TipTuesday_LightBulb.png
TUESDAY
CMOST AI Tips

April 23rd, 2019

For faster history matching and optimization, use CMOST AI's Proxy Analysis tools to better understand parameter-objective function interactions.

Tools such as Effect Estimates, Morris Analysis, and Sobol Analysis are found under ‘Proxy Analysis’ in the Objective Functions portion of the CMOST AI dashboard.
CMOST AI Tip Tuesday Graphic

The table below lists key differences between the methods, depending on your analysis requirements:
Effect Estimates
Morris Analysis
Sobol Analysis
Applicable for polynomial regression proxy ✓ ✓ ✓
Applicable for neural network proxy n/a ✓ ✓
Indicate effects in terms of absolute values ✓ n/a n/a
Indicate negative/positive effects ✓ n/a n/a
Indicate effects in terms of absolute mean and standard deviation n/a ✓ n/a
Indicate effects in terms of percentage of variance n/a n/a ✓
Aggregate all effects (linear, non-linear, interaction) into one indicator n/a n/a ✓
March 5th, 2019
Using CMOST AI, test different variable changes to know what will happen before the run is submitted.


  1. Build preliminary proxy models to begin predicting reservoir behavior
  2. Use the proxy dashboard to test a “What-if Scenario”
  3. Submit the experiment to the simulator to validate the results of the “What-if Scenario”