The fault tree analysis has rapidly gained favor with reliability
analysts of complex systems. The main feature of the fault tree
technique is the versatility in the degree of detail in which the
analysis can be carried out. Also the analyst has options for
qualitative and quantitative analysis. And further, the simple logic
of the fault tree approach makes it a visibility tool for both
engineering and management. These general properties of the fault
tree technique are discussed in Section 1.
The first step in reliability analysis is the identification of all system
failure modes. This step is called the Failure Mode and Effect
Analysis (FMEA). During FMEA all component failures are
hypothesized and the possible adverse effects on the system are
determined by investigating how the system responds to each failure
and failure combination. Section 2 contains a brief discussion about
this preceding step for the actual fault tree analysis.
In Section 3 the general procedure for constructing a fault tree is
dealt with. Fault tree is a tool by which failures that can contribute to
an undesired event in the system (e.g. fire, explosion) are organized
deductively and represented pictorially. Fault tree is so one way to
diagram and relate the information developed in the preceding
FMEA. The resulting arrangement is a treelike logical structure with
information flows from the branches (component failures) to the top
of the tree (the undesired event, system failure). The general
construction principle of the fault tree and the commonly used fault
tree symbols - logic gates and fault events - as well as the main steps
involving in the tree construction are included in the discussion of
Section 3.
Section 4 deals with the analysis of a fault tree. Two different
approaches are possible: qualitative or quantitative analysis. In the
qualitative analysis the fault tree is inspected in order to determine
all the combinations of component failures that can lead to the
undersired event (minimal cut set evaluation). In the quantitative
analysis such quantitative reliability measures as
- the probability of occurrence of the undesired event
- the failure rate for the undesired event
- the expected number of occurences of the undesired event during
the time interval from 0 to t
- the expected number of undesired event occurences per unit time
can be obtained. Similar information can also be determined for the
minimal cut sets and primary events (component failures). There is
a great number of methods available for carrying out the fault tree
evaluation, both analytical and simulation methods. These methods
are also touched upon in Section 4.
The construction and evaluation of a fault tree is illustrated in
Section 5 by means of a simplifield system. The system is a lawn
mower internal combustion engine. The engine can be started by
batter power or by a pull cord. The undesired event is "Engine does
not start". One possible fault tree for the undesired event of this
system is presented. For the fault tree of the example both qualitative
and quantitative analysis are carried out.
(Tutkielmia ja tutkimusraportteja 1976, Publications of The
Turku School of Economics and Business Administration,
Series A1-2:1976, 243-269.)