Discriminator-Based Inner Layer Forcing Model for Wall Bounded Flows over Rough Walls

Simulations of turbulent flows over rough walls are usually linked to prohibitively high computational cost, hence various simplified models are often used to consider the effects of roughness on the outer flow. However, the model performance strongly depends on the proper selection of parameters linked to the roughness geometry. We present an inner layer forcing model based on training a discriminator that can distinguish between well-resolved flow fields over rough walls and under-resolved ones over smooth walls. The discriminator is integrated as an inner layer forcing model by introducing a body force term to Navier-Stokes equation. The under-resolved flow fields on a coarse grid are dynamically modified to mimic the well-resolved fields over rough walls. The discriminator-based inner layer forcing models are applied to three rough surface topographies (k+ = 15.0, 16.2, and 27.0) at Re_tau = 180. The predicted mean velocity and rms profiles on coarse grids show good agreement with roughness-resolved simulations. The relative error of Ub+ is less than 4% for low k+ cases when a proper forcing location is adopted.