Development and Application of a Genetic Algorithm Based Dynamic PRA Methodology to Plant Vulnerability Search
2011 (English)In: International Topical Meeting on Probabilistic Safety Assessment and Analysis 2011, PSA 2011, 2011, 559-573 p.Conference paper (Refereed)
The paper describes recent achievements in development and application of the Dynamic Probabilistic Risk Analysis (DPRA) methodology based on the Genetic Algorithm (GA). The aim of the GA-DPRA approach is to enable identification of safety vulnerabilities and quantification of accident risks related to operation of nuclear power plants (NPP). The approach combines a system code as a deterministic model of the plant and a GA search engine for the exploration of the plant scenarios space. A point in this space represents a scenario (transient) which is defined by unique combination of initial plant state and time dependent sequence of changes in the plant state parameters implemented in the system code input. The GA-DPRA is used to address two main types of safety analysis problems: (i) identification of a "worst case" scenario with most severe violation of safety limits (failure of safety barriers); (ii) identification of "failure domains" (subdomains in the space of plant scenarios where at least one of the safety limits (barriers) is violated). Safety critical parameters (safety limits) are used by GA as fitness functions to guide selection of the system code input parameters in process of the global optimum search. The GA controls selection of system code input parameters within predefined diapasons and time windows. Unlike "brute force" approaches or Monte Carlo type methods the GA-DPRA is much less demanding to computational resources due to intelligent and adaptive resolution in the exploration of the plant scenarios space. Stochastic properties of GA and Importance Sampling technique are applied to estimate probabilistic characteristics of the identified vulnerabilities. Solutions of benchmark problems and comparison with other methods are discussed in the paper.
Place, publisher, year, edition, pages
2011. 559-573 p.
IdentifiersURN: urn:nbn:se:kth:diva-53609ScopusID: 2-s2.0-80051997967ISBN: 978-1-61782-847OAI: oai:DiVA.org:kth-53609DiVA: diva2:470439
International Topical Meeting on Probabilistic Safety Assessment and Analysis 2011, PSA 2011. Wilmington, NC. 13 March 2011 - 17 March 2011
QC 201201042011-12-292011-12-292012-01-04Bibliographically approved