Multi-Variable Modeling And Feedback Control For MWD In Polymerization

This paper firstly introduces a new approach to model and control the molecular weight distribution (MWD) for polymerization, and then expresses a new method to control MWD, which is completed based on the estimation by the neural network model.MWD curve is the best parameter for polymer performance analyzing. As there's no effective way to measure it on line at present, it is hard to control MWD on line. So a neural network model should be built, so that the output of it can be instead of the MWD curve, which is the output of the real system. Polymerization process is a high nonlinear complex system with many variables, so the neural network model need to get close to the MWD curve and express the dynamic characteristic of the system at the same time. So in this paper B-spline neural network and dynamic feedback neural network are combined into a complex neural network, and then are used to...