基于软测量技术的间歇聚合过程质量控制  

Quality control for batch polymerization process based on soft sensor technology

作　　者：朱鹏飞[1] 夏陆岳[1] 潘海天[1] 

机构地区：[1]浙江工业大学化学工程学院,浙江杭州310014

出　　处：《计算机与应用化学》2015年第8期959-963,共5页Computers and Applied Chemistry

基　　金：浙江省自然科学基金资助项目(Z4100743)

摘　　要：针对间歇聚合过程质量指标的控制问题,提出了一种基于软测量技术的质量控制方法。将混合核函数偏最小二乘法(K2PLS)与人工神经网络(ANN)相结合,构建一种软测量模型,用于预测工艺变量与质量指标之间的定量关系;利用软测量技术和非线性规划方法,求解得到符合质量指标约束的最佳操作变量;根据离线质量指标分析值,利用间歇过程批次间重复的相似性特性,提出了一种偏差修正策略,用于调整操作变量和指导批次间的生产操作。将上述方法应用于氯乙烯聚合过程的质量指标控制研究中,结果表明:基于K2PLS-ANN的软测量模型具有优秀的预测性能,提出的质量优化控制策略,实现了聚氯乙烯质量指标的平稳控制,有助于降低生产消耗,可用于指导实际生产过程。For the quality index control of a batch polymer production process, a method based on the soft sensor technology is proposed. By combining mixtures of kernels partial least squares （K2PLS） with an artificial neural network （ANN）, a data-driven soft-sensor modeling method was proposed to predict the quantitative information between the process variables and the quality index. The optimal operational variable with quality index constraint was obtained by soft sensor prediction and nonlinear programming. According to the similarity of the batch polymerization process, a deviation elimination strategy was presented to adjust the operational variable by the offline analysis value. The application of the proposed method in the polyvinyl chloride （PVC） quality index control verified that the proposed K2PLS-ANN model performs excellent in predicting the quality index; the control method based on K2PLS-ANN soft sensor model can not only reduce the industrial polymer production cost, but also improve the stability of quality control. It is able to guide the PVC production process.

关 键 词：间歇聚合 混合核函数 预测 优化 质量控制 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]