A self-tuned fuzzy-neural-network-based BLDCM speed control strategy

To improve the control performance and control accuracy of brushless DC motor (BLDCM) and reduce the impacts of unmodeled dynamics, parameter variations and disturbances, a fuzzy neural network speed regulation system with online learning parameters without relying on the model is proposed in this paper. Different from the traditional fuzzy neural controllers, the structure of the proposed speed controller consists of two parts, a fuzzy neural network controller and a training controller based on sliding mode theory working in parallel. An online adaptive training algorithm is put forward to adjust the weights and membership functions. The tuning method is designed to make the learning error satisfy a stable energy equation rather than minimize an error-based function, thus avoiding the calculation of the Jacobian matrix of the control object and falling into the local minima, which simplifies the calculation, and guarantees the real-time of the control system. Experimental results based on digital signals processor (DSP) verify the effectiveness of the proposed scheme.