Evaluating the Thermohydraulic Performance of Microchannel Gas Coolers: A Machine Learning Approach
In this study, a numerical model of a microchannel gas cooler was developed using a segment-by-segment approach for thermohydraulic performance evaluation. State-of-the-art heat transfer and pressure drop correlations were used to determine the air and refrigerant side heat transfer coefficients and...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2025-06-01
|
| Series: | Energies |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1996-1073/18/12/3007 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | In this study, a numerical model of a microchannel gas cooler was developed using a segment-by-segment approach for thermohydraulic performance evaluation. State-of-the-art heat transfer and pressure drop correlations were used to determine the air and refrigerant side heat transfer coefficients and friction factors. The developed model was validated against a wide range of experimental data and was found to accurately predict the gas cooler capacity (Q) and pressure drop (ΔP) within an acceptable margin of error. Furthermore, advanced machine learning algorithms such as extreme gradient boosting (XGB), random forest (RF), support vector regression (SVR), k-nearest neighbors (KNNs), and artificial neural networks (ANNs) were employed to analyze their predictive capability. Over 11,000 data points from the numerical model were used, with 80% of the data for training and 20% for testing. The evaluation metrics, such as the coefficient of determination (R<sup>2</sup>, 0.99841–0.99836) and mean squared error values (0.09918–0.10639), demonstrated high predictive efficacy and accuracy, with only slight variations among the models. All models accurately predict the Q, with the XGB and ANN models showing superior performance in ΔP prediction. Notably, the ANN model emerges as the most accurate method for refrigerant and air outlet temperatures predictions. These findings highlight the potential of machine learning as a robust tool for optimizing thermal system performance and guiding the design of energy-efficient heat exchange technologies. |
|---|---|
| ISSN: | 1996-1073 |