Defining Statistical Test-Cases for Safety-Related Applications

摘要

With increased use of software-based or logic-based components in safety-related systems, the issue arises to assessthe dependability (availability or reliability) of such components. Dependability assessment for hardwarecomponents seems well established, however for software this is not the case. Currently, apart from formal proofs,the only way of quantifying software dependability is via statistical testing. In this paper we describe our experiencewith the design of statistical test-cases on nuclear industry safety-related components. We report on our observationsgathered, for example, on the benefits of using such tests and on the current limitations.