AI/ML Lithology Prediction

Description Generate lithology probability curves using machine learning.
Module(s) Lithofluids

This process accepts up to 9 input curves corresponding to different measurement types and returns probabilities for 9 different lithologies at each depth sample, as predicted by DUG Insight's Machine Learning model. It can also generate a Most Likely lithology curve that identifies the highest-probability lithology at each depth.

The value at each depth indicates the probability that the corresponding depth is composed primarily of this lithology. It does not indicate a relative volume like petrophysical interpretation curves do.

  1. In the Control Panel,go to View New Single Well View.
  2. In the Single Well window, open the Process tab.
  3. At the tab header, click the blue "+" icon.
  4. Click AI/ML Lithology Prediction.
  5. Type a name for the process and click OK.

Configure the AI/ML Lithology Prediction Process

  1. Input
    • Curves: Select the relevant input curves.
  2. Output
    • Enable custom name: If enabled, the specified name is used as a prefix for the output curve names. The custom name should be unique for each AI/ML Lithology Prediction process.
    • Extent: Determines the sampling grid for the output curves. All input data is resampled onto the output grid before being passed through the inference model.
  3. Most likely threshold: Sets the minimum probability a lithology must reach to be recorded as the most-likely lithology at a given depth. Use the slider or enter a value between 0 and 1. Depths where no lithology meets the threshold are left undefined. Default: 0.8.

Note: The output curves of the AI/ML Lithology Prediction process are placed in individual data classes named "Probability (lithology)" in the Curves tab. The most-likely lithology curve is placed in a separate data class named "Most Likely".

Display Probability Curves as Stacked Fills

DUG Insight provides an easy way to visualise the predicted lithology probabilities on a single track:

  1. In the Process tab, right-click the AI/ML Lithology Prediction process, and select Add outputs to new Stacked Fill Track.
  2. A stacked fill curve containing all probability curves is displayed in a new track.

For more configuration details, refer to Stacked Fill Curves.