Advanced Thermal Monitoring in Scaled Reactors using Deep Learning-Enhanced Drone IR Imaging

Abstract

For enhance surveillance capability within nuclear power plant and assist expert in hazardous region, this study presents a combined technology of drone-equipped infrared (IR) imaging and segregation deep learning for remote and smart monitoring of nuclear power plant. The test facility, designed to replicate the thermal-hydraulic phenomena of the OPR1000/APR1400 at lower pressures and temperatures, was monitored using drones equipped with dual vision and IR cameras. This setup allowed for the collection of detailed imagery under various test scenarios, including normal operation, transient operation such as all reactor coolant pumps trip and station blackout conditions. The captured images were analyzed using advanced deep learning algorithms, including object detection and instance segmentation models. These models were trained to identify normal states and abnormal states within the reactor model, providing enhanced insights into the dynamic thermal-hydraulic behavior of the system. The application of color-coded bounding boxes and segmentations in the imagery further facilitated the identification of subtle anomalies and component faults, which are challenging to detect through traditional monitoring methods. Our approach demonstrates an advancement in the use of AI and drone technology for real-time monitoring and analysis in safety-critical nuclear reactor environments. The combination of these technologies in a scaled IET setting offers a more comprehensive and automated method for thermal-hydraulic analysis, contributing to improved safety and efficiency in nuclear reactor operations.