Work

Research on variable-length control chart pattern recognition based on sliding window method and SECNN-BiLSTM

Year: 2025

Type: article

Abstract: Control charts, as essential tools in Statistical Process Control (SPC), are frequently used to analyze whether production processes are under control. Most existing control chart recognition methods target fixed-length data, failing to meet the needs of recognizing variable-length control charts in production. This paper proposes a variable-length control chart recognition method based on Sliding Window Method and SE-attention CNN and Bi-LSTM (SECNN-BiLSTM). A cloud-edge integrated recognition system was developed using wireless digital calipers, embedded devices, and cloud computing. Different length control chart data is transformed from one-dimensional to two-dimensional matrices using a sliding window approach and then fed into a deep learning network combining SE-attention CNN and Bi-LSTM. This network, inspired by residual structures, extracts multiple features to build a control chart recognition model. Simulations, the cloud-edge recognition system, and engineering applications demonstrate that this method efficiently and accurately recognizes variable-length control charts, establishing a foundation for more efficient pattern recognition.

Source: Scientific Reports

DOI: 10.1038/s41598-025-86849-4

Authors: Tao Zan, Xiaolong Jia, Xiaoyu Guo, Min Wang, Xiangsheng Gao, Peng Gao

Institution: Beijing University of Technology

Language: English

Topic: Advanced Statistical Process Monitoring

Subfield: Statistics, Probability and Uncertainty

Field: Decision Sciences

Domain: Social Sciences

Open Access status: gold

Estimated APC paid: $2,690

Funder: National Natural Science Foundation of China