Effects of Intrinsic Material Length Scales and Grain Boundaries on Void Density Distributions of Shock-Induced Spall Damage: A 3D Tomography Study

The characterization of spall microstructural damage metallic samples is critical to predicting and modeling modes of failure under blast, ballistic, and other dynamic loads. In this regard, a key step to improve models of dynamic damage is making appropriate connections between experimental characterization of actual damage in the form of discrete voids distributed over a given volume of the specimens, and the output of the models, which provide a continuous measure of damage, for example, void fraction as a function of position. Hence, appropriate homogenization schemes to estimate, e.g., continuous void fraction estimations from discrete void distributions, are key to calibration and validation of damage models. This project seeks to analyze 3D tomography data to relate the homogenization parameters for the discrete void distributions, i.e., homogenization volume size and step, as well as representative volume element size, to the local length scales, e.g., grain size as well as void size and spacing. Copper disks 10 mm in diameter and 1 mm thick with polycrystalline structures were subjected to flyer plate impacts resulting in shock stresses ranging from 2 to 5 GPa. The spall damage induced in samples by release waves was characterized using X-ray tomography techniques. The resulting data is thresholded to differentiate voids from the matrix and void fraction is obtained via homogenization using various parameterization schemes to characterize void fraction distributions along the shock and transverse directions. The representative volume element is determined by relating void fraction for varying parameterized window sizes to the void fraction in the overall volume. Results of this study demonstrate that the optimal representative volume element (RVE) to represent void fraction within 10% error of the overall sample void fraction for this Hitachi copper sample is .2304 mm3. The RVE is found to contain approximately 255 grains. Statistical volume elements of 1300 µm3 or smaller are used to quantify void fraction as a function of position and while the results along the shock direction, i.e., the presence of a clear peak at the expected location of the spall plane, are expected, the void fraction along the transverse direction show oscillatory behavior. The power spectra and predominant frequencies of these distributions suggest the periodicity of the oscillations relates to multiples of local material length scales such as grain size. This demonstrates that the grain size in the samples, about 120 µm, is too large compared to the sample size to try to capture spatial variability due to applied loads and the microstructure, since the microstructure itself produces variability on the order of a few grain sizes. These results may play a role for the design of experiments to collect real-world 3D damage data for validating and enhancing the accuracy and definition of simulation models for damage characterization by providing frameworks for microstructural strain variability when modeling spall behavior under dynamic damage.