Comparative study of influential factors for punching shear resistance/failure of RC slab-column joints using machine-learning models. (November 2022)

• Record Type: Journal Article
• Authors: Liang, Shixue; Shen, Yuanxie; Ren, Xiaodan
• Is Part Of: Structures. Volume 45(2022)
• Page Start: 1333
• Page End: 1349
• Publication Date: 2022-11
• Subjects: RC slab-column joint -- Punching shear resistance -- Failure mode -- Machine learning -- SHAP -- PDP; Structural engineering -- Periodicals; 624.1
• DOI: 10.1016/j.istruc.2022.09.110
• Languages: English
• ISSNs: 2352-0124

Abstract

Abstract: The study of punching shear resistance and failure mode for reinforced concrete (RC) slab-column structures has been constantly highlighted, due to a number of catastrophic punching shear failures in engineering applications. However, mechanics-based punching shear resistance models and failure mode identification models have problems of prediction accuracy and influential factors quantification. In this paper, the prediction models based on machine learning (ML) are proposed to predict the punching shear resistance and failure mode. Compared with other seven ML models, extreme gradient boosting (XGBoost) is selected as the best model, which performance is examined by several performance measures. The prediction process of XGBoost is explained by shapley additive explanation (SHAP) and partial dependence plot (PDP). A comparative SHAP and PDP analysis reveals the relationships between punching shear resistance/failure and influential factors, which cannot be quantified through traditional experimental and theoretical analysis. Furthermore, suggestions for choosing the influential factors are provided to avert the brittle punching shear failure and improve the punching shear resistance of RC slab-column joints.