Software reliability estimation of nuclear critical safety system using statistics of the extremes

Applying statistics of the extremes to the reliability estimation of nuclear safety software is proposed to improve alternately the reliability evaluation methodology. The software of nuclear safety functions such as reactor protection system and heat removal system frequently unable to produce pro-operation testing data and field experienced failure data for at least two reasons; the safety software is the one-of-a-kind for special purpose usage and it rarely fails in the testing. Therefore, the conventional Gaussian reliability model cannot be used for this purpose. Instead statistics of the extremes which is developed for analysis of rare events may be preferred to applying on this. This gas such a advantage that it does not require a prior assumptions concerning the distributional form of failure history data but it concentrates on the tails of the data. It is presented in this paper the theoretical background and its applying methodology