Research On Fault Diagnosis Of High-speed Train Bogies Based On Feedback DS Theory

The research on the faults of high-speed train bogies is mainly based on monitoring data.Among them,the selection of features and the optimization of parameters directly affect the accuracy of fault identification.In recent years,this has been a research hotspot in this field.Because a large number of sensors are arranged in different azimuths of high-speed trains,there are correlations among the features extracted by sensors of different channels,and the integration of features of all the channels for analysis becomes an effective method for fault identification.However,the fusion of a large number of channel data has caused characteristics such as high dimension of state feature and a large number of data,which has increased the difficulty of feature selection and parameter optimization.The characteristics of some insensitive channels also lead to data redundancy and reduce the accuracy of fault diagnosis.Based on this problem,this paper discusses the application of feedback-type DS theory in fault diagnosis from the perspective of feature selection and parameter optimization,and carries out fault diagnosis for different fault types and different degrees of faults.The specific research work is as follows:1.In the paper,failure features are sorted using single evaluation criteria such as Mahalanobis distace,Fisher's ratio,and Fuzzy entropy.The DS theory was used to fuse the results of multiple feature rankings,and then to construct an efficient sequencing of the fault features of high-speed train bogies.The obtained effective feature set is added one-dimensionally to input of the parameter optimization process of the fault diagnosis model.2.In the parameter optimization process of fault diagnosis model,three common parameter optimization algorithms,such as genetic algorithm,particle swarm optimization algorithm and grid search algorithm,are used to optimize the penalty factor and kernel function parameters of SVM.And for the situation that the grid search method is easy to fall into a local optimum,a method to improve the grid search method to optimize the parameters is given.The experimental results shows that the improved grid search method is better than the three commonly used optimization methods,and improves the diagnostic accuracy.3.Based on the feedback control theory and combined with the optimal fault diagnosis model obtained by parameter optimization,a feedback-adjusted DS evaluation ranking result was proposed to obtain the final optimal fault diagnosis model.This method uses the improved grid search method to optimize the acquisition of the optimal fault diagnosis model,and feedback-adjusts the result of the original DS theoretical feature ranking.The new feature evaluation ranking result is again combined with the parameter optimization,and the final optimal fault diagnosis model is obtained through multiple feedback adjustments.Finally,aiming at the lack of qualitative analysis of fault damage degree in the research of high-speed train faults,the feedback DS theory and improved GS parameter optimization methods are applied to the high-speed train fault diagnosis,and carried out analysis of different types of faults and different levels of faults.