A Comparison Between ANNs and Ant Tracking Outputs for Fault and Fracture Detection

Abstract

Abstract Knowing the fractures and faults in a field or reservoir leads to more precise exploration and makes development strategies less uncertain. To better identify these events, in this study, the outputs of multilayer perceptrons (MLP), and unsupervised vector quantizers (UVQ), were compared with the outputs of ant-tracking. Although the specific seismic attributes and ant-tracking allow for the depiction of faults and fractures, the combination of the shortlisted attributes in the ANNs, especially MLP, reveals small-scaled and, in some areas, vertical faults and fractures more continuously and precisely than previous studies that used only conventional ant-tracking. The chaos has some deficiencies in identifying the faulted and fractured zones. Ant-tracking has a much lower performance than the MLP output. The output of ant-tracking attributes shows fault detection with less accuracy than the MLP output or even that of UVQ. To be clearly described, for the given seismic survey data, finding small-scale and in some areas vertical fractures using the integration of seismic attributes were more reliable than the output of the ant-tracking.