Deep convolutional generative adversarial networks for the generation of numerous artificial spectrum‐compatible earthquake accelerograms using a limited number of ground motion records

Abstract

Deep learning (DL) methodologies have been recently employed to solve various civil and earthquake engineering problems. Nevertheless, due to the limited number of reliable data in the field of earthquake engineering, it is not convenient to obtain accurate results using DL. To tackle this challenge, the generative adversarial network (GAN) approach may be considered a reliable possible candidate. GANs have been introduced as an efficient way to train generative models. GANs exhibited their capabilities as well as versatility in the field of image production. For nonlinear dynamic analyses of structures, artificial ground accelerograms that are compatible with a target response spectrum are usually generated. In this paper, an efficient algorithm is proposed by which numerous artificial spectrum-compatible earthquake accelerograms are generated using a few ground motion records. For this purpose, a specific wellestablished generative model, namely, the deep convolutional GAN (DCGAN), is adopted for the first time and used. It is shown that DCGAN can easily generate desirable artificial ground accelerograms by having a limited number of seismic records as input to train the network. To quantitatively demonstrate the quality of the artificial ground accelerograms generated by the DCGAN, several computer experiments are presented, among which the robustness and feasibility of the proposed method are examined by using only four earthquake accelerograms as the worst scenario. Moreover, the efficiency of the DCGAN is illustrated by comparing various seismic parameters and the spectral response of the generated accelerograms with those of the actual accelerograms. The outcomes illustrate the efficiency and robustness of the presented DCGAN. INTRODUCTIONFor the aseismic design of structural systems, response spectra may be employed as a convenient tool to evaluate the maximum responses of these systems. A useful © 2022 Computer-Aided Civil and Infrastructure Engineering.response spectrum is the design spectrum provided by seismic codes. Time-history dynamic analysis is frequently employed to design critical/important structures like high-rise buildings, for which several ground accelerations should be available in conducting nonlinear dynamic

Self‐training approach for crack detection using synthesized crack images based on conditional generative adversarial network

“…They proposed an imageto-image translation method using GAN and completed a method that could increase the accuracy of road surface damage detection even at night. Matinfar and Khaji (2023) used a deep convolutional GAN to augment acceleration data recorded when an earthquake occurred. Shim et al (2023) used adversarial learning to detect damage on the concrete wall inside a tunnel, and they proposed a technology that could automatically check damage by installing this detection algorithm in a robot system.…”

“…They proposed an image‐to‐image translation method using GAN and completed a method that could increase the accuracy of road surface damage detection even at night. Matinfar and Khaji (2023) used a deep convolutional GAN to augment acceleration data recorded when an earthquake occurred. Shim et al.…”

Artificial intelligence in civil infrastructure health monitoring—Historical perspectives, current trends, and future visions

“…It enables the generation of a large amount of data in a limited time. Large-scale synthetic data can also be produced by generative adversarial networks (GANs) (Matinfar et al, 2022). It also permits automatic annotation of the data, which saves a lot of time and effort which go into the manual data annotation process.…”

A response‐compatible ground motion generation method using physics‐guided neural networks

“…Corresponding to these three characterizations of GMs, several methods for generating artificial GMs have been proposed by researchers. Matinfar et al (2023) used deep convolutional GANs to generate large amounts of response spectra-compatible artificial GMs from a small number of natural GMs. Lin and Ghaboussi (2001) proposed new stochastic neural networks that can generate multiple artificial GMs from a single response spectrum.…”