Efficiency of Data-Driven Hybrid Algorithms for Steel-Column Base Connection Failure Mode Detection

Abstract

Base connections join the column to the foundation, thereby providing a superstructure fixation to the foundation, and play a major role in the steel structure’s ductile behavior. Seismic damage to these connections can dramatically increase the cost of restoration and the risk of destruction. The purpose of this research was to evaluate the effectiveness of three advanced hybrid models, which combine the particle swarm optimization (PSO) algorithm, Runge–Kutta optimizer (RUN), and sparrow search algorithm (SSA) with an artificial neural network (ANN), to recognize the failure modes of the steel-column base plate (SCBP) connection. Data from prior experiments were used as inputs to the models. A comparison was performed between the results of the proposed models (PSO-ANN, RUN-ANN, and SSA-ANN) and the previous studies that utilized different machine learning algorithms, such as support vector machine and naive Bayes, for the failure mode identification of the SCBP connections. Examination of all models showed that the hybrids RUN-ANN, PSO-ANN, SSA-ANN, and decision tree perform better than the others models and can predict the failure mode with an accuracy of 95%, 92%, 90%, and 91%, respectively. The SHapley Additive exPlanation methodology is also used in this study to demonstrate the importance and contribution of the components that influence SCBP connections failure mechanism identification.