EUVIMEDEUVIMED entdecken
Crossref

Hybrid Metaheuristic–AI Modeling for Accurate Prediction of CO₂ Solubility in Amine–Ionic Liquid Hybrid Solvents

Soudabeh Khalilian, Alireza Goodarzi‑Mohammadi

Journal of Machine Learning Advances · 10.11.2026 · 2 (1) · 36-64

DOI: 10.66879/jmla.2026.37 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Machine Learning Advances, online erschienen am 10.11.2026. Verfasst von Soudabeh Khalilian und 1 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Reducing CO₂ emissions from fossil-fuel-based processes requires accurate and reliable predictive tools for designing advanced solvent systems for post-combustion carbon capture. In this study, an extensive experimental database of CO₂ solubility in hybrid amine–ionic liquid (IL) solvents was compiled and modeled using four hybrid metaheuristic–machine-learning frameworks: Histogram-Based Gradient Boosting combined with the Multi-Verse Optimization Algorithm (HistGBoost-MVOA), Tabular Neural Network combined with the Moth-Flame Optimization Algorithm (TabNet-MFOA), Support Vector Regression combined with the Improved Whale Optimization Algorithm (SVR-IWOA), and Extreme Learning Machine combined with the Jellyfish Optimization Algorithm (ELM-JFOA). The input variables included temperature, amine concentration, IL concentration, molecular weights of the amine and IL, and CO₂ partial pressure, while the CO₂ mole fraction in the liquid phase (xCO₂) was considered the target variable. To ensure robust model development, outlier detection was performed using the Isolation Forest algorithm, followed by Robust Scaling for data normalization. Model performance was evaluated using repeated 10-fold cross-validation and independent testing based on R², RMSE, MAE, MAPE, and PBIAS, together with Taylor diagrams, centered RMSE (cRMSE), SHAP-based interpretability analysis, and bootstrap uncertainty quantification. All developed models demonstrated excellent predictive capability (testing R² > 0.96), with SVR-IWOA consistently achieving the highest prediction accuracy and the narrowest uncertainty intervals. SHAP analysis identified the molecular weight of the ionic liquid and temperature as the most influential variables governing CO₂ solubility. The proposed modeling framework provides a robust and interpretable data-driven approach for accurately predicting CO₂ solubility and supports the design and optimization of hybrid amine–ionic liquid solvents for carbon capture applications.

Zitieren

APA

Khalilian, S., & Goodarzi‑Mohammadi, A. (2026). Hybrid Metaheuristic–AI Modeling for Accurate Prediction of CO₂ Solubility in Amine–Ionic Liquid Hybrid Solvents. Journal of Machine Learning Advances, 2(1), 36-64. https://doi.org/10.66879/jmla.2026.37

Vancouver

Khalilian S, Goodarzi‑Mohammadi A. Hybrid Metaheuristic–AI Modeling for Accurate Prediction of CO₂ Solubility in Amine–Ionic Liquid Hybrid Solvents. Journal of Machine Learning Advances. 2026;2(1):36-64. doi:10.66879/jmla.2026.37

BibTeX
@article{Khalilian2026,
  author = {Khalilian, Soudabeh and Goodarzi‑Mohammadi, Alireza},
  title = {Hybrid Metaheuristic–AI Modeling for Accurate Prediction of CO₂ Solubility in Amine–Ionic Liquid Hybrid Solvents},
  journal = {Journal of Machine Learning Advances},
  year = {2026},
  volume = {2},
  number = {1},
  pages = {36--64},
  doi = {10.66879/jmla.2026.37}
}
Kostenlos starten →