Autonomy: Risk Assessment 24.2 Risk of Failure 531
Part B | 24.2
architecture. The failure model end effect analysis
(FMEA) is a well-established technique for identifying critical components or phases of the system design
and operation [24.18]. However, this technique does not
accommodate quantitative estimation of the system’s
reliability and therefore it will not be discussed in this
chapter. Event trees and fault trees are two basic methods for system reliability quantification. Event trees use
forward logic. They begin with an initiating event (an
abnormal incident) and propagate this event through the
system under study by considering all possible ways
in which it can affect the behavior of the subsystems.
Event trees are useful for accident sequence analysis in
which the aim is to estimate the likelihood of a sequence
of potential functioning or malfunctioning events. Thus
event trees cannot be used for estimating the probability of failure of a system. This figure can be estimated
with the support of fault trees. In a fault tree analysis,
one attempts to develop a deterministic description of
the occurrence of an event, the top event, in terms of the
probability of occurrence of other (intermediate) events.
Intermediate events are also described further until, at
the finest level of detail, the basic events are reached.
When the top event is failure of a system then the basic events are usually failures of components. A fault
tree can be considered as an expression in a Boolean
model, which implements predicate logic. A fault tree
contains the same operators, for example, conjunctions
Calculation
error
Payload
config error
Pressure
vessel leak
Human error
Ice jams
dive planes
Clutch jams
Human
failure
Ballasted
incorrectly
Failure to
launch AUV
G 1
G 3
G 4
G 5
G 2
4
6
Wrong
density used
5
7
Human error
not noticed
Buoyancy
material
damaged
8
9
Premature
emergency
weight drop
Dive plane
mech problem
3
2
1
10
Fig. 24.2 Probability of failure during AUV deployment through an ice hole. Fault tree presented in [24.19]
and disjunctions. However, instead of operating with
bits, where 0 corresponds to false and 1 corresponds
to true, fault trees operates with probabilities – varying from 0 to 1 [24.20]. Figure 24.2 presents a fault
tree devised by Griffiths and Brito for estimating the
probability of failure during the deployment of an AUV
through an ice hole [24.19]. Graphically a conjunction
is represented using AND and a disjunction is represented using an OR gate. In Fig. 24.2, the operators
named G 1 , G 2 , G 3 , and G 4 are OR gates. The G 5 operator is an AND gate. A fault tree injects transparency
into the fault analysis process. Failure modes are clearly
presented and how they propagate in the system is visually displayed.
There is a unique probabilistic model associated
with each fault tree, which enable us to calculate the
probability of failure for the top event. The probability
of failure to launch, P10, is calculated using the following expression:
P10 D G 1
D 1 .1 G 2 / .1 P3/ .1 P4/
.1 G 3 /
D 1 ..1 P1/ .1 P2//
.1 P3/ .1 P4/
.1 .1 .1 G 4 / .1 G 5 /// :
(24.1)
Part B | 24.2
architecture. The failure model end effect analysis
(FMEA) is a well-established technique for identifying critical components or phases of the system design
and operation [24.18]. However, this technique does not
accommodate quantitative estimation of the system’s
reliability and therefore it will not be discussed in this
chapter. Event trees and fault trees are two basic methods for system reliability quantification. Event trees use
forward logic. They begin with an initiating event (an
abnormal incident) and propagate this event through the
system under study by considering all possible ways
in which it can affect the behavior of the subsystems.
Event trees are useful for accident sequence analysis in
which the aim is to estimate the likelihood of a sequence
of potential functioning or malfunctioning events. Thus
event trees cannot be used for estimating the probability of failure of a system. This figure can be estimated
with the support of fault trees. In a fault tree analysis,
one attempts to develop a deterministic description of
the occurrence of an event, the top event, in terms of the
probability of occurrence of other (intermediate) events.
Intermediate events are also described further until, at
the finest level of detail, the basic events are reached.
When the top event is failure of a system then the basic events are usually failures of components. A fault
tree can be considered as an expression in a Boolean
model, which implements predicate logic. A fault tree
contains the same operators, for example, conjunctions
Calculation
error
Payload
config error
Pressure
vessel leak
Human error
Ice jams
dive planes
Clutch jams
Human
failure
Ballasted
incorrectly
Failure to
launch AUV
G 1
G 3
G 4
G 5
G 2
4
6
Wrong
density used
5
7
Human error
not noticed
Buoyancy
material
damaged
8
9
Premature
emergency
weight drop
Dive plane
mech problem
3
2
1
10
Fig. 24.2 Probability of failure during AUV deployment through an ice hole. Fault tree presented in [24.19]
and disjunctions. However, instead of operating with
bits, where 0 corresponds to false and 1 corresponds
to true, fault trees operates with probabilities – varying from 0 to 1 [24.20]. Figure 24.2 presents a fault
tree devised by Griffiths and Brito for estimating the
probability of failure during the deployment of an AUV
through an ice hole [24.19]. Graphically a conjunction
is represented using AND and a disjunction is represented using an OR gate. In Fig. 24.2, the operators
named G 1 , G 2 , G 3 , and G 4 are OR gates. The G 5 operator is an AND gate. A fault tree injects transparency
into the fault analysis process. Failure modes are clearly
presented and how they propagate in the system is visually displayed.
There is a unique probabilistic model associated
with each fault tree, which enable us to calculate the
probability of failure for the top event. The probability
of failure to launch, P10, is calculated using the following expression:
P10 D G 1
D 1 .1 G 2 / .1 P3/ .1 P4/
.1 G 3 /
D 1 ..1 P1/ .1 P2//
.1 P3/ .1 P4/
.1 .1 .1 G 4 / .1 G 5 /// :
(24.1)