The maintenance optimization strategy for key components of train bogies based on DQN

Abstract The rapid expansion of subway systems not only enhances convenience in daily life but also presents significant challenges for maintenance. The existing bogie maintenance strategies are characterized by high costs, prolonged maintenance durations, and difficulties in adapting the strategy based on the state of components. To address these issues, this paper develops a deep Q-network (DQN)-based strategy for maintaining key components of train bogies. Firstly, this paper fits the reliability function of the key components of the bogie with a two-parameter Weibull distribution and constructs the imperfect maintenance and cost model of the components based on the age reduction theory. Secondly, a DQN strategy is proposed to obtain a possible solution for the maintenance model, and then the algorithm is optimized to adjust the maintenance strategy of the key components. Finally, this paper validated the proposed strategy using failure statistics from City Metro Line 1. The results show that this method is applicable to the maintenance optimization of different components, and the cost of this method is about 20% lower than that of the traditional genetic algorithm and 16.4% lower than the Particle swarm optimization (PSO) algorithm.