A self-supervised temporal temperature prediction method based on dilated contrastive learning. (December 2022)

• Record Type: Journal Article
• Authors: Lei, Yongxiang; Chen, Xiaofang; Xie, Yongfang; Cen, Lihui
• Is Part Of: Journal of process control. Volume 120(2022)
• Page Start: 150
• Page End: 158
• Publication Date: 2022-12
• Subjects: Contrastive self-supervised learning (CSSL) -- Aluminum electrolysis industry -- Long–short time memory (LSTM) -- Temperature prediction -- Dilated contrastive learning; Process control -- Periodicals; Fabrication -- Contrôle -- Périodiques; Process control; Periodicals; Electronic journals; 660.281
• DOI: 10.1016/j.jprocont.2022.11.005
• Languages: English
• ISSNs: 0959-1524

Abstract

Abstract: Due to the scarcity of the labeled data, traditional supervised learning methods have a limited application scope, which caused the supervised-based model performance will greatly be decreased. In this paper, we propose a promising model based on self-supervised learning. To update the weight and the contrastive relation in the features, a new self-supervised loss, is introduced. First, the convolution neural network is used in the proposed network to extract the deep feature in the first processing. Second, the self-supervised long–short time memory (LSTM) sequential is constructed for further deal. At last, the teacher net and student net have coordinately fine-tuned the credibility of the temperature prediction. By the experimental comparison, our proposed CNN-SSDLSTM is competitive with other supervised and semi-supervised methods. The evaluation experiments achieve state-of-the-art performance in aluminum electrolysis temperature prediction applications. Graphical abstract: Highlights: A novel self-supervised architecture for temperature identification is proposed. The model achieves great accuracy with the constraint of the scarcity of labeled data. A new self-supervised loss is used in the proposed architecture. This loss can fully utilize the information contained in the unlabeled data, and cucumber the limit of the label data. A performance strategy for contrastive learning is used in the training of the model. The dual net of a student and teacher net is used to dilate the knowledge of the cosine loss.