A Multiscale Framework for Predicting Additively Manufactured Alloy Deformation

Thesis Abstract

Additive manufacturing (AM) of metals enables rapid customization but introduces complex deformation behavior across multiple length scales that challenges predictive reliability for engineering components. This study addresses the problem of accurately predicting deformation of AM alloys by developing a multiscale framework that integrates microstructural physics with macroscopic constitutive modeling to capture the interplay between laser-material interactions, solidification-induced grain structures, and post-build thermal histories. The aim is to develop a validated multiscale predictive tool capable of forecasting residual stresses, distortions, and local plastic deformation in AM builds to inform design and process optimization. Specific objectives are to (i) characterize the microstructure evolution of a nickel-based superalloy (IN718) during laser powder bed fusion using in-situ and ex-situ techniques, (ii) calibrate a hierarchy of scale-bridging models, including a crystal plasticity finite element (CPFEM) model for grain-scale deformation, a mesoscale phase-field model for defect and grain boundary evolution, and a thermomechanical finite element model for macro-scale residual stress and distortion, (iii) couple these models through transfer quantities such as texture descriptors, dislocation density, and phase fractions, and (iv) validate the integrated framework against experimental measurements of residual stress, distortion, and hardness from a controlled set of builds. The study adopts an explanatory sequential mixed-methods design. The population comprises commercial-grade IN718 powder and additively manufactured coupons produced on a standardized laser powder bed fusion platform. A sample of 60 coupons is prepared under four distinct process parameter sets (laser power, scan speed, hatch spacing, and layer thickness) to generate a broad deformation response. Data collection combines quantitative techniques—neutron diffraction for residual stress profiling, 3D coordinate measuring machine (CMM) for distortion assessment, electron backscatter diffraction (EBSD) for grain orientation and texture, nanoindentation for local hardness—with qualitative process-condition metadata. Instrument validity and reliability are established through calibration standards, repeat measurements (n=3 per coupon), and cross-validation against independent measurements reported in the literature. The CPFEM sub-model is calibrated using crystallographic slip systems and single-crystal plasticity parameters drawn from published alloy databases, while the phase-field sub-model uses diffusion coefficients and interface energies sourced from high-resolution microscopy studies. The macroscopic model employs a viscoplastic constitutive relation with temperature-dependent yield criteria and incorporates thermal histories from finite element process simulations. Analytical methods include multiscale data fusion to map microstructural descriptors to macro-scale fields, Bayesian calibration to quantify parameter uncertainty, and regression analysis to identify key drivers of distortion. The CPFEM simulations (up to 1.2 million elements per run) are validated against EBSD-derived slip activity indicators and neutron diffraction residual stresses. The study uses ANOVA to compare deformation metrics across parameter sets and nonparametric Spearman correlations to examine relationships between grain texture and local yield strength. The conceptual framework integrates the principles of crystal plasticity theory, phase-field modeling of solidification, and thermo-mechanical consolidation theory, with explicit reference to established theories such as Taylor-type work hardening and Hall-Petch grain-size strengthening to interpret results. Expected findings include a robust mapping from process-induced microstructure to macro-scale deformation, with quantified contributions of grain texture, defect density, and residual stress evolution to overall distortion. It is anticipated that the multiscale framework will reduce predictive error of final part geometry by at least 25% compared with conventional single-scale models and provide actionable insights into optimal process parameter windows that minimize distortion and residual stresses. The study contributes to knowledge by operationalizing a transferable multiscale framework that can be adapted to other AM alloys and bed configurations, advancing theoretical integration between crystal plasticity, phase-field, and thermo-mechanical modeling in the AM context. The main conclusion is that accurate deformation prediction in AM alloys requires coherent coupling across scales and explicit incorporation of process-structure–property relationships. Recommended directions include extending the framework to incorporate in-situ monitoring data for real-time model updating and exploring machine learning surrogates to accelerate predictive cycles for industrial design optimization.