TY - JOUR
T1 - Continual Semisupervised Learning of Echo State Network for Quality Prediction of Multimode Processes
AU - Yang, Chao
AU - Liu, Qiang
AU - Liu, Yi
AU - Cheung, Yiu-ming
N1 - This work was supported in part by the National Natural Science Foundation of China under Grant 62161160338, Grant U23A20328, Grant 61991401, and Grant U20A20189, in part by the National Natural Science Foundation of China (NSFC)/Research Grants Council (RGC) Joint Research Scheme (N_HKBU214/21), the General Research Fund of RGC under Grant 12202924, Grant 12202622, and Grant 12201323, in part by the RGC Senior Research Fellow Scheme under Grant SRFS2324-2S02, and in part by the China Scholarship Council under Grant 202306440060. Paper no. TII-25-0230.
PY - 2025/6/12
Y1 - 2025/6/12
N2 - The successive switching nature of multimode processes, coupled with data scarcity, challenges traditional quality prediction models. Specifically, the difficulty of simultaneously collecting abundant labeled datasets from all modes forces the model to update its parameters as modes switch. This leads to the forgetting of historical mode knowledge and hinders the aggregation of knowledge, thereby degrading generalization across modes. To this end, we propose a novel continual semisupervised graph echo state network (CS2GESN). First, a semisupervised graph echo state network (S2GESN) is designed based on the graph smoothing assumption to extract dynamic information from unlabeled samples within each mode. The S2GESN model then evolves into a continual model, CS2GESN, employing an elastic weight consolidation strategy for parameter importance estimation derived from pseudoinverse parameter optimization, facilitating the accumulation of historically learned knowledge. This manner alleviates performance deterioration from data scarcity and information forgetting, and enables more flexible modeling of successive arriving operating modes. The superiority and feasibility of the proposed method are demonstrated through its application to the Tennessee Eastman process and the three-phase flow facility process.
AB - The successive switching nature of multimode processes, coupled with data scarcity, challenges traditional quality prediction models. Specifically, the difficulty of simultaneously collecting abundant labeled datasets from all modes forces the model to update its parameters as modes switch. This leads to the forgetting of historical mode knowledge and hinders the aggregation of knowledge, thereby degrading generalization across modes. To this end, we propose a novel continual semisupervised graph echo state network (CS2GESN). First, a semisupervised graph echo state network (S2GESN) is designed based on the graph smoothing assumption to extract dynamic information from unlabeled samples within each mode. The S2GESN model then evolves into a continual model, CS2GESN, employing an elastic weight consolidation strategy for parameter importance estimation derived from pseudoinverse parameter optimization, facilitating the accumulation of historically learned knowledge. This manner alleviates performance deterioration from data scarcity and information forgetting, and enables more flexible modeling of successive arriving operating modes. The superiority and feasibility of the proposed method are demonstrated through its application to the Tennessee Eastman process and the three-phase flow facility process.
KW - Catastrophic forgetting
KW - continual learning (CL)
KW - echo state network (ESN)
KW - multimode processes
KW - quality prediction
KW - semisupervised learning (SSL)
UR - http://www.scopus.com/inward/record.url?scp=105008272885&partnerID=8YFLogxK
U2 - 10.1109/TII.2025.3575101
DO - 10.1109/TII.2025.3575101
M3 - Journal article
SN - 1551-3203
JO - IEEE Transactions on Industrial Informatics
JF - IEEE Transactions on Industrial Informatics
ER -