Monitoring multimode processes: A modified PCA algorithm with continual learning ability

• Published in: Journal of process control 2021-07, Vol.103, p.76-86
• Main Authors: Zhang, Jingxin, Zhou, Donghua, Chen, Maoyin
• Format: Article
• Language: English
• Subjects: Catastrophic forgetting; Continual learning; Elastic weight consolidation; Multimode process monitoring; Principal component analysis
• ISSN: 0959-1524; 1873-2771
• DOI: 10.1016/j.jprocont.2021.05.007

For multimode processes, one generally establishes local monitoring models corresponding to local modes. However, the significant features of previous modes may be catastrophically forgotten when a monitoring model for the current mode is built. It would result in an abrupt performance decrease. It could be an effective manner to make local monitoring model remember the features of previous modes. Choosing the principal component analysis (PCA) as a basic monitoring model, we try to resolve this problem. A modified PCA algorithm is built with continual learning ability for monitoring multimode processes, which adopts elastic weight consolidation (EWC) to overcome catastrophic forgetting of PCA for successive modes. It is called PCA–EWC, where the significant features of previous modes are preserved when a PCA model is established for the current mode. The optimal parameters are acquired by difference of convex functions. Moreover, the proposed PCA–EWC is extended to general multimode processes and the procedure is presented. The computational complexity and key parameters are discussed to further understand the relationship between PCA and the proposed algorithm. Potential limitations and relevant solutions are pointed to understand the algorithm further. A numerical case study and a practical industrial system in China are employed to illustrate the effectiveness of the proposed algorithm. •A novel PCA–EWC framework is proposed for monitoring multimode processes.•It builds one model with continual learning ability, without retraining from scratch.•The model is updated continually based on the learned knowledge and current mode data.•It is not required that data from all modes are available in advance.