A Neural-Network-Identifier and Fuzzy-Controller-Based Algorithm for Dynamic Decoupling Control of Permanent-Magnet Spherical Motor

This paper proposes a dynamic model of permanent-magnet spherical motor (PMSM) and puts forward a dynamic decoupling control algorithm of the motor, using fuzzy controllers (FCs) and a neural network identifier (NNI). PMSM is a multivariable nonlinear system with strong interaxis couplings. A computed torque method structure is applied to PMSM. There are such uncertainties as estimated errors of the model and external perturbations, which may influence the precision of the control system. A back-propagation algorithm with additional momentum term and self-adaptive learning rate applied to feed-forward neural network can approach nonlinear functions with a learning rate adjusted online, which helps to improve training speed. In this paper, an NNI is applied to identify the uncertainties online. An adaptive-neuro-fuzzy-inference-system-based FC is applied, which has self-adaptive ability and strong robustness. Simulation results preliminarily validate that the algorithm proposed in this paper can eliminate the influences of interaxis nonlinear couplings effectively to actualize dynamic decoupling control. Furthermore, the static and dynamic performances of the control system have been improved greatly with strong robustness to uncertainties. A hypothetical microprocessor system is proposed, and simple experiments of spinning operation are carried out as a foundation for further study.