Decoupling control of single winding bearingless switched reluctance motors based on support vector machine inverse system

The none-linearity and magnetic saturation are limiting factors for separating control torque and levitation force in single-winding bearingless switched reluctance motor (SWBSRM). Further analysis by finite element method and mathematical model show that currents generating torque and radical force are difficult to be independently controlled. This paper introduces a novel method based on least-squares support vector machine (LSSVM) and inverse system to establish the predictive model for driving current and radical force current. Firstly, the working principle and the nonlinear model of the motor are explained. Then, the predictive model is obtained by training LSSVMs with representative currents and radical displacement data. Meanwhile, the particle swarm optimization (PSO) with inertia weights is used to optimize parameters of LSSVMs. Finally, the proposed methodology is verified by the simulations based on internal model control (IMC) scheme.