Melt index prediction by fuzzy functions with dynamic fuzzy neural networks

Melt Index (MI) is an important variable in determining the product specifications and grades of polypropylene (PP), whose accurate prediction is crucial in the quality control of PP processes. In this paper, a novel predictive model of MI, fuzzy functions with dynamic fuzzy neural networks (FF-D-FNN), which combines the universal approximation property of fuzzy functions and the parsimonious structure with high performance of dynamic fuzzy neural networks (D-FNN), is first proposed. Research on the FF-D-FNN is further accomplished with the data from a real PP plant and the results are compared among the neural network, fuzzy neural networks, D-FNN and FF-D-FNN models. The research results show the accuracy and universality of the presented model for the online MI prediction.