Details

Autor(en) / Beteiligte

• Li, Zele
• Noori, Mohammad
• Wan, Chunfeng
• Yu, Bo
• Wang, Bochen
• Altabey, Wael A.

Titel

A Deep Learning-Based Approach for the Identification of a Multi-Parameter BWBN Model

Ist Teil von

• Applied sciences, 2022-10, Vol.12 (19), p.9440

Ort / Verlag

MDPI AG

Erscheinungsjahr

2022

Quelle

EZB Electronic Journals Library

Beschreibungen/Notizen

• A restoring-force model is a versatile mathematical model that can describe the relationship between the restoring force and the deformation obtained from a large number of experiments. Over the past few decades, a large body of work on the development of restoring-force models has been reported in the literature. Under high intensity cyclic loadings or seismic excitations, reinforced concrete (RC) structures undergo a wide range of hysteretic deteriorations such as strength, stiffness and pinching degradations. These characteristic behaviors can be described by the multi-parameter Bouc-Wen-Baber-Noori (BWBN) model, which offers a wide range of applicability. This model has been applied for the response prediction and modeling restoring-force behavior in structural and mechanical engineering systems, by adjusting the distribution range of this model’s parameters. However, a major difficulty in utilizing the multi-parameter BWBN model is the parameters’ identification. In this paper, a deep neural network model is used to estimate the hysteresis parameters of the BWBN model. This model is one of the most versatile and widely used general hysteresis models that can describe the hysteretic behavior of RC columns. The experimental data of the RC columns used in this paper are collected from the database of the Pacific Earthquake Engineering Research Center (PEER). Firstly, the hysteretic loop obtained from a physical experiment is described by the BWBN model, and the parameters of the BWBN model are identified via a genetic optimization algorithm. Then a neural network is established by a backpropagation (BP) algorithm for associating the identified BWBN model parameters with physical parameters of the RC column. Finally, the regression analysis of the identified parameters is carried out to obtain the regression characteristics of the RC columns. The trained neural network model can directly identify the parameters of BWBN model based on the physical parameters of RC columns, and is effective and computationally efficient for multi-parameter BWBN model identification. The proposed approach overcomes the difficult problem of identifying the parameters of BWBN model and provides a promising approach for a wider application of this multi-parameter hysteresis model.