Machine Learning Methods to Catalog Source from Diverse, Widely Distributed Sensors

Air Force Technical Applications Center (AFTAC) seeks robust and accurate solutions to monitor seismic signals and flag likely nuclear explosions for human analysts. The current system gets overwhelmed during large earthquake aftershocks, performs poorly for smaller events, and cannot dynamically incorporate new sensors. Global Technology Connection, Inc. and its university partners have demonstrated the feasibility of a deep machine learning tool that can fill these gaps by training on historical datasets with its high accuracy discriminator. In phase II, the team will develop an automated framework that contains a phase associator and search engine that will coordinate with the discriminator. The tool will enable the real-time operation to automate detection, association, geo-location, and event classification using incoming signals from a global network of seismic sensors. This robust and accurate tool would flag highly likely nuclear explosions in supporting AFTAC's mission to monitor nuclear treaty compliance.