Natural and forced ventilation of buoyant gas released in a full-scale garage: Comparison of model predictions and experimental data

An increase in the number of hydrogen-fueled applications in the marketplace will require a better understanding of the potential for fires and explosion associated with the unintended release of hydrogen within a structure. Predicting the temporally evolving hydrogen concentration in a structure, with unknown release rates, leak sizes and leak locations is a challenging task. A simple analytical model was developed to predict the natural and forced mixing and dispersion of a buoyant gas released in a partially enclosed compartment with vents at multiple levels. The model is based on determining the instantaneous compartment over-pressure that drives the flow through the vents and assumes that the helium released under the automobile mixes fully with the surrounding air. Model predictions were compared with data from a series of experiments conducted to measure the volume fraction of a buoyant gas (at 8 different locations) released under an automobile placed in the center of a full-scale garage (6.8 m × 5.4 m × 2.4 m). Helium was used as a surrogate gas, for safety concerns. The rate of helium released under an automobile was scaled to represent 5 kg of hydrogen released over 4 h. CFD simulations were also performed to confirm the observed physical phenomena. Analytical model predictions for helium volume fraction compared favorably with measured experimental data for natural and forced ventilation. Parametric studies are presented to understand the effect of release rates, vent size and location on the predicted volume fraction in the garage. Results demonstrate the applicability of the model to effectively and rapidly reduce the flammable concentration of hydrogen in a compartment through forced ventilation.

► Analytical model for predicting hydrogen concentration in a partially enclosed compartment.
► Comparison with experimental data for full-scale garages with and without an automobile.
► Effect of model parameters on predicted volume fraction.