Research On The Fault Diagnosis Of High-speed Rail Vehicle Suspension System

Suspension system is a key component of high speed train.Regular fault diagnosis to the suspension system is important that it not only can strengthen its safety,but also can decrease maintenance cost.This thesis,based on the Guang Xi Education Department funded project: research project of fault diagnosis technology of high speed train suspension system(No:KY2016KB247),researches the fault detection and isolation algorithm for high speed train suspension system based on data-driven.The thesis mainly including following contents:1.At the first place,introduction to the structure of vertical suspension system of high speed train is conducted.Meanwhile,the suspension system's mode is established and simulation platform of high speed train suspension system is constructed based on its structure and SIMPACK dynamics software.And the condition of irregular pathway(crosswise,vertical,and sidewinder)is considered,so as to provide relative experimental data for algorithm simulation of MATLAB.2.Then,KPCA algorithm and fuzzy neural network algorithm are employed to conduct fault detection toward suspension system.Based on the KPCA algorithm and FDA,the fault detection is conducted successfully toward suspension system.The results of comparing the false alarm rates and loss rates of KPCA algorithm and improved KPCA algorithm show that: the improved KPCA algorithm is significantly better than traditional KPCA algorithm,and the selection of kernel function parameter? is important to the detection precision.For the fault diagnosis of fuzzy neural network on suspension system,the fuzzy neural network is designed based on the acquired data of SIMPACK.The results show that such method is effective and possesses higher precision.3.The thesis researches the fault isolation of high speed train's suspension system based on PCA and SVM.The results show that the fault separation precision after employing PCA to process data is better than that of traditional data.In addition,the selection of kernel function greatly affects the separation precision-the performance of Gaussian kernel function is higher than that of linear kernel function and that of polynomial kernel function.4.The PSO fault isolation algorithm of support vector machine optimizing and supporting is researched.PSO is employed to optimize the two key parameterspenalty factor C and kernel function parameter? of support vector machine so as to improve the precision of fault separation results.The results show that comparing with SVM,such algorithm can greatly improve precision.