Research on Control Strategy of Maglev Train Levitation System Based on Multi-Sensor Fusion

Levitation control of high-speed maglev trains requires accurate state perception and effective suppression of composite disturbances. This paper proposes a composite levitation control strategy combining multi-sensor fusion, intelligent parameter optimization and active feedforward compensation. A dynamic Jacobian extended Kalman filter (EKF) fuses eddy current displacement and accelerometer data to estimate the gap and acceleration in real time, with an equilibrium-point linearized Kalman filter as the baseline. The dynamic EKF achieves lower gap and acceleration RMSEs than the baseline, with an acceleration variance reduction rate of 82.46% versus 64.03%. On a single-degree-of-freedom platform, the residual against the pre-processed observation is at the micrometre level, a consistency indicator rather than an absolute accuracy. Particle swarm optimization reduces the step-response overshoot of a cascade PID controller from 10.02% to 3.39%. A Hall sensor array is proposed to measure the normal component of the traction field, and the feedforward compensation reduces the maximum gap deviation under a 1180 N pulse from 0.14 mm to 0.01 mm, within 0.03 mm under 20% parameter mismatch. The cascade and feedforward modules were validated together under composite disturbance, and the estimator was validated on the measurement channels; a fully integrated vehicle-level test remains to be carried out.