基于最大矩阵元法确定聚类数的软测量建模方法

Soft-sensing modeling method based on the maximal matrix element to determine the clustering number

采用模糊C均值聚类算法(FCM)与支持向量机(SVM)相结合的多模型建模方法:较单一支持向量机软测量模型而言,可以有效解决复杂工业对象的强非线性和大工况范围的问题。但是传统的模糊C均值聚类算法必须依赖先验知识预先确定聚类个数。本文通过建立样本间的相似矩阵,利用模糊聚类最大矩阵元法确定FCM最佳聚类个数,再由FCM对训练样本数据进行聚类并用SVM构建组合软测量模型,得到多模型软测量系统。在对双酚A结晶单元工艺分析的基础上,将该方法:应用于结晶单元苯酚含量的软测量建模,仿真结果:证明该建模方法:提高了模型的估计精度,具有更好的可行性和有效性,能够满足工业生产的要求。

Multi-model modeling method based on Fuzzy C-Means Clustering （FCM） and Support Vector Machines （SVM） can deal with the properties of strong nonlinear and large range of operating conditions in the complex production process more effectively than single Support vector machine model. However, to determine the clustering number of the traditional Fuzzy C-Means Clustering algorithm must rely on the priori knowledge. In this paper, the clustering number is determined by the method of fuzzy clustering maximal matrix element made use of the similar matrix of the samples, and then to build the combination soft-sensor model combined with FCM and SVM to get the multi-model soft sensor system. The method is applied to a soft-sensing model to estimate the phenol content in a Bisphenol-A crystal production process based on the process analysis, and the simulation results show the feasibility and the effectiveness of the method, and can meet the requirements of industrial production.