WCSE

Predicting Process Quality Indicators Based on a Transfer Learning Approach That Combines TCN and GRU Hybrid Networks

Abstract—The evaluation of process quality indicators directly determines the product quality, and the prediction of process quality indicators plays a critical role in process industry product quality and production scheduling. In order to deeply explore effective information contained in massive time-series process data and improve the practicality of algorithm models under different working conditions, while ensuring the accuracy of process quality indicator prediction, this paper proposes a process quality indicator prediction algorithm based on a transfer learning approach that combines TCN (Temporal Convolutional Network) and GRU (Gated Recurrent Unit) hybrid networks. The algorithm constructs a continuous feature matrix as input by sliding windows over massive process data, operating data, and time data in a sequential manner. In terms of model construction, TCN is used to extract temporal features from sequence data and GRU is used to capture temporal dependencies in sequence data. To address the problem of training efficiency in actual production, a transfer learning strategy is introduced which greatly improves training efficiency while maintaining training accuracy. Furthermore, a case study using process data from a micro-cigarette production line in a tobacco factory verifies the correctness and effectiveness of the proposed algorithm.

Index Terms—Process industry; TCN-GRU; Hybrid model; Transfer learning

Bin Yi, Wenqiang Lin, Wenqi Li, Xiaohua Gao, Bing Zhou
Technology Center, China Tobacco Yunnan Industrial Co., Ltd., CHINA
Junjun Fang, Xiaoli Xu
Affiliation Yuxi Cigarette Factory, Hongta Tobacco (Group) Co., Ltd., CHINA
Jun Tang
Technology Center, China Tobacco Yunnan Industrial Co., Ltd., CHINA

[Download]

Cite: Bin Yi, Wenqiang Lin, Wenqi Li, Xiaohua Gao, Bing Zhou, Junjun Fang, Xiaoli Xu, Jun Tang, "Predicting Process Quality Indicators Based on a Transfer Learning Approach That Combines TCN and GRU Hybrid Networks" Proceedings of 2023 the 13th International Workshop on Computer Science and Engineering (WCSE 2023), pp. 327-335, June 16-18, 2023.