Multi-scale extension of PLS algorithm for advanced on-line process monitoring

Classification of multiple time signals using localized frequency characteristics applied to industrial process monitoring

“…Further, wavelet packets (a generalisation of wavelets which results in a much richer set of potential basis functions) have been used in modeling transient underwater signals (Learned and Willsky, 1995), sales data (Michis, 2009), wastewater filtering (Lee et al, 2009), sleep state modeling ) and wind speed prediction (Hunt and Nason, 2001;Nason and Sapatinas, 2002). The greater flexibility of wavelet packets comes at a price; the number of potential covariates given n observations on a single predictor variable is log 2 (n) when using the non-decimated wavelet transform, but 2n − 2 when using the non-decimated wavelet packet transform.…”

Variable importance in PLS in the presence of autocorrelated data — Case studies in manufacturing processes

“…This approach, although shown to be reasonable for process monitoring and control, results in very large unfolded matrices, which after modeling may pose challenges in model interpretation. A second approach using wavelet analysis on each process variable has also been proposed [3,4]. Two challenges associated with this method are the critical, but somewhat subjective, choice of wavelet; secondly, the complexity of this method might be an obstacle for new users.…”

Improved anomaly detection using multi-scale PLS and generalized likelihood ratio test

“…In fact, MSPLS model is able to remove the autocorrelations of variables by wavelet analysis and to remove correlations between variables with PLS transformation [11]. The author in [11] demonstrates that multiscale representation of data improved the FD abilities of conventional PLS. Thus, combining the advantages of MSPLS with those of generalized likelihood ratio (GLR) hypothesis testing should provide even further improvements in fault detection.…”

“…Wavelet based-multiscale representation of data has been used extensively in literature to ameliorate the effectiveness and robustness of fault detection strategies [9], [10]. Regarding multiscale PLS (MSPLS) modeling and monitoring, in [11], [12] the author employed multiscale representation to develop multiscale MSPLS in order to enhance the accuracy of PLS model. In fact, MSPLS model is able to remove the autocorrelations of variables by wavelet analysis and to remove correlations between variables with PLS transformation [11].…”

“…Regarding multiscale PLS (MSPLS) modeling and monitoring, in [11], [12] the author employed multiscale representation to develop multiscale MSPLS in order to enhance the accuracy of PLS model. In fact, MSPLS model is able to remove the autocorrelations of variables by wavelet analysis and to remove correlations between variables with PLS transformation [11]. The author in [11] demonstrates that multiscale representation of data improved the FD abilities of conventional PLS.…”