Time-dependent Reliability Analysis Of Key Components For EMU Traction Drive System

With the continuous improvement of EMU related technology,its operation speed and application scale have been greatly improved,and the risks and challenges such as reliability and safety are becoming increasingly prominent,which is directly related to the high-quality development of railway.As the main source of driving force of EMU,the quality characteristics and operation capacity of traction drive system directly determine the safety and reliability of EMU.The traditional reliability analysis of traction drive system fails to consider the influence of performance degradation and load randomness,and the results can not accurately reflect the time-dependent characteristics of structural reliability.Therefore,starting from the failure mode of each component of the traction drive system,this paper determines the key components of the system,establishes its time-dependent reliability model according to the failure criterion,and carries out the time-dependent reliability research of each key component combined with numerical simulation and surrogate model technology,so as to provide theoretical support for the reliability evaluation of the service performance of EMU.The main contents of this paper are as follows:Firstly,in order to determine the key parts of EMU traction drive system,the failure mode analysis is carried out around the main structural composition of the system,the typical failure mode of the system is selected,and the weak links and key parts of the system are determined by FMECA method.Aiming at the problem that the traditional FMECA method is too subjective and the hazard values of some fault modes are the same,an improved FMECA method based on fuzzy comprehensive evaluation and D-S evidence theory is proposed.The Gaussian membership function is used to weaken the subjectivity of expert scoring results,and the basic probability distribution of expert scoring results is calculated.The D-S evidence theory of matrix synthesis is applied to effectively integrate the evaluation results of various experts,and the hazard level of each failure mode is calculated from the centralized weight value of factors,so as to determine the weak links and key parts of the system.Secondly,taking the axle as the research object,considering the influence of strength degradation and load randomness on the performance of the axle,the adaptive double-layer nested surrogate model method is used to analyze its time-dependent reliability.According to the stochastic process model,the mean and variance of axle time-dependent stress and performance parameter degradation are calculated,and the axle time-dependent reliability model is established.Based on the solution principle of extreme value method,the given working time is discretized,and combined with active learning function,a double-layer nested Kriging surrogate model of inner time and function value,outer sample points and function extreme value is constructed.The update of outer sample points needs to call the inner surrogate model.Through iterative updating,the failure probability of the axle in the observation period is calculated from the outer sample points and the extreme value of the function,and then the time-dependent reliability result is obtained.Finally,taking the gear as the research object,the time-dependent reliability analysis of single / multiple failure modes of the gear is carried out.Considering the strength degradation and load randomness,the time-dependent reliability analysis model of gear single / multiple failure modes is constructed.In view of the characteristics of many input variables in the gear time-dependent reliability model,the single loop decoupling method integrating extreme value response is applied to determine the failure probability of each failure mode in given working time.Considering the characteristics of low gear failure probability and increased output response under multiple failure modes,the Kriging surrogate model of multidimensional output response is constructed.Combined with the subset simulation method,the failure probability of gear multiple failure modes in given working time is obtained.