Evaluation of structural creep reliability using entropy principles

Abstract

The principle of maximum entropy is used to obtain the prior probability distribution functions for critical structural creep strain and structural creep rupture at known high temperatures and uniaxial stresses. From the prior distribution function obtained, reliability function which is simply the probability of successful operation, can be derived for specified critical creep-strain and creep-rupture modes of failure. An attempt is made to derive the reliability functions from prior considerations of the mechanics of failure, and the mechanical and physical characteristics of engineering materials. This work assumes that mechanical creep design reliability functions for creep-rupture and critical creep-strain modes of structural elements can have values such that the failure of the elements can occur either by any of the modes of failure or by the assumed combined modes of failure. It is also pointed out that the prior probability distribution functions from which the reliability functions are derived can be improved by the use of Baye's theorem in order to obtain a posterior probability distribution function, whenever more creep data are made available. The posterior probability distribution functions can then be used to derive more accurate reliability functions. Finally, these considerations and procedures which yield the reliability design criteria are illustrated by an application to the stress analysis of a structural member, with given mechanical, physical, environmental and creep properties.