Artificial Neural Network Model to Predicting the Boiling Heat Transfer Coefficient of R1234yf Inside the Horizontal Multiport Mini-Channel Tube

Data-based models, such as those based on machine learning, have been an interesting topic of discussion, particularly with regard to predicting the heat transfer coefficient. In this study, an application of machine learning to predict the boiling heat transfer coefficient of R1234yf inside a multiport mini-channel tube is proposed. An artificial neural network (ANN) model based on machine learning is trained and tested using data on R1234yf, considering a hydraulic diameter of 0.969 mm, saturation temperature of 6°C, mass flux of 50–500 kg/m2 s, and heat flux of 3–12 kW/m2 . Experimental data on flow boiling of R1234yf are divided into 80% for training and 20% for testing. A total of 10 dimensionless numbers (Bd, Frv , Frvo , Frlo , Rev , Revo , Relo , Wev , Wevo , Welo ) are used as input data, including the mass flux and heat flux. The predictions provided by the ANN model with hidden layer (96, 48, 24, 12) is compared with those provided by the superposition, asymptotic, and flow pattern models. The results shows that predictions obtained using the ANN model are more accurate than those provided by the superposition, asymptotic, or flow pattern models, with a mean deviation of 8.36%. The ANN model can serve as a convenient alternative for predicting the boiling heat transfer coefficient of the refrigerant in a multiport mini-channel tube