Wheel skid prediction and antiskid control of high speed trains

Abstract

Timely and accurate prediction of wheel skid is the precondition for effective antiskid control of high speed trains (HSTs) to ensure its safe operation. Since wheel skid is usually accompanied with significant change of the wheel speed or/and other related system states, the wheel skid identification can be considered as a classification problem where different states can be distinguished before and after wheel skid occurs. Motivated by this observation, the authors establish in this work a wheel skid prediction method by using the support vector machine (SVM) technique that has been proven effective for linear and nonlinear classification and prediction. It is shown that by modifying the bias parameter of the SVM classifier, the proposed method can predict the trend of wheel skid rapidly before wheel skid is upcoming, thus an active antiskid control can be activated in advance to avoid wheel skid. The effectiveness of the proposed strategy for skid prediction and antiskid control is theoretically analyzed and validated via numerical simulations.