Computational Fluid Dynamics (CFD) Validated Surrogate Modelling of Drag Reduction in Multiphase Non-Newtonian Turbulent Pipe Flow

Turbulence-induced pressure losses significantly increase pumping power in pipeline transport, motivating the use of minute quantity of polymer-based drag-reducing agents (DRAs) such as polyacrylamide (PAM), polyethylene oxide (PEO),and polyisobutylene (PIB),mitigate these losses by aligning the flow direction, stretching under shear stress, and damping turbulent eddies, thereby, suppressing turbulence within the viscous sublayer, buffer layer, and logarithmic region. Despite extensive research on Computational Fluid Dynamics (CFD) simulations, polymer turbulence models, and Artificial Neural Network (ANN) surrogates individually, prior studies rarely integrate layer-resolved polymer mechanisms with predictive data-driven modeling for multiphase flows. The present study presents a novel hybrid methodology combining high-fidelity CFD, experimental validation, and an ANN surrogate specifically tailored for polymer-induced drag reduction in multiphase non-Newtonian turbulent flows. CFD simulations use very refined polyhedral meshing with Shear Stress Transport (SST k-ω) 2-equation turbulence model, which outperforms the features of Realizable k-ε, Renormalizable k-ε, and standard k-ε with a smaller number of equations (i.e., 2-equations), computational power (i.e., <128 GB), and more control on turbulence by increasing turbulence dissipation rate (ω) and increasing damping turbulence, providing superior accuracy for near-wall resolution. The ANN surrogate, trained on experimental data and validated, including 80:20 training to validation ratio, predicts friction factor (f), pressure drop (∆p), velocity profile, and drag reduction (DR) within ±10-15% for most of the dataset, accounting for layer-specific polymer effects. This integrated framework offers mechanistic insight and rapid correlated predictions, bridging the gap between empirical observations and practical pipeline design. Resolving polymer-turbulence interaction across all near-wall layers enables the energy-efficient optimization system.