The use of wavelet-based optical flows for wall-bounded turbulent flows

Abstract

Experimental investigations of turbulent boundary layer flows requires the ability to resolve the velocity at high spatial resolution down to the wall. Particle image velocimetry (PIV) has been the workhorse of experimental fluid mechanics for many decades. However, the principles of cross-correlation based PIV often make it challenging to resolve the viscous sublayer with sufficient vector resolution, while still being able to resolve a significant portion of the velocity in the outer boundary layer regions. Optical flow is an alternative approach to velocimetry; it describes the apparent displacement of brightness intensity patters in an image sequence, and it is capable of determining the velocity field with impressive vector resolution down to the pixel scale. Originally developed by the computer vision community, variants of optical flow have been adopted within the fluid mechanics research community. However, in comparison to PIV, optical flow is far-less developed and applied within experimental fluid mechanics, since it is often more intimidating to learn and develop. Thus, the capabilities of optical flow and the understanding of how to best develop optical flow for accurate velocity estimations is often lacking. In this seminar, we will evaluate the capacity of wOF to resolve a turbulent boundary layer flow field with high accuracy and impressive spatial resolution. We will first assess the sensitivity of wOF findings to the regularization parameter (λ). This analysis will describe the effect of under-regularization or over-regularization for accurate measurements of velocity and wall shear stress, the latter of which is important for calculation of non-dimensional boundary layer parameters. This analysis is performed on synthetic DNS data of a turbulent channel flow. wOF findings will be compared with PIV as a benchmark to discuss wOF’s performance against the state-of-the-art. wOF will then be applied and assessed to experimental particle images acquired within a turbulent boundary layer. We will discuss the optimization of wOF in the absence of ground-truth data, and the capability of wOF to extend beyond PIV as well as particle tracking velocimetry, when wOF parameters are chosen appropriately.