Unifying PCA and multiscale approaches to fault detection and isolation

Abstract Process signals represent the cumulative effects of many underlying process phenomena. Multiresolution analysis is used to decompose the cumulative process effects. The decomposed process measurements are rearranged according to their scales, and PCA is applied to these multi scale data to capture process variable correlations occurring at different scales. Choosing an orthonormal mother wavelet allows each principal component to be a function of the process variables at only one scale level. The proposed method can identify when a multiscale approach is needed. The conventional PCA as well as MSPCA models are shown as the limiting cases of the proposed model. A procedure for both fault detection and isolation is presented. The proposed method is discussed and illustrated in detail using simulated data from a CSTR system. A comparison study is done through Monte Carlo simulation. The proposed method significantly enhances FDI performance by using additional scale information.