Skip to the content

S-Cube exists because current seismic inversion technology is not good enough. Our XWITM solution solves the 6 well-known and long-standing industry problems related to Full Waveform Inversion to deliver all-important automation and repeatability to the technique for wide-scale adoption.

Notably these are: (1) requiring high-quality starting models; (2) requiring artificially long offsets; (3) requiring prohibitively costly simulations to account for elasticity; (4) requiring error-prone manual interpretation; (5) misconvergence due to non optimal parameterisation; and (6) lack of compute to use the full bandwidth.

To solve these, XWITM delivers: (1) automatic convergence with AWI; (2) deep reflection-driven automatic convergence with RWI; (3) neural-network based acoustic-to-elastic mapping with GANs; (4) total variation constraints for preserving sharp boundaries with CWI; (5) a human-machine intelligent UI to tune hyperparameters mid-run; and (6) A combination of a vast pool of autonomous EC2 hardware orchestrated in the cloud with an intelligent UI.

1. Automatic Convergence

Problem: Current technology requires artificially high-quality starting models or artificially low useable frequencies.

Our Solution: AWI for increasing convexity around the basin of attraction to start from featureless model without tomography.

Aerial view through model without any evidence of low-velocity zone

Aerial view through model without any evidence of low-velocity zone

Aerial view through AWI intermediate model (1)

Aerial view through AWI intermediate model (1)

Aerial view through AWI intermediate model (2)

Aerial view through AWI intermediate model (2)

Aerial view through AWI intermediate model (3)

Aerial view through AWI intermediate model (3)

Aerial view through AWI intermediate model (4)

Aerial view through AWI intermediate model (4)

Aerial view through AWI final model

Aerial view through AWI final model

2. Automatic Convergence at Target Depths

Problem: Current technology requires artificially long offsets to automatically converge at target depths. 

Our Solution: RWI for explicit sensitivity to reflection moveout for macro model update below diving wave zone with multiple starting points automatically converging to the same point. 

Starting model 1

Starting model 1

XWI recovered model from starting model 1

XWI recovered model from starting model 1

Starting model 1

Starting model 1

XWI recovered model from starting model 1

XWI recovered model from starting model 1

3. Beyond Acoustic

Problem: Current technology requires prohibitively costly simulations to account for non acoustic effects. 

Our Solution: GANs for acoustic to elastic domain translation in shot and gradient domains. 

Predicted starting gather

Predicted starting gather

Predicted XWI gather

Predicted XWI gather

Predicted XWI with GANs gather

Predicted XWI with GANs gather

Field observed gather.

Field observed gather.

4. Highly Contrasting Geology

Problem: Current technology requires error-prone manual processes to account for highly contrasting geology. 

Our Solution: AWI with total variation constraints for preserving boundaries and suppressing bandwidth related oscillations. 

Starting model for the two scenarios below.

Starting model for the two scenarios below.

FWI model - conventional solution/problem

FWI model - conventional solution/problem

XWI model - our solution.

XWI model - our solution.

5. Human and Machine Intelligence Out of Sync

Problem: Current technology are unpredictable with no means to intervene in a run to switch loss functions or tune hyper parameters. 

Our Solution: GWI with intelligent UI to rapidly identify the best performing residuals on a shot-by-shot iteration-by-iteration basis.

Residual loss function filters - start

Residual loss function filters - start

Residual loss function time filters - XWI

Residual loss function time filters - XWI

Predicted data - starting shot gather

Predicted data - starting shot gather

Field data - shot gather

Field data - shot gather

Predicted data - XWl shot gather

Predicted data - XWl shot gather

Residual loss function time shift maps - start

Residual loss function time shift maps - start

Residual loss function time shift maps - XWI

Residual loss function time shift maps - XWI

6. Full Bandwidth

Problem: Solve all the problems of current technology not just at low frequencies but with increasing bandwidth, taking advantage of compute resources.

Our Solution: Unlock the best hyper parameters for example, solving the density unknowns are each bandwidth. 

Starting model

Starting model

Hyper parameter tuning for density at seabed

Hyper parameter tuning for density at seabed

High frequency (40Hz) XWI model

High frequency (40Hz) XWI model

 Predictor of unseen drilling logs

The power of our combined solution - predictor of unseen drilling logs like the industry has never seen. This is now being put to the test within Infrastructure-Led-Exploration (ILX) programmes in the North Sea. Using a vast existing portfolio of seismic and well data to automate and accelerate seismic imaging,  XWITM is combining deep learning with classical optimisation: the overall goal is to take the predictor for unseen drilling logs from a test scale level to a regional scale for ILX, CCS and the energy transition in a single unified optimisation loss function loop.

Standard FWI - well misconvergence

Standard FWI - well misconvergence

XWI - Predictive power with blind sonic measurements

XWI - Predictive power with blind sonic measurements

S-Cube Cloud

Run Full Waveform Inversion on the Cloud

Use XWITM on AWS to discover a step change in accuracy and resolution of your velocity model.