Classification of metallic alloys for fatigue damage accumulation: A conservative model under strain control for 304 stainless steels

Fully reversed uniaxial tests performed under total strain and stress control on 304 stainless steels specimens show that, under strain control the fatigue damage for High-Low (H-L) cycling is more significant than that using Miner's rule, but under stress control opposite results are obtained. This has been attributed to opposite effects of pre-hardening under strain and stress control. Classical non linear damage accumulation models are not able to take into account this difference in sequence effect. Smith-Watson-Topper (SWT) and Fatemi-Socie (FS) criterion combined to linear damage accumulation can take into account this difference in sequence effect through the presence of maximum stress. However these models require an elastic-plastic constitutive law which is difficult to propose due to the presence of high cycle secondary hardening observed on 304 stainless steel. A conservative model for damage accumulation under variable amplitude strain control loading is thus proposed, which does not require a constitutive law. Linear damage accumulation is used, while sequence effect is taken into account using the elastic-plastic memory effect through cyclic strain-stress curves (CSSC) with pre-hardening. This modeling classifies metallic alloys in two groups for damage accumulation, with a stable (independent to pre-hardening) CSSC as for aluminum alloys and with an unstable (dependent to pre-hardening) one as for austenitic stainless steels. For the former case the modeling is identical to Miner's rule. The modeling is approved based on a large number of tests on 304 stainless steel and is compared with SWT and FS models. In presence of mean stress the modeling permits in a qualitative way to explain the fact that tensile mean stresses in constant amplitude strain control tests are more detrimental than for constant amplitude stress control tests. Moreover it is shown that the SWT model is not always able to predict accurately the fatigue life in presence of a mean stress. Finally, it is concluded that for a 304 stainless steel, in order to take into account the mean stress in fatigue life, the mean stress effect has to be decomposed into two parts: maximum and 'intrinsic' mean stress effects. (authors)