An operating company's experience with computational fluid dynamics explosion and dispersion modeling in ethylene plants

摘要

History has shown that vapour cloud explosions (VCEs) constitute a major risk to ethylene plants. A VCE is caused by an inadvertent release of flammable materials that creates a dispersing cloud in or near the facility and a subsequent ignition of the flammable cloud. In the past VCEs have been analyzed by means of simplified mathematical models, such as the TNT equivalency, TNO multi-energy and Baker-Strehlow methods. Thanks to developments in computer technology, computational fluid dynamics (CFD) VCE modeling has become a tool that is readily available to designers and risk analysts. The beauty of the CFD methodology is that a much more detailed explosion simulation can be achieved as long as values for a limited number of input variables are available. These input variables include facility geometry, dispersing cloud and mixture reactivity data. During the design of the new third ethylene facility for the Joffre site (Alberta, Canada) it became abundantly clear that CFD was the preferred tool for analyzing risks and optimizing designs. For design optimization purposes it is important to realize that structural parameters, defined as a result of this type of simulation, are not fixed quantities but merely point values representing a certain probability distribution. Therefore a significant number of sensitivity studies were performed. In order to obtain robust designs, insensitive to random variations, the occurrence of modeling errors, as well as variations and potential errors in software, were taken into account prior to the detailed project definition stage. A major concern was the dispersing cloud input data file that had been generated with regular firmware. This concern was "semi-validated" through CFD studies on a number of different facilities and by the application of the CFD methodology to cloud dispersion, which yielded surprising results. Moreover in a recent CCPS (2002) publication Hanna and Britter assert that practically all (non-CFD) models currently available for dispersion calculations implicitly assume that the depth of any vapour cloud is considerably greater than the height of nearby buildings and obstacles. For releases near and within an ethylene plant, this is unlikely to be the case. This paper discusses the benefits that can be accrued from an integrated dispersion and explosion CFD simulation effort; e.g., better maximum foreseeable loss data. It also touches upon the obstacles that can get in the way of a successful application.