Fatigue behaviour of AA6082 friction stir welds under variable loadings

Abstract Welded components are often subjected to variable amplitude service loads, demanding fatigue life prediction methods that can take into account fatigue damage accumulation. In previous work the fatigue behaviour of friction stir welding of aluminium alloys under constant and variable amplitude loadings was analysed under the stress ratio R = 0. The objective of this work is to extend the study of the fatigue strength of friction stir welds in AA6082-T6 under constant and variable amplitude loadings under the stress ratio R = −1 and analyse the validity of Miners´s and Manson–Halford damage sum rules for both stress ratios R = 0 and R = −1. Fatigue tests were carried out in a servo-hydraulic testing machine using typified Gassner amplitude spectra, ranging the correspondent shape exponent between 1.5 and 5. Constant amplitude fatigue tests revealed that crack initiation occurs internally from tunnel defects or at the surface, near stress concentration due to shear lips, leading to fatigue life reduction relative to the base material. Tunnel defects showed to be more detrimental for fatigue resistance than the stress concentration created near shear lips. As expected, a significant mean stress influence was observed. Therefore, the assumption in design codes of no mean stress effect, based in the presence of high residual stresses close to the yield stress, seems not adequate to be applied for friction stir butt welds. Although residual stresses were not measured in this work, the small test samples (160 × 15 × 4 mm) tested transversely implies that they must be very small. The comparison of experimental fatigue lives with predictions calculated with both Miners´s Linear Damage Rule and Manson–Halford Double-Linear Damage Rule, using two stress ratios and four spectrum shape factor values, revealed a good agreement for R = 0. Under R = −1 both damage predictions methods were, in general, unconservative, with the Double-Linear Damage Rule being less unconservative. Therefore, the application of the Double-Linear Damage Rule can be considered advantageous as it requires no further input information than the required by the Linear Damage Rule and takes into account both loading level and loading sequence effects.