Artificial intelligence as ally in hazard analysis. Issue 3 (16th February 2021)

• Record Type: Journal Article
• Authors: Garvin, Thomas; Kimbleton, Scott
• Is Part Of: Process safety progress. Volume 40:Issue 3(2021)
• Page Start: 43
• Page End: 49
• Publication Date: 2021-02-16
• Subjects: ai -- artificial_intelligence -- hse -- machine_learning -- safety -- text_analytics -- unstructured_data; Chemical plants -- Management -- Periodicals; 660
• DOI: 10.1002/prs.12243
• Languages: English
• ISSNs: 1066-8527

Abstract

Abstract: Hazard analysis techniques have been around for many years, and have proven effective in the prevention of incidents and no doubt the saving of lives. Process hazard analysis (PHA) is now fairly robust and regulated, focused on overarching risks associated with the safe handling of hazardous materials and approaches to engineer‐out such risks. Occupational hazard analysis (OHA) is keenly focused on human activity, and personal protection in hazardous working conditions. Both approaches are critical ‐ but are often carried out separately, by different parts of an organization, which could result in an incomplete picture of the full set of operational risks in the field. Developing a holistic picture of both past and present dangers calls for a deep exploration of evidence. HAZOPs, PHA's, incident records and investigations provide expert analysis of hazards and mitigating strategies. Near‐miss reports and safety observations add a large amount of information as well; the reporting frequency of these "leading indicators" can be both a blessing and a curse, as time and available resources constrain the ability to analyze and detect hazard signals within. As important as analyzing the historical record is for lessons learned, the more recent observations could indicate new hazards or highlight concerning trends. These could feed valuable "real time" information back to operations and maintenance teams to improve risk assessments and task planning. Enter artificial intelligence (AI) as a means to analyze the large amount of written hazard analyses, reports and observations to quickly extract insights around hazardous conditions, activities, incident causes and risk mitigation measures. Trained to understand concepts and contexts in both process and personal safety, AI can provide a natural‐language information exploration environment for scanning thousands of documents in seconds and present common themes and related records. Not unlike us humans, AI learns from the past, informs the present and can help reduce risks in the future.