Subchannel analysis of fuel assemblies is critical for the development of lead-bismuth reactors.

This study aims to modify and optimize the COBRA subchannel program to make it suitable for lead-bismuth reactors and validate its performance.

Modifications were made to the COBRA subchannel program, involving adjusting physical properties, convective heat transfer models, friction models, and turbulence mixing models. The performance of the modified program was evaluated by comparing its numerical calculation results to experimental data. To optimize results over a wide range of mass flow rate conditions, an optimization method based on a subchannel model and coupled with a neural network was proposed, and the influence of mass flow rate on calculation accuracy was analyzed.

The comparison results demonstrate that the modified subchannel program performs well under experimental conditions, with an error of no more than 5% compared with experimental results and no more than 3% compared with FLUENT results. The application of neural networks is found to improve accuracy and reduce errors by an order of magnitude.

The optimized subchannel analysis method, derived from the modifications and neural network coupling, can accurately predict outlet temperatures for lead-bismuth reactors under a wide range of mass flow rate conditions. This method provides valuable guidance for the design of such reactors.