TESTING PROCEDURES FOR THE DESIGN AND LIFE ESTIMATION OF FATIGUE-SENSITIVE STRUCTURES

Abstract The demands on applicability and reliability of experimental techniques for the estimation of allowable stresses and the corresponding fatigue life depend directly on the degree of reliability which can be attained in establishing the design stresses of a fatigue-sensitive structural component. In most cases the uncertainties involved in establishing the stresses expected in service, even when employing up-to-date statistical techniques, are such that the current reliability criteria, which have presumably been far too unconservative, must be replaced by more adequate criteria. Upon a critical examination of the existing techniques, it is realized that the allowable stresses can be estimated with satisfactory reliability only in rather infrequent cases, where the stress-time history can be described by a stationary random process. Satisfactory results can also be obtained for similar stress-time histories with constant mean stress and for those resulting from individual events of random or non-random nature. The situation is quite different in the case of stress-time histories where incremental stresses of a purely random nature are superimposed on a basic stress which is varying quasi-statically. The numerous parameters of such stress-time histories are the primary reasons that have made it impossible up to now to establish a technique yielding design data that can be generalized. In the present paper some work is mentioned that aims at overcoming the difficulties in this respect. The related problems will presumably not be completely solved for years to come; however, there are definite rules for calculation, design, and production that have to be observed in order to avoid premature failure of the fatigue-sensitive components of a structure.