Stinville Research Group
Materials Science and Engineering
Techniques & AI Science Tools
We develop and combine high-fidelity experimental methods with AI Science tools to observe materials, quantify their local response, learn from complex spatial datasets, and accelerate the prediction of macroscopic performance. Explore the experimental and computational methods that underpin our research.
From experiments to materials intelligence
Our experimental and AI capabilities are designed as one connected workflow: high-fidelity measurements capture material states and their evolution; AI tools encode and interpret those states; predictive models connect them to performance and guide the next experiment or materials-design decision.
Experimental Techniques
High-resolution, large-field-of-view and extreme-environment measurements designed to capture microstructure, deformation, damage and material-state evolution across relevant length and time scales.
High-Resolution Digital Image Correlation
SEM-based full-field measurements of heterogeneous plastic deformation at the microstructure scale, enabling quantitative mapping of localization and strain evolution.
Cryogenic Mechanical Testing
Mechanical testing from conventional temperatures to liquid-nitrogen and liquid-helium regimes for understanding material response in extreme environments.
High-Throughput Characterization & Testing
Experimental workflows that combine rapid mechanical testing and large-area microstructure characterization to evaluate many material states efficiently.
Very High Cycle Fatigue Testing
Ultrasonic fatigue methods for probing damage and failure at very large cycle counts, including conditions relevant to long-life structural components.
2D / 3D Microstructure Characterization
Spatially resolved characterization of crystallography, morphology, defects and microstructural neighborhoods in two and three dimensions.
Transmission Scanning Electron Microscopy
Electron microscopy approaches for resolving nanoscale deformation structures and linking dislocation-scale processes to larger-scale mechanical behavior.
Heaviside Digital Image Correlation
DIC formulations designed to quantify sharp displacement discontinuities and localized deformation events associated with slip and interfacial processes.
Time-Resolved High-Resolution DIC
Time-resolved full-field measurements for tracking the nucleation, accumulation and evolution of deformation events during loading.
AI Science Tools
Machine-learning methods developed to encode complex material states, detect physical events, connect multimodal measurements, predict performance, and identify the microstructural features and interactions that govern materials behavior.
Materials Spatial Intelligence
A framework that learns directly from spatial maps of microstructure and local material response to predict macroscopic properties and discover governing features.
Self-Supervised Microstructure Encoding
Variational autoencoders, contrastive approaches and related representation-learning methods for discovering compact descriptions of heterogeneous microstructures.
Predictive Representation Learning
Predictive and joint-embedding learning strategies for representing material states and learning relationships among evolving spatial observations without reconstructing every pixel.
Automated Deformation-Event Detection
Computer-vision workflows that automatically identify and quantify localized plastic deformation events from high-resolution experimental fields.
Spatial Property Prediction
Encoding and mapping local deformation fingerprints to rapidly predict strength, fatigue and other macroscopic properties from short-duration experiments.
Accelerated Fatigue Prediction
High-throughput experimental and learning workflows for extracting fatigue information across multiple material states at dramatically reduced testing burden.
Interpretable & Explainable AI
Feature-attribution and controlled-perturbation strategies that identify the microstructural regions, features and spatial interactions governing predicted performance.
Multimodal Spatial Data Fusion
Registration and joint learning from crystallography, microstructure, chemistry, deformation and property maps while preserving their physical spatial relationships.
