Data-driven engineering descriptor and refined scale relations for predicting bubble departure diameter

The accurate prediction of bubble departure in subcooling flow boiling is challenging because the complex physical conditions make it difficult to predict. In particular, it is difficult to accurately predict the bubble departure diameter, which plays an essential role in subcooled flow boiling. This work, assisted by data-driven machine learning, identifies new engineering descriptors in algebraic expressions and establishes a refined scaling relation for predicting bubble departure diameter. The accuracy of the new engineering descriptor is superior to that of the single descriptor highlighted previously. Compared with previous models, the new sure independence screening and sparsifying operator (SISSO)-refined scaling relation has a better prediction performance. The results show the Mean Squared Error (MSE) with 0.0093 and the Root Mean Squared Error (RMSE) with 0.0963 for the whole database. Our trained model realizes the prediction of the orientation angle between the heat and flow, different working fluids (excluded from the database), and different heating methods of the bubble departure diameter, which is the ability that the previous model did not have. The work shows that the SISSO-refined scaling relation becomes a robust new predicting tool for bubble departure diameter.