PSI - Issue 75

Jan Papuga et al. / Procedia Structural Integrity 75 (2025) 289–298 Author name / Structural Integrity Procedia (2025)

297

9

3. The two longest lifetimes result in significant scatter manifested by low R 2 . Use of the factor 1.4 mentioned above to obtain the fatigue strength for the torsion curve would again cause the tension and torsion curves to overlap, but the bending curve is much lower. The best fit dependency for each lifetime is used to predict the local fatigue strengths for all specimen configurations. Its results are provided in the first row of Table 5. The comparison with RSG and TCD methods here is not fair in such a case, because all available experimental data were used to generate the output, while RSG asks only for the S-N curve of the unnotched specimen in push-pull, and the TCD methods for one more curve for one notched specimen. To get closer in comparison of the output quality of CV approach with TCD and RSG without exaggerated demands on the number of data inputs, the dependency between the local fatigue strength amplitude and the critical volume was defined as a simple power law from cases with the smallest and largest critical volumes in the case of the bending load mode. This load mode was deliberately chosen, because its position in Fig. 3 left is the lowest one and thus its use will generate intrinsically conservative results. Additionally, rotating bending loading remains the most economical way of generating the S-N curve and even medium-sized companies can build and operate the Moore machine with minor investments. For this analysis, the most typical ratio of z =0.9 was chosen, and the results are provided in the second line of Table 5. It is apparent that the best performing RSG methods result in better quality of estimates, but even this basic version of CV approach application provides results clearly comparatively better than any variant of the TCD method.

Fig. 3. CV vs local fatigue strength at 200,000 cycles and z factor of 0.85 (left). Coefficient of determination of the power law regression of this dependency for various number of cycles (right).

Table 5. Statistics of ERRσ relative errors of chosen CV solutions using the von Mises stress (100,000-500,000 cycles).

Variant

Average Maximum Minimum

Range 59.6% 56.9%

Standard deviation

From best fit

-0.9% 2.0%

30.0% 28.0%

-29.6% -28.9%

16.4% 16.8%

From bending mode z =90%

5. Conclusion The unique data set comprising of fatigue experiments with 7 different configurations of notched and unnotched specimens from ČSN 411523 structural steel tested in 3 different load modes (push-pull, torsion and plane bending) was analysed using multiple variants of the relative stress gradient approach, the theory of critical distances in point and line variants and newly proposed critical volume approach. The obtained results show that: • Addition of two more notched configurations with very sharp and very blunt notch magnified differences in quality of results delivered by different calculation methods if compared to the original paper [2].