Argonne National Laboratory develops new monitoring system that can accurately diagnose hidden faults inside nuclear reactors

U.S. researchers have developed a new monitoring system that combines fiber-optic temperature sensors with artificial intelligence (AI) to identify hidden internal channel blockage risks before they cause costly reactor shutdowns. The technology was developed by Argonne National Laboratory (ANL), part of the U.S. Department of Energy (DOE), and is specifically designed for key heat exchanger components in molten-salt-cooled reactors (MSCRs).

As the core hub for transferring thermal energy out of the reactor core, the heat exchanger is indispensable to the safe operation of a nuclear power plant. In molten-salt-cooled reactors, high-temperature molten salt transfers the thermal energy released by the reactor core to the secondary-loop working fluid, which is then used to produce high-temperature steam and drive a steam turbine to generate electricity. However, once the temperature of the molten salt medium approaches its freezing point too closely, the salt fluid becomes highly prone to localized solidification, which can partially or completely block the thousands of tiny flow channels inside the heat exchanger and restrict coolant flow. This has prompted the research team to explore precise methods for early detection of flow channel blockage.

Argonne National Laboratory noted that conventional matrix-type heat exchangers typically contain a dense arrangement of 2,000 to 4,000 microchannels, and larger configurations may have even more. At present, the vast majority of industrial heat exchange systems only have temperature and flow monitoring points at the overall inlet and outlet, making it difficult to capture flow-state changes occurring within individual internal channels, which often allows minor blockage risks to go undetected in time. To address this, the research team innovatively proposed the engineering concept of integrating a distributed fiber-optic temperature sensing system into the load-bearing and support structures of the heat exchanger.

The system collects massive amounts of real-time thermal temperature data, which is then analyzed by dedicated AI algorithms to identify abnormal fluctuations in the thermal field, so as to detect signs of flow channel blockage caused by molten salt solidification. Dr. Alex Heifetz, a principal electrical engineer at Argonne National Laboratory and co-author of the study, pointed out that the system can achieve early detection, precise localization, and severity assessment when microchannels become partially or completely blocked, much like providing an ultra-high-resolution monitoring lens for troubleshooting internal faults in heat exchange systems.

In terms of engineering layout, the sensing optical fibers are attached and fixed to the surface of the equipment's mechanical support structures rather than being inserted directly into the flow channels and immersed in the medium. Therefore, there is no need for destructive penetration of the heat exchanger's pressure-containing shell wall, and it does not interfere with the normal thermal-hydraulic circulation of the internal working fluid. At the same time, the AI monitoring model features high explainability (Explainable AI), and when triggering an alarm, it can simultaneously provide nuclear power plant operators with the basis for fault diagnosis and the chain of decision-making reasoning, significantly enhancing the transparency of fault alarms for nuclear-grade equipment.