Interpretation of geometrical effects in consequence modelling. Comparison study between the commercial consequence assessment tools FLACS and PHAST for flammable gas dispersion

Computational fluid dynamics and integral consequence modeling is today a crucial part of safety assessments worldwide. The models are widely used to predict the effects of ventilation patterns, hazardous gas dispersion, flammable potential, toxic exposure or fire and explosion loads. In addition, the consequence assessment is combined with likelihood considerations in order to build-up risk assessment and hence to compare to risk-acceptance criteria. All these aspects are one way or another analyzed and open for optimization when considering consequence or risk based design of industrial facilities. Implicitly, discharge considerations (prior to dispersion) and explosion modeling aspects (subsequent to flammable gas dispersion in case of ignition) are kept in mind when analyzing the comparisons performed since the consequence assessment cannot only be related to gas dispersion and must be seen as a complete threats identification process. Of course, it is recognized that there are significant differences between an integral model and a CFD approach. Simplified integral approaches (like the one included in the DNV PHAST tools but also like the ones promoted in the TNO Yellow Books [9] or in other commercial packages for safety assessment) consider simplifications and empirical assumptions. Mainly, they do not include a straightforward way of dealing with a detailed and complex geometry (influence of vessels, piping, walls etc ) nor they can consider terrain and topography. The integral approaches require less computational time, are easier to handle and are much more cost efficient (considering use within the range of applicability ), thus by all means, less labor intensive but thus also less case specific. On the other hand, CFD codes (like FLACS, CFX, KFX ) are taking into account the effects of complex geometries within the dispersion simulation process. CFD modelling requires significant additional efforts in terms of simulation time, hardware capabilities, skills, man-hours and thus overall cost. The optimal goal of the safety engineer is to use an approach that is suited for the problem to be solved. Also, and by all means, not only accuracy but also conservatism are evaluated within safety engineering. Subsequently, and despite the intrinsic differences listed above, CFD tools (like FLACS) and simplified tools (like PHAST) are observed to serve the same purposes in consequence analysis within the industry. Practically, this study progressively examines and compares the results given by the CFD model FLACS version 10 and the integral model PHAST version 6.6 for a set of various hazardous gas dispersion scenarios. The following aspects have been analyzed; An academic investigation has primarily been performed: ideal steady state free field gaseous jets have been simulated within PHAST and FLACS in a range of conditions where both tools should perform well. Experiments from the large scale Kit Fox trials have subsequently been simulated with both the models for the trial 0604, making it possible to compare the tools with real scale releases. Following this, one has tried to represent / simulate impinging jets and releases into a congested geometry consisting of pipes resembling a ventilated offshore module, moving slightly away from the theoretical range of applicability (but nevertheless sticking to what is observed in the industry). Attempts to represent the geometrical effecs in PHAST have been made. A set of various flow rates and resulting gas clouds have been considered and analyzed. Eventually, releases in a realistic process module have been simulated. Among all the previously described cases, comparison of the centerline concentration decays and resulting flammable potential (as the flammable mass within the gas cloud) have been assessed with varying mass flows and wind conditions. The initial simulations for free field jets in FLACS and PHAST gave overall good agreement between the models in terms of concentration decay along a centerline profile and for flammable mass. The effect of ground has been studied including releases at 10 meters and at 1.5 to 2 meters above ground level. The dispersion model in PHAST (when user defined discharge was used in PHAST) was most comparable with the FLACS results for concentration decay along a centerline profile. The results were within 20% deviation for almost all cases for centerline profile concentration decay. However, simulations run by the FLACS model gave the largest flammable plumes for most of the scenarios run. The comparison effort concludes that the free jets are within a range that may give the same risk assessment for the open free field flammable scenarios. To comfort this initial statement and to move slightly towards some concepts of introducing effects of geometry over dispersion, an experimentation used in the two tools validation framework was also analyzed in order to give more robustness of the comparison and also to get insight of the “validation range” concepts. For the particular studied case, FLACS performed very well in both near and far field, on the conservative side. PHAST was still in relatively good agreement with the experiments while very close to the +/-30 % deviation threshold. Subsequently, the level of congestion in which the gas cloud disperses was slightly increased. Jets simpinging into pipework were run at a relatively low mass flow rate (0.59 kg/s, high likelihood) and at a relative large mass flow rate (9.48 kg/s, lower likelihood). Observations show that small releases are quite dependent of their immediate environment as the impingement may fully control the behavior of the jet (a tunneling and an obstruction effect were seen). The large jet (9.48 kg/s) completely filled the geometry and thus represented a different phenomenon. However, even though the geometrical effect influences the gas cloud shape for the smaller release, the contour of the potential consequences were not that different at a consequence assessment level and neither was the flammable mass. For the largest flow rate interacting strongly with the geometry, three times as high flammable mass was observed in FLACS compared to the PHAST model, indicating different consequences assessments by the models and a large violation of the threshold for the use of the tool outside its main range of application. The thesis is written down as a technical safety scientist would perceive the situation at hand. Advanced details of the chosen models are therefore ruled out of the thesis scope. The focus will be on applicability and the domain where the different models give comparable results. Recommendations on optimized use of the tools available are eventually suggested