基于PCA和LS-SVM的丁苯橡胶的门尼粘度预测  

Prediction for the mooney-viscosity of SBR based on PCA and LS-SVM

作　　者：刘美[1,2] 黄道平[1] 孙宗海[1] 

机构地区：[1]茂名学院自动化系,广东茂名525000 [2]华南理工大学自动化科学与工程学院,广东广州510640

出　　处：《计算机与应用化学》2008年第11期1317-1320,共4页Computers and Applied Chemistry

基　　金：国家自然科学基金项目(60704012);广东省自然科学基金项目(06300232);广东省科技计划项目(2004B10301035);茂名市科技计划项目(201095)．

摘　　要：门尼粘度是合成橡胶生产的主要质量指标,如何在线监测门尼粘度,并实现质量的自动监控是橡胶生产工业亟待解决的问题。本文应用主元分析和最小二乘支持向量机法建立生产过程门尼粘度预测模型。结合工艺机理分析,找出影响橡胶门尼粘度的主要参数并做主元分析,确定最少辅助变量,简化支持向量机结构,建立基于PCA LS-SVM的门尼粘度预测模型。仿真结果,门尼粘度预测值与实际值最大相对误差为5.78%,预测模型精度高,泛化能力强,运行速度快,可以指导生产。Mooney-viscosity is the main quality index of Synthetic Rubber. Monitoring the mooney-viscosity effectively and realizing automatic quality control of production process is an urgent problem in rubber industry. This paper proposes a prediction model of the mooney-viscosity based on PCA and LS-SVM. First, several major parameters affecting the mooney-viscosity were chosen based on mechanism analysis. Principal components were extracted by PCA and used as secondary variables of LS-SVM. Then, a prediction model for the mooney-viscosity was established by LS-SVM. The simulation results show that the maximum relative error of mooney-viscosity was 5.78%. The model features high accuracy, good approximation and generalization ability. This data may be used to guide efficiently production.

关 键 词：丁苯橡胶 门尼粘度 预测 最小二乘支持向量机 主元分析法 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]