Improvement of diffusion layer model for steam condensation in the presence of non-condensable gas under turbulent free convection

Abstract

Steam condensation plays a crucial role in predicting the pressure behavior and hydrogen distribution in nuclear power plant containment during a postulated loss-of-coolant accident or a severe accident. The stagnant film model and diffusion layer model are two commonly used heat and transfer analogy models for predicting film-wise condensation in the presence of non-condensable gas on a vertical wall in the nuclear engineering field. This study proposed a general formula for the equivalent condensation thermal conductivity, which can be used to convert between the traditional diffusion layer model and the stagnant film model. An improved diffusion layer model is established, which can be used to analyze steam condensation in the presence of non-condensable gas on a vertical flat surface or outer surface of condenser tubes under turbulent free convection. The density difference of water vapor at the bulk and at the gas–liquid interface is the driving force for steam condensation in the improved model, the traditional diffusion layer model’s assumption, e.g., diffusion layer thickness, constant gas mixture density, and the application of modified Clausius-Clapeyron equation, are no longer used. The newly established model also divides the condensation process into liquid film heat transfer, latent heat transfer, and sensible heat transfer. The three parts of heat transfer processes are linked together by the equivalent heat flux on both sides of the gas–liquid interface. The effects of gas compressibility, suction, fog formation, and liquid film waviness have been considered in this study. The influences of gas compressibility, the thermal resistance of condensate film, and sensible heat transfer on modeling steam condensation were discussed. Compared with various sources of experimental data (619 data points), more than 97% of the model predicted results are within the ±25.0% error band, and the mean absolute relative error is 10.2%, which indicates that the newly proposed model has high accuracy and widely applicable ranges.