PSI - Issue 42

T. Koščo et al. / Procedia Structural Integrity 42 (2022) 1600 – 1606

1603

4

Koščo, T., Chmelko, V. / Structural Integrity Procedia 00 (2019) 000 – 000

Results from partial view (Fig. 1b) has been compared with finite element solution for 2D plane strass and for 3D case as well. As can be seen on Fig. 2 3D FEM solution results slightly different strain distribution on the cylindrical part of the path. In both cases the contour has been smoothed using 7x7 local regression, that doesn’t affect the strain results significantly. Smoothing spline has been performed with grid reduction factor of 2.5.

Fig. 2. Comparison of two smoothing algorithms with different parameters for partial view, (0 to 5mm of the path is on the cylindrical surface, 5 to 12 mm is on the planar surface)

Partial view (Fig. 1b) reported problems with the filtering on the edge between the planar surface and the cylindrical shape of the hole. Strain function must be continuous. The strain function (component in the direction of loading) is relatively constant with zero gradient on the cylindrical surface in the root of notch (see Fig. 2). On the planar surface, there is significant gradient. This gradient difference on the edge causes severe problems for both filters. Smoothing spline tends to minimize the infinite curvature on the edge and therefore lowers the values in the notch close to the edge. Local regression exhibits the same problem. Therefore, only values from cylindrical shape are relevant and comparable with analytical or numerical solution. Strain component in the direction of loading along the grey path (see Fig. 1b) is on Fig. 2. Smoothing spline results steeper strain gradient by larger curvature restriction, what results from the nature of the method. On the other hand, local regression with larger kernel sizes (21x21 and larger) tends to converge to the FEM solution further from the notch edge. Mean of the strain in last 3 mm of the observed path with 21x21 kernel size resulted 1.255*10 -3 what equals to stress concentration of 3.23 and absolute error with respect to the analytical solution of 3.1*10 -5 (6.4 MPa) and 31x31 size resulted 1.247*10 -3 with stress concentration factor of 3.2 (absolute error of 2*10 -5 , 4 MPa in the linear elastic region). Therefore, authors came with another configuration to fully utilize the strength of the 3D DIC and measure only on the cylindrical surface (see Fig. 3). This led to the previously mentioned measurement with symmetric camera configuration. Results has been significantly less dependent on the smoothing parameters. Absence of the results on the planar surface does not affect the feasibility of the measurement on general notch. Result postprocessing on the smooth surface without edges is in better accordance with the smoothing methods.

Fig. 3. Symmetric view

Results do not vary significantly through the specimen thickness. Potential strain gradient is caused by additional bending. Inverse white on black speckle pattern has been applied on the specimen surface. Pixel/mm ratio in the notch root has been 40.9 and for 5mm radius its 204.5 pixels per notch radius. If the absolute measurement error remains constant, relative error varies with loading. Stress concentration factor for both cases has been calculated for 20 kN load step with 80 MPa of nominal stress and 252 MPa in the notch root calculated by using analytical solution by Howland (1930). Results has been filtered to obtain relative spatial