Fault Planes from OpendTect’s 3D Unet Fault Predictor The goal of this workflow is to automatically extract fault planes from a fault cube. In this project, the Fault Likelihood...
Fault Planes from OpendTect’s 3D Unet Fault Predictor
The goal of this workflow is to automatically extract fault planes from a fault cube. In this project, the Fault Likelihood was computed first, but the thinned derivative (TFL) showed some unwanted features.
The radial patterns in the gif are an artifact caused by missing traces. The horizontal stripes in TFL reflect the acquisition footprint. This footprint was successfully removed with the MGS-Destriping plugin https://lnkd.in/d86PCCc (compare time-slice before and after).
Next, OpendTect’s pre-trained 3D Unet Fault Predictor was applied to an MGS-Destriped and Edge-Preserved-Smoothed volume. As the 3D Unet Fault Predictor only predicts Fault Likelihood and not Fault Dip and Fault Strike, the output cannot be used “AS IS” for automated fault plane extraction in OpendTect’s Faults & Fractures plugin.
The fault planes in the gif are a selection of fault planes extracted from a modified fault cube with the required three components:
1. Fault Likelihood from the 3D Unet Fault Predictor, 2. Dip from Fault Likelihood, 3. Strike from Fault Likelihood.
Author: Frances Cooke PhD student Tromsø University
