The effective traction of heavy-haul locomotives is contingent on adhesion forces generated at the wheel-rail interface.Traditional control methods struggle with efficient adhesion control due to the nonlinear dynamics,variable coupling,and difficulty in detecting the optimal creep speed in the adhesion process of these locomotives.To address this issue,a model-free integrated predictive control approach for adhesion was proposed.Initially,issues with traditional model-free methods were analyzed,and a novel expanded ultra-local model for locomotive creep speed was established.An extended sliding mode observer was designed to estimate the uncertain components of the ultra-local model,which uniquely estimated the current adhesion coefficient and creep speed in a "one-step" manner,differing from traditional observation methods.An improved post-modulated,oscillation-free extremum seeking method with convergence analysis was introduced to optimize the desired creep speed in real-time,based on effective adhesion coefficient observation.Subsequently,a discrete creep speed predictive controller was designed based on the extremum seeking of the optimal creep speed,mitigating noise amplification issues associated with differential speed in traditional control law designs and proving system stability under closed-loop control.Results are drawn as follows.Simulations can demonstrate the proposed control framework's effectiveness in reducing complexity in adhesion control strategy design and enhancing the smoothness of control torque.The improved extremum seeking method precisely identifies optimal creep speeds under sudden track conditions changes and locks in the optimum across various predefined scenarios.The proposed discrete adhesion control law accurately modulates the drive torque,significantly reducing torque jitter compared to traditional control laws incorporating creep speed differentials.The model-free adhesion predictive control method can effectively optimize adhesion system performance,and achieve efficient control under transient operating conditions.
维普
万方数据