摘要
针对单一软测量模型难以精确描述复杂非线性的化工生产过程的问题,为提高软测量模型的预测精度,基于多模型建模思想,提出一种基于二次判别分析的支持向量机多模型建模方法。首先依据样本输出空间的值区间把样本集合分为若干子集,并分别对每个子集建立基于支持向量机算法的子模型,多个子模型采用"开关切换"方式连接。对于未知类别的输入数据,依据各子集的先验类别信息,用二次判别分析算法判断其所属类别,并以输入向量所属类别的支持向量机模型的输出作为多模型的最终输出。工业仿真实例表明,该建模方法建立的多支持向量机模型比单一支持向量机模型具有更高的预测精度。
To problem that it is difficult to describe the complex nonlinear chemical production process precisely by a single soft sensing model,a multi-model modeling method based on quadratic discriminant analysis and support vector machine is proposed to improve the prediction precision of the soft sensing model.The sample data are divided into several subsets by taking the value interval of the object variable,and the sub-models are trained by the support vector machine for each subset.The sub-models are combined in a＂switch＂way.For the unclassified input data,its classification can be determined by the quadratic discriminant analysis algorithm based on the prior classification information and the final-output of the multi-SVM model is represented by the output of the corresponding sub-model according to the input data.The industrial simulation shows that the multi-SVM model has a prediction precision higher than the one of a single SVM model.
出处
《控制工程》
CSCD
北大核心
2010年第5期662-664,685,共4页
Control Engineering of China
基金
国家自然科学基金资助项目(60674092)
江苏省高技术研究资助项目(BG20060010)
关键词
多模型
二次判别分析
支持向量机
软测量
multi-model
quadratic discriminant analysis
support vector machine
soft sensor
