PSI - Issue 33

Sabrina Vantadori et al. / Procedia Structural Integrity 33 (2021) 773–780 S. Vantadori, C. Ronchei, D. Scorza, A. Zanichelli / Structural Integrity Procedia 00 (2021) 000–000

778

6

- For infinite life fatigue tests under biaxial loading (Fig. 2(b)), the above criterion correctly captures run-out conditions, whereas it does not estimate the experimental failures, but rather a condition of incipient failure; - For finite life fatigue tests (Fig. 3), the estimations (except one, are totally in agreement with the experimental outcomes.

50

75

150

(a)

Failure Run-out R=

-3 -1 0 0.5

(b)

Failure Run-out R=

-1 0

(c)

Failure Run-out  =

0° 90°

125

50

100

25 SHEAR STRESS AMPLITUDE, C a [MPa]

75

25

50

25

0 300 EQUIVALENT NORMAL STRESS AMPLITUDE, N eq,a [MPa] 0 100 200

0 300 EQUIVALENT NORMAL STRESS AMPLITUDE, N eq,a [MPa] 0 100 200

0 300 EQUIVALENT NORMAL STRESS AMPLITUDE, N eq,a [MPa] 0 100 200

Fig. 3. Fatigue strength evaluation for finite life fatigue tests under: (a) tensile, (b) torsional and (c) biaxial loading.

Regarding finite life fatigue tests, the number of loading cycles to failure is theoretically determined only for the failed specimens, and the comparison between experimental, exp N , and estimated, cal N , fatigue life is plotted in Fig. 4. Independent of the examined loading condition, the agreement between experimental data and estimated fatigue lives is quite satisfactory, being 63% and 97% of theoretical results included into scatter band 2x and 3x, respectively.

10 7

Tension Torsion Biaxial

10 6

10 5

N exp [cycles] 10 4

Scatter band 2x Scatter band 3x

10 3

10 3

10 4

10 5

10 6

10 7

N cal [cycles]

Fig. 4.

exp N vs

cal N for finite life fatigue tests.

Finally, the results in terms of fatigue lifetime determined by means of the present criterion are compared with those derived through a well-known stress invariant-based criterion, that is, the Crossland criterion (Suresh (2008)). Such a comparison is performed by computing the value of the mean square error, RMS T , as was proposed by Walat and Łagoda (2014). From Fig. 5, it can be observed that the present criterion provides considerably better results than those deduced by employing the Crossland criterion, and this holds true independent of the applied loading conditions.