Akademik Veri Yönetim Sistemi
International Journal of Scientific Advances , cilt.2, sa.1, ss.39-58, 2021 (Hakemli Dergi)
In the current study, the
performance of a synthetic wavelet CNN (Wave-CNN) was tested and then compared
with geophysical field data; two
separate synthetic studies were conducted for this purpose. In the first
synthetic application, the success of the Wave-CNN method for the separation of
regional-residual dipole structures was tested. In the second synthetic
application, using prismatic structures with magnetic properties, the success
of the Wave-CNN method to determine boundaries was compared with classical
methods. It was found that the Wave-CNN method could not be repeated as many
times as required. At first, it was possible to distinguish between regional
and residual anomalies and successfully determine boundaries. As the number of
repetitions increased, it was possible to clarify structures with stronger
magnetic properties and to filter out structures with weak magnetic properties.
The Wave-CNN
using a vertical component of the
magnetic field method
was then applied to magnetic
anomaly data from
three Avnik iron mines located in the Bingöl Province, East Anatolia, Turkey:
these areas are called, from south-to-north, Gonactepe, Heylandere and Miskel.
The Wave-CNN outputs were also compared to drilling results from the study
areas. Using this method, Wave-CNNs can be used to solve geophysical problems
by detecting surface layer boundaries. In this Avnik field application, the Wave-CNN results showed high
anomaly values indicating probable an iron ore deposits. In addition, represantive anomaly values were
selected from the Wave-CNN outputs of each subarea and forward modeling was
performed by applying Differential Evolution (DE) method. Thus, it has been
shown that a good relation is obtained with the geological cross-section and
geophysical model structures.
Keywords: wavelet cellular neural network; image processing; iron ore; forward
modelling