Utility-grade safety intelligence for thermal power: knowledge graph-augmented multimodal monitoring

Q: What does this research mean for the field?

A knowledge graph-augmented multimodal monitoring system significantly improves the accuracy and response time of safety hazard identification in thermal power plants compared to traditional pipelines. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

Xinrong Yan; Chongbo Zhou; Jiaxu Qian; Zhengtao Ding

doi:10.1631/jzus.a2500399

September 7, 2026Journal of Zhejiang University. Science A

Utility-grade safety intelligence for thermal power: knowledge graph-augmented multimodal monitoring

View Full Paper

Ask AI

Bookmark

Share

Key Points

Develop a rapid, rule-interpretable multimodal safety monitoring system to resolve delayed hazard identification and poor traceability in thermal power coal handling operations.
Built a safety knowledge graph from operating rules and incident logs to guide multimodal data fusion across video, sensor streams, and text logs from a 235,000-sample dataset (2018–2024).
Employed low-rank adaptation, safety-aware proximal policy optimization, and knowledge distillation to deploy an 8-bit quantized 7B edge model compressed from a 72B teacher model in an edge-cloud setup.
Generated 3,652 alerts during a two-week operational field trial, yielding 90%–100% expert-verified precision for severe safety hazards.
Outperformed conventional pipelines on representative hazards, including missing hard hats (98.7% vs. 76.4%) and conveyor belt blockages (96.8% vs. 84.3%).
Achieved a median alert response time of approximately 10 s (95th percentile ≤ 20 s), representing a 67% speed improvement over baseline while retaining auditable records.

Implication

Field validation demonstrates high-accuracy hazard detection in thermal power plants, highlighting the utility of knowledge graph-augmented multimodal systems.

Abstract

No abstract yet. The journal hasn't published one we can show.

Discussion

No takes yet. Share an insight, caveat, or question.

Cite This Study

Yan et al. (2026) studied this question.

synapsesocial.com/papers/6a9fd7c658e84d0ff5b46d15 https://doi.org/10.1631/jzus.a2500399

View Full Paper

Ask AI

Bookmark

Share

Utility-grade safety intelligence for thermal power: knowledge graph-augmented multimodal monitoring

Key Points

Implication

Abstract

Discussion

Cite This Study

Utility-grade safety intelligence for thermal power: knowledge graph-augmented multimodal monitoring

Key Points

Implication

Abstract

Discussion

Cite This Study

Keep reading

More related papers

More from Journal of Zhejiang University. Science A

Keep reading

More related papers

More from Journal of Zhejiang University. Science A