Research on a brushless DC motor control method based on an improved three-layer BP neural network

To address the problems of nonlinearity, time-varying parameters, and complex dynamic coupling in brushless DC motor (BLDCM) speed regulation systems, conventional proportional–integral–derivative (PID) control and single-neuron PID control exhibit deficiencies such as weak adaptability, poor disturbance rejection capability, and insufficient steady-state accuracy. This paper proposes a PID self-tuning control method based on an improved three-layer backpropagation (BP) neural network. By exploiting the strong nonlinear approximation capability and self-learning characteristics of the neural network, the PID parameters can be adjusted online in real time. A mathematical model of the BLDCM is established, and a speed-loop controller based on the improved three-layer BP neural network is designed to achieve adaptive tuning of PID parameters. A dual closed-loop speed control simulation platform is developed in MATLAB/Simulink, and comparative experiments are carried out against conventional PID control and single-neuron PID control. The results show that, compared with conventional PID control, the proposed method improves the dynamic response by 6.7% and the steady-state accuracy by 14.6%. Under abrupt load disturbances, it yields smaller speed fluctuations and faster recovery. The proposed method effectively enhances the dynamic performance, steady-state accuracy, and robustness of the BLDCM speed control system and is, therefore, suitable for high-precision servo drive applications.