Modeling Particle Sedimentation in Drilling Fluids using Gaussian Process Regression: A Computational Approach for...

Area of Science:

• Petroleum Engineering
• Chemical Engineering
• Computational Science

Background:

• Sedimentation is crucial in the oil industry, requiring accurate particle concentration and velocity profiles for simulation.
• Complex phenomenological and constitutive models limit practical simulation times.
• Understanding particle behavior in drilling fluids like Br-Mul is essential for operational efficiency.

Purpose of the Study:

• To apply Gaussian Process Regression (GPR) for regularizing experimental data in particle sedimentation models.
• To predict volumetric concentration profiles of solids in Br-Mul drilling fluid over time and height.
• To assess the computational efficiency of GPR compared to traditional simulation methods.

Main Methods:

• Utilized Gamma-ray attenuation technique to monitor solid concentration over 500 days.
• Employed Gaussian Process Regression to model and predict concentration profiles.
• Compared computational cost with algebraic-differential model simulations.

Main Results:

• GPR provided good estimates for concentration profiles across different spatial and temporal points.
• Experimental data fluctuations can lead to physically implausible profile predictions.
• The GPR approach demonstrated significantly lower computational requirements.

Conclusions:

• Gaussian Process Regression offers a computationally efficient alternative for modeling particle sedimentation in oil industry applications.
• While effective, careful handling of experimental data is needed to ensure physically realistic model outputs.
• This method holds promise for improving the practicality of sedimentation simulations in petroleum engineering.