Second-order component analysis for fault detection. (December 2021)

• Record Type: Journal Article
• Authors: Peng, Jingchao; Zhao, Haitao; Hu, Zhengwei
• Is Part Of: Journal of process control. Volume 108(2021)
• Page Start: 25
• Page End: 39
• Publication Date: 2021-12
• Subjects: Fault detection -- Process monitoring -- High-order neural network -- Orthogonal constraint -- Riemannian manifold; Process control -- Periodicals; Fabrication -- Contrôle -- Périodiques; Process control; Periodicals; Electronic journals; 660.281
• DOI: 10.1016/j.jprocont.2021.10.011
• Languages: English
• ISSNs: 0959-1524

Abstract

Abstract: Process monitoring based on neural networks is getting more and more attention. Compared with classical neural networks, high-order neural networks have natural advantages in dealing with heteroscedastic data. However, high-order neural networks might bring the risk of overfitting, which learning both the key information from original data and noises or anomalies. Orthogonal constraints can greatly reduce correlations between extracted features, thereby reducing the overfitting risk. This paper proposes a novel fault detection method called second-order component analysis (SCA). SCA rules out the heteroscedasticity of process data by optimizing a second-order autoencoder with orthogonal constraints. In order to deal with this constrained optimization problem, a geometric conjugate gradient algorithm is adopted in this paper, which performs geometric optimization on the combination of Stiefel manifold and Euclidean manifold. Extensive experiments on the Tennessee-Eastman benchmark process show that SCA outperforms the compared state-of-the-art methods with missed detection rate (MDR) and false alarm rate (FAR). Highlights: Second-order component analysis (SCA) is proposed for fault detection. SCA uses the structure of autoencoder and adds second-order terms to improve nonlinear mapping capabilities. SCA adopts orthogonal constraints to reduce the overfitting problem. SCA adopts a geometric conjugate gradient (GCG) algorithm to deal with constrained optimization problem.