Simultaneous fault detection and isolation using semi‐supervised kernel nonnegative matrix factorization

Abstract

This paper presents a monitoring approach for nonlinear processes based on a new semi‐supervised kernel nonnegative matrix factorization (SKNMF). Different from the existing nonnegative matrix factorization (NMF) and kernel nonnegative matrix factorization (KNMF), SKNMF is a semi‐supervised matrix factorization algorithm, which takes advantages of both labelled and unlabelled samples to improve algorithm performance. Labelled samples refer to the samples whose memberships are already known, while unlabelled samples are a set of samples whose memberships are unknown. In fact, both NMF and KNMF are unsupervised algorithms, and they cannot make full use of labelled samples to improve algorithm performance. More importantly, we explain the reasons why labelled samples can improve algorithm performance even if the amount of labelled samples is small. Last but not least, SKNMF induces a simultaneous fault detection and isolation scheme for online processes monitoring. Case studies of a numerical example and a penicillin fermentation process (PFP) demonstrate that the proposed process monitoring approaches outperform the existing process monitoring approaches.

Issue Highlights

“…Moreover, the semi-supervised kernel nonnegative matrix factorization algorithm is developed in an all-new manner so that it can take advantage of labelled and unlabelled samples simultaneously to train a model for fault detection and isolation. [2] The study proposed a new model for fault diagnosis of the Tennessee Eastman process. Support vector machine recursive feature elimination in the model was used for feature selection, which can reduce the training time of the probabilistic neural network and improve classification performance.…”

“…The fault detection and isolation approach is developed with the aid of a semi‐supervised kernel nonnegative matrix factorization algorithm. Moreover, the semi‐supervised kernel nonnegative matrix factorization algorithm is developed in an all‐new manner so that it can take advantage of labelled and unlabelled samples simultaneously to train a model for fault detection and isolation …”

Fault clustering by small‐entropy nonnegative matrix factorizations

“…Zhai and Jia established a kernel NMF model using both labelled and unlabelled samples and proposed a fault detection and isolation method for nonlinear processes. [ 24 ] Due to the use of unlabelled samples in the model training phase, the proposed fault detection and isolation method show better performance than traditional NMF‐based fault detection and isolation methods, especially in the case of insufficient labelled samples. Jia et al proposed a simultaneous fault detection and isolation method based on transfer semi‐NMFs.…”

Fault Detection and Isolation of Industrial Fermentation Process Based on Semi-supervised Convex Nonnegative Matrix Factorizations

“…In the training stage, a set of labeled samples is used to train the model. In the evaluation stage, the trained model is used to infer the patterns of the new samples (Alcala and Qin, 2011;Chen and Patton, 2012;Zhai and Jia, 2019).…”