Research and Application of the Pellet Grate Thickness Control System Base on Improved CMAC Neural Network Algorithm

摘要

Based on the analysis of the structure parameters of CMAC neural network, the concept of association membership is proposed, and the mapping relation between quantization space and output layer is established. Thus, a novel Association Membership CMAC (AM-CMAC) is designed and its convergence proof is given. According to the fitting and the approximation of nonlinear function and the industrial control that can be solved by data and experience, AM-CMAC identification implement and feedfor-ward controller are also designed. The simulation result shows that the generalization ability and learning accuracy are improved. And the algorithm of AM-CMAC is applied in the pellet grate thickness control system. The application result shows that the pellet grate speed of regulation of AM-CMAC algorithm is smoother, and can better track the change of raw material quantity.