Event Tree Analysis

Having identified a set of top events and estimated their initial frequencies as discussed at the previous section, it is necessary to consider the range of possible consequences that could occur after the original failure, and to estimate the probabilities of alternative outcomes. This can be done with using an "event tree analysis (ETA)". An event tree is a logic diagram in which all of possible outcomes of a single initiating event are described.

Instantaneous release

Immediate Formation Pool ignition of pool ignition

Initiating event

Cloud Delayed denser than ignition

Explosion Incident or fire outcomes

Explosion Fireball Pool Fire VCE

Flash Fire Harmless

VCE Flash Fire Harmless Harmless

Jet Fire Pool Fire VCE Flash Fire Harmless VCE Flash Fire Harmless Harmless

Fig. 4.5Event tree diagram of LH2 releases [17, 2] 4.4.3.1 Event Trees Diagram for Hydrogen Release

A hydrogen release may have many event outcomes, depending on the timing and type of ignition. For example, a released substance may be ignited immediately at the point of release, or it may be ignited after the cloud has been dispersing for a certain amount at time, or it may not ignite at all. Fig. 4.5 and 4.6 show event tree diagrams to develop incident outcomes from hydrogen releases (in liquid and gaseous forms). For a given release of hydrogen an early explosion, fireball, or jet fire outcomes will be occurred if it is followed by immediate ignition. Otherwise, a pool (pool fire may occurs if it ignites) may be formed or the substance may be evaporated and disperse away from the release centre following the wind direction. If concentration of the hydrogen clouds is within its flammability limits (4-75%/vol) and an ignition source exists around the clouds, then a delayed outcome (such as a vapour cloud explosion, or flash fire) may occur. Frequencies of the event outcomes (such as explosion, fireball, etc.) for a given scenarios are calculated by multiplying the initial frequency with the associated branch probabilities of the event tree diagram.

Initiating

.event_

Instantaneous release

Immediate Cloud Delayed ignition denser than ignition

Explosion or fire

Incident outcomes

Explosion Fireball VCE Flash Fire Harmless VCE Flash Fire Harmless

Jet Fire

Flash Fire

Harmless

Flash Fire

Harmless

Fig. 4.6 Event tree diagram of GH2 release [17, 2]

4.4.3.1 Conditional Probabilities

In order to calculate the frequencies of the event outcomes the probabilities for each of the branches of the event tree diagram have to be first determined. They can be estimated by using fault tree analysis or are developed based on numbers of accidents in the past. In the Canvey Study, for example, it is quoted that covering 59 incidents involving small spills of LPG and flammable liquids gave probability of ignition of 0.9 [36]. On the other hand, the LPG study carried out by [187] set the probability of ignition occurring for road transport as shown in Table 4-6.

 Scenario Immediate ignition Delayed ignition No ignition Broken pipe hole 0.1 0.05 0.85 Instantaneous release of tank contents 0.4 0.5 0.1

Based on a limited number of hydrogen accidents, [36] gave some probabilities for hydrogen and other combustible materials released from road tankers that generally carry up to 30.3 m3 inventories. These values are presented in Tables 4-7 and Table 4-8, and are generally applicable to tank truck and station releases. Table 4-8 shows that the immediate ignition of hydrogen release is a very likely event. It gives the probability is 0.9 for large spill, and 0.5 for small spill. The same ways for delayed ignition gives probability of 0.09 for a large release, and 0.45 for a small release. The small spills meant that it involves 10% of tank inventory, large spill involve 100% of tank inventory.

 Fuel Small spill Large Spill Total Hydrogen G.G6 G.G2 G.G8 Propane G.G75 G.G25 G.G1 Gasoline G.G9 G.G7 G.16 Ethyl alcohol G.G9 G.G6 G.15

Table 4-8 Conditional probabilities of immediate ignition for given a spill [36]

Immediate Ignition Delayed Ignition

Small spill Large Spill Small spill Large Spill

Table 4-8 Conditional probabilities of immediate ignition for given a spill [36]

 Hydrogen G.5G G.9G G.45 G.G9 Propane G.25 G.75 G.68 G.23 Gasoline G.15 G.5G G.G4 G.G5 Ethyl alcohol G.2G G.6G G.G4 G.G4

Table 4-9 Conditional probability of hydrogen release used in the study

Event Instantaneous Continuous Sources

Table 4-9 Conditional probability of hydrogen release used in the study

Event Instantaneous Continuous Sources

 Immediate ignition G.9G G.5G Expert opinion A pool is formed? G.4G G.2G Relationship Pool ignited G.8G G.8G Relationship Cloud denser than air? G.2G G.2G Relationship Delayed ignition G.G9 G.45 Expert opinion Explosion rather than fire G.2G G.2G Historical data

Several references [36, 17, 96] concluded that an unconfined space release explosion is a very unlikely event, because an explosion requires several circumstances, such as the presence of obstacles and a very strong source of ignition (>1000 Joule) (see Chapter 2). According to the evaluation given by several authors based on explosion accidents in the past, only a small portion of the energy of a hydrogen cloud is expected to be liberated in an open air explosion; it is estimated to be in the range of 0.1 - 10%, mostly < 1% [96]. Statistical evaluation to 25 selected hydrogen accidents recorded by UNEP, OECD, MHDAS, BARPI (see Chapter 2) give a fire contribution for about 60%, explosion 30%, and 10% for both fire and explosion. Based on the above information it can be estimated that the probability is 0.2 for explosion, and 0.8 for fire. Table 4-10 give a summary of the incident outcomes probabilities used in the study.

Living Off The Grid

Get All The Support And Guidance You Need To Be A Success At Living Off The Grid. This Book Is One Of The Most Valuable Resources In The World When It Comes To When Living Within The Grid Is Not Making Sense Anymore.

Get My Free Ebook