Optimization of Wheel Maintenance Strategy for Railway Freight Train Based on Remaining Useful Life Prediction

Abstract

Wheels are essential parts of railway freight cars, thus scheduling reasonable maintenance time can improve transportation efficiency and reduce maintenance cost, while ensuring operation safety. In this paper, an opportunity-centralized maintenance strategy is presented based on Remaining Useful Life (RUL) prediction. First, a degradation model of wheel tread wear is constructed with non-linear Wiener process, while model parameters are estimated using Gibbs sampling algorithm and wheel degradation data. With the updated model, real-time RUL of wheels are predicted. Next, an opportunity-centralized maintenance strategy based on the different failure time of all the wheels is proposed in order to make a reasonable maintenance schedule, which is difficult to set due to the significantly varying individual degradation processes on different wheels. The decision of maintenance strategy is transformed into a multi-objective optimization problem, in which maintenance times and cost are the optimization objectives. A cost function reflecting the wheel operation cost is constructed, while the average cost per time unit is defined as an optimization objective. Furthermore, NSGA-II algorithm is employed to solve the optimization problem. And a semi-randomized method is proposed to increase the diversity of the initial population. Finally, the effectiveness of the presented algorithm is verified using 500-days on-line monitoring data of wheel tread wear in a 54 fixed group of railway freight cars. The proposed method can be applied to maintenance strategy formulation of any equipment with many similar parts.