A Hybrid Approach Using Lattice Boltzmann and Machine Learning for Predicting the Transient Dynamics of Slug Flow in Non-Newtonian Fluids

Slug flow is a prevalent and crucial flow regime in many industrial applications, such as petroleum and chemical processing, where accurate modeling and prediction are essential for optimizing system performance and ensuring operational stability. This study presents a comprehensive numerical investigation of two-phase slug flow (Newtonian and non-Newtonian) in a horizontal channel using the Lattice Boltzmann Method (LBM). A color-gradient model is employed to capture slug formation, and a non-Newtonian power-law model is used to represent the shear-thickening behavior of one of the fluids. The simulations explore the influence of key parameters, including the flow index (n) and inlet velocities, on slug formation, propagation, and pressure drop. Results show that increasing the power-law index (n)—indicating higher shear-thickening behavior—significantly modifies the velocity profile, reduces apparent viscosity near the wall, and leads to a positive pressure drop compared with Newtonian cases. Additionally, this study introduces a hybrid machine learning approach combining Proper Orthogonal Decomposition (POD) and Long Short-Term Memory (LSTM) networks for predicting transient slug flow behavior. The POD method reduces the high-dimensional LBM data by capturing the dominant flow dynamics, while LSTM predicts the temporal evolution of these modes with minimal error. The POD+LSTM framework demonstrates significant reduction in computational cost for real-time prediction, with high accuracy, particularly when trained with larger datasets. Validation of the LBM simulations against benchmark tests, as well as comparisons with analytical and numerical solutions from the literature, confirms the reliability of the proposed LBM solver for simulating complex two-phase fluid flows. This research provides valuable insights for both fundamental understanding and practical applications of slug flow in multiphase systems, enabling real-time monitoring and optimization of two-phase flow systems.