Manual Tagging-Guided Interactive AI Fault Detection

Fault detection is a staple of reservoir prediction because faults and associatedfractures are important to hydrocarbon migration and reservoir properties.Conventional methods can credibly detect large faults, but they are not sufficient forsubtle or small faults that are important to the deepening hydrocarbon exploration anddevelopment. Moreover, machine learning is introduced into fault detection, but itssuccess is closely tied to seismic data quality and fault tags. We present a method ofmanual tagging-guided interactive AI fault detection. This new method incorporatesseismic data preconditioning to improve data quality through fault enhancement basedon the theory of anisotropic diffusion and manual fault tagging based on RGB fusionof fault-sensitive attributes combined with geologic knowledge. Field applicationsshow improved intelligent fault detection for subsequent fault-controlled reservoirprediction.