Machine Learning–Based Hysteretic Lateral Force-Displacement Models of Reinforced Concrete Columns
Caigui Huang, Yong Li, Quan Gu, Jiadaren Liu
Abstract
Hysteretic lateral force-displacement (HLFD) models are important for efficient structural analysis under cyclic loading (e.g., earthquakes). This paper proposes a novel machine learning (ML)-based HLFD model, referred to as ML-HLFD, to characterize the relationship between lateral force and displacement of reinforced concrete (RC) columns with different properties (e.g., geometry, and material properties). To this end, a database including 498 experimental results is collected for model training, validation, and testing purposes. The ML-HLFD first uses a support vector machine (SVM) to classify the different failure modes (i.e., flexure failure, flexure-shear failure, and shear failure). After that, an artificial neural network (ANN) is trained for obtaining the implicit mapping between inputs (i.e., the properties of RC column) and outputs (i.e., the crucial parameters of selected HLFD models). The performance of the ML-HLFD models is studied by (1) cross-validation; and (2) comparisons with experiments, a classical fiber-element model, and an existing analytical model, which demonstrate the accuracy and efficiency of ML-HLFD models under a wide range of scenarios.