Optimal threshold-based network-level transportation infrastructure life-cycle management with heterogeneous maintenance actions

Transportation infrastructure life-cycle management deals with maintenance decision making of transportation facilities such as pavement, bridges, and railways under budget constraints. In practice, transportation agencies adopt threshold-based rules for maintenance planning because they are intuitive and easy to implement. However, the thresholds are often determined based on engineering judgment without any systematic approach. Therefore, maintenance budgets cannot be used effectively and facility conditions are not optimized. This research uses hybrid dynamic models to represent threshold-based maintenance for transportation infrastructure in a realistic manner. Hybrid dynamic models combine continuous states such as pavement roughness and age with discrete states such as maintenance history. These models are also capable of considering multiple maintenance actions with heterogeneous effects. Based on facility conditions and maintenance thresholds, corresponding maintenance actions are selected automatically and the facility switches between deterioration modes to reflect the effects of the chosen action. Furthermore, to consider users’ reactions to maintenance actions and accurately predict deterioration for a network of facilities, threshold-based maintenance is formulated as an upper-level problem, and user response is incorporated as a lower-level problem. This leads to a bi-level programming problem where maintenance thresholds are decision variables, which is solved with a modified tabu search algorithm. The proposed methodology is validated with the road network of an urban area and the generated maintenance thresholds are reasonable and robust, which shows that the methodology has great potential to support transportation infrastructure life-cycle management in practice.

► Hybrid dynamic models simulate state-of-the-practice threshold-based maintenance.
► Heterogeneous maintenance actions and maintenance history are represented by multiple modes.
► Models automatically switch between modes based on facility condition and maintenance thresholds.
► User responses to maintenance are considered to predict facility deterioration at the network level.
► A solution method is proposed to solve the bi-level problem and find optimal thresholds.