The Research on CMAC Network Model Based on Rough Sets for Flatness Control

It was hard to set up an accurate mathematics model in cold rolling process and the structure of neural network was hard to confirm, so a CMAC neural network controller based on rough sets theory was proposed for flatness control. The original model of CMAC neural network flatness control was set up. Simultaneity dynamic learning rate was improved in the error correction algorithm of this neural network to speed up the network convergence rate, to overcome disadvantage of network with fixed learning rate. Rough set theory was used to optimize the structure of the neural network, and eliminate redundant linked weights and nodes of the network to solve the problem of the needs of a large space of CMAC. The neural network´s training speed and precision was improved. The simulation result shows that the control effect of CMAC neural network controller based on rough sets theory was good, and could get the precision of the intelligence on-line control for shape.