WM2008 Conference, February 24 -28, 2008, Phoenix, AZ Using Fuzzy Logic as a Complement to Probabilistic Radioactive Waste Disposal Facilities Safety Assessment -8450 F. L. De Lemos CNEN- National Nuclear Energy Commission; Rua Prof. Mario Werneck, s/n, BH – CEP: 30123-970, Brazil T. Sullivan BNL-Brookhaven National Laboratory, USA ABSTRACT Probabilistic calculation assumes randomness in events and processes, i.e., it assumes that all the observations in a probability distribution have the same possibility of occurrence. The only difference from one event to another is their respective frequency. To avoid the situation that a very low probability event drives the decision making process, there must be a criterion for the predicted dose as a function of probability to be considered acceptable by regulatory authorities. For example, it can be required that the predicted dose at 2 standard deviations above the mean value for dose to be no more than three times the regulatory standard. This paper proposes the use of possibility analysis, as a complement to the probability analysis. In this approach two separate performance analyses, probabilistic and possibilistic, are performed and the results are used to complement each other. A case example is provided to illustrate the methodology. INTRODUCTION According to the 1989 International Atomic Energy Agency (IAEA) report, [1], the uncertainties can be classified as type A, aleatoric, and type B, epistemic. Aleatoric uncertainty is generated by occurrence of random and independent events, while epistemic uncertainty is generated by factors such as lack of data, ignorance and high complexity of the system. Probability theory is used to model aleatoric uncertainty. A number of methodologies exist to deal with the epistemic uncertainty and they are called non-probabilistic methods. The two types of uncertainty, aleatoric and epistemic, are sometimes mixed and modeled as probabilistic. This happens because the input distributions of data are also used to represent lack of understanding of some processes. For example, in high-level waste performance assessment, the fuel dissolution rate or input data related to canister failure contain both types of uncertainty [2]. This made it very difficult to evaluate the impact of the epistemic uncertainty in the Total System Performance 1