An Optimal Life Cycle Reprofiling Strategy of Train Wheels Based on Markov Decision Process of Wheel Degradation

Wheels are vital components on railway vehicles and may encounter various problems during long-term operation, e.g., wear and defects. The existing preventative reprofiling strategy with fixed reprofiling interval and cutting depth causes huge maintenance costs. In order to reduce life cycle cost (LCC), this paper proposes an optimal life cycle reprofiling (LCR) strategy for risk-based preventative maintenance of train wheels. First, the probabilistic models of wheel wear, reprofiling, and defects are built. Then, Markov decision process is constructed to model a complete wheel degradation process in its life cycle with sequential decision-making of reprofiling. Meanwhile, the reward matrices are designed considering LCC and safety-related costs. Further, a two-loop optimization scheme is proposed to pursue the optimal LCR strategy minimizing LCC. In the inner loop, policy iteration is adopted to pursue the optimal reprofiling policy. In the outer loop, gradient descent with momentum is employed to optimize the overhaul limits. This scheme achieves fast convergence and global optimization of maintenance decision-making for both routine and overhaul reprofiling. Finally, the mechanism of optimal reprofiling policy is exemplified, including the principles of prolonging, protecting, and abandoning. The main trade-offs in cutting depth and overhaul limits provide the feasibility for optimization. A case study and field tests demonstrate that through controlling wheel degradation and life span, LCR strategy can not only reduce the LCC of wheel maintenance, but also improve the reliability and safety of wheel operation.