(9b) Effects of the Primary Explosion Site (PES) and Bulk Cloud in VCE Prediction: A Comparison with Historical Accident Data

Unconfined vapor cloud explosions (VCEs) can occur when large quantities of flammable mixtures are released in an open area, typically in areas where congestion is present such as chemical process facilities.  These explosions can result in significant damage to buildings, injury to personnel and loss of life.  In these facilities, loss prevention and risk assessment tasks require accurate estimates of potential blast loads associated with VCEs.

Traditional approaches to estimating the damage potential of VCEs, such as the Baker-Strehlow-Tang (BST) methodology or the TNO multi-energy methodology, focus on the contribution of a Primary Explosion Site (PES).  The PES is a region of congestion and/or confinement within the flammable cloud that is typically responsible for the maximum overpressures that are generated.  Most large VCEs, however, span multiple regions of congestion or confinement, all which can contribute to the overall damage to the site.  These contributions are either ignored in simple models, or a detailed analysis of the site, defining multiple PESs and explosion zones is performed.  This type of analysis is highly dependent on the release scenario that is considered as it is extremely sensitive to the shape of the overall vapor cloud.  In addition, realistic estimates should account for the variation in mixture composition within the cloud, which adds further complexity to the problem. To perform VCE studies over many sites, or to consider several different release scenarios, creates a need for an approximate technique to produce quick, reasonable estimates for VCE damage potential without requiring a detailed site analysis. 

In the present study, a simple model for predicting VCE blast loads is described taking into consideration the contribution of a PES as well as the bulk effect of the total cloud (based on averaged properties within the cloud).  This combined model only requires the definition of a single PES and the total mass of fuel released.  The results of this combined model are then compared with blast damage indicators from historical accident investigation studies.

Four historical events are examined: the Flixborough accident of 1974, the Texas City accident of 2005, the Phillips 66 accident of 1989 and the Buncefield incident of 2005.  In addition to the combined model, the individual components of the model, as well as the TNT equivalence method and PES-based methods such as the BST methodology, are compared with the historical data.  For each accident, the model predictions are compared with pressure and impulse damage indicators gathered by site investigations following the accidents.

Throughout the case studies where pressure and impulse damage indicators are available, the combined model produces results that match the observed damage indicators well.  Without considering the total cloud, the PES model under predicts the pressure damage indictors at far field, for pressures typically between 0.07-0.35 bar (1-5 psi), and significantly under predicts the impulse through the cloud.  This emphasizes the importance of accounting for the contribution of the entire vapor cloud to the blast damage. 

It was also found that existing methodologies considering only a single PES, such as the BST methodology, produced, in general, significantly lower overpressure and impulse compared with the damage indicators seen in the accidental explosion.  To properly apply a PES based methodology which does not consider the total cloud, requires the definition of multiple explosion zones to produce results similar to the combined model and the results of the accident investigations.

Since the combined model produces results consistent with the notion of multiple explosion zones but with fewer required inputs, it can be used to quickly produce consistent predictions across multiple different sites.  Also, while the notion of multiple explosion zones may reproduce the near field damage in regions where multiple congested areas are present over a large area, this method would not be able to reproduce the damage caused by large vapor cloud explosions that do not involve large areas of congestion and confinement.  For these scenarios, even a detailed multiple PES analysis of the few areas of congestion would not be able to produce the observed damage contours.