DETECT FAULTS AUTOMATICALLY
Interpreters spend a lot of time identifying faults in seismic data by picking 2D sections from a 3D seismic volume. This critical process can now be automated with AI Fault Detection in Paradise®, which uses deep learning and machine learning processes to generate fault volumes for fault interpretation. Check out the short video to the right to see results from AI Fault Detection, including for complex fault regimes or noisy seismic data. More examples of Fault Detection results are shown below. Select the technical papers below to learn more about fault interpretation and automatic Fault Detection.
Interpreters spend a lot of time identifying faults in seismic data by picking 2D lines. This critical process can now be automated with AI Fault Detection in Paradise, which uses deep learning and machine learning processes to generate fault volumes for fault interpretation. Check out the short video to the right to see results from AI Fault Detection, including for complex fault regimes or noisy seismic data. More examples of Fault Detection results are shown below. Select the technical papers below to learn more about fault interpretation and automatic Fault Detection.
Supporting resources on fault detection using Paradise
Learn how geoscientists are using the AI Fault Detection in Paradise to produce a refined fault volume, ready to import into an interpretation system.
Built on 3D CNN technology, the Deep Learning (DL) Fault Detection application is equipped with robust synthetic fault models supporting a wide range of seismic data. Since the fault models are pre-trained, geoscientists are not required to pick lines, avoiding bias and the computational training cost. AI Fault Detection uses a combination of supervised Deep Learning (DL) and unsupervised Machine Learning (ML) technologies to produce a refined fault volume, ready to import into an interpretation system and generate fault planes. Fault Detection offers these advantages:
- Pre-trained, robust 3D synthetic fault model – dramatically saves compute time
- Frees geoscientists from having to pick faults manually – eliminates bias
- Rapid conversion of fault probability results
- Eliminates uncertainty in manually selecting faults to build a valid machine-learning model
- Enables semi-supervised machine learning fault interpretation
A semi-supervised Machine Learning process that orchestrates multiple tools to delineate faults.
Semi-supervised Machine Learning – fault classification via SOM. The SOM has brought out the main stratigraphic packages, even in the dimmest part of the line – red box – this would be a very challenging section to interpret
Semi-supervised Machine Learning – Fault Geobodies extracted from SOM neurons.
Fault planes generated in Petrel using Fault Geobodies as input to Ant Tracking. These faults were generated using program defaults but can be selectively processed based on which are desired in the geomodel.
“The results from the project…were successful in isolating major and minor faults within the 3D survey to a level of detail not previously achieved.”
– Laudon, et al., An enhanced fault detection workflow combining machine learning and seismic attributes yields an improved fault model for Caspian Sea asset, First Break, 2021