The cost effective benefits of process monitoring will never be over emphasised.Amongst monitoring techniques, the Independent Component Analysis (ICA) is anefficient tool to reveal hidden factors from process measurements, which follownon-Gaussian distributions. Conventionally, most ICA algorithms adopt thePrincipal Component Analysis (PCA) as a pre-processing tool for dimensionreduction and de-correlation before extracting the independent components (ICs).However, due to the static nature of the PCA, such algorithms are not suitablefor dynamic process monitoring. The dynamic extension of the ICA (DICA), similarto the dynamic PCA, is able to deal with dynamic processes, howeverunsatisfactorily. On the other hand, the Canonical Variate Analysis(CVA) is anideal tool for dynamic process monitoring, however is not sufficient fornonlinear systems where most measurements follow non-Gaussian distributions. Toimprove the performance of nonlinear dynamic process monitoring, a state spacebased ICA (SSICA) approach is proposed in this work. Unlike the conventionalICA, the proposed algorithm employs the CVA as a dimension reduction tool toconstruct a state space, from where statistically independent components areextracted for process monitoring. The proposed SSICA is applied to the TennesseeEastman Process Plant as a case study. It shows that the new SSICA providesbetter monitoring performance and detect some faults earlier than otherapproaches, such as the DICA and the CVA. (C) 2010 Elsevier B.V. All rightsreserved.
展开▼