Comprehensive risk assessment of flammable refrigerants in HVAC systems based on Bayesian networks and consequence modeling

Abstract

Promoting environmentally friendly refrigerants is a global imperative. However, their excellent thermal properties often come with flammability risks, raising concerns about HVAC safety. Existing studies focus on basic flammable characteristics and concentration distribution, neglecting comprehensive risk assessment for refrigerant leakage, diffusion, and fire. This paper fills this gap by constructing a comprehensive risk assessment model. The model integrates Bayesian networks to analyze causes of refrigerant leakage and employs CFD to evaluate distribution and residence time of flammable clouds, ignition sources, and consequences of ignition. Validation using real incident events quantitatively assesses risks associated with leaks, flammable cloud migration, ignition probability, and fire hazard. Critical failure points like improper welding and compressor exhaust are identified. Increasing ventilation speed from 0 to 1 m/s and 2 m/s reduces fire probability from 1.65 × 10−9 to 1.0 × 10−9 and 2.5 × 10−10, respectively. Effective measures to reduce leakage and fire risk are proposed based on accident experience and risk acceptability. The model is applicable not only to R290 chillers but also to other vapor compression cycles, aiding safety measures and emergency response plans for accidents involving flammable refrigerants in HVAC systems.