26th November 2020
Abercus presents at Hazards 30 on the 10-4 /yr criterion for blast overpressure - an alternative comparative approach for safer design
The probabilistic explosion approach outlined in NORSOK Z-013 is now widely used in the oil and gas sector for determining blast loads for design. This process involves three steps:
1. CFD simulations - use computational fluid dynamics (CFD) to simulate a large number of deterministic gas dispersion and explosion consequences to form a database of representative scenarios
for pre and post (delayed) ignition behaviour following a loss of containment of flammable material.
2. Probabilistic analysis - consider the probabilities of release and ignition for each gas dispersion/explosion scenario within the database so that exceedance data can be derived for the explosion blast loads.
3. Determine the design blast loads - from the exceedance data, retrieve the blast load corresponding to the acceptability criterion – typically a frequency of 10-4 /yr.
Abercus is aware of several concerns relating to this probabilistic approach from industry and shares these concerns. This paper focusses on steps 2 and 3 – on how the CFD predictions are interpreted and used for deriving blast loads for design.
Specifically, when exceedance data is presented, it is usually only a single set of exceedance data that is provided. The sensitivity of the exceedance data to the uncertainties in the assumptions used to derive it are usually not presented,
or may not even have been considered at all. Abercus has previously argued that these sensitivities should be transparent and properly understood.
Two important uncertainties, for example, are the release frequency data assumed for the gas dispersion analysis and the ignition frequency model assumed for the explosion analysis. Indeed, Abercus has recently reviewed a number of probabilistic studies
where depending upon the assumptions for these frequencies, both of which have come from supposedly credible sources, the design blast load may vary from several barg to zero (because the relevant exceedance curve entirely falls below the 10-4 /yr criterion).
With this level of sensitivity to the underlying assumptions, the current probabilistic approach is simply not credible. To address this, Abercus proposes a shift away from using absolute acceptance criteria (such as the 10-4 /yr criterion),
in favour of a relative criterion. It is proposed that for any particular facility, the probabilistic approach should consider two separate models:
1. the actual model of the facility that already exists or as will be built
2. a model of a notional facility that does not and will not exist, but is designed according to certain criteria representative of its age and type.
The notional model will be used to derive blast loads for a representative facility and then, a percentage improvement upon this in terms of reduced blast impact (not just load but also structural response) will be required for the actual model/design,
which is more in line with the ALARP approach. The benefit of this relative approach is that any uncertainties associated with the input assumptions, such as release frequencies and ignition model, will be inherent in both models,
so that when they are compared to check whether the proposed design is OK, these errors will essentially cancel out. So long as industry continues to pursue an absolute approach, these errors will continue to have a significant impact
on the exceedance data and the approach will continue to lose credibility.
There is an existing precedent for this relative approach - the National Calculation Method (NCM) that is used for compliance for energy performance of buildings in the UK.
Download the full presentation here