Research On Fault Diagnosis Of Railway Freight Car Brake System Based On Hidden Markov Model

With the continuous improvement of railway freight capacity,the issue of railway freight car operation safety has been paid more and more attention.As an important part of railway freight car,the brake system plays an important role in maintaining the running safety.Because the brake system is a complex system,the harsh working environment causes brake system faults occurrence from time to time.Therefore,how to diagnose the brake system fault accurately and efficiently has become a research hotspot of railway freight car.This dissertation focused on the fault diagnosis research of the railway freight car brake system.By analyzing the common fault modes of the brake system during the actual line operation of the railway freight car,a fault diagnosis method of the freight car brake system based on the hidden Markov model was proposed.The method was based on the wind pressure signal of the railway freight car brake system,and established the hidden Markov model of the brake system different states by extracting the effective characteristics of the wind pressure data to realize the state recognition and fault diagnosis of the brake system.Following research work were completed in this dissertation:(1)The structure and working principle of the railway freight car brake system with the 120-1 air brake as the core were analyzed.Based on the freight cars line operation fault data,the three common faults and their failure mechanisms were analyzed in terms of improper release,natural release and brake cylinder leakage.And summarized the characteristics of brake failure data.(2)A multi-dimensional feature extraction and optimization method of railway freight car brake wind pressure signal was proposed.First,the multi-dimensional feature parameters of the brake wind pressure signal was extracted from the time domain,frequency domain,time-frequency domain and correlation;then performed feature dimensionality reduction based on the combination of feature selection and feature fusion.That is,selected sensitive features and weighted based on the compensation distance evaluation technology,and performed feature fusion on the non-linear part of the feature vector by KPCA processing.(3)By analyzing the characteristics of brake system fault data,the continuous HMM was innovatively applied to freight car brake fault diagnosis,and a fault diagnosis method of freight car brake system based on discrete HMM was proposed.Based on the multi-dimensional feature extraction and optimal selection of brake system faults,the feature vector was scalar quantized,and then different states discrete HMMs were trained using the Baum-welch algorithm to form a model library for fault diagnosis.(4)In order to avoid the lack of effective information caused by data discretization,continuous HMM was applied to the freight car brake fault diagnosis,and a fault diagnosis method for brake system based on continuous HMM was proposed.Aiming at the problem of continuous HMM initial parameter optimization,the K-Means clustering algorithm was used to complete the optimization of the initial model,and then different states continuous HMMs were trained to form a model library for fault diagnosis.This dissertation conducted experiments on the Daqin railway line operation data of the freight car brake system.Through the comparison of different methods,the experimental results showed that the discrete HMM and continuous HMM methods proposed could effectively identify the brake system faults which had certain advantages in aspect of the model training,diagnosis accuracy,classification dispersion.Through comparison with common algorithms such as SVM and RF,it showed the superior performance of HMM in the freight car brake fault diagnosis.Among them,discrete HMM had the advantages of simple and efficient,fast training speed,while continuous HMM had a long training time,but other experimental evaluation indexes were optimal,which can be better applied to fault diagnosis of railway freight car brake system under the premise of satisfying the diagnosis efficiency.