Advertisement

A Bayesian Network Approach to Accident Management and Estimation of Source Terms for Emergency Planning

  • Michael Zavisca
  • Heinrich Kahlert
  • Mohsen Khatib-Rahbar
  • Elizabeth Grindon
  • Ming Ang
Conference paper

Abstract

An approach to accident management and estimation of radiological releases for application to emergency planning is developed using the Bayesian Belief Network (BN) method, which is coupled to the Accident Diagnostics Analysis and Management (ADAM) system. This approach uses a BN to diagnose the initial and boundary conditions of the accident using the actual observable and sensor data from a plant. The topology and conditional probability values in the main part of the network are derived mainly from a level-2 PSA model that also includes the impact of various severe accident prevention and mitigation features that are included in current Severe Accident Management Guidelines (SAMGs). In this approach, the network links represent causality and logical consistency relationships. The plant instrumentation database and operator are queried in order to enter findings into the input nodes of the network (observables), after which posterior values of belief at the output (i. e., accident initial and boundary condition) nodes are post-processed (i. e., eliminating redundant or very similar scenarios) into a list accident scenarios ranked in accordance with te the inferred likelihood of occurrence. These results include the characteristics of the accident initiator and the availability of various systems, which are in turn used as inputs to ADAM to calculate the various phenomenological outcomes and their resulting radiological source terms for the top-ranked scenarios.

Keywords

Source Term Hide Node Emergency Planning Bayesian Belief Network Core Damage 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. 1.
    Grindon E, et al. A Rapid Response Source Term Indicator Based on Plant Status for Use in Emergency Response. FISA 2003, Luxembourg, 10-13 November 2003Google Scholar
  2. 2.
    Grindon E and Kinniburgh C. Probabilistic Source Term Predictions for Use with Decision Support Systems. Off-site Nuclear Emergency Management Symposium, Salzburg, September 29-October 3 2003Google Scholar
  3. 3.
    Dutton LMC, et al. The Use of Probabilistic Safety Analysis to Assist Emergency Response. Proc of INucE International Conference on the Commercial and Operational Benefits of PSAs, Edinburgh, October 1997Google Scholar
  4. 4.
    Esmaili H, et al. ADAM: An Accident Diagnostics, Analysis, and Management System. In: Advances in Safety and Reliability, C. G. Soares (ed.), Pergamon, U.K., 1997Google Scholar
  5. 5.
    Norsys Software Corp. Netica Application for Belief Networks and Influence Diagrams — User’s Guide. March 1997Google Scholar
  6. 6.
    Cooper G. Probabilistic Inference Using Belief Networks is NP-Hard. In: Artificial Intelligence, 42: 393–405, 1990CrossRefMATHMathSciNetGoogle Scholar

Copyright information

© Springer-Verlag London 2004

Authors and Affiliations

  • Michael Zavisca
    • 1
  • Heinrich Kahlert
    • 1
  • Mohsen Khatib-Rahbar
    • 1
  • Elizabeth Grindon
    • 2
  • Ming Ang
    • 2
  1. 1.ERI Consulting & Co.Rotkreuz
  2. 2.NNC LimitedKnutsford, CheshireUK

Personalised recommendations