119 The Methodology to Build the Network Used in a Bayesian Belief Net Approach (PSAM-0068)

In using the Bayesian belief net (BBN) method, there are two phases: the first step is to build the graphical structure and the second step is to define the node probability table, or the distribution for each node of the network. In spite of the fact that the graph has a great importance for the BBN approach, most of the studies in the literature pay attention only on the quantification phase. In these studies, the networks are presented and used and there is no specification about how they have been built. This paper presents the methodology to build the network used in a BBN approach. The procedure has several steps, starting with the construction of the model in a natural language, the transformation of the model into an operational one, the implementation and the identification of the decision variables. The result of the procedure is a prototype of the network which is tested in order to be improved. The methodology can be applied also for building the graphical model of a fault tree or event tree. The paper contains an example of network used in order to estimate the distribution of the human damage produced by a fire in a building.