Machine learning of accelerogram data for analyses, modeling and prediction
Files
Date
2022
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Open Access Color
OpenAIRE Downloads
OpenAIRE Views
Abstract
Bu tezde, kuvvetli yer hareketi istasyonundan kaydedilmiş ivmeölçer kayıtlarının evrişimsel ağlarda işlenmesi ile merkez üssü konumu tahmini sağlanmıştır. Deprem ivmeölçer kayıtlarının spektrogram tabanlı sahte renk gösterimi önerilmiş ve bu gösterimin evrişimsel ağlarda uygulanması tartışılmıştır. Kamuya açık kırk iki binden fazla deprem kaydı kullanılarak, 5 saniyelik yüzbinlerce sahte renk spektrogramı ile bir merkez üssü kümelemesi yapılmış, ve benzer kümelerdeki depremlerin benzer gösterimler yarattığı gözlenmiştir. Elde edilen bu merkez üssü kümeleme ile farklı yıllara ait farklı kayıtlar kullanılarak evrişimsel ağ eğitilmiştir. Eğitilen bu ağ ile herhangi bir deprem olayına ait, merkez üssü ve derinlik bilgilerini tahmin etmek amaçlanmıştır. Eğitimler sonucunda, tek istasyondan kaydedilen ivmeölçer verileri ile yaratılan spektrogramların evrişimsel ağlarda kullanılabildiği ve ivmeölçer verilerinin merkez üssü tespit etmede potansiyeli olduğu gözlemlenmiştir.
In this thesis, the earthquake epicenter coordinate prediction is provided by processing the accelerometer records recorded from the strong motion station using convolutional networks. Spectrogram-based false color representation of earthquake accelerometer records is proposed and its application in convolutional networks is discussed. Using more than forty-two thousand publicly available earthquake records, an epicenter cluster has been made with hundreds of thousands of 5-second false color spectrograms, and earthquakes in similar clusters were observed to produce similar impressions. With this epicenter clustering, the convolutional network is trained by using different records from different years. Using this trained network, it is aimed to predict the epicenter and depth information of any earthquake event. As a result of the trainings, it has been observed that the spectrograms created with single station accelerogram data can be used in convolutional networks and accelerograms have potential to detect epicenters.
In this thesis, the earthquake epicenter coordinate prediction is provided by processing the accelerometer records recorded from the strong motion station using convolutional networks. Spectrogram-based false color representation of earthquake accelerometer records is proposed and its application in convolutional networks is discussed. Using more than forty-two thousand publicly available earthquake records, an epicenter cluster has been made with hundreds of thousands of 5-second false color spectrograms, and earthquakes in similar clusters were observed to produce similar impressions. With this epicenter clustering, the convolutional network is trained by using different records from different years. Using this trained network, it is aimed to predict the epicenter and depth information of any earthquake event. As a result of the trainings, it has been observed that the spectrograms created with single station accelerogram data can be used in convolutional networks and accelerograms have potential to detect epicenters.
Description
Keywords
Earthquake Accelerograms, Epicenter Clustering, Convolutional Neural Networks, Spectrogram, Deprem İvmeölçer Kayıtlar, Merkez Üssü Kümeleme, Evrişimsel Ağlar, Spektrogram
Turkish CoHE Thesis Center URL
Fields of Science
Citation
Çıkış, Melis (2022). Machine learning of accelerogram data for analyses, modeling and prediction / Analizler, modelleme ve tahmin için akselerogram verilerinin makine öğrenimi. Yayımlanmış yüksek lisans tezi. Ankara: Çankaya Üniversitesi, Fen Bilimleri Enstitüsü.
WoS Q
Scopus Q
Source
Volume
Issue
Start Page
1
End Page
87
Sustainable Development Goals
3
GOOD HEALTH AND WELL-BEING

4
QUALITY EDUCATION

5
GENDER EQUALITY

8
DECENT WORK AND ECONOMIC GROWTH

9
INDUSTRY, INNOVATION AND INFRASTRUCTURE

10
REDUCED INEQUALITIES

11
SUSTAINABLE CITIES AND COMMUNITIES

14
LIFE BELOW WATER

17
PARTNERSHIPS FOR THE GOALS
