Adhesion control of electric locomotives based on machine learning rail surface state recognition

The key to traction control in locomotive systems is adjusting the axle traction torque through adhesion control. In traditional adhesion control systems, due to the inability to detect the peak adhesion coefficient of the rail surface in realtime, locomotives are prone to spinning and sliding phenomena under different rail surface conditions. This reduces the utilization of adhesion and can even lead to irreversible damage such as tread detachment and rail wear. Therefore, this paper proposes a machine learning-based rail surface recognition method to identify the rail surface condition online and set the peak adhesion coefficient for different rail surfaces. The output torque of the traction motor is dynamically adjusted based on the set reference peak adhesion point to fully utilize the current adhesion force between the rail surface and wheels. A locomotive multi-axis adhesion control model is established to validate the effectiveness of the rail surface recognition algorithm. The results show that the improved adhesion control method effectively prevents wheel spinning, improves the stability of locomotive operation, and reduces the loss of adhesion utilization.