This preview shows page 1. Sign up to view the full content.
Unformatted text preview: ty estimates and
outcome dollar values. NASA Systems Engineering Handbook
Management Issues in Systems Engineering
Probabilistic Risk Assessment (PRA). A PRA seeks
to measure the risk inherent in a system's design and
operation by quantifying both the likelihood of various
possible accident sequences and their consequences. A
typical PRA application is to determine the risk
associated with a specific nuclear power plant. Within
NASA, PRAs are used to demonstrate, for example, the
relative safety of launching spacecraft containing RTGs
(Radioisotope Thermoelectric Generators).
The search for accident sequences is facilitated
by event trees, which depict initiating events and
combinations of system successes and failures, and
fault trees, which depict ways in which the system
failures represented in an event tree can occur. When
integrated, an event tree and its associated fault tree(s)
can be used to calculate the probability of each accident
sequence. The structure and
Probabilistic Risk Assessment Pitfalls
Risk is generally defined in a probabilistic risk assessment (PRA) as the expected value of a consequence
R = Σ PS CS
S where PS is the probability of outcome s, and CS is the
consequence of outcome s. To attach probabilities to
outcomes, event trees and fault trees are developed.
These techniques have been used since 1953, but by
the late 1970s, they were under attack by PRA
practitioners. The reasons include the following:
• • • Fault trees are limiting because a complete set
of failures is not definable.
Common cause failures could not be captured
properly. An example of a common cause failure is one where all the valves in a system have
a defect so that their failures are not truly independent.
PRA results are sometimes sensitive to simple
changes in event tree assumptions
Stated criteria for accepting different kinds of
risks are often inconsistent, and therefore not
appropriate for allocating risk reduction resources.
Many risk-related decisions are driven by...
View Full Document