Wheel-rail adhesion control model by integrating neural network and direct torque control during traction under low adhesion

Abstract

In the realm of train traction, achieving optimal utilization of wheel-rail adhesion is of utmost importance. The motor’s efficiency plays a significant role in this process. However, there has been limited research on adhesion optimization for motor control in recent years. Therefore, this paper proposes a neural network controller based on the Levenberg–Marquardt (LM) algorithm to improve adaptive regulation ability. This approach integrates the direct torque control (DTC) method, which utilizes a three-phase asynchronous motor to output torque and speed. By integrating these techniques, we mitigate the significant slip occurrence during complex low-adhesion scenarios. MATLAB/Simulink simulations are conducted using three different rails: dry, greasy, and wet, each with distinct characteristics. The obtained results demonstrate that the proposed strategy optimizes adhesion utilization while mitigating excessive slip, and exhibits excellent robustness and self-regulation capabilities throughout the adhesion optimization process.