Physics-based modelling
Computational methods for studying electromagnetic behaviour in underground environments.
Underground-imaging SciML workflow
A computational approach to subsurface interpretation using physics-based simulation, scientific machine learning and validation-aware analysis.
About the project
Underground imaging is relevant across construction, transportation, energy, defence, utilities, archaeology and infrastructure assessment. These applications often require an understanding of what lies below the surface without extensive excavation or direct physical access.
The main difficulty is not limited to collecting signals. Subsurface measurements are affected by ground conditions, material interfaces, environmental variation and noise, making interpretation a complex problem.
Scientific machine learning provides a way to combine computational modelling with data-driven interpretation. The objective is to support faster and more informative analysis while keeping the results connected to the underlying physics.
How it works
The workflow combines physics-based electromagnetic modelling, structured simulation data and scientific machine-learning methods for subsurface interpretation.
Computational models are used to study how electromagnetic signals interact with different underground conditions. The resulting information can support the development and assessment of data-driven methods for estimating subsurface features.
Validation remains an important part of the workflow. Model behaviour is checked against suitable numerical references and available evidence before the results are used for interpretation.
Subsurface conditions → electromagnetic response → scientific ML → validation → interpretation
What the work produced
The work contributed to a reusable approach for combining electromagnetic simulation and data-driven subsurface analysis.
Scientific-ML interpretation
A workflow for exploring data-driven methods while maintaining a connection to physical modelling.
Validation-aware analysis
A structured approach for assessing model outputs against suitable reference evidence.
Let’s talk about what you’re building.
I’m always open to conversations about engineering, computation, deeptech products and mentoring people building or learning in these areas.