Performance prediction of electronic fan and water pump of engine cooling system based on joint simulation and machine learning
Abstract The performance of fans and pumps is pivotal to the efficiency and responsiveness of the engine cooling system. In this study, a joint simulation model incorporating a detailed engine cooling system was developed and calibrated using vehicle road cycle tests, and the predictive capabilities...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2025-05-01
|
| Series: | Scientific Reports |
| Subjects: | |
| Online Access: | https://doi.org/10.1038/s41598-025-00313-x |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract The performance of fans and pumps is pivotal to the efficiency and responsiveness of the engine cooling system. In this study, a joint simulation model incorporating a detailed engine cooling system was developed and calibrated using vehicle road cycle tests, and the predictive capabilities of four different machine learning models for water pump and fan speeds were systematically evaluated. Calibration results indicate that the simulated speeds deviate from experimental values by less than 2%, accurately reflecting real-world driving conditions. Moreover, the instantaneous fuel consumption closely mirrors the experimental data, and the error in cumulative fuel consumption under the NEDC is limited to 1.4%. The simulation outcomes for the cooling system remain within a 5% error margin, thus meeting the designated calibration criteria. Among the water pump speed prediction models, the SVR model demonstrates the highest accuracy and reliability, achieving a test mean square error (MSE) of 0.03986. In contrast, for fan speed prediction, the RF model delivers superior performance with a test MSE of 0.00052, aligning closely with experimental observations without overfitting. Consequently, the RF model proves to be the most accurate and reliable approach for fan speed prediction. These findings provide a solid theoretical basis and a robust modeling foundation for intelligent management of engine cooling systems. |
|---|---|
| ISSN: | 2045-2322 |