were used as input parameters in the calibration of the HDM-4 RD models. With
these grouping criteria, 57 groups were formed and each was treated as a single pavement section
in the calibration analysis. The group sample size varied significantly, from a minimum of one to a
maximum of 606 individual pavement sections, with an overall average size of 92 sections.

4.5.2 Performance History


Only roughness was available for the analysis of the Queensland road network. A roughness
history of up to 18 years (from 1987 to 2005) provided a large sample size for the calibration of the
roughness RD model.

4.5.3 Maintenance Intervention


The data set also contained the annual maintenance sealing cost and the treatment age for each
section. However, maintenance history was not needed for this analysis as the method of
estimating the underlying rates of deterioration (Appendix B) excluded deterioration data that were
assumed to be affected by maintenance interventions.

Austroads 2008

— 27 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

Table 4.7: Characteristics of the Queensland pavement sections considered for analysis
Road type1 Pavement Soil reactivity3 Climate No. of SNC Pavement AADT (avg)#
type2 group4 sections (average)# age
(average
years)#
3.14 ~ 9 ~ 31 545 ~ 3497
Dry 6
3.54 (3.45) (13.2) (2916.5)
3.24 ~ 13 ~ 30 2523 ~ 4612
Non-reactive Temperate 17
3.74 (3.63) (17.6) (3176.6)
3.24 ~ 4 ~ 22 2523 ~ 4241
Wet 9
3.48 (3.33) (12.7) (2947.7)
AMGB
Dry 1 3.14 31 545
2.87 ~ 11 ~ 30 310 ~ 4612
Reactive Temperate 7
3.35 (2.94) (17.7) (924.6)
2.96 ~ 7 ~ 18 537 ~ 2523
Wet 4
3.24 (3.17) (14.8) (2026.5)
5 ~ 38 310 ~ 4657
Dry 188 3.29
(14.5) (1454.3)
2.87 ~ 4 ~ 37 310 ~ 4817
Non-reactive Temperate 254
3.74 (3.3) (14.8) (2175.5)
NH
3.24 ~ 3 ~ 24 2045 ~ 4828
Wet 97
3.48 (3.34) (14.6) (2820.1)
STGB
2.78 ~ 3 ~ 42 242 ~ 4255
Dry 225
3.72 (3.12) (14.3) (746.3)
2.8 ~ 3.7 6 ~ 29 242 ~ 4257
Reactive Temperate 86
(3) (12.8) (806.1)
2.8 ~ 3.7 6 ~ 38 242 ~ 4036
Wet 128
(2.95) (14.3) (612.3)
2.87 ~ 4 ~ 17 310 ~ 4173
Temperate 7
3.48 (3.17) (8.9) (2248.7)
Non-reactive
3.31 ~ 2572 ~ 4334
Wet 3 3 ~ 10 (6)
3.32 (3.32) (3746.7)
STSB
Dry 1 3.01 9 772
Reactive
2.87 ~ 5 ~ 20 310 ~ 4612
Temperate 12
3.51 (3.16) (13.4) (2168.6)
3.1 ~ 3.1 14 ~ 28 1523 ~ 1523
AMGB Non-reactive Wet 5
(3.1) (23.2) (1523)
2.64 ~ 4 ~ 36 167 ~ 4451
Dry 606
3.65 (2.98) (15.7) (538.1)
2.81 ~ 4 ~ 38 214 ~ 3348
Non-reactive Temperate 107
3.67 (3.21) (16.4) (1502.1)
2.9 ~ 3.46 4 ~ 40 532 ~ 1652
SS Wet 42
(3.19) (15.8) (1320.3)
STGB
2.64 ~ 3.6 4 ~ 58 167 ~ 2321
Dry 425
(2.98) (13.6) (497.2)
4 ~ 38 167 ~ 1652
Reactive Temperate 27 3.05
(13.3) (731.1)
2.85 ~ 4 ~ 33 301 ~ 1652
Wet 29
3.46 (3.03) (17.5) (654.1)
2.75 ~ 217 ~ 1235
SS STSB Non-reactive Dry 21 3 ~ 9 (6)
3.47 (3.26) (954.6)

Austroads 2008

— 28 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

Road type1 Pavement Soil reactivity3 Climate No. of SNC Pavement AADT (avg)#
type2 group4 sections (average)# age
(average
years)#
Temperate 1 3.37 7 3326
2.93 ~ 751 ~ 1523
Wet 6 5 ~ 17 (9)
3.37 (3.04) (1086.8)
2.88 ~ 4 ~ 25 340 ~ 1213
Dry 10
3.47 (3.07) (9.3) (787.7)
Reactive 2.88 ~ 4 ~ 25 399 ~ 518
Temperate 6
3.03 (2.9) (21.2) (418.8)
Wet 1 3.04 7 672
2.96 ~ 3.7 8 ~ 45 348 ~ 3303
Non-reactive Dry 6
(3.37) (32) (2068.8)
AMGB
7 ~ 23 76 ~ 3742
Reactive Dry 5 2.68
(10.4) (809.2)
2.41 ~ 4 ~ 61 14 ~ 3313
Dry 260
3.63 (2.76) (17.6) (543.3)
2.37 ~ 4 ~ 31 14 ~ 3916
Non-reactive Temperate 181
3.71 (2.89) (13.8) (734.6)
RR
2.45 ~ 4 ~ 46 37 ~ 4237
Wet 419
3.38 (2.95) (13.6) (541)
STGB
2.41 ~ 3 ~ 40 14 ~ 4118
Dry 302
3.63 (2.74) (11.2) (309.1)
2.36 ~ 4 ~ 36 24 ~ 1409
Reactive Temperate 126
3.47 (2.94) (9.9) (265.7)
2.29 ~ 4 ~ 38 37 ~ 1077
Wet 132
3.32 (2.75) (13.8) (194.4)
2.66 ~ 4 ~ 15 113 ~ 689
Non-reactive Dry 18
3.37 (3.2) (7.7) (477.8)
2.97 ~ 4 ~ 12 730 ~ 3072
Temperate 29
3.36 (3.32) (9.7) (1070.1)
RR STSB 2.81 ~ 4 ~ 15 183 ~ 2260
Wet 7
3.51 (3.29) (8.6) (1661.7)
2.71 ~ 9 ~ 11 160 ~ 730
Reactive Temperate 17
3.36 (3.23) (10.6) (596.6)
Wet 1 2.81 3 226

Austroads 2008

— 29 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

Road type1 Pavement Soil reactivity3 Climate No. of SNC Pavement AADT (avg)#
type2 group4 sections (average)# age
(average
years)#
2.51 ~ 5 ~ 49 188 ~ 3018
Non-reactive Dry 16
3.37 (2.65) (24.1) (595.5)
Wet 1 3.30 44 3971
AMGB
Reactive Dry 1 2.87 30 66
2.5 ~ 2.5 0 ~ 41
Wet 6 30 ~ 30 (30)
(2.5) (13.7)
2 ~ 3.63 3 ~ 58 36 ~ 4923
Non-reactive Dry 469
(2.81) (16.3) (778.8)
2.3 ~ 3.5 4 ~ 43 26 ~ 4807
Temperate 122
(2.84) (12.5) (673.9)
2.3 ~ 3.55 4 ~ 53 29 ~ 3971
Wet 233
(2.85) (17.2) (330.6)
STGB
2.28 ~ 5 ~ 38 29 ~ 3275
DR Reactive Dry 196
3.58 (2.77) (14.4) (468.2)
2.34 ~ 2 ~ 61 29 ~ 348
Temperate 101
2.87 (2.6) (14.3) (52.4)
2.3 ~ 3.37 5 ~ 41 29 ~ 2932
Wet 224
(2.69) (21) (179.3)
2.66 ~ 400 ~ 2628
Non-reactive Dry 8 1 ~ 14 (6)
3.12 (2.91) (1145.3)
Temperate 1 2.81 4 744
2.78 ~ 4 ~ 18 72 ~ 1500
STSB Wet 8
3.09 (2.99) (9.3) (666.8)
1567 ~ 1567
Reactive Dry 4 0 ~ 0 (0) 9 ~ 9 (9)
(1567)
2.81 ~ 5 ~ 16 72 ~ 102
Wet 13
3.02 (2.82) (6.7) (99.7)
Notes:
1. Road types are defined as per QDMR classifications.
2. Pavement types are defined as per HDM-4 classifications (Appendix A).
3. Soil reactivity is as per QDMR classifications.
4. Climate group: dry (Thornthwaite Index, TI, < 0), temperate (0 < TI < 50), wet (TI > 50). # Values in the brackets represent the average values.

4.5.4 Calibration Results


As noted earlier, only time series roughness deterioration data were available for the calibration
exercise. As the HDM-4 roughness progression model is a function of cumulative changes in
roughness due to structural rutting and cracking, the HDM-4 default value of 1.0 was assigned for
the calibration factors of these components (i.e. Kcpa and Krst = 1.0) before performing the
roughness calibration.

The available computing power (hardware and software) was sufficient to estimate the underlying
rates of deterioration for the 5327 individual sections within a reasonable time frame. However, the
calibration exercise for these sections was based on grouping the sections, as noted in Section
4.5.1, into the 57 group means (or average values) of the estimated deterioration rates.

Austroads 2008

— 30 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

Table 4.7 below shows some pavement characteristics that are part of the primary input
parameters required for calibration and Table 4.8 lists the calibration results, including the
estimated roughness deterioration rates and their corresponding calibration factors (Kgm), for each
group. Similar to the results for the other states, Kgm values were generally less than the default
HDM-4 value of 1.0 and varied only moderately across various groups. It appeared that the sub-
grade soil type had more influence on the calibration coefficient than any other parameters, as
indicated by K values becoming greater than 1.0 for a significant number of groups with reactive
soil.

Figure 4.4 plots the relationships between the calibration coefficients and their corresponding
roughness deterioration rates. It important to note that each data point plotted represents a group
mean, not the individual sections except for the eight groups where a single data point represented
a group listed in the above table.

The correlation obtained between the calibration coefficients and their corresponding roughness
deterioration rates was quite reasonable (r2 = 0.50), despite the fact that the group mean values
were used as input parameters for the calibration.

Table 4.8: Calibration results of the pavement groups in Queensland


Road type1 Pavement Soil reactivity3 Climate No. of Roughness Calibration coefficient 5
type2 group4 sections progression (Kgm)
(NRM/year)
2.28 ~ 8.3
Dry 6 1.15
(4.51)
0.33 ~ 4.01
AMGB Non-reactive Temperate 17 0.17
(1.21)
0.1 ~ 2.27
Wet 9 0.27
(0.94)
Dry 1 1.50 0.39
0.21 ~ 8.21
Temperate 7 0.93
Reactive (3.64)
0.52 ~ 3.74
Wet 4 0.52
(1.9)
STGB 0.02 ~ 10.19
Dry 188 0.37
(1.81)
0 ~ 7.91
Non-reactive Temperate 254 0.22
(1.18)
NH
0.02 ~ 7.65
Wet 97 0.19
(1.35)
0.1 ~ 12.2
Dry 225 0.74
(2.02)
0.09 ~ 7.34
Reactive Temperate 86 0.34
(1.49)
0.02 ~ 4.67
Wet 128 0.34
(1.24)
STSB 0.21 ~ 4.65
Temperate 7 0.25
(1.52)
Non-reactive
0.47 ~ 1.03
Wet 3 0.16
(0.81)
Dry 1 3.58 0.75
Reactive 0.23 ~ 3.88
Temperate 12 0.31
(1.39)

Austroads 2008

— 31 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

Road type1 Pavement Soil reactivity3 Climate No. of Roughness Calibration coefficient 5
type2 group4 sections progression (Kgm)
(NRM/year)
0.52 ~ 1.43
AMGB Non-reactive Wet 5 0.44
(0.86)
0 ~ 12.16
Non-reactive Dry 606 0.78
(1.75)
0 ~ 6.71
Temperate 107 0.22
(1.2)
0 ~ 10.33
Wet 42 0.57
(1.68)
STGB
0 ~ 11.83
Reactive Dry 425 1.38
(2.2)
0.21 ~ 6.27
Temperate 27 0.39
(2.1)
SS
0.6 ~ 8.57
Wet 29 0.48
(2.69)
0.25 ~ 6.58
Non-reactive Dry 21 0.50
(1.92)
Temperate 1 0.56 0.12
0 ~ 6.15
Wet 6 0.33
(2.46)
STSB
0.21 ~ 4.96
Reactive Dry 10 0.38
(2.14)
0.01 ~ 6.01
Temperate 6 1.08
(3.31)
Wet 1 6.30 1.55

Austroads 2008

— 32 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

Road type1 Pavement Soil reactivity3 Climate No. of Roughness Calibration coefficient 5
type2 group4 sections progression (Kgm)
(NRM/year)
0.96 ~ 7.99
Non-reactive Dry 6 0.12
(2.83)
AMGB
1.3 ~ 5.48
Reactive Dry 5 0.40
(2.97)
0.02 ~ 16.75
Non-reactive Dry 260 0.66
(2.51)
0.02 ~ 13.13
Temperate 181 0.85
(1.82)
0 ~ 18.4
Wet 419 1.02
(2.35)
STGB
0 ~ 21.42
RR Reactive Dry 302 1.42
(2.56)
0.07 ~ 11.54
Temperate 126 1.00
(2.28)
0.08 ~ 16.4
Wet 132 1.06
(3.13)
0.11 ~ 7.15
Non-reactive Dry 18 0.70
(3.24)
0.04 ~ 7.91
STSB Temperate 29 0.70
(2.94)
0.8 ~ 10.08
Wet 7 0.86
(4.3)
0.96 ~ 7.99
Non-reactive Dry 6 0.12
(2.83)
AMGB
1.3 ~ 5.48
Reactive Dry 5 0.40
(2.97)
0.02 ~ 16.75
Non-reactive Dry 260 0.66
(2.51)
0.02 ~ 13.13
Temperate 181 0.85
(1.82)
0 ~ 18.4
Wet 419 1.02
(2.35)
STGB
0 ~ 21.42
Reactive Dry 302 1.42
(2.56)
RR 0.07 ~ 11.54
Temperate 126 1.00
(2.28)
0.08 ~ 16.4
Wet 132 1.06
(3.13)
0.11 ~ 7.15
Non-reactive Dry 18 0.70
(3.24)
0.04 ~ 7.91
Temperate 29 0.70
(2.94)
STSB 0.8 ~ 10.08
Wet 7 0.86
(4.3)
1.03 ~ 6.22
Reactive Temperate 17 0.40
(2.51)
Wet 1 0.52 0.14

Austroads 2008

— 33 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

Road type1 Pavement Soil reactivity3 Climate No. of Roughness Calibration coefficient 5
type2 group4 sections progression (Kgm)
(NRM/year)
0.18 ~ 6.82
Non-reactive Dry 16 0.55
(1.84)
Wet 1 0.61 0.12
AMGB
Reactive Dry 1 11.26 2.04
0.53 ~ 1.94
Wet 6 0.67
(1.09)
0 ~ 12.54
Non-reactive Dry 469 0.72
(1.69)
0.01 ~ 11.24
Temperate 122 0.90
(1.92)
0 ~ 15.06
Wet 233 0.24
(1.81)
STGB
0.01 ~ 14
DR Reactive Dry 196 0.44
(2.22)
0.18 ~ 11.82
Temperate 101 1.32
(3.14)
0 ~ 12.27
Wet 224 0.74
(2.67)
0.08 ~ 6.7
Non-reactive Dry 8 0.65
(2.22)
Temperate 1 0.92 0.25
0.06 ~ 8.15
STSB Wet 8 0.25
(1.49)
1.77 ~ 4.9
Reactive Dry 4 1.53
(3.3)
0 ~ 2.95
Wet 13 0.17
(1.1)
Notes:
1. Road types are defined as per QDMR classifications.
2. Pavement types are defined as per HDM-4 classifications (Appendix A).
3. Soil reactivity is as per QDMR data.
4. Climate group: dry (Thornthwaite Index, TI, < 0), temperate (0 < TI < 50), wet (TI > 50).
5. Revised calibration results obtained using the calibration approach described in Section 3.2.

Austroads 2008

— 34 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

2
Revised calibration
1.8
Linear (Revised R 2 = 0.498
1.6 calibration)

1.4
Calibration coefficient, Kgm

1.2

0.8

0.6

0.4

0.2

0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0

Roughness progression (NRM/year)

Figure 4.4: Relationship between rate of deterioration and RD model coefficients for Queensland road network

4.6 New Zealand State Highway Network


4.6.1 Road Sections for RD Model Calibration
The inventory and time series condition data for 63 calibration sections on the New Zealand state
highway network established in 2001 specifically for the long-term pavement performance study,
were made available by TNZ for the RD model calibration exercise.

All sections have a length of approximately 300 m, most are on two-lane two-way rural roads with
both lanes being monitored on a yearly basis. The sections were split into two groups, sterilised
and un-sterilised, based on the maintenance policy applied to them. The sterilised sections, with a
prefix ‘CAL’ on the section identification label, were meant to be kept free from maintenance except
for safety or emergency reasons, whereas the un-sterilised sections, labelled with a prefix ‘CS’,
were subject to normal routine maintenance.

The yearly monitoring was conducted independently in both lane directions for most sites, one in
increasing chainage and the other in decreasing chainage as identified by the suffix ‘I’ and ‘D’,
respectively on the section label. This allowed the two lane sections to be treated separately,
resulting in the number of individual calibration sections increasing from 63 to 124.

A comprehensive set of deterioration data was collected from the above sites using an ARRB
Walking Profiler (WP) for roughness measurements, a Transverse Profile Beam (TPB) for rut depth
measurements across the lane width, a Falling Weight Deflectometer (FWD) for deflection
measurements, and a Stationary Laser Profilometer (SLP) for texture measurements. For visual
condition rating, major forms of surface defects and all types of structural cracking were recorded
at each 50 m subsection.

Austroads 2008

— 35 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

As part of data preparation for the analysis, all data collected from the subsections of their main
300 m section were averaged to produce a single value representing each section.

The initial analysis identified and selected the sections experiencing surface condition deterioration
over a minimum period of two consecutive years on the basis that a minimum of three data points
were required for estimating deterioration. As both cracking and rutting have an influence on
roughness, a section for calibration must therefore show positive deterioration trends in cracking,
rutting and roughness. However, a review of all the cracking data (narrow plus wide cracks)
revealed that the percentage of cracking for the majority of the 300 m sections was less than 1%.
Also, the estimation of the rate of cracking deterioration indicated that the rate of cracking
deterioration was virtually nil. Consequently, no calibration of the RD model for cracking occurred
because it would not contribute to roughness deterioration.

After the initial analysis, only 52 out 124 sections were estimated to experience both rutting and
roughness deterioration. Given that calibration for each section took 15 to 30 minutes on average
for both roughness and rutting, it was possible to complete the calibrations for all 52 sections
individually within a reasonable time fame. On this basis all 52 sections were calibrated.

In addition to the performance data and inventory data, other data were supplied by TNZ including
pavement characteristics, rainfall and traffic, and maintenance history data.

4.6.2 Performance History


An almost complete set of performance data including roughness, rutting, cracking and deflection
collected on a yearly basis since the establishment of the calibration sections up to 2005 (or early
2006) for the New Zealand state highway network, were made available by TNZ for the study.
Despite a short monitoring period, a reasonable number of sections, that met the criteria for
estimating deterioration and the decision rules described in Appendix B, were available for
calibration.

4.6.3 Maintenance Intervention


Maintenance treatment data were available and described for each individual section. As noted
earlier, maintenance data were not needed because the method of estimating the underlying rate
of deterioration (Appendix B) excluded performance data that were suspected of being influenced
by maintenance intervention. However, maintenance history such as the years of rehabilitation and
resurfacing were important when both pavement and seal ages were used as part of the input data
for calibration.

4.6.4 Calibration Results


Calibration estimated the rutting structural deterioration coefficients (Krst) and roughness
progression coefficients (Kgm) for the New Zealand road sections. As noted previously, cracking
models were not calibrated due to insufficient sections with significant rates of cracking. Hence, a
HDM-4 default coefficient for cracking progression, Kcpa, of unity was assumed for all sections and
used as an input parameter for the calibration procedure described in Section 3.2.

Austroads 2008

— 36 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

Although the selected pavement sections were calibrated individually during the calibration, they
were reported in groups in the summary tables, Table 4.9 and Table 4.10, where descriptive
statistics of the group (mean, minimum and maximum value and sample size) for pavement
characteristics and calibration coefficients were presented. This form of result presentation was
used in order to maintain consistency with those of the Australian states where the grouping
needed to be done before processing because of the large number of sections used. Grouping of
the calibration sections was based on the following hierarchical structure: pavement type,
sterilisation, traffic and climate. As a result of grouping, the 52 calibrated sections were reduced to
only 12 groups.

Figure 4.5 plots the relationship between the estimated rutting and roughness progression rates
and the Krst and Kgm values, respectively. The data points plotted for both rutting and roughness
progressions represent individual pavement sections and not the group means as summarised in
Table 4.9 and Table 4.10.

The correlation between the rates of deterioration for rutting (r2 = 0.70 and roughness (r2 = 0.85)
and the K values for rutting and roughness were good, although they resulted from the removal of
two outliers representing the two asphalt on granular base sections (AMGB). A post-analysis
investigation of these data revealed that one section experienced very low rates of deterioration
and the other section showed unusually high rates of deterioration for both rutting and roughness
over the deterioration period. This was a sound basis on which to treat these two data points as
outliers for the correlation purpose.

From Table 4.10, both Krst and Kgm values appear to be within typical and expected ranges of the
calibration results achieved for the Australian states.

Austroads 2008

— 37 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

Table 4.9: Characteristics of the New Zealand pavement section groups for analysis
Pavement Sterilisation2 Traffic Climate No. of Pavement SNP Rutting Roughness
type1 group3 group4 sections age progressio progression
(years) n (NRM/year)
(mm/year)
31 ~ 40 4.45 ~ 10 0.12 ~ 3.83 0 ~ 7.62
AMGB S Heavy Sub-humid 2
(35.5 ) (7.23 ) (1.98 ) (3.81 )
12 ~ 12 1.9 ~ 3.51 0.29 ~ 0.53 0.03 ~ 0.53
Humid 2
(12 ) (2.71 ) (0.41 ) (0.28 )
4 ~ 21 0.7 ~ 5.85 0.26 ~ 1.07 0.11 ~ 9.75
Low Sub-humid 11
(12 ) (2.93 ) (0.74 ) (2.17 )
4 ~ 40 1.45 ~ 4.05 0.34 ~ 0.73 0.26 ~ 3.44
Semi-arid 5
(20 ) (2.53 ) (0.51 ) (1.7 )
NS
4 ~ 22 1 ~ 8.5 0.24 ~ 1.95 0.17 ~ 2.13
Humid 5
(7.6 ) (3.86 ) (0.93 ) (1.17 )
2.55 ~ 4.9 0.42 ~ 0.62 0.67 ~ 3.76
Medium Sub-humid 4 4 ~ 8 (5 )
(3.74 ) (0.57 ) (2.41 )
4 ~ 40 0.55 ~ 2.65 0.22 ~ 0.37 0.39 ~ 1.06
STGB Semi-arid 2
(22 ) (1.6 ) (0.3 ) (0.73 )
4 ~ 40 0.15 ~ 4.05 0.02 ~ 0.61 0.01 ~ 1.99
Humid 7
(20.71 ) (1.03 ) (0.24 ) (0.55 )
4 ~ 22 0.25 ~ 2.35 0.2 ~ 0.55 0.44 ~ 3.37
Low Sub-humid 5
(7.6 ) (1.27 ) (0.37 ) (1.82 )
40 ~ 40 0.3 ~ 0.4 0.17 ~ 0.25 0.39 ~ 1.87
S Semi-arid 2
(40 ) (0.35 ) (0.21 ) (1.13 )
4 ~ 23 2.05 ~ 4.55 0.39 ~ 1.07 0 ~ 4.73
Sub-humid 6
(13.17 ) (3.13 ) (0.87 ) (1.9 )
Medium
15 ~ 15 5.73 ~ 5.73 1.22 ~ 1.22
Semi-arid 1 NA
(15 ) (5.73 ) (1.22 )
Notes:
1. Road types are defined as per QDMR classifications.
2. Pavement types are defined as per HDM-4 classifications (Appendix A).
3. Soil reactivity is as per QDMR data.
4. Climate group: dry (Thornthwaite Index, TI, < 0), temperate (0 < TI < 50), wet (TI > 50).
5. Revised calibration results obtained using the calibration approach described in Section 3.2.

Austroads 2008

— 38 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

Table 4.10: Calibration results for the pavement groups for the New Zealand road network
Pavement Sterilisation2 Traffic Climate No. of RD model coefficients5
type1 group3 group4 sections Rutting Roughness
(Krst) (Kgm)
4.45 ~ 10 0.12 ~ 0.78
AMGB S Heavy Sub-humid 2
(7.23 ) (0.45 )
1.9 ~ 3.51 0.12 ~ 0.17
Humid 2
(2.71 ) (0.15 )
0.7 ~ 5.85 0.12 ~ 2.49
Low Sub-humid 11
(2.93 ) (0.63 )
1.45 ~ 4.05 0.15 ~ 1.4
Semi-arid 5
(2.53 ) (0.68 )
NS
1 ~ 8.5 0.12 ~ 0.39
Humid 5
(3.86 ) (0.25 )
2.55 ~ 4.9 0.17 ~ 1.16
Medium Sub-humid 4
(3.74 ) (0.74 )
0.55 ~ 2.65 0.17 ~ 0.6
STGB Semi-arid 2
(1.6 ) (0.39 )
0.15 ~ 4.05 0.11 ~ 0.26
Humid 7
(1.03 ) (0.14 )
0.25 ~ 2.35 0.14 ~ 0.85
Low Sub-humid 5
(1.27 ) (0.4 )
0.3 ~ 0.4 0.12 ~ 0.12
S Semi-arid 2
(0.35 ) (0.12 )
2.05 ~ 4.55 0.13 ~ 1.37
Sub-humid 6
(3.13 ) (0.53 )
Medium
5.73 ~ 5.73
Semi-arid 1 NA
(5.73 )
Notes:
1. Pavement types are defined as per HDM-4 classifications (Appendix A).
2. Sterilisation is used to flag whether or not the section is subject to the routine maintenance
3. Traffic group: low (AADT < 5000), medium (5000 < AADT < 15000), heavy (AADT > 15000).
4. Climate group: semi-arid (Rainfall (RF), 300<RF <800), sub-humid (800 < RF< 1600), humid (1500<RF<3000) as per HDM-4 moisture classification.
5. Revised calibration results obtained using the calibration approach described in Section 3.2.

Austroads 2008

— 39 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

12.0
Revis ed calibration
Outliers
10.0 Linear (Revis ed calibration)

8.0
R 2 = 0.7025
K rst values

6.0

4.0

2.0

0.0
0.0 1.0 2.0 3.0 4.0
Rutting progression (mm/year)

4.0
Revised calibration
Outliers
Linear (Revised calibration)

3.0

R2 = 0.8477
K gm values

2.0

1.0

0.0
0.0 2.0 4.0 6.0 8.0 10.0 12.0

Roughness progression (NRM/year)

Figure 4.5: Relationship between rate of deterioration and RD model coefficients for
New Zealand State Highway network

Austroads 2008

— 40 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

4.7 Summary of Calibration Factors


4.7.1 Victorian Road Network
A total of 55 asphalt and sprayed seal pavement sections were used for RD model calibration in
Victoria based on their observed deterioration using the objective estimation approach outlined
above. These sections were grouped according to road type, pavement type, traffic loading and
climate condition. The results from the current RD model calibration were compared with the
values obtained from a previous calibration study. Mean values of the calibrated coefficients (Krst
and Kgm) were estimated for each group of pavements. The mean values of the calibration
coefficients derived from the current and previous studies were comparable, although the current
study generally estimated lower mean values compared to those from the previous study.

When comparing the calibration coefficient (K) values with the estimated rate of pavement
deterioration, the results from the current calibration study showed an improved correlation
compared to the results from the previous calibration study. The correlation coefficients (r2) from
the current study were greater than 0.66, while those from the previous study were only 0.39.

4.7.2 Tasmanian Road Network


A total of 58 pavement sections from the Department of Infrastructure, Energy and Resources
(DIER), Tasmania, road network were calibrated by this study. Similar to the Victorian network, the
selected sections were grouped according to road type, pavement type, traffic loading and climate
condition in order to estimate and compare mean calibration coefficient (K) values from the current
calibration study with those from a previous calibration study. A comparison of the mean K values
for each group of pavements indicated that the current study generally yielded lower K values
compared to the K values from the previous study.

The calibration results from the current study were good because the correlation between the
calibration K values and the estimated rate of deterioration was high. The current study estimated
lower K values (Krst and Kgm) because the observed rates of deterioration were estimated to be
lower. The correlation coefficients (r2) between the calibration values and the estimated rates of
deterioration for the revised method were greater than 0.82, while the correlation coefficients (r2)
for the previous study were less than 0.13.

4.7.3 South Australian Road Network


Of the 129000 pavement sections with deterioration data available, only 14765 sections were
estimated to be experiencing deterioration concurrently with cracking, rutting and roughness and
used in the current study. Because of limited computational capacity, the calibration of all the
14765 sections was not undertaken. Instead, pavement sections were grouped in terms of road
type, pavement type, traffic loading and climatic conditions prior to calibration. The mean values of
observed and predicted deterioration for each group of pavements were estimated or assumed and
calibration was based on these mean values.

Austroads 2008

— 41 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

The calibration coefficients for cracking, rutting and roughness (Kgm, Krst and Kgm) for each group of
South Australian pavements are presented in this report. It should be noted that these K values are
applicable to pavement sections whose characteristics, in terms of road and pavement type, traffic
loading and climatic conditions, are similar to the mean values used to group the pavement
sections. The correlation between the calibration coefficient K values and the estimated rate of
deterioration was high for rutting (r2 = 0.89), practically nil for cracking, and reasonable for
roughness (r2 = 0.45). With future improvements in computational capacity and an efficient
deterioration algorithm, it may be possible to estimate the deterioration of all individual pavement
sections in a large road network so that the mean calibration coefficient K values for various
groups of roads could be estimated more accurately.

4.7.4 Queensland Road Network


Roughness deterioration, traffic and inventory data for over 28600 pavement sections, each 1 km
in length, were made available by the Queensland Department of Main Roads (QDMR) for the
study. Of these, more than 5300 sections had satisfied the selection criteria which included the
following: (1) traffic with AADT less than 5000; and, (2) deterioration identified and estimated by
the principles and decision rules of the underlying rate of deterioration approach outlined above.

Due to the large number of sections representing the network, the number of sections for the
calibration study was reduced by grouping the sections based on the usual criteria of road type,
pavement type, traffic loading and climate conditions. Group mean values of observed and
predicted deterioration, not the values of individual sections, were used as input data for
calibration.

The roughness calibration of 57 section groups, based on the mean values of estimated and
predicted deterioration, produced a range of calibration coefficient, K values, generally less than
the HDM-4 default value of 1.0. The correlation between group means of the estimated rate of
deterioration and the K values was reasonable (r2 = 0.50) given that there was some reduction in
the quality of input data due to using mean values of deterioration.

4.7.5 New Zealand State Highway Network


A large data set from 63 pavement sections, each 300 m in length and monitored on a yearly basis
to date, was provided by Transit New Zealand (TNZ). Performance data from 2001 to 2005
including roughness, rutting, cracking and deflection were collected either from each lane of a two-
lane two-way road or from one of the lanes on a divided carriageway. As each lane was treated
independently, the total number of lane sections available for the analysis increased to 124.

As for the Australian states, the selection of pavement sections for calibration was based on the
estimation of underlying rate of deterioration to represent the observed deterioration. It was found
that 54 of the 124 sections showed concurrent increases in rutting and roughness whereas only a
small number of sections were experiencing cracking deterioration. Consequently, RD models
were calibrated only for rutting and roughness using input data from the individual sections.

The calibration coefficients for rutting and roughness (Krst and Kgm) were plotted against and
correlated with their corresponding estimated rates of deterioration. The correlations between the
calibration coefficients and the estimated rates of deterioration were strong (r2 = 0.70 and 0.85 for
rutting and roughness respectively), indicating that the calibration approach produced reliable
estimates.

Austroads 2008

— 42 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

4.7.6 Comparison of Calibration Factors


The HDM-4 RD models require many independent variables. Calibration of all the HDM-4 RD
models demands detailed and high quality long term data. Because of the limited availability of all
the necessary data from the five road networks, usually only the roughness and rutting RD models
for each SRA were calibrated using this data, except for South Australia. Therefore, when based
on local performance data from limited sections, the calibrated RD models do not fully account for
all the factors (variables) that influence pavement deterioration predictions in other regions. It is
therefore not advisable to apply the RD model calibration coefficients obtained from this analysis to
road networks other than the ones they represent.

Consequently, the HDM-4 RD models should be calibrated for the local conditions of each SRA
separately. Standardised national HDM-4 RD models are not likely to be appropriate for use in
Australia and New Zealand.

Austroads 2008

— 43 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

5 WORK SUMMARY
The calibration of a range of HDM-4 RD models was completed using the data made available by
the SRAs in 2005/06 and 2006/07. The following outcomes were achieved:
ƒ estimation of calibration coefficients for roughness, rutting and, where possible, cracking for
road networks in Victoria, Tasmania, South Australia and Queensland
ƒ estimation of calibration coefficients for roughness and rutting for the State Highway network
in New Zealand.

Table 5.1 provides an overall summary of the RD calibrations based on the data made available
from the states and New Zealand for the asphalt and sealed granular pavement types for a range
of road types, climates, and traffic loads.

Table 5.1: Summary of HDM-4 RD calibration results


AMA Location Pavement type Rutting coefficient Roughness Cracking
(Krst) range coefficient (Kgm) coefficient (Kcpa)
range range
Victoria. asphalt 0.3–4.9 (1.4) 0.9–2.0 (0.8) –
sealed granular 0.5–3.6 (1.6) 0.2–0.8 (0.5) –
Tasmania asphalt 0.3–1.9 (1.1) 0.2–0.8 (0.5) –
sealed granular 0.2–3.0 (1.2) 0.1–0.7 (0.3) –
South Australia asphalt 0.2–8.8 (2.0) 0.2–1.2 (0.7) 0.1–1.3 (0.3)
sealed granular 0.2–5.2 (1.9) 0.3–1.0 (0.6) 0.2–0.7 (0.4)
Queensland asphalt – 0.1–2.0 (0.9) –
sealed granular – 0.1–1.4 (0.6) –
New Zealand asphalt 4.5–10.0 (5.0) 0.1–0.8 (0.4) –
sealed granular 0.2–8.5 (4.0) 0.1–2.5 (0.9) –

Austroads 2008

— 44 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

6 CONCLUSIONS AND RECOMMENDATION

6.1 Conclusions
HDM-4 road deterioration (RD) models for roughness and rutting were calibrated to suit road
sections in Victoria, Tasmania, South Australia and New Zealand. Only calibration for roughness
was carried out for Queensland as this was the only performance data of this type available for the
study. Cracking deterioration models were not calibrated for Victoria, Tasmania and Queensland
due to the low quality of the available data. The New Zealand cracking data was not used because
the data did not have significant cracking. The quality of the cracking data from South Australia
appeared reasonable, although calibration of the cracking model showed no correlation between
calibration coefficients and estimated deterioration rates of cracking.

An objective approach for estimating the observed rate of underlying deterioration as part of the
calibration process was documented in this report. This process calibrated the cracking, rutting
and roughness RD models to match the estimated observed rate of underlying deterioration. The
objective approach for estimating underlying rate of deterioration was based on previous Austroads
work.

The calibration results based on the objective approach for estimating deterioration were compared
with those from a previous method for results where they were available. The comparison showed
that the calibrations from both methods were comparable, although the objective approach yielded
generally lower calibration values compared to those from the previous method. It should be noted
that the RD calibration coefficients from the objective approach produced strong correlation with
the estimated rate of deterioration, as expected.

6.2 Recommendation
With future improvements in computational capacity and an efficient deterioration algorithm, it may
be possible to estimate the deterioration of all individual pavement sections in a large road network
so that the mean calibration coefficient K values for various groups of roads can be estimated more
accurately.

Austroads 2008

— 45 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

REFERENCES
Bennett, CR & Paterson W 2000, A guide to calibration and adaptation, vol. 5, highway development and
management, HDM-4 series of publications, World Bank, Washington DC, & PIARC, Paris, France.

Freeme, CR 1983, Evaluation of pavement behaviour for major rehabilitation of roads, ITRR technical report
RP/19/83, CSIR, South Africa.

GEIPOT 1982, Research on the interrelationships between costs of highway construction, maintenance, and
utilisation (PICR), final report, Enpresa Brasileria de Planejamento de Transportes (GEIPOT), Ministry
of Transport, Brasilia.

Hoque, Z & Martin, T 2005, ‘Development of interim deterioration (RD) models for sealed granular roads’,
contract report RC4202-3, ARRB Group, Vermont South, Victoria.

Hoque, Z & Martin, T 2006a, ‘Selection of additional sites for inclusion in Austroads long term pavement
performance study – final report’, contract report REAT 1064, ARRB Group, Vermont South, Victoria.

Hoque, Z & Martin, T 2006b, ‘Development of interim deterioration (RD) models for sealed granular and
asphalt roads’, contract report REAT1064, ARRB Group, Vermont South, Victoria.

Martin, T, Choummanivong, L, George, M & Hoque, Z 2006, Development of road condition performance
profiles: model documentation, contract report RC4387-2, ARRB Group, Vermont South, Victoria.

Martin, T & Hoque, Z 2005, ‘Methodology for desk-top review to identify ‘good’ and ‘poor’ pavement
performance’, draft contract report RC4309, ARRB Group, Vermont South, Victoria.

Martin, T & Hoque, Z 2006, ‘Under-performing pavements, identification, classification, inspection and
causes’, contact report REAT-1067, ARRB Group, Vermont South, Victoria.

Morosiuk, G, Riley, M & Odoki, JB 2001, HDM-4 modelling road deterioration and works effects, vol. 6,
highway development and management, HDM-4 series of publications, World Bank, Washington DC,
& PIARC, Paris, France.

Odoki, JB & Kerali, HGR 2000, HDM-4 analytical framework and model descriptions, vol. 4, highway
development and management, HDM-4 series of publications, World Bank, Washington DC, & PIARC,
Paris, France.

Toole, T, Michel, N & Roper, R 2004a, ‘Implementation of HDM-4 in Tasmania: part 1 – project summary and
results of strategy analysis’, contract report RC3296-1, ARRB Transport Research, Vermont South,
Victoria.

Toole, T, Roper, R & Martin, T 2004b, ‘Improved HDM-4 model calibration factors and application guidelines
for sealed Roads in Victoria, final project’, contact report RC2464 - part 1, ARRB Transport Research,
Vermont South, Victoria.

Austroads 2008

— 46 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

APPENDIX A HDM-4 PAVEMENT CODES,


CLASSIFICATIONS AND DETERIORATION
MODELS
A.1 Pavement Classifications
The HDM-4 pavement classification system is shown in Table A 1 and generic bituminous
pavement types are shown in Table A 2.

Table A 1: HDM-4 Bituminous pavement classification system


Surface Surface Base Base Pavement
type material type material type
AC CRS
GB AMGB
HRA GM
PMA AB AB AMAB
RAC CS
AM SB AMSB
CM LS
PA TNA
AMAP
SMA AP FDA
Xx
CAPE CRS
GB STGB
DBSD GM
SBSD AB AB STAB
ST SL CS
SB STSB
PM LS
Xx TNA STAP
AP
FDA

Notes: AM Asphalt mix GB Granular base


ST Surface treatment AB Asphalt base
AC Asphaltic concrete AP Asphalt pavement
HRA Hot rolled asphalt SB Stabilised base
PMA Polymer modified asphalt CS Crushed stone
RAC Rubberised asphalt concrete GM Natural gravel
CM Soft bitumen mix (cold mix) AB Asphalt base
PA Porous asphalt CS Cement stabilisation
SMA Stone mastic LS Lime stabilisation
CAPE Cape seal TNA Thin surfacing (< 100 mm)
DBSD Double bituminous surface dressing FDA Thick surfacing (> 100 mm)
SBSD Single bituminous surface dressing PM Penetration macadam
SL Slurry seal Xx User defined

Austroads 2008

— 47 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

Table A 2: Generic HDM-4 bituminous pavement types


Pavement Surface Base Description
type type type
AMGB AM GB Asphalt mix on granular base
AMAB AM AB Asphalt mix on asphalt (dense bitumen macadam) base
AMSB AM SB Asphalt mix on stabilised base
AMAP AM AP Asphalt mix on asphalt pavement
STGB ST GB Surface treatment on granular base
STAB ST AB Surface treatment on asphalt (dense bitumen macadam) base
STSB ST SB Surface treatment on stabilised base
STAP ST AP Surface treatment on asphalt pavement

A.2 Deterioration Models


A.2.1 Crack Modelling
The HDM-4 cracking model considers two components of cracking:
ƒ structural cracking
ƒ transverse thermal cracking – occurs in mainly freeze/thaw conditions. Hence not covered
here.

Separate models are used for each component and they are predicted as a function of the factors
that are known to contribute to their development and progression.

Structural cracking
Structural cracking is load and age/environment associated cracking. Two types of structural
cracking are considered – all structural cracking and wide structural cracking. In this report only all
structural cracking was considered. Therefore models related to wide structural cracking are not
included here.

Initiation of all structural cracking


Crack initiation is assumed to occur when 0.5% of the carriageway surface area is cracked.
Initiation of all structural cracking depends on the base:
ƒ Stabilised base
If HSOLD = 0 (that is, new surfacing)
ICA = Kcia {CDS2 a0 exp[a1 HSE + a2 loge CMOD + a3 loge DEF + a4 (YE4) A.1
(DEF)] + CRT}
If HSOLD > 0 (that is, overlays or reseals)
ICA = Kcia {CDS2 [(0.8KA + 0.2KW) (1+0.1HSE) + (1-KA)(1-KW) a0 * exp[a1 A.2
HSE + a2 loge CMOD + a3 loge DEF + a4 (YE4) (DEF)] ]+ CRT

Austroads 2008

— 48 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

ƒ Other bases
If HSOLD = 0 (that is, new surfacing)
ICA = Kcia {CDS2 a0 exp[a1 SNP + a2 (YE4/SNP2)] + CRT} A.3
If HSOLD > 0 (that is, overlays or reseals)
For all surface materials except CM, SL and CAPE
ICA = Kcia {CDS2 [max(a0 exp[a1 SNP + a2 (YE4/SNP2)] * max(1- A.4
PCRW/a3,0), a4HSNEW)] + CRT}

ƒ For surface materials - CM, SL and CAPE


ICA = Kcia {CDS2 [max(a0 exp[a1 SNP + a2 (YE4/SNP2)] * max(1-PCRA/a3,0),a4)]+ A.5
CRT}

where:

ICA = time to initiation of all structural cracks (years)


CDS = construction defects indicator for bituminous surfacings
YE4 = annual number of equivalent standard axles (millions/lane)
SNP = average adjusted structural number of the pavement
DEF = mean Benkelman Beam deflection in both wheel paths (mm)
CMOD = resilient modulus of soil cement (GPa) (in the range between 0 and 30 GPa
for most soils)
HSNEW = thickness of the most recent surfacing (mm)
HSOLD = total thickness of previous underlying surfacing layers (mm)
PCRA = area of all cracking before latest reseal or overlay (% of total carriageway
area)
PCRW = area of wide cracking before latest reseal or overlay (% of total carriageway
area)
KW = min(0.05 max (PCRW – 10, 0), 1)
KA = min (0.05 max (PCRA – 10, 0), 1)
HSE = min (100, HSNEW + (1-KW) HSOLD)
Kcia = calibration factor for initiation of all structural cracking
CRT = crack retardation time due to maintenance (years)
a0 to a4 = model coefficients, default values are given in Table A 3.

Austroads 2008

— 49 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

Table A 3: Default coefficient values for initiation of all structural cracking models
Pavement type Surface material HSOLD value Equation a0 a1 a2 a3 a4
All 0 A.3 4.21 0.14 -17.1
AMGB All except CM >0 A.4 4.21 0.14 -17.1 30 0.025
CM >0 A.5 13.2 0 -20.7 20 1.4
0 A.3 4.21 0.14 -17.1
AMAB All
>0 A.4 4.21 0.14 -17.1 30 0.025
AMAP All >0 A.4 4.21 0.14 -17.1 30 0.025
0 A.1 1.12 0.035 0.371 -0.418 -2.87
AMSB All
>0 A.2 1.12 0.035 0.371 -0.418 -2.87
All 0 A.3 13.2 0 -20.7
STGB All except SL, CAPE >0 A.4 13.2 0 -20.7 20 0.22
SL, CAPE >0 A.5 13.2 0 -20.7 20 1.4
All 0 A.3 13.2 0 -20.7
STAB All except SL, CAPE >0 A.4 4.21 0.14 -17.1 20 0.12
SL, CAPE >0 A.4 4.21 0.14 -17.1 30 0.025
STAP All >0 A.4 4.21 0.14 -17.1 20 0.12
0 A.1 1.12 0.035 0.371 -0.418 -2.87
STSB All
>0 A.2 1.12 0.0.35 0.371 -0.418 -2.87

Progression of all structural cracking


dACA = Kcpa [CRP/CDS] ZA [(ZA a0 a1 δtA + SCAa1)1/a1 – SCA] A.6

Progression of all structural cracking commences when δtA > 0 or ACAa > 0
where
If ACAa > 0 δtA = 1; otherwise δtA = max {0, min[(AGE2 – ICA), 1]}
If ACAa >= 50 then ZA = -1; otherwise ZA = 1
ACAa = max(ACAa, 0.5)
SCA = min[ACAa, (100 - ACAa)]
Y = a0 a1 ZA δtA + SCAa1 A.7
If Y < 0 then dACA = Kpca [CRP/CDS] (100 - ACAa) A.8
If Y ≥ 0 then dACA = Kpca [CRP/CDS] ZA (Y(1/a1) – SCA) A.9
If ACAa < 50 and (ACAa + dACA) > 50 then
dACA = Kpca [CRP/CDS] (100 – C1(1/a1) - ACAa] A.10

Austroads 2008

— 50 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

where
C1 = max{[2(50a1) - SCAa1 - a0 a1 δtA], 0] A.11
dACA = incremental change in area of all structural cracking during the analysis
year (% of total carriageway area)
ACAa = area of all structural cracking at the start of the analysis year
δtA = fraction of analysis year in which all structural cracking progression
applies
AGE2 = pavement surfacing age since last reseal, overlay, reconstruction or new
construction (years)
ICA = time to initiation of all structural cracking (years)
Kcpa = calibration factor for progression of all structural cracking
CRP = retardation of cracking progression due to preventive treatment, given by
CRP = 1 – 0.12 CRT
a0 & a1 model coefficients, default values are given in Table A 4.

Table A 4: Default coefficient values for progression of all structural cracking


Pavement type Surface material HSOLD value All cracking
a0 a1
All 0 1.84 0.45
AMGB All except CM >0 1.07 0.28
CM >0 2.41 0.34
AMAB All 0 1.84 0.45
>0 1.07 0.28
AMAP All >0 1.07 0.28
AMSB All 0 2.13 0.35
>0 2.13 0.35
STGB All 0 1.76 0.32
>0 2.41 0.34
All 0 1.76 0.32
STAB All except SL, CAPE >0 2.41 0.34
SL, CAPE >0 1.07 0.28
STAP All >0 2.41 0.34
STSB All 0 2.13 0.35
>0 2.41 0.34

A.2.2 Rut Depth Models


Rutting is defined as the permanent (unrecoverable) deformation within pavement layers. The rut
depth model is based on there being four components to rutting:
ƒ initial densification
ƒ structural deterioration
ƒ plastic deformation
ƒ wear from studded tyres.

Austroads 2008

— 51 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

The rut depth at any time is the sum of the four components. Rut depth modelling will be
performed after the values of all the surface distresses (i.e., cracking, ravelling, potholing and edge
break) at the end of the year have been calculated.

Initial densification
The initial densification depends upon the degree of compaction of the base, sub-base and
selected subgrade layers, i.e. COMP.

The initial densification is:

RD0 = Krid [a0 (YE4 106)(a1 + a2 DEF) SNPa3 COMPa4] A.12

where
RD0 = rutting due to initial densification (mm)
DEF = Benkelman Beam deflection (mm)
YE4 = annual number of equivalent standard axles (millions/lane)
SNP = average annual adjusted structural number of the pavement
COMP = relative compaction in percent
Krid = initial rut densification factor
a0 to a4 = model coefficients, default values are given in Table A 5.

The initial densification only applies to new construction or reconstruction that involve the
construction of a new base layer (i.e. from when AGE4 = 0), for a period of time of one year. AGE4
is defined as follows:

AGE4 = age since reconstruction (including base) or new construction (years).

Table A 5: Default coefficient values for initial densification model


Pavement type a0 a1 a2 a3 a4
AMGB, AMAB, AMSB, STGB, STAB, STSB 51740 0.09 0.0348 -0.502 -2.30
AMAP, STAP 0 0 0 0 0

Structural deterioration
Separate terms are proposed for structural deformation without cracking and structural deformation
after cracking as follows:

ƒ Structural deformation without cracking


ΔRDSTuc = Krst (a0 SNPa1 YE4a2 COMPa3) A.13

ƒ Structural deformation after cracking


ΔRDSTcrk = Krst (a0 SNPa1 YE4a2 MMPa3 ACXaa4)
P A.14

Austroads 2008

— 52 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

The total annual incremental increase in structural deformation is as follows:


If ACRA = 0 then ΔRDST = ΔRDSTuc A.15

If ACRA > 0 then ΔRDST = ΔRDSTuc + ΔRDSTcrk A.16

where

ΔRDST = total increment increase in structural deformation in the analysis year


(mm)
ΔRDSTuc = total increment increase in structural deformation without cracking in the
analysis year (mm)
ΔRDSTcrk = total increment increase in structural deformation after cracking in the
analysis year (mm)
MMP = mean monthly precipitation (mm/month)
ACXa = area of indexed cracking at the beginning of the analysis year (% of total
carriageway area)
SNP = average annual adjusted structural number of the pavement
YE4 = annual number of equivalent standard axles (millions/lane)
Krst = calibration factor for structural deformation
a0 to a4 = model coefficients, default values are given in Table A 6.

Table A 6: Default coefficient values for structural deformation model


Pavement type a0 a1 a2 a3 a4
Without cracking All pavement types 44950 -1.14 0.11 -2.3
After cracking All pavement types 0.0000248 -0.84 0.14 1.07 1.11

Plastic deformation
The plastic deformation model includes a variable, CDS, which indicates whether the surfacing is
prone to plastic deformation. The general plastic deformation model is given by:

ΔRDPD = Krpd CDS3 a0 YE4 Sha1 HSa2 A.17

where
ΔRDPD = incremental increase in plastic deformation in the analysis year (mm)
CDS = construction defects indicator for bituminous surfacings
YE4 = annual number of equivalent standard axles (millions/lane)
Sh = speed of highway vehicles (km/h)
HS = total thickness of bituminous surfacing (mm)
Krpd = calibration factor for plastic deformation
a0 to a2 = model coefficients, default values are given in Table A 7.

Austroads 2008

— 53 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

Table A 7: Default coefficient values for plastic deformation model


Surface type a0 a1 a2
AM 2.46 -0.78 0.71
ST 0 -0.78 0.71

Surface wear
The surface wear model is applied to environments where vehicles use studded tyres during the
freezing period.

ΔRDW = Krsw [a0 PASSa1 Wa2 Sa3 SALTa4 ] A.18

where
ΔRDW = incremental rut depth due to studded tyres, in mm
PASS = annual number of vehicles with studded tyres in one direction, in thousands
S = average traffic speed in km/h
SALT = variable for salted or unsalted roads (2 = salted; 1 = unsalted)
W = road width in m, (i.e., the carriageway width plus total shoulder width)
Krsw = calibration factor for surface wear
a0 to a4 = model coefficients, default values are given in Table A 8.

Table A 8: Default coefficient values for surface wear model


Pavement type a0 a1 a2 a3 a4
All pavement types 0.0000248 1.0 -0.46 1.22 0.32

Total rut depth


If AGE4 <= 1

ΔRDM = RD0 + ΔRDPD + ΔRDW A.19

otherwise
ΔRDM = ΔRDST + ΔRDPD + ΔRDW A.20

Standard deviation of rut depth


The standard deviation of rut depth is used in the roughness model. It is calculated from the mean
total rut depth as:

RDSb = max [0.3, (0.9 – 0.04 RDMb)] RDMb A.21

where
RDSb = standard deviation of rut depth at the end of the year (mm)
RDMb = mean rut depth at the end of the analysis year (mm)

Austroads 2008

— 54 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

A.2.3 Roughness
The roughness model consists of the predictions for each component of roughness (cracking,
disintegration, deformation, and maintenance). The total incremental roughness is the sum of
these components. The surface distress values used in predicting roughness are those that have
been adjusted so that the total damaged surface area plus the undamaged area equals 100%.

Structural
The structural component of roughness relates to the deformation in the pavement materials under
the shear stress imposed by traffic loading.

ΔRIs = a0 exp (m Kgm AGE3) (1 + SNPKb)-5 YE4 A.22

and

SNPKb = max [(SNPa – dSNPK), 1.5] A.23

and

dSNPK = Ksnpk a0 {min (a1, ACXa) HSNEW + max [min (ACXa – PACX, A.24
a2), 0] HSOLD
where
ΔRIs = incremental change in roughness due to structural deterioration during the
analysis year (IRI m/km)
dSNPK = reduction in the structural number of pavement due to cracking
SNPKb = adjusted structural number of pavement for cracking at the end of the
analysis year
SNPa = adjusted structural number of pavement at the start of the analysis year
ACXa = area of indexed cracking at the start of the analysis year (% of total
carriageway area)
PACX = area of previous indexed cracking in the old surfacing, i.e. 0.62 PCRA +
0.39 PCRW
HSNEW = thickness of the most recent surfacing (mm)
HSOLD = total thickness of previous underlying bituminous surfacing layers (mm)
AGE3 = pavement age since last overlay (rehabilitation), reconstruction or new
construction (years)
m = environmental coefficient, values are given in Table A 9
Kgm = calibration factor for environmental coefficient
Ksnpk = calibration factor for SNPK
a0 to a2 = model coefficients, default values are given in Table A10.

Austroads 2008

— 55 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

Table A 9: Roughness environmental coefficient ‘m’ by climate zones


Moisture Temperature classification
classification Tropical Sub-tropical hot Sub-tropical cool Temperate cool Temperate freeze
Arid 0.005 0.010 0.015 0.025 0.040
Semi-arid 0.010 0.015 0.025 0.035 0.060
Sub-humid 0.020 0.025 0.040 0.060 0.100
Humid 0.025 0.030 0.060 0.100 0.200
Per-humid 0.30 0.040 0.70

Cracking
The incremental change in roughness due to cracking is given by:
ΔRIc = a0 ΔACRA A.25

where
ΔRIc = the change in roughness due to cracking during the analysis year (IRI
m/km)
ΔACRA = incremental area of total cracking during the analysis year (% of total
carriageway area)
a0 = model coefficient, default value is given in Table A.10.

Rutting
The incremental change in roughness due to variation of rut depth is given by:

ΔRIr = a0 ΔRDS A.26

where:
ΔRIr = incremental change in roughness due to rutting during the analysis year
(IRI m/km)
ΔRDS = incremental increase in the standard deviation of rut depth during the
analysis year (mm) (= RDSb – RDSa)
a0 = model coefficient, default value is given in Table A.10.

Potholing
The potholing effect depends upon the number of vehicles which actually hit the pothole which in
turn depends upon the traffic volume and the freedom to manoeuvre. This is predicted using the
following equation:

FM = (max [min {0.25 (CW - 3), 1), 0]) max (1 - AADT/5000, 0) A.27

where:
FM = the freedom to manoeuvre
CW = carriageway width in m
AADT = the two-way traffic flow in vehicle/day.

Austroads 2008

— 56 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

The change in roughness is calculated as follows:


ƒ If nil patching (TFL = 1) or 100% patching policy options is specified, then
ΔRIt = a0 (a1 - FM) [{NPTa * TLF + (ΔNPT*TLF)/2}a2 - NPTaa2] A.28

Otherwise (partial policy options)


ΔRIt = a0 (a1 - FM) ΔNPT * (NPTa + NPT/2)a2 A.29

where:
ΔRIt = incremental change in roughness due to potholing during the analysis year
(IRI m/km)
FM = freedom to manoeuvre
CW = carriageway width (m)
AADT = annual average daily traffic (veh/day)
ΔNPT = incremental number of potholes per km during the analysis year
NPTa = number of potholes per km at the start of the analysis year
TLF = time lapse factor
a0 to a2 = model coefficients, default values are given in Table A.10.

Environment
The environmental component of roughness is due to factors which include temperature and
moisture fluctuations, and also foundation movements. This is given by:

ΔRIe = m*Kgm RIa A.30

where:
ΔRIe = incremental change in roughness due to environment during the analysis
year (IRI m/km)
RIa = the roughness at the start of the analysis year (IRI in m/km)
Kgm = calibration factor for the environmental component of roughness
m = environmental coefficient.

Total change in roughness

ΔRI = Kgp [ΔRIs + ΔRIc + ΔRIr + ΔRIt] + ΔRIe A.31

where
ΔRI = total incremental change in roughness during the analysis year (IRI m/km)
Kgp = calibration factor for roughness progression.

Austroads 2008

— 57 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

The roughness at the end of the year is given by:

RIb = min [(RIa + ΔRI), a0) A.32

The annual average roughness for a given analysis year is calculated as:

RIav = 0.5*(RIa + RIb) A.33


where
RIa = roughness of the pavement at the start of the analysis year (IRI m/km)
RIb = roughness of the pavement at the end of the analysis year (IRI m/km)
a0 = upper limit of pavement roughness, (default = 16 IRI m/km)
RIav = annual average roughness of the pavement for the analysis year (IRI m/km)

Table A 10: Default coefficient values for roughness component


Pavement type Roughness component Equation a0 a1 a2
Structural A.22 134
dSNPK A.24 0.0000758 63.0 40.0
All pavement types Cracking A.25 0.0066
Rutting A.26 0.088
Potholing A.28 0.00019 2.0 1.5

Austroads 2008

— 58 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

APPENDIX B ESTIMATION OF UNDERLYING RATE OF


DETERIORATION
B.1 Principles of Estimating Deterioration
The following principles were adhered to in estimating deterioration from the time series of distress
deterioration data for each pavement segment (Martin and Hoque 2006):
ƒ Deterioration estimates measure the underlying rate of the ‘latest’ deterioration, that is, the
deterioration of the pavement that is not influenced by the immediate impact of maintenance
treatments (Figure B 1 for roughness example).
ƒ Deterioration estimates are made from data post the latest rehabilitation treatment (Figure B
2 for roughness example).
ƒ Estimates are based on an absolute minimum of three consecutive ‘valid’ data points,
although four consecutive points are more reliable.
ƒ Some allowance for ‘noise’ in the data must be considered, including identifying outliers that
are errors in measurement and not related to previous or later measurements.

1.6

'early' deterioration trend

1.2
Roughness

'latest' deterioration trend

0.8

immediate impact of maintenance (or


data scatter?)
0.4
roughness data
'latest' deterioration estimate
'early' deterioration estimate

0.0
0 1 2 3 4 5 6 7 8
Time (years)

Figure B 1: ‘Latest’ deterioration estimate (roughness example)

Austroads 2008

— 59 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

1.4

1.2

decrease due
to rehab.
1
Roughness

0.8 'latest' deterioration


estimate
0.6

0.4

0.2 roughness data


'latest' deterioration estimate
0

0 1 2 3 4 5 6 7 8
Time (years)

Figure B 2: Estimated deterioration post rehabilitation (roughness example)

B.2 Decision Rules


The following decision rules were used for estimating the latest underlying rate of deterioration:
1. Deterioration estimation begins with a minimum of four of the latest data points with the
earliest time series data point being the reference for evaluation of the other data points.
2. The analysis considers all time series data provided they are considered ‘valid’.
3. To avoid unexpected increases in deterioration, or the effect of maintenance, and to allow for
‘noise’ in the data, two terms are defined: (a) the maximum limit (ML) is defined as the upper
limit of consecutive deterioration that can be acceptable; and, (b) the tolerance limit (TL) is
defined as the maximum improved condition allowed between two consecutive
measurements (for roughness and rutting it is the maximum decrease from the immediate
past recorded value). The limits assumed for ML and TL are shown in Table B 1
4. An increase of up to ML over a period of up to two years between two consecutive data
points was considered acceptable, that is, the data point was considered ‘valid’ for estimating
deterioration (Figure B 3).
5. A decrease of up to TL over a period of up to two years between two consecutive data points
was considered acceptable, that is, the data point was considered ‘valid’ for estimating
deterioration (Figure B 4).
6. An increase of more than ML over a period of up to two years between two consecutive data
points was considered unacceptable (identified suspected outlier) where this increase was
contrary to the trend line which includes more recent time series data (Figure B 5).
7. A decrease of more than ML over a period of up to two years between two consecutive data
points was assumed to be the result of either rehabilitation or a major maintenance
intervention activity.

Austroads 2008

— 60 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

8. A decrease between TL and ML over a period of up to two years between two consecutive
data points was considered unacceptable (identified suspected outlier) where the decrease
was followed by an increase in distress above the reference data point (Figure B 6).
9. A decrease between TL and ML over a period of up to two years between two consecutive
data points was considered unacceptable (identified suspected outlier) where the decrease
was followed by an increase in distress which remained below the reference data point. In
this case the reference point was an outlier, but the sequential point was considered to be
valid (Figure B 7).
10. An ‘outlier’ is considered a ‘valid’ data point when the difference between the ’outlier’ and the
trend line, which excludes the ‘outlier’, is less than or equal to TL (Figure B 8). If no
deterioration estimation is possible using four ‘valid’ data points, a less reliable estimate with
three ‘valid’ data points is made.
Table B 1: Assumed maximum and tolerance limits

Performance parameter Maximum limit (ML) Tolerance limit (TL)


Roughness (NRM) 25 5
Rutting (mm) 5 1
Cracking (%) 15 3

'valid' data point


Performance parameter

ML

1 - 2 years

data evaluation from


earliest data point

Time

Figure B 3: Acceptable increase (or deterioration) in observed data (≤ ML)

Austroads 2008

— 61 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

1 - 2 years
Performance parameter

≤ TL

data evaluation from


earliest data point 'valid' data point

Time

Figure B 4: Acceptable decrease (or improvement) in observed data (≤ TL)

suspected outlier trend line (initially excludes


suspected outliers)

1 - 2 years
Performance parameter

> ML

> TL

1 - 2 years

data evaluation from earliest data point suspected outlier

Time

Figure B 5: Suspected outlier above and below the trend-line (>ML and > TL)

Austroads 2008

— 62 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

'valid' data point

≤ ML

reference point
Performance parameter

trend line (includes


'valid' data points)

data evaluation from


earliest data point
> TL but ≤ ML suspected outlier

1 - 2 years 1 - 2 years

Time

Figure B 6: Suspected outlier below the trend line (> TL but ≤ ML)

Suspected outlier (&


reference point)

> TL but ≤ ML
Performance parameter

trend line (includes


'valid' data points) ≤
ML

'valid' data point

1 - 2 years 1 - 2 years
data evaluation
from earliest data
point

Time

Figure B 7: Suspected outlier above the trend line (> TL but ≤ ML)

Austroads 2008

— 63 —
Development of HDM-4 Road Deterioration (RD) Model Calibrations for Sealed Granular and Asphalt Roads

> TL
'valid' data point > ML
Outlier

≤ TL
Performance parameter

data evaluation trend line (excludes


from earliest data > ML
'valid' data points)
point

≤ TL
1 - 2 years

1 - 2 years 'valid' data point

Time

Figure B 8: Suspected outliers compared with the trend line and valid data points decided

Austroads 2008

— 64 —
INFORMATION RETRIEVAL

Austroads, 2008, Development of HDM-4 Road Deterioration (RD) Model


Calibrations for Sealed Granular and Asphalt Roads, Sydney, A4, 75pp,
AP-T97/08

Keywords:

Road deterioration (RD), pavement performance, RD modelling, sealed and


asphalt pavements, HDM-4 roughness, rutting and cracking models, RD model
calibration.

Abstract:

Improved RD models for sealed granular and asphalt pavements are needed.
Pavement performance history data from state road authorities and New
Zealand (SRAs) were used for calibrating RD models. The calibration and
refinement of RD models is expected to continue up to 2007/08. This report
presents the outcome of using an objective method to estimate pavement
deterioration from the performance history of the Victorian, Queensland,
Tasmanian, South Australian and New Zealand sealed and asphalt road
network.

The objective method of estimating deterioration and calibrating HDM-4


roughness, rutting and cracking models is presented. It is recommended that
this approach be used to analyse performance data from other SRAs in
2007/08 for RD model calibration.

Você também pode gostar