The probabilistic safety assessment (PSA) for gas-cooled nuclear power plants has been investigated where the operational data are deficient, because there is not any commercial gas-cooled nuclear power plant. Therefore, it is necessary to use the statistical data for the basic event constructions. Several estimations for the safety margin are introduced for the quantification of the failure frequency in the basic event, which is made by the concept of the impact and affordability. Trend of probability of failure (TPF) and fuzzy converter (FC) are introduced using the safety margin, which shows the simplified and easy configurations for the event characteristics. The mass flow rate in the natural circulation is studied for the modeling. The potential energy in the gravity, the temperature and pressure in the heat conduction, and the heat transfer rate in the internal stored energy are also investigated. The values in the probability set are compared with those of the fuzzy set modeling. Non-linearity of the safety margin is expressed by the fuzziness of the membership function. This artificial intelligence analysis of the fuzzy set could enhance the reliability of the system comparing to the probabilistic analysis. (C) 2009 Elsevier Ltd. All rights reserved.