基于Isomap-MPA-LSSVM软测量模型的PVC聚合过程转化率预测

针对聚氯乙烯（PVC）聚合过程转化率难以实时获取，提出了一种基于Isomap-MPA-LSSVM软测量模型预测氯乙烯单体（VCM）转化率的方法。首先，为消除数据间共线性，采用等度量映射（Isomap）算法对原始数据特征进行提取，得到模型的输入变量。其次，用海洋捕食者算法（MPA）优化最小二乘支持向量机（LSSVM）的参数，并建立模型。最后，进行仿真实验。结果表明，软测量模型相较于MPA-LSSVM与LSSVM模型，预测的平均绝对误差分别降低了43.03%和59.92%，均方根误差分别降低了30.27%和52.24%，显著提升了VCM转化率的预测精度。所提模型能精确预测VCM转化率，对提高PVC的产品质量具有重要参考意义。

Conversion rate prediction of PVC polymerization process

In view of the difficulty in obtaining real-time conversion rate of polyvinyl chloride polymerization process, a soft sensing model based on Isomap-MPA-LSSVM was proposed to predict the conversion rate of vinyl chloride monomer. Firstly, in order to eliminate the collinearity between the data, isometric mapping (Isomap) algorithm was used to extract features from original data and obtain the input variables of the model. Secondly, marine predators algorithm (MPA) was used to optimize the parameters of least squares support vector machine (LSSVM) to establish the model. Finally, the simulation experiment was carried out. The results show that compared with MPA-LSSVM and LSSVM models, the mean absolute error of the proposed soft-sensor model decreases by 4303% and 5992%, respectively, and the root mean square error decreases by 3027% and 5224%, respectively, which significantly improves the prediction accuracy of VCM conversion rate. The model can accurately predict the conversion rate of VCM, which provides an important reference for improving the product quality of PVC products.