Surface contamination caused by insects or debris can introduce discrete roughness elements (DREs) on aerodynamic surfaces, generating high-amplitude disturbances that promote boundary-layer transition and subsequently a turbulent wedge. We investigate whether a distributed surface pre-texturing can mitigate these disturbances and delay/avoid transition. To optimize the vast, high-dimensional parametric space describing the pre-texture geometries (such as heights, wavelengths, and spatial patterns), we introduce a novel framework for data-driven optimization utilizing Deep Reinforcement Learning (DRL) paired with neural network surrogate models. Three-dimensional simulations are performed over a flat plate with a Blasius profile as inflow for two different settings: a baseline case consisting of a DRE and another with a DRE surrounded by uniformly distributed cubical surface roughness (texture). In the baseline configuration, the DRE generates elongated low-speed streamwise streaks that amplify and interact downstream, leading to the development and lateral spreading of a turbulent wedge. This behavior is accompanied by strong velocity fluctuations and elevated turbulence intensity. In contrast, the distributed surface roughness/pre-textured case substantially weakens the DRE-induced disturbances. After an initial transient, the flow remains laminar over the computational domain, and the turbulent wedge observed in the baseline case is absent. Upstream of the DRE, pre-texture reduces the local velocity incident on the DRE and consequently lowers its effective roughness Reynolds number. Downstream of DRE, the texture-induced near-wall structures interact with the DRE wake and suppress further disturbance amplification.
This work is licensed under the Creative Commons Attribution-NonCommercial 4.0 International License. Any reuse must credit the author(s) and provide a link back to this page.