Exploration of digital volume correlation in nominally homogeneous metals

The ability to conduct digital volume correlation (DVC) in nominally homogeneous metals using X-ray computed tomography (XCT) is examined by a combination of synthetic and experimental techniques. DVC requires markers to track within the volume and we first discuss methods by which the grayscale variations due to inherent inhomogeneities in nominally homogeneous metals can be enhanced in X-ray scans. Using these enhanced scans we then validate the predictions of the DVC algorithm via a combination of synthetically imposed rigid motions and complex displacement fields. The rigid body motions are captured very easily as the fields are dominated by motion of the specimen boundaries where a high grayscale contrast exists between the metal and air. The local deformation fields where there is no boundary motion require high quality tomographic scans, and we explore the range of hyperparameters that give high fidelity predictions. The study is then extended to real displacement fields obtained from experiments. A key conclusion is that hyperparameters optimised by imposing synthetic displacement fields are often inappropriate. This is because the changes to the image grayscales that occur due to the actual deformation of metals are different from those assumed in the algorithms used to impose synthetic deformations. Using a combination of different specimen geometries and known physical behaviour of the specimens, we reoptimise the hyperparameters in a global DVC algorithm that naturally uses boundary information. Finally, we also explore “hardware” methods to improve the DVC predictions. Our results show promise and suggest routes for further improvements.