UC Santa Barbara

This dissertation presents an experimental investigation of turbulent boundary layer flows arising over arrays of identical, two-dimensional bluff obstacles, motivated by contemporary engineering and geophysical flow applications. At sufficiently high Reynolds numbers, rough-surface boundary layers are primarily parameterized by the geometry of the individual obstacles, as well as by their spacing s, normalized by the obstacle height k. When the roughness enhances the turbulent fluxes between the near wall region and the outer flow, the log-law profile of the mean velocity is shifted closer to the wall. This shift is empirically determined, and is traditionally encoded by the equivalent sand grain roughness k_s, such that larger values of k_s/k represent surfaces that induce greater fluxes, for a given obstacle height.

Although square bar geometries and their effects on the flow have been studied extensively, arrays consisting of thin, plate-like obstacles have been understudied. For square bars, it has been demonstrated that, as the spacing decreases below s/k~7, the flow enters a different regime with drastically reduced turbulent fluxes, and with values of k_s/k that are over one order of magnitude smaller. However, it is not known whether plate-like obstacles may delay this transition to smaller spacings, or whether they may experience a less significant drop in k_s/k. This would be highly valuable in applications, as it would allow obstacles to be more densely packed without sacrificing fluxes between the obstacles and overlying flow.

The first part of this dissertation presents a review of the underlying physics for smooth and rough surfaces before discussing investigations of various roughness geometries, ultimately arguing that plate roughness is understudied. The experimental and processing methods are then described. Particle-Image Velocimetry in a water channel is used to obtain time-resolved velocity fields over bar and plate obstacles for 1< s/k < 10. It is found that plate obstacles can be packed closer together while still yielding a k_s/k that is one order of magnitude higher than for square bar geometries.