Flow over Ice Roughness Patch

This talk reports a DNS dataset of flow over rime and glaze ice roughness and their modeling in the context of Reynolds-averaged Navier-Stokes (RANS) and wall-modeled large-eddy simulation (WMLES). The flow configuration is a channel, with the inflow generated from a plane-channel precursor. The ice roughness patch comes from scans of realistic icing on airfoils, and the roughness height is set to be about 1/6 of the half-channel height. The DNS results show that the mean flow is largely parallel and that the dispersive stress is negligible outside the ice roughness. A delayed response is observed in the Reynolds stress, but the mean flow seems to attain equilibrium rapidly irrespective of the type of ice roughness. Encouraged by these observations, we invoke equilibrium-type modeling assumptions. Specifically, we formulate a mixing length model to close the RANS equations and adopt the equilibrium rough-wall model for LES wall modeling. Both models require equivalent roughness height information, which can be obtained through roughness correlations using morphological parameters as inputs. In this study, two correlations by Bornhoft et al. and Flack & Schultz are considered and applied locally to account for the streamwise variation of the ice roughness. Both the RANS and WMLES results are promising, capturing, to some extent, the streamwise variation of the skin friction in the flow. In summary, we conclude the following: Firstly, we do not observe a strong history effect in the mean flow—at least for the roughness patches and length scales considered here. Secondly, equilibrium-type models provide reasonable skin friction predictions—again, at least for the roughness and length scales considered here.