Shear Strength Prediction of Unusual Interior Reinforced Concrete Beam-Column Joint Using Multi-Layer Neural Network: a Data Collection by Digital 3D Finite Element Simulation
Christ John L. Marcos, Dario Landa-Silva
Abstract
One of the controversial topics in the literature on structural engineering is retrofitting existing substandard interior reinforced concrete beam-column joints. However, these retrofitting methods gave an unusual shape to the joints, causing the unpredictability of their strength. A machine learning application was developed to predict the shear strength of unusual joint, farther finite element analysis was utilized to generate 3D samples as a training dataset. The paper presented detailed methodologies and discussions of the two disciplines. Powerful digital technologies and computer systems shown dominance by presenting the performance and regression analysis through different trained neural network models. Sensitivity analysis was conducted utilizing connection weights algorithm to determine the relative importance factor.