Development of a global digital image correlation approach for fast high-resolution displacement measurements.

摘要

Digital Image Correlation (DIC) is one of the non-contact full-field measurement techniques widely used in experimental mechanics due to its simplicity and low experimental costs. The abundance of knowledge provided by the full-field measurement, has led to emerging applications in different areas of research. These applications include, but not limited to, mechanical properties identification, verification of numerical simulations, strain mapping and pinpointing local phenomena (e.g. onset of plasticity or cracks). DIC is based on establishing spatial relations between two digital images acquired from the surface of a specimen in two different deformation states. Furthermore, if volume images are acquired using X-ray tomography, for instance, the same image correlation method in 3D can be used to correlate the 3D texture within the volumes. The latter method is called Digital Volume Correlation (DVC) and leads to the measurement of internal strains, thus further extending the application range of the technique.;The earliest approach of DIC was based on the successive correlation of image subsets. The resulting motion vectors were thus obtained independently from each other, both in terms of computation and the kinematic basis. This nonconformity of the deformation mechanisms among neighboring subsets, however favorable for low-cost computations, may downgrade the measurement reliability, especially at high resolutions. This issue justified the development of global approaches, where one deformation mechanism was sought for the whole region of interest using a large number of Degrees of Freedom (DOF) and through correlating the whole region once and for all. However, the implemented algorithms became computationally expensive, especially at high resolutions. The problem became even more prominent for DVC measurements, in which a huge amount of data should be treated.;The main purpose of this thesis was to develop and validate an improved global approach that can reconcile the accuracy and computational efficiency needed for high-resolution measurements in 2D and 3D.;First, a global DIC approach based on Fourier decomposition for the sought displacement field was adopted and subsequently improved. The improvement consisted firstly in modifying the Fourier-based kinematic to achieve more rapid convergence and lower uncertainties for a wider class of displacement fields. Also, a special strategy was developed for properly correcting for the edge effects stemming from periodic basis functions. The developed algorithm was tested using computer-generated experiments, the results of which proved the functionality of the introduced modifications. Furthermore, it was fairly compared to the local approach. It was shown that the improved spectral approach outperformed the subset-based method in equal conditions due to the global framework as well as the high capacity of the method in estimating complex displacements.;Second, potential applications of the developed approach were investigated for composite materials at micro-scale. Scanning Electron Microscopy (SEM) images from the surface of a Fiber-Reinforced Polymer (FRP) at micro-scale were taken from the literature. Suitable measurement parameters were identified using a priori measurements and evaluations on artificial experiments, i.e. Finite Element (FE) simulated displacement field of the exact microstructure artificially applied to the SEM images. The study also took the effect of the recorded noise into account. The evaluation on the simulated experiments proved the high potentials of the approach in measuring full-field strains in the fiber scale. Comparing the obtained experimental results with those previously reported for the same experiment in the literature further revealed this capability.;Finally, the developed global approach was extended to 3D leading to an improved spectral DVC approach. The potentials of the developed DVC technique were demonstrated by several artificial experiments simulating the volume images of composites at particle-scale recorded before and after deformation. Proper imaging resolution for accurate capturing of strain heterogeneities was estimated as a function of particle size and displacement resolution using a useful analogy presented in the 2D approach.;The main contribution of this thesis was the development of a global approach for accurate measurement of high-resolution full-field strains in 2D and 3D. This approach is very promising for strain mapping at micro-structural levels since it reconciles low computational costs and high sought resolutions. Especially, there are very limited approaches in 3D by now that enable measurements at this level of accuracy.