Based On Fuzzy Neural Network, Nonlinear System Modeling And Industrial Applications

A type of compound fuzzy neural network that has a function network and a rule network is proposed in this paper for building nonlinear system models of chemical process. The local behavior of the process is approximated by the function network, with its improved fuzzy neural network structure, and based on process knowledge, the process operation is partitioned into a number of local operating regions by the rule network. For the improved fuzzy neural network and the compound fuzzy neural network, simulations have been made with some nonlinear function models, and the results of which show that the compound fuzzy neural network is better than the ordinary fuzzy neural network in train precision and predication precision, and shows an excellent ability in predication and generalization. Simulation has been made with some white noise models, which results show that the compound fuzzy neural network has a good robust and resist disturb ability. Moreover, a preliminary discuss has been made with the compound fuzzy neural network with outer feedback dynamic structure, and simulation has been made with a two-order dynamic nonlinear function model. The result shows that it has some dynamic ability. Finally, in a urea high press circulation process of LUTIANHUA Group Incorporated, the compound fuzzy neural network has been applied to three important level systems. A lot of real time data have been collected from the field, and the compound fuzzy neural network is applied to the modeling and prediction of the level systems, which results are satisfactory. This study shows that the compound fuzzy neural network not only possesses evident dominant in modeling precision, but also possesses a good prospect in the application of production process.