Loading...
Description
Earthquake records can be encoded as images to classify their resulting damage states to structures based on convolutional neural networks (CNNs). Different encoding techniques such as Recurrence Plot (RP) and Wavelet Transform (WT) can be used to transfer earthquake records to images. Presently, no consensus has been reached on the understanding of the most suitable encoding technique for CNN-based seismic damage classification. In this study, we develop a new encoding technique based on Time-series Segmentation (TS) and compare it to state-of-the-art RP and WT techniques. These techniques are mainly compared based on their classification accuracy and computation efficiency.
Presentation Date
11 Aug 2021, 10:30-10:40 am
Meeting Name
INSPIRE-UTC 2021 Annual Meeting
Department(s)
Civil, Architectural and Environmental Engineering
Document Type
Poster
Document Version
Final Version
File Type
text
Language(s)
English
Encoding Time-series Ground Motions as Images for Convolutional Neural Networks-Based Seismic Damage Evaluation
Earthquake records can be encoded as images to classify their resulting damage states to structures based on convolutional neural networks (CNNs). Different encoding techniques such as Recurrence Plot (RP) and Wavelet Transform (WT) can be used to transfer earthquake records to images. Presently, no consensus has been reached on the understanding of the most suitable encoding technique for CNN-based seismic damage classification. In this study, we develop a new encoding technique based on Time-series Segmentation (TS) and compare it to state-of-the-art RP and WT techniques. These techniques are mainly compared based on their classification accuracy and computation efficiency.
Comments
Financial support to complete this study was partially provided in part by the U.S. Department of Transportation, Office of Assistant Secretary for Research and Technology under the auspices of Mid-America Transportation Center at the University of Nebraska, Lincoln (grant no. 00059709).