Prediction and optimization of condensation heat transfer coefficients and pressure drops of R134a inside an inclined smooth tube

In this study, an adaptive neuro-fuzzy inference system (ANFIS) is proposed for the prediction and optimization of condensation heat transfer coefficient and pressure drops along an inclined smooth tube. The performance of three ANFIS structure identification methods, grid partitioning (GP), a subtractive clustering method (SCM), and fuzzy C-means (FCM) clustering, were examined. For training the proposed ANFIS model, an in-house experimental database was utilised. Three statistical criteria, the mean absolute error (MAE), mean relative error and root mean square error were used to evaluate the accuracy of each method. The results indicate that the GP structure identification method has the lowest number of training errors for both the pressure drop, i.e., MAE = 6.4%, and condensation heat transfer coefficient, i.e., MAE = 2.3%, models. In addition to the ANFIS model, numerical simulations were also conducted to assess the accuracy and capability of the proposed model. The comparison shows that the CFD simulation results have better accuracy for the specified operating conditions. However, the errors of both the CFD and ANFIS methods were within the uncertainties of the experimental data. It was therefore concluded that the ANFIS model is useful in obtaining faster and reliable results. Finally, the optimization results showed a possible optimum point at a mass flux of 100 kg/m2 s, saturation temperature of 36.2 °C, downward inclination angle of −15° and a vapour quality of 0.48. At this condition the pressure drop is almost zero.