Integrating Software into PRA: A Test-Based Approach

摘要

Probabilistic Risk Assessment (PRA) is a methodology to assess the probability of failure or success of a system's operation. Results provided by the risk assessment methodology can be used to make decisions concerning choice of upgrades, scheduling of maintenance, improvement to the design, etc. Current PRA modelling concentrates on representing the behaviour of hardware systems, humans and their contributions (to a limited extent) to risk. This neglects the current technological reality, i.e. most complex systems are software intensive or will be. It is thus imperative to consider and model the impact of software on risk if one wishes PRA to reflect the risk in current and future systems. The following paper presents a methodology to systematically integrate software contributions into PRA and as such consists of one of the first efforts to resolve this is-sue. Our previous research has established and validated a software-related failure mode taxonomy. This taxonomy can be used to identify potential software contributors to risk. The failure mode taxonomy encompasses four classes of failures: input failures, failures of the software itself, support failures and output failures. In this paper we show how these contributions can be modelled and quantified within the frame of a classical PRA analysis, i.e. within a modelling environment that uses primarily combinations of Event Sequence Diagrams (ESDs) or Event Trees (ET) and Fault Trees (FTs). The methodology proposed is based on the following three elements: the software functional decomposition, a software input tree and a software safety testing strategy. The assumptions underlying the methodology are also discussed.