A machine-learning model for predicting sand production onset under true triaxial stresses based on CFD-DEM simulation dataset

Predicting and managing sand production has become a critical research focus in geo-energy systems such as natural gas hydrates, petroleum extraction, geothermal energy, and underground hydrogen storage due to the safety risks associated with sand production during and after extraction processes. This work identifies the dependence of the critical drawdown pressure gradient of sand production on particle size distribution, true triaxial stresses, and slot width. Based on Particle Flow Code (PFC) software, a validated computational fluid dynamics and discrete element method (CFD-DEM) numerical model was applied for fluid-sand flow simulation behind the sand control equipment, e.g., slotted liner or wire-wrapped screen. A machine-learning model based on the CFD-DEM simulation dataset was developed using SPSS to predict the critical drawdown promptly. We found that some extreme values (D0, D10, and D100), the mid-value (D70) of particle size distribution, and the stress along the short axis of the slot are the key factors affecting the critical drawdown pressure gradient. Overall, the machine-learning model presented in this study offers a fast and reliable prediction of sand production. The model improves sand management and well productivity by allowing adjustments in wire spacing, slot width, and mechanical strength. It can also be easily installed on mobile devices and used at oil and gas development sites.