Detecting and Identifying Nuclear Reactor Operational Events using Seismo-acoustic Signals

Conventional facility monitoring involves a direct connection between the observer and the phenomena of interest. However, traditional methods of monitoring become unreliable when attempting to examine occurrences that occur in heavily obscured areas and under obfuscating conditions. Seismo-acoustic waves are less hindered by physical barriers. Thus, developing new approaches using seismo-acoustic can advance monitoring operations - all without any additional input from the operator. Past applications have used seismic and acoustic waves for monitoring the occurrence of remote and concealed nuclear explosions as part of the Comprehensive Nuclear-Test-Ban Treaty. Here, we apply similar methods and scientific understanding to detect and categorize seismo-acoustic signals created by operations conducted at a nuclear research reactor. Specifically, we analyze signals from logged operations at the High Flux Isotope reactor (HFIR) hosted at Oak Ridge National Lab (ORNL) and construct a catalog of operational events. Since July 2017, the seismology and acoustics team at ORNL has been collecting seismo-acoustic data near HFIR. Recordings have been obtained using a 3-component seismic and 3-component acoustic recording station, WACO, located near HFIR. We use a Power Spectral Density (PSD) detector to identify operational events that generate recordable seismo-acoustic signals. Comparing to the traditional short-term-averaging over long-term-averaging (STA/LTA) technique, the PSD detector can detect the associated frequencies in the seismo-acoustic signals in addition to the time of occurrence. Preliminary results show a good match between PSD detections and operational logs for certain events. Additionally, we plan to document missed detections to understand the limitations and potential errors that exist when monitoring nuclear reactors with this method.