CFD-based simulation of flammable gas dispersion in a complex geometry

Abstract

Quantification of time and spatial concentration profiles of a gas-air cloud caused by accidental leakage of flammable gas from damaged equipment can be of great importance for the correct prediction of the occurrence of dangerous situations. Knowing the extent of leakage and time dependence of generation of hazardous concentrations may be a good tool in the design of mitigation equipment (e.g., gas detection alarms) or in determining further measures (e.g., the extent and method of ventilation of the premises).

The paper describes the dispersion of leaked methane (natural gas) from a damaged low-pressure gas pipeline, in a confined space with a geometrically complex environment and the formation of local explosive concentrations at the lower explosion limit (LEL) of the methane-air mixture. The measurements were carried out in a middle-scale model of a building with a volume of approximately 2.7 m³. A new type of sensing system has been developed for the detection of individual concentrations, allowing the required concentrations to be recorded continuously and with greater accuracy.

The velocity of propagation of leaked gas, local explosive concentrations of a gas-air mixture, and also duration of formations of these local concentrations, could be determined using successful verification of a CFD simulation with experiments. A comparison of experimental data with simulation data demonstrated that CFD technology can be an effective aid to describe the process of flammable gas dispersion and can also predict the tendency of gas dispersion and gas distribution. The key to a successful application of CFD in dispersion simulation lies in the accuracy with which the effect of turbulence is generated, due to the complex geometry of the monitored space. Within the simulations of turbulent flow, the accordance of experimentally determined values with the values calculated in dependence on the mesh density and used turbulence model was observed. RANS equations with six turbulent models were used (Laminar; k-ε Standard; k-ε RNG; k-ε Realizable; k-ω Standard and k-ω SST). The best match of the numerical simulation results with the performed experiments was achieved using the mathematical model of turbulence k-ε RNG.