Deep learning of interfacial curvature: a symmetry-preserving approach for the volume of fluid method

Asim Önder, Philip Li-Fan Liu

Published 2022-06-13Version 1

Accurate estimation of the curvature of the fluid-fluid interfaces is essential for the success of Volume of Fluid (VOF) methods in surface-tension dominated flows. The present study employs artificial neural networks with deep multilayer perceptron (MLP) architecture to estimate the interfacial curvature from volume fraction fields on regular grids. Using input normalization, odd-symmetric activation functions and bias-free neurons, we construct a cost-effective MLP model that conserves the symmetries in the curvature fields. MLP models are implemented into a flow solver (OpenFOAM) and tested against conventional schemes in the literature. The symmetry-preserving MLP model shows superior performance compared to its standard counterpart and has similar accuracy and convergence properties with the state-of-the art conventional method despite using smaller stencil.