The Research On Life Prediction Method Of Safety Relay

As a part of railway system,the safety relay has a very important role.Its life value directly affects the safe and reliable operation of the whole railway system.At present,the rapid development of the railway has higher requirements on the quality and quantity demand of the relay.Only relying on the regular maintenance and replacement of the relay cannot meet many requirements,in this way,the fault relay would not be found in time,the service life of relay that required periodic replacement would not be prolonged.Thus,considering from the point of economic and security,the research on life prediction method of safety relay is badly need of development,and it can provide support to improvement of relay's repair class and repair system.At present,there is no research on the life prediction methods of safety relay at home and abroad,but it can learn from the research methods of electromagnetic relays in aerospace and other fields.The service life of the relay is divided into mechanical life and electrical life.In general,the mechanical life value is much larger than the electrical life value.So,the life prediction method of this thesis is mainly aimed at the prediction of electrical life.In addition to having predictive models,predicting electrical life values also needs to determine the predictive variable and its failure threshold,but not all of them can use the same standard.Therefore,the failure process of the relay should be analyzed concretely.According to the failure mechanism,making classified discussion.And the prediction model,the predictive variable and the failure threshold of different failure types are obtained.The failure state of the safety relay is mainly caused by the contact failure.In this thesis,through analyzing characteristic change of contact during its failure process,and determining the three failure forms which caused by the transfer of the contact surface material.At the same time,aiming at the closing process and releasing process of contact,introducing the six characteristic parameters of the contact,and these parameters including five time parameters and contact resistance.Several characteristic parameters are quantitatively analyzed based on principal components analysis method,and 18 sets of contacts are classified into three kinds of failure forms with the aid of the Mahalanobis distance discriminant method and the principal components characteristics of the characteristic parameters.The characteristic parameter features of the contacts with different failure forms are studied by using Fisher discriminant criterion.Electing characteristic parameters which conform with change features of different failure forms as the electrical life predictive variables.And according to the experimental data,the respective failure thresholds of predictive variables are determined.Elimination method,moving average method and wavelet denoising method are used to deal with each predictive variable.On the basis of it,a variety of curve fitting models and time series prediction models based on BP neural network are established.Then,the predictive simulations of all models are carried out by MATLAB.Through simulation,respective fitting errors and the predictive results are obtained,and the predictive accuracy is calculated by the deviation of the predicted value from the known actual electrical life value.Finally,the characteristics of simulation curve and the predictive accuracy are analyzed synthetically,and the predictive models that are suitable for predictive variables with various failure forms are determined.In this thesis,the characteristic parameter data used in the simulation are provided by the electrical life test system platform of Shenyang Railway Signal Plant.