Semester 1: Problem Definition & Data Formatting

Applying DAS to UNE monitoring introduces a challenge, as all historical nuclear test recordings exist as seismometer velocity data, not as the strain-rate measurements compatible with DAS systems. This project's research foundation has been established through reviewing relevant literature spanning seismic wave physics, nuclear monitoring systems, DAS signal processing, and machine learning-based event classification. A basic data processing system was built and validated on two real DAS earthquake datasets: the 2016 M5.8 Pawnee, Oklahoma earthquake (SGY format), and the 2023 M5.0 Cook Inlet, Alaska earthquake (HDF5 format), demonstrating that the system correctly handles different file formats, removes faulty sensors, isolates seismic frequencies through filtering and automatically detects seismic events using the STA/LTA algorithm.

Semester 2: Model Training & Validation

We plan to implement the GTUNE conversion algorithm, training and validating the ML classifier and integrating source localization for the system.

Documentation & Deliverables!

Semester 1 Final Report (.PDF)

Semester 1 Final Presentation (.PDF)

View GitHub Repository