Expert adaptive ANN-PID control in digital invert power supply

For multi-parameter coupling nonlinear welding process, a new algorithm is presented in this paper, which combines expert rules and neural network to achieve optimal matching of PID parameters. Almost all parameters of BP network are researched and analyzed to improve the convergence rate and stability. Five expert rules, including network structure optimization, activation function improvement, weight and learning rate adjustment, are proposed according to mutual relationship between the parameters. Then expert adaptive ANN-PID algorithm is researched. Simulation experiments demonstrate that the welding system based on the algorithm has faster convergence rate and stronger robust than fuzzy-PID, GA-PID and so on. The new algorithm can avoid shortcomings of BP algorithm and meet online real-time control of welding current.