中文摘要：

针对聚氯乙烯粒径分布在线软测量问题,提出了一种基于机理分析和神经网络的混合建模方法,并将该建模方法应用于聚氯乙烯粒径分布建模研究中。混合模型由机理模型和误差补偿模型所组成。通过机理分析建立氯乙烯悬浮聚合过程的单体液滴群体平衡（Population Balance Equation,简称PBE）模型,由于聚氯乙烯成粒过程的复杂性和强非线性,单纯的机理模型预测与实际分析值相比仍存在一定偏差,因此利用人工神经网络建模方法建立了基于BP神经网络的单体液滴群体平衡模型修正模型,对单体液滴群体平衡模型的输出进行修正,由此建立起聚氯乙烯粒径分布混合模型。由于混合模型既能按照液滴分散与聚并机理对聚氯乙烯颗粒的成长过程进行描述,同时又充分利用了生产现场数据对模型误差进行修正,应用到聚氯乙烯生产过程的测试结果表明,与单纯机理模型相比,聚氯乙烯粒径分布混合模型具有更佳的预测效果。

英文摘要：

A parallel hybrid modeling method was proposed for online soft sensoring of polyvinyl chlorideparticle size distribution. The hybrid model was consisted with a dynamic mechanism model and an errorcompensation model. By analyzing the mechanism of vinyl chloride suspension polymerization processes, ageneralized population balance equation （PBE） model was developed. Due to the complexity and highlynon-linear of the polyvinyl chloride granulation process, PBE model was unable to accurately predict theparticle size distribution. Therefore, the error compensation model constituted by BP neural network waspresented to compensate the problem when using the PBE model. This hybrid model can be used to study theevolution mechanism of the polyvinyl chloride particle size distribution, and take the full advantage of the fielddata. The application of the hybrid model in predicting particle size distribution of an industrial vinyl chloridepolymerization process verifies that it is more effective comparing to the dynamic pure mechanism model.