Hybrid remaining useful life prediction method. A case study on railway D-cables

This paper develops a hybrid remaining useful life (RUL) prediction method and explores the feasibility for complex system equipment, using one of transmission equipment D-cables in high-speed railways as an example. RUL prediction is a promising way to reduce high maintenance costs for high-speed railways. However, there is no sufficient actual life-cycle data due to the lack of sensors, and no mature physics-of-failure model of the equipment in high-speed railways, which make it difficult to predict RUL. To solving this problem, firstly the failure modes, mechanisms, and effects of the D-cables are first analyzed, and accelerated life tests are run under different thermal stresses in Ansys to obtain the life-cycle data. Based on the life-cycle data, particle filtering (PF) method predicts the RUL based on Paris-Law model, meanwhile feedforward neural network (FNN) predicts the RUL under the same thermal stress with PF method, finally a hybrid RUL prediction method that combines model-based and data-driven methods is developed. The results are verified using simulation. •A hybrid method that balances model-based and data-driven methods.•Failure modes, mechanisms, causes, and effects of D-cables were analyzed.•Accelerated life tests were designed to obtain lifecycle datasets.•Particle filtering method and feedforward neural network predict RULs.•Hybrid method predicts more accurate RUL predictions under different stresses.