Dynamic Inspection and Maintenance Planning for Cracked Pipes using Reinforcement Learning with Risk-informed Rewards

Pipeline integrity management involves crack inspections and subsequent repair or other operation actions to prevent leakage or burst failures. While inspection and maintenance (I&M) strategies mitigate failure risks, they need to make a balance between failure risk reduction and I&M costs. Given the dependence of repair decisions on inspection outcomes, determining an optimal inspection schedule that minimizes lifecycle costs is necessary. The selection of both optimal repair methods and inspection timings constitute a complex sequential decision-making problem, where repair methods influence future crack states and maintenance costs. To address this challenge, this study develops a dynamic I&M planning framework based on reinforcement learning (RL). The proposed framework focuses on two distinct scenarios in this study. In Scenario I, a dynamic maintenance plan following fixed periodic inspections is derived, which allows operators to choose an optimal repair method from multiple repair methods (e.g., coating, replacement, reducing operating pressure) depending on observed condition of cracked pipe. In Scenario II, a dynamic I&M plan is derived, which allows operators to choose both optimal repair method and the next inspection time depending on the observed condition of the cracked pipe. For the purpose of demonstration, I&M planning for a single cracked pipe joint is considered, and the dynamic I&M planning based on RL is compared with conventional static I&M planning. Findings illustrate the superiority of the RL-based dynamic I&M, which can support decision-making for condition-based maintenance (CBM) to reduce lifecycle costs.