Research On The Soft-Sensing Of Vinyl Acetate Polymerization Rate Based On Neural Networks Optimized By Genetic Algorithm

This paper aims at currently the vinyl acetate (VAC) polymerization rate not having effectively real-time and on-line detection method. Through researching and analysising of VAC polymerization process, raising to use the soft-sensing technology applied in on-line detection rate of VAC polymerization. In a comprehensive studies of the soft-sensing technique principle, components and technological process, Focus on three aspects of soft-sensing technique contenting secondary variables selection, data preprocessing and soft-sensing modeling. Then promotes researching on the soft-sensing of VAC polymerization rate, The concrete research works are follows:1,Firstly, selecting VAC polymerization rate as master variable, the initiator addition ratio, methanol ratio, polymerization temperature and VAC acting degree as auxiliary variables; using data normalization transformation and Pauta rule for original data processing; using MATLAB7.1 for researching on the VAC polymerization rate soft-sensing based on BP network. In the process of researching, discussing the principle, structure and the algorithm of the BP network in detail. During the modeling based on BP neural network, using the error analysising combining with programming method to determine the topology of BP network. In the network training processing, using the Levenberg-Marquardt optimization algorithm. Then according to the testing and training results, evaluating the project of the VAC polymerization rate soft-sensing based on BP network.2,According to local optimal defects of BP network, deciding to use genetic algorithm to optimize the BP neural network. This paper comprehensive elaborates the principle and realization process of genetic algorithm, especially the way of genetic algorithm combining with BP neural network. Deciding to optimize BP network by using genetic algorithm to determine the connection weight of BP network in order to established the genetic algorithm- BP (GA-BP) network. Finally, using GA-BP network for the modeling of VAC polymerization rate soft-sensing. From the results of simulation researching, The training and testing results of GA-BP network has greatly improving comparing with BP network, achieving the requirement.The results proves soft-sensing of VAC polymerization rate based on neural networks optimized by genetic algorithm is feasible and has very important significance for combining soft-sensing technology, neural network and genetic algorithm to solve the technical problem and realize the on-line detection of VAC polymerization rate.