Prediction of Condensation Heat Transfer of Passive Cooling Systems in Nuclear Power Plants through Machine Learning

The accurate prediction of steam-air mixture condensation plays crucial role in managing severe accidents such as main steam line break (MSLB) and loss of coolant accident (LOCA) using passive cooling systems within nuclear power plant containment. However, this prediction task remains challenging due to the intricate characteristics of two-phase flow phenomena and the absence of comprehensive underlying physics. While numerous empirical correlations have been proposed to estimate steam condensation heat transfer coefficients (HTC), there is still a need for more precise and reliable models because of limited range of applicability. This study addresses this challenge through a multidisciplinary approach that leverages machine learning techniques on a large dataset collected from existing literature. A comprehensive dataset consisting of 2658 experiments collected from published sources is employed to develop the model. Five different machine learning models (ML), namely artificial neural networks (ANN), k-nearest neighbors (KNN), support vector regression (SVR), random forest (RF) and Adaboost are utilized to predict steam condensation HTC. The findings of this investigation illustrate the superiority of ML models, particularly the ANN, in accurately predicting steam condensation HTC with wide range of applicability. These ML models outperform traditional empirical correlation commonly used for such predictions. By leveraging the power of ML, this study provides a more accurate and reliable approach to estimating steam condensation HTC under air-inclusive conditions.