Research On Performance Degradation And Life Prediction Of Electro-mechanical Braking Units For Railway Vehicles

Electro-Mechanical brake(EMB)is a new generation braking technology for railway transportation,which use electricity instead of compressed air as source power.EMB system has achieved the innovative development of a fully electrified braking system.Braking system is the last defense line for railway operational safety.The Electro-Mechanical Brake Units(EMBUs)are actuators of the EMB system,which are essential for the normal perform of the braking function.The fully electrified system architecture makes intelligent condition monitoring of the EMBU possible.The failure research of EMBU on rail vehicles is a brand-new field.The existing related researches for other transport vehicles suffer from inconsistent operating conditions,inapplicable sensor layouts,and mismatched monitoring strategies,and thus cannot be directly transplanted.For EMBUs on rail vehicles,the characteristics of non-stationary operation and electro-mechanical coupled non-linear transmission would increase the complexity of degradation performance and fault localization under fault conditions.And the fault detection of components alone cannot meet the safety monitoring demand of braking systems.Accordingly,a full study on the performance degradation and life prediction of EMBUs on rail vehicles from components to the equipment was carried out in this paper,aiming to provide theoretical support and technical guidance to guarantee the reliability and safety of the braking system.The fault characteristics and health indicators are extracted from online multi-parameter intelligent measurement from motor drivers and the built-in force sensor signal from the braking cylinder.And then,the EMBU internal fault detection and trend tracking,as well as Remaining Useful Life(RUL)prediction for braking performance requirements are realized.The main results are as follows.(1)Research on EMBU dynamic simulation modeling and non-stationary operating phase identification method.Firstly,from the EMBU composition and working principle,an EMBU dynamic simulation model was established.And then,the identification method for the friction parameters and the calibration method for the polynomial coefficients of the stiffness curve were proposed.Experiment results showed that the simulation model can accurately reflect the dynamic characteristics of the whole operating process.According to the characteristics of multi-stage variation during the operating process,the braking and releasing process was divided into a combination of quasi-steady-state and transient operating phases.For the problem that the force sensor signal had zero drift and noise,and cannot be directly used for phase identification,an EMBU operating phase identification method based on zero-phase filtering of the force slope,and the threshold selection method based on the stiffness curve were proposed.The accuracy of the identification method was verified by the experiment.The identification method can be used for preprocessing of fault feature and health indicator extraction.(2)Analysis on typical faults of EMBUs.Fault modeling for non-stationary operating characteristics and simulation needs,and analysis on performance degradation.Based on the working principle,the fault mode and effect analysis,fatigue test results and the temperature test of the braking cylinder,the typical faults of EMBUs was obtained,which are intern short circuit of motor stator,rotor eccentricity,the increased wear and increased friction of mechanical transmission components.For the non-stationary operating characteristics and dynamic simulation needs,the fault models of the above faults were established,based on equivalent circuit method,magnetic field analysis,modified winding function analysis,multi-coordinate system dynamics modeling and kinematic modeling of braking unit.The degradation performance of EMBUs in the quasi-steady-state operating phase,as well as the fault evolution and propagation characteristics due to the operating mode and coupled transmission characteristics of EMBUs were analyzed through fault simulation.(3)Research on fault feature extraction,fault detection and trend tracking based on quasi-steady-state operating characteristics of EMBUs.To address the problems of non-stationary factors affecting the feature accuracy and transmission characteristics affecting the fault thresholds,the methods of fault feature extraction and fault threshold selection for intern short circuit fault were proposed,combining Motor Current Signature Analysis(MCSA)and Load Torque Signature Analysis(LTSA).The features were extracted from the clearance elimination phase and the force standstill phase.By the methods,the fault detection adapted to different braking levels was realized.The method of trend tracking for eccentricity fault was proposed based on MCSA.The feature was extracted from the clearance elimination phase.The specific sideband harmonic components of the EMBU torque motor with special pole-slot were corrected by experiment.To address the problem that the internal wear backlash and the external friction pairs clearance cannot be distinguished under normal operating conditions,an internal wear backlash testing method based on transmission characteristics was proposed.The backlash detection was realized by the feature extraction from displacement-force curve of the releasing phase.Based on the phenomenon that friction increase affects performance of multiple characteristics,a modified Euclidean distance and statistical distance construction method were proposed.The distance was modified to overcome the effects of vector element variation and random fluctuation.The trend tracking of increased friction was realized through health/fault-state base point distance comparison.All four faults were verified by experiments.(4)Research on change point detection of health stages and RUL prediction for braking system performance requirements.Taking the train safety operation perspective as the entry point,the pushout force,clearance elimination speed and releasing response time were proposed as the system-level health indicators of EMBU.The fatigue life test of EMBU was designed and carried out.From the test results,the degradation characteristic of 3 health stages was proposed for the first time.And the degradation process was modeled based on the Wiener process.According to the characteristics of health stage evolution,the sliding window change point detection methods based on variance ratio hypothesis test,mean difference hypothesis test,and Maximal Overlap Discrete Wavelet Transform-Multi-Resolution Analysis(MODWT-MRA)were proposed respectively.For the problem that random fluctuations during the degradation process may lead to false detection,a continuous abnormal window detection strategy was proposed.Two RUL prediction methods based on Maximum Likelihood Estimation(MLE)and MODWT-MRA were proposed.According to the adaptability of different prediction methods to health indicators,an EMBU integrated RUL prediction method that fused multiple health indicators was further proposed.The normalized root mean square error evaluation index of this fusion method is smaller than that of the prediction methods based on MLE or MODWT-MRA only.