Residual Life Prediction Based On Dynamic And Static Parameters Of High Speed-Railway Catenary

Catenary is an important part of the traction power supply system,so its service performance impacted directly the operation of high-speed railway.Catenary is outdoor and no reserved equipment which is disturbed by external factors badly.Once faulty happen the normal transportation of high-speed railway and the people's life property safety will be damaged.Therefore,it's important to maintain the catenary before malfunction according to degradation performance identification and residual life prediction based on dynamic and static parameters of catenary.In this dissertation,the catenary which is taking as research object is located in Wuhan-Guangzhou high-speed railway,Qingyuan to Guangzhou North section.Degradation performance identification and residual life prediction are studied by Continuous Hidden Semi-Markov Model(CHSMM).Fristly,the state division of catenary.The eigenvalues are extracted which can characterize degradation process from dynamic and static parameters of catenary system.Based on this,the number of degradation stages to be experienced during the degradation of catenary system is settled by K-means algorithm and cross validation algorithm for modeling,analysis and forecasting.Secondly,the system-level modeling of catenary based on CHSMM.First of all,the parameters of Continuous Hidden Semi-Markov Model are modified according to the operation and eigenvalues of catenary.Then study problems include model training,underflow in algorithm,multi-observation input and parameter initialization.At last,the theoretical framework of degradation performance identification and residual life prediction are established.Thridly,the system-level residual life prediction of catenary.The residual life prediction algorithm is improved to enable it fit the actual situation.In actuality,there will be several sizes of maintenance and multiple degradation states between different sample data based on the characteristic that during the service of the catenary.In the end,verify the measured data from scene.Study degradation performance identification and residual life prediction of the model by import the eigenvalues of the measured data from scene into model.Verify the accuracy of modal and the feasibility of this method to characterize the degradation of catenary by comparing the output of model and the report of on-site maintenance.