Study On Rail Wear Prediction And Profile Optimization Of High-Speed Railway

Low energy consumption,light pollution,large transport capacity,high speed,small land occupation and comfort and safety are the comprehensive advantages of high-speed railways,which have developed rapidly all over the world.In recent years,with the rapid development of China’s high-speed railways,the operating mileage and road network area are constantly increasing.The geographical environment encountered in high-speed railway lines is complex and quite different,and there are many small radius curves that bring great difficulties to train passability.Therefore,it is necessary to predict rail wear.Because the traditional rail wear prediction method needs a lot of time,this paper proposes a rail wear prediction method based on proxy model prediction.When the train passes through the unconventional section,the rail abrasion will be more serious due to the irregularity of the track subgrade,such as rail corrugation,rail weld,road-bridge transition section and seasonal frozen area.Therefore,it is necessary to study the rail abrasion in the unconventional section and analyze the impact on the safety,stability and comfort of the train passing through the unconventional section due to the abrasion of the rail profile.In the section with small radius curve,due to the insufficient guiding ability,the wheelset produces a large amount of lateral movement,which leads to frequent contact between the wheel rim and the inner side of the rail,and the rail wear at the section with small radius curve is aggravated due to the large amount of lateral movement of the wheelset.Therefore,according to the load-bearing requirements of curve working conditions,rail profile optimization is carried out by rail grinding.The main work of this paper is as follows:(1)Build a vehicle-rail coupling dynamic model and a rail wear calculation model.When calculating rail wear,we choose Archard material wear model,Kik-Piotrowiski algorithm for wheel-rail normal contact and Fa Strip algorithm for wheel-rail tangential contact,and compile a rail wear prediction program,which lays a foundation for the follow-up research by combining the vehicle-rail coupling dynamic model.(2)In order to save time and improve efficiency,the traditional wear simulation calculation method needs a lot of time.In this section,a rail wear prediction method based on proxy model is proposed,and several key factors affecting rail wear are analyzed,with speed,curve radius,friction coefficient,transition curve length and total weight as input variables and the whole rail wear profile as output target.The traditional numerical prediction method of rail wear is used to provide training data,NURBS curve is used to construct rail profile,kriging is used to build proxy model to predict rail wear profile,and finally the error is analyzed,and the two methods are compared with the measured data.(3)Because the rail wear of unconventional road sections is more serious,which has a greater impact on the dynamic performance of vehicles,we select common rail corrugation,rail weld,road-bridge transition section and seasonal frozen area to quantitatively analyze the changes of rail wear depth under unconventional road sections.We input four kinds of unconventional road sections as track irregularities into the dynamic model to carry out the wear evolution of irregular road sections to study the safety,stability and comfort of vehicles passing through unconventional road sections.(4)In the case of serious rail wear in small radius curve,this section studies the optimization of rail profile in small radius curve,and establishes a multi-objective mathematical model for rail profile optimization of high-speed railway.In order to consider the rail grinding amount at the small radius curve,we select CN60 rail profile as the reference profile,parameterize the rail profile through NURBS curve,because NURBS curve has good local control,take the weight of NURBS curve control points in the wear area as the optimization variable,and randomly sample the optimization variable through the optimal Latin hypercube.The random rail profile obtained by using random weights is input into the vehicle-track dynamic model,and the vehicle dynamic performance is obtained,which is taken as the optimization goal together with reducing the wear.Because of the high accuracy of the neural network proxy model,the neural network is selected to build the relationship between the optimization variables and the objective function,and finally the optimal rail profile is obtained through particle swarm optimization,so as to reduce the rail wear of small radius curves and improve the dynamic performance.