Research On Catenary Health Status Assessment And Remaining Useful Life Prediction Based On Data Fusion

The catenary is the power source of high-speed railway system,and ensuring the safety and reliability of catenary is the basis for the stable flow and on-time operation of high-speed trains.Under the combined effect of internal and external factors,the performance of the catenary system inevitably shows a trend of degradation,and ultimately makes the whole system fail.Therefore,it is particularly important to complete the accurate assessment and key monitoring of the health status of catenary and the accurate prediction of the remaining useful life of catenary through prognostics and health management technology before the fault occurs,so as to achieve the goal of scientific and intelligent operation and maintenance management of railway infrastructure.In this paper,the fault maintenance time point is taken as the zero point,and the contact force,guide height,pull-out value and other data of the catenary fault point are found by time flashback.Then,the noise points are removed,the abnormal values are excluded,and the whole life cycle data formed by data fusion technology such as unified dimension and feature extraction are carried out,and the research is carried out on this basis.First,select the appropriate catenary parameters to determine the health status evaluation index system and evaluation level;kernel principal component analysis is used to extract the features of the indexes,and the reconstructed features are input into the fuzzy C-means clustering algorithm for classification.The clustering center is used to improve the traditional normal distribution membership function.The combination weight and fuzzy judgment matrix are used for product addition operation to quantify the health status,and the construction of the catenary health status evaluation model is realized.Second,the correlation degree of each algorithm error is calculated by Pearson correlation coefficient,so as to select the algorithm with good prediction effect and large difference as the base learner to build the remaining useful life prediction model of Stacking integrated algorithm.At the same time,aiming at the problem that the model contains a large number of super parameter settings,the Bayesian method is used to optimize the super parameters of each algorithm to improve the overall prediction effect of the model.The results of the case study show that the health index evaluated by the improved fuzzy comprehensive evaluation method and the real remaining useful life show an approximate exponential function distribution,which conforms to the general law of equipment degradation and matches the actual catenary degradation condition,which shows the effectiveness of the method.At the same time,the Stacking ensemble learning model including XGBoost,DNN,SVM and KNN obtained the prediction error evaluation index RMSE of 0.068,R~2 of 0.957 and MAE of 0.053 after Bayesian optimization.Compared with a single traditional machine learning algorithm,its prediction results are more accurate and have significant application value in the remaining useful life prediction of catenary.