A Systematic Literature Review on Artificial Intelligence-based Industrial Safety

Objective: This study aims to systematically analyze the current status and trends of artificial intelligence-based industrial safety research published in Korea over the past five years, identifying research field distribution, applied technologies, research purposes, and suggesting future research directions.<br/>Background: With the advent of the Fourth Industrial Revolution and the implementation of the Serious Accident Punishment Act, interest in artificial intelligence-based industrial safety technologies has rapidly increased. However, systematic analysis of the overall status and trends of related research in Korea is very limited.<br/>Method: A systematic literature review methodology applying PRISMA guidelines was used. Academic papers from KCI-indexed journals published between 2020 and 2025 were searched using &quotindustrial safety artificial intelligence&quot and &quotindustrial safety AI&quot as keywords in the DBpia academic database. From the initial 148 search results, 29 duplicate papers and 51 low-relevance papers were excluded, resulting in 68 papers for final analysis.<br/>Results: By research field, construction (35.3%), transportation/aviation/railway (25.0%), energy/nuclear power (22.1%), and manufacturing (17.6%) showed the most active research. Applied AI technologies were deep learning (38.2%), computer vision (33.8%), machine learning (19.1%), and IoT/edge computing (8.8%), while research purposes were monitoring/alarm system development (39.7%), risk prediction (32.4%), automation (19.1%), and education/profiling (8.8%). Research showed concentration in 2021 and 2024.<br/>Conclusion: I-based industrial safety research is concentrated in specific high-risk fields such as construction (35.3%), transportation/aviation/railway (25.0%), and energy /nuclear (22.1%), with deep learning and computer vision as mainstream technologies. However, most studies focus on technology development and system construction, lacking verification of practical field applicability and economic analysis. Research on system interoperability, standardization, and ethical issues is also insufficient.<br/>Application: Future research needs to expand to various industrial sectors including chemical, food, and service industries, verify practical applicability through long-term field operations, establish performance evaluation standards and certification systems for AI safety systems, and conduct human factors research considering human-AI interaction.