Representation, mining and analysis of unsafe behaviour based on pan-scene data.

Area of Science:

• Industrial Safety Engineering
• Data Mining and Analytics
• Human Factors in Engineering

Background:

• Industrial processes generate vast amounts of data, crucial for understanding and preventing accidents.
• Unsafe behaviors are a primary cause of industrial accidents, necessitating detailed analysis.
• Existing safety rule descriptions often lack a comprehensive, data-driven approach to behavioral analysis.

Purpose of the Study:

• To propose association rules for unsafe behavior based on pan-scene industrial data.
• To describe and structurally transform unsafe behavior using eight key dimensions.
• To explore the distribution and interaction of unsafe behavior dimensions.

Main Methods:

• Scene data theory applied to define unsafe behavior across eight dimensions.
• Apriori algorithm utilized for multidimensional association rule mining.
• Empirical analysis using SPSS Modeler on subway construction pan-scene data.

Main Results:

• Identified strong association rules between work type, construction stage, working time, and unsafe actions.
• Demonstrated the frequent occurrence of irregular binding of lifting objects by machine operators during excavation.
• Highlighted specific unsafe actions linked to particular construction stages and working times.

Conclusions:

• The proposed pan-scene association rules effectively describe and analyze unsafe behaviors in industrial settings.
• Findings provide targeted insights for safety management and accident reduction in construction.
• Data-driven identification of behavioral patterns can proactively mitigate risks.