Multiscale Simulation and Machine Learning Optimization of Cuttings Transport in Riserless Pipelines Under Bubble...

Area of Science:

• Petroleum Engineering
• Multiphase Flow Dynamics
• Computational Fluid Dynamics

Background:

• Deepwater riserless drilling faces challenges with cuttings transport, leading to operational issues like stuck pipes.
• Existing models often neglect bubble dynamics in gas-liquid-solid systems, limiting accuracy in high-pressure, high-viscosity conditions.

Purpose of the Study:

• To develop an integrated framework for understanding bubble dynamics and cuttings transport in riserless drilling.
• To improve the predictive accuracy of cuttings concentration and transport rate using advanced modeling and machine learning.

Main Methods:

• Integrated framework combining visualization experiments, CFD-DEM coupled with Population Balance Model (PBM) simulation, and machine learning.
• High-fidelity quantification of bubble dynamics and cuttings interaction using the CFD-DEM-PBM model.
• Training Backpropagation (BP) and Radial Basis Function (RBF) neural networks with experimental data for physics-informed prediction.

Main Results:

• Bubble coalescence suppresses sedimentation, while bubble breakup promotes particle resuspension, with their synergy governing transport efficiency.
• The coupled CFD-DEM-PBM model accurately captures bubble dynamics and cross-scale interactions with cuttings.
• RBF neural network achieved superior predictive accuracy (R² = 0.91182), outperforming BP networks and empirical models.

Conclusions:

• The integrated framework provides a reliable basis for real-time parameter optimization and decision-making in riserless drilling.
• This study establishes a new paradigm for analyzing complex multiphase transport phenomena.
• Understanding bubble coalescence and breakup is critical for optimizing cuttings transport efficiency.