Application of machine learning to the prediction of the boiling heat transfer coefficient of R32 inside a multiport mini-channel tube

Abstract

The possibility of using machine learning to predict the heat transfer coefficient is becoming more evident. In fact, artificial neural networks (ANN) are widely used in heat transfer coefficient research. In this study, an ANN was used in the dataset training and testing of the boiling heat transfer coefficient of R32 inside a horizontal multiport mini-channel tube with a hydraulic diameter of 0.969 mm and an aspect ratio of 0.6. A mass flux range of 50–500 kg m−2 s−1, heat flux of 3–6 kW m−2, saturation temperature of 6 °C, and vapor quality up to 1 were applied as experimental conditions. The superposition, asymptotic, and flow pattern models were used to assess the experimental data. The ANN model with hidden layers (96,72,48,24) and 16 input parameters (Revo, Relo, Bd, Frvo, Wevo, Frlo, Welo, Rev, Frv, Rel, Wel, Prv, Xtt, Co, Prl, and Bo) was included in the prediction of the boiling heat transfer coefficient of R32 inside a horizontal multiport mini-channel tube and achieved better results than the empirical correlation models with a mean deviation of 6.35%. Results indicate that ANN models can be applied to improve the prediction accuracy of the boiling heat transfer coefficient, especially in multiport mini-channel tubes.