Overview:
• Not many error sources are Gaussian (Normal) distributed
• Mostly have some form of heavy-tail exponential behaviour
• When sources are combined the effect is reduced (central limit theorem) – but still significant
• Confidence levels are lower than predicted with Gaussian, especially at higher sigma levels
• Can we use the same propagation mathematics but use input (1 sigma) based on 99.7% confidence and divide by 3?
Basis for Study
• Magnetic Field study showed that declination errors are highly skewed and cant be treated as Gaussian (normal)
• Are other sources of error similar & how does this affect confidence levels?
• To look at survey comparison data to examine the spread of differences and observe the distribution type
• 2 Survey Comparisons of the same section of hole, commonly tied
• Unit/1000 unit separation, compared Lateral and Highside Axis
• Attempt to isolate some basic sources of error
• Like Gyro/Inrun outrun comparison for drift/inclination errors
• MWD vs EMS for Sag error and Magnetic interference
• Magnetic vs Gyros from BP Alaska, Inclination Model
• Interested in 1 Dimensional confidence because most requirements
are 1D
....
View the entire Presentation:
Tall Heavy Distributions & Confidence Levels for Directional Survey Measurements
Jerry Codling - Halliburton / Landmark