Until now, the discussion on the QRA methodology has been very generic. The following dis- cussion is intended to provide more insight into the actual QRA methodology.
An electronic database of approximately 1200 materials is available to the PHAST / PHAST RISK software, with the material properties regularly reviewed.
The PHAST RISK risk modelling software requires the following inputs to be able to pro- duce risk results:
• Import an electronic map of the study area, on which individual fatality risk contour re- sults may be produced
• The electronic map may be programmed in PHAST RISK to:
− Superimpose all on-site and off-site populations within the study area by location, and specifying the day / night number of people for each location
− Superimpose all potential ignition sources within the study area, which may cause de- layed ignition of a flammable release
− Delayed ignition sources may be specified as point sources (e.g. flares and fired heaters etc.), area sources (e.g. parking area etc.) or line sources (e.g. traffic roads etc.). Each ignition source carries additional specification in terms of presence factor and ignition source strength (probability of ignition per unit time, when in contact with a flammable vapour cloud between LFL and UFL. The actual delayed ignition prob- ability of any release is calculated by PHAST RISK, based on the dispersion model- ling results and event duration
− The immediate ignition probability associated with each failure case of flammable fluid is a risk analyst programmed value, based on historical ignition data, which varies with leak size and release phase (gas / liquid / 2-phase) (the larger the leak vapour flow rate, the higher the ignition probability, typically varying from 1% to 30%, unless above auto ignition, then 100%)
• Prepare and import weather class, wind speed and wind direction probability data for the study area. For day and night due to lack of data the same weather class, wind speed and wind directional probability are used in PHAST RISK
• Enter all identified failure cases, which are defined in terms of: Location, Material re- leased, Quantity released (or release duration), Temperature, Pressure, Leak size, Leak direction (e.g. horizontal, vertical), Leak elevation, Leak frequency and Immediate igni- tion probability
• Each failure case calculation in PHAST RISK starts with discharge modelling. Based on release duration and release phase (gas, liquid, 2-phase), PHAST RISK directs the dis- persion and consequence calculations to one of 4 alternate, built-in consequence out- come event trees (continuous vapour release, continuous release with rain-out, instanta- neous vapour release, instantaneous release with rain-out), where each event tree branch probability carries default values, which may be re-programmed by the risk ana- lyst.
• PHAST RISK will then calculate all alternate consequence outcomes (e.g. jet fire, explo- sion) of the event tree selected, in terms of hazard range and event duration (where ap- plicable), for each weather class / wind speed combination
• So far the calculations performed in PHAST RISK only relate to the alternate conse- quence outcomes and the consequence hazard ranges, for each specified failure case. To produce risk results, PHAST RISK will perform impact frequency calculations, using
the failure case specified leak frequency as starting point. Frequency aspects of the risk calculations relate to the:
− Defined failure case leak frequency
− Weather class, wind speed and wind directional probability, for each of the 8 wind directions
− Specified immediate ignition probability and PHAST RISK calculated delayed igni- tion probability. The delayed ignition probability calculation is based on the strength and location of all specified ignition sources and the failure case dispersion hazard range, combined with vapour cloud persistence (duration).
− Selected event tree branch probabilities, for each alternate consequence out come. − Impact probability for each alternate consequence outcome. This is based on the
PHAST RISK calculated magnitude of each consequence and the PHAST RISK de- fault impact probability criteria or risk analyst specified impact criteria for that type of consequence.
− Location and number of people (or equipment) within hazard area for societal risk results, with separate calculations for day and night, indoors and outdoors.
• PHAST RISK performs its individual and societal risk calculations based on a 200 x 200 grid (40,000 points), with the grid point spacing automatically varied, based on the con- sequence hazard range results.
• For each release case, PHAST RISK takes the failure case release frequency as initial input, multiplies this by the first weather class / wind speed probability, for the first of 8 wind directions.
• PHAST RISK takes this result and multiplies it by the immediate ignition probability, while also separately multiplying this result by the PHAST RISK calculated delayed igni- tion probability.
• These 2 results are multiplied by the first of the event tree consequence branch prob- abilities, relating to immediate or delayed ignition branch path.
• PHAST RISK takes the PHAST calculated consequence hazard range and verifies which grid points are within the consequence hazard area. For each grid point within range PHAST RISK then calculates the magnitude of the consequence at each grid point (e.g. explosion overpressure at a particular grid point may be 3 psi (200 mbar)).
• The calculated consequence magnitude at each grid point is then compared to the PHAST RISK programmed impact criteria level, and the likelihood of fatality or damage calculated, based on the impact probability criteria specified in PHAST RISK, for the type of consequence and the magnitude of the consequence.
• This calculation is repeated for each event tree alternate consequence outcome at each grid point, for that weather class / wind speed and wind direction, and the result added to the previous risk level, at each grid point.
• The above calculations are then repeated for each of the 8 wind directions, cumulatively adding to the risk level at each grid point.
• The above calculations are repeated for all day / night weather classes, wind speeds and wind directions, cumulatively adding these risk results at each grid point.
• Once all risk calculations at these grid points have been completed for the first failure case, the next failure case will be calculated, again adding all results cumulatively at each grid point. This is repeated until all failure cases have been calculated, while
PHAST RISK also tracks the risk contribution made by each failure case at each grid point.
• Once completed, PHAST RISK produces individual risk contour results by linking points of equal risk, based on the pre-specified individual fatality risk (or equipment damage) criteria levels, and using linear interpolation between relevant grid points. The risk con- tour results are super imposed on the electronic site map, entered in the PHAST RISK software.
• PHAST RISK can also produce societal risk results by comparing the calculated level of individual risk at all 40,000 grid points, and combining this with the number of people in- doors and outdoors, entered by the risk analyst by location.
The above discussion demonstrates that the meteorological data, ignition data and population data entered into the PHAST RISK software are critical to the risk results.
Note that by default the risk modelling within PHAST RISK aims to produce offsite fatality risk results. This is achieved by the build-in but programmable parameter settings, which include:
• Indoor & outdoor people fatality impact criteria levels, for each alternate consequence outcome. For flammable releases the alternate consequences would be spill fires, fire balls, jet fires, flash fires and vapour cloud explosions (VCEs), each with pre-defined values for the impact levels that will affect people. For fire ball exposure this is based on the Eisenberg Probit equation. For spill fires and jet fires a single criterion flammable dose of 250 KJ is used, with the radiation impact level set at 9.8 kW/m2, corresponding to 1 % lethality in 20 seconds. For flash fires the 0.5 LFL envelope is used and for VCE overpressure two impact criteria levels are used, 1.5 psi (0.1 barg) and 5 psi (0.34 barg). • 4 built-in event trees (Continuous No Rain Out; Continuous With Rain Out; Instantane-
ous No Rain Out; Instantaneous With Rain Out) that are automatically selected based on the type of material and the release conditions. Each event-tree assigns a ‘split’ be- tween alternate consequence outcomes (spill fires, fire balls, jet fires, flash fires, VCEs and no hazard), based on the immediate ignition, delayed ignition and no ignition prob- abilities.
• People vulnerability criteria, which pre-determines the fraction of fatalities resulting in- door & outdoor from being exposed to specific consequence outcomes for a specified duration, or to one or more specified criteria levels. Most of these values go back to data provided in the Dutch "Purple Book" [5].
Figure 28: SAFETI Default Vulnerability Parameters
By default PHAST RISK uses (programmable) blast overpressure levels of 1.5 psi (110 mbar) and 5 psi (approx. 345 mbar) for assessing indoor and outdoor offsite fatality impact. These overpressure levels correspond to light damage and total destruction of ordinary brick (resi- dential) housing. The default blast criteria levels need to be modified for assessing the impact of overpressure on prefabs, brick and concrete buildings. In particular, it is known:
• Wood or corrugated asbestos panels shatter at 1 – 2 psi (0.07 – 0.14 barg) overpres- sure,
• Partial collapse brick housing at 2 psi (0.14 barg) overpressure, • Concrete walls (non-reinforced) shatter at 2 – 3 psi (0.14 – 0.21 barg), • Complete destruction of houses at 5 – 7 psi (0.34 – 0.48 barg),
• Blast proof concrete control buildings fail at 10 psi (0.68 barg) overpressure.
Consistent with these data, Linde (expert knowledge from DNV) set the PHAST RISK explo- sion parameters to achieve the following overpressure risk results:
• 1.5 psi (0.1 barg) - Prefabs severely damaged by overpressure, • 3 psi (0.2 barg) – Brick buildings severely damaged by overpressure,
• 5 psi (0.34 barg) – Reinforced concrete buildings (non blast proof) severely damaged by overpressure.
The default PHAST RISK hydrocarbon vapour cloud explosion efficiency is set to 10 percent, consistent with the objective to produce conservative offsite fatality risk estimates. This is an overestimation compared to historical data. The Dutch Government Coloured Book Risk As- sessment Guidelines reports historical Hydrocarbon Explosion Efficiencies ranging from 2 per- cent to 5 percent. To be less conservative than the PHAST RISK defaults, DNV has used the upper estimate of 5 percent explosion efficiency in the PHAST RISK Explosion Parameters.
Early ignition vapour cloud explosions are modelled in PHAST RISK at the centre of release. For delayed ignition the PHAST RISK explosion modelling results are (conservatively) based on locating the centre of the explosion at the maximum (theoretical) displacement distance.
For delayed ignition the PHAST RISK explosion modelling takes into account the defined igni- tion sources. PHAST RISK will first perform discharge calculations and dispersion modelling. Where a dispersing vapour cloud contacts one or more of the defined ignition sources, PHAST RISK will then calculate the time varying ignition probability at that time. The explosion mass is calculated based on the mass of vapour between UFL and LFL at that time.
PHAST RISK allows to choose the location for the epicentre of a delayed explosion, where the location is defined in relation to the dimensions of the cloud. If Cloud Front (LFL Fraction) is chosen, the epicentre will be located at the furthest location downwind at which the concentra- tion is equal to the LFL fraction to finish, set in PHAST RISK's Flammable Parameters. If one chooses Cloud Centroid, the epicentre will be located at the centre of the cloud for an instanta- neous release, and at the cloud centroid for a continuous release, where the downwind (x) lo- cation of the centroid is obtained by taking a weighted average of the centre point of each re- lease segment. If one chooses Cloud Front (LFL), the epicentre will be located at the furthest location downwind at which the concentration is equal to the LFL.
All three options are very conservative when compared to historical explosion incidents, where explosion damage is most often reported around the area of the release. If PHAST RISK de- fault options are used, the explosion centres would (most often) be modelled as located out- side the process units in open areas, which are very unlikely to give rise to explosions. To ob- tain more accurate and realistic explosion modelling results for this QRA, ignition location was set to the cloud centroid.