Conference Tools for 2024 TMS Annual Meeting & Exhibition

Help

About this Abstract
Meeting	2024 TMS Annual Meeting & Exhibition
Symposium	Materials Informatics to Accelerate Nuclear Materials Investigation
Presentation Title	Revealing the Story of Defects from Coupled Extreme Environments with Autoencoders and Dense Neural Networks
Author(s)	Kory Durell Burns, Khalid Hattar, Nan Li, Caitlin Kohnert
On-Site Speaker (Planned)	Kory Durell Burns
Abstract Scope	In situ transmission electron microscopy (TEM) allows one to observe the structure of a material at unprecedented spatial resolution, while probing non-equilibrium phenomena with an external stimulus. However, with the dynamic frame rate of cameras now collecting over 1000 frames per second, an individual experiment can easily create 100,000 frames to process. While it simply isn’t feasible for a human to extrapolate metrics from this amount of information, data-driven approaches emerge to mitigate this issue. In this study, we use few-shot image classification from in situ ion beam irradiation experiments to classify the generation, mobility, and interaction of dislocation loops and stacking fault tetrahedra. Next, we use a variational autoencoder to segment Helium bubbles during a thermal anneal by training on synthetic data based on distorted circles. Ultimately, we shed light on how data-driven approaches can probe defects in time-series micrographs for nuclear materials application.
Proceedings Inclusion?	Planned:
Keywords	Characterization, Machine Learning, Nuclear Materials

OTHER PAPERS PLANNED FOR THIS SYMPOSIUM

Accelerating Characterization of Radiation Driven Processes using Machine Learning Tools

Characterizing Microstructures in Aluminide Coatings Captured in SEM Image with Convolutional Neural Networks

Deep Neural Network for Porosity and Microstructure Analytics of a High Burnup U-10Zr Metallic Solid Fuel

Defect Evolution in Multi-principal Chemically Disordered Alloys from Multiscale Simulations

E-25: Feature Engineering for Construction of High-accuracy Thermal Conductivity Prediction Model for Uranium Compounds

E-26: Materials Genomics Search for Possible Helium-absorbing Nano-phases in Fusion Structural Materials and Experimental Validation

Emergent Molecular Structure and Dynamics of Tetrahedral Liquids Revealed by Neural Network Forcefield Simulations and Neutron Spin Echo Experiments

Few-shot Machine Learning for Automated Analysis of TEM Images of Nuclear Materials

Influence of Empirical Potentials on Data Quality in Computational Studies of Zr Alloys

Inverse Uncertainty Quantification of Dispersion Analysis Research Tool (DART) Parameters Necessary for the Calculation of Fission Gas Swelling in U-Mo Fuel

Machine Learning Enhanced Kinetic Monte Carlo Modeling of Molten Salt Corrosion of Ni-Cr Alloys

Machine Learning for Predicting Reactor Pressure Vessel Embrittlement

Materials Genomics Search for Helium-absorbing Nano-phases in Fusion Structural Materials

Modeling Cascade Damage in Tungsten Using Machine Learning SNAP Interatomic Potential: Electron-Phonon Interaction Model

Neural Networks of Defect Kinetics in Refractory Alloys

Optimizing Thermal Conductivity Prediction of Uranium Compounds using Balanced Multiclass Classification

Probing Radiation Induced Interface Metastability Using Deep Learning Object Detection

Putting Artificial Intelligence into Action to Quantify Radiation Effects in Materials

Quantitative Insight to Fission Gas Bubble Distribution and Lanthanide Movement in Irradiated Annular U-10Zr Metallic Fuel Using Deep Learning

Revealing the Story of Defects from Coupled Extreme Environments with Autoencoders and Dense Neural Networks

Scaling Ductility from Microscale to Bulk by Coupling Crystal Plasticity Simulations with 3D Convolutional Neural Networks

Scanning-TEM (STEM) 3D Tomography for Quantification of Radiation Damage in Neutron Irradiated 316L Stainless Steel

Synthetic Data Driven Materials Informatics Methods for Nuclear Materials Characterization

Utilizing Mechanistic Modeling and Uncertainty Analysis to Support Nuclear Fuel Qualification

Questions about ProgramMaster? Contact programming@programmaster.org