Strains in Shallow and Deep Notches Using Two DIC Algorithms

Abstract

In the current work, a high magnification 3D stereoscopic digital image correlation (DIC) system and a 3D micro-stereoscopic DIC system were used to determine elastic and elastic-plastic strain distributions in the proximity of shallow and deep notches. Specimens used were a plate with a shallow circumferential notch under bending load, and two deep U-Notch specimens under tensile load. The data analysis employed two methodologies. One was the direct application of a conventional and commercial DIC analysis tool. Another, applied to the same gathered data, using a novel combination of image analysis technique with a meshless numerical method. The Scale Invariant Feature Transform (SIFT) computer vision algorithm was used to extract distinctive points or features of the captured images. Then, the displacements were obtained by tracking the positions of successfully matched SIFT points of the undeformed-deformed pair of images. In sequence, the points provided by SIFT were selected as nodes of a meshless formulation to generate a numerical approximation of the displacement fields and their derivatives. For that, the moving least square method was employed. The advantages of the novel proposed method were the automaticity in the process correlation and the good performance when low and high strain gradient strain fields co-existed in the same analysis. The SIFT-Meshless strain distributions consistently matched with those obtained from the traditional DIC subset-step type of analysis applied to the same sets of images and with results obtained from finite element models. An important conclusion was that the SIFT-Meshless method provided reliable measurements, especially at the vicinity of the notches where the maximum strain values were expected.