Probabilistic explosion analysis

The objective of a probabilistic explosion analysis is to generate realistic overpressures for an area based on probabilistic arguments. This means that one has to look at ventilation, gas leaks and dispersion as well as gas explosions, by:

A) Establishing probable explosion scenarios 
B) Performing explosion simulations 
C) Establishing probability of exceedance curves 
 
These are described below.

A) Establish probable explosion scenarios
The objective of this task is to 

1. establish a reasonable estimate of the largest possible gas cloud (worst case for explosion simulation)
2. generate a distribution of gas cloud sizes with associated probabilities of occurrence

Ventilation conditions may be assessed on the basis of ventilation simulations using FLACS.
All dispersion simulations are based on a table combining a number of different leak rates, ventilation conditions, leak directions and leak locations. Symmetry approximations as well as using the "frozen cloud" concept will be applied to reduce the number of dispersion simulations to 100-150. The exact number will always be case dependent.
The simulations may also include the effect of e.g. blowdown and isolation on the leak profile and hence the dispersion. The time development of each leak is accounted for by each second registering the cloud size as an equivalent stoichiometric cloud, which is categorized in a set of size groups with associated ignition probabilities. A time dependent ignition model will be used.
 
B) Perform explosion simulations
The objective of this task is to predict maximum probable explosion overpressures as well as a distribution of overpressures with associated probabilities of occurrance. A set of cloud size categories will be used for the explosion simulations, varying cloud and ignition location (typically 50-100 simulations).
During the explosion simulations overpressure and impulse will be measured on all walls and decks as well as critical equipment, which upon failure may lead to escalation. Also 2D cross-sectional plots for any variable in the simulation may be produced. In high-velocity zones drag will also be measured.

C) Establish probability of exceedance curves
Each modelled explosion event will have its associated frequency. A set of probability of exceedance curves for overpressure for specific areas or structures (decks, walls) may hence be generated for different leak profiles for each rate (e.g. no mitigation actions taken, initiating blowdown, or effect of ESD on leak profile). If required, surfaces showing probability of exceedance in the (p,I)-plane can also be produced.

There are several ways to utilise overpressure exceedance curves. One way is to base design loads on a probabilistic argument. It is then necessary to have probabilistic acceptance criteria. This could for example be that the probability of a wall or deck suffering unacceptable damage (e.g. penetration or deflection above a specified limit) shall be less than 10-4/year.

Another way is to accept that a probabilistic overpressure analysis gives valuable information, but not define acceptance criteria as such. It is still possible to compare exceedance curves based on different assumptions, e.g. resulting from applying mitigation measures. An example could be:

how different are the curves from

  • the base scenarios
  • scenarios applying mitigation by walls, or
  • mitigation by using general area deluge,

and which solution gives best value for money?


Contact
Lars Rogstadkjernet
Phone no: +47 55 57 43 22
Email: larsr(at)gexcon.com