Using Artificial Intelligence and Machine Learning Techniques to Analyze Incident Reports

Hazards conference paper applying AI/ML to an oil and gas incident-report database to reduce human bias in risk-rating and uncover causal relationships. Incidents are classified against CCPS PSM elements and then used as inputs to a Bayesian Network to identify influential factors affecting incident occurrence, including Hazard Identification & Risk Analysis. For Process Safety Management, this demonstrates a practical pathway to turn routine incident data into actionable insights: improving hazard awareness, strengthening risk ranking consistency, and focusing improvement efforts on the PSM elements that most strongly influence incident frequency.