En Büyük Yer İvmesi Azalım İlişkilerinin Yakın Dönemde Türkiye’de Meydana Gelen Depremlerle Performanslarının Değerlendirmesi

Abstract

The realistic prediction of the peak ground acceleration (amax or PGA) by considering possible earthquake scenarios for the engineering applications in regions having high seismic activity is very important. Hundreds of amax prediction equations were developed by using different databases in different regions for this purpose. However, it is another fact that high prediction performances could not be achieved despite these many equations in the literature. In this study, the performances of some widely known attenuation relationships for the prediction of the peak ground acceleration were tested and evaluated by using recent earthquake data in Turkey. In order to integrate the results of the two equations developed by İnan et al. (1996) and Akkar and Bommer (2010), which stand out in performance evaluations, it was first tried to sum the amax results of these two equations with different weights. In the second stage, an artificial neural network (ANN) modelling was carried out by using the moment magnitude (Mw) and the distance from the epicenter (Repi) of the earthquake in addition to the amax results of these two equations as input parameters. In this ANN-based approach, it was aimed to ensure that the two estimation equations produce a variable weighted result based on Mw and Repi for integrating amax results. As a result of the performance evaluations, a series of charts has been prepared to ensure the practice of using the ANN model, which is the model having the best performance. In the final part of the thesis, by using the developed charts amax distribution maps of an area in the Black Sea region was produced for two scenario earthquakes with the magnitude of Mw=7.5 and Mw=6.5 on the North Anatolian Fault.