Network-Based Optimization System for Pavement Maintenance Using a Probabilistic Simulation-Based Genetic Algorithm Approach

Abstract

Forecasting future pavement performance is one of the main elements of an infrastructure management system. Performance modeling, however, is not a simple task, particularly due to (1) the unforeseen parameters (other than age) that could affect the deterioration rate of an asset, (2) the serious lack of historical data even for known parameters that affect asset deterioration, and (3) the variability in performance behavior even among similar assets. To overcome these challenges and improve performance prediction accuracy, this paper introduces an optimization transition probability matrix (TPM) to be used by a Markov chain (MC) approach to forecasting pavement performance by generating an average deterioration curve using the Long-Term Pavement Performance (LTPP) database. Then, a TPM using the MC approach is developed that optimizes a customized deterioration curve coinciding with the average deterioration curve generated from the historical LTPP data. The proposed model is tested on different pavement sections and demonstrates its ability to predict the performance of pavement using the Pavement Condition Index throughout its service life. Moreover, a multiobjective pavement maintenance optimization problem using genetic algorithms on the network level is introduced in this paper. Two-objective optimization functions, including minimizing the cost of maintenance and maximizing the condition for the utilized road network, are presented. The suggested model will help road engineers provide maintenance plans with the best conditions and lowest costs.