Synthetic seismic
data sets

The training accumulates and analyzes information from thousands of synthetic datasets. This way the algorithm is fully prepared to look for faults, noise data, specific geobodies, etc.

It is used to correct Machine Learning results in the areas where the user can specify complex faults, geobodies, seismic facies. This way the user gets to combine learning on synthetic models and real data.

Together these three anchors make Geoplat AI a unique product for geoscience.

Machine leaning is a well-known approach that Geoplat has managed to adapt to the needs of seismic survey. Our method is based on three anchors:

With different conditions: including noise, different kind of faults, horizontal and dipping reflections, geobodies, etc.
Over the years over 1 mln datasets have been generated.

Transfer learning

Machine Learning approach in Geoplat AI

Pre-trained neural networks

Initial Data

Significant increase of Signal-to-noise ratio for further detailed interpretations

Ability to detect thin layers in case of low data resolution

Precise determination of fault zones and real amplitude shifts up to several meters

AI

Seismic Data Enhancement

Significant increase of Signal-to-noise ratio for further detailed interpretations

Ability to detect thin layers in case of low data resolution

Precise determination of fault zones and real amplitude shifts up to several meters

Fault Determination

Geobodies Detection

Significant increase of Signal-to-noise ratio for further detailed interpretations

Ability to detect thin layers in case of low data resolution

Precise determination of fault zones and real amplitude shifts up to several meters

Significant increase of Signal-to-noise ratio for further detailed interpretations

Ability to detect thin layers in case of low data resolution

Precise determination of fault zones and real amplitude shifts up to several meters

Seismic Facies Identification