Research On Freight Train Braking Model And Performance Based On The Intelligent Platform Of Braking System

Freight trains have become one of the mainstream freight transportation methods in the world because of their characteristics of high load and low cost.With the development of freight trains in the direction of high speed,long formation and high load,the longitudinal impact between trains will also become larger,which leads to the increased possibility of train accidents.And the performance of the train braking system directly affects the longitudinal impact force.Therefore,it is of good practical significance to study the performance of train braking system.In this paper,the Segmented Electro-Pneumatic(SEP)braking system is used as the research object,and the co-simulation model of braking model,control module and dynamics model is established by using computer simulation technology for studying the performance of the train braking system.Meanwhile,in order to ensure the reliability of the SEP brake system performance,a brake fault diagnosis method is proposed for the common fault types of the brake system.The main work of this paper is as follows:1.Study the brake system performance by using numerical simulation technology.Firstly,according to the principle and structure of brake valve,AMESim is used to establish a single car brake system simulation model,and the common braking conditions and relieving conditions are simulated and studied,meanwhile,based on this,a SEP brake system simulation model of long formation trains is established.Secondly,the train longitudinal dynamics model is analyzed,and the train braking dynamics model is established by SIMPACK.Thirdly,control modules were designed in Simulink platform,including a functional variable universe fuzzy controller optimized by genetic algorithm(GA)and a computational module for calculating braking force and basic resistance,respectively,for controlling the electronic control device in the SEP braking system and feeding back to the dynamics model.Finally,the co-simulation model is formed by combining the braking system model,the braking dynamics model and the control module through SIMAT and AME2 SLCosim,and the co-simulation model is used to study the braking performance at different common braking conditions under a straight line.The results show that the braking performance of the SEP braking system with the functional variable universe fuzzy controller optimized by GA is better,and its braking wave speed is increased by about 4 times compared with the pure air braking system,and the braking distance is also reduced accordingly,by4.17% at the minimum depressurization and 23.03% at the maximum depressurization.2.To guarantee the braking performance,a fault diagnosis method based on hybrid stacked autoencoder is proposed.The method includes two parts: nonlocal and local structure preserving stacked autoencoder(NLSP-SAE)and stack autoencoder(SAE).NLSP-SAE mainly enhances the small sample braking fault data,while SAE classifies faults based on data features.The results show that the proposed method has a higher fault type identification accuracy than other algorithms such as supporting vector machine(SVM),random forest(RF)and data enhanced stacked autoencoder(DESAE),up to 93.28%.