基于改进核主元分析的TE故障诊断

TE Sfault diagnosis based on improved Kernel Principal Component Analysis

摘要:

PCA、KPCA作为常用的多变量统计监控算法,一般适用于定常过程。针对实际工业过程的时变、非线性特性,提出一种基于分块的改进KPCA算法。该方法通过采用随时间更新的核矩阵代替固定核矩阵用于主元模型的建立,使非线性监控模型能够在线更新,从而提高KPCA的检测正确率。与KPCA方法相比,该方法的运算复杂度明显降低。将该方法应用于TE(Tennessee Eastman)过程,仿真结果显示,该方法具有较好的监测性能,且所需时间大大减小,说明了本算法的有效性。

Abstract:

As widely used process monitoring techniques, principal component analysis(PCA) and kernel PCA are limited to the application in time-invariant systems. To handle the time-varying and nonlinear characteristics of real processes, an improved KPCA based on block was proposed(BKPCA). It using a time-varying kernel matrix instead of a fixed kernel matrix to establish the principal component model, which was suitable for online model updating and much lower computational complexity. When applied to TE process monitoring, computer simulation demonstrates the effectiveness and efficiency of the proposed method.