TABLE OF CONTENTS Michael Oberguggenberger [1] and Wolfgang Fellin [2]
Summary. In this article, we scrutinize basic issues concerning the interpretation of probability in the probabilistic safety concept. Using simple geotechnical design problems we demonstrate that the failure probability depends in an extremely sensitive way on the choice of the input distribution function. We conclude that the failure probability has no meaning as a frequency of failure. It may supply, however, a useful means for decision making under uncertainty. We suggest a number of alternatives, as interval probability, random and fuzzy sets, which serve the same purpose in a more robust way.
This paper addresses the role of probabilistic modeling in geotechnical problems. There is a general awareness of the uncertainties in all questions of geotechnical engineering; by now, it has also become clear that the uncertainties themselves have to be modeled. Common practice is to use a probabilistic set-up to achieve this task. The probabilistic format should be seen as supporting the decision process under uncertainty, as formulated e. g. by Einstein [6]. It helps structuring the problem and aids in obtaining qualitative judgments.
The numerical values thus obtained, like failure probabilities and safety factors, play an important role in comparative and qualitative studies. The point we wish to make, however, is that these numerical values do not allow quantitative assertions about reality.
In particular, contrary to common language, the failure probability cannot be interpreted as a frequency of failure. This fact was already pointed out by Bolotin [2]...
