Research On Methods Of Health Condition Assessment And Prediction Of Railway Turnout

Railway turnout is a mechanical and electrical integration equipment that works intermittently.The large number of railway lines,frequent conversion,high failure efficiency and difficulty in fault prevention directly affect driving efficiency and driving safety.The current periodic plan to repair,there are the following deficiencies:the switch is going to fail to get priority maintenance;It is difficult to avoid the problem of surplus repair,even leading to"repair failure";Labor efficiency is difficult to improve;It is difficult to control the labor safety risk in the process of overhaul.The implementation of preventive repair of turnout will be the direction of development.The premise of preventive repair is to evaluate the health condition of the turnout as timely and accurately as possible.At present,the research on theories,methods and technologies to improve the utilization quality of switches is still focused on diagnosis after failure,and there are still few studies on the healthy state division of switches,evaluation criteria,parameter collection and prediction methods for preventive repair,and the depth of theoretical research is still insufficient.In this paper,based on the theory and method foundation needed for the implementation of turnout prevention repair,and based on the theory of fault prediction and health management,the method of health state evaluation and prediction of railway turnout is studied and verified on site.The main innovation points of this paper are as follows:1.For preventive repair,the turnout state is divided into three health levels:failure,sub-health and health;By adding time factor,four corresponding maintenance grade divisions are proposed,and the boundary parameters are summarized.Based on statistical analysis of a large number of data and feature extraction,a health assessment and prediction method based on curve similarity and turnout conversion time is proposed,and the relationship between Index,Pearson correlation coefficient rxy and Euclidean distance dxy is established to form a numerical standard for evaluating the deterioration degree of turnout conversion process.2.By means of notch mark,the problem of affecting the precision of the image boundary due to pollution when directly shooting the turnout to indicate the notch is solved.An adaptive gray threshold image boundary recognition algorithm based on CMOS array technology is proposed to improve the identification accuracy of turnout representation gap.A probabilistic latent semantic(PLSA)based fault feature data recognition method is proposed to mine semantic relevance of turnout data and improve the utilization rate of historical turnout data.3.Starting from the analysis of the morphological characteristics of the parameter curves of ZYJ7 switch conversion process,which is typical representative of the conversion process,a method based on mode value analysis is proposed to identify the intermediate stage and the parameters at both ends of the conversion process,which improves the accuracy of the three stages of the start,conversion and inspection of the switch conversion process.The method proposed in this paper is used to refine the health status assessment of the three stages of the switch transition process,which provides theoretical support and method for the fine repair of the switch.4.In view of the problem that turnout fault data are small samples and the serious imbalance between normal data and fault data,support vector machine(SVM)and support vector description(SVDD)are used to evaluate the turnout health state respectively.According to the characteristics of the turnout periodic adjustment,the influence of external emergencies and the temporal transmission,polynomial regression,multiple regression and LSTM were used to predict the change trend of health status.Based on the forecast results,the list of maintenance priority recommendations provides an effective reference for the prevention of turnout repair.