Researchers of the Oak Ridge National Laboratory (ORNL) explored the possibility of using seismic and acoustic data recorded 50 meters away from a nuclear research reactor to predict whether the reactor was on or off with 98% accuracy.
Using ORNL’s high flux isotope reactor, the researchers placed remote sensors near the facility and continuously recorded data. Their published results allowed them to determine:
- whether the reactor was on or off with 98% accuracy;
- reactor power levels with an accuracy of about 66%;
- sources of seismic and acoustic activity (whether it came from reactor-specific operations or other sources, such as equipment vibrations from a nearby cooling tower).
The researchers compared machine learning algorithms to find out what worked best for estimating the reactor power level from specific seismoacoustic signals. The algorithms were programmed for using seismic-only, acoustic-only, and both types of data collected within a year. They revealed that combined data provided the best results.
The new seismoacoustic technique can be used as a protective measure for sensitive facilities and nonproliferation applications and as a monitoring tool for building structural health.
According to Oak Ridge National Laboratory