Saturated flow boiling frictional pressure drop inside smooth macro and mini/micro channels: a new predictive tool using CatBoost and XGBoost
Reliable prediction of the saturated boiling frictional pressure drop is fundamental to optimum evaporator design. Hence, frictional pressure drop is a critical parameter for improving the efficiency of the evaporator. The main constraints of the empirical or semi-empirical predictive tools, updated...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-09-01
|
| Series: | Case Studies in Thermal Engineering |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2214157X25009153 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Reliable prediction of the saturated boiling frictional pressure drop is fundamental to optimum evaporator design. Hence, frictional pressure drop is a critical parameter for improving the efficiency of the evaporator. The main constraints of the empirical or semi-empirical predictive tools, updated using non-linear statistical regression techniques, lie in their reduced predictive effectiveness when applied to unseen data. New intelligent tools can contribute to help overcome these constraints by improving accuracy and the generalization capability of the tool. In this study, a new predictive tool for estimating the frictional pressure drop combining XGBoost and CatBoost is built using a database of 6199 high quality points collected from 56 experimental studies reported in the literature. The database involves a wide range of operating conditions, hydraulic diameter from 0.15 to 14 mm, saturation temperature from −40 to 145 °C, heat flux from 3 to 1100 kW/m2, mass velocity from 30 to 1200 kg/m2s, thermophysical properties with 24 fluids, capturing reduced pressure from 0.00060 to 0.70 and incorporating different geometries. A total of 22 input parameters were selected for the training process by performing a comparative analysis between several feature selection techniques. The result indicate that the new predictive tool captures 89.52 % of the points within ±30 % error bands with technical statistical metrics MAE = 1.901, MSE = 11.956, RMSE = 3.458, MAPE = 14.204, R2 = 0.992. The proposed predictive tool demonstrates a powerful high performance and accurately captures the physical trend of frictional pressure drop versus vapor quality. |
|---|---|
| ISSN: | 2214-157X |