Theoretical framework for incorporating routine maintenance long-term performance into pavement management using Markov chains

Abstract

Abstract this paper proposes three different models for investigating the long-term performance of routine maintenance works using the discrete-time Markov model. The three models are called (M1, M2 & M3) models representing specific forms of the transition probability matrix. The (M1) model only incorporates the routine maintenance variables (Mi,i−1) under the assumption of only one-state upgrade. The (M2 & M3) models incorporate both routine maintenance variables (Mi,i−1), and major rehabilitation variables (Qi,1) under the assumption of upgrade to condition state (1). Two different modelling methods are proposed for pavement long-term performance investigation in the presence of maintenance and rehabilitation (M&R) variables. The first one is the project-level approach as applied to small pavement networks which assumes that all M&R works can be performed in a short-time period. The second one is the network-level approach which requires the M&R works to be spanned over the entire year as applied to large pavement networks. A model cost-effectiveness index is proposed to evaluate the long-term performance of potential M&R schedules under unconstrained annual budget, while M&R variable cost-effectiveness indices are proposed as useful parameters for yielding optimal M&R schedules under constrained annual budget. The sample results indicated the unrealistic performance of the (M1) model, and the usefulness of the (M2 & M3) models in yielding reliable long-term M&R schedules. The unconstrained approach indicated the optimal M&R schedule is the one associated with highest M&R variable values, and the constrained approach revealed that the optimal solutions are dominated by M&R variables with lowest cost-effectiveness indices.