AbstractThe track is the basis of railway and subway train operations. Under the repeated action of traffic load and the influence of line structure, geographical environment, and other factors, the condition of the track facility deteriorates, affecting driving safety, shortening the service life of the facility, and increasing maintenance costs. Therefore, it is of great significance to investigate the optimization of rail facility network life cycle repair decisions and devise a reasonable repair plan for scientifically controlling the repair cost under the prerequisite of ensuring operational safety. This study focused on optimizing track facility network life cycle repair decisions comprehensively considering the heterogeneity, uncertainty, and linkage of rail facility degradation. The minimum life cycle cost of the facility network was considered as the optimization objective, and an adaptive learning (AL) mechanism-based maximum likelihood estimation (MLE) method was developed. A network-level life cycle repair decision optimization model based on an AL-Markov decision process (AL-MDP) was constructed and solved by using the Lpsolve toolkit in MATLAB. The Beijing Metro facility network, which is composed of small-radius curved rail units, was considered as an example, and based on a simulation of the network state using the Monte Carlo method, the proposed network-level AL-MDP model and network-level MDP model without AL were used to optimize repair decisions with a planning cycle of 10 years. The results were compared and analyzed, demonstrating that the proposed network-level AL-MDP model can effectively improve the quality of facility network repair decisions compared with the MDP model and that it has higher practicability.