Refined extraction of crack characteristics in large-scale concrete experiments based on digital image correlation

The accurate extraction of the crack patterns and measurements of crack kinematics are essential for understanding the mechanical behaviour in experiments on structural concrete as well as in the validation and further development of sound mechanical models. This paper presents important refinements of the authors' recently published automatic crack detection and measurement procedure (ACDM) based on surface displacement measurements obtained with digital image correlation (DIC). The proposed refinements are crucial for reliably assessing the crack behaviour in large-scale experiments with complex crack patterns, since the original methods of ACDM may fail or result in biased measurements at locations with closely spaced cracks, crack intersections or cracks with high morphological curvature. The main refinements are (i) a Canny edge-based crack detector, which is applied on the DIC major principal strain field and (ii) enhancements in the crack kinematic measurement to assess the reliability of the results. The latter includes the automatic selection of optimum reference points used in the crack kinematic measurement to increase its reliability and remove uncertain results. The refined ACDM procedure is validated using several large-scale 2.0 × 2.0 m shear panel experiments with highly complex crack patterns. Compared to the original ACDM, significantly thinner cracks can be detected with a much higher reliability of crack locations and crack kinematic measurements, particularly close to crack intersections and at closely spaced cracks. Additionally, two approaches for the statistical consolidation of the large amount of gathered data into characteristic crack properties in large-scale homogeneous concrete element experiments are proposed and compared. The results show that the statistical consolidation of the ACDM data using a 95%-quantile match well with the direct extraction of the best-fit homogeneous crack properties from the full-field DIC displacements. The consolidated data provides highly valuable insight into the mechanical behaviour, especially regarding crack phenomena.ISSN:0141-029