Soft sensor model based on improved fuzzy neural network

By changing the consequents of the fuzzy rules using wavelet basis function, an improved fuzzy neural network is introduced for soft sensor. In order to improve the system convergence, an efficient initialization is used by the selection of wavelet base and the orthogonal least-square (OLS) algorithm. The parameters of the model are trained by the steepest gradient decent method and least-square estimation. Finally a soft sensor model of the concentration of hydrochloric acid for a chemical plant based on the proposed method is presented which has fast convergence and prediction precision