Maximum Likelihood Identification For Adhesion Performance Parameters Of Heavy Duty Locomotive

Heavy duty locomotive is the best bulk freight channel under middle and long distance.Whether the traction and braking force of locomotive can meet the constraints of rail environment and provide stability depends on the adhesion between wheels and rail.Locomotive running in a complex and varied external environment,which causes locomotive adhesion condition to change at any time,often needs adjustment to meet the requirement.For this reason,this paper aims to accurately estimate the adhesion model parameters between wheel and rail,and designs an online estimation algorithm based on maximum likelihood estimation method,which is based on two common and practical models.It tries to provide a basis for control and optimization of locomotive longitudinal force.It can better ensure the safe operation of locomotive under severe environment such as steep slope and rainfall,and provide better passing ability.The study is shown as follows:In view of the frequent changes in the parameters of the locomotive wheel and rail adhesion model,a real-time online estimation algorithm for the parameters of the rail surface adhesion model is proposed.Based on the Kiencke creep adhesion model,we estimate the descriptive parameters online under the maximum likelihood estimation method.The algorithm uses the maximum likelihood estimation model parameter estimation framework and combines the Kiencke model to get the likelihood function.Then,we use the quadratic programming method to solve the likelihood function in real-time,and finally get the online algorithm of parameter estimation.Considering that the change of wheel rail surface environment brings the abrupt change of model parameters,an time-varying forgetting factor is designed to adapt to the sudden change of parameters.The simulation results show that the algorithm can track the changes of the wheel rail environment in time,and effectively estimate the parameters of the adhesive model,and be able to adapt to the difficulty of the insufficient sample in the adhesion problem.In order to further increase the accuracy of the algorithm,an online estimation algorithm for describing parameters of the adhesion model is designed based on the Burckhardt creep adhesion model.Based on the analysis of the key factors that influence the relationship between the wheels and rail,an identification model is established under the framework of maximum likelihood parameter identification.And then the likelihood function is designed.In order to solve the likelihood function in real-time,an improved differential evolution algorithm be devised to solve the likelihood function.Then the online estimation value of the nonlinear adhesion model is obtained.Finally,several numerical simulations under different working conditions are carried out.The experimental results show that the algorithm has good adaptability to the nonlinear model,the uncertainty of the rail surface environment and the insufficient sample.The algorithm can accurately estimate the description parameters of the creep-adhesion model.It also has good tracking effect on adhesion parameters such as maximum adhesion coefficient and optimal creep rate.