Investigation of boiling heat transfer coefficients of different refrigerants for low fin, Turbo-B and Thermoexcel-E enhanced tubes using computational smart schemes

The design and manufacture of highly efficient evaporators and heat exchangers in cooling machinery need an accurate estimation of the boiling heat transfer coefficient of the refrigerants. In the present study, the boiling heat transfer coefficients of different refrigerants were predicted using machine learning methods and were compared to the existing empirical correlations. For this purpose, four models of ANN, ANFIS, ELM and SVM were developed by using MATLAB functions. 320 data were collected on the boiling heat transfer coefficient of the refrigerants in the plain and enhanced (low fin, Turbo-B, Thermo excel-E) tubes. The percent deviation from the actual value was between − 2.05 and 1.36% for the ANN, − 4.97 and 8.72% for the ANFIS, − 38.11 and 83.01% for the ELM and − 17.35 and 78.37% for the SVM. The results show that the proposed ANN and ANFIS models are reliable models for predicting the boiling heat transfer coefficient. They have a better performance than the ELM and SVM models. The values of RMSE, AARD and R² for the best model were 74 W m⁻² K⁻¹, 0.399% and 0.99993 for the ANN, 306 W m⁻² K⁻¹, 1.117% and 0.99883 for the ANFIS, 2163 W m⁻² K⁻¹, 15.539% and 0.94191 for the ELM and 2212 W m⁻² K⁻¹, 14.905% and 0.93921 for the SVM. The intelligent algorithms of ANN and ANFIS have more accurate predictions than empirical correlations.