Condition-based maintenance for multi-component systems under spatio-temporal degradation propagation

For complex industrial systems, a commonly witnessed phenomenon is that degradation occurring in one component propagates spatially over time, aggravating the degradation of adjacent components, which is referred to as spatio-temporal degradation propagation. In this paper, we propose a maintenance optimization framework for systems where component degradation is jointly governed by intrinsic generation and propagation processes: the Wiener process is adopted to model the degradation generation process, while a convolution-based model is employed to characterize the degradation propagation process. The maintenance decision problem is formulated as a Markov decision process (MDP), and further transformed into a factored Markov decision process (FMDP) by decomposing the system’s factor structure to reduce computational complexity. We formulate the maintenance optimization problem under an expected discounted cost criterion and derive structural properties of the optimal policy. Numerical examples show that the proposed approach achieves a lower expected discounted cost than alternative policies. It is further demonstrated that the degradation propagation process exerts a significant impact on the optimal maintenance policy, highlighting the importance of preventive maintenance for components strongly affected by propagation effects.