About this Abstract

Meeting

MS&T24: Materials Science & Technology

Symposium

Machine Learning and Simulations

Presentation Title

EBSD Geometry Calibration Through SE(3) Lie Group Optimization

Author(s)

Zachary Varley, Marc De Graef, Gregory Rohrer

On-Site Speaker (Planned)

Zachary Varley

Abstract Scope

We present a novel method for electron backscatter diffraction (EBSD) detector geometry calibration that parametrizes the full 6-degrees-of-freedom detector pose via the special Euclidean group SE(3). Central to this approach is the usage of physics-based forward modeling of backscatter electron signal generation. The core innovation lies in jointly indexing orientations and estimating detector pose by iteratively comparing projected simulation patterns with experimental patterns. Optimization is performed in the Lie algebra tangent space using many experimental patterns from across the region of interest. The algorithm is implemented in PyTorch and validated on standard Nickel EBSD datasets with varied noise levels to quantify the method's noise robustness. The basin of attraction and optimization landscape are compared across metrics. By integrating Lie algebra based optimization with physics-based simulations, a more general model of detector geometry may be fitted, enabling enhanced EBSD measurements compared to standard pattern projection center methods.

OTHER PAPERS PLANNED FOR THIS SYMPOSIUM

A Machine Learning Approach to Predict Solute Segregation Energy in Ni Grain Boundaries

A Machine Learning Based Computational Method for Accurate Prediction of Equilibrium Cation Distribution in Complex Spinel Oxides

Assessing GPR Models for Steel Hardness Prediction in Production Environments

B-2:Forecasting Nutrient Flows Using Terrain Elevation-Aware Spatial-Temporal Graph Neural Networks

Decoding the Structural Genome of Silicate Glasses

EBSD Geometry Calibration Through SE(3) Lie Group Optimization

End-To-End Differentiability and Tensor Processing Unit (TPU) Computing to Accelerate Materials’ Inverse Design

Estimation of Thermal Hysteresis in Zirconia Using Machine Learning Molecular Dynamics and Transition State Modelling

Forward Prediction and Inverse Design of Additively Manufacturable Alloys via Autoregressive Language Models

Generation of Machine Learning Interatomic Potentials for Boron Carbide with Comparison to the Analytic Angular Dependent Potential

Graph Neural Networks for Rapid Continuum Damage Modeling of Semi-Crystalline Polymers

Machine Learning in Nuclear Waste Glass Formulation and Property Model Development

Multi-Fidelity Gaussian Process Models for Time-Series Outputs

New Machine–Learning Interatomic Potentials (MLIPs) for Si-C-O-H Compounds Enabling Atomistic Simulations of Complex Chemical Transformations

Predicting the Dynamics of Atoms in Liquids by a Surrogate Machine-Learned Simulator

Understanding Grain-Boundary Structure Using Strain Functional Descriptors and Unsupervised Machine Learning

Questions about ProgramMaster? Contact programming@programmaster.org | TMS Privacy Policy | Accessibility Statement