Algorithm Development in Materials Science and Engineering: Multiscale Algorithms for Crystal Plasticity and Damage Mechanics I
Sponsored by: TMS Materials Processing and Manufacturing Division, TMS Functional Materials Division, TMS Structural Materials Division, TMS: Computational Materials Science and Engineering Committee, TMS: Integrated Computational Materials Engineering Committee, TMS: Phase Transformations Committee, TMS: Solidification Committee, TMS: Chemistry and Physics of Materials Committee
Program Organizers: Adrian Sabau, Oak Ridge National Laboratory; Ebrahim Asadi, University of Memphis; Enrique Martinez Saez, Clemson University; Garritt Tucker, Colorado School of Mines; Hojun Lim, Sandia National Laboratories; Vimal Ramanuj, Oak Ridge National Laboratory
Tuesday 2:30 PM
March 21, 2023
Room: Aqua 310A
Location: Hilton
Session Chair: Hojun Lim, Sandia National Laboratories
2:30 PM Invited
A New AI/ML Framework for Materials Innovation: Surya Kalidindi1; 1Georgia Institute of Technology
A novel information gain-driven Bayesian AI/ML (artificial intelligence/machine learning) framework is presented with the following main features: (i) explicit consideration of the physics parameters as inputs (i.e., regressors) in the formulation of process-structure-property (PSP) surrogate models needed to drive materials improvement workflows; (ii) information gain-driven autonomous workflows for training efficient AI/ML surrogates to otherwise computationally expensive physics-based simulations; (iii) versatile feature engineering for multiscale material internal structure using the formalism of n-point spatial correlations; (iv) amenable to a broad suite of surrogate model building approaches (including Gaussian Process regression (GPR), convolutional neural networks (CNN)); and (v) Markov chain Monte Carlo (MCMC)-based computation of posteriors for physics parameters using all available experimental observations (usually disparate and sparse). The benefits of this framework in supporting accelerated design and development of heterogeneous materials will be demonstrated using multiple case studies.
3:10 PM
A Non-local Formulation of the Elastoplastic Self-consistent Crystal Plasticity Model: Applications to Modeling Deformation and Recrystallization: Zhangxi Feng1; Miroslav Zecevic2; Ricardo Lebensohn2; Marko Knezevic1; 1University of New Hampshire; 2Los Alamos National Laboratory
This work extends a standard mean-field elastoplastic self-consistent (EPSC) formulation to consider intragranular misorientation distributions and stress fluctuations resulting from the second moments of the stress and lattice spin fields inside the grains. The extended formulation calculates the geometrically necessary dislocation (GND) density along with the underlying backstress fields that influence the activation of slip systems. Introducing GNDs in the evolution of slip resistance renders the formulation non-local. Accuracy and performance of the extended EPSC model is compared with the extended viscoplastic self-consistent (VPSC) formulation [1] in terms of predictions of texture evolution during deformation and recrystallization of a cold drawn copper wire and a compressed iron-carbon alloy.[1] M. Zecevic, R.A. Lebensohn, R.J. McCabe and M. Knezevic: "Modelling recrystallization textures driven by intragranular fluctuations implemented in the viscoplastic self-consistent formulation". Acta Materialia, 164, 530-546 (2019).
3:30 PM
A Peridynamic-based Approach to Study the Influence of Oxide on Impact and Bonding in Cold Spray: Baihua Ren1; Jun Song1; 1McGill University
In many metal cold spray depositions, a key factor in determining bond strength and final coating quality is the oxide-free metal-to-metal contact between the powder and the substrate. In this study, a peridynamics (PD) based numerical approach was developed and employed to model the deformation behaviors of oxide-coated metal particles during cold-spray deposition. The disruption, fracture, and subsequent removal of oxide films during impact were studied. The amount and morphology of the residual oxide, and locations of intimate metal contact were found to be consistent with the experimental observations. The effects of oxide thickness and fracture toughness on particle deformation/bonding were investigated, and the coefficients of restitution for different particle impact velocities with varying adhesion energies between the particles and the substrate were also calculated. This simulation method provides a more realistic numerical framework for the more accurate study of cold spray impact in the presence of oxide films.
3:50 PM
Crystal Plasticity Finite Element Analysis of Crystalline Thermo-mechanical Constitutive Response: Anderson Nascimento; Akhilesh Pedgaonkar1; Curt A Bronkhorst1; Irene Beyerlein2; 1University of Wisconsin-Madison; 2University of California, Santa Barbara
Constitutive thermo-mechanical models allow for a more accurate description of temperature sensitive processes and often provide a pathway for a more rigorous thermodynamic description of crystal plasticity, given the importance of temperature in the dislocation density of crystalline materials. Commonly, however, the thermo-mechanical description of the materials response is limited to phenomenological macroscopic models with limited interpretability. In this work, thermal expansion is included in the single crystal constitutive response via an eigenstrain contribution to the deformation gradient and a fully coupled implicit thermo-mechanical formulation is implemented into the variational framework of the crystal plasticity finite element method. Such thermodynamically consistent model with temperature and micromechanical fields is used to investigate the response of different materials under thermal-stress boundary conditions.
4:10 PM Break
4:25 PM
Data-Driven Bayesian Model-Based Prediction of Fatigue Crack Nucleation in Ni-based Superalloys: Somnath Ghosh1; George Weber2; Maxwell Pinz1; 1Johns Hopkins University; 2NASA Langley
This paper develops a Bayesian inference-based probabilistic crack nucleation model for the Ni-based superalloy Ren'e 88DT under fatigue loading. A data-driven, machine learning approach is developed to identify the underlying mechanisms driving crack nucleation. An experimental set of fatigue-loaded microstructures is characterized near crack nucleation sites using SEM and EBSD images for correlating the grain morphology and crystallography to the location of crack nucleation sites. A concurrent multiscale model that embeds experimentally acquired polycrystalline microstructural RVEs in a homogenized material, is developed for fatigue simulations. The RVE domain is modeled by a crystal plasticity finite element (CPFE) model. A Bayesian classification method is introduced to optimally select the most informative state variable predictors of crack nucleation and constructs a near-Pareto frontier of models. From this principal set of state variables, a simple scalar crack nucleation indicator is formulated.
4:45 PM
Data-driven Plastic Anisotropy Predictions Using Crystal Plasticity and Deep Learning Models: Hojun Lim1; Taejoon Park2; David Montes de Oca Zapiain1; Farhang Pourboghrat2; 1Sandia National Laboratories; 2The Ohio State University
Traditional methods of characterizing plastic anisotropy in metal alloys require iterative experiments or high-fidelity computational simulations. To avoid expensive anisotropy characterization procedures, a novel data-driven anisotropy prediction model is developed from a large dataset of crystal plasticity (CP) calculations. A deep learning (DL) model was trained by analytical CP calculations of normalized yield stresses and plastic strain increments. The validity and accuracy of the DL model was assessed by additional validation dataset from CP calculations. Quantitative comparisons of CP and DL predictions show that the DL model accurately and efficiently links material’s initial crystallographic texture to plastic anisotropy. DL-based predictions were further assessed by performing finite element simulations of cup drawing using the non-quadratic yield function parameterized from CP, CP-FEM and DL anisotropy predictions. The DL-based simulation showed excellent agreement CP and CP-FEM, demonstrating that accurate anisotropy information was obtained without the need of performing expensive high-fidelity simulations.
5:05 PM
Exascale Fracture Mechanics with Peridynamics: Sam Reeve1; Pablo Seleson1; 1Oak Ridge National Laboratory
Peridynamics is a nonlocal reformulation of classical continuum mechanics ideal for simulation of material failure and damage, successfully demonstrated as a tool capable of capturing complex fracture phenomena across many applications. However, the nonlocal nature of peridynamics makes it highly computationally expensive compared to classical continuum mechanics, hindering large-scale fracture simulations. We present CabanaPD, a new performance portable and exascale-capable peridynamics application designed to run on U.S. Department of Energy supercomputers by building on top of the Cabana and Kokkos libraries. We will discuss design of CabanaPD, highlighting the performance critical force and communication kernels, explore performance across particle counts and neighborhood sizes, and show scalability for benchmark test problems.
5:25 PM
Finite Element Implementation of a Dislocation Thermo-mechanics Model: Application to Study Dislocation Structure Evolution during Laser Scanning: Gabriel Lima Chaves1; Manas Upadhyay1; 1Ecole Polytechnique
Recently, a thermo-mechanical model of field dislocations (T-FDM), was proposed by Upadhyay, JMPS 145 (2020) 104150. The motivation behind the development of this model was two-fold: (i) to study the dynamics of dislocations under transient heterogeneous temperature fields such as those occurring during metal additive manufacturing, welding, quenching, and other thermomechanical processes, and (ii) to capture the temperature changes induced via dislocation dynamics. In the present study, the partial differential equations (PDEs) of the T-FDM model are numerically solved using the finite element (FE) method. The model is implemented in FEniCS, which is an open-source platform built to solve PDEs via the FE method.In this talk, the governing equations of the T-FDM model will be presented. Their numerical implementation will be followed by some illustrative examples. Specifically, dislocation patterning induced during laser scanning of a metal microstructure, similar to that occurring during additive manufacturing, will be shown.