Page 278 - CHEF Guide

Layers of Protection Analysis

a larger group of people may be present. Personnel in adjacent units will also need to be considered if the event is large
enough to affect more than the immediate area. The probability of personnel presence must be independent of the scenario
being evaluated. This is often not the case when operator response to an alarm is involved or the presence of an operator
would be required for the initiating event to occur, such as an error made during a loading or unloading operation.

Probability of Ignition - The LOPA conditional probability of a flammable vapor, explosible dust cloud or combustible
mist igniting or an uncontrolled reaction (such as an explosive decomposition) initiating is treated in various ways by different
companies. A probability of ignition or initiation may be associated with two basic scenario types:
• Igniting a flammable or explosible atmosphere inside process equipment, resulting in an internal combustion
reaction (usually a deflagration; transition to a detonation may be possible under the right conditions) that may or
may not breach the primary containment.
• Igniting flammable vapors or an ignitable dust cloud external to process equipment, resulting in a flash fire, pool
fire, jet fire and/or vapor cloud explosion.
The probability of ignition of a flammable or explosible atmosphere inside process equipment will not always warrant a
conditional modifier. Scenarios involving ignition inside process equipment must be carefully evaluated on a case-by-case
basis. When there is normally a flammable or explosible atmosphere inside the process equipment, and an internal
deflagration will be initiated as soon as a sufficiently energetic ignition source is present, then the presence of the ignition
source is the initiating event (such as failure of the grounding and bonding system or where failure of mechanical
components may result in hot surfaces).
Many of the same internal ignition considerations also apply to ignition outside of process equipment, such as following
an atmospheric release or a loss of primary containment (LOPC) event. Other factors needing to be considered when
assessing the probability of ignition include:
• Size of the flammable vapor cloud or ignitable dust cloud (larger releases may extend beyond electrical
classification boundaries and involve more potential ignition sources)
• Duration of the release (since probability of ignition by e.g. vehicle traffic will increase as the duration of the release
is extended)
• Overall number and strength of potential ignition sources
• Minimum ignition energy of the flammable vapors or ignitable dust.
• Particle size of ignitable dust.
The approach used in CHEF and RAST is based on two primary considerations: the probability that the flammable
cloud reaches an ignition source, and the probability of ignition from the source. The larger the flammable cloud footprint
(or greater the release rate and corresponding cloud size), the more likely reaching an ignition source will be. In RAST,
categories for the “Ease of Ignition” is also taken into account [78]. Releases of pyrophoric materials are assumed to always
ignite (POI=1) independent of release quantity or strength of ignition sources while materials such as ammonia or methylene
chloride are categorized as “Low” Ease of Ignition and likely need a very strong source (such as open flame) for ignition to
occur. Most flammable materials (such as methane, propane, etc.) would be considered for “normal” Ease of Ignition. The
concept of an area source is described in the literature, with an example for an LPG release shown in Figure 17-4 [79].

The values for the ignition probability of known ignition sources once the flammable cloud has reached the source are
shown in Table 17-2 [80, p. 50]. The source strength for point sources are much higher than 0.1 and might be rounded to
1 if using “order of magnitude” values in a Layers of Protection Analysis. If there is not a known ignition source within the
flammable cloud, an approach based on probability of reaching an ignition source may be preferred rather than assume no
ignition occurs. This approach of considering both likelihood that the flammable cloud reaches an ignition source and the

Page 238