生产状态的熵值评估谱聚类分析

Production state spectral clustering analysis based on entropy ranking

实际生产过程呈现多模态,通过聚类分析可以了解生产状态,进行生产故障诊断或有针对性的质量检测,传统的线性分析方法难以有效提取非线性特性。谱聚类是较为先进的聚类方法,但常规的谱聚类分析是按照特征值的大小来进行特征选择的,而特征值的大小表示数据在特征向量上的方差信息;实际生产过程数据分布复杂,将熵值估计引入谱聚类特征选择中,并应用于生产过程状态的聚类分析中,分别利用标准数据、TE生产过程数据对方法的有效性进行验证。验证结果表明熵值评估谱聚类方法取得了更优的聚类结果,可以更加有效了解生产过程状态。

The real production process data are always multi-model. The clustering method with process data is used to acquire the production status. The traditional linear methods can not extract the nonlinear relationship among the data. Spectral clustering is one of the advanced methods, where the eigenvector is selected based on the eigenvalue. The eigenvalue only can express the variation information of the data. In real field, there are the complex distribution data. The entropy ranking is introduced to spectral clustering, which is used to do the production state clustering. The benchmark data, Tennessee Eastman（TE） process data are used for model validation, as a result the proposed method has better performance on clustering to acquire the production state, compared with the classical spectral clustering.