Nonlinear Dynamic Process Monitoring Using Canonical Variate Analysis and KernelDensity Estimations

摘要

The Principal Component Analysis (PCA) and the Partial Least Squares (PLS) aretwo commonly used techniques for process monitoring. Both PCA and PLS assumethat the data to be analysed are not self-correlated i.e. time-independent.However, most industrial processes are dynamic so that the assumption of time-independence made by the PCA and the PLS is invalid in nature. Dynamicextensions to PCA and PLS, so called DPCA and DPLS, have been developed toaddress this problem, however, unsatisfactorily. Nevertheless, the CanonicalVariate Analysis (CVA) is a state-space-based monitoring tool, hence is moresuitable for dynamic monitoring than DPCA and DPLS. The CVA is a linear tool andtraditionally for simplicity, the upper control limit (UCL) of monitoringmetrics associated with the CVA is derived based on a Gaussian assumption.However, most industrial processes are nonlinear and the Gaussian assumption isinvalid for such processes so that CVA with a UCL based on this assumption maynot be able to correctly identify underlying faults. In this work, a newmonitoring technique using the CVA with UCLs derived from the estimatedprobability density function through kernel density estimations (KDEs) isproposed and applied to the simulated nonlinear Tennessee Eastman Process Plant.The proposed CVA with KDE approach is able to significantly improve themonitoring performance and detect faults earlier when compared to other methodsalso examined in this study.