RÉSUMÉ

Flares are most common in the oil industry. To access the safety threats of such flares is necessary to predict the thermal radiation flux that would he received around the flame. Empirically based prediction techniques are the current standard methods for such evaluations, hut their validity is circumscribed. Computational fluid dynamics (CFD), instead, is based on first principles and should therefore be able to solve a more general class of problems. Many of the submodels required to provide closure to the descriptive equations are empirically based, particularly those related to turbulence. Besides, although each thermal radiation model may be based on fundamental principles, many of the fluid properties that determine the incident radiation at a surface are not. For these reasons a validation activity is necessary in order to explore the limits and potentialities of such an approach. The experimental data used for the validation have been obtained from a wind-tunnel study [1, 2] of a laboratory-scale flare. Temperatures and radiation intensity were measured. The validation was performed with two CFD codes: Fluent version 6.3, a commercial code, and Kfx version 2007, a code developed in a joint industrial project where Eni is involved. Many different approaches have been tested with the two codes, trying both to have a correct comparison between the two (using the same models, same number of grids, etc.) and to explore their capacity.