Development of a probabilistic based, integrated pavement management system

Abstract

Accurate prediction of pavement structural and functional deterioration plays an essential role in the pavement management process and investment planning at both project and network levels. The investigation described in this study was primarily concerned with development of systematic concepts of pavement management and other type of infrastructure network management, such as highway bridges, airfield pavements and oil or gas pipelines. At present, there are still research needs for improving on the existing models and developing new methodologies of pavement preformance prediction.

This thesis describes the development of a probabilistic based , integrated pavement management system (PMS), which can assist pavement engineers or haighway agencies to make strategic investment decisions in programming pavement maintenance and rehabilitation (M&R) projects for the preservation of a road network. The system developed has three major components: 1) using non-homogeneous (i.e., time-related) Markovian prediction models to forecast pavement deterioration, 2) employing stochastic theory and Monte Carlo simulation technique to establish the Markovian transition probability matrices (TPMs) for individual pavements, and 3) utilizing cost-effectiveness based prioritization program to select the optimal multi-year pavement M&R projects and action years.

The non-homogeneous Markov prediction models were established through a process of system conversion from deterministic to probabilistic model. The basic process of performing the prediction model system conversion is described. Each element of the time-related Markovian TPMs is calculated using Monte Carlo simulation. The validation and efficiency of the time-related Markov prediction models are demonstrated by a number of example application.

A Bayesian technique is employed to update the predicted TPMs for accurate prediction of pavement deterioration carried out on a yearly basis through observed pavement performance data.

The determination of a set of standardized M&R treatment strategies for the preservation of a road network is based on the time-related Markov prediction models. Each of the standardized M&R strategies is defined in terms of work content, treatment effect, cost and structural improvement on the existing pavement. The main purposes of standardizing M&R treatments are to: a) provide the highway agency with a list of cost-effective alternatives, b) modify efficiently the established TPMs after each M&R treatment is applied, and c) facilitate the cost-effectiveness based system optimization analysis.

Outputs of the non-homogeneous Markov prediction model include a series of time related TPMs, probability distribution vectors of the predicted pavement condition state in each year and pavement dynamic performance graphs for individual pavements. The year-by-year based integer programming is used to determine the optimal M&R projects and investments for the road network preservation. The optimality criterion is to maximize the effectiveness/cost ratio of total selected M&R treatment projects in each programming year. The key feature of the developed optimization model is the ability to integrate a set of standardized M&R treatment strategies with the predicted multi-year pavement performance into the network optimization analysis.

Both the proposed non-homogeneous Markovian prediction model and the integrated performance-treatment optimization model were tested using examples from Ontario pavement network. Reasonable results were produced in comparison with other existing methods.