MétaCan
Menu
Back to cohort
Record W7089626961 · doi:10.1016/j.rcim.2025.103162

Proactive safety reasoning in human-robot collaboration in disassembly through LLM-augmented STPA and FMEA

2025· article· en· W7089626961 on OpenAlexafffund

Bibliographic record

VenueRobotics and Computer-Integrated Manufacturing · 2025
Typearticle
Languageen
FieldHealth Professions
TopicOccupational Health and Safety Research
Canadian institutionsInstitut de recherche Robert-Sauvé en santé et en sécurité du travailPolytechnique Montréal
FundersNatural Sciences and Engineering Research Council of Canada
KeywordsFailure mode and effects analysisHazard analysisHazardProcess (computing)System safetyTracingControl (management)Risk assessment

Abstract

fetched live from OpenAlex

• RAG+KG framework enables real-time, standards-compliant safety analysis in HRC. • GPT-4.1 delivers the most accurate and compliant safety recommendations in tests. • STPA and FMEA integration supports both control-level and component-level risks. • EV battery case study confirms hazard detection and mitigation across risk types. • Six safety-oriented evaluation metrics introduced, achieving 92 % hazard recall. Disassembly tasks in human–robot collaboration (HRC) environments present safety challenges due to hazardous materials, control system variability, and physically demanding operator tasks. To address these challenges, we propose an AI-augmented risk assessment framework integrating System-Theoretic Process Analysis (STPA) and Failure Mode and Effects Analysis (FMEA). This framework is implemented in four configurations: Term Frequency–Inverse Document Frequency (TF-IDF), Fine-tuned Large Language Models (LLMs), Retrieval-Augmented Generation (RAG), and RAG with a structured Knowledge Graph (KG) built from safety standards. The system supports real-time, standards-compliant safety reasoning by generating interpretable, context-specific recommendations. We evaluate these configurations across GPT-3.5 TURBO, GPT-4o, GPT-4.1, and open-source LLMs Qwen2.5 (3B) and Ministral (3B). Among all, RAG+KG with GPT-4.1 achieved the highest results across language-based metrics (BLEU: 68.3, ROUGE-L: 72.0, Semantic Similarity: 81.1, BERTScore (F1): 90.0) and safety-specific metrics (Hazard Recall: 92, Compliance Precision: 97, Safety Violation Rate: zero). Six safety-oriented metrics were introduced to assess compliance, hazard coverage, interpretability, and robustness. A case study on electrical vehicle (EV) battery module disassembly demonstrated the system’s effectiveness in identifying unsafe control actions, tracing failure modes, and recommending targeted mitigation strategies for mechanical, electrical, and chemical hazards, and ergonomic considerations. This framework offers a scalable, explainable approach to real-time safety analysis, advancing AI-enabled risk assessment in dynamic HRC disassembly tasks and supporting the vision of human-centered Industry 5.0 manufacturing. LLM-Augmented Risk Analysis in Human–Robot Disassembly: From Safety Challenges to Standards-Compliant Real-Time Recommendations

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.

How this classification was reachedexpand

Full frame machine prediction

Teacher imitation

Not calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.

metaresearch head score (Codex)0.002
metaresearch head score (Gemma)0.008
Version: metacan-v3-hybrid-931329e0061cValidation status: machine_predicted_unvalidated
Candidate categoriesnone
Consensus categoriesnone
DomainCandidate signal: none · Consensus signal: none
Study designCandidate signal: Simulation or modeling · Consensus signal: none
GenreCandidate signal: Empirical · Consensus signal: none
Teacher disagreement score0.013
Threshold uncertainty score0.026

Distilled classifier scores by category (both heads)

CategoryCodexGemma
Metaresearch0.0020.008
Meta-epidemiology (narrow)0.0020.000
Meta-epidemiology (broad)0.0010.002
Bibliometrics0.0020.001
Science and technology studies0.0000.001
Scholarly communication0.0010.002
Open science0.0020.002
Research integrity0.0010.002
Insufficient payload (model declined to judge)0.0040.001

Machine scores (provisional)

The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.

Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.

Opus teacher head0.032
GPT teacher head0.407
Teacher spread0.375 · how far apart the two teachers sit on this one work
Validation statusscore_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it

Classification

machine, unvalidated

Machine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.

The models applied no category: nothing in the taxonomy fit this work.
Study designSimulation or modeling
Domainnot available
GenreEmpirical

How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".

Quick stats

Citations6
Published2025
Admission routes2
Has abstractyes

Explore more

Same venueRobotics and Computer-Integrated ManufacturingSame topicOccupational Health and Safety ResearchFrench-language works237,207