It may be helpful to classify fire hazards according to their potential for harm. This may be done by assessing what is in place either in a design or an existing facility and determining the classification. It is preferable to actively manage the fire hazards such that steps are taken to actively lower their classification by reducing the severity of the effects. This may be done by following the principles for inherently safer design (see Section 2.7) or by applying or optimising hazard management measures to minimise the size of releases, their location, effects on the facility, the rate of release and the duration. One method of classification follows.
Catastrophic: As the name suggests, these events would overwhelm an installation and it would be impractical to counteract the effects such that the lives of those on board could be saved. This type of event should be designed out or very high integrity preventative measures provided such that the likelihood is minimised.
Evacuation/Extreme: This type of event would have a major impact upon a large part of the installation such that the effects upon people, both physical and psychological, would be such that evacuation would be necessary. It would also apply to those events where the potential for escalation is widespread including structural, process, safety systems or the impairment of muster and escape routes. Typically these events are those which would give prolonged effects beyond the source module, in particular external flaming and dense smoke effects. In some cases it may be possible to suppress the widespread effects of these fires reducing the categorisation to the lower, controllable, level. If not, the effects must be fully understood and premature catastrophic escalation delayed, and personnel protected from smoke and heat until evacuation has been completed. By their nature these are inherently low frequency events, requiring a significant sized release from a major inventory and/or its combination with safety system failures such as ESD.
Controllable: These events have the potential for local fatalities and may also be capable of escalation to a scale requiring evacuation. However, the moderate scale of the effects should allow these events to be controlled such that further escalation is prevented and evacuation is not essential to preserve life. Typically, the prolonged effects of these events will be limited to one module or process area and will be of finite duration. They would be associated with smaller releases from moderate inventories. In these cases effective control of the source inventory and the prevention of escalation will be critical. It may be practical to extinguish some of these events but in other cases, this may not be possible or may be dangerous in which case, they should burn out under controlled conditions.
Minor: These events are of a very small scale. They may cause local injuries but would not have either the scale or duration to cause critical escalation. They may lead to damage to plant causing financial loss but not major loss of life. These events can be managed by limiting
2.6.3.1 Causes and likelihood of hydrocarbon releases
The causes of hydrocarbon releases are numerous and it is essential that a full causation is carried out so that effective preventative measures can be put in place. These causes can generally be broken down into three categories:
1. human or procedural error;
2. plant or equipment failure;
3. systemic failure; i.e. inherent weaknesses in the business processes and infrastructure supporting design and operation.
Lack of maintenance, particularly over long periods may distort the understanding of the underlying causes of failures. Effective maintenance regimes are essential to determining the likelihood of plant failures.
The likelihood of an event is a function of the propensity of the causes; e.g. the corrosivity of the fluids, the number of times containment is deliberately breached or the number of weak points such as flanges or tappings. It is also a function of the understanding of those causes and the effectiveness of measures which are put in place to manage them.
Statistical data form a good start point from which to list causes and to determine likelihood.
This should then be augmented with the knowledge of engineers, technicians and operators to give a more accurate picture for each facility. HAZOP procedures will give a rigorous identification of process causes but the overall examination should be sufficiently broad to address external and human effects. This examination should be fully documented so that there can be assurances that preventative measures are suitable and sufficient.
For statistical data, the most frequent sources of the hazard as given by the history of releases experienced to date are documented as follows.
The HSE document OTO 2001 055 [2.5] states that for the UK sector of the North Sea:
61 % of all releases are from pipework systems 11 % of all releases are from small bore piping 15 % of all releases are from flanges
14 % of all releases are from seals and packing Of the causes;
11 % are due to incorrect installation
26 % from degradation of materials (excluding corrosion and erosion) 11 % of all releases are due to vibration/fatigue
19 % of all releases are due to corrosion and erosion
Avoidance of potential leak sources in design therefore needs to consider these above issues in particular. The importance of operational aspects is also shown in proportion of leaks attributable to poor inspection and monitoring.
Sources of release data include WOAD [2.6], OREDA [2.7] release statistics published annually by the HSE and UKOOA [2.8, 2.9, 2.10 and 2.11]. The Minerals Management Service (MMS) of the US also publishes data on incidents on the Gulf of Mexico.
2.6.3.2 Ignition causes and probability
The probability of ignition will depend upon the following factors.
• The rate and duration of the release and the size of the consequent gas cloud
• The location of the release
• The type of fuel and the proportion of gas or volatile vapours which is generated in the short term
• The nature of the release; whether high or lower pressure. The turbulence caused by high pressure gas releases will cause effective mixing with the air to give a well defined flammable cloud. High pressure liquid releases will encourage fine droplet formation and increase the vaporisation of any light ends.
• The flammability characteristics of the gases and vapours. Each different gas or vapour has a specific flammability range, from a lower flammability limit; through stoichiometric and rising to a higher limit above which ignition should not occur. Very large releases may have a non flammable rich core but will be surrounded by a flammable region which may engulf ignition sources as it spreads away from the point of release. Each gas or vapour will also have a specific auto ignition temperature ranging from 200 - 550 °C. such that contact with hot surfaces such as an exhaust turbocharger would cause ignition
• The dispersion characteristics; whether there are heavy vapours which will descend or lighter gases which should rise.
• The confinement of the escaping vapours and gases by floors, ceilings or walls. These may also cause flammable gases to be directed towards areas without flameproof equipment
• The ventilation characteristics in the areas, whether forced or natural and the variation of those characteristics with wind strength and direction
• The characteristics of the fluid and its release, where this might build up static
• The presence of sulphurous impurities in the fluids which might lead to the formation of pyrophoric scale
• The number of fixed ignition sources and the standard of their maintenance, if designed for use in flammable atmospheres, including the presence or not of Ex equipment.
• The proximity of the release to areas which are classified as “safe” and therefore are
• The gas detection philosophy and the local and wider shutdown of ignition sources upon detection
• The detection of gas ingress at the air intakes to enclosures such as accommodation or equipment rooms and the closure of dampers.
• The hot work philosophy on the facility, the number of these activities and the effectiveness of their control.
• The possibility of ignition being caused by the action of personnel carrying out emergency response actions such as plant shutdown causing sparks at electrical breakers.
Ignition probabilities have been widely studied and this work is summarised in recent work for UKOOA studying ignition probabilities [2.12]. The probability of ignition should be determined using that guidance together with an assessment of the characteristics listed above. As with the likelihood of release, it is possible to influence the probability of ignition by design, good maintenance and operational controls.
2.6.4 Likelihood
The likelihood of an explosion will depend upon the likelihood of occurrence of a gas cloud and delayed ignition. The following parameters will influence the potential likelihood of an explosion:
• hazardous inventory complexity, i.e. the number of flanges, valves, compressors and other potential gas leak sources;
• the type of flanges, valves or pipework, some generic types of flange tend to have lower leak frequencies associated with them, e.g. hub type flanges;
• the number of ignition sources within the potential gas cloud;
• the ventilation regime;
• the equipment reliability and the maintenance philosophy.
The likelihood considerations tend to align closely with the consequence factors in that the low consequence installations will tend to be small and therefore less complex. Large installations will have more potential leak and ignition sources and therefore a greater requirement for intervention and maintenance.
Low event likelihood installations and compartments will have a low equipment count. The frequency intervention of 6 weeks or more is also recommended as a criterion as this will be a surrogate for equipment count and reliability as well as a measure of maintenance risk with respect to explosion.
Medium event likelihood is suggested by an NUI with equipment count greater than for the
‘low’ case. Similarly, where the planned frequency of maintenance/intervention is greater than a 6-weekly basis then this suggests a higher or less reliable class of equipment with medium
Where there is doubt regarding the category into which an installation should fall, it is recommended the category with next higher likelihood is used.