prediction of the pressure drop of r1234yf boiling flow using machine learning techniques

In this study, it is proposed to use machine learning algorithms (mlas) to predict the pressure drop of r1234yf two-phase flow. two methods of mlas are designed and trained using a total of 108 experimental data points collected from the literature. these methods are support vector regression (svr) and multi-layer perceptron network (mlp). mass flux, vapor quality, saturation temperature, and heat flux are used as input variables, while the corresponding pressure drop is considered the output variable. the results demonstrate that both methods could successfully predict the pressure drop with a correlation coefficient (r) higher than 99%. the svm model reaches less mean square error (mse) as compared to the mlp model. also, svr and mlp models predict 99% and 100% of data within relative deviations of ±20%, respectively. by comparing the results of mlas with the literature correlations, the remarkable effect of machine learning methods in improving the pressure drop prediction accuracy is shown.