Research On Fault Isolation And Fault Identification Of High Speed Vehicle Suspension System

The suspension system is vital important te the safety operation of the high speed vehicle,because it undertakes the load of car body and bogie,acts as a buffer to vehicle vibration and wheel-rail impact,and guides the vehicle running.With the development of high speed railway in our country,higher requirements to the safety and reliability of the suspension system are raised because of high intensity running of the high speed vehicle.Based on the research of fault isolation on high speed vehicle suspension system,precise localization of different faults is obtained,which provides guidance to inspection and maintenance.Based on the research of fault identification on high speed vehicle suspension system,monitoring on parameters of key components is obtained,which contributes to the detection of parameter attenuation and sudden failure.In order to implement fault isolation and fault identification of high speed vehicle suspension system,research on feature extraction,feature dimension reduction,fault isolation and parameter estimation is carried out.The main research in this paper is as below.(1)Fault feature extraction and dimension reduction algorithms based on zoom spectrum analysis are studied in this paper.In the process of fault feature extraction research,a power spectrum feature extraction method based on zoom spectrum analysis is proposed,which contributes to the supplement and perfection of time domain and frequency domain features,and improves the quality of fault feature sample.In the process of fault feature dimension reduction research,KPCA is applied for dimension reduction of high-dimensional nonlinear fault feature sample,which overcomes the disadvantage that PCA is practicable only to linear dimension reduction.A parameter optimization algorithm for Kernel function is proposed,which improves the dimension reduction efficiency of KPCA.Based on the fault simulation platform built in this paper,the efficiency and reliability of the algorithms talked above are verified.(2)Fault isolation algorithms based on fuzzy intelligence are studied in this paper.In the process of fault isolation research based on FPCM,reasonable dimension of the input samples is discussed,optimization algorithm for deciding optimal clustering number and initialized clustering centers is proposed,which improves stability and accuracy of fault isolation using FPCM.In the process of fault isolation research based on BP neural network,optimal design on network structure and parameters is carried out.Considering that FPCM is less dependent on the priori knowledge of sample distribution,a hybrid algorithm is proposed,which optimizes the selection of training sample for BP neural network,and improves its training efficiency.Therefore,the stability and accuracy of fault isolation using hybrid algorithm is higher than BP neural network.Based on the result of fault feature extraction and dimension reduction,the efficiency and reliability of the algorithms talked above are verified.(3)Fault identification algorithms based on nonlinear filtering are studied in this paper.Aimed at fault identification of the suspension system,parameter estimation algorithms,like PF and RBPF,are studied for parameter attenuation monitoring of key components.An evaluation index called error ratio is introduced for comparing the identification speed and accuracy of the two algorithms,and proves that RBPF is more practicable when it comes to parameter estimation of high-order system.An RBPF algorithm based on repeat uniform sampling strategy is studied,which implements parameter estimation under sudden failure condition,and further implements fault identification of key components.Based on the high speed vehicle suspension system lateral dynamic model built in this paper,the efficiency and reliability of the algorithms talked above are verified.