Nonlinear finite element and machine learning modeling of tubed reinforced concrete columns under eccentric axial compression loading

Abstract

There is still insufficient data on the behavior of tubed-reinforced concrete columns (TRCCs) under the eccentric compression. Thus, this research work comprehensively examines the eccentric compression behavior of TRCCs using nonlinear finite element modeling and machine learning (ML). To do this, numerical simulation and parametric analysis based on existing investigations were conducted. In addition to the existing 22 specimens with limited test variables, additional 188 specimens were developed to cover a wide range of parameters, including the load eccentricity, transverse reinforcement spacing, columns’ slenderness ratio, yield strength of steel, and outer steel tube diameter. Additionally, six ML models were created to estimate the ultimate load results. The results indicated that increasing the outer steel tube yield strength and diameter, and reducing the load eccentricity, slenderness ratio, and spacing of the transverse reinforcement enhanced the load-carrying capacity of the columns. The Gaussian process regression model demonstrated superior performance metrics in comparison to other ML models, with the highest R2 values (0.998613 in training and 0.99823 in testing stages) and lowest root mean square error values (0.007213 in training and 0.008471 in testing stages). To save money, time, and resources compared to laboratory testing, an online-based prediction program is finally presented to predict the columns’ ultimate load.

Keywords

• Finite element method
• Tubed-reinforced concrete column
• Reinforced concrete
• Machine learning
• Ductility
• Compression performance