Signal and Noise Detection in Magnetotelluric Data by the Artificial Neural Network Method

In this study artificial neural network method was used to classify noisy components in the MT method data. For this purpose a multi-layered, feed-foorward and back propagation model was employed. Noisy time windows were determined to an accuracy of 89% depending on the training data set. In addition, when all types of noise in the data are defined (synthetic data), all noisy time windows can be sellected and eliminated by artificial neural network method. Test results from synthetic and field data indicate that artificial neural network classification is succesfull in identifying and eliminating the noisy data windows compared to both visual inspection and conventional assessment methods.Original Abstract: Bu calisma kapsaminda manyetotelluerik yoentem verisindeki guerueltue bilesenlerini siniflamak icin yapay sinir agi yoentemi kullanilmistir. Bu amacla cok katmanli, ileri beslemeli ve geri yayilimli bir model olusturulmustur. Secilen egitim setine bagli olarak guerueltuelue zaman pencereleri % 89 dogrulukla belirlenmistir. Ayrica verideki guerueltue tuerlerinin hepsi tanimlandiginda (yapay veri), tuem guerueltuelue pencereler yapay sinir agi yoentemi ile secilip elenebilmektedir. Yapay veri ve arazi verisi ile yapilan uygulamalar sonucunda, hem goersel denetlemeye hem de geleneksel degerlendirme yoentemlerine goere, guerueltuelue veri pencerelerini siniflayip elemede yapay sinir agi yoenteminin daha basarili oldugu goesterilmistir.