Integrated machine learning model for condensation flow heat transfer in smooth and enhanced tubes

This paper introduces an integrated machine learning (ML) model that combines K-means clustering (KMC), random forest (RF), and artificial neural networks (ANNs) to predict the condensation heat transfer coefficient (HTC) in horizontal enhanced tubes under various experimental conditions. A comprehensive database of 5332 experimental data points was constructed, covering a wide range of conditions. Given the important role of the flow regime in determining the thermal characteristics of this problem, the model begins by using KMC to categorise the data into five distinct flow patterns, which are then visualised. Following this categorisation, a RF model is employed to predict the probability of each data point belonging to a specific flow pattern, achieving excellent average precision, recall, and an F1 score of 0.99. Based on the flow pattern classification, five separate ANN regressors are trained and evaluated for each flow pattern, with each regressor demonstrating excellent performance metrics. The final integrated model shows good performance in interpretability, predictive accuracy, and generalisation, achieving a mean relative deviation of 1.4 %, a mean absolute relative deviation of 8 %, and an R2 of 0.98 in predicting the HTC, far surpassing empirical fitting methods. Overall, the integrated ML model exhibits good predictive performance, and can serves as an effective tool for predicting and assessing the condensation characteristics in heat exchange tubes.