Process Production Quality Prediction Algorithm Based on Mar-G LSTM

In response to the characteristics of strong continuity and complex temporal coupling in process production, traditional neural networks lack long-term memory capabilities and tend to encounter issues such as parameter disaster and gradient explosion during deep network training. This paper proposes a combined prediction model based on Markov optimization, integrating Gated Recurrent Units (GRU) and Long Short-Term Memory networks (LSTM), termed Mar-G LSTM. Firstly, a gated mechanism is incorporated into the recurrent neural network structure to construct a deep LSTM neural network model that selectively remembers temporal data information, learning the information dependencies of temporal data sequences, thereby addressing the gradient explosion issue during training. Simultaneously, the prediction results of the GRU-LSTM model are corrected and optimized using Markov chains, further improving the prediction accuracy while reducing model complexity. Finally, analysis and verification are conducted using the process data from a specific production line, showing that the Mar-G LSTM algorithm improves prediction accuracy by 37.42%, 21.32%, 17.91%, and 12.56% compared to the Random Forest model, GRU model, LSTM model, and CNN-GRU model, respectively. The proposed Mar-G LSTM algorithm can achieve accurate predictions of process production quality, providing ideas and implementation pathways to reduce the completion time of process parameter control tasks.

Yin Yanchao (1977-)

Female, from Anyang, Henan, Professor at Kunming University of Science and Technology, PhD, PhD supervisor, research direction: Intelligent Manufacturing, Industrial Big Data, etc.

E-mail:[email protected]

Su Yifan (1997-)

Female, from Zhengzhou, Henan, Master’s student at Kunming University of Science and Technology, research direction: Machine Learning, Intelligent Algorithms, Industrial Big Data, Corresponding Author.

E-mail:[email protected]

Tang Jun (1984-)

Male, from Guilin, Guangxi, Senior Engineer at Yunnan Tobacco Industrial Co., Ltd., PhD, research direction: Processing Technology and Equipment, Data Mining, etc.

E-mail:[email protected]

Lin Wenqiang (1984-)

Male, from Guigang, Guangxi, Engineer at Yunnan Tobacco Industrial Co., Ltd., Master’s student, research direction: Processing Technology and Equipment, Data Processing, etc.

E-mail:[email protected]

Pu Haoran (1995-)

Male, from Guangyuan, Sichuan, Master’s student at Kunming University of Science and Technology, research direction: Big Data Mining, Knowledge Graph.

E-mail:[email protected]

Wang Linyu (1998-)

Male, from Shangrao, Jiangxi, Master’s student at Kunming University of Science and Technology, research direction: Machine Learning, Data Processing, Intelligent Algorithms.

E-mail:[email protected]

Yin Yanchao, Su Yifan, Tang Jun, Lin Wenqiang, Pu Haoran, Wang Linyu. Process Production Quality Prediction Algorithm Based on Mar-G LSTM. Computer Integrated Manufacturing Systems, 2024, 30(3):942-957.

DOI:10.13196/j.cims.2023.0126