Abstract
Modified partial least squares (MPLS) is an efficient tool widely used in multivariate statistical process monitoring. To properly describe slow time-varying processes, the method commonly used in model updating is data expansion. However, when the number of lagged variables grows, the modeling order and computational load increase significantly. A recursive structure that has a low computational complexity is an efficient way to update models. However, the recursive structure fails to reduce the false alarm rates (FARs) of quality-unrelated faults when model updating. To cope with this problem, a recursive MPLS (RMPLS) detection approach based on orthogonal signal correction (OSC), named OSC-RMPLS here, is proposed. OSC-RMPLS removes the variation orthogonal to the output space Y from the input space X before MPLS modeling, decomposes X into two orthogonal subspaces, and further updates the RMPLS model adaptively. Compared with the data expansion method, the proposed algorithm has a lower computational load and more robust performance. Numerical simulation experiments and Tennessee Eastman experiments (TEP) are used to verify the effectiveness of the proposed algorithm.
Highlights
Data-driven process monitoring techniques [1]–[4] are very popular in industrial process safety detection due to their easy implementation and low requirements for the underlying model
Principal component analysis (PCA) [5] and partial least squares (PLS) [6] are two typical statistical methods that are used in data-driven processes
The fault false alarm rate (FAR) and fault detection rate (FDR) in reference [16] are used as indicators to measure the monitoring performance of the algorithm
Summary
Data-driven process monitoring techniques [1]–[4] are very popular in industrial process safety detection due to their easy implementation and low requirements for the underlying model. To update the model efficiently and take into account the ability to distinguish the two spaces, we derive an MPLS-based recursive model update structure that replaces the raw data from the data expansion with historical model parameters and new test data This method reduces the computational complexity of the model update while orthogonally dividing the input data into two subspaces. Postprocessing such as recursive MPLS (RMPLS) cannot completely remove information that is useless for the prediction quality in PCS. The proposed approach removes the large variation that is orthogonal to Y from X It divides the filtered X into two orthogonal subspaces using MPLS and updates model adaptively by RMPLS.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
