Parameter Calculation Method of Porous Media Based on BP Neural Network

Abstract

There are a large number of equipment densely arranged in liquids in nuclear power plants, such as fuel assemblies, steam generator heat transfer tubes, spent fuel storage and transportation racks, etc. These equipment are complex in shape and compact in arrangement and have strong fluid-structure coupling effects under excitations, so the calculation is computationally intensive. For such complex structures, the use of porous media models is an important means of structure simplification. The parameters of porous media are often calculated by CFD modeling, and the calculation process is complicated and time-consuming. BP neural network has strong nonlinear mapping capability and can be used to calculate the parameters of porous media. For different racks designs, the gap arrangement is different, and the fluid-structure coupling parameters are also different. Therefore, it is necessary to study the fluid-structure coupling parameters of square tube bundles such as racks. Taking porous storage racks as an example, by building different CFD models, 1366 sets of valid data were obtained for training. This paper uses BP neural network to study the porous medium parameters required for fluid-structure interaction of porous racks. Compared with the CFD calculation method of fine modeling, the calculation error of the additional mass of the porous media model established by the porous media parameters predicted by the neural network is controlled at about 10%. The research results provide a reference for the fast calculation of porous media parameters and fluid-structure interaction.