Research of Chemical Modeling Method Based on Rough-Fuzzy Inference System

Based on the rule derived from the Rough Sets methodology, we designed the fuzzy neural networks. Using the regular parameter and the default value that are estimated by the discretization results, the network is up to an optimal value rapidly by large amount of training. When applied to model the solvent dehydrating tower in the PTA complex process, the performance is superior to the common feed-forward neural network. The fuzzy neural network can eliminate redundant information of the decision system and reduce modeling complexity. In the practical application, the system represents its dominance at quickly convergence and powerful generalization.