Machine learning-enhanced fully coupled fluid–solid interaction models for proppant dynamics in hydraulic fractures

Dennis Delali Kwesi Wayo, Sonny Irawan, Lei Wang, Leonardo Goliatt

2025

Full text Open access: Gold Metadata (CSL-JSON)
DOI 10.1038/s41598-025-15837-5
Data confidence 70 medium
How this score is calculated

It is built from checkable properties of the record itself: whether the article has a DOI, how many independent sources confirmed its data, whether the required fields are filled in, and how recent the latest confirmation is. Disagreement between sources and a missing identifier lower it. A value of 80 or above means the record was published without further checking; 60 to 79 means it is published at reduced confidence, with a manual check advisable. A record verified by an editor by hand is never pushed below 90 by a recalculation. This scores our data about the article, not the article itself.

Abstract

This study presents a hybrid modeling framework for predicting proppant settling rate (PSR) in hydraulic fracturing by integrating symbolic physics-based derivations, parametric simulations, and ensemble machine learning. Symbolic expressions were formulated using Stokes' law, drag equations, and pressure-gradient dynamics. A symbolic dataset was synthetically generated by sampling realistic physical ranges: proppant density [Formula: see text], fluid viscosity [Formula: see text], and particle diameter [Formula: see text]. Complementary CFD-informed datasets were simulated to represent complex flow behavior. Both datasets were used to train stacked ensemble regressors comprising five base learners: Random Forest, Extra Trees, Gradient Boosting, XGBoost, and Support Vector Regression (SVR), combined with a RidgeCV meta-learner. Numerical analysis validated the physics consistency of the symbolic model. ODE-based simulations revealed terminal velocity of ∼0.39 m/s reached within 0.5 s, while parametric studies showed velocity reductions up to 40% for strain [Formula: see text]. Pressure-gradient analysis showed a 45% reduction in settling depth as [Formula: see text] increased from 0.1 to 1.0 bar/m. Model performance was evaluated across symbolic, CFD, and combined datasets. The symbolic model achieved R[Formula: see text] = 0.9934, RMSE = 0.0436; the CFD model yielded R[Formula: see text] = 0.9941, RMSE = 0.2033. The hybrid ensemble outperformed both with R[Formula: see text] = 0.9970, RMSE = 0.1801. This framework enables interpretable, accurate, and computationally efficient prediction of PSR, eliminating the need for full-scale CFD-DEM simulations. It is well-suited for decision support in multiscale fracture design and proppant transport analysis.