Machine learning methods for compression capacity prediction and sensitivity analysis of concrete-filled steel tubular columns: State-of-the-art review

Concrete-filled steel tube (CFST) column is commonly utilized in modern construction and bridge engineering because of exceptional mechanical properties and cost-effectiveness. While traditional CFST column design and evaluation methods rely heavily on empirical formulas and design specifications, they face limitations when dealing with intricate conditions and advanced materials. In recent years, the emergence of machine learning algorithms has provided a new approach for predicting CFST column performance. This review paper examines the application of machine learning algorithms in predicting the compression capacity of CFST columns, with a focus on gene expression programming (GEP), artificial neural network (ANN), gradient tree boosting (GTB), support vector machine (SVM) and random forests (RF) approaches. The study demonstrates that these machine learning algorithms can accurately forecast the compression capacity of CFST columns, showcasing superior performance in certain scenarios compared to traditional methods. Additionally, the paper conducts a sensitivity analysis of machine learning algorithms. Despite their promise, machine learning techniques encounter challenges related to data quality, model interpretability, overfitting, and computational resource requirements. Therefore, future research should concentrate on expanding the machine learning model database, exploring advanced algorithms, and integrating machine learning predictions with conventional engineering software to improve the design and analysis efficiency of CFST column in real-world engineering applications.