PSI - Issue 76

Xabat Orue et al. / Procedia Structural Integrity 76 (2026) 3–10

9

As can be seen in Fig. 4, the experimental results (red dots) are very aligned in the range of 10 ÷ 100 µm, as suggested by Murakami with a slope of 1/6. Its fitting results with a material dependent position factor ( C ) of 600 for internal defects ( Y = 0,50) that were considered with equation (7) are very good. Chapetti’s model considering an α lath width of 11 µm according to (8,9) presents good results too, especially in the case of H-H direction. In this direction, El Haddad’s model with equations (2,3) is slightly conservative, and the original K-T diagram with (1) slightly unconservative. In the case of the H-V direction on the other hand, all models are too conservative above 34 µm of √ . To draw further conclusions related to the directional issue, an in-depth metallographic study is required, where the propagation direction of fatigue cracks of HCF specimens would be evaluated. Likewise, due to the zero-defect parameter optimization carried out [2], no defects were found above 100 µm, which is promising compared to the literature [21]. However, more data above this range is needed to properly characterize the long crack regime and draw better conclusions about the convenience of each model. For this purpose, artificial defects would have to be induced by EDM, which is out of the scope of this paper and will be considered for future works. Based on the results obtained, it could be stated that the interaction between defects and microstructure is not negligible. However, these results depend on certain aspects that are worth mentioning: • The intrinsic fatigue limit of the material free of defects is the parameter that determines the upper bound of the K-T diagram. Here, it was considered a value of 675 MPa based on the relationship (4) with the mean Vickers hardness along the height of the microstructural samples (381,35 kg/mm 2 ) and the Walker’s model with (5,6) for R = 0,1. However, this parameter might be unrealistic, as all materials contain defects, especially those obtained through AM technologies. Consequently, the upper bound of the K-T diagrams presented might not be representative of the material obtained and a more precise estimation is needed. • The influence of mean stresses is considered through different models when R ≠ -1. Here, Walker ’s model with expression (5) was considered for R = 0,1 of the tests presented in this paper, where α = 0,26 according to expression (6) based on the mean Vickers hardness of 381,35 kg/mm 2 . This is consistent with some preliminary tests carried out at R = -1 for the same batch (not shown here). Besides, it is consistent with Murakami’s approach and thus the differences between models are reduced. However, a more detailed testing campaign at different stress ratios is required to evaluate the influence of mean stresses and determine an appropriate model for the material obtained. • The knee-point ( N k ) in a S-N curve is representative of the fatigue limit, i.e., the number of cycles where the region of finite life ends and infinite life begins. Consequently, it should be considered for the extrapolation of the fractures of HCF tests in K-T diagrams. The fatigue limit range obtained with the up-and-down method according to Dixon (598,5 MPa) is influenced by defects. Consequently, the results present a broader dispersion than the intrinsic of fatigue phenomenon, and the corresponding knee-point (2,18·10 7 cycles) was not considered for the extrapolation of fractures of HCF tests. In absence of a clearly defined N k in the S-N curve (Fig. 3) and with conservative purposes, 50 million cycles were set according to the design procedure for non-ferrous materials of helicopters [19]. However, more HCF tests are required to identify N k , extending beyond the established 10 million cycles for the run-out. • The microstructural characterization and the corresponding post-processing of images should be considered as preliminary. Here, t he length of α -laths was considered as a dimensional characteristic of the microstructure, but more samples and images are required to identify a more representative parameter of the microstructure obtained, despite achieving good fitting results with Chapetti’s microstructural model. Likewise, a directional distinction is needed to identify a microstructural parameter in each direction and justify the different FCG behaviors that were observed. 4. Conclusions In this work, K-T diagrams have been obtained with models based on defects and microstructure for fatigue assessment of Ti-6Al-4V manufactured by DED-LB/CW. For this purpose, microstructural analysis, as well as Vickers hardness, tensile, HCF and FCG tests were undergone. To minimize the effect of process defects, a set of optimized parameters was employed following a zero-defect philosophy. Consequently, the interaction between microstructure and defects is highlighted, and microstructure-based models take more relevance. Furthermore,