Structure Reliability Analysis Of Bogie Frame Based On BP Neural Network

Bogie frame is the key component of high-speed trains,not only to install the high-speed train braking unit,power unit and related sensors and other components,as well as the main bearing parts in the process of the train operation,due to the high speed railway passenger car running conditions,to improve the speed,the bogie frame in the actual operation process,the problems exposed a series of structural strength direct threat to the safe and reliable operation of the train,so consider vehicle uncertainty factors in the production and operation analysis,the reliability is the key to ensure the safety of the train moving.In this paper,a type of bogie frame is taken as the research object.The static strength reliability and fatigue life reliability were analyzed respectively based on the genetic chaos particle swarm optimization(GACPS-BR-BP)neural network surrogate model and the active learning Br-BP neural network surrogate model.The specific contents are as follows:Firstly,the finite element model of the bogie frame was established by Hyper Mesh software.Six abnormal conditions of the bogie frame were determined based on UIC614 and EN13749 standards.The maximum Von.Mises stress of the bogie frame was calculated by ANSYS finite element analysis software and compared with the allowable stress of the material.Then,13 simulated operating conditions of the frame were determined and fatigue strength analysis was carried out on the bogie frame.All welds on the frame were selected to evaluate whether the welds met the fatigue strength requirements based on Goodman fatigue evaluation limit chart.Then,the cross-validation technique was used to determine the optimal training parameters of BP neural network,the initial weights and thresholds of BP neural network were optimized by GACPSO algorithm,and the optimized network was trained by Bayesian Regularization algorithm(BR).A reliability analysis method based on GACPSO-BR-BP neural network surrogate model was proposed.Agent based on the improved model,according to the structure of static strength analysis of the results,pick the largest stress condition as reliability analysis conditions,and then to extract more significant influence on frame maximum stress sensitivity analysis of load,the size of the plate,material properties as by analysis of design variables,establishment of maximum stress GACPSO-BR-BP neural network model,and using Monte Carlo method for static strength reliability of the architecture.Finally,considering the contribution of sample point quality to failure probability,an active learning BP neural network surrogate model was proposed,and the active learning surrogate model was combined with the nominal stress method,Miner’s cumulative damage theory and Monte Carlo method to analyze the fatigue reliability of the welds of bogie frame.Consider structure of external load,the size of the plate,material properties and the influence of random factors on the bogie frame will be defined as the input variables,determine its distribution,based on fatigue failure criterion,based on cumulative damage ratio is less than the critical damage than fatigue reliability evaluation standard,establish a bogie frame fatigue reliability analysis of the limit state equation,using Monte Carlo method for fatigue reliability of the architecture.