Change search
ReferencesLink to record
Permanent link

Direct link
Development of methodology for identification of failure domains with GA-DPSA
KTH, School of Engineering Sciences (SCI), Physics, Nuclear Power Safety.ORCID iD: 0000-0002-0683-9136
2012 (English)In: 11th International Probabilistic Safety Assessment and Management Conference and the Annual European Safety and Reliability Conference 2012, PSAM11 ESREL 2012: Volume 3, 2012, 2480-2489 p.Conference paper (Refereed)
Abstract [en]

Methods of PSA/PRA play important role in understanding of threats to Nuclear Power Plants (NPPs) safety. However, static logic of PSA has difficulties in considering the dynamic nature of physical processes and their interaction with stochastic events. Different Dynamic PSA (DPSA) methods have been proposed to resolve the influence of timing and order of events in safety analysis of NPP. In this work we discuss a DPSA approach which employs global optimum search methods (particularly genetic algorithm (GA)) for the exploration of the uncertainty space (the space of plant accident scenarios and uncertain parameters) and a system code as a deterministic model of the plant. The GA is used to generate the system code input for probing the uncertainty space. Safety important parameters (e.g. peak cladding temperature etc.) are used by GA as objective functions (which are also often called fitness functions in GA) to guide selection of the system code input to find conditions at which safety limits are exceeded (failure domains (FD) in the uncertainty space). The biggest challenges in the problem of FD identification are (i) difficulties caused by enormous dimensionality of the space, (ii) large variations in sensitivity of the fitness function to different input parameters, (iii) significant cross correlations between input parameters, and (iv) non-monotonic behavior of the fitness function in the whole uncertainty space. In this paper we report most recent developments of the GA-DPSA methodology. Specifically we investigate the influence of the selection of GA internal parameters on the efficiency of failure domain identification. A method for probability estimation based on the neural networks is discussed. A test case for DPSA methods is proposed based on the LOCA scenario for a PWR model distributed along with the RELAP5 code. Presented test case reveals intricate dynamic interactions in different accident scenarios despite relative simplicity of the model.

Place, publisher, year, edition, pages
2012. 2480-2489 p.
Keyword [en]
Dynamic PSA, Genetic algorithm, Safety analysis
National Category
Physical Sciences
URN: urn:nbn:se:kth:diva-118408ScopusID: 2-s2.0-84873161376ISBN: 978-162276436-5OAI: diva2:606140
11th International Probabilistic Safety Assessment and Management Conference and the Annual European Safety and Reliability Conference 2012, PSAM11 ESREL 2012, 25 June 2012 through 29 June 2012, Helsinki

QC 20130218

Available from: 2013-02-18 Created: 2013-02-18 Last updated: 2013-02-18Bibliographically approved

Open Access in DiVA

No full text


Search in DiVA

By author/editor
Kudinov, Pavel
By organisation
Nuclear Power Safety
Physical Sciences

Search outside of DiVA

GoogleGoogle Scholar
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

Total: 28 hits
ReferencesLink to record
Permanent link

Direct link