@article {
author = {Eshaghzadeh, Ata and Seyedi Sahebari, Sanaz and Kalantari, Roghayeh Sadat},
title = {2-D Anticlinal Structure Modeling Using Feed-Forward Neural Network (FNN) Inversion of Profile Gravity Data: A Case Study from Iran},
journal = {Journal of the Earth and Space Physics},
volume = {46},
number = {4},
pages = {79-91},
year = {2021},
publisher = {Institute of Geophysics, University of Tehran},
issn = {2538-371X},
eissn = {2538-3906},
doi = {10.22059/jesphys.2021.286888.1007148},
abstract = {The Anticlines are the main hydrocarbon traps on land or at sea. This structure is considered as the target of the many projects of gravity exploration all over the world. Artificial neural networks (ANNs) are used in order to solve prediction, estimation, and optimization problems. In this paper, the feed-forward neural network (FNN) is applied for modeling the anticlinal structure using gravity anomaly profile and the back propagation algorithm is used for artificial neural network training. Moreover, the scalene triangle model is employed to describe the geometry of anticlinal structure in analyzing gravity anomalies. In terms of neural network training, eight features among the synthetic gravity field variations curves along 22500 profiles are defined. These gravity profiles are computed based on different values of the scalene triangle parameters consisting of the top depth, bottom depth, limb angles and density contrast. The defined neural network contain three layers, eight neurons (the number of features) in the input layer, 30 neurons in the hidden layer and six neurons (the number of scalene triangle parameters) in the output layer. In order to evaluate the performance of the trained neural network, the specified features related to a synthetic model, with and without random noise, are applied as the input data to train neural network. The parameters estimation error by FNN is negligible. The proposed method is illustrated with a real gravity data set from Korand region, Iran. The inferred anticlinal structures are compared with the interpreted map of the seismic data.},
keywords = {Anticlinal structure,Feed-forward neural network (FNN),Gravity,Scalene triangle,Iran},
title_fa = {2-D Anticlinal Structure Modeling Using Feed-Forward Neural Network (FNN) Inversion of Profile Gravity Data: A Case Study from Iran},
abstract_fa = {The Anticlines are the main hydrocarbon traps on land or at sea. This structure is considered as the target of the many projects of gravity exploration all over the world. Artificial neural networks (ANNs) are used in order to solve prediction, estimation, and optimization problems. In this paper, the feed-forward neural network (FNN) is applied for modeling the anticlinal structure using gravity anomaly profile and the back propagation algorithm is used for artificial neural network training. Moreover, the scalene triangle model is employed to describe the geometry of anticlinal structure in analyzing gravity anomalies. In terms of neural network training, eight features among the synthetic gravity field variations curves along 22500 profiles are defined. These gravity profiles are computed based on different values of the scalene triangle parameters consisting of the top depth, bottom depth, limb angles and density contrast. The defined neural network contain three layers, eight neurons (the number of features) in the input layer, 30 neurons in the hidden layer and six neurons (the number of scalene triangle parameters) in the output layer. In order to evaluate the performance of the trained neural network, the specified features related to a synthetic model, with and without random noise, are applied as the input data to train neural network. The parameters estimation error by FNN is negligible. The proposed method is illustrated with a real gravity data set from Korand region, Iran. The inferred anticlinal structures are compared with the interpreted map of the seismic data.},
keywords_fa = {Anticlinal structure,Feed-forward neural network (FNN),Gravity,Scalene triangle,Iran},
url = {https://jesphys.ut.ac.ir/article_79564.html},
eprint = {https://jesphys.ut.ac.ir/article_79564_34c83ae75b026c7ec80f444eec9af24a.pdf}
}