What Shape for Frequency Distribution of Subjective Variables?
M. E. Soffiu, L. Lauria, and G. Marsili, Istituto Superiore di Sanit, viale Regina Elena 299, 00161 Rome – Italy Consequence analysis in Probabilistic Risk Assessment (PRA) of major industrial hazards needs the uncertainty affecting its results is evaluated. Consequently individual risk, as well as input variables, should be expressed as their frequency distributions rather than as single values. Although the frequency distribution of risk is very sensitive to the frequency distributions of the input variables, many of these last can difficulty be estimated from historical data because of the rarity and the heterogeneity of the events. As a consequence PRAs are usually carried out by adopting empirical or conservative values of the input variables. This approach, from an hand, produces a loss of information which can difficulty be quantified and, on the other hand, it cannot take in account the specificity of each case. In example, considering the diameter of the release orifice, which
