Diagonal Recurrent Neural Network Control of Active Magnetic Bearing Based on Multi-Physics Field Coupling

Multi-physical field factors impact the dynamic modeling of active magnetic bearings (AMBs), a crucial concern in the control of AMBs. In this paper, the focus is on modeling the thermal, electromagnetic, fluid, and structural multi-physics field dynamics of the AMB-rotor. A virtual prototype model (VPM) based on this coupling is proposed, and neural network control of Diagonal Recurrent Neural Network (DRNN) is utilized to address the nonlinear and multi-physics field issues of the AMB-rotor system. This approach effectively achieves system torque decoupling. Finally, the effectiveness of the control algorithm, along with the accuracy and reliability of the multi-physical field coupled VPM, is verified through comparative simulations and experiments of the PID controller and the DRNN controller.