Kayma Dalgasi Hizinin Sugeno Bulanik Mantik ve Yapay Sinir Aglari Modelleri ile Tahmin Edilmesi

Yıl 2007, Cilt: 3 Sayı: 2, 1 - 9, 01.08.2007

Eray Yıldırım Cengiz Sertkaya Can Karavul

Öz

Sismik deneylerin yapilamadigi ya da sinirli miktarda yapildigi yerler için zemin dinamik özelliklerini
belirlemek zordur. Simdiye kadar ki çalisma larda, baska araziler üzerinde yapilmis deney sonuçlarindan
faydalanarak istenilen arazideki zemin özellikleri tahmin edilmeye çalisilmistir. Bu amaçla yapilan arastirmalar
sonucunda çesitli korelasyonlar, ampirik bagintilar gelistirilmistir.
Bu çalismada, Kuvaterner Alüvyonal zemin özellikleri tasiyan Adapazari’nda 33 ölçüm yerinde yapilan 44 SCPT
(Sismik Koni Penetrasyon Testi) ölçümlerinden elde edilen Kayma Dalgasi Hizlari (Vs) ile 64 Sondaj noktasindan
elde edilen SPT-N degerleri arasindaki iliskiler incelenerek modeller olusturulmustur. Model olarak Yapay Sinir
Aglari ve Sugeno – BM kullanilmistir. Olusturulan modellerin Ortalama Mutlak Hata Yüzdesine, Ortalama Hata
Karelerinin ve Regresyon Katsayilarina göre karsilastirmalari sonucunda Sugeno – BM modelin in en dogru
sonuçlari verdigi görülmüstür. Olusturulan bu model Adapazari’nda veya zemin özelligi benzer olan yerlerde çesitli
sebeplerle sismik deneylerin yapilamadigi durumlarda zemin dinamik özellikleri hakkinda yaklasik fikir edinmek
veya sinirli sayida sismik deneylerin uygulanabildigi durumlarda ise, ölçülen hiz degerlerini kontrol etmek ve sismik
deney programini desteklemek amaciyla kullanilabilir.

Anahtar Kelimeler

Yapay Sinir Aglari, Sugeno – BM, Kayma Dalgasi Hizi, SPT-N

Estimation of Shear Wave Velocity Using Sugeno Fuzzy Logic and Artificial Neural Networks Models

Öz

It is quite hard to determine the dynamic properties of the soils, where there is no or limited data on
seismic experiments. Most of the recent research uses the previous experimental results obtained from the similar
soil profiles to estimate the dynamic soil properties for the existent site to be analyzed. Hence, it is possible to come
across numerous empirical relations and correlations that were developed for these dynamic properties and
estimations in the literature.
This study focuses on the analysis of 44 Seismic Cone Penetration tests and 64 SPT tests performed on a Kuvaterner
Alluvial soil profile in 33 different measurement locations in city of Adapazari with different models. The relations
between SPT-N values and shear wave velocity values measured from the seismic cone penetration tests are
evaluated and a model study based on this relation is assessed. Artificial neural network and Sugeno-Fuzzy Logic
models are adopted for the estimation of dynamic properties. Evaluations of the models based on probabilistic
approaches using correlation coefficient, average absolute error and mean squared error have revealed that the
Sugeno-Fuzzy Logic model was superior to artificial neural network model. This soil model performs well to
estimate the dynamic properties of soil profiles in Adapazari or similar soil profiles where the seismic tests were not
performed at all. The model also proposes a useful tool to cross-check the shear wave veloc

Anahtar Kelimeler

Artificial neural netorks, Sugeno Fuzzy Logic, shear wave velocity, SPT-N