Development of Geiger-Müller network for anomaly detection and tracking for accident preparedness at KAERI

Monitoring environmental radiation around nuclear facilities is critical for safety and regulatory compliance. Traditional methods, such as environmental radiation monitoring using high-pressure ion chambers and thermoluminescent dosimeters, have limitations with regard to cost, complexity, and response time. To address these issues, we developed a compact Geiger-Müller (GM) counter-based detector network for real-time radiation monitoring at the Korea Atomic Energy Research Institute (KAERI). The developed GM detector module is operated using a battery and a solar panel to ensure maintenance-free operation and is equipped with LTE wireless communication. The Daejeon KAERI site spans approximately 1.42 km², where a total of 50 GM modules were installed, forming a high-resolution radiation monitoring network. In addition, convolutional neural network-based radiation anomaly detection and source-tracking models were developed to enhance the monitoring capabilities. The anomaly-detection model achieved an accuracy of 0.9999 and an area under the receiver operating characteristic curve of 0.9999, effectively distinguishing between normal and anomalous radiation. The source-tracking model predicted source locations with an average error of 3.44 m for the test set. In field experiments using a low-intensity ¹³⁷Cs source, the average error was 54.73 m. The proposed cost-effective, high-resolution radiation mapping solution can be easily deployed and maintained, ensuring comprehensive coverage and timely detection of radiation anomalies.