Reliable models for calculating the condensation heat transfer inside smooth helical tubes of different flow directions utilizing smart computational techniques

Condensers with helical tubes have received much attention in diverse industries. The optimal design of the mentioned equipment necessitates predictive tools for calculating the condensation heat transfer coefficient (HTC). However, literature models are applicable only to specific operational and geometrical conditions. The current study aims at developing reliable models for the condensation HTC within smooth helical tubes at all flow directions. Two machine learning (ML) techniques, namely Support Vector Machine (SVM) and Gaussian Process Method (GPM) were implemented to accomplish this target. To design and validate the models, 563 HTC data, encompassing a wide spectrum of conditions, were gathered from 10 experimental studies. While both SVM and GPM tools provided excellent predictions, the latter achieved the highest accuracy with mean absolute percentage error (MAPE) and R² value of 3.36% and 99%, respectively, for the testing dataset. Also, more than 96% of the HTC values calculated by the GPM model were situated within a ± 5% error margin. The accuracy of the literature correlations was also analyzed based on the collected data, and it was found that all of them showed MAPE values exceeding 25% from the experimental data. Moreover, unlike the previous models, the novel ML tools allowed the prediction of HTC for all flow directions with adequate precision. Also, they were capable of describing the physical variations of the condensation HTC versus operational factors. Finally, the dominant dimensionless parameters governing the two-phase Nusselt number in helical tubes were identified based on a sensitivity analysis.