A new approach to quantitative assessment of reliability of passive systems

Abstract

The objective of this paper is to show how probabilistic reliability can be assessed for complex systems in the absence of statistical data on their operating experience, based on performance evaluation of the dominant underlying physical processes. The approach is to distinguish between functional and performance probabilities when dealing with the quantification of the overall probability of a system to perform a given function in a given period of time (reliability). In the case of systems where sufficient statistical operating experience data are available, one can focus the quantitative evaluation entirely on the assessment of the functional probability for a given active item (e.g. a pump) by assuming that the specification, layout, construction and installation is such that the item is providing the assigned performance, e.g. in the form of generating the required flow rate. This is how traditional probabilistic safety assessments (PSAs) focus the reliability analysis for the various safety features on the calculation of values for the availability per demand. In contrast, for various systems relevant in advanced technical applications, such as passive safety features in innovative reactor designs, it is essential to evaluate both functional and performance probabilities explicitly and combine the two probabilities later on. This is of course due to the strong reliance of passive safety systems on inherent physical principles. In practice, this means that, for example, in case of a passive cooling system based on natural circulation of a given medium, one has to evaluate and to assess the probability to have a medium condition and a flow rate such that a cladding temperature, represented by a probability distribution, can be hold at a required level. A practical example of this method is given for the case of the reliability assessment of a residual passive heat removal system. General conclusions are drawn regarding reliability estimation of complex, interconnected systems in the absence of statistical performance data, such as for infrastructures.