Internal Model Control Via RBF Neural Network In Polymerization

This thesis aims at primarily the studies of nonlinear internal model control (IMC) strategies based on radial basis function (RBF) network models designed for bulk polymerization of styrene. It primarily studies these problems: the importance of polymerization process control; constructing the model of bulk polymerization of styrene; building up the models via RBF neural network for the linear system; making use of the RBF neural network to control the SISO system of bulk polymerization of styrene; the research in controlling MIMO system of bulk polymerization of styrene.As the RBF neural network can approximate to the nonlinear system perfectly, we can train the models with RBF neural network. Using an improved RBF algorithm, we can get a better neural network structure. At the same time, the internal model controller (the inverse of the model) can also obtained by training the RBF neural network. After the work above all, internal model control system for SISO model can be obtained. It is proved that the internal model control system has a strong robust performance and can reject the disturbances. As there is static error during building up the models with RBF neural network, the static data can be added to the training data, which effectively improved the verification accuracy of the model, and also improved the control accuracy of the internal model controller and eliminated system residual error as well. This thesis still studied internal model control based RBF neural network for MIMO system, and also extended to the control of molecular weight distribution. This experiment acquired good control results.The simulation results show that the strategies of internal model control based RBF neural network are satisfied.