Study Of Bogie Failure Prediction And Maintenance Of Urban Rail Train Based On Reliability Analysis

Bogie is one of the main parts of locomotive, which is critically important for the safety operation of urban rail train. Therefore, it’s necessary to study the reliability of metro bogie. Through the investigation of the line two of a metro, this paper studies the reliability of bogie system of urban rail train, then predicates its failure rate and optimizes the maintenance plan.First, this paper makes a statistical analysis on the bogie failure data of line two, using the method of survival analysis to decide the best life distribution model as well as the reliability function of each subsystem: the bogie frame, the spring device, connecting device, wheel and axle, drive mechanism and the foundation brake rigging. Then by analyzing the reliability of each subsystem, the weak link of bogie is concluded. The results show that, the life distribution model of each bogie subsystem obeys the Weibull distribution and the reliability of wheel set and axle box system as well as spring system is relatively low.Then, optimized neural network models are used to predict the failure of each subsystem of bogie. In this paper, taking the wheel and axle system whose reliability is relatively low as example, BP neural network, genetic algorithm-based RBF neural network and particle swarm algorithm-based RBF neural network are used to establish the failure rate prediction model. The results show that particle swarm algorithm-based RBF neural network can predict the failure rate better and the precision is obviously higher than that of BP and GA-BP neural network model.Finally, this paper uses proportional hazards model to research the bogie system maintenance plan according to conditions. The reliability and failure rate of each subsystem of bogie are used as covariates to build proportional hazard model, screening out significant covariates that effect bogie running safety. Then based on cost optimization, the threshold and the control limits of bogie system maintenance according to condition are concluded.At last, the effectiveness of the model is verified with predicted failure rate as input. It turns out that the maintenance plan output by PHM model is basically consistent with the actual maintenance plan, which provides theoretical foundation for the preventive maintenance of bogie system.