Artificial neural network-based hysteresis model for circular steel tubes. (April 2021)

• Record Type: Journal Article
• Authors: Yang, Chuanchang; Fan, Jian
• Is Part Of: Structures. Volume 30(2021)
• Page Start: 418
• Page End: 439
• Publication Date: 2021-04
• Subjects: Circular steel tube -- Hysteresis model -- Neural network -- Stiffness degradation -- Strength degradation -- Restoring force curve; Structural engineering -- Periodicals; 624.1
• DOI: 10.1016/j.istruc.2021.01.021
• Languages: English
• ISSNs: 2352-0124

Abstract

Highlights: An innovative hysteretic model of circular steel tubes is provided. The model considers the effects of kinematic hardening and progressive yield. The model considers the damage degradations of strength and stiffness. The key parameters of the model are predicted by artificial neural network (ANN). The ANN model can predict the hysteretic behavior of components accurately. Abstract: Hysteresis models of structural members form a basis for the seismic analysis of structures. In this study, a new hysteresis model for circular steel tube (CST) members, which includes a skeleton curve part, a Bauschinger part, a strength and stiffness degradation part, and an unloading part, is proposed. The stiffness and strength degradation of the members are considered from the perspective of energy dissipation. The skeleton curve also includes the softening range due to local buckling. The key parameters of the hysteresis model, such as yield rotation and elastic stiffness, are strongly nonlinearly related to the geometric and material parameters of the members. Therefore, an artificial neural network method is employed to establish the nonlinear mapping relation between the parameters of the hysteresis model and the geometric and material parameters by training a large number of samples. To verify the validity of the model, the restoring force curves of the CST members predicted by the proposed hysteresis model are compared with those calculated using the finite element method. The results show that the artificial neural network-based model has relatively high accuracy and generalization ability and can effectively and accurately simulate the restoring force curve of the CSTs.