LNG - Leakage, spreading and fire

Abstract

This master s thesis continues the line of LNG related theses in cooperation with ComputIT AS. The Phoenix series large-scale LNG pool fire experiments have been simulated using the CFD software Kameleon FireEx (KFX) for validation purposes of the software to experimental data.

A number of simulations were performed, and the main investigations of the experiment thermal radiation, pool spreading and flame morphology have been compared to results from KFX, using a range of different simulation scenarios. Scenarios included simulating the LNG as either a liquid release or as a gaseous release of methane from a circular equivalent of the quasi-steady LNG pool area obtained on the experiment. Other variations included transient or constant release, where the transient release was based on an approximated flow rate from the experiment, and the constant release was based on the average flow rate over the duration of the experiment. Due to limitations in the pool model in KFX, the liquid releases were simulated on a flat ground instead of on a pool of water, with various heat transfer coefficients to account for the heat transfer between the water and the LNG.

The radiative heat fluxes from the simulations compared to experimental values were initially low by factors between two and four, regardless of simulation scenario. The soot model was investigated, and found to give an excess of relatively cold soot outside of the flame, acting as a radiation screen, thus reducing the thermal radiation from the fire to the surroundings. The soot model was modified, and subsequent simulations produced very comparable values for the radiative heat fluxes for the majority of simulation scenarios.

For all liquid, transient simulations, the LNG pool spread much faster than in the experiment, and only the simulations with adiabatic conditions between the LNG and ground reached the same maximum area as the experiment. All simulations with non-adiabatic conditions had a maximum pool area below that of the experiment. As a result, it was concluded that the LNG vaporized too quickly, and consequently the duration of the fire was about 250 s shorter in the simulations than in the experiment. As a consequence of the high vaporization rate in the simulations, the flame grew to heights far above the experiment, and at some instances the flame in the simulations exceeded the flame height in the experiment of over 100 m.

A new, simple model for LNG spreading through water was introduced, and initial computations in MATLAB produced highly promising results.