Vision-Based Monitoring of Worker Operation Sequences During Blowout Preventer Flange Bolt Installation
Vision-based monitoring has shown potential for industrial process monitoring, but tracking continuous manual operations against prescribed procedures remains challenging. This study presents a vision-based method for monitoring worker operation sequences during blowout preventer (BOP) flange bolt installation, where occlusion, visually similar bolts, and object-scale variation complicate operation identification. The proposed method integrates three stages: Multi-Scale Visual Perception, Specific Bolt Operation Identification, and Spatiotemporal Operation Sequence Monitoring. It detects operation-relevant objects, associates hand and torque-meter observations with physical Bolt IDs using Bolt Operation Regions, and reconstructs operation sequences through spatial filtering, temporal consistency verification, and logical consolidation. The reconstructed sequences are compared with the prescribed tightening rule to identify rule-based sequence deviations. The perception model achieved a mean average precision (mAP) of 57.1% on the Oilfield Surveillance Object Detection Dataset comprising 3303 images. On 218 four-view installation recordings, the proposed method achieved an Average Longest Common Subsequence (LCS) Accuracy of 95.06% and a Mean Normalized Edit Distance (NED) of 7.46%. The Precision and Recall for sequence-deviation recognition were 97.02% and 96.96%, respectively, with a False Positive Rate (FPR) of 1.02%. These results support Bolt ID-level operation sequence reconstruction, rule-based tightening-sequence deviation monitoring, and process traceability within the evaluated BOP installation setting.
Authors
- Yingxue Liu (ORCID: https://orcid.org/0000-0002-9585-7159)
- Sha He
- Yanjin Luo
- Jinrong Ma
- Ruihao Yuan
- Shiqiang Wang
- Qiang Feng
- Yonggang Deng
- Rong Wang
Institutions
- China University of Petroleum, Beijing (CN)
- State Administration for Market Regulation (CN)
- CNPC Chuanqing Drilling Engineering Company Limited (China) (CN)
- China National Petroleum Corporation (China) (CN)
Publication Details
- Journal
- Processes
- Published
- 2026-11-17
- DOI
- https://doi.org/10.3390/pr14223699
- Primary Topic
- Industrial Vision Systems and Defect Detection
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- Science Foundation of China University of Petroleum, Beijing
- National Key Research and Development Program of China