44 Developments in Importance Measures for Risk-Informed Ranking and Other Applications (PSAM-0049)

New developments and roles of different importance measures are described in two groups of activities:

• Making decisions about permissible permanent and temporary configurations and allowed configuration times (ACT) for regulation, technical specifications and on-line risk monitoring

• Ranking safety significance of systems, structures, components and human actions (SSCH) for preventive safety assurance activities (PSAA) such as safety classification, testing and inspections, quality assurance, preventive maintenance, training, procedures and human factors development

Risk Increase Factor (RIF) is most appropriately used to assess the risk and acceptability of a situation when a component is already known to be failed, or when planning to take a component out of service for maintenance or in an exceptional situation. Several ad hoc suggestions have been made in the past on how to define and calculate a RIF for a component that is modeled with multiple failure mode basic events, including common cause failures, and when information available about the actual configuration is uncertain. This paper introduces a precise definition for a multi-failure configuration, and shows how it can be analyzed under uncertainties. A general weighted average method (WAM) is developed and compared to several other candidate methods. Examples demonstrate that WAM makes sense and yields correct values in benchmark cases. Conditions have been found under which several methods yield equal results. The basic idea is the definition and calculation of Risk Gain (RG) for prediction, when a momentary configuration is known, and only partially known. In some cases success states can be taken into account, and the risk gain can also be less than 1. Relationships of RG to the conventional importance measures are described. Solutions can be obtained using normal fault tree techniques or directly from basic event importances.

The rationale of preventive activities is that effective PSAA improves the reliability characteristic of components (SSCH). The objective of ranking SSCH for PSAA is to make sure that the effect is proportional to the risk importance of different parts of SSCH. It follows from this that a useful risk importance measure must depend strongly on the reliability characteristic of the component being ranked. The Criticality Importance (CI) seems to be most appropriate, unlike Birnbaum importance (BI) or RIF. BI and RIF have different dimension and meaning for process pipes and safety pipes, which makes these measures alone unsuitable to rank both consistently with a common scale. It is shown that CI (Fussell-Vesely) has several advantages; it can rank both types of segments with the same scale, and yields exactly the same ranking as a recently suggested Differential Importance Measure (DIM). CI is routinely computed by most fault tree codes and it allows arbitrary partition of pipelines to welds, bends, sections etc. With CI the importance is partitioned proportionally to the failure rate, preserving correct ranking.

Traditional reliability importance measures are defined for basic events that are mutually statistically independent, and fault tree software often makes this assumption. At the same time many practitioners assume important common cause basic events to be mutually exclusive, without realizing that no event can be both at the same time. This paper develops some new expressions and relations under the mutual exclusivity assumption. These have implications on the methods that can be used in risk-informed configuration control.