ANN-based correlation for frictional pressure drop of non-azeotropic mixtures during cryogenic forced boiling

•ANN accurately calculates pressure drop during forced boiling of N2-hydrocarbons.•ANN inputs are easily measured parameters commonly used in fluid mechanics.•Model has proven applicability to laminar, transitional and turbulent flow.•The model robustness in predicting pressure drop during boiling was highlighted.•ANN greatly outperforms three correlations previously selected as most accurate. A crucial aspect of Joule-Thomson cryocooler analysis and optimization is the accurate estimation of frictional pressure drop. This paper presents a pressure drop model for boiling of non-azeotropic mixtures of nitrogen with hydrocarbons (e.g., methane, ethane, and propane) in microchannels. These refrigerant mixtures are important for their applicability in natural gas liquefaction plants. The pressure drop model is based on computational intelligence techniques, and its performance is evaluated with the mean relative error (mre), and compared with three correlations previously selected as most accurate: Awad and Muzychka; Sun and Mishima; and Cicchitti et al. Comparison between the proposed artificial neural network (ANN) model and the three correlations shows the advantages of the ANN to predict pressure drop for non-azeotropic mixtures. Existing correlations predict experimental data within mre = 23.9–25.3%, while the ANN has mre = 8.3%. Additional features of the ANN model include: (1) applicability to laminar, transitional and turbulent flow, and (2) demonstrated applicability to experiments not used in the training process. Therefore, the ANN model is recommended for predicting pressure drop due to accuracy and ease of applicability.