HEA 2023: High-throughput and Machine Learning I
Program Organizers: Andrew Detor, DARPA/DSO; Amy Clarke, Los Alamos National Laboratory
Tuesday 1:40 PM
November 14, 2023
Room: Riverboat
Location: Omni William Penn
Session Chair: George Spanos, TMS
1:40 PM Introductory Comments
1:45 PM Invited
BIRDSHOT: An Accelerated Program for the Discovery and Optimization of Refractory High Entropy Alloys: Raymundo Arroyave1; Brent Vela1; Danial Khatamsaz1; William Trehern1; Ibrahim Karaman1; Weiwei Zhang2; Douglas Allaire1; Paul Mason2; Duane Johnson1; Prashant Singh3; Axel van de Walle4; Miladin Radovic1; Ankit Srivastava1; 1Texas A&M University; 2ThermoCalc; 3Ames National Laboratory; 4Brown University
The Refractory High Entropy Alloy (RHEA) space is vast and it is impossible to explore using conventional approaches to materials discovery. In this talk, we present the Batch-wise Improvement in Reduced Design Space using a Holistic Optimization Technique (BIRDSHOT) framework. BIRDSHOT incorporates the strengths of ICME and combinatorial methods, while addressing all their drawbacks, as it: (i) employs novel machine learning (ML) and data-driven search algorithms to identify efficiently the feasible regions amenable to optimization; (ii) exploits correlations to fuse simulations and experiments to obtain efficient ML models for predicting PSPP relations; (iii) uses Bayesian Optimization (BO) to make globally optimal iterative decisions regarding which region in the RHEA space to explore/exploit, leveraging existing models and data; (iv) is capable of carrying out multiple optimal parallel queries to the design space. We show how we have been using BIRDSHOT to search for next generation refractory alloys for turbine engine applications.
2:15 PM
Yield Strength-plasticity Trade-off and Uncertainty Quantification in ML-based Design of Refractory High-entropy Alloys: Stephen Giles1; Debasis Sengupta1; Hugh Shortt2; Peter Liaw2; 1CFD Research Corp; 2University of Tennessee, Knoxville
Development of process-structure-property relationships in materials science is an important and challenging frontier which promises improved materials and reduced time and cost in production. Refractory high entropy alloys (RHEAs) are a class of materials that are capable of excellent high-temperature properties. However, due to their multi-component nature, RHEAs have a vast composition space which presents challenges for traditional experimental exploration. Here, quantitative models of compressive yield strength and room-temperature plasticity are developed through a deep learning approach. Uncertainty quantification is performed through a variety of statistical validation techniques. Model predictions are experimentally validated through collection of recent literature and the synthesis and experimental characterization of two new, unreported RHEAs: AlMoTaTiZr and Al0.239Mo0.123Ta0.095Ti0.342Zr0.201. Finally, through the application of model interpretability, features having the greatest impact on both the mechanical property and uncertainty of the deep learning models are revealed, and shown to agree well with current physics and materials science theory.
2:35 PM
Accelerating the Discovery of Lightweight High Entropy Alloys for Extreme Conditions: Michael Miller1; Jianliang Lin1; Mirella Vargas1; John Macha1; 1Southwest Research Institute
High entropy alloys (HEAs) represent a vast, largely unexplored space of material systems with potentially revolutionary combinations of properties such as strength at high temperatures and resistance to oxidation and corrosion. Compositions in these material systems have the potential to surpass the operational temperature limits of current state of the art high temperature alloys. However, the search space for candidate alloys is huge, making efficiency and optimization of the exploratory process crucial. A multifaceted alloy discovery approach has been undertaken that couples advanced computational materials modeling, synthesis of combinatorial thin-film libraries, and high throughput experimentation to rapidly assess the properties of synthesized alloys and provide feedback loops to further refine future modeling and alloy synthesis. The present focus is on discovering lightweight high entropy alloys (LHEAs), a subset of HEAs with potentially unique properties. By including elements with low atomic numbers, we have discovered LHEAs exhibiting stable, single phase domains.
2:55 PM Cancelled
Automated Characterization and Bayesian Prediction of High Strength Multi-principal Element Alloys: Eddie Gienger1; Maitreyee Sharma Priyadarshini2; Denise Yin1; Lisa Pogue1; Justin Rokisky1; Paulette Clancy2; Christopher Stiles2; 1JHU APL; 2Johns Hopkins University
Multi-principal element alloys (MPEAs) are an active research area for their desirable corrosion resistance, strength-to-weight ratios and high temperature survivability. However, design of new MPEAs with tailored performance is complex, and traditional Edisonian discovery methods are slow. AI-guided discovery is promising for these materials, but there is limited training data. Here, we demonstrate a high-throughput characterization pipeline enabling registration of mechanical properties from nano-indentation with material microstructure. These results, combined with literature data, are used to inform a novel physics-based material discovery framework, PAL 2.0. This new method leverages the advantages of Deep Learning with Bayesian Optimization, making it accessible to the data-scarce field of MPEAs. MPEA recommendations from PAL2.0 are synthesized and characterized to determine yield strength and hardness. This “in-the-loop" computational-experimental approach is one of the first of its kind and has a disruptively attractive potential in the field of materials discovery.
3:15 PM Break
3:35 PM
Fusing Analytical Models and Hardness Experiments for Accelerated Optimization of Yield Strength in RHEAs: Brent Vela1; Danial Khatamsaz1; Cafer Acemi1; Prashant Singh2; Douglas Allaire1; Raymundo Arroyave1; Ibrahim Karaman1; Duane Johnson2; 1Texas A&M University; 2Ames Laboratory
Refractory high entropy alloys (RHEAs) have gained attention as potential replacements for Ni-based superalloys in gas turbine applications. Improving their properties, such as their high-temperature yield strength, is crucial to their success. Unfortunately, exploring this vast chemical space using only experimental approaches is impractical due to the cost of testing of candidate alloys at operation-relevant temperatures. The lack of reasonably accurate strength models makes traditional Integrated Computational Materials Engineering (ICME) methods inadequate. We address this challenge by combining machine-learning models, easy-to-implement physics-based models, and inexpensive proxy experiments to develop robust and fast-acting models via Bayesian-updating. The framework combines data from one of the most comprehensive databases on RHEAs with a widely used physics-based strength model for BCC-based RHEAs into a compact predictive model that is significantly more accurate that the state-of-the-art. This model is amenable to ICME frameworks that screen for RHEAs with superior high-temperature properties.
3:55 PM
Charge-density Based Convolutional Neural Networks for Stacking Fault Energy Prediction in Concentrated Alloys: Jacob Fischer1; Gaurav Arora2; Serveh Kamrava3; Pejman Tahmasebi3; Dilpuneet Aidhy1; 1Clemson University; 2Fermi Lab; 3Colorado School of Mines
A descriptor-less machine learning (ML) approach based only on charge density extracted from density functional theory (DFT) is developed to predict stacking fault energies (SFE) in concentrated alloys. Often, in ML models, textbook physical descriptors such as atomic radius, valence charge and electronegativity are used which have limitations because these properties ‘adjust’ in concentrated alloys due to varying nearest neighbor environments. We illustrate that, within the scope of DFT, the search for descriptors can be circumvented by charge density, which is the backbone of the Kohn-Sham DFT and describes the system completely. The descriptors are captured by charge-density inherently. The model is based on convolutional neural networks (CNNs) as one of the promising ML techniques. The performance of our model is evaluated by predicting SFE of concentrated alloys with an RMSE and R2 of 6.18 mJ/m2 and 0.87, respectively, validating the accuracy of the approach.
4:15 PM
Design Metastability in High-entropy Alloys by Tailoring Unstable Fault Energies: Chenyang Li1; Xing Wang2; Wei Xiong2; Wei Chen1; 1Illinois Institute of Technology; 2University of Pittsburgh
Metastable alloys with transformation-/twinning-induced plasticity (TRIP/TWIP) can overcome the strength-ductility trade-off in structural materials. Originated from the development of traditional alloys, the intrinsic stacking fault energy (ISFE) has been applied to tailor TRIP/TWIP in high-entropy alloys (HEAs) but with limited quantitative success. Here, we demonstrate a strategy for designing metastable HEAs and validate its effectiveness by discovering seven alloys with experimentally observed metastability for TRIP/TWIP. We propose unstable fault energies as the more effective design metric and attribute the deformation mechanism of metastable face-centered cubic alloys to unstable martensite fault energy (UMFE)/unstable twin fault energy (UTFE) rather than ISFE. Among the studied HEAs and steels, the traditional ISFE criterion fails in more than half of the cases, while the UMFE/UTFE criterion accurately predicts the deformation mechanisms in all cases. The UMFE/UTFE criterion provides an effective paradigm for developing metastable alloys with TRIP/TWIP for an enhanced strength-ductility synergy.
4:35 PM
High Throughput Multi-objective Optimization of FCC Complex Concentrated Alloys for Extreme Conditions: Ibrahim Karaman1; Raymundo Arroyave1; James Paramore2; Brady Butler2; Trevor Hastings1; Danial Khatamsaz1; Daniel Lewis1; Mrinalini Mulukutla1; Nicole Person1; Daniel Salas1; Wenle Xu1; Matthew Skokan1; Douglas Allaire1; George Pharr1; 1Texas A&M University; 2Army CCDC Army Research Laboratory
The FCC Complex Concentrated Alloys (CCA) compositional space is very broad, making it almost impossible to cover using conventional approaches to materials discovery. This talk will present an implementation of a framework combining an iterative multi-constrain multi-objective Bayesian optimization technique with CALPHAD-based phase stability predictions and machine learning-based modeling. This framework has been implemented for efficiently exploring the compositional space of FCC CCAs containing three or more elements among Co, Cr, Fe, Ni, V and Al, searching for alloys with the most optimum mechanical properties under extreme conditions. A total of 5 iterations (80 alloys) were produced and characterized in less than 9 months. These 80 alloys, which are only 0.15% of the total pool of 53124 alloys, were sufficient for achieving a clear picture of the objective mechanical properties Pareto front, demonstrating the high efficiency of the applied framework as compared to traditional approaches to materials discovery.
4:55 PM
High-entropy Materials Design by Integrating the First-principles Calculations and Machine Learning: A Case Study in the Al-Co-Cr-Fe-Ni System: Guangchen Liu1; Songge Yang1; Yu Zhong1; 1Worcester Polytechnic Institute
The first-principles calculation is widely used in high-entropy materials. However, this approach may consume many computational resources for complex systems, limiting the development of property maps for the related materials across the whole composition range. This work chooses the most prevalent Al-Co-Cr-Fe-Ni system (FCC and BCC) for our investigation. A comprehensive database of properties (e.g., phase stabilities and elastic properties) was established by combining the first-principles calculation results and machine learning: starting from unary, binary, ternary, and quaternary, then extending into quinary systems. A comparable software program was also developed by utilizing this database. Furthermore, the information/mechanism that underlies the database was thoroughly studied by screening and statistical analysis.