Prediction Of Metro Wheel Wear Based On Integrated Data-Model-Driven Approach

In the actual operation of vehicles,the complex interaction between wheel and rail will lead to wheel wear,cause the change of wheel profile parameters,lead to the decline of wheel-rail matching performance,and then endanger the safety of the entire vehicle system.With the explosive development of urban rail transit in China,frequent repairs to wheels will inevitably increase maintenance costs.Therefore,it is necessary to predict the changing trend of the shape of the metro wheel through scientific and effective methods.In order to describe the wheel wear quantitatively,the theory of wheel wear based on creep mechanism has been well applied in railway practice.At present,Jendel material wear model based on creep mechanism is widely used to calculate the wheel wear.When choosing the wear coefficient k_A in Jendel model,there are mainly experimental and empirical methods.The cost of experimental statistical methods is higher and generalization ability is weaker,while empirical selection increases the uncertainty of calculation and prediction.The technology of prognostics and health management(PHM)is one of the core technologies in the field of reliability and system engineering,which can predict the functional status of the system and make reasonable maintenance decisions based on the predicted information.Current prediction methods of PHM mainly include physical model-based method and data-driven method.In recent years,in view of the shortcomings of the above two methods,scholars have put forward the failure prediction technology of complementary fusion of the two methods,which greatly improves the prediction accuracy and efficiency,and reduces the prediction cost.In this paper,the data-model fusion method in PHM is introduced into wheel wear prediction,and a prediction method of metro wheel wear based on integrated data-model-driven approach is proposed,which optimizes the wear coefficient k_A in Jendel wear model,improves the accuracy and generalization ability of the model,and reduces the data and experimental costs.At the same time,the influence of line conditions and load conditions on wheel wear is analyzed.The main contents and innovations of this paper as follows:(1)From the point of view of data-driven and physical model-based,the current situation of wheel wear prediction by scholars at home and abroad is analyzed and summarized,and a method of wheel wear prediction based on integrated data-model-driven approach is proposed.(2)A prediction model of metro vehicle wear based on data-model hybrid is constructed.Firstly,according to the measured wear data for a specific running mileage,the least squares algorithm is used to analyze and calculate the difference between the measured wear value and wheel wear value derived through simulation calculation,with the minimum value of the difference taken as an objective function.By means of Genetic Algorithm,the wear coefficient k_A in Jendel Wear Model is optimized,so that an optimized Jendel Wear Model is obtained.Secondly,through combined application of the vehicle system dynamics model,wheel-rail contact model and optimized Jendel Wear Model,the metro wheel wear in other running mileage is simulated and predicted.(3)The wheel wear of metro vehicles on line A and B in a city is simulated and predicted by using the hybrid driving model.The model and prediction method are verified by comparing with the measured wheel wear data.(4)In order to study the wheel wear of Metro more comprehensively,the influence of different line composition and different load states on wheel wear is analyzed by physical model simulation,which provides theoretical guidance for improving the service life of wheel and increasing the safety of vehicle operation.The results show that the data-model fusion driven Metro wheel wear prediction method is more accurate and more generalization than the traditional wheel wear prediction method.The method effectively reduces the uncertainty of choosing wear coefficient by experience,reduces the cost of experimental data,and improves the accuracy of wear prediction,it represents an effective application of AI method in wheel wear prediction.At the same time,line composition and load states of vehicles also have a significant impact on the wheel wear.The more the curves are and the more the passenger loads are,the worse wheel wear will be.In this paper,scientific and reasonable wear reduction measures and suggestions are put forward through numerical simulation.