Neuro-adaptive anti-slip brake control of high-speed trains

This paper investigates the problem of anti-slip brake control of high-speed train. An adaptive neural network control strategy without using precise system parameters is proposed to counteract modeling uncertainties and unexpected disturbances. Moreover, various anomaly factors such as varying and uncertain operation environment, the unknown nature of the wheel/rail contact surface and unexpected geological hazards are taken into consideration in control design. Adaptive variable structure observer is constructed for estimating the adhesion force. Reference slip ratio generation algorithm using fuzzy logic is developed to determine the desired slip ratio for a large adhesion force. The resultant control algorithms are not only independent of the dynamic model, but also robust against the modeling uncertainties and external disturbances. The performance and robustness of control scheme is evaluated through computer simulation. © 2013 TCCT, CAA.