Session: 15-05-01: Models and Methods for Probabilistic Risk Assessment

Paper Number: 173198

Integrated Cfd–dynamic Event Tree Framework for Quantitative Risk Assessment of Hazardous Gas Leaks 

Highly toxic gases such as hydrogen fluoride (HF), chlorine (Cl₂), and phosphine (PH₃) are essential in semiconductor manufacturing and chemical processing industries. While these substances are critical for core operations, accidental releases pose significant risks to personnel and equipment. HF’s high toxicity—evidenced by stringent limits like AEGLs and IDLH—means even small leaks can cause severe harm, highlighting the need for realistic, comprehensive risk assessments.

Traditional static probabilistic risk assessment (PRA) methods often fail to capture the dynamic nature of gas dispersion and the time-dependent aspects of worker exposure. These approaches typically use average conditions and may not fully account for the interactions among leak characteristics, ventilation effectiveness, and safety systems. To address these limitations, this study introduces an integrated risk assessment framework that combines dynamic event tree (DET) analysis with transient computational fluid dynamics (CFD) simulations. This methodology enables more detailed and quantitative evaluation of accidental release scenarios involving highly toxic gases, consistent with current safety standards and best practices in industry.

In this study, ISO/ASME standard piping systems were modeled to simulate leaks at typical vulnerable points such as flanges, valves, and joints. Key leak parameters—including orifice diameter (1–5 mm), internal pressure, and flow rate—were systematically varied to represent realistic conditions. Transient CFD simulations were performed to generate time-concentration profiles for critical locations, such as worker breathing zones and equipment. These time-resolved concentration data were incorporated into the DET model to define branch points based on exposure durations exceeding IDLH values and the operational status of ventilation systems. Branch probabilities were estimated using published reliability data and widely accepted industry assumptions, enabling a comprehensive estimation of potential exposures and equipment impacts across multiple event sequences.

Preliminary results indicate that, under typical ventilation rates of 6–12 air changes per hour (ACH), the individual annual risk remains below the widely accepted threshold of 1 × 10⁻⁶ per year. However, scenarios involving ventilation system failure or substantial leaks could exceed this limit, highlighting residual risks that require careful consideration. Sensitivity analysis confirmed that enhancing ventilation reliability and implementing redundant systems can significantly reduce the likelihood of critical exposure events. These findings reinforce the importance of effective ventilation design, regular maintenance, and integrated safety measures in facilities handling hazardous gases.

 

The proposed framework advances conventional PRA by directly incorporating time-dependent concentration profiles into risk calculations, providing a more realistic foundation for decision-making. Although this study focuses on HF, the methodology can be extended to other highly toxic gases such as Cl₂ and PH₃ and adapted to various piping configurations. The outcomes are expected to inform decisions on system design, safety instrumented system (SIS) implementation, and emergency response planning.

Future research will explore the integration of AI-based surrogate models to accelerate simulations and enable real-time risk monitoring. Overall, this approach aims to improve safety performance and support risk-informed management in industries where handling highly toxic gases is essential.

Presenting Author: Yewon Kim Seoul National University

Presenting Author Biography: The author is a researcher in the Theoretical and Computational Soft Matters laboratory at the Department of Chemical and Biological Engineering, Seoul National University, and concurrently serves as a safety engineer at Samsung Electronics. With over six years of hands-on experience in industrial process safety and chemical reaction hazard management, the author's current research focuses on developing advanced risk assessment methodologies. This dual role enables the author to directly integrate practical, real-world engineering challenges with advanced computational modeling techniques, such as CFD and dynamic event tree analysis, to create more realistic and effective safety solutions.

Authors:

Yewon Kim Seoul National University
Won Bo Lee Seoul National University