Predicting subgrain size and dislocation density in machining-induced surface microstructure of nickel using supervised model-based learning

Abstract

Microstructure transformations during severe shear deformation induced by large strain machining experiments are examined on high-purity (99.99%) nickel plastically deformed to strain (1–5) and strain-rate (102–104/s). Deformation conditions are created using plane strain machining (PSM). They are characterized via in-situ techniques, which are then juxtaposed with orientation imaging microscopy (OIM) via electron backscattered diffraction (EBSD), and the dislocation densities are quantified using the broadening of X-ray diffraction peaks of crystallographic planes. We parameterize the variation of microstructure response by measuring the subgrain size as a function of seven variables involved in the cutting process: cutting speed, rake angle, temperature, strain, strain-rate, Zenner-Holloman parameter, and a dependent rate parameter (R). This parametrization was based on supervised model-based learning. One of them used Principal Component Analysis (PCA) within a linear regression, which produces good predictions. We use the PCA model in addition to a FEM simulation of the PSM to predict the subsurface subgrain size. Furthermore, the Principle of Similitude (PS) is incorporated for predicting the dislocation densities in both the deformed chip and the machined subsurface. The proposed framework (FEM simulation-PCA model-PS relation) is shown to offer opportunities for creating multifunctional surface microstructure in an array of machining manufacturing processes.