Fault detection of chemical processes using improved generalized likelihood ratio test

Abstract

In this paper, we address the problem of fault detection (FD) of chemical processes using improved generalized likelihood ratio test. The improved GLRT is the method that combines the advantages of the exponentially weighted moving average (EWMA) filter with those of the GLRT method. The idea behind the developed EWMA-GLRT is to compute a new GLRT statistic that integrates current and previous data information in a decreasing exponential fashion giving more weight to the more recent data. The FD problem will be addressed so that the kernel partial least square (KPLS) is used as a modeling framework and the generated residuals are evaluated using the developed EWMA-GLRT chart. The KPLS model is capable of dealing with high dimensional input-output nonlinear and multivariate data. Therefore, in this paper, KPLS-based EWMA-GLRT method will be utilized in practice to help improve FD of chemical processes. The FD performance is assessed and evaluated in terms of false alarm rate, missed detection rate and ARLi values.