Buncefield: Where are we now?

Author : Kehinde Shaba, DNV GL

15 May 2014

The 78th FABIG technical meeting titled “Dispersion and Explosion Characteristics of Large Vapour Clouds (Buncefield Explosion Mechanism JIP – Phase 2)” took place in London on the 4th of March 2014. The core aim of the session was to share the key learnings of the second Phase of the Joint Industry Project (JIP).

The first phase had concluded (amongst other things) that the most likely explanation was that a deflagration occurred (driven by the vegetation in the area which provided the requisite congestion) and this subsequently transitioned to a detonation. A key goal of Phase 2 was to confirm the feasibility of this hypothesis of the deflagration to detonation transition (DDT). Seven presentations in total were delivered. Some takeouts from the session are as follows.

Rethinking barriers as linked rather than independent

To understand this event properly, one must look at the event itself as well as the circumstances that led to it. Viewing the accident from an orthodox respective, it is clear that several barriers failed for the accident to occur. Traditionally, such barriers are thought to be independent, but new thinking on the matter indicates that such an outlook can be optimistic and as the barriers tend to be governed by the same system (e.g. the plant safety management system), a failure of one barrier most often means that the others will be in a degraded state. Basically, this means more attention needs to be paid to any interrelationships that exist between seemingly independent barriers.

Are detonations a feasible outcome?

With regards to the event itself, full scale experiments conducted as part of the phase 2 JIP have demonstrated that a DDT is possible with vegetation providing the necessary congestion. Flame speeds of up to 2000m/s were recorded as well as overpressures in excess of 20 bar. Particularly striking is that a standard flammable mixture of propane and air was used in the experiments; such high overpressures are traditionally associated with much more highly reactive materials such as acetylene and hydrogen. The final phase 2 report will include recommendations on the optimum vegetation density to mitigate the potential for congestion.

Calm conditions yielding large vapour clouds?

The research has also shown that large vapour clouds can form as a result of an overfilling incident in calm conditions (i.e. no wind). A technical note (FABIG Technical Note 12) describing an assessment method for calculating the rate at which the volume of a vapour cloud increases during an overfilling incident has also been published. The inability to model calm conditions is a key weakness in many empirical dispersion models (both simple and CFD based). More research is needed in this area.

How good are our current predictive models?

Comparisons of the experimental data for the Buncefield scenario with predictions from both advanced (i.e. CFD based) models and simple models e.g. the multi-energy and TNT models show a very mixed picture – good in some cases, significantly disparate in others. The reasons for the lack of agreement are legion, but factors relating to the experimental set up and inherent model deficiencies are key considerations. The lack of agreement suggests that some work is required to improve the predictive capability of these tools for this particular scenario.
A noteworthy factor is the flexible nature of elements of the vegetation (e.g. branches) which means they will probably streamline with the blast wave and create less drag. How to best capture this influence in predictive models is yet to be resolved.

Summary

Whilst many more questions remain, the work done by the JIP has helped give a measure of clarity to many of the uncertainties surrounding the event. Kudos to all the organisations and individuals involved – the work that has been done is definitely world class and the knowledge generated will go a long way to preventing a similar accident. The global Oil and Gas/Process industry is thankful and hugely indebted to the JIP sponsors for undertaking this initiative and more importantly for sharing the findings publicly.

Thanks to Kehinde Shaba of DNV GL for this update.

The final report of the Dispersion and Explosion Characteristics of Large Vapour Clouds JIP is now available on the Fire and Blast Information Group (FABIG) website at www.fabig.com