The Impact of Real Roughness Features on Boundary Layer Transition

In this work, the effects of realistic roughness scales on boundary layer transition are investigated using high-resolution scale-resolving simulations. This is in contrast to most of the roughness-induced transition studies reported in the literature based on ordered and well-defined surface roughness elements. Two highly irregular surface roughness patterns are generated from a given roughness patch by selectively filtering out the higher frequencies. The transitional behavior of a laminar boundary layer developing over these roughness scales is examined at \(\text {Re}_{\delta _{in}^{*}}=360\) and 540, defined in terms of the inflow velocity and inlet displacement thickness. The impact on transition is explored by examining the instantaneous and time-averaged flow fields. The results show that the transition onset is sensitive to the roughness scales: the inclusion of finer scales reduces the spacing between roughness features thereby constraining the lateral development of the flow. The streaks are weaker due to the mutual sheltering effect and the finer scales are shown to promote spanwise inhomogeneity of flow in the transition region. This effect is found to be much more prominent at low Reynolds numbers. In contrast, filtering out the finer roughness scales results in an earlier transition onset, caused by strong streaks developing from the horseshoe vortices wrapping around sparsely packed roughness features. Further cases are studied to investigate the spatial features of roughness that are important for transition. A series of high-fidelity simulations using selective retention of roughness features are performed at \(\text {Re}_{\delta _{in}^{*}}=360\). The transition onset can be predicted satisfactorily by retaining the dominant scales (20 tallest peaks in this study) from the original rough surface while the valleys and fine-scale features are shown to have minimal effect. In addition, we demonstrate that modifying the roughness patch or Reynolds number during the simulation alters the transition onset, which gets quickly established within \(\approx 1725k_{rms}/U_{in}\). These findings have the potential to reduce the computational cost and further aid in improving the transition correlations employed in low-fidelity simulations.