Application of machine learning models in the capacity prediction of CCFST columns

Authors

Khaled Megahed, Department of structural Engineering Mansoura University Mansoura, EgyptFollow
Nabil Said Mahmoud, Department of structural Engineering Mansoura University Mansoura, Egypt
Saad Elden Mostafa Abd-Rabou, Department of structural Engineering Mansoura University Mansoura, Egypt

Subject Area

Civil and Environmental Engineering

Article Type

Review

Abstract

Circular concrete-filled steel tubular (CCFST) columns are widely utilized in structural engineering due to their impressive load-bearing capabilities and ductility. Existing design standards often yield disparate outcomes when applied to structural columns with identical properties, introducing uncertainty for engineering designers. This study introduces an innovative technique to address these challenges using two machine learning (ML) models: Gaussian process regression (GPR) and extreme gradient boosting (XGBoost). These models consider various input variables, including the geometric and material properties of CCFST columns, to estimate the compressive strength. The models undergo training and evaluation using two datasets comprising 1004 axially loaded CCFST columns and 515 eccentrically loaded CCFST columns. Furthermore, a unitless output variable, termed the strength index, is introduced to enhance model performance that can capture the physical properties of CCFST. To further assess the performance of the introduced models, two additional ML models are trained, including support vector regression optimized by the Jaya optimization method (JSVR) and an artificial neural network (ANN). Evaluation metrics indicate that the GPR model outperforms the remaining models. In addition, the Shapley additive interpretation (SHAP) technique is employed for feature analysis. This analysis shows that column length and load end-eccentricity have the most negative impact on compressive resistance. These findings can guarantee that the ML models can accurately predict the compression capacity of CCFST columns, providing structural engineers with reliable and valuable tools for their work.

Keywords

Circular concrete-filled steel tubular (CCFST) columns, Gaussian process (GPR), machine learning (ML) models, Extreme gradient boosting model (XGBoost), Shapley additive interpretation, Support vector regression (SVR), Artificial neural network (ANN).

Creative Commons License

This work is licensed under a Creative Commons Attribution 4.0 License.

Recommended Citation

Megahed, Khaled; Mahmoud, Nabil Said; and Abd-Rabou, Saad Elden Mostafa (2024) "Application of machine learning models in the capacity prediction of CCFST columns," Mansoura Engineering Journal: Vol. 49 : Iss. 3 , Article 5.
Available at: https://doi.org/10.58491/2735-4202.3178