A study on fatigue behavior of MIG-welded Al-Mg alloy with different filler-wire materials under mean stress

Cylindrical specimens of ϕ5 mm were cut from the weld bead of 50 mm thick MIG-welded Al-Mg alloy (Al-5083) plates with two different filler materials: Al-5183 (commonly used) and Al-5.8%Mg (special hardened new alloy). Load-controlled fatigue tests were done at different stress-ratios (R=-1,-0.5,0.1,0.5) using sinusoidal wave at 30 Hz and the performance of three popular finite-life empirical models: Gerber's parabola, Smith-Watson-Topper's model and Walker's model, was studied using the experimental data. Fatigue lives and endurance limits of both materials did not differ significantly, but they got reduced when the R-ratio was increased and the endurance limit followed the Gerber's parabola. A non-monotonic trend in the slope of S-N curves was observed: first increases and then decreases. Fracture surface observations revealed majority of surface-initiated failures at lower R-ratios (-1,-0.5), while for higher R-ratios (0.1,0.5), defect-induced failures were predominant. This shift in crack initiation site was attributed to the local cyclic plasticity due to stress-concentration and demonstrated using elastic-plastic finite element (FE) simulations. The detrimental effect of mean-stress on fatigue lives was attributed to combined effects of reduction in crack-nucleation time: due to stress concentration at defects and the crack closure effects. The three fatigue life models predicted monotonic decrease in slopes of S-N curves and thus did not give best results. A new model, based on Walker's model, was proposed which was able to capture the observed non-monotonic trend of the slopes of S-N curves and predicted fatigue lives better than these models.