EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Machine Learning-Enhanced Modeling of Heavy Metal Adsorption onto Coal Fly Ash-Derived Zeolite P

Benito A. Hernández-Guerrero, Lorena Martínez, Gabriel Peña-Rodríguez, Fernando Trejo

Water · 02.04.2026 · 18 (7) · 857

DOI: 10.3390/w18070857 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Water, online erschienen am 02.04.2026. Verfasst von Benito A. Hernández-Guerrero und 3 weiteren Autor:innen. Die Arbeit wurde bisher 7-mal zitiert (Crossref) und nennt 119 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Instituto Politécnico Nacional.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Zeolite P was synthesized by hydrothermal treatment of coal fly ash and applied to the individual removal of six heavy metals (Pb2+, Ni2+, Cu2+, Cr3+, Hg2+, Cd2+) from aqueous solutions. Characterization by SEM-EDS, FTIR, BET, XRD, zeta potential, and XPS revealed a BET surface area of 30 m2/g, Si/Al ratio of 1.63, and pHpzc of 3.2. Batch experiments at the natural solution pH of 3.9 in all cases (C0 = 10, 100, 200 mg/L; t = 1–60 min) yielded an apparent selectivity sequence at C0 = 200 mg/L of Hg2+ (10.47 mg/g) > Pb2+ (9.12) > Ni2+ (2.18) > Cr3+ (2.05) > Cu2+ (1.82) > Cd2+ (1.26), where Hg2+ and Pb2+ reached near-equilibrium while the remaining metals were still approaching it at t = 60 min. Weber–Morris and Boyd analyses confirmed three sequential diffusion stages with a concentration-dependent shift from film to intraparticle diffusion control through the narrow GIS channels (3.1 × 4.5 Å). Ion exchange was identified as the dominant mechanism based on convergent kinetic, diffusion, XPS, and selectivity–electronegativity evidence (r = +0.76). A leakage-free machine learning framework combining physicochemical descriptors with experimental variables was tested under three cross-validation strategies of increasing stringency. Gradient Boosting achieved R2 = 0.979 ± 0.043 (repeated K-Fold) and R2 = 0.880 on six completely held-out kinetic curves. An ablation study confirmed that physicochemical descriptors are essential (experimental-only models yielded negative R2). SHAP feature importance rankings were consistent with established ion exchange selectivity theory. This work demonstrates that group-level validation, physics-informed descriptors, and systematic ablation testing are able to identify both the capabilities and the boundaries of small-dataset ML when testing for metal kinetics prediction.

Förderung

  • Instituto Politécnico Nacional

Zitieren

APA

Hernández-Guerrero, B. A., Martínez, L., Peña-Rodríguez, G., & Trejo, F. (2026). Machine Learning-Enhanced Modeling of Heavy Metal Adsorption onto Coal Fly Ash-Derived Zeolite P. Water, 18(7), 857. https://doi.org/10.3390/w18070857

Vancouver

Hernández-Guerrero BA, Martínez L, Peña-Rodríguez G, Trejo F. Machine Learning-Enhanced Modeling of Heavy Metal Adsorption onto Coal Fly Ash-Derived Zeolite P. Water. 2026;18(7):857. doi:10.3390/w18070857

BibTeX
@article{HernndezGuerrero2026,
  author = {Hernández-Guerrero, Benito A. and Martínez, Lorena and Peña-Rodríguez, Gabriel and Trejo, Fernando},
  title = {Machine Learning-Enhanced Modeling of Heavy Metal Adsorption onto Coal Fly Ash-Derived Zeolite P},
  journal = {Water},
  year = {2026},
  volume = {18},
  number = {7},
  pages = {857},
  doi = {10.3390/w18070857}
}
Kostenlos starten →