PSI - Issue 57

Jan Papuga et al. / Procedia Structural Integrity 57 (2024) 79–86 Author name / Structural Integrity Procedia 00 (2019) 000 – 000

81

3

Fig. 1. Left: Drawing of the cavity specimen described in this paper; Right: Platforms with the built specimens from top view. The cavity specimens are those on the right side. The recoater moves from right to left and back, the inert gas flows from right to left.

2.2. Testing The specimens were all tested at IST, Lisbon using Instron 8874 biaxial servohydraulic fatigue testing system with the load range ±25 kN in push-pull and 100 Nm in torsion. The typical testing frequencies were within the range of 5 to 10 Hz depending on the applied load level and the load regime to control. The whole experimental campaign was conducted under load control with the intention to cover the typical high-cycle fatigue range between10,000 to 1,000,000 cycles for all test load setups. The load regimes chosen to be tested were: • Fully reversed push-pull loading • Fully reversed torsion • Fully reversed in-phase (IP) push-pull and torsion loading • Fully reversed out-of-phase (OP) push-pull and torsion loading with a phase shift of 90 degrees. All load regimes were run under a load ratio of R = -1. The ratio between applied axial stress and shear stress for the multiaxial load regimes was chosen to be 1.73. The experiments which were run to one million cycles or more were claimed to be run-outs. When a specimen achieves the established run-out stage the experiment was stopped, and then the same load regime was applied at a stress level increased by factor of 2.0 at least. 2.3. Results The obtained results of the experimental analysis can be found graphically depicted in Fig. 2. The regression analysis uses the Kohout- Věchet (K&V) formula, Kohout and V ěchet (2001): = ∙ [ ( + ) + ] (1) with four regression parameters a , B , C and  It was adopted to show the individual trends, because it also describes well the transition to horizontal asymptotes at the quasi-static and fatigue limit regions. Its adoption allows us to deal later with reasonable representations of the general fatigue trends also in the regions, where the common power law would fail. The power law with a single slope could not cover the lifetimes above 200,000 cycles otherwise. Apart from the K&V formula parameters, Table 2 gathers the following statistical parameters of regression quality: R 2 coefficient of determination, s logN standard deviation of logarithms of cycles and s S standard deviation of stresses. The substantially larger s logN parameter for the in-phase multiaxial loading is caused by the nature of the K&V regression, which must be run towards stresses and not towards numbers of cycles due to the transition to the horizontal asymptotes. Here, the s S parameter better represents the obtained quality of regression. Coefficient of determination R 2 shows that all obtained K&V curves represent the observed data well.