集成主角及其在故障检测中的应用

Ensemble principal angle and its application to fault detection

为了解决稀疏化造成的核主角不稳定问题,提出了集成主角方法.集成主角求再生核希尔伯特空间的多组近似基,将核主角问题极值向量的解空间限定在近似基张成的空间求核主角,然后集成特征值.利用集成主角(ensemble principal angle,EPA)可以对复杂环境下的多变量工业过程进行在线故障检测.最后本文通过在Tennessee Eastman数据集上的实验,对集成主角在故障检测中的应用进行了说明.

Ensemble principal angle is proposed to deal with the instability of a sparse kernel principal angle. Groups of approximate basis are found in the Reproducing Kernel Hilbert Space. Eigenvectors for the principal angle problem are limited to the spaces spanned by the approximate basis. Eigenvalues in different subspace are integrated to make up for the sparsity. Ensemble principal angle （EPA） can be applied to online multivariable process for fault detection in complicated conditions. An example is given to illustrate the application in fault detection by performing experiments on the tennessee eastman data set.