Machine learning modeling of ceramic beads: Impact of PVA and water on porosity and strength

Basit Öğe Kaydı
dc.authorid0000-0003-4577-028X
dc.contributor.authorBaydir, Enver
dc.contributor.authorSen, Hasan
dc.contributor.authorAras, Omuer
dc.date.accessioned2026-02-08T15:15:23Z
dc.date.available2026-02-08T15:15:23Z
dc.date.issued2025
dc.departmentBursa Teknik Üniversitesi
dc.description.abstractIn this study, ceramic beads with adjustable apparent porosity and compressive strength were successfully produced using the Na-alginate gelation method, with kaolin and zeolite as the raw materials. The effects of varying PVA and water ratios on apparent porosity, water absorption, and compressive strength were thoroughly examined. Zeolite exhibited a maximum apparent porosity of 33.93 %, while kaolin reached 55.06 %. Regarding compressive strength, zeolite showed superior performance with a value of 10.49 MPa, compared to 0.44 MPa for kaolin. Zeolite-based ceramic beads exhibited lower porosity but significantly higher compressive strength than kaolin-based beads, demonstrating enhanced mechanical properties. Characterization using X-ray diffraction (XRD), scanning electron microscopy (SEM), and Fourier-transform infrared (FT-IR) spectroscopy provided insights into the crystalline structure, morphology, and chemical composition of the beads. These techniques revealed detailed structural properties of the ceramic beads. To predict key properties such as apparent porosity (AP), water absorption (WA), and compressive strength (CS), machine learning models including XGBoost, Random Forest, Support Vector Regression (SVR), and Artificial Neural Networks (ANN) were employed. The performance of these models was assessed using R2, MAE (Mean Absolute Error), MSE (Mean Squared Error), and RMSE (Root Mean Squared Error) metrics. Among these models, SVR provided the most reliable predictions, while XGBoost also yielded competitive results. To improve interpretability, SHAP (Shapley Additive Explanations), LIME (Local Interpretable Model-agnostic Explanations), and Type II ANOVA (Analysis of Variance) were used. SHAP and Type II ANOVA identified 'Ceramic' as the most impactful variable, with LIME revealing its local effects, enhancing transparency in model decision-making.
dc.identifier.doi10.1016/j.mtcomm.2025.112844
dc.identifier.issn2352-4928
dc.identifier.scopus2-s2.0-105005488691
dc.identifier.scopusqualityQ2
dc.identifier.urihttps://doi.org/10.1016/j.mtcomm.2025.112844
dc.identifier.urihttps://hdl.handle.net/20.500.12885/5757
dc.identifier.volume46
dc.identifier.wosWOS:001499656400002
dc.identifier.wosqualityQ2
dc.indekslendigikaynakWeb of Science
dc.indekslendigikaynakScopus
dc.language.isoen
dc.publisherElsevier
dc.relation.ispartofMaterials Today Communications
dc.relation.publicationcategoryMakale - Uluslararası Hakemli Dergi - Kurum Öğretim Elemanı
dc.rightsinfo:eu-repo/semantics/closedAccess
dc.snmzWOS_KA_20260207
dc.subjectAdjustable porosity
dc.subjectCeramic beads
dc.subjectCharacterization
dc.subjectModelling
dc.subjectMachine Learning
dc.titleMachine learning modeling of ceramic beads: Impact of PVA and water on porosity and strength
dc.typeArticle

Dosyalar

Koleksiyon

WoS İndeksli Yayınlar Koleksiyonu
Scopus İndeksli Yayınlar Koleksiyonu