Machine Learning (ML) and Deep Learning (DL) Based Models for Prediction of Ultimate Strength of Circular Concrete-Filled Steel Tube Eccentrically Loaded Columns

Concrete-filled steel tubes (CFSTs) offer many advantages that make them popular in structural applications. Extensive research exists to determine their ultimate strength under different loading conditions. Determination of CFST column capacity is the most crucial engineering parameter for designing economic structural elements. However, the interaction between compression and bending induces complex failure mechanisms, making accurate predictions of ultimate strength more challenging. In the current paper, machine learning methods (ML) are adopted to predict the ultimate strength of circular CFSTs under eccentric loading conditions. A database of 1804 experimental tests from literature are employed to develop models. Nine deep learning and machine learning models are considered to evaluate the ultimate compressive strength. The study employed a range of machine learning models, including ensemble, regression, and neural network methods. Performance of considered models is assessed through analysis. Moreover, design formulas in three different codes are explored while showing the limitations related to their application. The study revealed that ML models are able to achieve different levels of predictive accuracy contributing to the sustainability of structural design by enabling more efficient material use and performance optimization. CatBoost and XGBoost were found to be the best performing models.