Aug 07, 2025
NEW RESEARCH BY ALEXANDER METELKIN & BERNHARD VOWINCKEL IN PHYSICS OF FLUIDS
The snapshot of the DNS data set used for training.
A recent study by Alexander Metelkin & Bernhard Vowinckel presents a data-driven model to predict hydrodynamic forces on spherical particles in flow. Using local volume fractions as input to a neural network trained on pr-DNS data, the method improves accuracy over position-based models and adapts to particles of varied size and shape.