PSI - Issue 57

8

Lewis Milne et al. / Procedia Structural Integrity 57 (2024) 365–374 Lewis Milne et al. / Structural Integrity Procedia 00 (2019) 000 – 000

372

20Hz test data 20Hz Basquin Fit 20kHz Corrected - Average Discrepancy 20kHz Corrected - Bach Model (a)

200 220 240 260 280 300 320 340 360 380 400 1,E+04 1,E+05 1,E+06 1,E+07 1,E+08 1,E+09 Stress Amplitude (MPa) Number of Cycles to Failure 20Hz test data 20Hz Basquin Fit 20kHz Corrected - Average Discrepancy 20kHz Corrected - Bach Model (b)

320

300

280

260

240

220

Stress Amplitude (MPa)

200

1,E+04

1,E+06

1,E+08

1,E+10

Number of Cycles to Failure

Figure 8 - Comparison between corrected UFT SN curves and conventional frequency test data for (a) S355JR and (b) Q355B

From Figure 7, a general trend of increased frequency sensitivity with ferrite content is observed, agreeing with the observations by Bach et al. (2018). It is not possible to discern a stronger correlation than this, however, as there is significant variation between reported results. This scatter between different investigations stems from differences in test methods, with test parameters such as test frequency and specimen geometry varying between each investigation. Comparing the materials in the current investigation to the literature data, the results of the S355JR steel appear to match most closely to the steels with similar ferrite contents, whereas Q355B shows a much lower frequency sensitivity than the similar steels. It is proposed that this is due to the aforementioned size effects, as the conventional frequency specimens tested by both Bach et al. (2018) and Gorash et al. (2023) used cylindrical specimen geometries with larger risk volumes when testing at conventional frequencies, and as such, size effects were likely present. This suggests that the difference in specimen geometries used at different frequencies may be causing the frequency sensitivity of ferritic steels to be overestimated, again highlighting the need to use consistent specimen geometries at all tested frequencies. 9. Applying Comparison Methods As Correction Factors In addition to evaluating the frequency sensitivity of the tested materials, the two comparison methods were also applied to produce a corrected ultrasonic curve which is equivalent to conventional frequency results. If successful, this would provide a method of better comparison across different fatigue test frequencies, and allow conventional frequency curves to be extended beyond the traditional fatigue limit using ultrasonic data. For the Average Discrepancy method, the average discrepancy magnitude was simply subtracted from the ultrasonic curve. The resulting corrected UFT curve is plotted in red alongside the conventional 20Hz data in Figure 8(a) for S355JR and in Figure 8(b) for Q355B. It was observed that the corrected ultrasonic curve matched up well with the conventional frequency test results for S355JR, but not for Q355B. This is because the gradient in the Q355B SN curve observed at 20Hz was considerably steeper than that observed at 20kHz. As such, when the UFT curve is shifted down, it underestimates the fatigue life at high stress amplitudes, and overestimates the life at low stress amplitudes. It is therefore concluded that this model cannot be used to produce corrected SN curves in its current form. The Bach Model was also applied to produce a corrected low-frequency SN curve based on the UFT data. To achieve this, equation 1 was re-arranged to predict the conventional frequency stress amplitude at a given number of cycles to failure, based on the equivalent ultrasonic frequency stress amplitude and the average frequency sensitivity, ̅ . This re-arrangement is presented in equation 2, where , and , are the stress amplitudes and and are the test frequencies of the conventional and ultrasonic fatigue tests, respectively. , = , ( ) ̅ (2)