QUT Digital Repository: http://eprints.qut.edu.au/26874

(1)

QUT Digital Repository:

http://eprints.qut.edu.au/26874

Piyatrapoomi , Noppadol and Kumar, Arun and Robertson , Neil and Weligamage , Justin (2004) Investment decision framework for infrastructure asset management:

a probability-based approach. CRC for Construction Innovation, Brisbane

The Participants of the CRC for Construction Innovation have delegated authority to the CEO of the CRC to give Participants permission to publish material created by the CRC for Construction Innovation. This delegation is contained in Clause 30 of the Agreement for the Establishment and Operation of the Cooperative Research Centre for Construction Innovation. The CEO of the CRC for Construction

Innovation gives permission to the Queensland University of Technology to publish the papers/publications provided in the collection in QUT ePrints provided that the publications are published in full. Icon.Net Pty Ltd retains copyright to the

publications. Any other usage is prohibited without the express permission of the

CEO of the CRC. The CRC warrants that Icon.Net Pty Ltd holds copyright to all

papers/reports/publications produced by the CRC for Construction Innovation.

(2)

Investment Decision

Framework for Infrastructure Asset Management: A Probability- based Approach

By: Noppadol Piyatrapoomi and Arun Kumar In Collaboration with

Neil Robertson and Justin Weligamage

Project Partners RMIT University

Queensland Department of Main Roads Queensland Department of Public Works John Holland Pty Ltd

Ove Arup Pty Ltd

Queensland University of Technology University of Western Sydney Project: 2001-010-C

Investment Decision Framework for Infrastructure Asset Management Research Program: C

Delivery and Management of Built Assets

Report No. 2001-010-C/011 June 2004

(3)

LISTS OF TABLES AND FIGURES

Table 1 Calibration factors (Kgp and Kgm) for the annual rates of change in road pavement roughness

Figure 1 Investment Decision-Making Framework for Infrastructure Asset Management

Figure 2 Cumulative distribution of pavement deflection sample data for outer wheel path and inner wheel path for a 92-kilometre national highway of Queensland Figure 3 Cumulative distribution of deflection data set for 1000-meter interval of a 92- kilometre national highway of Queensland

Figure 4 A typical cumulative distribution of budget/cost estimate

Figure 5 Percentage differences between mean budget/cost estimates and the 95^th percentile budget/cost estimates for 5-, 10-, 15-, 20- and 25-year periods

Figure 6 The cumulative probability distribution of annual rate of change in road pavement roughness between the years 2002-03, 2001-02 and 2000-01

Figure 7 Comparison between the cumulative probability distributions of actual and simulated annual change in roughness for pavement thickness

Figure 8 Mean values of each kilometre of a 92-kilometre National highway of Queensland

Figure 9 Standard deviations of each kilometre of a 92-kilometre National highway of Queensland

Figure 10 A typical cumulative distribution of Structural Number representing pavement strength sampled by Latin Hypercube Sampling Technique of one kilometre section for a 92-kilometre national highway

Figure 11 Cumulative distribution of budget/cost estimate for 25-year roadwork cost of a 92-kilometre National highway of Queensland (roadwork includes maintenance and rehabilitation)

Figure 12 An illustration of plotting two budget scenarios into risk map

Figure 13 An illustration of plotting social, environmental economic and political impacts for budget scenario 1

Figure 14 An illustration of plotting social, environmental, economic, and political impacts for budget scenario 2

(5)

PREFACE

This report forms part of a research project titled “Investment Decision Framework for Infrastructure Asset Management” conducted under The Cooperative Research Centre for Construction Innovation (CRC CI) Project No. 2001-010-C. The aim of this project is to develop a systematic investment decision framework for infrastructure asset management that takes into account economic justification, social and environmental consideration.

To accomplish the project goals, a procedure and three research tasks were identified under the current research study and another had been identified for future research. These research tasks include:

Research Task 1: The Development of Optimisation Procedure for Pavement Deflection Data Collection

Research Task 2: The Development of a Methodology for Risk-Adjusted Assessment of Budget Estimates in Road Maintenance and Rehabilitation

Research Task 3: An Assessment of Calibration Factors for Road Deterioration Prediction Models

Future Task: The Development of a Procedure to Incorporate Social, Environmental and other Related Issues in the Decision-Making Framework

This report presents summaries of the methodologies and benefits of Research Tasks 1, 2 and 3 to the industry.

The authors wish to acknowledge the Cooperative Research Centre for Construction Innovation (CRC CI) for their financial support. The authors also wish to thank the staff at Road Asset Management Branch in the Department of Main Roads Queensland for providing technical data and support.

(6)

v

EXECUTIVE SUMMARY

Australia’s civil infrastructure assets of roads, bridges, railways, buildings and other structures are worth billions of dollars. Road assets alone are valued at around A$

140 billion. As the condition of assets deteriorate over time, close to A$10 billion is spent annually in asset maintenance on Australia's roads, or the equivalent of A$27 million per day.

To effectively manage road infrastructures, firstly, road agencies need to optimise the expenditure for asset data collection, but at the same time, not jeopardise the reliability in using the optimised data to predict maintenance and rehabilitation costs.

Secondly, road agencies need to accurately predict the deterioration rates of infrastructures to reflect local conditions so that the budget estimates could be accurately estimated. And finally, the prediction of budgets for maintenance and rehabilitation must provide a certain degree of reliability.

A procedure for assessing investment decision for road asset management has been developed. The procedure includes:

• A methodology for optimising asset data collection;

• A methodology for calibrating deterioration prediction models;

• A methodology for assessing risk-adjusted estimates for life-cycle cost estimates.

• A decision framework in the form of risk map.

1. A Methodology for Analysing Optimal Data Collection

As mentioned, road authorities may need to optimise expenditure for asset data collection. This method is used for analysing optimal data collection.

A case study was conducted using the developed method for identifying optimal intervals for pavement strength data collection for road asset management.

Pavement strength data are usually collected at close intervals of 100 metres or 200 metres, which is an expensive exercise. Road authorities worldwide find it cost- prohibitive to collect data at the network level. The method developed in this project will produce a more cost effective asset data acquisition practice.

The method is used to examine the stochastic properties of road pavement strength data over extensive lengths of road network. The results found that road authorities could reduce strength test sampling rates to 20% to 25% of current practice without losing any statistical relevance for network applications. This means a chosen network strength testing strategy could be achieved at roughly one quarter of current costs, or conversely four to five times the length of data collection investment.

Benefits to the industry

• For Queensland Department of Main Roads, a potential 75% savings on pavement strength data collection costs, or increase data collection length by four times.

(7)

• The method is generic and could be used for analysing optimal data collection for other types of physical infrastructures.

2. A Methodology for Calibrating Deterioration Prediction Models

This method is used for calibrating deterioration prediction models for local conditions.

The variability in road condition data may arise from the variability in climatic conditions, soil conditions, user vehicles, etc. When the prediction functions do not show a strong correlation or relationship with recorded data, these functions provide less confidence in predicting the deterioration rate for local conditions.

This calibration method is based on probability-based assessment. The method has been used in calibrating pavement deterioration prediction models for road networks of 1688 km Bruce Highway, and 1033 km Landsborough Highway in Queensland as case studies.

Benefits to industry

• There is a global need for an improved model calibration procedure in predicting deterioration rates of road assets for local conditions.

• The method yields calibrated models that closely replicate the actual variability in network condition.

• The method is unique in world practice in that it utilises actual variability in asset data in calibrating prediction models.

• The method is generic and could be used for calibrating any types of functions.

3. A Methodology for Risk-Adjusted Assessment of Budget Estimates for Road Maintenance and Rehabilitation

Previous methods for assessing risk-adjusted of errors in budget estimates have been found to be impractical. The method developed in this study is based on Latin Hypercube Sampling Technique. This technique can simulate a small number of data, resulting in a more practical method by substantially reducing input data preparation efforts and computer time, and therefore, overcoming the limitation of previous methods.

Stochastic property of pavement strength at the network level, accurate deterioration prediction models, and the method of risk-adjusted assessment will enable road authorities to reliably assess life-cycle costing for road asset maintenance and rehabilitation.

Benefits to industry

• The method can assess the levels of risk associated with the budget estimate.

For example, road authorities could produce budget estimates for a project life cycle cost with 5% probability of exceeding.

• The method can be used for any critical data input to assess risk association.

Consequently, risk associated with investment decisions can be assessed.

(8)

vii 4. An Investment Decision Framework

An investment decision framework in the form of risk map in which social, environmental and political issues could be incorporated in decision-making has been developed. This risk map can be used as a tool to manage risk and adjust project allocations based on cost-benefit and risk. In risk mapping, the levels of risk can be quantified qualitatively or quantitatively. A conceptual decision-making framework in the form of risk mapping was discussed in which the life-cycle budget/cost was considered in conjunction with social, environmental and political impacts.

In conclusion, the procedure and methodologies developed in this study will yield affordable, comprehensive, relevant, and quality asset data, including accurate prediction models, and analysis tools in assessing life-cycle costing.

(9)

1. INTRODUCTION

Effective investment decision support relies on comprehensive, relevant and quality data of asset conditions; asset condition prediction modelling; and reliability assessment in life-cycle costing.

Research in the first stage of a CRC research project titled “Investment Decision Framework for Infrastructure Asset Management” concentrated on developing a procedure and methodology for effectively assessing life-cycle budgets/costs for road asset management. Figure 1 shows the framework. There are three important tasks in the framework, namely;

Research Task 1: Optimisation of data collection. The method is used for analysing optimal amount of data collection. A case study was conducted to identify optimal intervals for pavement strength data collection. The result showed that for the same budget, pavement strength could be collected fivefold of currently collected data.

Research Task 2: Calibration of road pavement deterioration prediction models. The method is used for calibrating deterioration prediction models of asset condition for local conditions. Accurately predicting the rate of asset deterioration would result in accurately predicting fund allocation for maintenance and rehabilitation.

Research Task 3: Risk-adjusted assessment for life-cycle costing estimates for maintenance and rehabilitation. The method is used for assessing risk of errors in budget/cost estimates.

This report presents concepts and methodologies of these three tasks and links them to the life-cycle budget/cost estimate for road asset management.

A conceptual decision-making framework in the form of risk mapping in which the life- cycle budget/cost investment could be considered in conjunction with social, environmental and political impacts is also presented for illustrating the concept.

(10)

Risk-Adjusted (or Reliability) Assessment

Calculation Tools, (e.g.

Highway Development and

Management Computer Software, HDM4) Assessment of

Calibration Factors for Road Deterioration Prediction Models

Research Task 1:

Research Task 2

Research Task 3

Using this technique, for example pavement strength data collection for the network analysis could be collected at least fivefold of currently collected data of 200m intervals for the same amount of budget.

Model developed for risk and reliability for management decision (potential for use in building and other sectors) Accurate prediction models for

road deterioration prediction

Outcome

(Piyatrapoomi and Kumar, 2004)

Figure 1 Investment Decision-Making Framework for Infrastructure Asset Management Data Collection:

Optimisation of Data Collection

Risk Assessment in

Life Cycle Costing for

Road Pavement

Asset Management

Outcome Outcome

Investment Decision Framework in

the Form of Risk Map

Social, Environmental

Political issues

(11)

2. METHOD OF OPTIMISING & RELIABILITY ASSESSMENT OF ASSET DATA COLLECTION

This section presents a method of optimising asset data collection and reliability assessment in using optimal data for predicting life-cycle budget/cost estimates.

A case study was conducted for identifying optimal intervals for road pavement strength data collection. Currently, pavement strength data were collected at 100m or 200m intervals which is time consuming and expensive. Many road authorities worldwide find it cost-prohibitive to collect data at the network level.

Falling Weight Deflectometer (FWD) deflection tests were used to collect the data (QDMR 2002). The data were tested at 200m spacing for outer and inner wheel paths from a 92km section of national highway in a tropical region of northeast Queensland.

It is hypothesised in the method that “if the statistical characteristics (i.e. mean, standard deviation and probability distribution) of data sets were quantifiable, and if different sets of data possessed similar means, standard deviations and probability distributions, these data sets would produce similar prediction outcomes”.

Optimisation analysis was carried out by eliminating data from the original data set to create new sets, which were in turn, tested to see whether they had similar mean, standard deviation and probability distribution to those found in the original data set.

If the new data set possessed similar mean, standard deviations and probability distribution, the new data set would provide similar prediction outcomes.

Figure 2 shows the cumulative probability distributions of the original data set of 200m spacing intervals for both inner and outer wheel paths. The means and standard deviations for out and inner wheel paths are given below. The probability distributions were log-normally distributed.

Ln(Deflection in microns) = N(6.05, 0.805) for outer wheel path Ln(Deflection in microns) = N(5.95, 0.817) for inner wheel path

After the analysis, Figure 3 shows the cumulative probability distribution of 1000m intervals of the same 92km pavement strength data. The cumulative probability distribution of the data was fitted by a log-normal distribution with the mean and standard deviation of Ln(Deflection in microns) = N(5.913, 0.795).

The results indicate that the mean, standard deviation and the probability distribution of the data set of 1000m intervals are similar to the means, standard deviations and the probability distributions of the data set of 200m intervals. From the hypothesis, it is assumed that these two sets of data would provide similar prediction outcomes.

Next, it is necessary to test the reliability in using the pavement strength data set of 1000m intervals in predicting life-cycle costing.

In the reliability assessment, the term “reliability” is defined as the percentage of discrepancy between the 95^th percentile budget/cost estimates and the budget/cost estimates calculated from the pavement strength data of 1000-metre intervals (Piyatrapoomi et al, Oct. 2004). The 95^th percentile value is a value that is commonly

(12)

4 selected to provide an appropriate level of confidence (Ang and Tang 1975, Billinton and Allan 1992).

Figure 2 Cumulative distribution of pavement deflection sample data for outer wheel path and inner wheel path for a 92-kilometre national highway of Queensland

Figure 3 Cumulative distribution of deflection data set for 1000-meter interval of a 92-kilometre national highway of Queensland

The performance data of 92-kilometre national highway segment was used in the analysis. Maintenance and rehabilitation budget/cost estimates for 5, 10, 15, 20 and 25-year periods were calculated starting from 2003. In the study, Highway Development and Management (HDM-4) System software package was employed in the analysis. HDM-4, developed by the International Study of Highway Development and Management (ISOHDM 2001), is a globally accepted pavement management system.

Cumulative Distribution of Deflection Data

0 0.2 0.4 0.6 0.8 1

0 2 4 6 8 10

Deflection, Ln(micron)

F(x) Outer Wheel Path

Inner Wheel Path

Cumulative Distribution of 1000m-Interval Deflection Data

0 0.2 0.4 0.6 0.8 1 1.2

0 2 4 6 8 10

Deflection, Ln(microns)

F(x) 1000m- Interval Data

Ln(Do)=N(5.913,0.795)

(13)

In this study, only the discrepancy influenced by the pavement strength was considered and compared. Thus, the variability of pavement strength was used in the calculation, while other input variables remained deterministic in the budget/cost estimates. The 95^th percentile budget/cost estimate is obtained from the budget/cost estimates at the 95% probability of occurrence from the probability distribution as illustrated in Figure 4.

A comparison was made, for a range of analysis periods, between life cycle maintenance cost estimates at the 95^th percentile confidence level, and the equivalent life cycle cost estimate obtained using the reduced data set of pavement deflection data obtained at the optimal spacing of 1000 metre intervals.

Figure 5 shows the discrepancies in percentage between the budget/cost estimates at the 95^th percentile and the budget/cost estimates calculated from the optimal pavement deflection data of 1000-metre intervals. The differences between the 95^th percentile budget/cost estimates and the budget/cost estimates calculated from the optimal data of 1000-metre intervals were calculated to be 12.23, 3.58, 2.85, 1.74 and 1.47, per cent for 5- and 10, 15-, 20- and 25-year periods, respectively.

Figure 4 A typical cumulative distribution of budget/cost estimate

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Budget/cost Estimate (A$) Cumulative Probability Distribution, F(x)

Fx of Budgets/ Costs Fx-Theoretical

95th percentile budget/cost 0.95

(14)

6 Differences between the 95th percentile budget estimates and the

budget estimated from the optimal interval of 1000-metre

0%

2%

4%

6%

8%

10%

12%

14%

0 5 10 15 20 25 30

Year of Budget Estimate

% of Difference of Budget Estimates

Figure 5 Percentage differences between mean budget/cost estimates and the 95^th percentile budget/cost estimates for 5-, 10-, 15-, 20- and 25-year periods It can be observed that the discrepancy between the cost estimated from the optimal pavement strength data of 1000-metre intervals and the 95^th percentile for a life-cycle cost of maintenance and rehabilitation is very small (1.47%). In simple terms, this work demonstrated that applying a reduced sample of deflection data collected at the optimal spacing (1000 metre intervals) to a life cycle analysis of future maintenance investment yielded estimates of life cycle cost with acceptable accuracy, while also providing a measure of confidence in the estimate. Reduced sampling plans for deflection data collection have only a marginal impact on the reliability of long term (that is, 10 year or more) cost estimates for maintenance works.

Details of the method can be found in Piyatrapoomi & Kumar, Jun. 2003 and Piyatrapoomi, et al, Oct. 2004).

3. CALIBRATING DETERIORATION PREDICTION MODELS

This section presents a method for calibrating deterioration prediction models for local conditions.

The variability in road condition data may arise from the variability in climatic conditions, soil conditions, user vehicles, etc. However, when the functions do not show a strong correlation or relationship with recorded data, these functions provide less confidence in predicting the deterioration rate for local conditions.

In this study the probability-based method (Ang & Tang 1975) and Monte Carlo simulation technique (Gray & Travers 1978) have been adopted for calibrating road deterioration prediction models. The methodology was used for calibrating road pavement roughness prediction models for Queensland conditions as a case study.

One of such deterioration model for predicting the rate of change in road pavement roughness suggested by The International Study of Highway Development and Management (ISOHDM 2001) is given below:

(15)

∆RI = Kgp (∆RIs + ∆RIc + ∆RIr + ∆RIt) + Kgm x m x RI (1) Where;

Kgp = calibration factor, Default value = 1.0

Kgm = calibration factor for environmental condition

∆RI = total annual rate of change in road pavement roughness

∆RIs = change in roughness resulting from pavement strength deterioration due to vehicles

∆RIc = change in roughness due to cracking

∆RIr = change in roughness due to rutting

∆RIt = change in roughness due to pothole

The last term in the right hand side of the equation takes into account environmental condition.

Where;

Kgm = Calibration factor for environmental condition m = a constant taking into account environmental effects RI = road pavement roughness of the start of the analysis year

Figure 6 illustrates the cumulative probability of annual rates of deterioration in road pavement roughness for three-year periods (i.e. 2000-01, 2001-02, and 2002-03). In this method, the input variables in Equation 1 are expressed in terms of the probability distribution. The rate of change (∆RI) in Equation 1 will result in a probability distribution.

In the calibration, the probability distribution of the rate of change obtained from Equation 1 and the actual rate of change obtained from the recorded data are compared while the calibration factors are adjusted so that the two cumulative probability distributions achieve best fit.

Figure 7 illustrates the result of a comparison between the cumulative probability distributions of the actual rate of change of road pavement roughness and the rate of change of road pavement roughness obtained from Equation 1. Table 1 shows examples of calibration factors (Kgp and Kgm) of Equation 1.

The calibration factors (Kgm) are given for different percentiles that reflect the actual variability of the recorded data. The method yields calibrated models that closely replicate the actual variability in road network condition.

Details of this method can be found in Piyatrapoomi and Kumar, Jun. 2004.

(16)

8 0

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5

Annual Change in Roughness (IRI)

Cumulative Probability (Fx) ^2003-2002

2002-2001 2001-2000

Figure 6 The cumulative probability distribution of annual rate of change in road pavement roughness between the years 2002-03, 2001-02 and 2000-01

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

-6 -4 -2 0

Ln(Total Annual Change in Road Roughness)

Cumulative Probability (Fx)

Actual Annual Rate of Change in Road Roughness Simulated Annual Rate of Change in Road Roughness

Figure 7 Comparison between the cumulative probability distributions of actual and simulated annual change in roughness for pavement thickness

Table 1 Calibration factors (Kgp and Kgm) for the annual rates of change in road pavement roughness

Calibration Factor (Kgm)

Calibration Factor (Kgm) Calibration

Factor (Kgp)

50^th

Percentile 70^th

Percentile 80^th

Percentile 90^th Percentile

0.20 1.0 1.20 1.70 2.90

(17)

4. RISK-ADJUSTED ASSESSMENT IN LIFE-CYCLE BUDGET/COST ESTIMATES

The preceding two sections addressed two important issues in assessing life-cycle budget/cost for asset management, namely;

(a) Optimising data collection for monitoring asset conditions (b) Predicting the rate of change in asset deterioration

When it is affordable, overall current conditions of assets could be monitored and the rate of change in road deterioration is accurately predicted for local condition. The variability of budget/cost estimates arising from the variability of asset conditions is still remaining an important issue.

This section presents a method that takes into account the variability of asset condition in assessing budget/cost estimates.

To demonstrate the methodology, only the variability of pavement strength was considered in the analysis. To assess the effect of the variability of input parameters on the output budget/cost estimates, the simplest method is to simulate representing values from the probability distributions of input variables and assess the variability of the output parameter.

The Latin hypercube sampling technique, as extensively studied by Iman and Conover (1980), appears to provide a satisfactory method for selecting small samples of input variables so that good estimates of the means, standard deviations and probability distribution functions of the output variables can be obtained. A practical method in assessing risk-adjusted budget/cost estimates is to adopt the Latin Hypercube Sampling Technique.

CASE STUDY

The performance data from the 92-kilometre national highway segment located in a tropical region of northeast Queensland in Australia was used in the risk-adjusted assessment of budget/cost estimates for road maintenance and rehabilitation as a case study.

The steps in the analysis are given below:

1. Establish probability distributions and statistical information (means, standard deviations) of input variables. In this case study, only the variability of pavement strength has been statistically quantified.

2. Simulate sampled data from the probability distributions of pavement strength to represent its variability in the analysis.

3. Conduct a series of analyses using Highway Development Management System (HDM-4) to obtain the statistics of the output budget/cost estimates (ISOHDM 2001).

4. Establish probability distributions and determine the statistical information of the output budget/cost estimates.

5. Assess the interested probability of occurrence of the budget/cost estimates from the output probability distributions.

(18)

10 Figures 8 and 9 show the variability expressed in terms of mean and standard deviation of pavement strength for the 92km national highway. The probability distribution of the pavement strength was log-normally distributed.

For the analysis, the 92-kilometer road segment was divided into 92 sections of 1000-meter in length. Each 1000-meter section has its own pavement strength characteristic. For each kilometre, the variability of pavement strength characteristic is represented by the probability distribution of the Structural Number (SN). The variability of the pavement strength was taken into account in the analysis by using the Latin Hypercube sampling technique. In this study, forty data points of the Structural Numbers (SN) for each kilometre were simulated and used in the analysis.

Figure 10 shows a typical cumulative probability distribution of pavement strength sampled by the Latin hypercube sampling technique for one kilometre.

Figure 8 Mean values of each kilometre of a 92-kilometre National highway of Queensland

Figure 9 Standard deviations of each kilometre of a 92-kilometre National highway of Queensland

Plot of Mean Values for Each Kilometre of a 92-kilometre National Highway of Queensland

0 2 4 6 8

0 20 40 60 80 100

Distance (Kilometres) Mean Values of Structural Numbers

Mean Values

Standard Deviations of Each Kilometre of a 92-kilometre National Highway of Queensland

0 0.5 1 1.5 2 2.5 3

0 20 40 60 80 100

Distance (Kilometres) Standard Deviation of Structural Number

Standard Deviations

(19)

Figure 10 A typical cumulative distribution of Structural Number representing pavement strength sampled by Latin Hypercube Sampling Technique of one kilometre section for a 92-kilometre national highway

A series of analyses were conducted to obtain the statistical output of the life-cycle budget/cost estimates. HDM-4 computer software was used for this purpose. Details of the analysis can be found in Piyatrapoomi and Kumar (Aug. 2003).

Figure 11 shows the cumulative distributions budget/costs estimate for maintenance and rehabilitation for a 25-year period. The probability distribution of the budget/cost estimates was log-normally distributed.

It can be seen from Figure 11 that the probability distribution can be explained by forty data points.

Twenty Five-year Budget Estimate for Maintenance and Rehabilitation of a 92km National Highway of Queensland

0 0.2 0.4 0.6 0.8 1 1.2

95 100 105 110 115

Budget Estimate of Roadwork (A$ Million) Cumulative Probability Distribution, F(x)

Fx of Roadwork Costs Fx-Theoretical 0

0.2 0.4 0.6 0.8 1 1.2

0 1 2 3 4 5 6 7

Structural Number (SN)

F(x)

(20)

12 Figure 11 Cumulative distribution of budget/cost estimate for 25-year roadwork cost of a 92-kilometre National highway of Queensland (roadwork includes maintenance and rehabilitation)

The term “risk adjusted” in budget/cost estimate is defined as the budget/cost that is specified with a level of probability of occurrence. Asset managers could also make use of this information to select the budget/cost estimate of an appropriate probability of occurrence or with the probability of occurrence that they are comfortable with.

For example they may choose a budget/cost estimate of 5% probability of exceeding (or the 95^th percentile budget/cost estimate).Figure 11 shows that the budget/cost estimates which have 5% probability of exceeding was calculated to be A$109 million for a 25-year budget/cost estimate.

Details of this method can be found in Piyatrapoomi & Kumar, August 2003.

5. RISK ASSESSMENT INVESTMENT DECISION FRAMEWORK FOR INFRASTRUCTURE ASSET MANAGEMENT

An investment decision framework in the form of risk map in which social, environmental, political issues and other risk related issues could be incorporated in the assessment was introduced by Piyatrapoomi and Kumar (Piyatrapoomi et al 2004, Piyatrapoomi & Kumar, Jan. 2003).

The concept of risk mapping will be discussed in this section and illustrated using the quantitative risk information presented in the preceding section. In order to provide life-cycle budget/cost for maintenance and rehabilitation for the 92km national highway for a 25-year period, there may be two budget scenarios, namely:

Scenario 1: The government could provide a budget of A$115 million Scenario 2: The government could provide a budget of A$ 95 million

The probability distribution of life-cycle budget/cost estimate shown in Figure 11 can be used to assess the probability of occurrence of these two budget scenarios.

These two scenarios can be plotted in the risk map according to the probabilities of occurrences. From Figure 11, the quantitative measure of risk can be calculated as follows:

ℜ = P(1-Pr) Where:

ℜ = quantitative measure of risk expressed in terms of probability Pr = cumulative probability

Figure 12 shows the probabilities of occurrences resulting from the budget of scenario 1 (A$115 million) and from the budget of scenario 2 (A$ 95 million). The probability that the project would be successful with a budget of A$ 95 millions was very low (Pr ≈ 0). The risk was almost certain (ℜ ≈ 1) that the project would not be successful. While the probability that the project would be successful with a budget of A$115 million was very high (Pr ≈ 1). Thus, the risk was low.

(21)

This risk map can be used as a tool to manage risk and adjust project allocations based on cost-benefit and risk. In risk mapping, the levels of risk can be quantified qualitatively or quantitatively. The X-axis is the magnitude of the resultant consequences, which range from being insignificant to highly significant. The Y-axis is the level of risk, which ranges from rare chance to certain chance of occurrence. It could be expressed in terms of quantitative measures of probability of occurrence.

The intolerable region is the region where risks are high and the impact of the consequences is significant. The tolerable region is the region where risks are low and the impact of the resultant consequences is low. The moderate region is the region where risks and the impact of the consequences are at moderate levels.

0.0000 1 0.0001 0.001 0.01 0.1 1

Pro b ab ility ( R is k)

Ins igni fic ant Signific ant Co ns eq uenc e R a re

U n likely L ike ly

Mo d e ra te C e rta in

A $115 m

A $95m

In s ign ifica n t Min o r Mo d e ra te Ma jo r C a ta s to p h y In to le ra b le R e g io n

Mo d e ra te R e g ion

To le ra b le R e g ion

Figure 12 An illustration of plotting two budget scenarios into risk map

Decision-makers may need to incorporate other factors into consideration. These factors may include social, environmental, or political issues. Figures 13 and 14 illustrate this by plotting risks and consequences related to social, environmental and political issues into the risk map for the two budget scenarios.

In Figure 13, by providing a budget of A$115 million for maintenance and rehabilitation of the 92-kilometre National Highway for a life-cycle cost of a 25-year period, risks and consequences on social, environmental and political issues were expected to be low.

Social issues that may involve traffic noises and accident rates could be kept at a minimum level since pavement conditions are kept in good condition. Road users would be happy with the service provided by the government. Political issue might not be a major concern. Environmental risk and consequences would be kept at a

(22)

14 low level resulted from less fuel consumption, less tyre-wear and less energy consumed for vehicle operation, maintenance and repairs.

In Figure 14, by providing a budget of A$95 million for maintenance and rehabilitation, risks and consequences on social, environmental, economic and political issues are expected to be high.

Social issues on accident rates and the level of noise could be high since pavement conditions could not be maintained in good condition. Political issue might become a big issue. Environmental risk and consequences would be high since high fuel consumption due to delays, higher tyre-wear and higher costs for operation, vehicle maintenance and repairs.

0.0000 1 0.0001 0.001 0.01 0.1 1

U n likely L ike ly

A $115 m

In s ign ifica n t Min o r Mo d e ra te Ma jo r C a ta s to p h y S oc ial

E n v.

P olitic al

In to le ra b le R e g io n

Figure 13 An illustration of plotting social, environmental economic and political impacts for budget scenario 1

(23)

0.0000 1 0.0001 0.001 0.01 0.1 1

U n likely L ike ly

S oc ial

A $95m

In s ign ifica n t Min o r Mo d e ra te Ma jo r C a ta s to p h y P olitic al

E n v.

In to le ra b le R e g io n

Figure 14 An illustration of plotting social, environmental, economic, and political impacts for budget scenario 2

Different scenarios would provide an in depth information for decision-makers to trade-off between cost and benefit. Risk mapping provides information on risk levels and consequences of the decision-making.

In Figures 13 and 14, social, environmental and political issues were presented only for illustrating the concept of the risk map. Further assessment needs to be conducted to address risks and consequences of these issues.

Details of the concept of risk map can be found in Piyatrapoomi & Kumar Jan. 2003 and Piyatrapoomi et al 2004.

6. CONCLUSION

This report presented a methodology for assessing life-cycle budget/cost estimates for road asset management. Three gaps have been identified and need to be addressed. These gaps included asset data collection, calibration of road deterioration prediction models and risk-adjusted assessment in budget/cost estimates. The report presented methodologies to address these three issues.

The optimisation and reliability assessment method can be used in analysing optimal investment in asset data collection. By using this method, expenditure on data collection could be optimised or more data could be collected for the same amount of budget.

The calibration method can be used for calibrating deterioration prediction models in predicting deterioration rates of road infrastructure to suit local conditions.

(24)

16 The risk-adjusted assessment in budget/cost estimates can be used in assessing the variability of budget/cost estimates arising from the variability and uncertainty of critical input variables.

In conclusion, the procedure and methodologies developed in this study will provide tools for assessing affordable, comprehensive and quality asset data collection, for adjusting prediction models to reflect local conditions, and for considering risk assessment within life-cycle cost estimates.

The report also presented a risk map concept for investment decision-making in which social, environmental and political issues could be incorporated for consideration.

The methodology and concept could be used for other types of infrastructure asset management investment.

ACKNOWLEDGEMENT

The authors wish to acknowledge the Cooperative Research Centre (CRC) for Construction Innovation for their financial support. The authors also wish to thank staff at Road Asset Management Branch in the Queensland Department of Main Roads in Australia in providing technical data and support. The views expressed in this report are of the authors and do not represent the views of the organisations.

REFERENCES

1. Ang, A. H-S. & Tang, W.H. (1975) Probability Concepts in Engineering Planning and Design: Volume I and Volume II, John Wiley & Sons, Inc., New York, USA.

2. Billinton, R., & Allan., R.N. (1992) Reliability Evaluation of Engineering Systems: Concepts and Techniques. Plenum Press, New York, USA.

3. Harrington, K. H., & Rose, S. R. (1999), Using Risk Mapping for Investment Decisions, Chemical Engineering Press.

4. Imam, R.L., and Conover, W.J. (1980) Small Sample Sensitivity Analysis Techniques for Computer Models, with an Application to Risk Assessment. Communication in Statistic, A9 (17), 1749-1842.

5. ISOHDM (The International Study of Highway Development and Management) (2001), Highway Development Management (HDM4) version 1.3, University of Birmingham, UK.

6. Gray, K. G. and Travers, K. J. (1978) The Monte Carlo Method. Stipes Publishing Company, Illinois, USA.

7. O’Connor, P. D. T. (1985) Practical Reliability Engineering. John Wiley and Sons, New York, USA.

8. QDMR, Road Asset Management Branch (2002) Managing Road Asset Data: Investing in Our Future. Queensland Government Department of Main Roads, Australia.

9. Piyatrapoomi, N., Kumar, A., & Setunge, S. (2004), Framework for Investment Decision-Making under Risk and Uncertainty for Infrastructure Asset Management, in: Bekiaris, E. & Nakanishi, Y.J. Economic Impacts of Intelligent Transportation Systems: Innovation and Case Studies, Research in Transportation and Economics, Volume 8, 193-209, Elsevier Ltd.

(25)

10. Piyatrapoomi, N., Kumar, A., Robertson, N., & Weligamage, J. (Oct. 2004)

‘Reliability of Optimal Intervals for Pavement Strength Data Collection at the Network Level’ In: Proceedings of the 6^th International Conference on Managing Pavements, Oct. 19-24, Brisbane, Queensland, Australia.

11. Piyatrapoomi, N., & Kumar, A. (Jun. 2004) Assessment of Calibration Factors for Road Deterioration Models, CRC CI Report No. 2001-010- C/009, The Cooperative Research Centre for Construction Innovation, Queensland University of Technology, Brisbane, Queensland, Australia.

12. Piyatrapoomi, N., & Kumar, A. (Aug. 2003) A Methodology for Risk- Adjusted Assessment of Budget Estimates in Road Maintenance and Rehabilitation, CRC CI Report No. 2001-010-C/008, The Cooperative Research Centre for Construction Innovation, Queensland University of Technology, Brisbane, Queensland, Australia.

13. Piyatrapoomi, N., Kumar, A., Robertson, N., & Weligamage, J. (Sept.

2003) ‘A Probability-Based Analysis for Identifying Pavement Deflection Test Intervals for Road Data Collection’ In: Proceedings of the International Conference on Highway Pavement Data Analysis and Mechanistic Design Application, Sept. 7-10, Columbus, Ohio, USA, pp.

291-302.

14. Piyatrapoomi, N., & Kumar, A. (Jun. 2003) The Development of Optimisation Procedure for Pavement Deflection Data Collection, CRC CI Report No. 2001-010-C/007, The Cooperative Research Centre for Construction Innovation, Queensland University of Technology, Brisbane, Queensland, Australia.

15. Piyatrapoomi, N., & Kumar, A., (Jan. 2003), Investment Decision-Making under Risk (Reliability) and Uncertainty for Infrastructure Asset Management, CRC CI Report No. 2001-010-C/006, The Cooperative Research Centre for Construction Innovation, Queensland University of Technology, Brisbane, Queensland, Australia.

Author Biography

Noppadol Piyatrapoomi obtained his Ph.D. degree from the University of Melbourne.

He has practiced as a civil and structural engineer for ten years before he joined the CRC research project on Investment Decision-Making Framework for Infrastructure Asset Management. His research interests include the application of risk and reliability in decision-making for infrastructure asset management; assessment of public risk perception on engineering investments; risk and reliability assessment of structures; seismic risk and reliability assessment of structures; the application of an evolutionary method for data analysis. He developed an evolutionary method of data analysis during his Ph. D. study. This method can be used to refine existing functions and develop new formulas by using probability, statistical and risk assessment theories in the analysis. The method provides a more fine-grained analysis and yields more accurate results and better fitness of data than commonly used methods, such as regression or correlation analyses.

QUT Digital Repository: http://eprints.qut.edu.au/26874