Self-Learning Sparse PCA for Multimode Process Monitoring

This article proposes a novel sparse principal component analysis algorithm with self-learning ability for multimode process monitoring, where the successive modes are learned in sequential fashion. Different from traditional monitoring methods, small set of data collected when mode arrives. The proposed method remembers knowledge by selectively slowing down changes parameters important previous modes, importance measure is estimated synaptic intelligence. sufficient condition fault detectability proved to provide comprehensive understanding method. Besides, computation and storage resources saved long run, because it not necessary retrain model scratch frequently discarded once they have been learned. More importantly, furnishes excellent interpretability catastrophic forgetting problem further alleviated owing sparsity parameters. In addition, hyperparameters discussed understand comprehensively computational complexity analyzed. Compared several state-of-the-art approaches, numerical case, practical pulverizing system adopted illustrate effectiveness algorithm.