EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Experimental insights and Bayesian-optimized explainable AI prediction of mechanical properties in hybrid fiber RCA concrete

Jawad Tariq, Kui Hu, Syed Tafheem Abbas Gillani, Imtiaz Iqbal, Adnan Khan, Jiahao Zhang, Masooma Zehra Syeda, Tingyi Zhang

REVIEWS ON ADVANCED MATERIALS SCIENCE · 23.01.2026 · 65 (1)

DOI: 10.1515/rams-2025-0267 ↗

Kurz gefasst

Fachartikel in der Zeitschrift REVIEWS ON ADVANCED MATERIALS SCIENCE, online erschienen am 29.07.2026. Verfasst von Jawad Tariq und 7 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 87 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The increasing demand for concrete, coupled with the depletion of natural aggregates, has accelerated the use of recycled coarse aggregates (RCA). However, the high porosity of RCA adversely affects mechanical performance. This study develops an integrated experimental and interpretable machine learning framework to improve and predict the mechanical properties of hybrid fiber-reinforced recycled aggregate concrete. Hybrid glass/polyvinyl alcohol fibers (0.05–0.15 % by volume, 10–20 mm) were incorporated into ternary binder systems (OPC/FA/GGBS) with 0 %, 50 %, and 100 % RCA replacement. Elastic modulus (EM) and axial compressive strength (ACS) were measured from laboratory specimens, while predictive models were trained on datasets comprising 302 EM samples and 201 ACS samples, drawn from both experiments and the literature. An artificial intelligence framework incorporating six machine learning (ML) models was optimized using Bayesian optimization with Gaussian Process and Random Forest surrogates and interpreted through SHapley Additive exPlanations (SHAP). Experimental results indicate that 0.1 % hybrid fibers increased EM by 36.96 % and ACS by 38 % at full RCA replacement. Among the models, CatBoost-BO-RF achieved the highest predictive accuracy, with R 2 values of 0.9783 for EM and 0.9751 for ACS in training, while maintaining values above 0.90 during validation. SHAP analysis identified fiber length (14.2 %), RCA content (14 %), and water absorption (10.6 %) as dominant factors for EM, whereas WARA (19.9 %), NFA (15.2 %), and Wa (14.1 %) governed ACS. The results confirm that hybrid fiber incorporation significantly enhances mechanical performance and that interpretable ML provides a robust basis for the design and prediction of sustainable RCA-concrete.

Zitieren

APA

Tariq, J., Hu, K., Gillani, S. T. A., Iqbal, I., Khan, A., Zhang, J., Syeda, M. Z., & Zhang, T. (2026). Experimental insights and Bayesian-optimized explainable AI prediction of mechanical properties in hybrid fiber RCA concrete. REVIEWS ON ADVANCED MATERIALS SCIENCE, 65(1). https://doi.org/10.1515/rams-2025-0267

Vancouver

Tariq J, Hu K, Gillani STA, Iqbal I, Khan A, Zhang J, et al. Experimental insights and Bayesian-optimized explainable AI prediction of mechanical properties in hybrid fiber RCA concrete. REVIEWS ON ADVANCED MATERIALS SCIENCE. 2026;65(1). doi:10.1515/rams-2025-0267

BibTeX
@article{Tariq2026,
  author = {Tariq, Jawad and Hu, Kui and Gillani, Syed Tafheem Abbas and Iqbal, Imtiaz and Khan, Adnan and Zhang, Jiahao and Syeda, Masooma Zehra and Zhang, Tingyi},
  title = {Experimental insights and Bayesian-optimized explainable AI prediction of mechanical properties in hybrid fiber RCA concrete},
  journal = {REVIEWS ON ADVANCED MATERIALS SCIENCE},
  year = {2026},
  volume = {65},
  number = {1},
  doi = {10.1515/rams-2025-0267}
}
Kostenlos starten →