Scholarlink

In the context of Industry 4.0, Digital Twin (DT) technology offers significant potential to enhance industrial safety. However, its adoption in Process Safety Management (PSM) remains limited, especially in delivering real-time, predictive capabilities. This paper proposes a scalable and adaptive Process Safety Digital Twin (PS-DT) architecture for real-time risk prediction and integrity barrier monitoring in oil and gas operations. The architecture integrates sensor data with a trained Bayesian Network (BN) and a dynamic Bow-Tie model to enable continuous monitoring of failure risks and predictive hazard analysis. The PS-DT system is aligned with major industry standards, including API 754 for Tier 3 safety indicators, API 521 for pressure-relieving scenarios, OSHA 1910.119 for compliance, and the IBM Bow-Tie framework for barrier-based risk visualization. These standards address the limitations of current static risk assessments by enabling dynamic monitoring of barrier degradation and early warnings of hazard escalation. To validate the architecture, a prototype use case is implemented using a Pressure Relief Valve (PRV)—a critical Safety Critical Equipment (SCE)—to simulate barrier impairments and overpressure scenarios. Synthetic time-series data is generated in Python to train the BN, which estimates the probabilities of overpressure events, Loss of Process Containment (LOPC), and other consequences. A dynamic Bow-Tie visualization tracks evolving risk pathways, while interactive dashboards present Tier 3 KPIs, barrier health, and recommended actions to support decision-making. This research demonstrates how the PS-DT system can enhance operational safety by shifting from reactive to proactive risk management. The proposed architecture is modular, ISA-95-aligned, and designed for easy integration with existing plant systems, making it extensible across multiple SCEs and applicable to broader process industries.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Architecture for Dynamic Process Safety Management Using Digital Twin and Machine Learning

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Ibrahim Aljahwari,
  • Vittaldas Prabhu

摘要

In the context of Industry 4.0, Digital Twin (DT) technology offers significant potential to enhance industrial safety. However, its adoption in Process Safety Management (PSM) remains limited, especially in delivering real-time, predictive capabilities. This paper proposes a scalable and adaptive Process Safety Digital Twin (PS-DT) architecture for real-time risk prediction and integrity barrier monitoring in oil and gas operations. The architecture integrates sensor data with a trained Bayesian Network (BN) and a dynamic Bow-Tie model to enable continuous monitoring of failure risks and predictive hazard analysis. The PS-DT system is aligned with major industry standards, including API 754 for Tier 3 safety indicators, API 521 for pressure-relieving scenarios, OSHA 1910.119 for compliance, and the IBM Bow-Tie framework for barrier-based risk visualization. These standards address the limitations of current static risk assessments by enabling dynamic monitoring of barrier degradation and early warnings of hazard escalation. To validate the architecture, a prototype use case is implemented using a Pressure Relief Valve (PRV)—a critical Safety Critical Equipment (SCE)—to simulate barrier impairments and overpressure scenarios. Synthetic time-series data is generated in Python to train the BN, which estimates the probabilities of overpressure events, Loss of Process Containment (LOPC), and other consequences. A dynamic Bow-Tie visualization tracks evolving risk pathways, while interactive dashboards present Tier 3 KPIs, barrier health, and recommended actions to support decision-making. This research demonstrates how the PS-DT system can enhance operational safety by shifting from reactive to proactive risk management. The proposed architecture is modular, ISA-95-aligned, and designed for easy integration with existing plant systems, making it extensible across multiple SCEs and applicable to broader process industries.