Part 1 Design Requirements: A Comparison of Vapor Cloud Explosion Models and the Importance of Properly Assessing Potential Incident Impact Part 2 Analysis of the Buncefield Oil Depot Explosion: Explosion Modeling and Process Safety Perspective

摘要

This paper discusses the various explosion models that are frequently used in support of facility layout and risk assessments and examines the results that each model type will produce for comparative purposes. It is shown that while empirical based models, such as the TNT equivalency and TNO Multi-Energy Method, may be simple in concept, that proper application of the model can be complex. In addition the use of these models provides results that result in a total loss of spatial specificity. Most empirical models assume a uniform hemisphere that changes solely in a radial fashion, the specifics of a plant layout and the impact that localized congestion may play in the development of a flammable cloud and its ignition can be extremely important in the proper design of a facility or in conducting an appropriate hazard review. The difficulty in assigning appropriate levels of congestion and in determining whether appropriate spacing between equipment exists when applying the model, can result in significant under or over prediction of the potential impact. These conceptually simple concepts are often quite difficult to implement and highlight some of the complexities of applying empirical models broadly. A brief overview on modeling is provided in order to highlight the purpose of modeling and the necessity of ensuring that the appropriate model has been constructed or used in assessing the situation. Both recent events, Buncefield, as well as previous events, in which these tools were applied, clearly show that incidents in which a large vapor cloud release followed by a delayed ignition are foreseeable. Even if the specific chain of events is deemed improbable, the end result, namely a vapor cloud release and subsequent ignition, is sufficiently predictable that an appropriate application of the common tools of the industry should be able to adequately capture the incident impact. A review of each of the models is presented along with the assumptions that accompany the model. Specific examples from the RIGOS test series are examined to compare the MEM, TNT, and the CFD tool FLACS in order to illustrate the differences that each model will produce when attempting to assess the same situation. The differences in these results is then discussed from the context that a change in requirements specification and scenario development can have a major impact when a foreseeable incident occurs, but the initial tools used to assess the scenario don’t capture the necessary level of detailed information or the tools are used inappropriately.