EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Prediction of Bond Strength in Corroded Reinforced Concrete Using SVM and XGB Methods

Zhi-Qiang Chen, Zhuang Chen, Ying-Zi Zhong iD

Materials · 08.05.2026 · 19 (10) · 1928

DOI: 10.3390/ma19101928 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Materials, online erschienen am 08.05.2026. Verfasst von Zhi-Qiang Chen und 2 weiteren Autor:innen. Beteiligte Einrichtung: School of Architecture and Civil Engineering, Chengdu University, Chengdu 610106, China. Die Arbeit wurde bisher 3-mal zitiert (Crossref) und nennt 47 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Key Research and Development Program of China.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The bond strength of corroded reinforced concrete (CRC) structures is critical for structural safety and long-term durability. However, the corrosion-induced bond degradation process is influenced by multiple, coupled factors and exhibits complex, nonlinear behavior, making it difficult for traditional theoretical models to provide accurate predictions. To address this challenge, this study proposes a novel, unified prediction framework based on machine learning techniques. A total of 391 experimental datasets were collected and compiled, covering key parameters including bond strength, reinforcing bar diameter, yield strength, concrete cover thickness, concrete compressive strength, mass loss rate due to corrosion, and the presence of stirrups. Support Vector Machine (SVM) and Extreme Gradient Boosting (XGBoost) algorithms were employed to develop predictive models for bond strength. Model training and testing were performed using 10-fold cross-validation. Furthermore, the SHapley Additive exPlanations (SHAP) approach was introduced to enhance model interpretability and quantitatively assess the influence of each input feature, revealing that mass loss rate and bar diameter are the dominant factors. This study effectively bridges the research gap between high-precision black-box algorithms and the need for physical interpretability in engineering. The results demonstrate that (1) the proposed XGBoost model significantly outperforms traditional empirical formulations, achieving a high coefficient of determination (R2 = 0.893) and a much lower coefficient of variation (25.85%) on the testing set, and (2) the SHAP analysis reveals that the machine learning predictions are highly consistent with established physical mechanisms, successfully capturing the negative impact of splitting tensile stresses caused by rust expansion and the positive confinement effect of stirrups. Overall, the proposed models demonstrate superior accuracy, robustness, and generalization capability, providing an effective tool and theoretical basis for evaluating bond behavior and designing durable CRC structures with broad engineering applicability.

Schlagwörter

Stichwörter der Autor:innen

SHAP methodbond strengthcorroded reinforced concreteextreme gradient boostingsupport vector machine

Autor:innen und Institutionen

  • School of Architecture and Civil Engineering, Chengdu University, Chengdu 610106, China.
  • China Railway Eryuan Engineering Group Co., Ltd., Chengdu 610031, China.
  • School of Civil Engineering, Southwest Jiaotong University, Chengdu 610031, China.

ORCID-Profile: Ying-Zi Zhong

Förderung

  • National Key Research and Development Program of China

Zitieren

APA

Chen, Z. Q., Chen, Z., & Zhong, Y. Z. (2026). Prediction of Bond Strength in Corroded Reinforced Concrete Using SVM and XGB Methods. Materials, 19(10), 1928. https://doi.org/10.3390/ma19101928

Vancouver

Chen ZQ, Chen Z, Zhong YZ. Prediction of Bond Strength in Corroded Reinforced Concrete Using SVM and XGB Methods. Materials. 2026;19(10):1928. doi:10.3390/ma19101928

BibTeX
@article{Chen2026,
  author = {Chen, Zhi-Qiang and Chen, Zhuang and Zhong, Ying-Zi},
  title = {Prediction of Bond Strength in Corroded Reinforced Concrete Using SVM and XGB Methods},
  journal = {Materials},
  year = {2026},
  volume = {19},
  number = {10},
  pages = {1928},
  doi = {10.3390/ma19101928}
}
Kostenlos starten →