Machine Learning

  • AI/ML for Automated Critical Materials Resource Mapping

    Ahamed – INL R&D Staff Scientist

    Research Overview

    Efficient exploration and development of critical mineral resources increasingly depend on integrating large volumes of geologic and geophysical data. Traditional log‑analysis workflows for solution‑mined deposits require detailed interpretation of lithology, stratigraphic boundaries, fluid‑flow‑controlled alteration zones, and grade‑related signals. These processes are often labor‑intensive and subject to interpreter variability.

    This project seeks to apply modern machine learning methodologies to automate essential components of mineral resource mapping. Students will work with representative geophysical and lithologic datasets to:

    • Clean and preprocess well‑log data (e.g., resistivity, natural gamma, spontaneous potential)
    • Correct depth misalignments between lithologic and geophysical logs
    • Build ML models for lithology prediction, stratigraphic correlation, and geologic zonation analysis
    • Investigate pattern recognition methods to identify alteration fronts and mineralization indicators
    • Explore workflow expansion using physics‑informed ML or sparse sensing approaches to improve mapping under limited‑data conditions

    The overarching goal is to develop scalable data‑driven methods that enhance subsurface interpretation for critical materials and solution‑mining applications.

    Student Research Experience

    Students will engage in interdisciplinary work spanning geoscience, machine learning, and data engineering. Activities include:

    • Data Integration & Preprocessing – preparing and aligning multi‑source well‑log datasets, correcting sample‑lag effects, and constructing geologically meaningful features.
    • Machine Learning Model Development – implementing algorithms such as PCA, dynamic time warping, tree‑based models, gradient‑boosting methods, and neural networks to support automated interpretation.
    • Geologic Interpretation – learning how lithologic variation, stratigraphic architecture, and fluid‑alteration signatures relate to mineral resource distribution within solution‑mined systems.
    • Model Evaluation & Visualization – generating correlation panels, prediction plots, and zonation maps to assess machine‑learning results in a geologic context.

    Students will collaborate with INL staff and gain experience applying AI/ML techniques to real‑world subsurface problems relevant to critical materials, energy security, and advanced resource extraction technologies.

    Skills & Techniques

  • Classical Molecular Dynamics Simulations of Nuclear Materials

    Russell – University of Idaho

    Research Overview

    Atomistic simulations are increasingly valuable for industrial needs to predict thermodynamic and kinetic properties of systems at the nanoscale. Design and fitting of interatomic potentials for application to nuclear materials is also an area of fundamental interest. Simulation and modeling can provide understanding of mechanisms that occur far from equilibrium, which can be difficult to directly observe by experiment.

    Accurate prediction of melting temperature and elastic properties at finite temperatures are challenging because interatomic potentials are typically fit to zero temperature electronic structure results. New interatomic potentials based on explainable artificial intelligence concepts are needed that combine physics based functional forms with advanced machine learning techniques to fit parameter sets.

    Student Research Experience

    Working in collaboration with INL’s Collaborative Computing Center, students will learn fundamentals of computational atomistic simulation and modeling. Students will obtain a user account on the NSF ACCESS cyberinfrastructure program and utilize the Bridges-2 supercomputer hosted by the Pittsburgh Supercomputing Center (PSC).

    Students will learn to use high performance computers, the LAMMPS code, visualization with VMD code, scripting, and post processing techniques. They will also learn fundamentals of machine learning to fit parameters and the physics behind interatomic potential design. Advanced students may learn programming languages and parallel computing techniques.

    Skills & Techniques