Mechanism analysis of thermal stratification in large domain of fast reactor and research on intelligent accelerated prediction scheme

The phenomenon of thermal stratification affects the core residual heat dissipation capability. To explore its underlying mechanisms and develop efficient prediction methods, this paper proposes a rapid intelligent prediction model for thermal stratification based on entropy production analysis. The model first analyzes the impact of entropy production rates from different angles on transient thermal stratification and defines a dimensionless number for the self-preservation of thermal stratification (Ssp) in the upper plenum during accident transients to evaluate the stability of the thermal stratification state. The prediction of transient thermal stratification at different moments is achieved by combining Convolutional Neural Networks (CNNs) with Long Short-Term Memory Networks (LSTMs). The results indicate a direct relationship between the one-dimensional axial temperature difference heat transfer entropy production rate and thermal stratification, with the Ssp number effectively reflecting the intensity of the flow field's influence on thermal stratification at different times. Additionally, the entropy production rate data helps the CNN focus on the thermal stratification regions, effectively extracting thermal stratification features and reducing computational complexity. Compared to three-dimensional Computational Fluid Dynamics (CFD) methods, this CNN-LSTM model can accurately predict the location, contour, and intensity of thermal stratification at different moments while simultaneously reducing computational resources and time consumption.