Automating Process Safety Barrier Management Using AI: A Case Study on GCC Onshore COMAH Site

Abstract This paper presents the development and implementation of an artificial intelligence (AI) solution aimed at automating the identification and verification of process safety barriers and generating performance standards. The goal is to enhance process safety management and operational efficiency in the energy industry. Our approach involved developing an AI system that leverages Retrieval-Augmented Generation (RAG) and the latest Large Language Models (LLMs) to identify Health, Safety, Environment, and Critical Equipment Systems (HSECES) equipment based on inclusion rules defined in primary engineering deliverables, HSECES management guidelines, Bow Tie, and COMAH reports. The AI system extracts and populates the HSECES register at the equipment level, referencing source documents. It then generates relevant performance standards, compares them with existing standards, and provides recommendations for improvement. The AI verifies implementation in the computerized maintenance management system (CMMS), reviews completed work orders, and highlights discrepancies and areas for improvement via an integrated dashboard. The system also performs reliability analysis to identify trends and maintenance optimization opportunities. The incorporation of AI significantly reduced the required engineering effort and is scalable across existing assets and new sites. The work resulted in substantial improvements in the accuracy and efficiency of HSECES identification and performance standard generation. The AI successfully populated the HSECES register, ensuring accurate equipment identification and allocation of relevant barriers, verified by engineering review. The performance standards generated were comprehensive and aligned with site operator requirements and industry standards. Verification in CMMS ensured accurate task list development and implementation. Using AI to review completed work orders identified and addressed discrepancies, leading to improved compliance and practices. Reliability analysis highlighted areas for performance improvement and optimization, contributing to enhanced plant uptime and reduced engineering effort. The use of RAG and advanced LLMs facilitated real-time, contextually relevant data processing, ensuring robust control of major accident hazards. The case study demonstrates the potential of AI to transform safety and maintenance processes, offering valuable insights and advancements in the field. The automation of these processes reduces incidents, improves plant uptime, and reduces engineering effort by over 90%, setting a new benchmark for AI-driven safety management in the energy industry.