A new automated job hazard analysis tool developed by researchers at the University of South Australia (UniSA) could significantly reduce workplace accidents and improve safety in the construction industry, where 60,000 workers lose their lives annually.
The research team, led by Dr Sonali Pandithawatta, has created a “knowledge graph” to predict hazards on construction sites, addressing the critical need to automate job hazard analysis (JHA).
This innovation comes at a crucial time, as construction sites account for nearly 20 per cent of occupational deaths worldwide.
Dr Pandithawatta explained: “Traditionally, potential job hazards rely heavily on safety personnel identifying risks and control measures, a process that is prone to inefficiencies and human error.
“Our research addresses the critical need to automate job hazard analysis.”
The algorithm was designed using data from over 100 JHA documents and input from 18 industry experts.
It integrates information such as weather, job steps, hazards, and preventive measures into a comprehensive graph.
Professor Rameez Rameezdeen, co-author of the study, highlighted the tool’s exceptional accuracy of over 90 per cent.
The model can analyse primary and secondary hazards, weather conditions, workplace proximity, and atmospheric hazards in real time.
“We are proud to contribute to the global effort in making construction sites safer,” said Professor Rameezdeen, adding that the tool could potentially be adapted for use in other high-risk industries.
The research team plans to assess additional risk factors, including human and managerial influences, and integrate advanced machine learning techniques for broader application.
The study, published in the Journal of Engineering, Project and Production Management, was supported by the South Australian Water Corporation (SA Water).
This innovative tool represents a significant step forward in construction safety, potentially saving thousands of lives and improving working conditions for millions of construction workers worldwide.
