The influence of anisotropic slip and shear transformation on heterogeneous deformation based upon nanoindentation, crystal plasticity modeling, and artificial neural networks

Most technological relevant structural materials are polycrystals that contain many (millions) so-called grains. Plastic deformation in polycrystalline metallic materials is not homogeneous at the microscale, but sometimes strongly heterogeneous among grains and varies spatially within an individual grain. The ability to predict the magnitude and spatial distribution of plastic inhomogeneity, particularly near grain boundaries (GBs), is crucial as it often results in a stress concentration that could lead to nucleation of damage sites, which largely affect the lifetime of stressed components. Accurate modeling of such inhomogeneity depends on the reliable description of basic plastic micro-mechanisms, e.g., dislocation slip, and fine details such as interactions between slip and GBs. This study aims to facilitate the above need by 1) establishing consistent critical resolved shear stress (CRSS) values that form the basis of a phenomenological description of plasticity, 2) enhancing effective metric selection for slip transfer across GBs, and 3) improving the understanding of the kinematics of the phase transformation—an important thermo-mechanical treatment that strongly influences the microstructure for many hexagonal engineering alloys such as Ti and Zr alloys.The first part of the present work is to quantify the uncertainty of initial CRSS values determined from Inverse indentation analysis (IIA). This approach optimizes the adjustable parameters in a chosen constitutive description of crystal plasticity until the load–depth response and the residual surface topography match between real and simulated nanoindentation(s) into a particular grain. IIA was evaluated for hexagonal pure Ti (CP-Ti) and two Ti alloys (Ti-3Al-2.5V and Ti-6Al-4V) at different temperatures (ambient and 523 K) and is found to produce consistent CRSS values when the combined relative error is no more than 20 %.A novel approach to evaluate the effectiveness of slip transfer metrics (individually) and their combinations using a double-layer artificial neural network (ANN) is presented as the second part of the thesis. The considered metrics include the misorientation angle between two grains, m′αβ , SFα +SFβ, ∆bαβ, and some of their compounds. The accuracy of binary (slip transfer or not) classification reaches around 90 % based on data collected from pure Al oligocrystals deformed in tension, and it is around 80 % for tensile deformed polycrystalline Ti-5Al-2.5Sn samples. This approach extends the one- or two-dimensional projections formerly applied to analyze slip transfer and can be implemented into crystal plasticity model as an “intelligent” decision-maker for each individual slip–boundary interaction.Lastly, a method to calculate orientation and deformation gradient variants resulting from phase transformation between hexagonal α and body-centered cubic β phase is proposed based on a series of frame rotations and transformations. Furthermore, a cluster-based approach is presented to automate point-wise reconstruction of β orientations from α orientations in a large indexed area. This work will assist in analyzing research problems that often require historical information of the current microstructures, for instance, understanding variant selections during the α → β → α transformation. The proposed method will also facilitate the implementation of such transformation kinematics into continuum-based models as the deformation gradient that influences the transformation can be conveniently computed.