PSI - Issue 71

Haru Fujishima et al. / Procedia Structural Integrity 71 (2025) 18–25

21

Fig. 3 Shapes and dimensions of specimens (in mm) and stress gradients of specimens: (a) = 2 = 5 2 = 0.4 −1 , (b) = 2 + 2 = 1 2 0 + 10 2 = 0.4 −1 ǡ ƒ† ሺ ሻ = 2 = 10 2 = 0.2 −1 Ǥ

Fig. 4 Shape and dimensions of small artificial defect

The fatigue testing machine was a four-point RB fatigue testing machine. Two testing machines were used, one with a capacity of 15 Nm and the other with 100 Nm. The test frequency was either f = 1 Hz or 67 Hz. The nominal stress σ is the normal stress defined at the surface of the smallest cross-section of a specimen. The stress concentration factor for a deep-notched specimen (Fig. 3b) is K t = 1.139 (Nisitani et al., 1985). The nominal stress for this specimen was defined by the maximum stress amplitude multiplied by K t . The stress gradient for this specimen is the same χ = 0.4 mm -1 as for a smooth specimen with a diameter of 5 mm (Fig. 3a); therefore, if the nominal stress (maximum bending stress) is the same, the inward stress distributions near the surface of two specimens (Figs. 3a and 3b) are almost identical. The fatigue tests for χ = 0 were performed using a servo-hydraulic tension-compression fatigue testing machine at R = – 1 and f = 45 Hz. The nominal stress was σ = 210 MPa for all tests. The surface length of a crack emanating from a drilled hole as a function of the number of cycles was measured using the following replica method. A small piece of cellulose acetate film of 0.035 mm in thickness was put into methyl acetate solvent; then, the gelled film was put onto the specimen surface. After an appropriate time (usually a few minutes), the dried film was peeled off from the specimen surface. During a fatigue test, these replicas were taken regularly at the appointed number of cycles. Accordingly, the fatigue process was successively investigated on the surface of the collected replicas. The crack growth curve was obtained by measuring the length of a crack projected on the plane perpendicular to the tensile stress. The crack initiation life was defined as the number of cycles in which a crack first appeared at the hole edge.

3. Experimental results and discussion

The S-N data, as displayed in Fig. 5, shows that even for an identical shaped defect and the same nominal stress, there is a significant difference in fatigue life. Some experimentally obtained fatigue lives were close to the predicted S-N curve (predicted fatigue life: N f = 347,600 cycles), but others were far longer than the predicted life. The effects of stress gradient and test frequency are separately discussed in what follows.