PSI - Issue 39

Aljaž Litrop et al. / Procedia Structural Integrity 39 (2022) 41–46 Author name / Structural Integrity Procedia 00 (20 19) 000–000

43

3

DIC is a non-destructive measurement of surface strain. A digital camera is used to track the relative surface displacement between successive images as the sample deforms. There are many DIC analysis software programs, but most work on the same principle. The output or reference image is used to define the region of interest (ROI) from the entire field of view of the camera. ROI is then divided into small areas, referred to as subsets. If the surface color of the tested sample is homogeneous and glossy, such as the aluminum sample tested here, a random pattern must be applied. The pattern is shown in Figure 1.b. This random pattern ensures that each subset in ROI has a certain gray scale distribution level (usually gray scale images are taken). Random patterns on the surface of the shear samples were created by first applying a uniform layer of white paint and then using black paint to create a random speckle pattern. Both paints are spray cans with a matte finish that limits unwanted light reflections on the camera sensor. The main function of the DIC algorithm is to track the subsets through a series of images by optimizing the correlation criterion. Strain is then calculated as the relative displacement between subsets, as if the entire surface were covered with virtual strain gages. Ncorr is an open-source DIC analysis software widely used in strain measurement research. [16, 17] The subsets are circular and can overlap, controlled by the distance between the subsets, called the subset spacing. In the Ncorr software, the new position of the subset is sought by a combination of subset displacement and deformation, and both processes are controlled by the correlation optimization method. The optimization is done in two steps, first by normalized cross-correlation of the gray level function of the image and second by a nonlinear optimizer that refines the correlation at the subpixel level. The presence of cracks imposes a large discontinuity on the overlaying subsets. Therefore, the algorithm must be robust enough to handle these events when using DIC software to detect cracks. Ncorr enables the user to remove any subsets that fail to meet the correlation criterion, thus giving a clear indication of crack regions (can be seen in Figure 2 at N = 37 as lack of data in the central part of the specimen). The experimental DIC setup consisted of a single CMOS camera with a resolution of 0.5 MPx equipped with a 50 times magnifying lens. A dedicated white light source was placed above the camera at an angle to the sample surface to avoid direct reflection into the camera lens. The camera was mounted on a tripod with an additional extension arm that enabled camera positioning. An analog spirit level was used to bring the camera as close as possible to its natural horizontal orientation. 3. Experimental Results The DIC full-field strain measurement indicates cracked regions as high gradient strain, corresponding to the actual crack on the sample surface. The length of the crack was measured in px using image processing software and scaled to the 13-mm line, which is shown in Figure 2. The measurement is not precise because the camera resolution produces large DIC subsets that average the strain at the transition from crack to undamaged material. Short surface cracks cannot be detected due to the limitation of the DIC strain measurement. [18, 19] The DIC ROI must be slightly offset from the sample surface to avoid generating subsets that are not completely covered by a random pattern, as shown in Figure 2, where there is a clear gap between the sample surface and the DIC region of interest. In general, Figure 2 shows an overview of the entire life of the specimen. The DIC results clearly show the crack region due to the apparent strain gradient and loss of subset uniformity (subsets are cut-off due to poor correlation). [20, 21] With higher camera resolution, the DIC results could become even more sensitive to the discontinuity of patterns, as the currently used low-resolution camera introduces larger subsets that averages the boundary of strain gradient. [22, 23] All results are shown for specimens made of 6061 aluminum alloy. The results presented in Figure 3 show the measured crack length as a function of the number of cycles. For a given load amplitude (±0-3 mm), two cracks were followed, naturally named “left” and “right”. The left crack nucleates earlier but with a slightly lower growth rate as compared to the right crack. This asymmetry of crack growth could be attributed to several factors, such as inhomogeneity of material, surface defects that introduce stress concentration, un-parallel specimen positioning inside the clamping apparatus, etc.