Numerical simulation and data-driven study on the axial compression bearing capacity of steel reactive powder concrete columns

Abstract

<p>To address the issue of inaccurate and inadequate stability design methods for steel-reinforced reactive powder concrete columns (SRPCC), this study utilized machine learning (ML) techniques to accurately predict the axial compressive bearing capacity of SRPCC. A finite element model was employed to conduct numerical simulations on the axial compression behavior of the SRPCC. The validity of the model was rigorously assessed by comparing the load-displacement curve and peak load obtained from the actual testing with those from the simulation results. A parameterized modeling approach using Python scripts was implemented in ABAQUS to generate a comprehensive database of finite element models for SRPCC under axial compression. The database consisted of 3200 models with different design parameters. Considering the potential of ML models in handling nonlinear and complex problems, an ML model was developed for intelligent prediction of the bearing capacity of SRPCC under axial compression. The prediction results of the ML model were compared with those of existing design specification calculation methods to evaluate the performance of the ML model. The findings reveal that the predictive accuracy of the ML model significantly surpasses that of existing design specifications. A graphical user interface tool was created using the proposed ML model to facilitate structural design utilization of the model by engineers and researchers.<br></p>