Machine learning approach for predicting refrigerant two-phase pressure drop inside Brazed Plate Heat Exchangers (BPHE)

• Published in: International journal of heat and mass transfer Vol. 163; p. 120450
• Main Authors: Longo, Giovanni A., Mancin, Simone, Righetti, Giulia, Zilio, Claudio, Ceccato, Riccardo, Salmaso, Luigi
• Format: Journal Article
• Language: English
• Published: Oxford Elsevier Ltd 01.12.2020; Elsevier BV
• Subjects: Boiling; Boiling points; BPHE; Condensation; GBM; Heat exchangers; Machine learning; Performance prediction; Plate heat exchangers; Pressure drop; Refrigerants; GBM; BPHE; Pressure drop; Boiling; Condensation; Machine learning
• ISSN: 0017-9310; 1879-2189
• DOI: 10.1016/j.ijheatmasstransfer.2020.120450

•A GBM model for predicting two-phase frictional pressure gradient inside BPHE is presented.•The GBM model accounts for BPHE geometry, operating conditions and refrigerant properties.•A database of 1624 boiling and 925 condensation pressure gradient data-points is presented.•The database includes 16 plate geometries, 4 natural refrigerants, 6 low-GWP refrigerants, and 6 traditional refrigerants.•The GBM model predicts the whole database with a MAPE of 6.6%. This paper presents a Gradient Boosting Machines (GBM) model for predicting refrigerant two-phase frictional pressure gradient inside Brazed Plate Heat Exchangers (BPHE) based on an extensive database that includes 1624 boiling data-points, 925 condensation data-points, 16 different plate geometries, and 16 different refrigerants (including 4 natural refrigerants and 6 other low-GWP refrigerants). The model accounts for the effect of plate geometry, operating conditions and refrigerant properties. The model is able to reproduce the whole database with a Mean Absolute Percentage Error (MAPE) of 6.6%. The GBM model exhibits a better predictive performance than the state-of-the-art analytical-computational procedures for two-phase pressure drop inside BPHE available in the open literature. The characteristic parameters of the GBM model are thoroughly reported in the paper.