Materials Processing and Fundamental Understanding Based on Machine Learning and Data Informatics: AI-guided Processing Study
Program Organizers: Kathy Lu, University of Alabama Birmingham; Jian Luo, University of California, San Diego; Xian-Ming Bai, Virginia Polytechnic Institute and State University; Yi Je Cho, Sunchon National University
Monday 2:00 PM
October 10, 2022
Room: 311
Location: David L. Lawrence Convention Center
Session Chair: Xian-Ming Bai, Virginia Polytechnic Institute and State University
2:00 PM Invited
Unraveling the Process Fundamentals of Additive Friction Stir Deposition by Integrating Physics Simulation with Data-driven Approaches: Hang Yu1; 1Virginia Polytechnic Institute and State University
Additive friction stir deposition is an emerging solid-state additive process leveraging rapid plastic deformation at elevated temperatures to enable location-specific material deposition. By producing fully-dense materials with forging-like mechanical characteristics, it has shown significant potential in a wide range of defense, automotive, and aerospace applications, including structural repair and dissimilar material cladding, as well as material recycling and upcycling. However, with fully-coupled heat and mass transfer and complex nature of the process, predicting the history of critical thermomechanical variables like strain rate, strain, and temperature based on the processing parameters remains difficult. To enable fast and precise prediction of thermomechanical variables, here we propose combining physics modeling with data-driven techniques, such as Bayesian learning. We show that compared to pure data-driven approaches, this combined strategy requires a much smaller dataset for training and cross validation, while enabling transfer learning.
2:30 PM
Comparison of Data Driven and Physics-informed Machine Learning Models for Temperature Prediction of Shear Assisted Processing and Extrusion: Ethan King1; Colby Wight1; WoongJo Choi1; Zhao Chen1; Keerti Kappagantula1; Shenyang Hu1; Yulan Li1; Tegan Emerson1; Sarah Akers1; Henry Kvinge1; Eric Machorro1; Jenna Pope1; Erin Barker1; Eric Smith1; 1Pacific Northwest National Laboratory
Shear assisted processing and extrusion (ShAPE™) uses friction to plasticize feedstock with a rotating die, which is then extruded into a consolidated tube. The microstructure and properties of the extruded product are dependent on the extrusion temperature, controlled by the tool rotation and traverse rate. The effect of these process parameters is in turn modulated by the temperature of the deforming material. Thus control of process temperature depends on complex thermo-mechanical feedbacks that are challenging to model. In this talk, we discuss two approaches developed to model the relationship between ShAPE process parameters and temperatures, namely a data-driven approach called DeepTemp and a physics informed ML model. We compare the accuracy and data requirements between the data driven and physics-informed ML models and demonstrate that ML models can closely predict process temperatures during ShAPE, which can accelerate and improve ShAPE process optimization.
2:50 PM
Composition and Property Prediction of Polymer-derived Silicon Oxycarbides: Yi Je Cho1; Harrison Chaney2; Kathy Lu2; 1Sunchon National University; 2Virginia Tech
Polymer-derived silicon oxycarbide (SiOC) materials enable the formation of homogeneous microstructures and high temperature stable properties. However, the relationship between the processing parameters and microstructures/properties has not been clearly understood. In this study, a materials informatics approach was employed to analyze and estimate this relationship. The correlation analysis provided importance ranking of the process parameters, which can be later utilized for the fabrication process. Machine learning models with high accuracy were proposed using the ranked features obtained from the correlation analysis. ReaxFF simulations were performed to evaluate the fidelity of the machine learning models and to confirm the feasibility of the models for designing new SiOC materials. The data analytics workflow proposed in this study can be extended to different types of polymer-derived ceramics by incorporating various features and targets involved in the process, microstructures, and properties.
3:10 PM
Computational and Machine Learning Studies of DNA-templated Dye Aggregate Design: Lan Li1; 1Boise State University
Organic molecules, known as dyes, which can absorb and emit light, have various potential applications, such as biomedical imaging, organic photovoltaics, non-linear optics, and quantum information systems. These applications are controlled by dye properties, including extinction coefficient, transition dipole moment, and aggregation ability. Dye aggregate networks via deoxyribonucleic acid (DNA) templating exhibit exciton delocalization, energy transport, and fluorescence emission. DNA nanotechnology provides scaffolding upon which dyes attach in an aqueous environment. In order to control the process and optimize the properties, we have combined machine learning and multiscale modeling to identify ideal dye candidates and reveal their dye aggregate-DNA interactions and the dye orientations. We found that those structural features have a strong impact on the resultant performance of the DNA-templated dye aggregates. The computational results were also validated with experimental observations.
3:30 PM Break
3:50 PM Invited
Data-Driven Study of Shape Memory Behavior of Multi-component Ni-Ti Alloys: Shreyas Honrao1; Othmane Benafan2; John Lawson1; 1NASA Ames Research Center; 2NASA Glenn Research Center
Ni-Ti based shape memory alloys (SMAs) have found wide-spread use in aerospace, automotive, biomedical, and commercial applications owing to their favorable properties and ease of operation. Especially important for many NASA applications is the ability to tune the martensitic transformation temperature of Ni-Ti alloys by varying the composition and processing conditions. Recently, researchers at NASA have compiled an extensive database of shape memory properties of materials, including over 8,000 multi-component Ni-Ti alloys containing 37 different alloying elements. Using this dataset, machine learning models are trained to predict transformation temperatures, hysteresis, and transformation strain with extremely small errors. These models are used to learn relationships between shape memory behavior and input parameters in the composition and processing space. ML predictions are validated through new experiments. The combination of an extensive dataset and accurate learning models, together, make our approach highly suitable for the rapid discovery of novel SMAs with targeted properties.
4:20 PM
Machine Learning for Joint Quality Performance-determining Relationship between Intermetallic Properties and weld Microstructure of Al/steel Resistance Spot Welds: Moses Obiri1; Alejandro Ojeda1; Deborah Fagan1; Keerti Kappagantula1; Hassan Ghassemi-Armaki1; Blair Carlson1; 1Pacific Northwest National Laboratory
Resistance spot welding (RSW) is a welding technique used to join resistive metals such as aluminum and low carbon steel by applying pressure and heat from an electric current to the weld area. The properties of RSW of aluminum to steel are being studied to reduce vehicle weight and thus increase fuel efficiency. Previous research has described the properties of microstructure variables (fracture mode, hardness, and thickness) in the intermetallic layer formed by RSW of Al – steel welds. Joint performance is also well known to be dependent on the spot weld attributes developed during processing, and several factors influence the weld attributes during processing. The relationship between microstructure variables and joint performance, on the other hand, has yet to be thoroughly investigated. We categorize hardness and thickness profile curves using machine learning techniques and identify profiles that result in optimal weld performance.
4:40 PM
Addressing Data Needs for High Temperature Material Processing with Natural Language Processing: Amit Verma1; Benjamin Glaser1; Robin Kuo1; Jason Zhang1; Nicholas David1; Zhisong Zhang1; Emma Strubell1; Anthony Rollett1; 1Carnegie Mellon University
Data problems persists across many disciplines of materials science, with a particular extreme dearth for high temperature materials where most material attributes need to be determined experimentally. To address this challenge, we are working on two key ideas: 1) data retrieval; and 2) recognition systems for identifying key concepts and their dependencies, from published literature. The first aim to address the lack of open-access experimental data for various machine learning activities, while the second aim to encode the semantics of the domain for bridging various heterogenous data sources. Natural Language Processing (NLP) provide a host of solutions in this regard, and this talk focuses on how NLP is being used to develop the tools mentioned, with specific examples to support our vision.
5:00 PM
Machine Learning Based Prediction of Cation Distribution in Complex Spinel Oxides as a Function of Processing Temperature: Ying Fang1; Siming Zhang1; Guofeng Wang1; 1University of Pittsburgh
It has been found that both cation chemistry and degree of inversion play an important role in technically relevant properties of spinel oxides. In this study, we have developed and applied a machine learning based computational approach to predict the equilibrium cation distribution in multi-cation spinel oxides at high temperatures. The database was constructed to contain the density functional theory calculated energies of the spinel oxides with various cation distributions. We applied the machine learning techniques (i.e., linear regression and neural network) to find the relation between the system energy and structural features of the spinel oxides and performed the atomistic Monte Carlo simulations to predict the equilibrium cation distribution as a function of processing temperature for single spinel AB2O4 and double spinel AB2-xCxO4. Our predicted cation distributions for material systems of CoFe2O4, NiFe2O4, MgAl2O4, CuAl2O4, and MgAl2-xFexO4 are found to agree well with available experimental results.