No Overheating: Researchers Develop New Diagnostic System for Reactors and Microchips
Researchers at the National Research Nuclear University MEPhI have developed a method for early detection of changes in heat transfer.
The algorithms analyze even the slightest temperature fluctuations and other coolant parameters to warn operators of the risk of boiling before problems arise. The technology could provide the foundation for an intelligent assistant to help operate nuclear power plants and manage cooling in high-power electronics.
The Danger of Hidden Boiling
There is still no single, universally accepted mathematical model of liquid boiling in modern science. Yet understanding this process remains critically important for nuclear power. In VVER-type reactors, bulk boiling of the coolant can lead to a highly dangerous phenomenon known as a heat-transfer crisis.
This occurs when heat removal in the reactor core deteriorates sharply because of vapor blanketing around the fuel elements. A dense film of vapor forms on the metal surface, acting as a thermal insulator. The temperature of the heated component then rises rapidly, potentially causing its cladding to fail or melt. The consequences of an error in such circumstances can be severe.
Conventional protection systems respond after overheating has already occurred. MEPhI engineers propose shifting the focus to the earliest stage of the problem. Their method can identify warning signs of a heat-transfer crisis at a point when the process can still be safely halted or corrected using standard operating procedures.
Analyzing Invisible Fluctuations
The main challenge in diagnostics is the lack of direct measurement data. Pressure and coolant flow rate are continuously monitored at nuclear power plants, but the temperature of the fuel element surface itself is not measured in routine operation. Sensors are installed only on experimental research fuel assemblies.
MEPhI researchers have proposed a method based on analyzing temperature fluctuations at the heat-transfer surface. The algorithms search for hidden mathematical correlations between these fluctuations and parameters that are always available, such as liquid pressure and flow rate.
Identifying mathematical relationships between indirect and direct parameters requires substantial computing power. Engineers must filter out industrial noise and isolate useful signals against the background of natural equipment vibrations. The method predicts hazardous conditions in real time, before they physically develop, without requiring additional equipment to be installed in the reactor core.
A Digital Adviser for Reactor Operators
The ultimate goal is to develop a complete hardware and software system that will serve as an intelligent adviser to nuclear power plant operators. The software is intended to analyze streams of telemetry data and alert personnel to potential risks well before emergency protection systems are triggered. Predictive systems of this kind are expected to help prevent unplanned shutdowns, potentially saving millions of rubles.
The project is currently at the mathematical modeling and training-data development stage. Researchers are conducting a series of physical experiments at test facilities operated by industry partners OKB Gidropress and NIKIET. The method has already been tested on experimental rigs using water and liquid nitrogen, demonstrating that it works under different thermodynamic conditions.
Looking ahead, the MEPhI team plans to integrate artificial intelligence technologies into the system. Neural networks will be trained on large datasets to identify complex nonlinear relationships. This could turn the system into a fully fledged digital assistant capable of recommending optimal ways to prevent accidents.
From Fast Reactors to Microchips
Although the method was originally developed for nuclear power applications, its underlying physics is universal. Fluctuation theory applies to all liquids, opening the door to applications in related industries. Developing diagnostic algorithms domestically also reduces critical infrastructure's dependence on foreign systems.
The technology could be particularly valuable for the Proryv (Breakthrough) project, which uses liquid-metal coolants such as sodium, lead, and bismuth. These metals must not be allowed to boil, making continuous monitoring of their thermodynamic state a critical task. The method is also suitable for cooling systems used in superconducting magnets at large particle accelerators, where a failure in heat removal can cause superconductivity to be lost and lead to equipment failures or accidents.
Microelectronics is another promising application. Today's high-performance chips generate enormous amounts of heat, making air cooling insufficient. Liquid cooling systems are increasingly used instead. MEPhI's algorithms could help prevent localized processor overheating, improving the operational stability of supercomputers and data centers.
