Modeling And Optimization Of MI Prediction For Propylene Polymerization Process Based On Swarm Intelligent Optimization Algorithm

As one of the five most commonly used plastics,polypropylene(PP)is playing an increasingly important role in social life,industry,military,etc.Thus,the quality control of PP in the propylene polymerization process is quite crucial.Melt index(MI)is one of the most important indicators to determine the quality of PP,which makes the prediction of MI important.In this paper,the least square support vector machine(LSSVM)is implied to build the soft sensor MI prediction model,then swarm intelligent optimization algorithms are used to optimize the parameters of LSSVM model.Several improved algorithms are presented to enhance the performance of LSSVM model.These optimized models have great performance in the practical data from the industrial plant,thus there are more options for MI prediction.The major work and contributions of this article are listed as follows:(1)Based on the propylene polymerization process,several related variables are selected to build LSSVM model for MI prediction,meanwhile the principal component analysis is employed to reduce the dimension of input variables.The experimental result proves the validity of the LSSVM model.(2)An AMPSO-LSSVM model for MI prediction is proposed.Considering the particle swarm optimization(PSO)has the drawback of trapping into local optimum,the evolutionary state estimation technique and mutation operation are added into the basic PSO,then AMPSO is used to optimize the parameters of LSSVM model,the AMPSO-LSSVM model is developed.The experimental result shows the AMPSO-LSSVM model has better performance than PSO-LSSVM model,APSO-LSSVM model,which proves the validity of AMPSO.(3)A MSFLA-LSSVM model for MI prediction is proposed further.In order to avoid trapping into local optimum,an adaptive chaotic mutation is added into the process of global searching in the basic shuffled frog-leaping algorithm(SFLA),then the modified SFLA(MSFLA)is used to optimize the parameters of LSSVM model,the MSFLA-LSSVM model is developed.The experimental result shows the MSFLA-LSSVM model has better performance than SFLA-LSSVM model,which proves the validity of MSFLA.