Explosion Hazard Assessment
Accidental releases of flammable liquids or gases often result in the formation of a cloud of vapour that is dense relative to ambient conditions. If the cloud encounters an ignition source then a vapour cloud fire (VCF) may result. In the present context, VCF is taken to mean either a flash fire or a fireball. VCF's are important for two reasons:
- An intrinsic hazard, in the form of thermal radiation, assuming no or limited confinement/congestion, so that overpressures are not important.
- The possibility of escalation. It is highly likely that secondary fires may be started as a result of the flash fire / fireball and, there is a high probability that following a VCF there will be a steady fire, typically either a pool fire or jet fire (or a combination of the two).
- To identify areas of uncertainty in the characterisation of flash fires and fireballs;
- Identify where the fire hazard is significant in relation to other hydrocarbon hazards;
- Initiate research to increase knowledge and understanding in ill-defined areas of flash fire and fireball evaluation; and
- Promote the use of a consistent methodology for evaluation of fire hazards.
Current knowledge of hazards
An overview of the incidents, experimental data and the methods for estimating the characteristics of vapour is given in the 'Guidelines for Evaluating the Characteristics of Vapour Cloud Explosions, Flash Fires and BLEVE's' published (1994) by the Centre for Chemical Process Safety. Since 1994, HSE has been involved in two experimental projects relating to fireballs and flash fires.
The resulting size and shape of the fireball following the BLEVE failure of a vessel was dependent on the amount of fuel in the vessel and the mode of failure.
The resulting external radiation field and hence received dosage are dependent on fuel mass, wind speed and direction.
The duration of the fireball was seen to be dependent on the mass of fuel involved.
Surface emissive power is highest for the smallest release, because a smaller mass is superheated such that, it flashes to vapour most rapidly, producing a highly radiative flame.
The resultant fireballs gave their maximum power output before the fireballs reached their maximum volume and close to the lift off time.
Additional experimental work on flash fires was performed as part of a Joint Industry Project (CERC, 2001). Butler and Royle (2001) characterised the flash fires from turbulent, two-phase jet releases of propane (up to 4.9 kg s-1).
The presence of obstructions in the path of the vapour cloud was found to alter the concentration of LPG vapour in the cloud dramatically with, in this case, significant decreases in the vapour concentration downwind of the fence. The concentration of gas in the vapour clouds formed was generally low and the vapour cloud fires produced were relatively lean. The flames were therefore often invisible. Ignition of the cloud was observed at concentrations below the Lower Flammability Limit (LFL) of 2.2 vol.%. This is thought to be due to localised pockets of high concentration of gas at locations where the average concentration is measured as being below the LFL. In some cases, the cloud was ignited, but the flame did not propagate throughout the cloud, resulting in the formation of isolated pockets of ignition. In no cases were fireballs observed.
The characteristics of fireballs (diameter, height, lift off, duration) are usually modelled using empirical formula based on the mass of fuel released. The far field thermal radiation is usually estimated by a:
- Point source model, where it is assumed that a certain fraction (usually between 0.25 and 0.4) of the heat of combustion is radiated in all directions; or'
- a solid-flame model where the radiation received is calculated from the surface emissive power of the flames, the relative geometry of the target and fireball and the atmospheric attenuation.
Both types of modelling have their disadvantages. A point source model tends to overestimate the irradiance at distances below 5 fireball diameters and, for a solid flame model, the result obtained is very dependent on how the surface emissive power is defined and measured.
Vapour cloud fires
Vapour cloud fire models were review by Rew et al. (1995, 1996). The simplest form of vapour cloud model uses a gas dispersion model to define the flammable region and assumes that anyone in the flammable region will be killed. As part of the vapour cloud fire model (CERC, 2001), three models were analysed:
- Raj & Emmons (CCPS, 1994);
- CLOUDF (Cracknell and Carsley, 1997); and
- HSE Flash Fire, HSEFF (WS Atkins, 2000).
Areas of uncertainty
The conclusions from the CERC (2001) model assessment exercise were that, the application of models was limited to low momentum sources, there was little or no validation, and there were areas of disagreement in calculation of flame height and flame speed.
Industry practice in assessment of hazards
Vapour cloud fires are generally not considered off-shore as part of the safety assessment, unless the possibility of developing into a vapour cloud explosion exists.
Strategy development issues
- To develop simple models to predict the occurrence and effect of vapour cloud fires.
- To develop a greater understanding of the effect of flash fires on personnel.
- To increase awareness of where the flash fire encroaches on evacuation routes.
- To more fully define fireball hazards in close proximity of process equipment.
- To develop a methodology to identify scenarios where a flash fire develops into a vapour cloud explosion.
- To develop an understanding of the current capabilities of models to address fireballs and flash fires and to predict the consequences of each.