You are on page 1of 84

Predicting the Performance of Australia’s Arterial and Sealed Local Roads

ARR 390
Tim Martin and Lith Choummanivong
ARR 390 About the Authors
March 2018
Tim Martin
ISBN 978-1-876592-81-3 Dr Tim Martin is a Chief Scientist (Asset Manage-
ISSN 0158-0728 ment) with ARRB. Tim has graduate and post-grad-
uate qualifications in civil engineering and a PhD in
Any material may be reproduced predicting network level sealed granular pavement
without permission provided the deterioration. Tim has over 40 years of experience in
source is acknowledged. a wide range of civil engineering work. Tim’s research
at ARRB commenced in 1990 and involved leading
Although the report is believed to an extensive road track cost attribution study, the
be correct at the time of publica- design and implementation of a major observational
tion, the Australian Road Research study using long term pavement performance (LTPP)
Board, to the extent lawful, ex- and long term pavement maintenance (LTPPM) sites,
cludes all liability for loss (wheth- experimental studies with accelerated load testing,
er arising under contract, torte, and the development of a range of pavement dete-
statue or otherwise) arising from rioration and works effects models for arterial road
the contents of the Report or it’s sealed granular pavements in a life-cycle costing
use. Where such liability cannot approach to asset management. More recently, this
be excluded, it is reduced to the research has extended to the development of dete-
full extent lawful. Without limiting rioration and works effects models for unsealed and
the forgoing, people should apply sealed local roads as part of a national study across
their own skill and judgment all Australian states.
when using the information con-
tained in the report. Lith Choummanivong
Lith Choummanivong is a Senior Research Engineer
Wholly prepared and printed by with ARRB Asset Management Group. He holds a
the Australian Road bachelor of civil engineering from RMIT and a Master
Research Board of Engineering (pavements) degree from Deakin Uni-
500 Burwood Highway versity. Lith has over 20 years of research experience
Vernont South VIC 3133 in pavement-related areas including pavement con-
AUSTRALIA struction quality management, material characteri-
sation, pavement trials with accelerated load testing
and asset management. Over the last ten years, he
has been involved in a number of long-term pave-
ment studies including sealed and unsealed local
roads deterioration modelling, long-term pavement
performance (LTPP) and the effects of maintenance
on long-term performance of pavement (LTPPM).
Predicting the Performance of Australia’s Arterial
and Sealed Local Roads

Tim Martin and Lith Choummanivong

Australian Road Research Board


March 2018
ARR 390
SUMMARY
The prediction of pavement performance is a critical component in a well-structured pavement
management system (PMS) for estimating preservation and capital funding requirements for
road networks. In the past many of the performance prediction models, such as the World
Bank’s HDM models, used in PMSs in Australia were based on data collected on pavement
performance in Brazil, the Caribbean and Kenya from 1971 to 1984. However, most of these
models were unlikely to be well suited to Australasian conditions of construction, maintenance,
climate and traffic without significant calibration and adjustment to local Australian conditions.

In many cases the explanatory power of the early models used in PMSs was limited.

A questionnaire survey in 2008 of Australiasian road agency practitioners revealed that the rate
of structural deterioration of pavements was ranked as the highest priority by the majority of
practitioners. The second highest priority of practitioners was the functional rate of pavement
deterioration (roughness, rutting, cracking, etc.).

This research report is based on the observational and experimental research data collected by
the Australian Road Research Board (ARRB) between 1994 and 2017 aimed at producing for
sealed flexible pavements deterministic road deterioration (RD) models (roughness, rutting,
cracking and strength) and deterministic work effects (WE) models (asphalt overlays, granular
re-sheeting and mill and replace asphalt). This work was largely funded by Austroads and the
Institute of Public Works Engineering Australasia (IPWEA) as well as contributions from local
government.

Some reseach was also conducted on the performance of unsealed local roads, however, the
resulting outcomes of this research are not the subject of this report.

The arterial roads RD and WE models have undergone some refinements since they were first
published in 2010. This report is aimed at ensuring that these models are easily comprehended
and highly implementable for practitioners.

More recently pavement performance analysis was conducted on pavements experiencing rapid
deterioration as these pavements approach ultimate structural and functional failure. Some RD
models for rapid deterioration were developed and are included in this report, although they are
based on limited data and are not likely to be reliable predictors of rapid deterioration. However,
they contribute to extending the range of factors relevant to pavement performance modelling,
and a more comprehensive analytical framework.

The deterministic RD and WE models for Australia’s arterial and sealed local roads documented
in this report represent a substantial step change in the technology and the financial investment
in research over some 23 years. These models have the capacity to improve the decision-
making processes of all Australian road agencies under a variety of distresses, distress phases,
conditions of climate and traffic loading and maintenance regimes.

No models can be claimed to be 100% reliable; however, these RD and WE models, particularly
those deveoped for arterial roads, have a high degree of utility and have contributed to a wider
analytical framework for performance modelling in Australia. Some local calibration of the

Page | i
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
models may be needed to account for local factors such as reactive soils, drainage and material
variability. This calibration is particularly important for local roads.

More research in the future can be undertaken to improve the reliability, range and scope of
these RD and WE models for all roads using the complete network-wide suite of road condition
measurements and associated factors that are becoming available with advanced data collection
techniques.

Acknowledgements
The authors acknowledge the technical support of their colleagues Tyrone Toole and Thorolf Thoresen. The authors
also gratefully acknowledge ARRB survey and data processing teams across Australia. The development of the model
outcomes described in this report was funded by Austroads, the IPWEA and local government across Australia.

Page | ii
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Contents
1 Introduction ...................................................................................................................... 1

1.1 Background .............................................................................................................. 1


1.1.1 Arterial Road Performance ................................................................................ 1
1.1.2 Sealed Local Road Performance ...................................................................... 2
1.2 Scope of this Report ................................................................................................. 3
1.2.1 Three Stages of In-Service Pavement Performance ......................................... 3
1.2.2 Three Phases of Road Deterioration ................................................................. 4
1.2.3 Limit of Gradual Deterioration ........................................................................... 5
2 Review of Data for Model Development .......................................................................... 7

2.1 Observational and Experimental Data ...................................................................... 7


2.1.1 Arterial Roads – Observational Data ................................................................. 7
2.1.2 Arterial Roads – Experimental Data .................................................................. 7
2.1.3 Sealed Local Roads – Observational Data ....................................................... 8
2.2 Limits to RD Model Prediction .................................................................................. 9
2.3 RD Model Calibration ............................................................................................... 9
3 Arterial Road RD Models for Structural Deterioration .................................................... 11

3.1 Structural Deterioration Models for Gradual Deterioration...................................... 11


3.1.1 Pavement Strength and Traffic Load Capacity ................................................ 11
3.1.2 Structural Pavement Performance .................................................................. 12
3.1.3 Structural Pavement Deterioration .................................................................. 13
3.1.4 End of Structural Life ....................................................................................... 19
4 Sealed Local Road RD Models for Structural Deterioration .......................................... 21

4.1 Structural Deterioration Models for Gradual Deterioration...................................... 21


4.1.1 Pavement Strength and Traffic Load Capacity ................................................ 21
4.1.2 Structural Pavement Performance .................................................................. 22
4.1.3 Structural Pavement Performance Using Combined LTPP/LTPPM and LRDS
Data 24
5 Arterial Road RD Models for Surface Deterioration ....................................................... 26

5.1 RD Models for Gradual Deterioration ..................................................................... 26


5.1.1 Permanent Deformation – Rutting ................................................................... 26
5.1.2 Cumulative Rutting, ∑Δrut(t) ........................................................................... 27
5.1.3 Annual Incremental Rutting, Δrut(t)i ................................................................ 32
5.1.4 Cumulative Surface Cracking, ∑Δcrx(t) ........................................................... 34
5.1.5 Cumulative Roughness, ∑ΔIRI(t) .................................................................... 37

Page | iii
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
5.1.6 Annual Incremental Roughness, ΔIRI(t)i ......................................................... 41
5.2 RD Models for Rapid Deterioration ......................................................................... 43
5.2.1 Surface Cracking – Sprayed Seals ................................................................. 43
5.2.2 Permanent Deformation – Rutting ................................................................... 45
5.2.3 Roughness ...................................................................................................... 47
6 Sealed Local Road RD Models for Surface Deterioration ............................................. 49

6.1 RD Models for Gradual Deterioration ..................................................................... 49


6.1.1 Permanent Deformation – Rutting ................................................................... 49
6.1.2 Model for Cracking Deterioration – Thin Bituminous Surfacings ..................... 51
6.1.3 Model for Roughness Deterioration ................................................................. 53
7 WE Models .................................................................................................................... 57

7.1 WE Models for Arterial Roads ................................................................................ 57


7.1.1 Sampled WE Maintenance Treatments ........................................................... 57
7.1.2 Model Development ........................................................................................ 58
7.2 WE Model Outcomes .............................................................................................. 59
7.2.1 Asphalt Overlays ............................................................................................. 59
7.2.2 Granular Re-sheeting ...................................................................................... 61
7.2.3 Mill and Replace (M&R) .................................................................................. 62
8 Recommendations ......................................................................................................... 64
9 References .................................................................................................................... 65
APPENDIX A: Summary of LTPP and LTPPM Sites ........................................................... 68
APPENDIX B: Experimental Accelerated Load Testing Results ........................................... 72

Page | iv
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
1 Introduction

1.1 Background
The prediction of pavement performance is a critical element in using a well-structured pavement
management system (PMS) for estimating maintenance and capital funding requirements for road
networks. In the past many of the performance prediction models, such as the World Bank’s HDM
models, used in PMSs in Australia were based on data collected in Brazil, the Caribbean and
Kenya from 1971 to 1984 (Cox 1990). However, these models were not particularly well suited to
Australasian conditions of construction, maintenance, climate and traffic.

A questionnaire survey (Martin & Moffatt 2008) of road agency representatives revealed that the
rate of structural deterioration of pavements was ranked as the highest priority by the majority of
practitioners. The second highest priority of practitioners was the functional (roughness, rutting,
cracking, etc.) rate of pavement deterioration.

1.1.1 Arterial Road Performance


The US Strategic Highway Research Program (SHRP), which commenced in 1987 (Crawley
1997), establishing a long-term pavement performance (LTPP) program that collected pavement
performance data as one of its major research areas. In 1995 Australia participated in SHRP by
mean of an Austroads-funded project 1 (Long-term pavement performance (LTPP) observational
study), to monitor a range of test sections on a number of arterial roads in various Australian
states. These test sections were known as the SHRP-LTPP sites. At about the same time, a
number of test sections were set up in tandem with the accelerated loading facility 2 (ALF) for a
related experimental study; these test sections were known as the ALF-LTPP sites. These studies
were aimed at improving the understanding of the long-term behaviour of in-service pavements
under different loading and environmental conditions in order to develop road deterioration (RD)
and works effects (WE) models that would credibly predict the future functional and structural
condition of pavements.

In 1998, eight sites, known as the long-term pavement performance maintenance (LTPPM) sites,
were established for a separate Austroads-funded observational study on the effect of various
maintenance/surface treatments on pavement performance in Victoria, New South Wales,
Queensland and Tasmania. In addition, a large number of additional LTPP sites were set up in
2000 and 2002 in New South Wales (NSW), Queensland (QL) and the Australian Capital Territory
(ACT) to cover a larger range of pavement types and increase the overall sample size for the LTPP
study. In 2001/02 an Austroads project was created to combine the databases and monitoring of
the LTPP and LTPPM sites.

Pavement performance data (roughness, rutting, deflection (strength) and visual (surface distress)
was collected annually for the LTPPM sites. Data was collected at less frequent intervals for most
of the LTPP sites due to their slow rate of deterioration. A breakdown of the LTPP and LTPPM

1
Austroads is the association of Australian federal/state/territory road agencies, the Australian Local Government Association and the
New Zealand Transport Agency.
2
The Accelerated Loading Facility (ALF) is an experimental device that applies full-scale repetitive heavy vehicle axle loads to a test
pavement simulating a whole of life design traffic loading in a relatively short period of time.

Page | 1
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
sites according to study type, pavement type, jurisdiction, climate and traffic load is presented in
Table A.1 in Appendix A. The traffic load on the sites varies from 0.02 to 2.30 MESA/lane/year and
the Thornthwaite Moisture Index (Thornthwaite 1948), TMI, varies from –24 (dry) to 100 (wet). Both
rigid (concrete) and flexible (asphalt and sprayed seal unbound pavements) are represented,
although the majority (96%) of the pavements are flexible pavements. Consequently, the rigid
pavement performance observational data is not adequate for model development.

It can be seen from Table A.1 that a total of 39 LTPP and eight LTPPM sites have been
established since 1994/95. As each of the LTPPM sites contained five sub-sections, there were up
to 79 LTPP and LTPPM sections being monitored. Currently 25 LTPP and six LTPPM sites (55
sections) are being monitored as there was a net loss of 24 sections due either to major
rehabilitation works or lane re-alignment over the 21 years of monitoring. In 2013/14 two new LTPP
sites were established in South Australia (SA06 and SA07 in Table A.1).

Supplementing this observational data collection, experimental data collected during ALF trials was
used to estimate the relative performance factors for surface maintenance effects, rpfm, and
increased axle loads, rpfl (Martin 2010a, 2010b).These factors were applied to the observed values
of roughness and rutting to supplement the observational data by increasing the range of its
variables for maintenance and loading changes. The observational and supplementary data was
used to derive the arterial road roughness and rutting distress models described in this report.
Table B.1 in Appendix B shows the values of rpfm determined for various surface maintenance
treatment options. Appendix B also shows how estimates of rpfl for roughness and rutting under
increased axle loading were determined.

1.1.2 Sealed Local Road Performance


Most local roads in Australia are managed by local government agencies (LGAs). ARRB initiated
the local roads deterioration study (LRDS) in 2000 with the aim of developing RD models suitable
for Australian local road conditions for unsealed and sealed local roads. To develop robust RD
models, a sufficiently large sample of local road sites across Australia needed to be established.
The sites were selected to be representative of local road networks. They covered a wide range of
climatic and traffic conditions, pavement material and surfacing types, as well as construction and
maintenance practices.

Some 236 organisations participated with data supply and financially supported the study. As a
result, 500 sealed and 100 unsealed test sites were established across Australia. A breakdown of
these sites by road type and jurisdiction is shown in Table 1.1 (Martin et al. 2013).

Page | 2
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Table 1.1: Details of sealed and unsealed test sites by state/territory

Number of Total number of


Number of Number of
State/Territory participating sites (sealed and
sealed sites unsealed sites
organisations unsealed)
New South Wales 59 118 20 138
Victoria 53 144 25 169
Queensland 41 64 22 86
Western Australia 37 66 13 79
South Australia 24 47 20 67
Tasmania 17 34 - 34
Australian Capital
1 6 - 6
Territory
Northern Territory 4 21 - 21
Total 236 500 100 600

The aim of the LRDS was to develop both sealed and unsealed deterministic RD models suitable
for Australian local road conditions, using all the data collected during the monitoring period from
2002 to 2009 from all states and territories. Unsealed RD models are not the subject of this report.

As agreed by the participating organisations, the key performance parameters (dependent


variables) developed for the sealed RD models were:
• strength – deflection
• rutting
• cracking
• roughness.

These key performance parameters for the sealed roads were measured at two to three year
intervals to gain at least three-time series data points of deterioration.

1.2 Scope of this Report


This report provides documentation of the currently-available Australian RD and WE mechanistic-
empirical deterministic models in a simple algorithm form that can be used to predict the life-cycle
performance of flexible sprayed seal unbound granular pavements and asphalt pavements, the
pavement types used for the majority of Australia’s arterial and local roads.

1.2.1 Three Stages of In-Service Pavement Performance


Figure 1.1 shows, in general terms, the three main stages of in-service pavement performance for
which most RD and WE models apply. The first stage (1) is the usual road deterioration (RD) of
pavement condition, both functional and structural, post construction. The second stage (2) is the
impact of works effects (WE) on pavement condition. Typically, most works effects, such as the
applications of thin surfacings to correct rutting, do not fully restore the pavement to its original
surface condition and strength unless they are either a full reconstruction or a major rehabilitation.
The third stage (3) is the road deterioration (RD) of pavement conditions post works effects.
Usually the third stage deterioration is similar to the first stage deterioration, although, if the

Page | 3
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
pavement has not had its strength fully restored, the deterioration may be greater than the
deterioration in stage (1).

Figure 1.1: Three stages of pavement performance

1.2.2 Three Phases of Road Deterioration


The three phases of road deterioration, or distress, with the passage of time or traffic load are
shown in Figure 1.2. The first phase, initial densification, occurs within the first 12 months after the
pavement is opened to traffic. Initial densification is apparent in certain forms of permanent distress
such as rutting and roughness. Initial densification may also lead to an increase in pavement
strength, but this is not usually accounted for in deterministic RD modelling.

Figure 1.2: Three phases of road deterioration (RD)

The second phase of road deterioration, gradual deterioration, is the phase in which most in-
service pavements operate. As Figure 1.2 shows, the gradual deterioration proceeds at a low rate
and is virtually linear. Most of the RD models documented in this report (Martin & Choummanivong
2010, Choummanivong & Martin 2010, Martin et al. 2013) are confined to predictions in the gradual
deterioration phase.

Page | 4
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
The third and final phase of deterioration is the rapid deterioration phase. As shown in Figure 1.2, it
is difficult to predict as the pavement approaches ultimate failure. Some RD models were recently
developed for this phase (Choummanivong & Martin 2014, Martin & Choummanivong 2015);
however, they are based on limited data and are currently not fully reliable predictors of rapid
deterioration due to the limited observational data. These RD models are included in this report
because they can be used to predict the impact of reduced maintenance funding on road
conditions.

1.2.3 Limit of Gradual Deterioration


The three phases of deterioration were simulated by ALF testing of various surface maintenance
treatments. This formed the basis for defining the frontier limiting the gradual deterioration phase of
sealed granular pavements. This frontier was defined using a binary logistic regression analysis
(Mendenhall & Sincich 1996) of the total of 71 samples identified as undergoing either gradual or
rapid deterioration (Martin 2010a). The samples identified as undergoing rapid deterioration were
defined as those deteriorating at a rate of two to three times the samples undergoing gradual
deterioration.

Figure 1.3 compares the rutting (mm) and roughness (IRI) collected at the 53 samples defined as
experiencing gradual deterioration (open squares) and the 18 samples defined as experiencing
rapid deterioration (filled squares).

The frontier between gradual and rapid deterioration was defined as follows:

rutmax = 86.347 − 11.008 × IRI 1.1


where
rutmax = mean maximum vertical deformation from the original surface profile (mm) with an
absolute maximum value of 25 mm
IRI = International Roughness Index (m/km).

All the coefficients and the constant term in Equation 1.1 had relatively low values of standard error
(SE) and all were significant at the 95% confidence level. Equation 1.1 predicted the correct state
of the gradual and rapid deterioration for over 94% of the total of 71 samples.

Figure 1.3: Mean rut (mm) vs roughness (IRI) for ALF maintenance experiments
50
Onset of rapid deterioration (Logit model y = 0)
( rut = 86.347 − 11.008 × IRI )
40
Mean rut (mm)

30 Max. rut depth


= 25 mm

20

10
Gradual deterioration samples
Rapid deterioration samples
Frontier prediction by logistic regression
0
0 5 7.84 10 15 20
IRI (m/km)

Page | 5
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
While some uncertainty exists because the definition is related to the accuracy of the ALF
simulation, the distress limitations for the extent of the gradual deterioration phase correspond well
with the practical roughness limit (IRI of 7.84) and rutting limit (25 mm) currently used to manage
pavement conditions (Smith et al. 1996). The rate of pavement deterioration during the rapid
deterioration phase and the consequent levels of distress would not be acceptable to road users.

Page | 6
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
2 Review of Data for Model Development

2.1 Observational and Experimental Data


2.1.1 Arterial Roads – Observational Data
A total of 27 discrete sealed granular pavement segments with continuous time series deterioration
data were selected from the LTPP and LTPPM sites. These segments formed the database for RD
model development. There was a total of 140 samples, including:
• the experimentally-supplemented data (sealed granular pavement dependent deterioration
variables cumulative roughness, ∑ΔIRI, and cumulative rutting, ∑Δrut)
• the estimated independent variables for traffic load, MESA, climate, TMI
• the initial pavement strength, SNC0, and the annualised value of maintenance expenditure,
me, representing the amount of preventative maintenance effort applied to the pavement
surface.
Table 2.1 summarises the source and extent of the data from observations and supplementary
data (derived from the ALF experimental data) used for the RD model development. In all cases
the performance data for roughness, rutting and cracking was limited to the pavement samples
showing increased deterioration with time (see stage (1) in Figure 1.1). Measured improvements in
performance, or condition, achieved through major maintenance and rehabilitation intervention
were separately covered by the works effects (WE) models (stage (2) as shown in Figure 1.1).

The assessment of whether the available performance data was showing continuous deterioration
over time (see stage (1) of Figure 1.1) was made by estimating the linear deterioration rate (LDR)
of the time series of the performance data (roughness, rutting and cracking) using an approach
originally documented for the identification of under-performing pavements by Martin & Hoque
(2007). This approach filtered out and excluded ‘noisy’ data and data representing improvements
in pavement condition (see stage (2) of Figure 1.1).

2.1.2 Arterial Roads – Experimental Data


The measured observational data from the LTPP/LTPPM sites shown in Table 2.1 was
supplemented by experimental data from a series of full-scale simulation experiments using the
ALF. The ALF experimental data separately quantified the impact on pavement deterioration of
changes to surface maintenance treatments (Martin 2010a) and increased axle loads (Martin
2010b), as noted in Section 1.1.1, for model development.

Page | 7
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Table 2.1: Summary of observational and experimental data used in RD model development

Data Independent variable range Dependent variable range No.


source MESA(6) AGE(2) TMI(7) SNC0(8) me(9) ∑ΔIRI(10) ∑Δrut(11) ∑Δcrx(12) samples

0.007 to 1 to -43 to 3.0 to 0 to 0.08 to 68


LTPP 0 to 14.79 0 to 100
2.54 105(5) 99 16.7 3365 6.24 (14(3), 54(4))

0.04 to 1 to -43 to 3.9 to 995 to 0.26 to 72


LTPPM 0 to 5.94 0
2.09 64(5) 99 11.5 3286 25.14 (47(3),25(4))
SA
0.008 to -35 to
cracking 5 to 79 – – – – 0.4 to 53 1675
2.07 6
data(1)
Notes:
1. Data made available by Department of Transport, Energy and Infrastructure (DTEI), South Australia (Martin et al. 2006).
2. Pavement age at time of observation.
3. LTPP/LTPPM observational samples.
4. Supplementary experimental data.
5. Pavement age (years) was extended at LTPP/LTPPM sites to reach limit of gradual deterioration (see Equation 1) for changes in axle loads.
6. Traffic load in millions of equivalent standard axles (ESAs) per lane per year.
7. Thornthwaite Moisture Index.
8. Modified structural number for pavement/subgrade strength at AGE = 0 (see Section 3.1).
9. Annualised pavement maintenance expenditure ($/lane-km/year).
10. Cumulative roughness (IRI, m/km).
11. Cumulative rutting (mm).
12. Cumulative cracking (% total lane area).

The ALF experimental data was used to derive relative performance factors, rpfm, for changes to
surface maintenance treatments and relative performance factors, rpfl, for changes to axle loads
during the gradual deterioration phase. The relative performance factors, rpfm and rpfl, were
applied to the observational data to increase and supplement the range of the data for possible
future changes to maintenance and axle loads. The observational data was linearly extrapolated
and interpolated up to the limit of the gradual deterioration phase by the relative performance
factors, rpfm and rpfl. This limit, or frontier, is a function of roughness (IRI) and rutting (mm) found
from ALF experiments as shown in Equation (1.1).

This valid approach involved the application of the outcomes of experimental work to quantify the
effects of the variables thought to influence the dependent deterioration variable (Box et al. 1978,
Robinson 1994). Appendix B presents a summary of the estimates of the relative performance
factors, rpfm and rpfl, as well as documenting how they were applied to the observational
roughness and rutting data.

2.1.3 Sealed Local Roads – Observational Data


As part of data preparation for the LRDS analysis for sealed local roads, a brief summary of
deterioration trends for all types of distress (rutting, roughness, strength and cracking), was
produced using the data collected through field surveys from 2002 to 2009 (see Table 2.2). The
trend estimated is the slope of a line of best fit through three data points on a coordinate system of
distress versus time.

Page | 8
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Table 2.2: Deterioration trends by distress type for RD models
Distress type (with deterioration trend indicated by +ve and -ve signs)
Surface Total
Statistics Rutting Roughness SNC(4) Cracking (%)
type
+ve(1) -ve(2) +ve -ve +ve -ve(3) +ve -ve
No. sites 128 63 148 43 107 84 127 42 191
Percentage 67.0% 33.0% 77.5% 22.5% 56.0% 44.0% 71.8% 23.7% (38.2%)
Asphalt
Average
0.21 0.10 1.60
progression -0.123
(mm/yr) (IRI/yr) (%/yr)
rate
No. sites 202 106 194 114 176 133 152 71 309
Percentage 65.6% 34.4% 63.0% 37.0% 57.0% 43.0% 61.3% 28.6% (61.8%)
Sprayed seal
(SS) Average
0.37 0.09 1.29
progression -0.11
(mm/yr) (IRI/yr) (%/yr)
rate
No. sites 330 169 342 157 283 217 279 113 500
asphalt + SS
Percentage 66.1% 33.9% 68.5% 31.5% 56.6% 43.4% 65.6% 26.6% (100%)

Notes:
1. Deterioration trend is positive (+ve) when there is an increase in rutting, roughness and cracking.
2. Deterioration trend is negative (-ve) when there is a decrease in rutting, roughness and cracking.
3. A decrease in strength for SNC (+ve) sign and the (-ve) sign means an increase in strength for SNC.
4. Modified structural number for pavement/subgrade strength (see Section 3.1).

2.2 Limits to RD Model Prediction

A feature of mechanistic-empirical deterministic models is that they provide single-value


predictions of the dependent variable. However, these single-value predictions do not recognise
that the dependent variable has an error function, Є, associated with each independent variable in
the model. Deterministic model predictions are considered to be the average of future
performance, assuming a symmetrical scatter of errors around the average, which means that 50%
of the actual condition may be better or worse than the predicted one.

2.3 RD Model Calibration


In adapting RD models to road networks it is expected that some form of local model calibration is
needed to ensure model adaptation to local conditions such as climate and other local variables
not directly accounted for in the models. These variables include reactive subgrades and the
impact of drainage on the subgrade and the pavement materials.

In undertaking a model calibration, at least three measurements of road conditions over at least a
three year interval are advisable for a reliable assessment of deterioration. Reliable and consistent
measures of traffic, climate and pavement strength are also needed. As noted in Section 1.2.2, the
gradual deterioration phase, which accounts for most of the in-service performance for sealed
pavements, is essentially linear.Time series deterioration data over several years can include data
‘noise’ and be influenced by maintenance intervention and measurement inaccuracy. In order to
estimate the underlying rate of linear deterioration that is also appropriate for calibration purposes,

Page | 9
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
the heuristic 3 rule-based linear deterioration rate (LDR) tool was developed by Martin & Hoque
(2007) to account for these factors.

3
Heuristic – a practical approach to problem solving which in this case involved estimation of a positive deterioration rate from time
series data.

Page | 10
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
3 Arterial Road RD Models for Structural Deterioration

3.1 Structural Deterioration Models for Gradual Deterioration


3.1.1 Pavement Strength and Traffic Load Capacity
Pavement strength has been historically represented by the modified structural number, SNCi,
which is the sum of the strength of the pavement, SNi, and subgrade, SNsg, as follows:

SNCi = SNi + SNsg 3.1


where
SNi = structural number for pavement strength at the time ‘i’ of its assessment
n

= 0.0394 × Ʃ ai × hi (AASHTO 1972)


i=1

ai = material strength coefficient for the ith layer of the pavement (= 0.14 for unbound granular
materials and 0.45 for asphalt)
hi = thickness of the ith layer of the pavement (mm)
SNsg = subgrade contribution to pavement strength (Hodges et al. 1975)
= 3.51 × log10(CBR) − 0.85 × [log10(CBR) ]2 − 1.43
CBR = Californian Bearing Ratio.

The material strength coefficients, ai, of each pavement layer are expected to decrease in value
over time due to the effects of cumulative traffic and the environment.

Pavement strength, SNCi, at time ‘i’ is related to the annual traffic load capacity of the pavement, in
terms of millions of cumulative equivalent standard axles per lane per year (MESA), over its design
life. Examination of the LTPP database (Choummanivong & Martin 2016) for arterial road sites
enabled the development of an empirical relationship (Toole & Roper 2014) between the initial
modified structural number, SNC0, immediately post-construction at zero pavement age and the
cumulative traffic load capacity, CAP, based on an annual traffic load, MESA, over a service life
(SL) of 60 years at an annual growth rate of 2.5%. Values of SNCi were estimated from the
maximum deflection, D0, measured using a Falling Weight Deflectometer (FWD) and back-
calculation (see Section 3.1.3) using a SNCi deterioration relationship to estimate SNC0.

The relationship between the initial modified structural number, SNC0, and the cumulative traffic
load capacity of the pavement, CAP, is as follows:

SNC0 = 1.128 × CAP0.1033 3.2


where
CAP = traffic load capacity over a defined service life
= MESA × CGF × 106
MESA = Millions of equivalent standard axles per lane per year
CGF = cumulative growth factor for a defined service life (SL) of up to 60 years at 2.5% annual traffic
growth rate
= (1.02560 – 1)/0.025 = 136.

The Equation 3.2 relationship is shown in Figure 3.1.

Page | 11
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Figure 3.1: SNC0 vs design traffic load capacity CAP (arterial roads)

3.1.2 Structural Pavement Performance


The modified structural number, SNCi, has also been a useful parameter in predicting pavement
deterioration at a network level (Watanatada et al. 1987). The use of SNCi to predict pavement
deterioration was based on the assumption that SNCi adequately represented the performance of
various pavement structures and subgrade strengths. This means that, if the pavement systems
were different but had the same value of SNCi, then their performance was predicted to be
identical, all other things being equal.

The pavement/ strength, SNCi, at time ‘i’ is related to the maximum deflection, D0, measured using
a Benkelman Beam, as follows for sealed pavements (Paterson 1987):

SNCi = αi × D0-0.63 3.3


where
SNCi = modified structural number for pavement strength at the time ‘i’ of its measurement
αi = 3.2 for unbound granular pavements
αi = 2.2 for bound (asphalt) pavements
D0 = maximum Benkelman Beam deflection (mm).

The D0 Benkelman Beam deflection is usually taken to be equal to the FWD deflection, although
this is not strictly correct (Austroads 2008).

In the event that the current modified structural number, SNCi, is not available, a default SNC0
value can be calculated. The default SNC0 is calculated using the following expression derived
from NAASRA (1979) and Hodges et al. (1975) for unbound granular pavements. The expression
is based on the design life of the pavement, the traffic load it is expected to experience over the
nominal pavement design life and a relatively low estimate of CBR (5%):

SNC0 = 0.55 × Log10(MESADL / 120 × 106) + 0.6 3.4


where
MESADL = millions of cumulative equivalent standard axles per lane over nominal design life
(DL), allowing for annual growth (%).

Page | 12
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
This estimate of SNC0 can be seen as a lower-bound value, rather than a mean value, because it
is based on a conservative pavement design approach using a low estimate of CBR (5%) for the
subgrade. Figure 3.2 compares the Equation 3.2 and 3.4 relationships between SNC0 and CAP.

Figure 3.2: Comparison of Equations 3.2 and 3.4 for SNC0 vs design traffic load capacity
CAP

The estimate of traffic load capacity (Equation 3.4) always gives a much lower value of SNC0 for a
given traffic capacity than the SNC0 value from Equation 3.2. It should be noted that increasing the
assumed value of CBR in Equation 3.4 to a more realistic value will cause these relationships to
give similar estimates of SNC0, particularly in the lower capacity range. This suggests that using an
increased value of CBR in Equation 3.4 should be used as the default value for SNC0 estimation.

3.1.3 Structural Pavement Deterioration


Table 3.1 summarises the range of data variables used in developing the structural deterioration
models from the LTPP/LTPPM data. Table 3.2 summarises the overall estimated rates of
deterioration from the time series LTPP/LTPPM data.

It can be seen from Table 3.2 that over 70% of the data experienced a gradual increase in
pavement strength, SNCi, of 0.12 per year, that is, a positive deterioration rate (+ ve). This
increase corresponded to a decrease in TMI of 0.99 per year, indicating increased climatic drying
experienced by these samples. The remaining 30% of the data experienced a gradual decrease
pavement strength, SNCi, of 0.08 per year, that is, a negative deterioration rate (- ve), while the
TMI increased at a rate of 0.95 per year.

Table 3.1: Summary of data variables used in model development


Pavement age, AGE MESA (ESA x
Pavement type TMI SNCi
(years) 106/lane/year)
Unbound granular 8 – 57 -25 – 100 0.019 – 1.97 3.2 – 7.7
Bound (cement-stabilised) 2 – 45 11 – 63 0.013 – 2.17 2.3 – 13.4
Asphalt 3 – 45 -2 – 54 0.094 – 4.26 3.3 – 14.1

Page | 13
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Table 3.2: Deterioration rates for time series variables
Deterioration rate (+ ve) Deterioration rate (- ve)
Total LTPPM/LTPP
at LTPPM/LTPP sites at LTPPM/LTPP sites
Variable
No. of Mean No. of Mean No. of Mean
sections/sites rate/yr sections/sites rate/yr sections/sites rate/yr
SNC 50 +0.12 (1) 21 -0.08(2) 71 +0.063
(3) (4)
TMI 60 -0.99 11 +0.95 71 -0.691
Annual traffic
57 36,600 14 -7,590 71 27,910
growth (ESAs)
Notes:
1. Positive (deterioration) rate for SNC means an increase in strength.
2. Negative (deterioration) rate for SNC means a decrease in strength.
3. Negative (deterioration) rate for TMI means a drier climate.
4. Positive (deterioration) rate for TMI means a wetter climate.

Structural deterioration model for pavements

The structural deterioration dependent variable, expressed in terms of the structural ratio, SNCratio,
is defined as follows:

SNCratio = SNCi/SNC0 3.5

Complete documentation of the structural deterioration model is found in Choummanivong & Martin
(2010). Further modification of these models was made to ensure that an initial boundary condition
was met, that is, the SNCratio is unity at zero pavement age (SNCi = SNC0). The resulting
deterioration models are defined as follows:

SNCratio = 2 – EXP[1/ks × AGEi (0.00004413 × TMIi + 0.2581/SL )] (granular pavements) 3.6

SNCratio = 2 – EXP[1/ks × AGEi × (0.00001942 × TMIi + 0.2975/SL)] (asphalt pavements) 3.7

where
TMIi = Thornthwaite Moisture Index at the time t = i of its estimation
AGEi = pavement age at time t = i (years)
ks = structural deterioration calibration factor (default = 1.0)
SL = service life of pavement ranging from 40 years (lightly-trafficked arterial) up to 60 years
(heavily-trafficked arterial or freeway).

Equation 3.6 was based on 278 samples with a goodness of fit (r2) of 0.72 while Equation 3.7 was
based on 165 samples with a goodness of fit (r2) of 0.79. Both of these equations are statistically
significant and have independent variables that are statistically significant. Equations 3.6 and 3.7
do not contain a traffic load variable, MESA. It is postulated that this is largely because arterial
roads are designed to carry a well-defined design traffic load and therefore a traffic load variable
was not identified as being significant.

Equations 3.6 and 3.7 predict that structural deterioration is approximately linear, except under
extreme conditions of climate and traffic load. Equations 3.6 and 3.7 also predict that structural
deterioration starts once the pavement age, AGE, exceeds zero. These equations depend on the
pavement age, AGEi, the assumed service life, SL, and climate as measured by TMIi.

Page | 14
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Equations 3.6 and 3.7 can indirectly incorporate the impact of increased traffic load, MESA, by a
pro-rata reduction in the service life, SL, as follows:

original service cumulative traffic load (MESAd)


Reduced SL = SL × 3.8
increased service cumulative traffic load (MESAr)
where
MESAd = original design cumulative MESA estimated for the service life, SL
MESAr = revised design cumulative MESA estimated for the service life, SL

Estimation of SNC0

SNC0 was estimated by re-expressing Equations 3.6 and 3.7 in terms of SNCi as follows:

SNCi.
SNC0 = (granular pavements) 3.9
[ 2 – EXP[1/ks × AGEi × (0.00004413 × TMIi + 0.2581/SL)]]

SNCi
SNC0 = (asphalt pavements) 3.10
[ 2 – EXP[1/ks × AGEi × (0.00001942 × TMIi + 0.2975/SL) ]]

Equations 3.9 and 3.10 estimate SNC0 which is used subsequently as an independent variable in
the RD models for rutting and roughness.

SNCratio variation with calibration, ks

Figure 3.3 plots Equation 3.6 for granular pavements with various values of the structural
deterioration calibration factor, ks. It can be seen from Figure 3.3 that the predicted rate of
structural deterioration can be adjusted by the calibration factor to suit a wide range of observed
local conditions. In a similar fashion, Equation 3.7 can also be adjusted by the calibration factor for
asphalt pavements.

Page | 15
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Figure 3.3: SNCi/SNC0 vs AGEi (granular) variation with calibration, ks

SNCratio variation with climate, TMI

Figure 3.4(a) and Figure 3.4(b) respectively show the impact of climate variation, TMIi, on the
dependent structural deterioration variable, SNCratio, for a sealed unbound granular pavement and
an asphalt pavement with 60 year service lives using Equation 3.6 and Equation 3.7.

Figure 3.4(a) and Figure 3.4(b) also show that, for a 60 year service life, the SNCratio at the end of
service life varies from 0.4 (granular pavements) up to 0.62 (asphalt pavements) over the range of
climate, TMIi, examined. Figure 3.4(a) and Figure 3.4(b) show that the granular pavement is more
sensitive to climate variations than the asphalt pavement.

Figure 3.4(a): SNCi/SNC0 vs AGEi (granular) variation with TMIi


1.00

TMI = 40

TMI = 80
0.90
TMI = 20

0.80
SNCi/SNC0

0.70

0.60

0.50 Sealed unbound granular pavements (n = 278)


SNCi/SNC0 = 2 - EXP[AGEi × (0.00004413 × TMIi + 0.258 × 1/SL)] (r2 = 0.72)
SL = 60 years
0.40
0 10 20 30 40 50 60

AGE (years)

Page | 16
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Figure 3.4(b): SNCi/SNC0 vs AGEi (asphalt) variation with TMIi
1.00
TMI = 40

TMI = 80
0.90
TMI = 20

0.80
SNCi/SNC0

0.70

0.60

0.50 Asphalt pavements (n = 165)


SNCi/SNC0 = 2 - EXP[ AGEi × (0.00001942 × TMIi + 0.2975/SL)] (r2 = 0.79)
SL = 60 years
0.40
0 10 20 30 40 50 60

AGE (years)

SNCratio variation with traffic load, MESA

Figure 3.5(a) and Figure 3.5(b) show, using Equation 3.6 and Equation 3.7 with a constant climate
(TMIi = 20), the impact on the structural deterioration variable, SNCratio, of an increase in annual
traffic load from 1.5 MESA to 1.875 MESA (both at 2.5% growth) reduces the service life from 60
years to 48 years (Equation 3.8).

It can also be seen in Figure 3.5(a) and Figure 3.5(b) that, under the 1.5 MESA traffic load, the
structural deterioration variable, SNCratio, at 60 year service life (SL) is 0.64 for the granular
pavement and 0.62 for the asphalt pavement. When the traffic load increases to 1.875 MESA, the
SL reduces to 48 years (see Equation 3.8) and the structural deterioration variable, SNCratio, at 60
years is 0.54 for the granular pavement and 0.52 for the asphalt pavement. A further increase in
traffic load to 2.125 MESA reduces the SL to 42 years and gives a SNCratio at 60 years of 0.48 for
the granular pavement and 0.44 for the asphalt pavement. The increase in structural deterioration
due to the increased traffic load for both asphalt and granular pavements is the same, although the
SNCratio is slightly lower for the asphalt pavement than the granular pavement under the same
traffic load at the same service life.

Page | 17
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Figure 3.5(a): SNCi/SNC0 vs AGEi (granular) variation with MESA

Figure 3.5(b): SNCi/SNC0 vs AGEi (asphalt) variation with MESA

SNCratio variation with service life, SL

Figure 3.6(a) and Figure 3.6(b) respectively show, using Equation 3.6 and Equation 3.7 with a
constant climate (TMIi = 20) and no change in traffic load, the variation of the structural
deterioration, SNCratio, with varying service life, SL, for a granular pavement and an asphalt
pavement.

It can be seen from Figure 3.6(a) and Figure 3.6(b) that, at the end of service life, the structural
deterioration variable, SNCratio, varies between 0.63 and 0.66. The SNCratio for the asphalt
pavement varies from 0.63 to 0.64 and the granular pavement SNCratio varies from 0.64 to 0.66.
When the climate TMIi value increases, the structural deterioration variable, SNCratio, will decrease
for a given service life. The SNCratio for granular pavements for a given service life and constant
climate is slightly higher than the SNCratio for asphalt pavements, implying the granular pavements
deteriorate less than asphalt pavements.

Page | 18
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Figure 3.6(a): SNCi/SNC0 vs AGEi (granular) variation with SL
1.0

TMI = 20, SL = 60

TMI = 20, SL = 50
0.9
TMI = 20, SL = 40

0.8
SNCi/SNC0

0.7

SL = 40 0.66 0.65 0.64


0.6 SL = 50

SL = 60
0.5

Sealed unbound granular pavements (GN)


0.4
0 10 20 30 40 50 60
AGE (years)

Figure 3.6(b): SNCi/SNC0 vs AGEi (asphalt) variation with SL


1.0
TMI = 20, SL = 60

TMI = 20, SL = 50
0.9
TMI = 20, SL = 40

0.8
SNCi/SNC0

0.7
SL = 40 0.64 0.63 0.63
SL = 50
0.6
SL = 60

0.5

Asphalt pavements (AC)


0.4
0 10 20 30 40 50 60
AGE (years)

3.1.4 End of Structural Life


Using Equation 3.6 and Equation 3.7, the end of structural life is predicted when the pavement
age, AGEi, is equal to the service life, SL, for a given annual traffic load and climate, TMIi.
However, this approach gives different values for structural deterioration, SNCratio, as noted above
and shown in Table 3.3 for a range of service lives and climates. The SNCratio values in Table 3.3
vary from 0.31 (TMIi = 100) to 0.87 (TMIi = -50), with a mid-range value of 0.59.

Page | 19
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Table 3.3: SNCratio when AGEi equal SL (end of life)
Granular pavements Asphalt pavements
Assumed
Road type TMIi = -50 TMIi = 20 TMIi = 100 TMIi = -50 TMIi = 20 TMIi = 100
SL (years)
SNCi/SNC0 SNCi/SNC0 SNCi/SNC0 SNCi/SNC0 SNCi/SNC0 SNCi/SNC0
Freeway 60 0.87 0.64 0.31 0.73 0.62 0.49
High traffic
50 0.84 0.65 0.39 0.72 0.63 0.52
highway
Medium
traffic 40 0.82 0.66 0.46 0.70 0.63 0.54
highway
Low traffic
30 0.79 0.67 0.52 0.69 0.64 0.57
highway

If the end of service life value of SNCratio is limited to 0.59 and applied to all pavements, then
pavements with a predicted SNCratio lower than 0.59 will have a restricted service life, SL, to keep
the SNCratio from falling below 0.59. The restricted SL under these conditions is estimated by
Equation 3.11 and Equation 3.12, assuming that the restricted SL is equal to AGEi.

[ Ln( 2 – 0.59 ) – 0.2581 ]


SL = (granular pavements) 3.11
1/ks × 0.00004413 × TMIi

[ Ln (2 – 0.59 ) – 0.2975 ]
SL = (asphalt pavements) 3.12
1/ks × 0.00001942 × TMIi

Page | 20
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
4 Sealed Local Road RD Models for Structural Deterioration

4.1 Structural Deterioration Models for Gradual Deterioration


4.1.1 Pavement Strength and Traffic Load Capacity
In a similar manner for the arterial roads, the LRDS database for the sealed local road sites was
examined to enable the development of an empirical relationship between the initial modified
structural number dependent variable, SNC0, immediately post construction at zero pavement age
and the cumulative traffic load capacity, CAP, over a defined design life which was assumed to
cover a service life of 30 years with an annual traffic load, MESA, at an annual growth rate of 2.5%.
Values of SNCi were estimated from the maximum deflection, D0, measured using a FWD and
back-calculated (see Section 3.1.3) using a SNCi deterioration relationship to estimate SNC0.

The relationship of the initial modified structural number, SNC0, with the cumulative traffic load
capacity of the pavement, CAP, is as follows:

SNC0 = 2.465 × CAP0.0505 4.1


where
CAP = traffic load capacity over a defined service life
= MESA × CGF
MESA = millions of equivalent standard axles per lane per year
CGF = cumulative growth factor for a defined service life (30 years) at 2.5% annual traffic growth
rate
= (1.02530 – 1)/0.025 = 44.

The Equation 4.1 relationship is shown in Figure 4.1 and compared with the Equation 3.2
relationship between SNC0 and CAP for arterial roads. For most of the typical traffic load capacity
(1.0E+02 to 4.0E+06), the SNC0 value for arterial roads is lower than that for local roads. This
implies that, in order for local roads to achieve the same traffic load capacity as arterial roads, they
need a higher SNC0 value through greater pavement thickness to compensate for generally lower
standards of material quality and construction practice.

It should be noted that the LRDS deflection data was highly varied and the local roads SNC0 and
CAP relationship was produced using very scattered data with the consequence it was not as a
good a fit to the data as the arterial roads SNC0 and CAP relationship.

Page | 21
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Figure 4.1: SNC0 vs design traffic load capacity CAP (local roads)

4.1.2 Structural Pavement Performance


Based on the road segment samples used in the LRDS, a single structural deterioration model was
developed for both sealed unbound granular pavements and thin (≤ 40 mm) asphalt-surfaced
unbound granular pavements using the strength ratio, SNCratio, as the dependent RD model
variable. The relatively thin asphalt-surfaced unbound granular pavements were not likely to be
significantly affected by temperature any more than the thin bituminous sealed unbound granular
pavements.

The road segment samples used for structural RD model development were selected based on the
following criteria:

• the pavement age was greater than zero and less than or equal to 80 years
• the time series for pavement strength, SNCi, showed consistent reduction with time and
therefore the structural deterioration, SNCratio (= SNCi/SNC0), was less than or equal to one.
The range of the variables used in developing the dependent variable strength ratio, SNCratio, RD
model for sealed local roads is shown in Table 4.1.

Table 4.1: Summary of data variables used in model development


AGE MESA
SL (years) TMI ADT (veh/day) SNCi
(years) (ESA x 106/lane/year)
30 – 50 1 – 79 -36 – 78 9 – 18 600 0.0002 – 0.147 2.04 – 10.22

Structural deterioration model for pavements

The strength ratio, SNCratio, RD model was initially developed by Martin et al. (2013) and
subsequently modified to ensure that an initial boundary condition was met, i.e. the SNCratio was
unity at zero pavement age (SNCi = SNC0).

Page | 22
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
The resulting deterioration model is defined in Equation (4.2):

SNCratio = 2 – EXP[1/ks × AGEi × (0.222/SL + 0.047 × MESA) ] (granular and asphalt pavements) 4.2
where
MESA = millions of equivalent standard axles per lane per year
AGEi = pavement age at time ‘i’ (years)
ks = structural deterioration calibration factor (default = 1.0)
SL = service life of pavement (years).

Equation 4.2 predicts that structural deterioration starts once the pavement age, AGE, exceeds
zero. This equation depends on the pavement age, AGEi, the assumed service life, SL, and the
annual traffic load, MESA. The climate, in terms of TMIi, was an independent variable considered
for inclusion in Equation 4.2, but was found not to be statistically significant. The goodness of fit, r2,
of Equation 4.2 to the observational data was a high value of 0.91 for the sample size of 363. This
equation and all its independent variables were statistically significant.

Figure 4.2(a) shows, using Equation 4.2 with a constant service life (SL = 50), the variation of the
dependent structural deterioration variable, SNCratio, with varying traffic load (MESA = 0.005 to
0.1). Figure 4.2(b) shows, using Equation 4.2 with a constant traffic load (MESA = 0.05), the
variation of the structural deterioration, SNCratio, with varying service life (SL = 30 to 50).

It can also be seen in Figure 4.2(a) that Equation 4.2 is sensitive to changes in traffic load, MESA.
At the end of the service life of 50 years, the structural deterioration, SNCratio, varies between 0.42
and 0.74, with lowest value of SNCratio at the highest value of MESA (= 0.1). It can also be seen in
Figure 4.2(b) that Equation 4.2 is not quite as sensitive to changes in service life from 50 to 30
years where the SNCratio varies between 0.60 and 0.66 at the end of service life.

The sensitivity of Equation 4.2 to changes in traffic load implies that local roads are relatively
poorer performers compared to the arterial roads whose structural deterioration performance is
based on Equation 3.6 and Equation 3.7.

Figure 4.2(a): SNCi/SNC0 vs AGEi variation with MESA


1.0
MESA = 0.1

MESA = 0.005
0.9
MESA = 0.05

0.8
SNCi/SNC0

0.7

0.6

Sealed local road pavements (n = 363)


0.5
SNCi/SNC0 = 2 - EXP[ AGEi × (0.222/SL + 0.047 × MESA)] (r2 = 0.91)
SL = 50 years
0.4
0 10 20 30 40 50

AGE (years)

Page | 23
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Figure 4.2(b): SNCi/SNC0 vs AGEi variation with SL
1.0
SL = 30

SL = 50
0.9
SL = 40

0.8
SNCi/SNC0

0.7

0.6

Sealed local road pavements (n = 363)


0.5
SNCi/SNC0 = 2 - EXP[ AGEi × (0.222/SL + 0.047 × MESA)] (r2 = 0.91)
MESA = 0.05
0.4
0 10 20 30 40 50

AGE (years)

4.1.3 Structural Pavement Performance Using Combined LTPP/LTPPM and LRDS Data
Structural deterioration model

A structural deterioration model using the strength ratio dependent variable, SNCratio, was
developed by combining the LTPP/LTPPM data for arterial roads with the LRDS data for sealed
local roads to give a structural deterioration model based on a wider range of traffic load (0.0003 to
4.27 MESA). The resulting deterioration model is defined as follows:

SNCratio = 2 – EXP[1/ks × AGEi × (0.29169/SL + 0.000556 × MESA)] 4.3

All the variables are as defined previously.

The goodness of fit, r2, of Equation 4.3 to the observational data was 0.66 for the combined sample
size of 641. This model and all its independent variables were statistically significant.

Figure 4.3(a) shows, using Equation 4.3 with a constant service life (SL = 50), minor variation of
the structural deterioration, SNCratio, with a widely ranging traffic load (MESA = 0.05 to 1.0).
Figure 4.3(b) shows, using Equation 4.3 with a constant traffic load (MESA = 0.5), the variation of
the structural deterioration, SNCratio, with varying service life (SL = 30 to 50).

Figure 4.3(a) shows that Equation 4.3 is relatively insensitive to changes in traffic load, MESA. At
the end of the service life of 50 years, the structural deterioration, SNCratio, varies between 0.62
and 0.66, with lowest value of SNCratio at the highest value of MESA (= 1.0). Figure 4.3(b) shows
that Equation 4.3 is much more sensitive to changes in service life from 50 to 30 years where the
SNCratio varies between 0.64 and 0.65 at the end of service life while the rate of structural
deterioration increases with reduced service life.

Equation 4.3 appears to be appropriate for well designed and constructed flexible pavements.

Page | 24
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Figure 4.3(a): SNCi/SNC0 vs AGEi variation with MESA (LTPP/LTPPM and LRDS data)
1.0
MESA = 0.5

MESA = 1.0
0.9
MESA = 0.05

0.8
SNCi/SNC0

0.7

0.6

Sealed road pavements (n = 641)


0.5
SNCi/SNC0 = 2 - EXP[ AGEi × (0.29169/SL + 0.000556 × MESA)] (r2 = 0.66)
SL = 50 years
0.4
0 10 20 30 40 50

AGE (years)

Figure 4.3(b): SNCi/SNC0 vs AGEi variation with SL (LTPP/LTPPM and LRDS data)
1.0
SL = 30

SL = 50
0.9
SL = 40

0.8
SNCi/SNC0

0.7

0.6

0.5 Sealed road pavements (n = 641)


SNCi/SNC0 = 2 - EXP[ AGEi × (0.29169/SL + 0.000556 × MESA)] (r2 = 0.66)
MESA = 0.5
0.4
0 10 20 30 40 50

AGE (years)

Page | 25
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
5 Arterial Road RD Models for Surface Deterioration

5.1 RD Models for Gradual Deterioration


5.1.1 Permanent Deformation – Rutting
Rutting due to initial densification, R0

Initial densification rutting, R0, of pavements occurs during the first year after their construction due
to trafficking. Initial densification rutting is a function of the pavement strength, SNC0, traffic load,
MESA, and the density of the pavement. Equation 5.1 for R0 is from Morosiuk et al. (2004) and
estimates initial densification rutting (mm) at the end of the first year of trafficking.

( 0.09 + 0.0384 × 6.5 × SNC0−1.6 )


R0 = Krid × 51 740 × (MESA × 106) × SNC0-0.502 × 100-2.3 5.1

where
Krid = calibration coefficient = 1.0 (default value for all seals)
all other terms are as defined previously.

This equation for the initial rut depth, R0, appears to provide reasonable estimates based on its
independent variables.

Total rutting, rut(t)

Total rutting, rut(t), is the observed permanent deformation in the wheel paths of the pavement at
time ‘t’. As Figure 5.1 shows, total rutting is the sum of the initial densification rutting, R0, and the
cumulative rutting, ∑∆rut(t), at time ‘t’.

Figure 5.1: Total rut vs pavement age

Page | 26
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Therefore, the total rut depth, rut(t), at time ‘t’ is expressed as follows:

rut(t) = R0 + ∑∆rut(t) 5.2


where
∑Δrut(t) = cumulative rutting at time ‘t’ (mm)
R0 = initial rut due to densification at time t = 1 (mm).

5.1.2 Cumulative Rutting, ∑Δrut(t)


Cumulative rutting model

The following relationship for cumulative rutting, ∑Δrut(t), was found (Martin & Choummanivong
2010):

∑∆rut(t) = kr × ( AGEi − 1 )0.617 × { 0.022 × ( 100 + TMIi )/SNC0 + 0.594 × MESA 5.3
− 0.000102 × me }
where
∑∆rut(t) = cumulative rut depth (mm) at time at time t = i after initial densification at AGEi = 1
AGEi = number of years ‘i’ since construction or last rehabilitation
TMIi = Thornthwaite Moisture Index at time at time t = i
SNC0 = modified structural number for pavement strength (years) at AGEi = 0 (see Equations 3.9
and 3.10)
me = annualised pavement maintenance expenditure ($/lane-km/year, in Table 5.1 and Table
5.2)
= annual routine maintenance expenditure + annualised periodic maintenance expenditure
(reseals)
MESA = annual traffic load per lane in millions of equivalent standard axles per lane
kr = calibration coefficient for local conditions (default value = 1.0).

The goodness of fit (r2) of Equation 5.3 to the data was a fair value of 0.44, which reflected the high
variability of the observational data. However, Equation 5.3 was statistically significant based on
the likelihood ratio test statistic (Greene 1990).

All the coefficients of the independent variables were statistically significant at the 95% confidence
limit (p < 0.05). The statistical significance of the model and the explanatory power of the
independent variables through their statistical significance are more important than the goodness
of fit of the model to the data (Greene 1990). The rutting RD model is a function of initial pavement
strength, SNC0, traffic load, MESA, pavement age, AGEi, maintenance expenditure, me, and
climate, TMIi.

The me variable is a measure of the intensity and type of maintenance activity applied to the
surface and surrounds of a pavement. It has been shown that increasing me can quantifiably
reduce pavement deterioration (Martin 2010a). The me variable is the sum of the routine
maintenance and periodic maintenance (resealing) expenditures converted to an equivalent annual
uniform expenditure, EAUC.

Page | 27
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Typical ranges of annual routine maintenance expenditure and annual periodic maintenance
expenditures, based on a conversion of resealing costs to equivalent annual uniform costs are
given in Table 5.1 and Table 5.2 respectively.

The me variable can also be estimated from a previous analysis of maintenance expenditure
versus heavy vehicle road use (Martin et al. 2011) as follows:

me = α + 0.00309 × ESA/lane/year 5.4


where
α= routine maintenance cost (increased with traffic load range, Table 5.1)
ESA/lane/year = equivalent standard axles per lane per year (based on the fourth power law).

However, the maximum allowable value of maintenance expenditure, me, estimated from either
Equation 5.4, or Table 5.1 and Table 5.2, must be limited to the following value for use in
Equation 5.3 to prevent the cumulative rutting rate becoming negative:

0.594 × MESA
me(max) = 5.5
0.000102

It is expected that, for lightly-trafficked local roads, Equation 5.5 will be used more frequently.

Table 5.1: Typical annual routine maintenance costs


Traffic load range, MESA (/lane/year) Annual routine maintenance, rm ($/lane-km/yr)(1)
≤ 0.1 500
0.1 – 1.0 700
> 1.0 1 000
Note:
1. Routine maintenance cost rates are in Australian dollars as at the year 2000.

Because the Australian dollar values of annual routine and periodic maintenance (see Table 5.1
and Table 5.2) are based on the year 2000, they need to be escalated to the current year, using
appropriate escalation factors.

Page | 28
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Table 5.2: Typical annual routine maintenance costs
Typical annual
Scale Rate Reseal cost
Seal description periodic maintenance
factor $/sq.m ($/lane-km(1),(2))
cost ($/lane-km/yr)
SS = single seal 1.00 3.50 12 250 495 – 1 183
DS = double seal 1.75 6.13 21 438 966 – 2 070
GS = geotextile seal 2.00 7.00 24 500 990 – 2 365
SSSR = single seal with
1.40 4.90 17 150 693 – 1 263
scrap rubber
SSPMB = single seal with
1.50 5.25 18 375 881 – 999
polymodified binder
SSSC = single seal with
1.90 6.65 23 275 1 146 – 1 266
surface correction
DSSR = double seal with
2.45 8.58 30 013 1 903
scrap rubber
DSSAMI = double seal with
3.33 11.64 40 731 2 586
strain alleviating membrane
Notes:
1. Assuming an average sealed lane width of 3.5 metres.
2. Resealing rates and periodic maintenance costs are in Australian dollars as at the year 2000.

Impact of calibration on rutting model

Equation 5.3 can be calibrated to suit local observed rates of cumulative rutting. The range of
rutting rates (uncracked) for given values of traffic load, pavement strength, climate and
maintenance expenditure is shown in Figure 5.2.

Figure 5.2: Cumulative rutting variation with calibration, kr

Explanatory variable impact on rutting model predictions

The cumulative rutting model is influenced by pavement maintenance expenditure, me, climate,
TMI, traffic load, MESA, and pavement strength, SNC0, as shown in Figure 5.3(a), Figure 5.3(b),
Figure 5.3(c) and Figure 5.3(d) respectively. Again, it should be noted that these predictions are
only for the gradual deterioration phase that encompasses normal expected levels of service for
pavement conditions.
Page | 29
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Figure 5.3(a) shows the sensitivity of cumulative rutting to the variations surface maintenance
expenditure, me. The cumulative rutting predicted when me is zero is effectively rutting occurring
with cracking after the pavement and seal age is typically between 10 to 15 years. High levels of
me reduce the rutting rate substantially. Figure 5.3(b) shows that the cumulative rutting rate can
vary widely from a low rate in a dry climate (TMI = -25) to a high rate in a wet climate (TMI = 100).

Figure 5.3(c) shows the impact of increased traffic load on a pavement with a given strength,
climate and maintenance expenditure. The cumulative rutting rate varied from 0.06 mm/year when
the traffic load was 0.007 MESA up to a rate of 0.33 mm/year when the traffic load was
2.54 MESA.

Figure 5.3(d) shows the impact of changes in pavement strength for a given traffic load, climate
and maintenance expenditure. The cumulative rutting rate varied from 0.06 mm/year when the
pavement strength, SNC0, was 16.5 up to a rate of 0.18 mm/year when pavement strength, SNC0,
was 3.3.

Figure 5.3(a): Cumulative rutting with variation in me

Page | 30
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Figure 5.3(b): Cumulative rutting with variation in TMI

Figure 5.3(c): Cumulative rutting with variation in MESA

Page | 31
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Figure 5.3(d): Cumulative rutting with variation in SNC0

5.1.3 Annual Incremental Rutting, Δrut(t)i


An estimate of the annual incremental rutting, Δrut(t)i, is needed when there is an annual increase
in traffic load, ΔMESA, associated with traffic growth. Figure 5.4 demonstrates the general
approach used to estimate the annual incremental distress, ΔDil - i, due to an annual increase in
traffic load, ΔMESAi, at time ‘i’. Equation 5.6 is based on this approach.

As shown in Figure 5.4, the total distress under traffic growth at time ‘i’, Di’, is as follows:

Dil = Di + ΔDil - i 5.6


where
Di = total distress under no traffic growth at time at time t = i
ΔDi l
-i = annual incremental distress due to an incremental increase in traffic load, ΔMESA, with traffic
growth at time ‘i’

Figure 5.4: Estimation of annual incremental distress

Page | 32
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
The annual incremental rutting at time t = i, Δrut(t)il - i, due to an annual increase in traffic growth,
ΔMESAi, at time ‘i’ is defined as follows by Equation 5.7 based on Equation 5.3:

∆rut(t)il - i = k × [(AGEi – 1)0.617 × {0.022× (100 + TMIi)/SNC0 + 0.594 × MESAgi – 0.000102 × me} − 5.7

(AGEi – 1)0.617 × {0.022 × (100 + TMIi)/SNC0 + 0.594 × (MESAngi) – 0.000102 × me}]


where
AGEi = number of years ‘i’ since construction or last rehabilitation
MESAgi = annual traffic load per lane in millions of equivalent standard axles in year ‘i’ with traffic
growth
MESAngi = annual traffic load per lane in millions of equivalent standard axles each year ‘i’ with no
traffic growth
ΔMESAi, = MESAgi – MESAngi
TMIi = Thornthwaite Moisture Index for climate pavement conditions at time t = i
kr = calibration coefficient for roughness (default = 1.0).

Total rutting, rut(t)

Under annual traffic growth, ΔMESAi, the total rut value at time ‘i’, rut(t)il, is defined as follows in
Equation 5.8:

rut(t)il = rut(t)i + ∆rut(t)il - i 5.8


where
rut(t)i = rut value at time ‘i’ under no traffic growth
∆rut(t)i l
-i = annual incremental rutting at time ‘i’ due to an annual increase in traffic growth,
ΔMESAi.

The predicted impact of a 1% traffic load growth on rutting for a given pavement is shown in Figure
5.5. The rutting predictions in Figure 5.5 are up to the limit of the gradual deterioration phase for a
given pavement strength, SNC0, and climate, TMI, and maintenance expenditure, me.

Figure 5.5: Predicted rutting increase under traffic load growth

Page | 33
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
5.1.4 Cumulative Surface Cracking, ∑Δcrx(t)
Consistent and objectively measured cracking data, using the former RTA/NSW RoadCrack
measuring equipment, was made available by the Department of Transport, Energy and
Infrastructure (DTEI), South Australia (Martin & Choummanivong 2010). A total of 1675 samples of
cracking data were collected between 1999 and 2004 on a range of arterial roads in South
Australia. These samples covered both sprayed seal and asphalt pavements, although the range
of climate, TMI, was relatively limited. Data of this consistency and quality over a defined time
series were not available from other states or territories at the time of modelling. As a
consequence, it is likely that the following cracking deterioration models will need substantial
calibration for cracking deterioration in other Australian states.

Bitumen seal life

Bitumen seal life, Y, is based on estimating the time it takes for the bitumen viscosity to harden to
the point where it will initiate cracking. Equation 5.9 (Oliver et al. 2010) estimates Y (years) for
crack initiation:

0.158 × TMIN – 0.107 × R + 0.84 2

Y = Kci × 5.9
0.0498 × T – 0.0216 × D – 0.000381 × S2
where
T= (TMAX – TMIN)/2
TMAX = yearly mean of daily maximum temperature (0C)
TMIN = yearly mean of daily minimum temperature (0C)
D= ARRB Durability test (assumed to be 9 years)
S= nominal seal size (stone size, mm)
Kci = calibration factor for crack initiation under local conditions
R= risk factor which varies from 1 (very low risk) to 10 (very high risk)

There is some concern currently that the ARRB Durability test used in Equation 5.9 may not be the
most appropriate variable in estimating bitumen hardening and its potential for surface cracking
due to the now common use of polymer modified binders (PMB) and other additives to extend seal
life. This suggests that some calibration, using Kci, in Equation 5.9 may be needed for local
alignment with observed seal lives.

For the cracking initiation and progression models (Morosiuk et al. 2004) from the Highway Design
and Maintenance (HDM) Standards, the cracking initiation calibration coefficient, Kci, as a rule of
thumb is typically the reciprocal of the crack progression calibration coefficient, kcrx. However, this
has not been validated for the crack initiation and progression models of Equations 5.9 and 5.10.

Cracking deterioration model for sprayed seal pavements

A total of 290 cumulative cracking samples, ∑Δcrx, that were experiencing increased cracking with
time were available for predicting sealed granular pavement cracking. Equation 5.10 was
developed from this data (Martin & Choummanivong 2010):

Page | 34
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
∑Δcrx(t) = (100 − 200 × (1 + EXP((kcrx × 0.234 × crxAGE/((200 − TMIi)/25))3.5 ))-1 ) 5.10
where
∑Δcrx(t) = cumulative cracking (% total lane area) at time t = i
crxAGE = cracking age (elapsed time from the commencement of cracking, years)
= seal age – seal life (Y); see Equation 5.9
TMIi = Thornthwaite Moisture Index at time t = i
kcrx = calibration coefficient for crack deterioration under local conditions (default = 1.0)

The goodness of fit (r2) of Equation 5.10 to the data was not determined which reflected the high
variability of this type of observational data and relatively narrow range of the data available.
However, the independent variable coefficient, crxAGE, was statistically significant with a ‘t’ test
value of 11.7 (p < 0.05).

Figure 5.6(a) shows the variation of cumulative cracking, ∑Δcrx(t), with variation of TMI, whilst
Figure 5.6(b) shows the variation of cumulative cracking, ∑Δcrx(t), with variation of the calibration
coefficient, kcrx, at a constant TMI of 7. It can be seen from Figure 5.6(b) that Equation 5.10 is
sensitive to changes in the calibration coefficient, kcrx, which suggests that Equation 5.10 can fit
various rates of observed cumulative cracking in a given climate, TMI.

Figure 5.6(a): Cumulative cracking with variation in TMI (sealed granular pavements)

Page | 35
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Figure 5.6(b): Cumulative cracking with variation in kcrx (sealed granular pavements)

Cracking deterioration model for asphalt pavements

A total of 1384 cumulative cracking samples, ∑Δcrx(t), that were experiencing increased cracking
with time were available for predicting asphalt pavement cracking. Equation 5.11 was developed
from this data (Martin & Choummanivong 2010):

∑Δcrx(t) = (100 − 200 × (1 + EXP((kcrx × 0.682 × crxAGE/((200 − TMIi)/25))3.5 ))-1) 5.11


where
∑Δcrx(t) = cumulative cracking (% total lane area) at time t = i
crxAGE = cracking age (elapsed time from the commencement of cracking, years)
= seal age – seal life (Y); see Equation 5.9
TMIi = Thornthwaite Moisture Index at time t = i
kcrx = calibration coefficient for local conditions (default = 1.0)

The goodness of fit (r2) of Equation 5.11 to the data was a reasonable value of 0.74, which shows
that the model form suited the wider range of available observational data. Equation 5.11 was also
statistically significant based on the likelihood ratio test statistic, LRTEST.

Figure 5.6(c) shows the variation of cumulative cracking, ∑Δcrx(t), with TMI given by Equation 5.11
and its sensitivity to changes in TMI.

When comparing cracking predictions with variations in TMI as shown in Figure 5.6(a) for sprayed
seals and Figure 5.6(c) for asphalt surfacings, it is evident for the uncalibrated models that the
asphalt is predicted to crack at a greater rate than the sprayed seal.

Figure 5.6(d) shows the variation of cumulative cracking, ∑Δcrx(t), with variation of the calibration
coefficient, kcrx, at a constant TMI of 7. Figure 5.6(d) shows that Equation 5.11 is sensitive to
changes in the calibration coefficient, kcrx, which suggests that Equation 5.11 can fit a range of
observed cumulative cracking in a given climate, TMI, in a similar manner as Equation 5.10.

Page | 36
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Figure 5.6(c): Cumulative cracking with variation in TMI (asphalt pavements)

Figure 5.6(d): Cumulative cracking with variation in kcrx (asphalt pavements)

5.1.5 Cumulative Roughness, ∑ΔIRI(t)


A cumulative roughness deterioration model was developed using a total of 140 cumulative
roughness deterioration dependent variable samples, ∑ΔIRI(t), and their associated independent
variables (Table 2.1) from the LTPP/LTPPM sites. The model was based on the linear addition of
individual contributing cumulative roughness components from rutting, cracking, structural and
environmental deterioration for the gradual deterioration phase as defined by Equation 1.1.

Page | 37
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Cumulative roughness model

The component model for cumulative roughness deterioration, ∑ΔIRI(t), in terms of IRI (m/km), is
defined by Equation 5.12 (Martin & Choummanivong 2010):

∑ΔIRI(t) = kiri × [ 196.74 × STRUC + 0.016 × ∑∆crx + 0.25 × ∑∆rut + 0.972 × ENVIR] 5.12
where
∑∆IRI(t) = cumulative increase in roughness, IRI (m/km), at time t = i from the initial roughness, IRI0, at
zero pavement age, AGE0
STRUC = EXP [m × AGEi] × MESA × AGEi [1 + (SNC0 − 0.0000758 × ∑∆crx × B × S)]-5 (with
cumulative cracking, ∑Δcrx)
= EXP [m × AGEi] × MESA × AGEi [1 + SNC0]-5 (no cracking, ∑Δcrx = 0)
S = nominal maximum size (mm) of seal aggregate
B = factor for estimating the field layer thickness (FLT) of bitumen binder
= 0.6 for single seals
= 0.9 for double seals
∑Δcrx = cumulative percentage (%) area of surface cracking (0 to 100%) contribution to roughness
deterioration (see Equation 5.10 and Equation 5.11)
∑Δrut = cumulative rut depth (mm) after initial densification at AGEi = 1 (see Equation 4.3)
EXP = e raised to the power
ENVIR = m × IRI0 × AGEi
m = environmental coefficient
= 0.0197 + 0.000155 × TMIi (Martin 1996)
TMIi = Thornthwaite Moisture Index for climate pavement conditions at time t = i
AGEi = number of years ‘i’ since construction or last rehabilitation
SNC0 = pavement strength (years) at AGEi = 0
MESA = annual traffic load per lane in millions of equivalent standard axles
IRI0 = initial roughness, IRI (m/km), at zero pavement age (typical range 1.0 to 1.8
kiri = calibration coefficient for roughness (default = 1.0).

The term, STRUC, was derived from the HDM-4 incremental roughness model (Morosiuk et al.
2004) as it was found to be the most appropriate form for the traffic load component for cumulative
roughness.

The total roughness, IRI(t) (m/km), at pavement age, AGEi, is estimated as follows:

IRI(t) = IRI0 + ∑ΔIRI(t) 5.13


where
∑ΔIRI(t) = cumulative IRI, at time ‘t’
IRI0 = initial roughness at time t = 0.

The goodness of fit (r2) of Equation 5.12 to the data was a value of 0.54 which is reasonable
considering the range and variability of the data. Equation 5.12 was statistically significant based
on the likelihood ratio test statistic, LRTEST.

Equation 5.12 and all the coefficients of the independent variables were statistically significant at
the 95% confidence limit (p < 0.05) to provide sound explanatory power for cumulative roughness.
The most appropriate form for the environmental term, ENVIR, was a linear function of the initial

Page | 38
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
roughness, IRI0, as it provided a better fit to the data than alternative non-linear forms which
predict an exponentially increasing environmental contribution to cumulative roughness
deterioration with increasing pavement age. The non-linear form of ENVIR is found in the HDM
roughness component model (Morosiuk et al. 2004).

Impact of calibration on roughness model

Equation 5.12 (for cumulative roughness) can be calibrated to suit local observed rates of
cumulative roughness. Figure 5.7 shows the range of roughness rates (uncracked) for given values
of traffic load, pavement strength, climate and rutting.

Figure 5.7: Cumulative roughness with variation in kiri (sealed granular pavements)

Explanatory variable impact on roughness model predictions

The cumulative roughness model is influenced by pavement maintenance expenditure, me,


climate, TMI, traffic load, MESA, and pavement strength, SNC0, as shown in Figure 5.8(a), Figure
5.8(b), Figure 5.8(c) and Figure 5.8(d) respectively.

Figure 5.8(a) shows the impact of zero maintenance expenditure, me, on roughness which allows
the pavement to crack after 10 years for a constant climate TMI value of 25. Figure 5.8(b) shows
the impact of the variation in climate, TMI, on roughness and again shows the impact of zero
maintenance expenditure, me, on roughness which also allows the pavement to crack after 10
years. In this case the roughness deterioration is greater because the TMI value is 100.

Figure 5.8(c) shows the impact of increased traffic load on roughness for a given value of climate
(TMI = 25) and pavement strength (SNC0 = 5.5). Figure 5.8(d) shows the impact of changes in
pavement strength on roughness for a given value of climate (TMI = 25) and traffic load (MESA =
0.59).

Note that these predictions are only for the gradual deterioration phase that encompasses normal
expected levels of service for pavement conditions.

Page | 39
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Figure 5.8(a): Cumulative roughness with variation in me

Figure 5.8(b): Cumulative roughness with variation in TMI

Page | 40
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Figure 5.8(c): Cumulative roughness with variation in MESA

Figure 5.8(d): Cumulative roughness with variation in SNC0

5.1.6 Annual Incremental Roughness, ΔIRI(t)i


An estimate of the annual incremental roughness, ΔIRI(t)i, is needed when there is an annual
increase in traffic load, ΔMESA, associated with traffic growth. Figure 5.4 demonstrated the
general approach used to estimate the annual incremental distress, ΔDil - i, due to an increase in
traffic load, ΔMESA, at time ‘i’. The total distress Equation 5.6 was based on this approach.

Page | 41
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
The annual incremental roughness at time t = i, ΔIRI(t)il - i, due to an annual increase in traffic
growth, ΔMESAi, at time ‘i’ is defined as follows by Equation 5.14 based on Equation 5.12 for the
general case of a cracked pavement:

∆IRI(t)il - i = kiri × [ 196.74 × STRUC il - i + 0.016 × ∆crxil - i + 0.25 × ∆rut(t)il - i + 0.972 × ENVIRil - i ] 5.14
where
AGEi = number of years ‘i’ since construction or last rehabilitation
MESAgi = annual traffic load per lane in millions of equivalent standard axles in year ‘i’ with traffic
growth
MESAngi = annual traffic load per lane in millions of equivalent standard axles each year ‘i’ with no
traffic growth
ΔMESAi, = MESAgi – MESAngi
TMIi = Thornthwaite Moisture Index for climate pavement conditions at time t = i
STRUCi l
-i = EXP [m × AGEi] × MESAgi × AGEi [1 + (SNC0 − 0.0000758 × ∑∆crx × B × S)]-5
– EXP [m × AGEi] × MESAngi × AGEi [1 + (SNC0 − 0.0000758 × ∑∆crx × B × S)]-5
∆crxil - i = annual incremental cracking = 0 (no cracking case and not a function of ΔMESAi)
∆rut(t)il - i = annual incremental rutting (see Equation 5.7)
ENVIRil - i = annual incremental increase in ENVIR due to ΔMESAi,
= 0 (not a function of ΔMESAi)
kiri = calibration coefficient for roughness (default = 1.0)
all other terms as previously defined.

Total roughness, IRI(t)

Under annual traffic growth, ΔMESAi, the total roughness value at time ‘i’, IRI(t)il, is defined as
follows in Equation 5.15:

IRI(t)il = IRI(t)i + ∆IRI(t)il - i 5.15


where
IRI(t)i = IRI value at time ‘i’ under no traffic growth
∆IRI(t)i l
-i = annual incremental roughness at time ‘i’ due to an annual increase in traffic growth,
ΔMESAi.

Figure 5.9 shows the predicted impact of a 1% traffic load growth on rutting for a given uncracked
pavement. The roughness predictions in Figure 5.9 are up to the limit of the gradual deterioration
phase for a given pavement strength, SNC0, and climate, TMI, and maintenance expenditure, me.

Page | 42
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Figure 5.9: Predicted roughness increase under traffic load growth

5.2 RD Models for Rapid Deterioration


As noted in Section 1.2.2, the data available for RD model development in the rapid deterioration
phase for sprayed seals was limited and therefore the RD models are not fully reliable predictors of
rapid deterioration. However, the RD models are very useful in a context that needs quantification
of the road condition and the ultimate cost consequences of delayed maintenance intervention due
to a lack of maintenance funding. These RD models could be used as part of an operational PMS
where the quantification of rapid deterioration on road conditions is needed.

Equation 1.1 was used to determine the limit of gradual deterioration which varied with the rutting
value predicted. All of the following models are only applicable in the rapid deterioration phase.

5.2.1 Surface Cracking – Sprayed Seals


The cumulative cracking model, ∑Δcrx(t), form adopted for the analysis was as follows and only
includes the independent variables that were statistically significant based on 393 observational
samples (Martin & Choummanivong 2015) collected by the DPTI (SA) from 1997 to 2013:

∑Δcrx(t) = EXP(0.105 × crxAGE) – 1) × EXP[(((100 + TMIi)/100)2.874) × (2.107/SNC0)] 5.16


where
crxAGE = cracking age (elapsed time from commencement of cracking, years)
= seal age – seal life (start of cracking)
TMIi = Thornthwaite Moisture Index for climate pavement conditions at time t = i
SNC0 = pavement strength (years) at AGEi = 0.

Equation 5.16 is not a function of traffic load, but rather it is function of crxAGE, TMIi and SNC0.
This model shows that in the rapid deterioration phase pavement deterioration does not depend on
the traffic load, due to the unstable condition of the pavement as it approaches ultimate failure.
This outcome may be due to the nature of the data which, although highly distressed on specific
short road length segments, was relatively lightly loaded in a relatively benign climate (hot and
dry).

Page | 43
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
The following data selection criteria were used to select the data for the analysis:

 3 ≥ SNC0 ≤ 10
 ∑Δcrx(t) > 0
 crxAGE > 0.

As the models were developed from different data sets, the cumulative cracking predictions at the
boundary of the transition from gradual to rapid deterioration are unlikely to match, so Equation
5.17(a) was used to predict the cumulative cracking, ∑Δcrx(t)r+1, one year after the start of the
rapid deterioration phase:

∑Δcrx(t)r+1 = ∑Δcrx(t)ig + (∑Δcrx(t)igr+1 − ∑Δcrx(t)igr) 5.17(a)


where
∑Δcrx(t)ig = cumulative cracking estimated at the limit of the gradual deterioration phase pavement
AGE, ‘ig’, using Equation 5.10 (sprayed seal) or Equation 5.11 (asphalt)
∑Δcrx(t)igr = cumulative cracking estimated at the above AGE, ‘ig’, for rapid deterioration using
Equation 5.16
∑Δcrx(t)igr + 1 = cumulative cracking estimated at AGE, ‘ig + 1’, in the rapid deterioration phase using
Equation 5.16.

Once cumulative cracking continued after the first year of rapid deterioration, Equation 5.17(b) was
used to predict cumulative cracking, Δcrx(t)r+2, in the second year and subsequent years:

∑Δcrx(t)r+2 = ∑Δcrx(t)r+1 + (∑Δcrx(t)igr+2 − ∑Δcrx(t)igr+1) 5.17(b)


where
∑Δcrx(t)igr + 2 = cumulative rutting estimated at the rapid deterioration phase AGE, ‘ig + 2’, using
Equation 5.16
all other terms are as defined previously.

Figure 5.10 shows the predictions of Equation 5.16 under rapid deterioration with two values of
TMIi (50 and 10) and an SNC0 value of 3.5. Figure 5.10 shows a rapid increase in the rate of
cracking during rapid deterioration, as expected with a greater rate of deterioration with an
increase in TMI from 10 to 50.

Figure 5.10 also shows that the transition from gradual to rapid deterioration ranged from 22 years
(TMI = 50) to 30 years (TMI = 10) due to the impact of a decreased TMI on cracking in Equation
5.16.

Page | 44
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Figure 5.10: Cumulative cracking under rapid deterioration

5.2.2 Permanent Deformation – Rutting


A cumulative rutting model, ∑Δrut(t), was also developed for the rapid deterioration phase. The
model’s functional form was based on past experience and structured so that it satisfied boundary
conditions.

The following data selection criteria were used to select the data for the analysis:
• 3 ≥ SNC0 ≤ 10
• ∑Δcrx(t) > 0
• crxAGE > 0
• ∑Δrut(t) > 0.

The rutting model, is defined by Equation 5.18, and only includes the independent variables that
were statistically significant based on 412 samples (Martin & Choummanivong 2015):

∑Δrut(t) = (EXP(0.037 × AGEi) – 1) × EXP[(((100 + TMIi)/100)8.814) × (1/SNC0)10.56 5.18


× (279.6 × ∑Δcrx(t))]
where
AGEi = number of years ‘i’ since construction or last rehabilitation
TMIi = Thornthwaite Moisture Index for climate pavement conditions at time t = i
SNC0 = pavement strength (years) at AGEi = 0
∑Δcrx(t) = predicted cumulative cracking at time, t (years)
all other terms are as defined previously.

Equation 5.18 is also not a function of traffic load, but it is a function of AGE, ∑Δcrx(t), TMI and
SNC0. This model outcome again shows that in the rapid deterioration phase the deterioration of
the pavement no longer depends on the traffic load, due to the unstable condition of the pavement
as it approaches ultimate failure. This outcome may be due to the nature of the data which,
although highly distressed on specific short road length segments, was relatively lightly loaded in a
relatively benign climate (hot and dry).

Page | 45
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
As the models were developed from different data sets, the cumulative rutting predictions at the
boundary of the transition from gradual to rapid deterioration are unlikely to match, so Equation
5.19(a) was used to predict the cumulative rutting, ∑Δrut(t)r+1, one year after the start of the rapid
deterioration phase:

∑Δrut(t)r+1 = ∑Δrut(t)ig + (∑Δrut(t)igr+1 − ∑Δrut(t)igr) 5.19(a)


where
∑Δrut(t)ig = cumulative rutting estimated at the limit of the gradual deterioration phase pavement
AGE, ‘ig’, using Equation 5.3
∑Δrut(t)igr = cumulative rutting estimated at the above AGE, ‘ig’, for rapid deterioration using
Equation 5.18
∑Δrut(t)igr + 1 = cumulative rutting estimated at AGE, ‘ig + 1’, in the rapid deterioration phase using
Equation 5.18

Once cumulative rutting proceeded after the first year of rapid deterioration, Equation 5.19(b) was
used to predict cumulative rutting, ∑Δrut(t)r+2, in the second year and subsequent years:

∑Δrut(t)r+2 = ∑Δrut(t)r+1 + (∑Δrut(t)igr+2 − ∑Δrut(t)igr+1) 5.19(b)


where
cumulative rutting estimated at the rapid deterioration phase AGE, ‘ig + 2’, using
∑Δrut(t)igr + 2 = Equation 5.18
all other terms are as defined previously.

Figure 5.11 shows the predictions for cumulative rutting over the gradual and rapid deterioration
phases. Note the increase in rutting once cracking occurs in year 10 due to no resealing. Figure
5.11 shows the predicted cumulative rutting in the rapid deterioration phase increasing at an
accelerated rate with AGE as expected for given values of TMIi and SNC0.

Figure 5.11 also shows the same transition from gradual to rapid deterioration as for cumulative
cracking that ranged from 22 to 30 years due to the impact of decreased TMI on rutting in
Equation 5.18.

Figure 5.11: Cumulative rutting under rapid deterioration

Page | 46
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
5.2.3 Roughness
It was expected that the cumulative roughness model, ∑ΔIRI(t), would be the sum of the
roughness contributions of cumulative rutting, cumulative cracking, cumulative environmental
roughness and cumulative traffic-related roughness as is the case in the gradual deterioration
phase.

The following data selection criteria were used to select the data for the analysis:
• 3 ≥ SNC0 ≤ 10
• ∑Δcrx(t) > 0
• crxAGE > 0
• ∑ΔIRI(t) > 0
• ∑Δrut(t) > 0.

A number of model forms and data selection criteria combinations were trialled in the analysis.
Equation 5.20 only includes the independent variables that were statistically significant based on
392 samples (Martin & Choummanivong 2015):

∑ΔIRI(t) = 0.295 × ∑Δrut(t) + (∑Δcrx(t))0.17 + (∑ΔIRI(t)ENVIR)1.288 5.20


where
∑ΔIRI(t) = IRI(t) – IRIo
IRI(t) = lane roughness at time = t (years)
IRI0 = initial roughness (= 2.0 IRI assumed)
∑ΔIRI(t)ENVIR = cumulative environmental roughness
= (0.0197 + 0.000155 × TMI) × IRIo × AGE
all other terms are as defined previously.

Equation 5.20 is a function of crxAGE, AGE, TMI and SNC0 because these variables are used to
predict ∑Δrut(t) and ∑Δcrx(t)) via Equations 5.18 and 5.16 respectively. This outcome again shows
that in the rapid deterioration phase the accelerated deterioration of the pavement does not
depend on the traffic load under these unstable conditions.

Figure 5.12 shows the predictions for cumulative roughness over the gradual and rapid
deterioration phases. Again as the models were developed from different data sets, the cumulative
roughness predictions at the boundary of the transition from gradual to rapid deterioration did not
match so the following relationship was used to predict the cumulative roughness, ∑ΔIRI(t)r+1, one
year after the start of the rapid deterioration phase:

∑ΔIRI(t)r+1 = ∑ΔIRI(t)ig + ( ∑ΔIRI(t)igr+1 − ∑ΔIRI(t)igr ) 5.21(a)


where
∑ΔIRI(t)ig = cumulative roughness estimated at the limit of the gradual deterioration phase
pavement AGE, ‘ig’, using Equation 5.12
∑ΔIRI(t)igr = cumulative roughness estimated at the above AGE, ‘igr’, for rapid deterioration using
Equation 5.20
∑ΔIRI(t)igr + 1 = cumulative roughness estimated at, ‘igr + 1’, in the rapid deterioration phase using
Equation 5.20

Page | 47
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Once cumulative roughness proceeded after the first year of rapid deterioration, the following
relationship was used to predict cumulative roughness, ∑ΔIRI(t)r+2, in the second year and
subsequent years:

∑ΔIRI(t)r+2 = ∑ΔIRI(t)r+1 + ( ∑ΔIRI(t)igr+2 − ∑ΔIRI(t)igr+1 ) 5.21(b)


where
∑ΔIRI(t)igr + 2 = cumulative roughness estimated at the rapid deterioration phase age, ‘igr + 2’, using
Equation 5.20
all other terms are as defined previously.

Figure 5.12 shows the predicted cumulative roughness in the rapid deterioration phase, although it
is not increasing at an accelerated rate with AGE as expected for given values of TMI and SNC0.
This is largely due to the inability to fit a reasonable predictive model to cumulative roughness in
the rapid deterioration phase to the observational data.

Figure 5.12 also shows the same transition from gradual to rapid deterioration that ranged from 22
to 30 years as for cumulative rutting and cracking due to the impact of decreased TMI on rutting
and cracking in Equation 5.18 and Equation 5.16.

Figure 5.12: Cumulative roughness under rapid deterioration

Page | 48
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
6 Sealed Local Road RD Models for Surface Deterioration

6.1 RD Models for Gradual Deterioration


6.1.1 Permanent Deformation – Rutting
The observed rut depth (mm) was the average of the maximum rut depths measured in the inner
and outer wheel path of the lane for a given lane length. The rut depth was measured using a
multi-laser profiler (MLP) with a 2 metre simulated straight-edge across each wheelpath.

Similar to the treatment of structural deterioration for all unbound granular base pavements, the
data sets, whether surfaced with asphalt or sprayed seal, were pooled into a single data set and
analysed together to develop the rutting deterioration model for sprayed seal unbound granular
pavements.

The pavement total rut depth, ruti, as described by Equation 6.1, consists of two components: (i)
the initial densification rutting, Ro, which normally occurs during the initial densification period of
one year after the pavement construction or rehabilitation; and (ii) the cumulative rutting
deterioration, ∑Δruti, which develops after initial rutting to be the remaining part of the observed
total rut depth.

ruti = Ro + ∑Δruti 6.1


where
ruti = observed rut depth (mm) at time i
Ro = initial densification rutting (mm) value at the end of year 1 (see Equation 5.1)
∑Δruti = cumulative change in rut depth since initial densification (mm) at time ‘i’.

Information on the initial densification rut depth, Ro, was not available and therefore the initial rut
depth was estimated using Equation 5.1. The data selected for cumulative rut model development
was based on the following selection criteria:
• the measured rut depth was greater than the estimated initial rut depth (i.e. ruti > Ro)
• the rate of rutting progression was increasing over the monitoring period
• the pavement age was limited to 80 years (0 > AGEi ≤ 80).

A number of independent variables were considered for predicting cumulative rutting deterioration,
∑Δruti. These included the pavement age, AGEi, climate, TMIi, traffic loading, MESA, and initial
pavement strength, SNC0. A variable for the impact of maintenance on rutting was also included by
estimating the annualised expenditure on surfacing works. These independent variables were all
used in a range of pre-formulated models (Martin & Choummanivong 2010) involving linear and
non-linear functions. These functions were trialled and the analysis outputs reviewed.
Consequently, it was found that the following cumulative rutting deterioration model, Equation 6.2,
was the most suitable model fit to the highly variable observed cumulative rutting data, for the
pavement AGEi greater than 1 (Martin et al. 2013):

∑Δruti = kr × 4.003 × [krr × 0.0035 × AGEi + 0.18 × (100 + TMIi)/100 + EXP(5.853 × MESA 6.2
– 0.418 × SNC0)]

Page | 49
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
where
kr = local calibration factor for initial magnitude of rutting (default = 1.0)
krr = local calibration factor for rutting rate (default = 1.0)
AGEi = number of years ‘i’ since construction or last rehabilitation (whichever is the lesser)
SNC0 = initial pavement strength at AGE0 = 0
MESA = annual traffic load per lane in millions of equivalent standard axles
TMIi = Thornthwaite Moisture Index for climate pavement conditions at time t = i
and all other variables are as defined previously.

Although the goodness of fit, r2, was less than 0.1, all coefficients for the independent variable
were significant mainly due to a large sample size of 514. The ‘t’ values indicate that all
independent variables in Equation 6.2 had a similar degree of explanatory power for cumulative
rutting. Equation 6.2 has no independent explanatory variables for surface cracking and
maintenance history, possibly due to the lack of accuracy in estimating these variables.

Equation 6.2 predicts that cumulative rutting increases with pavement age. Most of the observed
cumulative rutting was less than 8 mm. Figure 6.1(a), Figure 6.1(b) and Figure 6.1(c) are plotted
using a calibration factor, krr, equal to 10. This was used because the default value of krr, equal to
1.0 gives a low rutting rate of 0.014 mm/year.

Figure 6.1(a) shows the variation in predicted cumulative rutting with traffic load, MESA. Figure
6.1(b) shows the variation in predicted cumulative rutting with climate, TMI. Figure 6.1(c) shows the
variation in predicted cumulative rutting with the initial pavement strength, SNC0.

From Figure 6.1(a), Figure 6.1(b) and Figure 6.1(c), Equation 6.2 for cumulative rutting appears to
show some sensitivity to all the independent variables, SNC0, MESA and TMI, although it is not
particularly sensitive to traffic load, MESA, as shown by Figure 6.1(a). However, the cumulative
rutting rate (mm/year) from Equation 6.2 is a constant value of 0.14 using the variable values
shown in the Figures 6.1(a) to 6.1(c). This equation is likely to require calibration for most roads for
both the rutting rate and the magnitude of rutting.

Equation 6.1 is used to calculate the total rut depth, ruti.

Figure 6.1(a): Cumulative rutting variation with MESA

Page | 50
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Figure 6.1(b): Cumulative rutting variation with TMI

Figure 6.1(c): Cumulative rutting variation with SNC0

6.1.2 Model for Cracking Deterioration – Thin Bituminous Surfacings


The thin bituminous surfacings for the local roads samples included both sprayed seals and thin
asphalt wearing courses. It was considered that these surfacings behave in a similar fashion with
respect to cracking.

Unlike most equipment-based methods of data collection, the visual assessment of pavement
cracking was carried out manually based on the methods described in (NAASRA 1987). Due to the
subjective nature of this type of fieldwork, however, it was anticipated that some degree of
inconsistency in the cracking data collected would arise from this method. As a result, data
selected for the analysis was examined and filtered through the following selection criteria:
• the cracking was greater than 0.5% (to ensure that cracking had occurred) but limited to
80% of the area observed (to exclude sites with excessive cracking)
• the seal age after crack initiation was less than or equal to 15 years
• the rate of progression of cumulative cracking was positive (cracking deterioration rate > 0).

Despite the stringent selection criteria, a scattered plot of cracking against time after crack initiation
showed no obvious trend of the data, making it difficult to form appropriate models that could fit the

Page | 51
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
data. Consequently, the existing cracking deterioration model form, Equation 5.10, was formulated
for an independent data set consisting of long-term data collected from various highways in
Australia and trialled for this local roads data set (Martin & Choummanivong 2010). This model
form consisted of a combination of non-linear functions (both exponential and power) that produces
a logistic S-curve shape that varied with the value of the power exponent.

At the end of the analysis, Equation 6.3 was found to be the most suitable cumulative cracking
deterioration model, ∑Δcrx, for the full range of cracking up to 100% (Martin et al. 2013):

∑Δcrx = 100 − 200 × [1 + EXP (kc × (crxAGE/((200 − TMI)/25))0.649)] -1 6.3


where
∑Δcrx = cumulative total cracking as a percentage of observed lane area (%)
kc = local calibration factor for cracking (default = 1.0)
crxAGE = cracking age, elapsed time after crack initiation (years)
= age of seal − seal life (estimate)
seal life = see Equation 5.9 (years)
and all other variables are as defined previously.

The selected model form of Equation 6.3 had a very poor fit to the collected data despite a
significant ‘t’ value for the TMI variable and a large sample size of 247.

Figure 6.2 plots cumulative cracking against a range of TMI values. As expected, the highest value
of TMI (= 100) reaches 72% cumulative cracking in 10 years, while the lowest value of TMI (= -50)
reaches 46% cumulative cracking in 10 years.

Page | 52
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Figure 6.2: Cumulative cracking with variation in TMI

6.1.3 Model for Roughness Deterioration


The observed roughness was measured in terms of the International Roughness Index (IRI, m/km)
using the multi-laser profiler (MLP). Roughness for a given lane length was the average of the
inner and outer wheel path roughness values. The total roughness, IRI(t), at any time ‘t’ is as
defined by Equation 5.13 as for sealed arterial roads.

In developing the cumulative roughness model, ∑ΔIRIi, a subset of the observed data was set up
by filtering the main database using the following criteria:
• the pavement age, was greater than zero and less than 80 years (0 > AGEi ≤ 80)
• the field measured roughness, IRIi, was greater than the assumed initial roughness, IRIo,
but limited to a maximum observed value of IRI = 10 m/km
• roughness was increasing over the monitoring period
• the field measured rut depth was greater than initial rut depth (i.e. ruti > Ro)
• cracking had occurred on the pavement surface.

The formulation of the cumulative roughness deterioration model, ∑ΔIRI(t), was based on the
linear additive component model form as used for sealed arterial roads (Martin & Choummanivong
2010). This is shown by Equation 6.4 that predicts the cumulative roughness deterioration with a
set of contributing component variables for rutting, cracking and climatic effects (Martin et al.
2013):

∑ΔIRIi = kiri × [ 1.393 × IRIenv + 0.09 × ∑Δruti + 0.029 × ∑Δcrx] 6.4


where
IRIenv = cumulative roughness due to climatic effects (m/km)
= m × IRIo × AGEi
m= 0.0197 + 0.000155 × TMIi (Martin 1996)
kiri = calibration coefficient for roughness (default = 1.0)
∑Δruti = cumulative rut depth after initial densification at AGEi = 1 (see Equation 6.2)
∑Δcrx = cumulative percentage (%) area of surface cracking (0 to 100%) (see Equation 6.3)
all other variables are as previously defined.

Page | 53
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Equation 6.4 consists of three contributing components: the roughness due climatic effects, the
roughness due to cumulative cracking and the roughness due to cumulative rutting. The climatic
component is a function of initial roughness, IRI0, which needed to be estimated for all sites except
for those where an initial roughness survey was conducted after pavement construction or
rehabilitation.

However, unlike Equation 5.12 for cumulative roughness on arterial roads, Equation 6.4 is not a
function of roughness due to traffic effects, which uses the structural component, STRUC. In
Equation 6.4 traffic effects are accounted for indirectly by the cumulative rutting, ∑Δruti.

With the three components used as input variables in Equation 6.4, the multiple linear regression
analysis outputs showed that all coefficients for the independent variables were highly significant,
as indicated by their ‘t’ value being greater than 1.7. The model also fitted into the selected
observational data reasonably well (r2 = 0.22) for a large sample size of 357.

In order to use Equation 6.4 to predict the cumulative roughness deterioration, an estimate of the
initial roughness, IRI0, is needed for the value of IRIenv. An estimate of IRI0 can be made using
Equation 5.13 together with Equation 6.4 that gives an explicit function for IRI0, assuming there is a
constant annual traffic load, MESA, since construction or the last rehabilitation. This explicit
function is defined by Equation 6.5:

IRIi – ( kiri × 0.09 × ∑Δruti + kiri × 0.029 × ∑Δcrx )


IRI0 = 6.5
( 1 + kiri × 1.393 × m × AGEi )

The estimated value of IRI0 must always be greater than zero and less than IRIi. Typically most
values of IRI0 range from 1.5 to 3.0 IRI for local roads.

It should be noted that Equation 6.4 for cumulative roughness, ∑ΔIRIi, is highly influenced by
climatic effects as indicated by the relative value of the coefficient for the climatic independent
variable compared to the value of the coefficients for the other two independent variables, ∑Δruti
and ∑Δcrx.

Figure 6.3(a), Figure 6.3(b) and Figure 6.3(c) show the variation of cumulative roughness with
climate, TMI, pavement strength, SNC0, and traffic load, MESA, respectively.

Page | 54
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Figure 6.3(a): Cumulative roughness with variation in TMI (cracking after 10 years)

Figure 6.3(b): Cumulative roughness with variation in SNC0

Figure 6.3(c): Cumulative roughness with variation in MESA

Page | 55
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Figure 6.4: Cumulative roughness with cracking and no cracking

Figure 6.3(a) shows the predicted variation of cumulative roughness with climate, TMI, using
Equation 6.4. Because the climate component is the highest contribution to cumulative roughness,
the variation with TMI is observable once cracking commences after 10 years.

Figure 6.3(b) shows the predicted variation of cumulative roughness with pavement strength,
SNC0, using Equation 6.4. Equation 6.4 predicts that the cumulative roughness variation with SNC0
is minimal even after cracking commenced after 10 years.

Figure 6.3(c) shows the predicted variation of cumulative roughness with traffic load, MESA, using
Equation 6.4. Equation 6.4 predicts that the cumulative roughness variation with MESA is minimal
even after cracking commenced after 10 years.

Figure 6.4 shows the predicted variation of cumulative roughness with and without cracking after
10 years using Equation 6.4 which predicts that the cumulative roughness increases substantially
with cracking.

Page | 56
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
7 WE Models

7.1 WE Models for Arterial Roads


7.1.1 Sampled WE Maintenance Treatments
Figure 1.1 shows the stage 2 impact of works effects (WE) with the immediate impact on pavement
condition (and structural capacity) due to the intervention of maintenance and rehabilitation
treatments.

Table 7.1, Table 7.2 and Table 7.3 show the sampling details of three typical maintenance
treatments, including their state of location and the number of samples of roughness (IRI) and
rutting made available for WE models. These samples were derived from data previously collected
under Martin (2007) and Martin et al. (2017).

Table 7.1: Asphalt overlay sample details


Overlay thickness range
State IRI samples (no.) Rut samples (no)
(mm)
Victoria (Vic) 20 – 100 18 -
Tasmania 30 – 50 5 -
Queensland (Qld) 20 – 45 36 9
New South Wales (NSW) 18 – 450 15 -
South Australia 25 – 40 42 30
Vic/Qld 25 – 40 6 6
Total 122 45

Table 7.2: Granular re-sheeting sample details


Re-sheet thickness range
State IRI samples (no.) Rut samples (no)
(mm)
Victoria 100 – 300 21 5
Queensland 180 5 5
New South Wales 200 1 1
South Australia 125 – 275 5 5
Total 32 16

Table 7.3: Mill and replace sample details


Mill and replace
State IRI samples (no.) Rut samples (no)
thickness range (mm)
New South Wales 40 – 320 10 10
NSW/Qld 50 – 80 6 6
Total 16 16

Page | 57
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
7.1.2 Model Development
Simple models linking pre-treatment, or before (subscript b) values of pavement condition variables
representing distress (roughness and rutting) to post treatment condition variables, or after
(subscript a) post treatment were developed to take account of WE treatment attributes by using
Equation 7.1.

distressa = distressb × ( 1 – β1 ) – thick1 × β2 – k 7.1

Equation 7.2 was derived by re-arranging the terms in Equation 7.1. Equation 7.2 was assembled
to estimate change in the magnitude of a distress variable, Δdistress, associated with a WE
treatment. Equation 7.2 is:

Δdistress = distressb − distressa = k + distressb × β1 + thick1 × β2 7.2


where
Δdistress = change in measured distress from before and after treatment
distressb = distress measured before treatment
distressa = distress measured after treatment
k= constant term
thick1 = treatment ‘1’ nominal thickness
β1, β2, = coefficients for independent variables.

Equation 7.2 was subject to multiple variable regression analysis to determine the regression
coefficients, β1 and β2, and constant term, k, which best fit the WE data in Table 7.1, Table 7.2 and
Table 7.3. Minor filtering of the data was undertaken to eliminate samples that did not have any
measured improvement, that is, reduction in road conditions as a result of WE.

Equation 7.2 is based on the measured distress from before and after the WE. There is an inherent
error is this approach if there is appreciable deterioration occurring before and after the WE when
the distress measurements were taken. Where there is significant elapsed time between the
measurements made before and after the WE, it is also a cause of further error in estimating the
change in distress due to WE.

Figure 7.1 illustrates these factors and shows that Equation 7.2 will under-estimate the distress
change due to WE. However, in most sealed pavements the deterioration before and after the WE
is relatively low. Consequently, the inherent error in using Equation 7.2 may not be significant
provided the elapsed time between the measurements is not too long. Further refinement of
Equation 7.2 would be possible if both the time between the before and after measurements and
the rate of deterioration were known. The data did not contain this information with the
consequence that Equation 7.2 will under-estimate the immediate impact of WE on pavement
condition.

In some respects the tendency for Equation 7.2 to under-predict the distress change due to WE
may be compensated for by including only data samples that always had a measurable
improvement in conditions, that is, data that showed a reduction in road conditions as a result of
WE.

Page | 58
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Figure 7.1: Definition of measured and actual distress change due to WE

7.2 WE Model Outcomes


7.2.1 Asphalt Overlays
Roughness reduction (all values of Hs ≥20 & Hs ≤ 450)

Based on the asphalt overlay samples contained in Table 7.1, multiple variable regression
analyses were conducted using the reduction in roughness, ΔIRI, as the dependent variable in the
following form:

ΔIRI = IRIb – IRIa = β1, × IRIb + β2 × Hs + k 7.3


where
IRIb = IRI (m/km) measured before treatment
IRIa = IRI (m/km) measured after treatment
Hs = nominal thickness of asphalt overlay (mm)
k= constant term
β1, β2, = regression coefficients for independent variables, IRIb, and Hs, respectively.

The following relationship for the roughness after an asphalt overlay was determined using all 111
samples of asphalt overlay thickness with a goodness of fit, r2, of 0.58:

IRIa = 0.672 + 0.505 × IRIb – 0.003 × Hs 7.4

The independent variables, IRIb and Hs, were statistically significant. Figure 7.2 plots Equation 7.4
showing substantial reduction in roughness after the before treatment roughness, IRIb, reached
1 IRI. Increasing the asphalt overlay thickness increased the reduction in roughness as shown in
Figure 7.2.

Page | 59
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Figure 7.2: Predicted roughness reduction with asphalt overlay (all Hs values)

Roughness reduction (Hs ≥ 40 & Hs ≤ 200)

The asphalt overlay thickness in this range is more typical of that used in practice. Re-analysis of
76 samples gave the following roughness relationship with a goodness of fit, r2, of 0.48:

IRIa = 0.672 + 0.566 × IRIb – 0.005 × Hs 7.5

Equation 7.5 is similar to Equation 7.4, except that the Hs coefficient is a higher value and its
impact on reducing roughness is greater as shown in Figure 7.3.

Figure 7.3: Predicted roughness reduction with asphalt overlay (Hs ≥ 40 & Hs ≤ 200)

Rutting reduction (all values of Hs ≥ 20 & Hs ≤ 450)

In a similar approach as for roughness reduction, the following relationship for the rutting after an
asphalt overlay was determined using 51 samples with a goodness of fit, r2, of 0.68:

ruta = 0.232 × rutb + 1.046 7.6


where
rutb = rutting (mm) measured before treatment
ruta = rutting (mm) measured after treatment.

The independent variable, rutb, was the only statistically significant variable. Figure 7.4 plots
Equation 7.6; it shows a significant reduction in rutting after the before treatment rutting, rutb,
reached 1.7 mm. Unlike Equations 7.4 and 7.5, Equation 7.6 does not have an independent
variable for the asphalt overlay thickness. Using Equation 7.6, an asphalt overlay is predicted to
reduce rutting more than an asphalt overlay reduces roughness as predicted by Equation 7.5.

Page | 60
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Figure 7.4: Predicted rutting reduction with asphalt overlay (all Hs values)

Rutting reduction (Hs ≥ 40 & Hs ≤ 200 )

The asphalt overlay thickness in this range is more typical of that used in practice. Re-analysis of
28 samples did not result in any improvement in the relationship found by Equation 7.6.

7.2.2 Granular Re-sheeting


Roughness reduction (all values of Hs ≥ 100 & Hs ≤ 300 )

The following relationship for the roughness after a granular re-sheet was determined using 32
samples with a goodness of fit, r2, of 0.65:

IRIa = 1.203 + 0.099 × IRIb 7.7

The independent variable, IRIb, was statistically significant. Figure 7.5 plots Equation 7.7 showing
substantial reduction in roughness, IRIa, after the before treatment roughness, IRIb, reached an IRI
of 1.44.

Figure 7.5: Predicted roughness reduction with granular re-sheeting (all Hs values)

Page | 61
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Rutting reduction (all values of Hs ≥ 100 & Hs ≤ 300 )

In a similar approach as for roughness reduction, the following relationship for the rutting after a
granular re-sheet was determined using 15 samples with a goodness of fit, r2, of 0.56:

ruta = 0.438 × rutb 7.8

The independent variable, rutb, was the only statistically significant variable. Figure 7.6 plots
Equation 7.8 showing a significant reduction in rutting for all values of the before treatment rutting,
rutb. Using Equation 7.8, a granular re-sheet is predicted not to reduce rutting as much as a
granular re-sheet reduces roughness as predicted by Equation 7.7.

Figure 7.6: Predicted rutting reduction with granular re-sheeting (all Hs values)

7.2.3 Mill and Replace (M&R)


Roughness reduction (all values of Hs ≥ 40 & Hs ≤ 320 )

The following relationship for the roughness after a mill and replace was determined using 12
samples:

IRIa = 0.815 × IRIb 7.9

The independent variable, IRIb, was statistically significant, although no goodness of fit, r2, was
determined. Figure 7.7 plots Equation 7.9 showing a reduction in roughness, IRIa, for all values of
the before treatment roughness, IRIb.

Page | 62
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Figure 7.7: Predicted roughness reduction with mill and replace (all Hs values)

Rutting reduction (all values of Hs ≥ 40 & Hs ≤ 320 )

In a similar approach as for roughness reduction, the following relationship for the rutting after a
mill and replace was determined using 11 samples with a goodness of fit, r2, of 0.28:

ruta = 0.465 × rutb 7.10

The independent variable, rutb, was the only statistically significant variable. Figure 7.8 plots
Equation 7.10 showing a significant reduction in rutting for all values of the before treatment rutting,
rutb. Using Equation 7.10, a mill and replace is predicted to reduce rutting more than a mill and
replace reduces roughness as predicted by Equation 7.9.

Figure 7.8: Predicted rutting reduction with mill and replace (all Hs values)

In comparing Figure 7.8 with Figure 7.7, the mill and replace WE produces a much greater impact
on rutting than it does on roughness. This is the reverse of the outcome to that of granular re-
sheeting where re-sheeting has a much greater impact on the reduction in roughness than the
reduction in rutting.

Page | 63
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
8 Recommendations

The deterministic RD and WE models for Australia’s arterial and sealed local roads documented in
this report represent a substantial financial investment in research over some 23 years. These
models have the capacity to improve the decision-making processes of all Australian road
agencies under a variety of distresses, distress phases, conditions of climate and traffic loading
and maintenance regimes.

No models can be claimed to be 100% reliable; however, these RD and WE models, particularly
those deveoped for arterial roads, have a high degree of utility and have contributed to a wider
analytical framework for performance modelling in Australia. Some local calibration of the models
may be needed to account for local factors such as reactive soils, drainage and material variability.

More research in the future can be undertaken to improve the reliability, range and scope of these
RD and WE models for all roads using the complete network-wide suite of road condition
measurements and associated factors that are becoming available with advanced data collection
techniques.

Page | 64
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
9 References

AASHO 1972, AASHO interim guide for design of pavement structures, American Association of State
Highway Officials, Washington, DC, USA.

Box, GE, Hunter, WG & Hunter, JS 1978, Statistics for experimenters: an introduction to design, data
analysis and model building, John Wiley & Sons, New York, NY, USA.

Choummanivong, L & Martin, T 2010, Predicting structural deterioration of pavements at a network level:
interim models, AP-T159-10, Austroads, Sydney, NSW.

Choummanivong, L & Martin, T 2014, Interim road deterioration cracking model during accelerated
deterioration, AP-T259-14, Austroads, Sydney, NSW.

Choummanivong, L & Martin, T 2016, Austroads LTPP/LTPPM study: summary report 2014-15, AP-T305-16,
Austroads, Sydney, NSW.

Cox, JB 1990, ‘Canadian experience in adapting HDM-3 for road management systems’, Australian Road
Research Board (ARRB) conference, 15th, 1990, Darwin, Northern Territory, Australian Road
Research Board, Vermont South, Vic, vol. 15, no. 3, pp. 261-93.

Crawley, A 1997, ‘Long term pavement performance program: reaping large rewards from small
improvements’, TR News, no. 188, pp. 33-7.

Greene, WH 1990, ’Econometric analysis’, Macmillan Publishing, New York, USA.

Hodges, JW, Rolt, J & Jones, TE 1975, The Kenya road transport cost study: research on road deterioration,
Transport and Road Research Laboratory (TRRL) laboratory report 673, TRRL, Crowthorne,
Berkshire, UK.

Jameson, G 2012, Guide to pavement technology: part 2: pavement structural design, 3rd edn, AGPT02-12,
Austroads, Sydney, NSW.

Martin, TC 1996, A review of existing pavement performance relationships, research report ARR 282, ARRB
Transport Research, Vermont South, Vic.

Martin, T 2007, Interim works effects models, AP-R300-07, Austroads, Sydney, NSW.

Martin, TC, 2008, Predicting sealed granular pavement deterioration at a road network level, PhD Thesis,
Department of Civil Engineering, Monash University, Clayton, Victoria.

Martin, T 2010a, ‘Experimental estimation of the relative deterioration of surface maintenance treatments’,
Journal of Transportation Engineering, vol. 136, no. 1, pp. 1-10.

Martin, T 2010b, ‘Experimental estimation of the relative deterioration of flexible pavements under increased
axle loads’, International Journal of Pavement Engineering, vol. 12, no. 1, pp. 37-45.

Martin, T & Choummanivong, L 2010, Interim network level functional road deterioration models, AP-T158-
10, Austroads, Sydney, NSW.

Martin, T & Choummanivong, L 2015, Interim road deterioration models during accelerated deterioration, AP-
T291-15, Austroads, Sydney, NSW.

Martin, T & Hoque, Z 2007, ‘Under-performing pavements: identification, classification, inspection and
causes’, IR-135-07, Austroads, Sydney, NSW.

Martin, T & Moffatt, M 2008, ‘Results of long-term pavement performance questionnaire survey of
practitioner needs’, internal document, ARRB Group, Vermont South, Vic.

Page | 65
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Martin, T, Byrne, M & Aguiar, G 2011, Establishment of a new pavement maintenance database: stage 1
and 2 analysis, AP-R394-11, Austroads, Sydney, NSW.

Martin, T, Choummanivong, L, Hoque, Z & George, M 2006, ‘Development of road condition performance
profiles: model documentation’, contract report, RC4387-1, ARRB Group, Vermont South, Vic.

Martin, T, Choummanivong, L, Thoresen, T, Toole, T & Kadar, P 2013, ‘Deterioration and maintenance of
local roads’, International public works conference, 2013, Darwin, Northern Territory, Institute of Public
Works Engineering Australia (IPWEA), Sydney, NSW, 11 pp.

Martin, T, Thoresen, T & Choummanivong, L 2017, Revised interim works effect models, AP-T322-17,
Austroads, Sydney, NSW.

Mendenhall, W & Sincich, T 1996, A second course in statistics: regression analysis, 5th edn, Prentice-Hall,
New Jersey, USA.

Morosiuk, G, Riley, MJ & Odoki, JB 2004, Modelling road deterioration and works effects, Highway
Development and Management series, vol. 6, ISOHDM Technical Secretariat, University of
Birmingham, UK.

National Association of Australian State Road Authorities 1979, Interim guide to pavement thickness design,
, NAASRA, Sydney, NSW.

National Association of Australian State Road Authorities 1987, A guide to the visual assessment of
pavement condition, National Association of Australian State Road Authorities, NAASRA, Sydney,
NSW.

Oliver, J, Choi, Y & Martin, T 2010, Asphalt and seal life prediction models based on bitumen hardening, AP-
T160-10, Austroads, Sydney, NSW.

Paterson, WDO 1987, Road deterioration and maintenance effects: models for planning and management,
Highway Design and Maintenance Standards Series, Johns Hopkins University Press, Baltimore,
USA.

Roberts, J, Moffatt, M, Martin, T & George, R 2008, Guide to asset management: part 5D: strength,
AGAM05D-08, Austroads, Sydney, NSW.

Robinson, GK 1994, Opinions about research into road maintenance practices, report DMS-D94/88, CSIRO
Division of Mathematics and Statistics, CSIRO, Clayton, Vic.

Smith, RB, Cerecina, B & Peelgrane, M 1996, ‘A methodology for developing maintenance strategies for a
highway system’, Combined 18th ARRB Transport Research conference and Transit New Zealand
Land Transport symposium, Christchurch, New Zealand, ARRB Transport Research, Vermont South,
Vic, vol. 18, no. 4, pp. 371-86.

Thornthwaite, CW 1948, ‘An approach toward rational classification of climate’, Geographical Review, vol.
38, no. 1, pp. 55-94.

Toole, T & Roper, R 2014, ‘Case study of the maintenance needs of the non-urban road corridors of the
National Land Transport network in Victoria’, contract report 007959, ARRB Group, Vermont South,
Vic.

Watanatada, T, Harral, CG, Paterson, WDO, Dhareshwar, AM, Bhandari, A & Tsunokawa, K 1987, The
highway design and maintenance standards model: volume 1: description of the HDM-III model, Johns
Hopkins University Press, Baltimore, USA.

Page | 66
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
APPENDIX A: Summary of LTPP and LTPPM Sites

Table A.1 summarises the long-term pavement performance (LTPP) sites and the long-term
pavement performance maintenance (LTPPM) sites used to develop the arterial road deterioration
models.

Table A.1: Details of LTPP and LTPPM sites


Test
Site TMI Pavement Traffic
State/Territory section Road name (2) Status(5)
(1) type type(3) load(4)

Federal
NS03 ALF Highway, 5 JPCP/LM 0.63 A
Collector
Foreshore
NS17 ALF 42 CRCP/CR 1.45 A
Road, Botany
Pacific
New South NS20 ALF Motorway, 49 Asphalt/M 1.26 A
Wales (NSW) Somersby
LTPP Pacific Highway,
NS24 ALF 37 SS/BCR 2.30 A
Tomago
Pacific Highway,
NS25 ALF 37 SS/BCR 2.30 A
Tomago
Monaro
ARRB2 NA(6) Highway, 19 SS/GS/HL 0.31 A
Cooma
Western Ring
VC01 SHRP Road, 13 Asphalt/CTCR 0.53 I
Tullamarine
Western Ring
VC02 SHRP 13 Asphalt/CTCR 0.53 I
Road, Jacana
Western Ring
VC03 SHRP Road, 13 Asphalt/CTCR 0.41 I
Broadmeadows
Western Ring
VC04 SHRP 14 Asphalt/CTCR 0.41 I
Road, Glenroy
Western Ring
VC05 SHRP 14 Asphalt/CTCR 0.40 I
Road, Fawkner
Victoria LTPP Hume Highway,
VC20 ALF -5 SS/UBCR 0.74 A
Benalla
Metropolitan
VCA01 NA Ring Road, 15 Asphalt/CTCR 0.28 I
Thomastown
Western Ring
VCA02 NA Road, Sunshine 2 Asphalt/CTCR 0.49 I
West
Western Ring
VCA03 NA 5 Asphalt/CTCR 0.44 I
Road, St Albans
Eastern
VCA04 NA Freeway, 22 Asphalt/CTCR 0.22 A
Donvale

Page | 67
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Test
Site TMI Pavement Traffic
State/Territory section Road name (2) Status(5)
(1) type type(3) load(4)

Bruce Highway,
QL02 SHRP 73 Asphalt/CR 0.49 A
Beerburrum
Warrego
Asphalt/CR/CT
QL04 SHRP Highway, 5 0.85 A
SB
Ipswich
Queensland Bruce Highway,
QL13 ALF 70 Asphalt/CTCR 0.58 I
LTPP Beerburrum
Bruce Highway,
QL14 ALF 70 Asphalt/CTCR 0.58 I
Beerburrum
Rainbow Beach
ARRB1 NA Road, Rainbow 28 SS/BEC 0.04 A
Beach
South Eastern
SA02 ALF Freeway, -15 Asphalt/CR 0.02 I
Callington
South Eastern
SA03 ALF Freeway, -15 Asphalt/CR 0.02 I
Callington
South Eastern
SA05 ALF Freeway, -15 Asphalt/CR 0.02 I
South Australia Callington
LTPP
Port River
SA06 NA Expressway, -11 Asphalt/CTSB 0.80 A
Port Adelaide
Dukes Highway,
SA07 NA -24 SS/UBCR 0.71 A
Cooke Plains
South Eastern
SA10 ALF Freeway, -15 Asphalt/CR 0.02 I
Callington
ACT01 – Uriarra Road,
NA 6 SS/UBCR 0.12 A
P1 Stromlo
ACT01 – Uriarra Road,
NA 6 SS/UBCR 0.12 A
C1 Stromlo
ACT02 – Drakeford Drive,
NA 5 Asphalt/UBCR 0.15 A
C2 Kambah
ACT02 – Drakeford Drive,
NA 5 Asphalt/UBCR 0.08 A
C3 Kambah
ACT03 – Isabella Drive,
Australian NA 5 Asphalt/UBCR 0.15 A
C1 Macarthur
Capital Territory
LTPP ACT03 – Isabella Drive,
NA 5 Asphalt/UBCR 0.15 A
C2 Macarthur
ACT04 – Yamba Drive,
NA 5 Asphalt/UBCR 0.17 A
P1 Isaacs
ACT04 – Yamba Drive,
NA 5 Asphalt/UBCR 0.17 A
P2 Isaacs
ACT05 – Yarra Glen,
NA 5 DS/asphalt 0.46 A
C1 Curtin
ACT05 – Yarra Glen,
NA 5 DS/asphalt 0.46 A
C2 Curtin

Page | 68
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
Test
Site TMI Pavement Traffic
State/Territory section Road name (2) Status(5)
(1) type type(3) load(4)

Monaro
ACT06 –
NA Highway, 6 DS/asphalt 0.35 A
P1
Fyshwick
Monaro
ACT06 –
NA Highway, 6 DS/asphalt 0.35 A
P2
Fyshwick
Stud Road,
LTPPM1 LTPPM Dandenong 31 Asphalt/UBCR 0.56 A
North
Western
LTPPM2 LTPPM Highway, -24 SS/UBCR 0.53 A
Victoria Gerang Gerung
LTPPM Princes
LTPPM3 LTPPM Highway West, 13 SS/UBCR 0.41 I
Heywood
South Gippsland
LTPPM4 LTPPM Highway, 3 SS/UBCR 0.11 A
Woodside
Bruce Highway,
LTPPM5 LTPPM 100 SS/UBCR 0.24 A
North of Ingham
Queensland
LTPPM Flinders
LTPPM6 LTPPM Highway, West 12 SS/UBCR 0.21 I
of Townsville
Great Western
NSW
LTPPM7 LTPPM Highway, 40 Asphalt/UBCR 0.56 A
LTPPM
Blacktown
Tasmania Esk Main Road,
LTPPM8 LTPPM 13 Asphalt/CTCR 0.11 A
LTPPM Fingal
Notes:
1 Test section number.
2 Thornthwaite Moisture Index (Thornthwaite 1948).
3 SS: sprayed seal; DS: deep strength; LM: lean mix; CR: crushed rock; UB: unbound; BEC: bitumen emulsion cement; B: bound; M: macadam;
GS: granulated slag; HL: hydrated lime; JPCP: jointed plain reinforced concrete pavement; CRCP: continuously reinforced concrete pavement.
CTCR: cement-treated crushed rock; CTSB: cement-treated subbase.
4 Traffic load in millions equivalent standard axles (MESA)/lane/year.
5 A = Active; I = Inactive.
6 NA = these LTPP sites were not established as part of the former ALF-LTPP or SHRP-LTPP studies, but were later included to broaden the pavement types.

Page | 69
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR 390
APPENDIX B: Experimental Accelerated Load Testing Results

Summary of Maintenance Testing Results for Relative Performance Factors


Table B.1: Summary of maintenance relative performance factors, rpfmrut and rpfmiri, for various treatment ratios and application to the
observed rutting and roughness deterioration rates (sealed granular pavements)
Rutting, ∑Δrut Roughness, ∑ΔIRI
No. SNC β1 × (MESA – 0.009)/log10SNC β2 × (MESA – 0.009)/log10SNC Treatment rpfmiri
Treatment rpfmrut
obser. range ratio
β1 ‘t’ value r2 ‘F’ value β2 ‘t’ value r2 ‘F’ value
21.4 133 12.1 51
Ckdss1 (w) 9 4.2 – 4.7 167.9 0.95 37.7 0.88
(p <0.05) (p <0.05) (p <0.05) (p <0.05) Ckdss (w) 2.06 1.96
10.6 37 6.8 17 Unckdss (w)
Unckdss2 (w)5 8 4.1 – 4.2 81.4 0.86 19.2 0.73
(p <0.05) (p <0.05) (p <0.05) (p <0.05)
7.8 57 5.9 9
Unckdgeo3 (w) 5 4.0 – 4.5 149.5 0.85 88.8 0.75
(p <0.05) (p <0.05) (p <0.05) (p <0.05) Unckdgeo (w) 0.62 0.85
17.7 95 9.5 10 Unckdds (w)
Unckdds4 (w) 7 3.7 – 3.8 240.2 0.95 104.8 0.66
(p <0.05) (p <0.05) (p <0.05) (p <0.05)
14.7 57 9.8 28
Unckdss (d)6 9 4.6 71.8 0.89 15.8 0.80
(p <0.05) (p <0.05) (p <0.05) (p <0.05) Unckdss (d) 0.88 0.82
Unckdss (w)
6.7 8 6.3 8
Unckdds (d) 8 4.3 – 4.8 54.8 0.57 8.8 0.58
(p <0.05) (p <0.05) (p <0.05) (p <0.05) Unckdds (d) 0.76 0.56
Unckdss (d)
Unckdds (w) Unckdds (w) 0.76 0.56
Unckdss (w)
Unckdgeo (w) 0.47 0.48
Unckdss (w)
Unckdgeo (d) 0.47 0.48
Unckdss (d)

Page | 70
Predicting the Performance of Australia’s Arterial and Sealed
Local Roads ARR390
Rutting, ∑Δrut Roughness, ∑ΔIRI
No. SNC β1 × (MESA – 0.009)/log10SNC β2 × (MESA – 0.009)/log10SNC Treatment rpfmiri
Treatment rpfmrut
obser. range ratio
β1 ‘t’ value r 2
‘F’ value β2 ‘t’ value r 2
‘F’ value
Unckdgeo (w) Unckdgeo (w) 0.47 0.48
Unckdss (w)
Notes:
1 Ckdss = cracked single seal (100% cracked).
2 Unckdss = uncracked single seal.
3 Unckdgeo = uncracked geotextile seal.
4 Unckdds = uncracked double seal.
5 (w) = ‘wet’ test.
6 (d) = ‘dry’ test.

Page | 71
Predicting the Performance of Australia’s Arterial and Sealed
Local Roads ARR390
B.1.1 Application of Relative Performance Factors for Maintenance Changes
The relative performance factors for changes in surface maintenance treatment, rpfm were derived from
ALF testing using a half standard single axle with dual tyres. The relative performance factor, rpfm, for
treatment ‘i’ relative to reference treatment ‘j’ can be applied to observed cumulative rutting and
roughness deterioration, ∑Δdj, for the reference treatment ‘j’ to predict the cumulative deterioration, ∑Δdi,
of treatment ‘i’ as follows:

∑Δdi = rpfm × ∑Δdj B1


where
∑Δdi = predicted cumulative rutting and roughness deterioration for treatment ‘i’
∑Δdj = observed cumulative rutting and roughness deterioration for treatment ‘j’
rpfm = relative performance factors for treatment ‘i’ relative to reference treatment ‘j’ (rpfmrut and
rpfmiri)
rpfmrut = relative performance factor for rutting for treatment ‘i’ relative to reference treatment ‘j’
rpfmiri = relative performance factor for roughness for treatment ‘i’ relative to reference treatment ‘j’.

A summary of the relative performance factors for rutting and roughness, rpfmrut and rpfmiri,
respectively for cracked and uncracked single seals and uncracked double seals and geotextile
seals is presented in Table B.1.

Impact of climate on single uncracked seals

On the basis of Equation B1, the cumulative rutting and roughness deterioration, ∑Δdl, of an
uncracked single seal under climatic condition, TMIl, based on the observed cumulative rutting and
roughness deterioration, ∑Δd, of an uncracked single seal under climatic condition, TMI, can be
estimated using the relative performance factors, rpfm unckdss (TMI/d) and rpfm unckdss (TMIl/d)
as follows (Martin 2008):

∑Δdl (TMIl) = rpfm unckdss (TMIl/d) × ∑Δd B2


rpfm unckdss (TMI/d)
where
rpfm unckdss (TMI/d) = relative performance factor for an uncracked single seal under climatic
condition, TMI, as defined by rpfmrut unckdss (TMI/d) and rpfmiri unckdss
(TMI/d) below
rpfm unckdss (TMIl/d) = relative performance factor for an uncracked single seal under climatic
condition, TMIl, as defined by rpfmrut unckdss (TMI/d) and rpfmiri unckdss
(TMI/d) below with TMIl instead of TMI
rpfmrut unckdss (TMI/d) = 1.047 + 0.000933 × TMI (for rutting of uncracked seals)
rpfmiri unckdss (TMI/d) = 1.073 + 0.00147 × TMI (for roughness of uncracked seals)
l
TMI and TMI = Thornthwaite Moisture Index
∑Δd = observed cumulative rutting and roughness

Impact of climate on single cracked seals

Similarly, using Equation B1, the cumulative rutting and roughness deterioration, ∑Δdll, of a
cracked single seal under climatic condition, TMIl, based on the observed cumulative rutting and
roughness deterioration, ∑Δdl, of a cracked single seal under climatic condition, TMI, can be
estimated using the relative performance factors, rpfm ckdss (TMI/d) and rpfm ckdss (TMIl/d) as
follows:

Page | 72
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR390
∑Δdll (TMIl) = rpfm ckdss (TMIl/d) × ∑Δdl B3
rpfm ckdss (TMI/d)
where
rpfm ckdss (TMI/d) = relative performance factor for an cracked single seal under climatic
condition, TMI, as defined by rpfmrut unckdss (TMI/d) and rpfmiri unckdss
(TMI/d) below
rpfm ckdss (TMIl/d) = relative performance factor for an cracked single seal under climatic
condition, TMIl, as defined by rpfmrut unckdss (TMI/d) and rpfmiri unckdss
(TMI/d) below with TMIl instead of TMI
rpfmrut ckdss (TMI/d) = 1.45 + 0.009 × TMI
rpfmiri ckdss (TMI/d) = 1.463 + 0.00927 × TMI.

All other terms are as defined previously.

Impact of climate on double and geotextile seals

No variation of cumulative rutting and roughness deterioration was assumed to occur for
uncracked double seals and geotextile seals with climate (see Table B1). In other words, the
relative performance of uncracked double seals and geotextile seals was considered to be
independent of climate. There were, however, no experiments undertaken to confirm this.

B.2 Relative Performance Factors for Increased Axle Loads


Relative performance factors define the relative deterioration under the ALF experiments for
increased axle loads. These factors were derived to be applied to the observed pavement
deterioration rates monitored at nominated sealed unbound granular pavement LTPP/LTPPM test
sites to predict the deterioration due to changes to the axle loads of heavy vehicles on these
pavements.

Relative performance factors for the impact of increased axle loads (60 kN and 80 kN) above a
given reference axle load (40 kN) were developed from the ALF data using equations that predict
experimental deterioration (rutting and roughness) under increased axle loads, relative to the
deterioration equation under the reference axle load.

The general definition of a relative performance factor, rpfl, under the same number of load cycles
for both the increased load and the reference load, is as follows for the increased axle load
experiments:

rpfl = ∑Δdi (increased wheel load > 40 kN) B.4


∑Δd (reference wheel load = 40 kN)

where ∑Δdi = cumulative distress (rutting and roughness) due to increased wheel load > 40 kN =
cumulative deterioration equation (wheel load > 40 kN); and, ∑Δd = cumulative distress (rutting
and roughness) at reference wheel load = 40 kN = cumulative deterioration equation (wheel load =
40 kN).

The relative performance factor, rpfl, as defined by Equation B.4 relates increased axle load
deterioration to the deterioration under a 40 kN reference wheel load for the half single axle with
dual tyres (Jameson 2012). The deterioration equations due to the wheel loads define the

Page | 73
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR390
predicted increases in deformation (rutting and roughness), with increasing load cycles during the
gradual deterioration phase where it is relatively easy to develop equations on a statistical basis.

The relative performance factor, rpfl, for any wheel load greater than the reference wheel load of
40 kN is found by substituting the wheel load in the deterioration equations in the numerator of
Equation B.4. Where the wheel load is 40 kN, the relative performance factor is unity as this
represents no increase in wheel load.

B.2.1 Relative Performance Factors for Rutting


Based on Equation B.4, the relative performance factor for rutting, rpflrut, under an increased wheel
load is defined as follows:

rpflrut = cumulative rutting equation (60 kN or 80 kN wheel load) B.5


cumulative rutting equation (40 kN reference wheel load)

Using the outcomes of the ALF experiments (Martin 2010b) it can be shown that Equation B.5
reduces to the following equation for the load ratio, LR, between the actual wheel load and the
reference wheel load (40 kN):

rpflrut = (LR)c B.6

where LR = load ratio = wheel load (kN)/reference load (= 40 kN); and, c = exponent (in Equation
B.6) = 2.248 for the Montrose crushed rock used in the ALF experiments.

Equation B.6 shows that, when relative rutting deterioration is predicted by an equation, the relative
performance factors, rpflrut, for rutting are a constant value for a given ratio of actual axle load to
reference axle load for a given strength, SNC, for any number of load cycles during the gradual
deterioration phase. This means that the relative performance (rutting deterioration) during the
gradual deterioration phase depends only on the LR and Equation B.3 is independent of the
number of load cycles for a given pavement.

B.2.2 Relative Performance Factors for Roughness


Based on Equation B.4, the relative performance factor for roughness, rpfliri, under increased wheel
loading is defined as follows:

rpfliri = cumulative roughness equation (60 kN or 80 kN wheel load) B.7


cumulative roughness equation (40 kN reference wheel load)
Roughness is regarded as a fundamental measurement of the functional performance of
pavements Gopinath et al. (1994). Using the outcomes of the ALF experiments (Martin 2010b) it
can be shown that Equation B.7 reduces to the following equation for the load ratio, LR, between
the actual wheel load and the reference wheel load (40 kN):

rpfliri = (LR)f B.8

where f = 3.536 for Montrose crushed rock used in the ALF experiments.

Equation B.8 shows that, when relative roughness deterioration is represented by an equation, the
roughness relative performance factors, rpfliri, are a constant value for a given ratio of actual axle

Page | 74
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR390
load to reference axle load for a given pavement strength, SNC, for any number of load cycles
during the gradual deterioration phase. This means that the relative performance (roughness
deterioration) during the gradual deterioration phase depends only on the LR and Equation B.8 is
independent of the number of load cycles for a given pavement.

Page | 75
Predicting the Performance of Australia’s
Arterial and Sealed Local Roads ARR390
Australian Road Research Board (ARRB)
ABN 68 004 620 651
Victoria Western Australia New South Wales Queensland South Australia
500 Burwood Highway 191 Carr Place 2-14 Mountain St 21 McLachlan St Level 11
Vermont South VIC 3133 Leederville WA 6007 Ultimo NSW 2007 Fortitude Valley QLD 4006 101 Grenfell St
Australia Australia Australia Australia Adelaide SA 5000
P: +61 3 9881 1555 P: +61 8 9227 3000 P: +61 2 9282 4444 P: +61 7 3260 3500 Australia
info@arrb.com.au arrb.wa@arrb.com.au arrb.nsw@arrb.com.au arrb.qld@arrb.com.au P: +61 8 8235 3300
arrb.sa@arrb.com.au

The Australian Road Research Board (ARRB) is creating knowledge for tomorrows transport challenges and solutions for today. We are driving innovation
to deliver an adaptable connected for all of Australia and across the globe.

You might also like