Wear Prediction Of Collector Strip And Contact Wire In Pantograph-catenary System On Optimized Multi-kernel Least Squares Support Vector Regression

As the traction equipment of locomotive,the state of pantograph collector strip affects the safety of locomotive operation.The existing research on pantograph collector strip is mainly in two fields.On the one hand,it is about the detection system of pantograph collector strip to confirm whether it can support the normal operation of locomotive.On the other hand,it is about the law of collector strip wear.However,due to the complexity environment of pantograph operation,the existing research on the wear law of pantograph collector strip is based on simulation and experiment.With the development of big data,the wear research of high-speed railway wheels has begun to use the actual detection value accumulated in locomotive operation.The wear law of wheels is studied and analyzed by using the actual measured values accumulated in locomotive operation,and good results are obtained.In this paper,after referring to the wheel wear prediction research,the support vector regression prediction method was used to the pantograph collector strip wear research,and pantograph collector strip wear prediction based on the detection data.Because the accuracy of single-kernel function support vector regression(SVR)model is not high,and SVR it is easy to fall into over-fitting or under-fitting.The least squares support vector regression(LSSVR)and multi-kernel support vector regression(MK-SVR)are used to construct model.In the process of model constructing,the types of parameters gradually increase,so the quantum genetic algorithm is used to carry out the parameters.Finally,the optimized multi-kernel least squares support vector regression(QGA-MK-LSSVR)model is obtained.In this thesis,four models are used to train and test the pantograph collector strip data which from Chengdu Metro and Shanghai Hongqiao High-speed Railway.The fitting accuracy of the QGA-MK-LSSVR can even reach 99.6%.After accumulating a amount of collector strip wear data and using the model to fit the data to build a prediction system.A data of mileage is input into the system,and the system can predict and output the corresponding predicted value.That improving the efficiency of the collector strip.The reliability,safety and economy of pantograph collector strip wear research have been improved.This paper has certain significance for pantograph slide wear research.