Connected Worker: An AIDT Framework for Safety Excellence in Downstream Petrochemical Operations

Abstract

Abstract This paper presents an Artificial Intelligence and Digital Technology (AIDT) based Connected Worker Framework developed to enhance operational, process, and workforce safety within downstream petrochemical operations. The framework integrates electronic work management, computer-vision monitoring, workforce visibility and accountability, hands-free connected field-support devices, and agent-assisted safety and product-stewardship intelligence within a unified operational model. The implementation was evaluated between November 2025 and June 2026 and included 18 AI-enabled cameras, 964 location-enabled devices deployed across approximately 950 personnel, 22 hands-free connected field-support devices, 100 electronic work permits, five emergency drills, and 1,500 incident records used for agent-assisted categorization and knowledge management. Workforce visibility deployment achieved 100% beacon installation completion, and five emergency drills demonstrated immediate dashboard-based availability of workforce accountability information compared with the previous manual reconciliation process of approximately 15 minutes. Computer-vision monitoring achieved an overall validated detection accuracy of 92% and enabled proactive identification of recurring unsafe behaviors. At the product distribution facility, targeted interventions based on AI-camera observations reduced worker pathway violations from 36 observations to 6 observations, representing an 83% reduction, while average detection rates decreased from 19 observations per day to 2observations per day, representing an 85% reduction. Hands-free connected field-support devices delivered measurable operational benefits during maintenance and turnaround activities. Inspection effort was reduced from 180 minutes to 60 minutes, while troubleshooting and specialist-support decreased from 240 minutes to 15 minutes through remote expert collaboration and real-time visual assistance. The Agent-Assisted Safety Intelligence Platform improved incident categorization, knowledge retrieval, reporting, and product-stewardship support. Response times for safety and product-stewardship enquiries improved from approximately 30 minutes to 5 minutes, significantly reducing manual information-gathering effort previously requiring support from multiple personnel. Classification consistency exceeded 90% when compared with human-reviewed outcomes. The broader work-management performance series showed compliance, increasing from 88% in 2023 to 99% in 2026 year-to-date, while permit non-compliance decreased from 12% to 1%. The results demonstrate that measurable safety and operational improvements can be achieved when digital technologies are integrated with operational workflows, governance processes, and human oversight. The Connected Worker Framework provides a scalable approach for enabling proactive risk management, improving workforce accountability, strengthening operational efficiency, and supporting data-driven safety decision-making in high-hazard industrial environments.