Experimental and numerical study of the turbulent boundary layer over shallow dimples.

• TU Delft drag-reduction concept by dimples under turbulent boundary layer is tested. • New experiment and Direct Numerical Simulation predict drag increase of at least 1%. • Design variations, tested by DNS, failed to resolve the issue. • Doubly-periodic boundary conditions with source term facilitated DNS. A very convincing experimental paper from TU Delft, of 2016, motivated this joint study. They directly measured a reduction in drag of up to 3% for the developed turbulent boundary layer flowing over an array of shallow dimples, of diameter comparable with the boundary-layer thickness. The Reynolds numbers were not high, but nor marginal either. Our experimental set-up has been well validated by previous riblets studies. The simulation approach contains more new features, which are described, and passed the essential tests. It is a Direct Numerical Simulation, at the correct Reynolds number. The key finding is that both of our approaches predict a drag increase of at least 1%, with an uncertainty of the order of 1%. Not having identified any flaw in the 2016 work even through discussions with the authors, we are currently unable to explain the contradiction. We do note however that these initial results had a Reynolds-number dependence which we consider puzzling, and also that a 2018 paper from TU Delft by almost the same set of authors presents some less favourable findings. [ABSTRACT FROM AUTHOR]