ABSTRACT
Risks are unavoidable and as such the key challenge in engineering risk analysis is to identify
the elements of the system or facility that contribute most to risk and associated uncertainties.
To identify such contributors, the common method used is the importance ranking. One of
the most useful outputs of a risk assessment, especially PRA, is the set of importance
measures associated with the main elements of the risk models such as phenomena, failure
events, and processes. These importance measures are used to rank the risk-significance of
these elements in terms of their contributions to the total risk (e.g., expected loss or hazard)
assessed in the PRA. Importance measures are either absolute or relative. The absolute
measures define the contribution of each risk element in terms of an absolute risk metric
(reference level), such as the conditional frequency of a hazard exposure given a particular
state of the element. Relative measures compare the risk contribution of each element with
respect to others. In most risk analyses, it is common to conclude that importance measures of
a small fraction of risk elements contribute appreciably to the total risk. That is, often the
Pareto rule applies in which less than 20% of the elements in the risk model contribute to more
than 80% of the total risk. Moreover, the importance indices of risk elements usually cluster in
groups that may differ by orders of magnitude from one another. As such, the importance
indices are radically different such that they are generally insensitive to the precision of the
data used in the risk model.