PSI - Issue 28

Stepan Major et al. / Procedia Structural Integrity 28 (2020) 561–576 Stepan Major/ Structural Integrity Procedia 00 (2019) 000– 0

568 8

Fig. 4. SEM images of reconstructed fracture surfaces of virgin specimens, individual samples are described by loading ratio of applied loading forces: (a) L R = 0; (b); L R = 0.23; (c) L R = 0.5; (d) L R = 0.73; (e) L R =1.

Fig. 5. Selection of areas for reconstruction made from high-resolution images. Size of studied squares is ͲǤʹ ൈ ͲǤʹ mm 2 . The reason for using two different nets was endeavor for determination of appropriate edge length of elementary square. The finer network ( a 05 = 5 μm) was used only in some specific areas, due to the higher computational complexity. Although the entire surface can be reconstructed from a large number of images with great magnification, the exact binding of these images is very difficult. We can say that, if the magnification of individual images increases (and thus the number of images from which the total reconstruction is performed), the number of errors (i.e. probability of error) in the binding area increases. This is reason, why a grid with a square side a 99 = 99 μm was used in this study to cover the entire fracture surface. These lengths of the edges of the elementary squares were chosen to correspond to the macroscopic level and microscopic level of observation, resp. so as to match the difference between "Global" and "Local" surface analysis.