Comparative study of life-cycle costing for sustainable infrastructure rehabilitation using Markov Decision Process and directed genetic algorithms

The Metro Systems (Subways) usually offer an attractive alternative for mass transit transportation system in most large cities. Such infrastructures require proper maintenance and rehabilitation (M&R) programs in order to maintain them within a sustainable level of operational and safety performance. Traditional Markov Decision Process (MDP) has been widely used to find the optimal M&R decision policy for infrastructures. A drawback of the traditional Markov approach is that it uses a discrete number of states in the analysis as well as a stationary probability transition matrix (TPM). Thus, MDP is based on the Markovian or "memory less" property, which is not necessarily true for aging infrastructures. This paper presents a new generic methodology for Maintenance and Rehabilitation Planning for Public Infrastructure (M&RPPI). The method aims at determining the optimal rehabilitation profile over a desired analysis period (the best type of rehabilitation intervention and its optimal timing). The new method is based on life-cycle costing (LCC) with probabilistic and continuous rating approach. Also, the M&RPPI uses a "dynamic" Markov chain to represent the deterioration mechanism of an infrastructure and utilizes a new directed Genetic Algorithms (GA) in order to find the optimal rehabilitation profile. A case study is presented on the Montreal Metro systems using two types of analysis: (1) with the traditional MDP method and (2) with the newly developed methodology. Results revealed that new method, which generates lower LCC, is found practical in providing a complete M&R Plan over a required study period, compared to a stationary decision policy with the traditional MDP. The GA is found useful in the optimization process where it overcomes the computational difficulties for computing large MDP problems. The new method is beneficial to researchers and practitioners since it provides a step towards a broader infrastructure management system, and capital budgeting problems. It also benefits metro management agencies and enhances the MDP practice by overcoming some downsides of the traditional methodology.