A novel local neighborhood standardization strategy and its application in fault detection of multimode processes

Abstract

Complex modern industrial processes often have several operating regions, and the multimode process data would follow different distributions. However, most multivariate statistical process monitoring (MSPM) methods, such as principal component analysis (PCA) and partial least squares (PLS), have a fundamental assumption that the operating data follow a unimodal distribution. These data-based MSPM methods cannot perform well when directly applied to multimode processes because the assumption becomes invalid. In this paper, a novel local neighborhood standardization (LNS) strategy is proposed as a data preprocessing method to address the challenges caused by the multimode characteristic of operating data. After a thorough analysis of LNS, a new method called LNS-PCA is developed for fault detection in multimode processes. Multimode data can be scaled to follow one single distribution by using LNS, approximately. Based on these scaled operating data, the monitoring model can be built more accurately by utilizing local information. The advantages of LNS-PCA are that only one model is required for multimode process monitoring and no process knowledge is needed. Finally, the validity and effectiveness of LNS-PCA are illustrated through a numerical example and the Tennessee Eastman process. The results show that the proposed data preprocessing method is very suitable for multimode data normalization and LNS-PCA is superior to traditional PCA for fault detection.