Saved in:
| Main Authors: | , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2409.07811 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866913905723310080 |
|---|---|
| author | Kuberan, Vijay Gedupudi, Sateesh |
| author_facet | Kuberan, Vijay Gedupudi, Sateesh |
| contents | Surface modification results in substantial improvement in pool boiling heat transfer. Thin film-coated and porous-coated substrates, through different materials and techniques, significantly boost heat transfer through increased nucleation due to the presence of micro-cavities on the surface. The existing models and empirical correlations for boiling on these coated surfaces are constrained by specific operating conditions and parameter ranges and are hence limited by their prediction accuracy. This study focuses on developing an accurate and reliable Machine Learning (ML) model by effectively capturing the actual relationship between the influencing variables. Various ML algorithms have been evaluated on the thin film-coated and porous-coated datasets amassed from different studies. The CatBoost model demonstrated the best prediction accuracy after cross-validation and hyperparameter tuning. For the optimized CatBoost model, SHAP analysis has been carried out to identify the prominent influencing parameters and interpret the impact of parameter variation on the target variable. This model interpretation clearly justifies the decisions behind the model predictions, making it a robust model for the prediction of nucleate boiling Heat Transfer Coefficient (HTC) on coated surfaces. Finally, the existing empirical correlations have been assessed, and new correlations have been proposed to predict the HTC on these surfaces with the inclusion of influential parameters identified through SHAP interpretation.
Keywords: Pool boiling, Thin film-coated, Porous-coated, Heat transfer coefficient, Machine learning, CatBoost, SHAP analysis |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2409_07811 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Modelling of nucleate pool boiling on coated substrates using machine learning and empirical approaches Kuberan, Vijay Gedupudi, Sateesh Applied Physics Materials Science Surface modification results in substantial improvement in pool boiling heat transfer. Thin film-coated and porous-coated substrates, through different materials and techniques, significantly boost heat transfer through increased nucleation due to the presence of micro-cavities on the surface. The existing models and empirical correlations for boiling on these coated surfaces are constrained by specific operating conditions and parameter ranges and are hence limited by their prediction accuracy. This study focuses on developing an accurate and reliable Machine Learning (ML) model by effectively capturing the actual relationship between the influencing variables. Various ML algorithms have been evaluated on the thin film-coated and porous-coated datasets amassed from different studies. The CatBoost model demonstrated the best prediction accuracy after cross-validation and hyperparameter tuning. For the optimized CatBoost model, SHAP analysis has been carried out to identify the prominent influencing parameters and interpret the impact of parameter variation on the target variable. This model interpretation clearly justifies the decisions behind the model predictions, making it a robust model for the prediction of nucleate boiling Heat Transfer Coefficient (HTC) on coated surfaces. Finally, the existing empirical correlations have been assessed, and new correlations have been proposed to predict the HTC on these surfaces with the inclusion of influential parameters identified through SHAP interpretation. Keywords: Pool boiling, Thin film-coated, Porous-coated, Heat transfer coefficient, Machine learning, CatBoost, SHAP analysis |
| title | Modelling of nucleate pool boiling on coated substrates using machine learning and empirical approaches |
| topic | Applied Physics Materials Science |
| url | https://arxiv.org/abs/2409.07811 |