Multi-modeling Soft-sensing Technique and Its Application Based on Neural Network

Based on multi-modeling idea, fuzzy clustering method is used to classify soft-sensing data. For each class, different modeling methods based on artificial neural network are used. Furthermore, RBF neural network is used to build the degree of membership of every sample in this paper. The degrees of membership are used for combing several models to obtain the final result. The method is applied to model a practical case of oxygen content of catalyst-reforming process in petrol refining.