Estimation of blast-induced peak response of concrete-filled double-skin tube columns by intelligence-based technique

Abstract

This paper aims to predict the peak response of concrete-filled steel tubes (CFSTs) under blast loads using regression-based machine learning (ML) algorithms considering the variability of input parameters. Twelve regression metrics that are commonly used are utilized to evaluate and compare the efficiency and effectiveness of the proposed ML models. The Monte Carlo approach is used to propagate the variability in the input space to the predicted output. The results showed that the optimal method varies with perspectives. The XGBoost algorithm has the highest final score, and the SVM algorithm exhibits the highest total score for standard deviation. Furthermore, the confidence interval increases with the peak response, and the single model presents a higher accuracy in predicting large displacements. However, ensemble models do the opposite. Sensitivity and uncertainty analyses are performed, indicating that the scaled distance, section size and ultimate moment capacity exhibit the highest contribution to the peak response of the CFST. Finally, a simple formula is developed based on the MLP model, which can cover three input parameters. The findings of this paper can be used for computer-aided dynamic response design of CFST columns subjected to blast loads.