ANN-Based Model for Predicting the Nonlinear Response of Flush Endplate Connections

Abstract

Predicting the moment-rotation response parameters of semirigid steel connections can be challenging given the many components contributing to the connection’s elastic and plastic deformations. This is the case for the popular flush endplate beam-to-column connections (FEPCs). The literature has highlighted the limitations of current analytical, mechanical, and traditional empirical models in providing accurate predictions of the FEPCs’ moment-rotation response. Considering this limitation, machine-learning methods have gained wide attention recently in structural engineering applications to address problems associated with complex structural deformation and damage phenomena. To that end, the superior nonlinearity of artificial neural networks (ANN) is employed herein to predict the response characteristics of FEPCs. A large data set of about 200 specimens, collected from past experimental programs, is utilized to train the ANN for predicting the bilinear response of FEPCs including strain hardening. The paper describes the deduction of the response parameters from test data using data fitting, the determination of significant geometric, material, and layout features, the ANN architecture and algorithms, and the accuracy metrics of the new model. The Shapley algorithm is used to explain the inner workings of the model. A computer tool as well as a descriptive guide to the mathematical construction of the ANN are provided to aid with model implementation in practice.