ESTIMATION OF SOIL LIQUEFACTION POTENTIAL IN THE DALAMAN RESIDENTIAL AREA USING A SUPERVISED MACHINE LEARNING MODEL

Orkun Türe; Ergun Karacan

Makale detayları

ESTIMATION OF SOIL LIQUEFACTION POTENTIAL IN THE DALAMAN RESIDENTIAL AREA USING A SUPERVISED MACHINE LEARNING MODEL

Yazarlar:

Orkun Türe | Ergun Karacan

Yayıncı:

Mugla Sitki Kocman University

Ana Başlıklar:

İnşaat Geoteknik Mühendisliği | İnşaat Mühendisliğinde Zemin Mekaniği | Mühendislik Jeolojisi | Uygulamalı Jeoloji

Dergi:

Mugla Journal of Science and Technology Vol.11 Issue 1, pp.28-36

Belge Tipi:

Makale

Yayın Tarihi:

6/2025

Belge Dili:

İngilizce

Document ID (Org No.):

2001352538

pISSN:

2149-3596

Anahtar kelimeler:

Özet:

 Liquefaction is a critical phenomenon in geotechnical engineering, especially in mixed depositional environments where different soil types coexist. In such environments, assessment of liquefaction potential may be challenging. However, machine learning techniques overcome these challenges. In this study, to estimate the liquefaction potentials of the soils in the Dalaman residential area which is situated in a mixed depositional environment, supervised Multilayer Perceptron MLP model has been generated by using seismic parameters from the Chi-Chi and Kocaeli earthquakes and the parameters of the soils affected by these earthquakes. Sensitivity, specificity, accuracy, precision F1 score and AUC have been calculated for training and testing phases in model generation stage and for Dalaman Region. These values have been found to be 77.9, 91.5, 85.7, 87.0, 0.822 and 0.930 in training phase; 80, 83.1, 81.8, 75, 0.774 and 0.930 in testing phase. For Dalaman residential area, these values have been found as 81.3, 86.2, 83.5, 87.25 and 0.841. When the values from training and testing phase are compared to the results of Dalaman Region, it can be said that the model accurately estimated the liquefaction potentials of the soils in the Dalaman residential area.

 Liquefaction is a critical phenomenon in geotechnical engineering, especially in mixed depositional environments where different soil types coexist. In such environments, assessment of liquefaction potential may be challenging. However, machine learning techniques overcome these challenges. In this study, to estimate the liquefaction potentials of the soils in the Dalaman residential area which is situated in a mixed depositional environment, supervised Multilayer Perceptron MLP model has been generated by using seismic parameters from the Chi-Chi and Kocaeli earthquakes and the parameters of the soils affected by these earthquakes. Sensitivity, specificity, accuracy, precision F1 score and AUC have been calculated for training and testing phases in model generation stage and for Dalaman Region. These values have been found to be 77.9, 91.5, 85.7, 87.0, 0.822 and 0.930 in training phase; 80, 83.1, 81.8, 75, 0.774 and 0.930 in testing phase. For Dalaman residential area, these values have been found as 81.3, 86.2, 83.5, 87.25 and 0.841. When the values from training and testing phase are compared to the results of Dalaman Region, it can be said that the model accurately estimated the liquefaction potentials of the soils in the Dalaman residential area.

 Liquefaction is a critical phenomenon in geotechnical engineering, especially in mixed depositional environments where different soil types coexist. In such environments, assessment of liquefaction potential may be challenging. However, machine learning techniques overcome these challenges. In this study, to estimate the liquefaction potentials of the soils in the Dalaman residential area which is situated in a mixed depositional environment, supervised Multilayer Perceptron MLP model has been generated by using seismic parameters from the Chi-Chi and Kocaeli earthquakes and the parameters of the soils affected by these earthquakes. Sensitivity, specificity, accuracy, precision F1 score and AUC have been calculated for training and testing phases in model generation stage and for Dalaman Region. These values have been found to be 77.9, 91.5, 85.7, 87.0, 0.822 and 0.930 in training phase; 80, 83.1, 81.8, 75, 0.774 and 0.930 in testing phase. For Dalaman residential area, these values have been found as 81.3, 86.2, 83.5, 87.25 and 0.841. When the values from training and testing phase are compared to the results of Dalaman Region, it can be said that the model accurately estimated the liquefaction potentials of the soils in the Dalaman residential area.

 Sıvılaşma, özellikle karmaşık zemin tiplerinin bir arada bulunduğu karmaşık çökelim ortamlarında, geoteknik mühendisliği açısından kritik bir olgudur. Bu tür ortamlarda sıvılaşma potansiyelinin değerlendirilmesi zorlayıcı olabilir ancak makine öğrenimi teknikleri kullanarak, bu zorluklar kolayca aşılabilir. Bu çalışmada, karışık çökelim ortamında bulunan Dalaman yerleşim alanındaki zeminlerin sıvılaşma potansiyellerini tahmin etmek amacıyla, Chi-Chi ve Kocaeli depremlerine ait sismik parametreler ile bu depremlerden etkilenen zeminlerin parametreleri kullanılarak, denetimli bir makine öğrenimi algoritması olan Çok Katmanlı Algılayıcı MLP modeli geliştirilmiştir. Modelin oluşturulma aşamasındaki duyarlılık, özgüllük, doğruluk, kesinlik, F1-değeri ve AUC değeri eğitim aşamasında 77,9, 91,5, 85,7, 87,0, 0.822 ve 0.930 olarak, test aşamasında ise 80, 83,1, 81,8, 75, 0.774 ve 0.930 olarak bulunmuştur. Oluşturulan modelin Dalaman yerleşim alanındaki performans ölçütleri ise 81,3, 86,2, 83,5, 87,25 ve 0.841 olarak bulunmuştur. Bu değerler kıyaslandığında oluşturulan modelin Dalaman bölgesindeki zeminlerin sıvılaşma potansiyelini başarılı bir şekilde tahmin ettiği görülmüştür.

Makale detayları

ESTIMATION OF SOIL LIQUEFACTION POTENTIAL IN THE DALAMAN RESIDENTIAL AREA USING A SUPERVISED MACHINE LEARNING MODEL

Yazarlar:

Orkun Türe | Ergun Karacan

Yayıncı:

Mugla Sitki Kocman University

Ana Başlıklar:

İnşaat Geoteknik Mühendisliği | İnşaat Mühendisliğinde Zemin Mekaniği | Mühendislik Jeolojisi | Uygulamalı Jeoloji

Dergi:

Mugla Journal of Science and Technology Vol.11 Issue 1, pp.28-36

Belge Tipi:

Makale

Yayın Tarihi:

6/2025

Belge Dili:

İngilizce

Document ID (Org No.):

2001352538

pISSN:

2149-3596

Liquefaction

Multilayer Perceptron Machine Learning Technique

Mixed Depositional Environment

DalamanSıvılaşma

Çok Katmanlı Algılayıcı Makine Öğrenimi Tekniği

Karışık Çökelim Ortamı

Dalaman

Sayı

Makale detayları

Yazarlar: Orkun Türe | Ergun Karacan

Yayıncı: Mugla Sitki Kocman University

Ana Başlıklar: İnşaat Geoteknik Mühendisliği | İnşaat Mühendisliğinde Zemin Mekaniği | Mühendislik Jeolojisi | Uygulamalı Jeoloji

Dergi: Mugla Journal of Science and Technology Vol.11 Issue 1, pp.28-36

Belge Tipi: Makale

Yayın Tarihi: 6/2025

Belge Dili: İngilizce

Document ID (Org No.): 2001352538

pISSN: 2149-3596

Sayı

Yazarlar:

Orkun Türe | Ergun Karacan

Yayıncı:

Mugla Sitki Kocman University

Ana Başlıklar:

İnşaat Geoteknik Mühendisliği | İnşaat Mühendisliğinde Zemin Mekaniği | Mühendislik Jeolojisi | Uygulamalı Jeoloji

Dergi:

Mugla Journal of Science and Technology Vol.11 Issue 1, pp.28-36

Belge Tipi:

Makale

Yayın Tarihi:

6/2025

Belge Dili:

İngilizce

Document ID (Org No.):

2001352538

pISSN:

2149-3596