10
th
Lecture: Bayesian Probabilistic Nets in Risk Assessment
Aim of the present lecture
The present lecture introduces Bayesian Probabilistic Nets (BPN’s) as a tool of general
applicability in engineering risk assessment and risk management. First the aspects of
causality are introduced through an example and thereafter the basic theory of BPN’s with
discrete states is introduced. Following this, examples are provided whereby the use of BPN’s
is illustrated for the purpose of risk assessment as well as for decision making. The examples
also show how classical fault tree and event tree analysis can be performed using BPN’s, and
it is highlighted in which ways the BPN’s allow, in important ways, for more general risk
assessments and sensitivity studies. Finally the topic of large scale risk assessment in regard
to the management of risks due to natural hazards is addressed. A framework for such risk
assessments, based on the methods introduced in Lecture 4, in conjunction with BPN’s and
Geographical Information Systems (GIS) is outlined and the use of this is illustrated by
examples considering risk management of buildings in larger cities subject to earthquake
hazards.
Based on the introduced material in this lecture it is aimed for that the students should acquire
knowledge and skills in regard to:
What is causality and how can causality be represented graphically?
What is a BPN and which are the principles underlying its functionality?
How can risk assessments be performed using BPN’s?
How to construct “AND” and “OR” gates by means of conditional probability tables?
How can decision analysis be performed using BPN’s?
In which way may generic BPN’s be formulated for the purpose of GIS supported large
scale risk management?