Faults and Fractures Detection Using a Combination of Seismic Attributes
by the MLP Artificial Neural Network in an Iranian Oilfield
Abstract
Faults and fractures play a significant role in drilling operations,
trapping hydrocarbon, and reservoir development in oilfields; exploring
faults quickly and accurately can help to reach the target more
manageable. In this approach, to improve faults and fractures detection,
applicable seismic attributes have been combined using a Multilayer
Perceptron (MLP) neural network and applied to a 3-D seismic cube of the
Changuleh oil field. First of all, high probabilistic faulted areas, as
an interesting area, have been identified using a hand-picking method on
a single seismic section. It is used as a pattern and one input set for
the MLP neural network. Then, some single seismic attributes (e.g.,
Similarity, Coherency, Curvature, Instantaneous, etc.) were applied to
the data. Next, the Multilayer Perceptron (MLP) neural network has been
used to assess and determine the most contributed attributes. The less
contributed ones are eliminated and the best seismic attributes, as
another input set, combined using the MLP. Finally, the outputs of the
MLP network will be two cubes named ‘faulted cube’ and ‘non-faulted
cube’. Differences between faulted zones and non-faulted zones on each
cube were conspicuous, and there was no need to be interpreted manually.
By comparing initial seismic sections and the MLP network’s outputs, it
is easy to see where the faulted and fractured zones are.