PSI - Issue 79

Martin Sladký et al. / Procedia Structural Integrity 79 (2026) 421–432

424

σ

σ

σ peak

σ n

Real notch

Fictitious notch

ρ *

x

x

2 α

2 α

ρ f = ρ + s ρ *

ρ

Fig. 1. Definition of the e ff ective notch stress by introducing a fictitious notch radius as an equivalent to stress averaging ahead of the notch root.

structures, Baumgartner (2017) observed that the corresponding FAT classes depend on the notch opening angle, which led Hobbacher and Baumgartner (2024) to specify distinct FAT classes for di ff erent crack initiation sites. Evaluating the results of the notch stress-based analyses, Sonsino et al. (2010) was among the first to report that the S–N curves of thin-walled welded structures tend to be flatter than those of thick-walled components, with the slope exponent increasing from approximately 3 to 5 under normal stress loading. Using a larger statistical dataset, Baumgartner et al. (2020) confirmed a consistent tendency for the slope exponent to increase as wall thickness de creases, despite considerable scatter among individual test series. These findings are reflected in the recommendations of Hobbacher and Baumgartner (2024), who specified an S–N curve exponent of m = 5 for structures under normal stress loading and wall thicknesses below 7 mm. This article aims to compare the fatigue life prediction performance of three distinct approaches across ten configu rations of thin-walled fillet-welded specimens featuring diverse geometries and loading conditions. The experimental dataset employed for this purpose comprises literature-sourced specimen configurations as well as two configura tions tested directly by the authors from the Czech Technical University in Prague. The subsequent analysis provides further insights into fatigue life prediction for thin-walled fillet-welded structures and proposes several refinements concerning their practical application. The experimental dataset analyzed in this study consists of constant-amplitude fatigue test results for ten configu rations of thin-walled fillet-welded specimens, as shown in Figure 2. The principal characteristics of these specimen configurations, encompassing various geometries and loading conditions, are summarized in Table 1. All specimens were tested in the as-welded condition, and crack initiation was predominantly observed at the weld toe. Fatigue tests conducted by the authors from the Czech Technical University in Prague, comprehensively described by Machacˇ et al. (2022), involved two circular hollow section (CHS) based configurations fabricated from structural steel AISI 4130-N. The first, a CHS lap configuration, was fabricated by fitting a CHS tube into two machined hollow sections, whereas the second, a CHS fillet configuration, corresponded to a representative CHS-CHS H-type joint. Both configurations were welded using the TIG process. Fatigue testing was conducted at a frequency of approxi mately 130 Hz for the CHS lap specimens and between 75 Hz and 90 Hz for the CHS fillet specimens. The newly generated experimental results were evaluated alongside data from two CHS-Plate fillet configurations reported by Jiao et al. (2013) and six additional configurations presented by Gurney (1997), which encompassed both rectangular hollow section (RHS) based joints and simple plate-based fillet weldments. The CHS-Plate specimens, consisting of very high strength (VHS) steel tubes fillet-welded to the center of thick steel plates using the Gas Tung sten Arc Welding (GTAW) process, were tested at a frequency of 35 Hz. The six remaining configurations included two RHS-RHS T-type joints and four simple plate-based configurations: Plate fillet T Ben, Plate fillet T Ten, Plate RHS fillet, and Plate fillet L, all fabricated from low-strength steel using the MIG welding process. Most of these 2. Methods 2.1. Experimental Dataset