Advancing computational evaluation of adsorption via porous materials by artificial intelligence and computational fluid dynamics
Abstract A combination of artificial intelligence (AI) and computational fluid dynamics was carried out to advance the modeling of adsorption separation processes. A comparative examination of three AI-based regression models including Gaussian Process Regression (GPR), Multi-layer Perceptron (MLP),...
Saved in:
| Main Authors: | , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2025-08-01
|
| Series: | Scientific Reports |
| Subjects: | |
| Online Access: | https://doi.org/10.1038/s41598-025-15538-z |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract A combination of artificial intelligence (AI) and computational fluid dynamics was carried out to advance the modeling of adsorption separation processes. A comparative examination of three AI-based regression models including Gaussian Process Regression (GPR), Multi-layer Perceptron (MLP), and Polynomial Regression (PR) was carried out to predict chemical concentrations of solute in a dataset with two input variables (x and y) and one output feature (C in mol/m3). Employing gradient-based hyperparameter optimization, the results reveal that MLP outperforms GPR and PR with a significantly higher R2 score (MLP: 0.999, GPR: 0.966, PR: 0.980) and lower RMSE (MLP: 0.583, GPR: 3.022, PR: 2.370). Moreover, MLP demonstrates the lowest Average Absolute Relative Deviation (AARD%) at 2.564%, compared to GPR’s 18.733% and PR’s 11.327%. Five-fold cross-validation confirms MLP’s reliability (R² = 0.998 ± 0.001, RMSE = 0.590 ± 0.015). These findings underscore the practical utility of machine learning models, especially MLP, for accurate chemical concentration in environmental monitoring and process optimization with particular application for adsorption process. |
|---|---|
| ISSN: | 2045-2322 |