PSI - Issue 39

Hannes Panwitt et al. / Procedia Structural Integrity 39 (2022) 20–33 Author name / Structural Integrity Procedia 00 (2019) 000–000

31

12

linear strain threshold value was used. However, a comparison with the a-N- curve obtained by the crack width method shows an underestimation of the crack length for the front side at high crack lengths (detail in Fig. 8a). This indicates that for the front side and the main crack the constant strain threshold value of th ε1 = 0.4 % is more suitable, as it provides the same, but less noisy a - N -curve as the crack width method (detail in Fig. 8b). In addition to the a-N- curves, the kinking angles φ 0 are also obtained using the extended script. Therefore, the orientations of the branches calculated by the original ACDM-script are aligned according to the definition of the kinking angle (Fig. 9a). The results for the front and back side of the specimen are shown in Fig. 9b. While the crack on the back grows with a nearly constant angle of 25°, the kinking angle on the front side is reduced with an increase in crack length and number of cycles.

50

y [mm]

∆ a n

25

∆ a 1

Front Back

ϕ 0, n

0

[°]

0

Notch tip

-25

ϕ 0,1

x [mm]

-50

Pre-crack

0

2

4

6

8

10

12

14

16

N [-]

10 5

Fig. 9: Development of the kinking angle under in-phase mixed mode loading: a) Definition of the kinking angle; b) Comparison of the front and back side.

5. Discussion The automation tool developed by Gehri et al. (2020) provides a very good basis for minimizing the high manual efforts required for the evaluation of DIC data in terms of crack detection. The enhancements presented in this paper also enable the extraction of both accurate a - N and associated kinking angle progressions. In addition, the program code now provides the capability for quantitative evaluation of branching cracks. Overall, the extended tool is thus a helpful instrument to investigate even fatigue cracks under mixed mode loading. However, it should be noted that the method for determining the crack path with morphological thinning may lead to false positive crack paths, if the final HSA has a shape with holes or spikes alongside the true path of the crack. These may be caused by an inhomogeneous speckle pattern or noisy correlation data. Furthermore, the method may lead to the detection of a single crack path, when two cracks are located close to each other. These effects on the accuracy on the crack path detection are discussed in detail by Gehri et al. (2020), but with proper DIC preparation and post processing the crack path detection is sufficiently accurate for the investigations in this paper. Moreover, it has to be mentioned that the crack path detection becomes more difficult, if the HSA is large in front of the crack tip. Therefore, a careful strain threshold value calibration has to be conducted to avoid crack patterns like in Fig. 6c. As the original ACDM method was developed for brittle materials such as concrete with small local deformation before fracture, the newly presented method also works best with small local deformations i.e., if linear elastic fracture mechanics is applicable. As the ε1 -method uses the same threshold values as the crack path detection, it is also reliant on a careful threshold value determination. Depending on the development of SIF at the crack tip an accelerated crack growth may occur, especially under constant force-controlled loading. Unfortunately, the crack growth rate can be different on both sides of the specimens even within the same specimen. To account for this, the increasing threshold value calibrated on a constant F max -controlled mode I test can be utilized for all tests to obtain conservative results for the crack lengths (as can be seen in Fig. 8). On the back side of the specimen shown in Fig. 8, this leads to an increasing underestimation of the crack length up to 5% at the end of the test compared to the crack width method and marker load technique, due to the almost constant crack growth rate.