Integration of support vector regression and grey wolf optimization for estimating the ultimate bearing capacity in concrete-filled steel tube columns

Abstract

Concrete-filled steel tube (CFST) columns are widely used in the construction industry. Prediction of the ultimate bearing capacity of CFST columns is complicated because it is influenced nonlinearly by many factors such as steel tube length, steel tube thickness, ratio length and column diameter, and concrete compressive strength. This study proposes an artificial intelligence (AI) model to predict the ultimate bearing capacity of CFST columns. The AI model was developed based on support vector regression (SVR) and grey wolf optimization (GWO). The GWO optimized the SVR configuration that produces highly accurate prediction results. A large experimental dataset with normal, high, and ultimate strength concretes was used to validate the model’s effectiveness through the learning and test phases. A k-fold cross-validation method was adopted to ensure the generalizability. The column diameter (D), thickness of steel tube (t), yield stress of steel, compressive strength of concrete, column length, D/t ratio were used as inputs for the model. Results show that the proposed SVR-GWO model was more effective than the compared models and empirical methods in the bearing capacity prediction of CFST columns. The SVR-GWO yielded the outstanding performance in which the accuracy improvements by the proposed model were ranged from 10.3 to 87.9% in the mean absolute percentage error and from 15.4 to 74.2% in the mean absolute error compared to baseline models and empirical methods. As contributions, the study suggested an AI-based tool for estimating the ultimate bearing capacity of CFST columns in structural design.