基于群智能优化算法的聚丙烯熔融指数预报建模优化研究

发布时间：2018-10-13 08:28

【摘要】：聚丙烯作为全球五大通用塑料之一,广泛地应用于日常生活、工业生产、航空航天等领域。因此,聚丙烯生产过程中的产品质量控制就尤为重要。熔融指数是决定聚丙烯产品质量的重要指标。本文针对聚丙烯熔融指数软测量预报问题,引入了最小二乘支持向量机(LSSVM),通过群智能优化算法对LSSVM模型的参数进行优化;同时,对群智能优化算法进行改进,有效地提高了模型的预报性能;本文提出的几种智能LSSVM模型成功地应用于工厂实际数据预报中,为实际工业生产中的熔融指数预报提供了更多的选择。本文的主要工作及贡献如下:(1)针对聚丙烯工业生产过程,选取了合适的输入输出变量,通过主元分析方法(PCA),对输入变量进行降维处理,建立了聚丙烯熔融指数预报LSSVM模型;实验结果证明了熔融指数预报LSSVM模型的有效性;(2)提出了一种熔融指数预报AMPSO-LSSVM模型。为了克服粒子群优化算法(PSO)易陷于局部最优值、收敛精度低等缺陷,根据粒子群进化状态引入了变异操作,提出了一种自适应变异PSO算法优化熔融指数预报LSSVM模型的参数;实验结果表明AMPSO-LSSVM模型具有更好的预报性能和推广泛化能力,证明了 AMPSO智能优化算法的有效性;(3)进一步提出了一种熔融指数预报MSFLA-LSSVM模型。针对标准混合蛙跳算法(SFLA)易陷入局部最优值等缺陷,在SFLA算法的全局搜索过程中引入了自适应混沌变异,提出了一种改进的MSFLA算法对LSSVM模型的参数进行优化,建立了熔融指数预报MSFLA-LSSVM模型;实验结果表明MSFLA-LSSVM模型的预报性能在SFLA-LSSVM模型的基础上有了进一步的提高,证明了MSFLA智能优化算法的有效性。
[Abstract]:As one of the five universal plastics in the world, polypropylene is widely used in daily life, industrial production, aerospace and other fields. Therefore, the product quality control in polypropylene production process is particularly important. Melt index is an important index to determine the quality of polypropylene products. In this paper, the parameters of the LSSVM model are optimized by using the swarm intelligence optimization algorithm, and the swarm intelligence optimization algorithm is improved by introducing the least square support vector machine (LSSVM),) to the soft sensor prediction problem of polypropylene melt index. Several intelligent LSSVM models proposed in this paper have been successfully applied to the actual data prediction of the factory, which provides more choices for the prediction of melting index in actual industrial production. The main work and contributions of this paper are as follows: (1) according to the production process of polypropylene industry, appropriate input and output variables are selected, the dimension reduction of input variables is processed by principal component analysis (PCA), and the LSSVM model of PP melt index prediction is established. The experimental results show that the melting index can predict the LSSVM model effectively. (2) A melting index prediction AMPSO-LSSVM model is proposed. In order to overcome the defects of particle swarm optimization (PSO) algorithm, such as local optimal value and low convergence precision, a mutation operation is introduced according to the evolution state of particle swarm optimization, and an adaptive mutation PSO algorithm is proposed to optimize the parameters of the fusion index prediction LSSVM model. The experimental results show that the AMPSO-LSSVM model has better prediction performance and generalization ability, and proves the effectiveness of the AMPSO intelligent optimization algorithm. (3) A fusion index prediction MSFLA-LSSVM model is proposed. Aiming at the defects of standard hybrid leapfrog algorithm (SFLA) which is easy to fall into local optimal value, adaptive chaos mutation is introduced into the global search process of SFLA algorithm, and an improved MSFLA algorithm is proposed to optimize the parameters of LSSVM model. The experimental results show that the prediction performance of the MSFLA-LSSVM model is further improved on the basis of the SFLA-LSSVM model, which proves the effectiveness of the MSFLA intelligent optimization algorithm.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TQ325.14;TP18

【相似文献】