DeVa project reports on Bayesian Belief Networks in safety assessment

These two technical reports were produced during the DeVa project ("Design for Validation", ESPRIT Long Term Research Project No. 20072, 1996-1999)

Bayesian Belief Network Model for the Safety Assessment of Nuclear Computer-based Systems by N.E. Fenton, B. Littlewood, M. Neil, L. Strigini, D.R. Wright (City University, London) and P.-J. Courtois (AVN, Brussels),
DeVa ESPRIT Long Term Research Project No. 20072 - 2nd Year Report, pp. 485-512, Dec, 1997.

Examination of Bayesian Belief Network for Safety Assessment of Nuclear Computer-based Systems by B. Littlewood, L. Strigini, D. Wright (City University, London) and P.-J. Courtois (AVN, Brussels)
DeVa ESPRIT Long Term Research Project No. 20072 - 3rd Year Report, pp. 411-448, Dec. 1998

Abstracts and full text are available here.

Bayesian Belief Network Model for the Safety Assessment of Nuclear Computer-based Systems

N.E. Fenton, B. Littlewood, M. Neil, L. Strigini, D.R. Wright
City University, London
P.-J. Courtois
AVN, Brussels

Abstract

The formalism of Bayesian Belief Networks (BBNs) is being increasingly applied to probabilistic modelling and decision problems in a widening variety of fields. This method provides the advantages of a formal probabilistic model, presented in an easily assimilated visual form, together with the ready availability of efficient computational methods and tools for exploring model consequences. Here we formulate one BBN model of a part of the safety assessment task for computer and software based nuclear systems important to safety. Our model is developed from the perspective of an independent safety assessor who is presented with the task of evaluating evidence from disparate sources: the requirement specification and verification documentation of the system licensee and of the system manufacturer; the previous reputation of the various participants in the design process; knowledge of commercial pressures;information about tools and resources used; and many other sources. Based on these multiple sources of evidence, the independent assessor is ultimately obliged to make a decision as to whether or not the system should be licensed for operation within a particular nuclear plant environment. Our BBN model is a contribution towards a formal model of this decision problem. We restrict attention to a part of this problem: the safety analysis of the Computer System Specification documentation. As with other BBN applications we see this modelling activity as having several potential benefits. It employs a rigorous formalism as a focus for examination, discussion, and criticism of arguments about safety. It obliges the modeller to be very explicit about assumptions concerning probabilistic dependencies, correlations, and causal relationships. It allows sensitivity analyses to be carried out. Ultimately we envisage this BBN, or some later development of it, forming part of a larger model, which might well take the form of a larger BBN model, covering all sources of evidence about pre-operational life-cycle stages. This could provide an integrated model of all aspects of the task of the independent assessor, leading up to the final judgement about system safety in a particular context. We expect to offer some results of this further work later in the DeVa project.
Full text

Examination of Bayesian Belief Network for Safety Assessment of Nuclear Computer-based Systems

B. Littlewood, L. Strigini, D. Wright
City University, London

P.-J. Courtois
AV Nuclear. 1998

Abstract

We report here on a continuation of work on the Bayesian Belief Network (BBN) model described in DeVa Tech. Report No 52. As explained in the previous deliverable, our model concerns one part of the safety assessment task for computer and software based nuclear systems. We have produced a first complete, functioning version of our BBN model by eliciting a large numerical node probability table (NPT) required for our `Design Process Performance' variable. The requirement for such large numerical NPTs poses some difficult questions about how, in general, large NPTs should be elicited from domain experts. We report about the methods we have devised to support the expert in building and validating a BBN. On the one hand, we have proceeded by eliciting approximate descriptions of the expert's probabilistic beliefs, in terms of properties like stochastic orderings among distributions; on the other hand, we have explored ways of presenting to the expert visual and algebraic descriptions of relations among variables in the BBN, to assist the expert in an ongoing assessment of the validity of the BBN.
Full text
Links to: DeVa Technical Report Series index / DeVa Home Page

Page maintained by: webweaver@csr.city.ac.uk
Last modified 15 October 1999.