Investigation of instability and turbulence effects on gas explosions: experiments and modelling

Safe design of industrial facilities requires models that can predict the consequences of accidental gas explosions in complex geometries with sufficient accuracy. Computa- tional fluid dynamics (CFD) models are widely used for consequence assessment in the process industries. The primary mechanism for flame acceleration during gas explo- sions in congested geometries is the positive feedback between turbulence generated in the reactant mixture, especially in shear and boundary layers from explosion-driven flow past obstacles and walls, and enhanced combustion rates. Additionally, a range of instability phenomena can significantly increase flame acceleration in gas explosions. This doctoral project addresses the following research question: "how can the sub- grid representation of flame acceleration mechanisms due to instability effects and flow past obstructed regions be improved in a CFD tool used for consequence assessment of gas explosions?". The thesis presents and validates new sub-grid models developed for the CFD tool FLACS, focusing on the following flame acceleration mechanisms: (i) the influence of the hydrodynamic and thermal–diffusive instabilities (intrinsic in- stabilities) on flame acceleration in the initial phase of a gas explosion, (ii) the influ- ence of thermal–diffusive effects on the rate of turbulent combustion for different fuels and mixture concentrations, (iii) the role of the Bénard–von Kármán (BVK) instability downstream of bluff-body obstacles in explosion-induced flow, (iv) how the Rayleigh– Taylor (RT) instability developing on a flame front that is accelerated over an obstacle or a vent opening may enhance the combustion rate, and (v) how flexible obstructions with very small components (in the form of vegetation) induce flame acceleration in gas explosions. There was a lack of available experimental work describing the rela- tive importance of several of the aforementioned effects. Therefore, three experimental campaigns were designed and conducted as part of the doctoral study. Experimen- tal findings thus constitute a significant part of the original scientific contribution of the present work; these can be used to develop sub-grid models for any consequence model system. Three additional campaigns, performed by other research groups, were simulated in order to validate the new sub-grid models. The first experimental campaign presented in the thesis concerned explosion exper- iments performed in a large-scale, empty, vented enclosure comprising two chambers separated by a doorway. Simulations indicate that model performance might improve by introducing the effect of the RT instability. Furthermore, the results suggest that the model should account for thermal–diffusive instability effects both in the initial phase of a gas explosion and in the regime of turbulent combustion. The onset and growth rate of intrinsic instabilities in gas explosions are closely linked to the value of the Markstein number of the fuel-air mixture. The second experi- mental campaign presented in the thesis was performed as part of the doctoral study to explore the effect of varying the fuel concentration on overpressures and flame speeds in a series of propane-air explosions. The variations in the fuel concentration effec tively change the Markstein number of the mixture. The dissertation compares exper- imental results with predictions from a version of FLACS that includes a Markstein number-dependent combustion model, developed as part of the doctoral project. For negative Markstein numbers, the new model version performs significantly better than the original model. This work addresses mechanisms (i) and (ii). The thesis elaborates on instability effects relevant for scenarios where explosion- induced flow interacts with partial confinement and obstacles, focusing on BVK and RT instabilities in particular, i.e. mechanisms (iii) and (iv). A third experimental cam- paign was performed as part of the present study to investigate the relative contribution of vortex shedding, caused by the BVK instability, to the overpressure generation in gas explosions with a single obstacle inserted. Vortex shedding was observed in the baseline laboratory-scale experiments, and then suppressed by two passive flow con- trol methods. The most effective configuration reduced the maximum overpressures by approximately 32 %. The results can be used to model the flame surface area in- crease downstream of obstacles in any consequence model system. However, further experimental work, building on the findings of the present doctoral study, is required to formulate a sub-grid model for mechanism (iii). A sub-grid model for flame wrinkling due to the RT instability (mechanism (iv)), based on the linearised growth rate of instabilities on an accelerated flame front, was implemented in FLACS. The thesis presents updated model results for two series of vented explosion experiments, obtained with a version of FLACS that includes both the Markstein number-dependent burning velocity model and the RT instability effect. The first test case involved a series of fuel-lean hydrogen-air explosions. The second test case revisited the campaign performed in the twin-compartment enclosure that was presented in the very beginning of the thesis. A third test case was included to inves- tigate whether the sub-grid models that were initially developed and validated for ex- plosions with a high degree of confinement and idealised obstacle configurations could produce improved results also for explosions in complex geometries with a low degree of confinement. Therefore, a series of large-scale natural gas-air explosions performed in a full-scale offshore module was simulated. All three test cases were performed by other research groups. The sub-grid models developed as part of the present doctoral study significantly improve the representation of several of these experiments. The fourth experimental campaign was motivated by a need for an improved ap- proach to sub-grid modelling of the effect of vegetation on gas explosions in CFD tools, i.e. mechanism (v). Recent accidents have shown that this is highly relevant for risk assessment for onshore process facilities. The presence of foliage on spruce branches notably enhanced maximum overpressures in the experiments, suggesting that obstruc- tions of very small dimensions may contribute significantly to the flame acceleration. Highly flexible, fractal-like obstructions require additional modelling considerations. Based on the experimental results, the effects of foliage and flexibility were included in FLACS simulations by constructing congestion blocks, representing the effective drag area that is expected to produce flame acceleration. This modelling approach success- fully reproduces the experimental trends. In conclusion, the results of the experimental campaigns presented in this thesis have improved the understanding of several important physical effects related to flame acceleration in industrial-scale explosions. Furthermore, the thesis demonstrates how this knowledge may be used to model gas explosions more accurately