Internal Model Control Based On LMBP Neural Network For The Polymerization Reaction

The paper discusses the application of the Internal Model Control Strategy in component's control of propylene polymerization reaction's production, which is based on the LMBP Neural Network (NN) algorithm. Using the approach ability of neural network to the nonlinear system, to extend the neural network control to the polymerization process quality control. The control strategy adopts the Internal Model Control (IMC), which is a new control technology with great developing potential. Selecting LMBP Neural Network as a training network, LMBP network has good performance as a Forward NN, it can ensure not only global convergence, but also the best approximation performance. Through the reaction mechanism, to establish the model (internal model) and the inverse model (controller model) of the propylene polymerization process, and to get the effective control of components. As a result of simulation, the predictive control strategy based on neural network has attained finer control quality. For the Internal Model Control based on the LMBP neural network for the propylene polymerization reaction, I have completed the work as follows:1. By consulting references, to understand the polymerization principle, be familiar with propylene polymerization process and various quality indicators of the polymers. Having mastered the characteristics,structure and algorithm of the Artificial Neural Network, and understood the conventional methods and structure of the Internal Model Control, then having studied and done the research in the Internal Model Control strategy based on the Neural Networks.2. The IMC with the LMBP neural network in the application of the propylene polymerization reaction has been researched. Mainly including the following aspects:①Building the experiment of the propylene polymerization. Getting the necessary training and testing data for the LMBP neural network modeling , analysing and processing the data necessarily.②The LMBP neural network has been used to train forward model (internal model) and the inverse model (controller model) of the object.③Researching and programming the Internal Model Control project by the LMBP neural network training model., and to achieve effective control of the components.