Neural Network Inverse System Decoupling Control for Six-Pole Radial-Axial Hybrid Magnetic Bearing Based on Deep Fuzzy Inference

To address the strong coupling and nonlinearity issues in the hybrid magnetic bearing (HMB) of magnetic levitation electric spindles, a deep fuzzy neural network inverse system (DFNNIS) decoupling strategy is proposed for the six-pole radial-axial hybrid magnetic bearing. Firstly, the structure and principle of HMB are introduced, and its mathematical model of suspension force is derived. Secondly, the displacement and current data are collected using the HMB mathematical model and normalized. The inverse system model is obtained by training with the deep fuzzy neural network (DFNN). According to the right inverse principle of the inverse system, the inverse system model is connected in series with the original system, thereby achieving decoupling of the HMB suspension force. Finally, through the simulation and experiments of rotor suspension and anti-interference, it is proved that the convergence speed and anti-interference performance of the DFNNIS decoupling model are superior to those of the traditional neural network inverse system model.