You are here: Homepage Documents test View document
|NKS Programme Area:||NKS-R|
|Research Area:||Risk analysis and probabilistic methods|
|Report Title:||Software reliability analysis for PSA: failure mode and data analysis|
|Authors:||Ola Bäckström, Jan-Erik Holmberg, Mariana Jockenhövel-Barttfeld, Markus Porthin, Andre Taurines, Tero Tyrväinen, |
|Abstract:||This report proposes a method for quantification of software reliability for the purpose probabilistic safety assessment (PSA) for nuclear power plants. It includes a failure modes taxonomy outlining the relevant software failures to be modelled in PSA, quantification models for each failure type as well as an analysis of operating data on software failures concerning the TELEPERM® XS (TXS) platform developed at AREVA.
Software related failure modes are defined by a) their location, i.e., in which module the fault is, and b) their effect on the I&C unit. For a processor the effect is either a fatal failure of the processor (termination of the function and no outputs are produced) or non-fatal failure where operation continues with possible wrong output values. Following cases are relevant from the PSA modelling point of view: 1) fatal failure causing loss of all subsystems that have the same system software, 2a) fatal failure causing loss of one subsystem, due to fault in system software, 2b) fatal failure in communication modules of one subsystem. 3) fatal failure causing failure of redundant set of I&C units in one subsystem, 4) non-fatal failure associated with an application software module. In the case 4, the failure effect can be a failure to actuate the function or a spurious actuation.
The failure rates for software fault cases 1 and 2, associated with the system software, are proposed to be estimated from general operational data for same system software. The probabilities on failure on demand for cases 3 and 4, associated with the application software, are a priori assumed to correlate with the complexity and degree of verification and validation (V&V) of the application. The degree of V&V is related to the safety class of the software system and the complexity can be assessed by analysing the logic diagram specification of the application. A priori estimates could be updated by operational data, which is demonstrated in the report.|
|Keywords:||PSA, Software reliability, Failure mode, Operational history data|
|Publication date:||01 Jul 2015|
|Number of downloads:||2221|