Neural Network Sliding Model Control of Radial Translation for Magnetically Suspended Rotor (MSR) in Control Moment Gyro

Abstract

For a magnetically suspended control moment gyro (MSCMG), the high-speed rotor is actively suspended by magnetic bearings of 5-DOF, but the nonlinearity of the magnetic suspension force is one of the main reasons for the poor accuracy of radial translation control of the magnetically suspended rotor (MSR). To solve this problem, here, the characteristics of the magnetic suspension force are analyzed, and the nonlinear dynamic model of MSR is established. A sliding mode control (SMC) based on a neural network is presented, and the radial basis function (RBF) neural network is adopted to approximate the nonlinear displacement stiffness and the current displacement stiffness to weaken the chattering in SMC to improve the control accuracy of the MSR. The stability of the neural network SMC for the MSR is analyzed based on Lyapunov functions, and the rules of updating network weights are presented based on adaptive algorithms. Compared with these existing classic control methods, the simulation and experimental tests performed on a single-gimbal MSCMG with an angular momentum of 200 N.m.s indicated that this neural network SMC for MSR’s radial translation can not only make its suspension more stable but can also make its position precision higher.