Atypical major hazards/scenarios and their inclusion in risk analysis and safety assessments

Several of the major incidents that have occurred within Europe in recent years have shown unexpected incident scenarios not considered by their site Seveso II Safety Case (e.g. Toulouse, Buncefield,...). For instance at Buncefield, the scenario of an overflowing petroleum storage tank generated a large flammable cloud and gave rise to a large explosion that would not have been kept by standard assessment techniques. A standard safety assessment of the Buncefield site (before 11 December 2005) would have predicted that the worst credible scenario would have been a large pool fire. We know this missed the actual accident that occurred. This is an example of how improving hazard identification in safety reports has become progressively more difficult to achieve, despite the considerable investment in response to the relevant European Directives and their implementation in member states. One of the reasons is that available hazard identification (HAZID) methodologies, such as Hazard and Operability studies (HAZOP), fault trees, ‘what if’ scenarios etc, take no notice of less likely or unknown events. Unidentified scenarios thus constitute a latent risk, whose manage-ment is nowadays extremely complex and open-ended. For this reason one task of the European research project “INteg-Risk” addresses the general issue of im-proving Hazard Identification, also including specific lessons learned from the Buncefield accident and impli-cations for industry. The purpose of this paper is thus analyzing and describing the progress of work made un-til now to accomplish the task and showing the preliminary results obtained. An in-depth analysis of some of these unusual incidents was performed, in order to outline general features of process plant in which they occurred, their causes, consequences and lessons learned. At first a historical incident analysis was carried out to find similarities among the atypical incidents analyzed and other cases in the past. On the basis of the findings, well-known HAZID methodologies were studied and applied, with the purpose to assist future HAZID processes to identify atypical scenarios. In order to determine all the causes and consequences of a loss of containment, without preconceptions on occurrence probability, bow-tie dia-grams were thus obtained following the methodology for the identification of major accident hazards (MI-MAH), which was developed as part of the ARAMIS project. This allowed the methodology to be reviewed and recommendations developed, which should help to consider warnings of possible atypical scenarios.