Study On Intelligent Prediction And Diagnosis Of The Contact System Failure

"High power,heavy load" traction power supply system in the rapid development and transportation to provide convenient at the same time,there are also urgent problems,how to improve its safety,reliability makes it efficient for the supply of energy for railway transportation.As an important source of power supply,the contact system sits affected by many factors,such as climate,geography,etc.,due to the special nature of the service environment.And because of its cumbersome construction,by a variety of functional components to work together,so that the analysis of its work performance can not simply study a component,but also to analyze the interaction between various incentives.Therefore,Advanced data-driven technology is used to study the fault of the contact system and predict its state by detecting data in the paper.First of all,through a number of on-site research and all kinds of related literature,from the root cause inducing the contact system failure,various triggers between the links are analyzed to solve essentially the problem of the contact system fault diagnosis.The correlation between various faults and triggers is verified by SPSS statistics software,and data support is provided for the characteristic selection of the subsequent diagnostic model.Secondly,according to the data obtained by the research and the analysis results of different faults,the eight characteristics of fault voltage,current,impedance angle,impedance,temperature,humidity,wind speed and weather are extracted,and the foreign body,locomotive,lightning strike,overload,ice cover fault is diagnosed,and the diagnostic effect of BP and LVQ is analyzed.Then,in order to contain the failure in the embryonic stage,reduce the incidence of failure,the detection data of the investigation is analyzed and collated,including contact pressure,wire height and other detection values and further seek its mean and variance as the input of the neural network,to contact the average wear ingress of the network as the output of the model prediction,through MATLAB programming to establish the contact network wear prediction model.Finally,this paper uses MATLAB GUI to make the contact system state prediction and troubleshooting interface design,eliminating the cumbersome code interpretation,staff can read the diagnostic results simply by performing simple and clear interface operation.The research in this paper shows that it is feasible to use neural network algorithm based on fault data-driven methods such as voltage,current and power angle,especially the LVQ neural network trained by cross-verification,which has a high accuracy,and the prediction ability of BP neural network in contact line wear and tear is also significantly improved by genetic algorithm optimization,which proves that the above method has some reference value for contact network state prediction and fault diagnosis theory research,and can also be used in the field.The contact network state prediction and troubleshooting interface based on MATLAB GUI reduces the time for staff to analyze code and makes troubleshooting more intuitive and convenient.