Improved Neural Network Based On The Research Of Brushless DC Motor Control

Brushless DC motor (BLDC) has a simple structure, high reliability and good speed characteristics. It also overcomes the shortcomings of mechanical brush motor commutation. It is gradually extended in various fields of modern industrial applications now. And it occupies an important position in industrial control. But brushless DC motor has the same problem of unstable speed and torque ripple as traditional motor. And it has complex, non-linear characteristics of strong coupling. Therefore, how to improve the efficiency of speed control is an important goal of research field.Firstly, the text introduces details of the neural network control theory, BP neural network algorithm and network structure, and pointes out the lack of BP neural network. By analyzing the impact of the BP algorithm’s learning rate and initial weights of. In aspect of learning rate, the text proposes a hierarchical improved method for learning rate adjustment. And in aspect of the initial weights, it uses particle swarm optimization algorithm. Then thesis introduces the details of Particle Swarm Optimization (PSO) algorithm for the basic principles and standard algorithm. And contrary to the phenomenon of Basic PSO algorithm optimizes high-dimensional complex problems prone to low precision and premature convergence. Through analyzing the factors of the influence algorithm performance (inertia weight and learning factors), this paper presents an adaptive inertia weight and segmented time-varying learning factor of Particle Swarm Optimization. The network weight is trained by using improved particle swarm optimization on the initial BP neural network. Simulation results show that:improved algorithm not only simplifies the initial weights BP neural network selection process, but also overcomes the shortcomings of neural network is easy to fall into local minima and slow convergence.Then contrary to the traditional PID controller used in uncertainty, time-varying, multivariable nonlinear complex systems, it shows slow adaptability, robustness weak and poor coordination shortcomings, applies the improved BP neural network and an improved PSO algorithm to optimize BP neural network to the PID controller design. The three parameters KP, KI KD are self-tuned online. Simulation results show that this method is effectiveness and superiority.Finally, According to the basic composition brushless DC motor theory and mathematical models, builds brushless DC motor speed control system simulation model by using MATLAB/Simulink. We do the comparative experiment between improved PSO algorithm to optimize BP neural network PID controller and the adoption of improved BP neural network PID controller. Results of experiments summarized the advantages and disadvantages of these two methods and give them more suitable occasions.