Within the chemical industry it has become common practice
to assign a so-called SIL-value, which stands for safety
integrity level and which is part of the international IEC
61508 standard. It is an expression of a barrier’s assumed
reliability on a three-point scale. The number actually reflects
the exponent x in the formula 10−x rounded-off to the lowest
integer. So a SIL-value of 3 corresponds to a probability
of failure on demand between 10−4 and 10−3. The
M-index resulting from the ARAMIS audit can be used to
modify the SIL-values of any barrier in any scenario under
investigation.
The procedure for calculating the M-index is still rather
experimental, in that it is not based on empirical evidence
on management influences or experience. First of all, a decision
has been made that the delivery systems ‘risk analysis
and barrier selection’ and ‘learning and change management’
should be excluded from the quantification phase (see above).
The remaining seven delivery systems however do contribute
to current barrier effectiveness.
These delivery systems first had to be brought back to one
single number, instead of a separate rating per step. Therefore,
several persons with a certain degree of expertise (some
members from the ARAMIS consortium and 13 delegates
from a master’s course in safety) judged all steps within
these delivery systems for their influence on barrier effectiveness.
This resulted in a split between delivery systems
having equally weighted steps and systems having unequally
weighted steps. For the equally weighted delivery systems,
a geometrical mean is calculated to arrive at their overall
numerical rating (please note that all steps have been rated
on a five-point scale). For the unequally weighted ones, however,
a different reasoning applies. An average is first taken,
but when steps within a particular delivery system are judged
more important than others, the final mean rating of such a
system cannot be higher than the lowest scoring important
step. Or put in other words, the quality of less important
steps cannot compensate for the lesser quality of important
steps