Information Technology Reference
In-Depth Information
Initiating
Event
Fire Spreads
Quickly?
Sprinkler Fails
To Work?
People Cannot
Resultant
Event
Scenario
Escape?
P = 0.5
1
Y
Multiple Fatalities
P = 0.3
Y
N
P = 0.1
2
Y
P = 0.5
Loss / Damage
Fire Starts
Frequency
= 1/yr
N
3
P = 0.7
Fire Controll d
N
4
P = 0.9
Fire Contained
Fig. 10.3 Example quantification of event tree for a building protected by a sprinkler system
(reproduced with permission from the Institution for Engineering and Technology)
tree. Fault trees can also be modified to include recovery actions; the resultant trees
are usually called recovery trees (van der Schaaf
1991
).
Figure
10.4
shows an example of a quantified fault tree for accidents at a
particular road junction. The legend at the left of the figure describes the proba-
bilities used in the figure. Cars driving too fast at this road junction, for example,
occur frequently, so it is given a probability of 0.1.
10.4.3 CREAM
The CREAM can be used both for the retrospective analysis of accidents and the
prospective analysis of possible errors in a system that is being designed. In both
cases, the method that is followed is essentially the same.
Here we will only focus on the simple retrospective use of the CREAM (as
shown in Fig.
10.5
). The process starts with the description of the initiating event
(which could be an accident or incident). This description needs to provide enough
detail to form the basis for the analysis. From this description it should be possible
to identify the error mode(s) (or phenotypes) associated with the event. The next
step is to try and identify the possible antecedents for that error mode, using the set
of tables of antecedents and consequents that lie at the heart of the CREAM method.
If a specific antecedent is found, or there are no general antecedents for the error
mode, then the analysis is complete. If a general antecedent is found then the next
step is to find the general consequent associated with that antecedent. If none can be
found, the analysis terminates, otherwise we use the general consequent as a spe-
cific consequent, and go through the loop again, this time trying to find a matching
Search WWH ::
Custom Search