Application Research On The Soft-sensing Of Vinyl Acetate Polymerization Rate Based On Least Squares Support Vector Machine

With the development of society, people’s environmental consciousness are raisingand the commodities globalization competition are intensifying, the market requires morestringent to the energe consumption and quality of the product. The Vinyl acetate(VAC) polymerization rate which is a very important parameter for the process ofVAC polymerization producting downstream chemical products, it has obvious influenceon energy consumption, the production process safety, economy and quality etc.. However, atpresent most factories of our country for VAC polymerization rate detection are only remainin the manual offline laboratory analysis detection phase, it can not achieve industry onlinerapid detection. Therefore, this paper uses the principle of soft measuring technology to studythe VAC polymerization rate detection, according to the characteristics of VACpolymerization, it uses soft online VAC polymerization rate measurement method based onleast squares support vector machine (LSSVM) as the core. The main work is as follows:1. This paper describes the process and mechanism of the polymerization of VAC, andalso analyses the relevant factors affecting the VAC polymerization rate, through theintroduction of the ideas of soft measuring technology, the major methods and generalimplementation steps, it uses soft sensing technology to solve the difficult problem ofonline detection VAC polymerization rate.2. The idea and the mathematical basis of support vector machine which has beendetailed introduced, through the simulation performance of support vector machine, settingup the soft measurement model of VAC polymerization rate by LSSVM; The MATLABsimulation results show that, LSSVM soft sensor model can be well track VACpolymerization rate of real value in fitting accuracy.3. Aiming at the problem that can be difficult to choose parameters which significantlyinfluence the performance of LSSVM model, and for the analysis found that the traditionalempirical method and conventional grid method which have defect of low efficiency inparameter optimization search, this paper uses a genetic algorithm for intelligent selectionof the optimal parameters of LSSVM; The paper improves the shortages ofthe genetic algorithm such as it is easy to fall into local solution, finally, it establishs softmeasurement model base on the improved adaptive genetic algorithm(IAGA) optimize theparameters of LSSVM, through the VAC polymerization rate simulation tests show that themodel has a high the prediction accuracy, adaptive parameter optimization performance andgeneralization performance. 4. According to the deficiency of LSSVM which can not meet the production process ofVAC online learning samples to update the model data requirements, the paper hasimproved the LSSVM and made online learning LSSVM, after the parameters areoptimized by IAGA, it establishs the soft measurement model based on an improved onlineleast squares support vector machine (IOLSSVM). The simulation results show that,IOLSSVM not only meet the demand of VAC polymerization rate soft measurement underonline production situation, but also has very high precision and generalization ability.The study of this paper provides a new solution to realize the implementation of VACpolymerization rate online detection, which has the important practical significance.