ICME 2023: AI/ML: Alloys
Program Organizers: Charles Ward, AFRL/RXM; Heather Murdoch, U.S. Army Research Laboratory
Tuesday 8:00 AM
May 23, 2023
Room: Boca I-III
Location: Caribe Royale
8:00 AM Invited
Batch-wise Improvement in Reduced Design Space Using a Holistic Optimization Technique (BIRDSHOT): Raymundo Arroyave1; 1Texas A&M University
The efficient exploration and exploitation of compositionally complex alloy spaces is extremely resource-intensive and most conventional approaches (e.g. traditional ICME and open-loop combinatorial methods) are not effective. Here, I present our recent work on the development of BIRDSHOT. 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) leverages the team’s newly developed batch modifications to BO that enable the parallel, iterative and optimal exploration/exploitation of materials spaces; and (v) is capable of simultaneously considering multiple objectives and constraints.
8:30 AM
Elastic Constants Predictions in Multi-principal Element Alloys from DFT and Machine Learning: Nathan Linton1; Dilpuneet Aidhy1; 1Clemson University
Multi-principal element alloys (MPEAs) present a paradigm shift in materials design and consist of multiple principal elements randomly distributed on a crystal lattice resulting in an enormous phase space. On the one hand this presents opportunities to unravel novel properties whereas on the other it presents a large challenge to survey the phase space, presenting a data-science challenge. We present PREDICT (PRedict properties from Existing Database In Complex alloys Territory), a machine learning framework coupled with electronic structure methods whereby properties in MPEAs could be predicted by learning from the binary alloys database. Specifically, we demonstrate predictions of stiffness constants, Young’s modulus, bulk and shear moduli, and Poisson’s ratio in ternary, quaternary, and quinary MPEAs with a high-level of accuracy. A major benefit of this is that for every new composition discovered, the mechanical properties can be computed using only the existing binary alloy database, bypassing the computationally expensive calculations.
8:50 AM
Charge-density Based Convolutional Neural Networks for Stacking Fault Energy Prediction in Concentrated Alloys: Gaurav Arora1; Serveh Kamrava2; Pejman Tahmesabi2; Dilpuneet Aidhy3; 1University of Wyoming; 2Colorado School of Mines; 3Clemson University
A descriptor-less machine learning (ML) model based only on charge density extracted from density functional theory (DFT) is developed to predict stacking fault energies (SFE) in concentrated alloys. Often, in most ML models, textbook physical descriptors such as atomic radius, valence charge and electronegativity are used which have limitations because these properties change in concentrated alloys when multiple elements are mixed to form a solid solution. 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 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 proposed approach.
9:10 AM
An Interpretable Machine Learning Model to Predict Molten Salt Corrosion of Compositionally Complex Alloys and Facilitate Understanding of Novel Corrosion Mechanisms: Bonita Goh; Yafei Wang1; Phalgun Nelaturu2; Michael Moorehead3; Dan Thoma2; Santanu Chaudhuri4; Jason Hattrick-Simpers5; Kumar Sridharan2; Adrien Couet2; 1Shanghai Jiaotong University; 2University Of Wisconsin Madison; 3Idaho National Laboratory; 4University of Illinois - Urbana-Champaign; 5University of Toronto
CCAs are of interest as structural materials in molten salt reactors because current alloys certified by ASME Sec(III) Div(5) code for their mechanical properties contain high Cr that are readily corrodible in molten halides due to the thermodynamic favorability of Cr corrosion in halides at nuclear reactor operating conditions. Corrosion of compositionally complex alloys (CCAs) in molten halides is not straightforward to predict because they do not possess one obvious base element whose kinetic and thermodynamic behavior provides the basis for prediction. The lack of prediction capability presents a bottleneck to search a quasi-infinite compositional space for a particular set of alloying elements for CCAs that could be suitable for molten salt reactor structural materials. We present a generalizable Random Forest Regressor (RFR) model trained and tested on 110 experimentally tested CCAs. Shapley analysis was used to interpret the model and extract alloy design parameters for optimizing corrosion resistance.
9:30 AM Break