PSI - Issue 28

F. Conrad et al. / Procedia Structural Integrity 28 (2020) 2195–2205 F. Conrad, A. Blug et al. / Structural Integrity Procedia 00 (2019) 000–000

2204 10

reproduced by looking at the shear strains within the FEM. Compared to DIC (blue circles in Fig. 8, right), the peaks differ in magnitude and size as expected because of two effects: On the one hand, a linear-elastic material behaviour was considered leading to an overestimation in the FEM simulation due to missing plasticity. On the other hand, the resolution of both evaluations differs. The FEM mesh uses an element size of 2.5 µm at crack tip while the length � for the DIC derivatives in equation (5) is 200 µm due to ROI spacing and Sobel kernels given in equation (4). However, the position of the peaks – the crack tip location - is clearly and uniquely defined by utilizing this criteria in both FEM and DIC. 6. Conclusion The aim of the work presented here is to turn the GPU-based DIC-system described in Blug et. Al. (2019), which was primarily designed for fast and marker-free strain-controlled applications, into a more comprehensive system for crack-growth investigations. In particular, it should be a single sensor being able to do both strain measurements, strain-controlled ones with high frame rates of at least 850 Hz and full-field evaluation for crack growth investigations at least once per cycle, in a single sensor. This is enabled by combining a fast CoaXPress camera with path independent and marker-free image correlation on the GPU. In the strain-controlled case, where four ROIs per image are required, the frame rate of the camera in the range of 1 kHz limits the measurement rate of the system whereas the GPU with a correlation rate of about 10 kHz defines the latency of about 2 ms (see Blug et. Al., 2019). For full-field evaluation under maximum load, the relevant images are selected in real-time and the evaluation with thousands of ROIs is evaluated only a few times per cycle. In this case, the evaluation time is limited by the GPU. For the biaxial strain field with 9000 ROIs, evaluation rate was 74 kHz per ROI corresponding to 121 ms per image. This is sufficient to do one full-field evaluation per load cycle in real-time, e.g. live images of the crack contour. This article reports first full-field results of uniaxial and biaxial crack-growth investigations. For uniaxial testing of CC-specimen, crack tip position was measured in two ways: the first one was thresholding the divergence of the displacement field, which effectively measures crack opening. The second one was localizing maximum rotation of the displacement vectors at the crack tip as maximum gradient in the rotation of the displacement field. Both measures agree well with crack growth measured by ACPD, but the value of the threshold is load dependent. Therefore, the maximum gradient in the rotation is preferred as a direction-independent measure for crack tip position. Within the biaxial measurements, the crack contour was measured by fitting a polynomial to the divergence of the displacement field. Along this crack contour, the gradient of the rotation of the displacement field looks very similar to the uniaxial case. However, its applicability to localize crack tip in biaxial results has to be proved in future experiments. So far, the crack contour from DIC is transferred to the mesh of the linear-elastic 2D FEM analysis to simulate displacement and shear strain under the load applied to the cruciform specimen. For the displacement, a very good agreement was found, i.e. a crack opening of 19.6 µm in the numerical model and 21.0 µm within the DIC measurement. Within shear strain, a qualitative agreement was found but the height of the poles differ due to linear elastic model on the one hand and due to the difference in spatial resolution for the calculation of the derivative (2.5 µm in the FEM mesh and about 200 µm in the DIC kernel) as expected. However, the position of the peak can be used in both FEM and DIC situation to identify the crack tip in a robust way. For all full-field measurements presented in this article, the Qioptiq lens with an image resolution of 4.5 µm/pixel was used, because the 8.5 µm/pixel of the telecentric lens were not sufficient for marker-free DIC. A comparison of zero-strain experiments on a large number of metal surfaces ranging from polished steel to cast aluminum showed that a resolution of about 5 µm/pixel is preferred to achieve marker-free and high-resolution full-field results. If this resolution is applied to a field-of-view of at least 7.5 x 7.5 mm², as it is required for strain-control with a base length � of 10 mm, the theoretical minimum for the image size is 2.25 MPixel. Therefore, the bandwidth of the CoaxPress camera of 2.1 GB/s still limits the frame rate to about 933 Hz. For available camera models, the achievable frame rate it is much lower. Therefore, a new CoaxPress 2.0 camera transferring more than 3 GB/s is currently implemented to combine strain-control and full-field evaluation in a single sensor. The good agreement between DIC and FEM analysis will be used as a basis to measure additional crack tip parameters such as J-integral, stress intensity factor (SIF) or crack opening displacement (COD) in a similar way in