Algorithm Development in Materials Science and Engineering: Microscale Algorithms and Their Applications - Solidification
Sponsored by: TMS Materials Processing and Manufacturing Division, TMS: Computational Materials Science and Engineering Committee, TMS: Integrated Computational Materials Engineering Committee, TMS: Phase Transformations Committee, TMS: Solidification Committee
Program Organizers: Mohsen Asle Zaeem, Colorado School of Mines; Mikhail Mendelev, NASA ARC; Garritt Tucker, Colorado School of Mines; Ebrahim Asadi, University of Memphis; Bryan Wong, University of California, Riverside; Sam Reeve, Oak Ridge National Laboratory; Enrique Martinez Saez, Clemson University; Adrian Sabau, Oak Ridge National Laboratory
Tuesday 4:00 PM
March 1, 2022
Room: 253A
Location: Anaheim Convention Center
Session Chair: Mohsen Asle Zaeem, Colorado School of Mines; Adrian Sabau, Oak Ridge National Laboratory
4:00 PM
Multiscale Modeling of Ni Alloys Selective Laser Melting Combining CalPhaD, Finite Elements, and Phase-field Methods: Seyed Mohammad Elahi1; Rouhollah Tavakoli2; Ahmed Kaci Boukellal2; Thomas Isensee1; Ignacio Romero1; Damien Tourret2; 1IMDEA Materials & Universidad Politecnica de Madrid (UPM); 2IMDEA Materials
This study aims to build an Integrated Computational Materials Engineering (ICME) framework to simulate additive manufacturing across scales, focusing on selective laser melting (SLM). We combine CalPhaD, Finite Elements (FE) and Phase-Field (PF) methods, to predict relevant features of printed parts at different length scales. The CalPhaD method is used to compute the phase diagram and temperature-dependent alloy properties (e.g. heat capacity, density) used in other models. At the macroscopic scale, FE are used to calculate residual stresses and part distortion, as well as the temperature profile used in PF simulations. Microstructure formation at the scale of the melt pool, within two-dimensional slices of the printed part, is simulated using a computationally-efficient polycrystalline multi-GPU parallelized PF model. We apply our computational framework to SLM of nickel-based superalloys such as Inconel-718 and Hastelloy-X.
4:20 PM
Microstructure and Porosity Predictions in Additively Manufactured Ti-6Al-4V Alloys Using a Hierarchical Modeling Approach: Bonnie Whitney1; Akshatha Chandrashekar Dixith1; Anthony Spangenberger1; Diana Lados1; 1Worcester Polytechnic Institute
The additive manufacturing (AM) technologies are rapidly changing the manufacturing paradigm by enhancing component design flexibility while lowering development time and cost. Computational tools are needed to optimize AM parts for high-integrity structural applications, which require accurate microstructure and porosity predictions. This research proposes a hierarchical computational methodology for laser powder bed fusion of Ti-6Al-4V to link key processing-structure parameters. The approach integrates component-scale thermal finite element simulations (spatial-temporal temperature histories), melt pool cellular automaton simulations (prior beta grain morphology), and solid-state phase field transformations (alpha phase morphology/orientation). In parallel, melt pool dimensions from thermal simulations were used to predict the morphology and distribution of lack-of-fusion pores using a geometric model. An extensive experimental build was judiciously developed and characterized to validate and fine-tune the microstructure and porosity predictions. The results will be presented and discussed from the perspective of future integration in mechanical properties modeling.
4:40 PM Invited
OpenMP GPU Offloading for Cellular Automaton Solidification Microstructural Model: Lang Yuan1; Adrian Sabau2; Jean-Luc Fattebert2; 1University of South Carolina; 2Oak Ridge National Laboratory
The state-of-the-art supercomputers implements a hybrid architecture where general-purpose host CPUs couple with specialized computing devices, GPUs, allowing significant acceleration of computations. In this study, a CPU-based solidification microstructure code that simulates dendritic growth was parallelized with Open MPI with implementation of OpenMP GPU offloading. This development allows the code to take the advantages of exascale computing without significantly restructuring the original code. The offloading strategy, unique data structure inherited to cellular automation methods and memory access will be discussed for the improvement of GPU utilization. Two cases, unconstrained dendritic growth under the equilibrium condition and constrained growth during rapid solidification, were examined in detail to evaluate its performance. Individual computational modules, e.g., nucleation, diffusion, interface capture, can be accelerated from 1 to 100 times, depending on the computational intensity. Overall performance was improved ranging from 2.5 times to 30 times based on the baseline conditions and solidification cases.