Active Adhesion And Anti-skid Based Traction Control Of High Speed Trains Via Online Identification Of Input Constraint

Recently, constructing the train transport system with safe, comfortable and high-efficient characters has become the important development direction of high speed railway. What is more, high speed trains (HSTs) have the characteristics of very fast running velocity, long distance, complex and changing operation environment, the problem of ensuring the safe and reliable operation of HSTs has become gradually serious, which has also been attracted more and more attention to. Traction system is one of the key sub-systems to control the operation of HSTs, the stability of traction system, as well as the performance of wheel-rail adhesion and anti-skid, has the direct effect on the operation safety of HSTs. Hence, in the traction system, traction control as well as adhesion and anti-skid control is the key part ensuring the safe and reliable operation of HSTs.In this paper, the major objective is to design control scheme simultaneously ensuring the performance of active wheel-rail adhesion and anti-skid as well as the performance of traction, and then the objective of active adhesion and anti-skid based traction control is finally realized. More specifically, the dynamic model of HSTs based on adhesion and anti-skid constraint condition is constructed; wheel skid prediction method of HSTs via support vector machine (SVM) is proposed; at the same time, active adhesion and anti-skid based traction control scheme of HSTs via SVM based wheel skid predictor is derived. The major achievements of the work are summarized as follows:Firstly, based on the wheel-rail adhesion mechanism, the law, which reflects how wheel-rail adhesion and anti-skid control has the effect on the process of traction control, is deeply analyzed; the equivalent form of wheel-rail adhesion and anti-skid constraint condition, which is able to be added to the dynamic model of HSTs, is researched. Adhesion and anti-skid constraint condition that ensures the performance of wheel-rail adhesion and anti-skid is converted to the traction torque constraint condition, and then the dynamic model of HSTs that contains wheel-rail adhesion and anti-skid constraint condition is established. It is the model with dynamic input constraint, which is able to be used as the model to design the adhesion and anti-skid constraint condition based traction control scheme ensuring the simultaneous achievement of the objectives of traction control as well as wheel-rail adhesion and anti-skid control. Secondly, in order to ensure the active adhesion and antiskid performance between the rail and wheel, wheel skid prediction method via SVM is studied. To our best knowledge, it is the first work trying to provide a feasible solution to wheel skid prediction problem in HSTs. Since wheel skid is usually accompanied with apparent changes of some related system states, the wheel skid identification problem can be viewed as a classification problem where wheel skid can be distinguished from wheel creep. Motivated by this observation, wheel skid prediction method is established by using SVM that has been proven effective for linear and nonlinear classification and prediction problem. The optimal separating hyperplane used for wheel skid identification is able to separate the wheel states into wheel creep and wheel skid, which has the same function of the peak points of wheel-rail adhesion characteristic curves. According to moving the peak points to wheel creep regions, the hyperplane is parallelly moved to wheel creep region, and then wheel skid prediction method is generated. Two ways for implementation are also provided. The wheel skid prediction method is able to identify the likelihood of wheel skid and a warning signal of wheel skid is generated before wheel skid appears. At the same time, it is unnecessary to fit adhesion characteristic curves. The efficiency of the wheel skid prediction is verified in a series of MATLAB simulations. At the same time, the widely used wheel skid prediction model via SVM of HSTs is established through the above simulations.Thirdly, on the basis of the dynamic model of HSTs with dynamic input constraint, combining the wheel skid prediction model that is used as wheel skid predictor with the corresponding control schemes, active adhesion and anti-skid constraint condition is able to be expressed by traction torque constraint condition, and then the corresponding dynamic model of HSTs is finally used as the model to solve the problem of active adhesion and anti-skid based traction control. The basic idea of control scheme, which includes traction control as well as active adhesion and antiskid control based wheel skid predictor, is derived. Since the implementation measures of the traction torque constraint condition are different, that is to say, the solutions to the dynamic input constraint via online identification are different; two kinds of schemes for implementation are designed. The simple double loop control scheme verifies the efficiency of the basic idea of the control scheme. Since traction torque compensation method may excessively restrict traction torque or not completely compensate the traction torque that is great than the limit value, robust and adaptive control scheme is proposed, which contains active adhesion and anti-skid control based wheel skid predictor as well as robust and adaptive control with the saturation function. The closed-loop stability issue of robust and adaptive control scheme is analyzed using a Lyapunov-based method. And the stability and efficiency of the control scheme is verified via MATLAB simulations. The simulation results prove that during the operation of HSTs, the objective of traction control is achieved, wheel skid never occurs, the traction system is stability all the time, and then the operation safety of HSTs is realized.Some conclusions and suggestions of the work are provided at the end of the paper.