Fig. 2 Two images of the same face of the sample with and without residual stresses. The square dot pattern glued on the post served for calibration purposes. The white arrows point out the deformation of the sample. Due to the large amount of the frames captured it was possible to choose two closely corresponding views of the sample before and after the needles were removed and to match them via DIC. Image processing was performed by considering a measurement grid of about 2000 evenly spaced points for each face. DIC was performed by using template and analysis subset sizes of 21¯ 21 pixels2 and 41¯ 41 pixels2, respectively. Only points pairs matched with a correlation coefficient > 0.98 were considered for subsequent analysis. Eight couples of stereo-systems were considered for reconstructing the sample lateral surface all over 360° (4 for edges and 4 for the faces). The accuracy of the data overlap between the contiguous reconstructed surfaces (of the order of 10-2 mm2) made not necessary to consider a larger number of stereo-views. Figure 3 shows the data point cloud obtained for the reference (stress-free) configuration. In this case, it was possible also to reconstruct the top and the bottom of the arterial segment. Being these two surfaces load-free they don’t need to be contoured also for the deformed configuration. Point data clouds so obtained were used as input to build the FE model. 2.4. Numerical reconstruction of residual strain fields Based on the experimental shape and displacement measurements described above, a finite element (FE) model is built in order to obtain the three-dimensional residual strain field as well as an estimation of the residual stress field within the piece Fig. 3 Point data obtained from the ‘whole-body’ DIC contouring of the stress-free configuration. The use of a single ‘cylindrical’ dot calibration pattern glued on the post (see Fig.2) allowed to automatically reconstruct and merge the point data of the 8 stereo-system in a single reference system. 36
RkJQdWJsaXNoZXIy MTMzNzEzMQ==