(70ad) Exploring Dependencies of Ignition Locations on Probabilistic Explosion Analysis

CFD based Probabilistic Explosion Analysis has become a common method for estimating structural design accidental load (DAL) specifications for offshore capital projects. This approach has been shown to yield more reasonable results than traditional analysis i.e. solely consequence based which generally result in highly conservative values.  As expected, results from probabilistic based approach may have dependencies based on chosen methodologies. Previous research and studies have generally focused on exploring methodology changes which are considered to result in significant differences i.e. transient vs. non-transient dispersion modeling, different levels of 3-D model detail, etc.  The purpose of this paper is to explore the dependency that can come from changes to ignition locations (often viewed as small adjustment) within a probabilistic explosion analysis. Commonly a limited number of ignition points are chosen for each simulated cloud location/size, per NORSOK Standard Z-013 [2] “Ignition can in principle take place anywhere in the gas cloud. Explosion simulations shall include at least three different ignitions: central, edge and other.”  This paper will be broadly divided in two parts where the first part will explore the use of synthesized ignition locations based on data from  CFD simulations conducted using a fewer number of ignition points. This will be compared to actual simulation data from a larger set of ignition locations for establishing convergence.  Having established criteria from the first part, the second part of the paper will explore the use of different ignition probability distribution functions (equal, linear, Gaussian, etc.) and compare the exceedance results on specified targets.  It is the intent that by understanding how the locations of ignition sources affect the sensitivities of the forthcoming QRA work performed to internally or externally demonstrate ALARP for most facility design.