A non-linear modification to Miner’s rule for damage accumulation is proposed to reduce the scatter between experimental fatigue life and fatigue life predicted according to the original Miner’s sum. Based on P-s-n probability distribution and design s-n curves, the modification satisfies the assumption of equality between the mean damage degree (at the critical level) and fatigue life random variables, which is not covered in the original formulation. The adopted formulation shows the discrepancies between the fatigue lives predicted according to the design s-n curves and the estimated probability distribution. It also proves that it is inappropriate to apply a normal distribution to fatigue life analysis and that the model becomes non-linear only for non-normal distributions. The predictions according to the established model were compared to the predictions obtained with Miner’s rule.
Welded joints submitted to variable loadings are, by their nature, experiencing a stress concentration resulting from the presence of a complex notch effect relating to geometrical and structural features of the joint. In the finite element method, modelling of the weld toe as a sharp notch reveals a non-linear relation of stresses dependent on the mesh density. In this paper, an approach for determination of the notch radius in a welded part was proposed, based on the application of finite element calculations with a non-linear elastic-plastic body model under employment of the linear hardening rule. The derived stresses are used in calculations of fatigue life, carried out according to algorithms provided for structural elements without the presence of welds.