Abstract
Resolvent analysis is utilized to model the temporal mean of experimental measurements of a turbulent boundary layer over streamwise aligned, spanwise alternating rough surfaces, where large counter-rotating rolls are observed. Without detailed knowledge of the roughness geometry, the resolvent modes are generated using a smooth wall boundary condition and a spatio-temporal mean extrapolated using an analytical model. A least squares fit is then performed using the resolvent modes on the streamwise time averaged hot wire measurements, achieving good agreement with the experimental results. The spanwise and wall-normal velocity components predicted from the resolvent modes achieves a qualitative agreement with particle image velocimetry (PIV) measurements, correctly predicting the location, shape, and direction of the secondary counter-rotating rolls while under-predicting its magnitude. An eddy-viscosity model is then added into the streamwise momentum equation in the resolvent framework to augment its predictive capabilities, significantly improving the predicted magnitude of the spanwise and wall-normal velocity components. Finally, all three velocity components from the PIV measurement are utilized in the least squares regression, achieving good agreement with the experiments while reducing the degree of freedom to 0.18% of the experimental data. This work showcases the ability to use resolvent modes that can be computed with effectively negligible cost to model flow features unavailable from the experimental data and to efficiently represent experimentally measured flow fields over rough surfaces of industrial interests.
