Forecasting of saturated boiling heat transfer inside smooth helically coiled tubes using conventional and machine learning techniques

Abstract

This study concerns to model the flow boiling heat transfer coefficient (HTC) in smooth helically coiled tubes. A dataset including 1035 samples was collected from 13 independent studies, enveloping a broad range of geometrical and operating conditions. The predictive capability of the earlier models was assessed for straight and coiled tubes by the analyzed database that they were not precise enough. Accordingly, a new empirical model based on the least square fitting method (LSFM) was constructed using seven input effective dimensionless factors. It was found that LSFM was not able to describe the complex and nonlinear nature of HTC in smooth helically coiled tubes. Furthermore, the intelligent method of genetic programming (GP) was utilized to obtain more accurate explicit correlation for HTC, which produced an acceptable average absolute relative error (AARE) of 17.35%. Finally, the machine learning approaches of multilayer perceptron (MLP), Gaussian process regression (GPR), radial basis function (RBF) was also implemented to model HTC in smooth coiled tubes. Although all intelligent based models provided excellent results, the GPR model outperformed the others with an average absolute relative error (AARE) of 5.93% for the tested dataset. In addition to the proposed models’ performance, the most influential factors in controlling the boiling HTC in coiled tubes were also detected.