ARTIFICIAL-INTELLIGENCE-BASED MULTIPHYSICS SIMULATION OF A SODIUM-COOLED FAST REACTOR WITH GEN-FOAM

抄録

The new generation of nuclear energy systems represented by liquid metal reactors has a complex process of multi-physics mutual coupling in the process of steady-state or transient operation. In addition to the neutron feedback of the coolant in the core, there are some complex flow heat transfer phenomena in the reactor. This will affect the thermal stress distribution of the structural components in the reactor, and then change the size and physical properties of the core flow channel. Therefore, it is often necessary to couple physics, thermal engineering, and structural mechanics in the design of new reactors. To quickly and accurately calculate and predict the operating state of nuclear reactors online, this paper uses machine learning combined with multi-physics coupling prediction for the digital twin of reactor operation to achieve high-speed precaution on neutron flux distribution, power distribution, and temperature fields. Then we build a multi-physics fast computing model based on model reduction technology and machine learning to achieve physical guidance. The predicted physical fields are proven to achieve high accuracy in a short time. This paper is of great significance for the design, development, operation, and subsequent realization of digital twin technology of new reactors.