A pressure-driven three-dimensional turbulent boundary layer documented with stereo particle image velocimetry

A three-dimensional turbulent boundary layer (3DTBL) over a flat plate is carefully documented with Stereo Particle Image Velocimetry (SPIV). At the first measurement station, the boundary layer is nearly two-dimensional with the following parameters: Red211 = 3972, delta = 78.45 mm, Cf = 0.0032 and Uinlet = 9.0 m/s. The secondary flow is generated via the removal of mass through the side wall of the high speed test section of the Mark V. Morkovin wind tunnel. Thereby, the developed 3DTBL is subject to an adverse pressure gradient due to the turning of the initially two-dimensional flow. The wall shear stress vector and its direction are measured directly using oil film interferometry.; The results indicate that the TKE increases slightly due to the three-dimensional mean flow, mainly through changes in <u'2 >. The Reynolds stresses are all affected by the mean flow three-dimensionality with turbulence production and the pressure rate-of-strain tensor playing the major role in the dynamics of the Reynolds stresses. Measurements of Townsend's structure parameter indicate that it is significantly altered by the mean flow three-dimensionality, especially near the wall, as a result of a reduction in <u'v' >. The eddy viscosity within the 3DTBL is highly anisotropic. In addition, the shear stress vector in the plane of the wall leads and lags the gradient vector in different parts of the flow. The lag of the shear stress vector is mainly due to the residual history of the upstream two-dimensional turbulent boundary layer.; The longitudinal and transverse integral length scales are altered considerably by the mean flow three-dimensionality. For most of the components, the size of the energy containing eddies decreases with increasing mean flow three-dimensionality except for those responsible for <w'2> which increase in size. Analysis of the joint PDF's of the Reynolds stress <u'v '> indicate that the ability of the turbulent motions to produce u'v' sweeps is reduced. Conditional averaging produces turbulent motions consistent with an asymmetric version of the horseshoe vortex typically observed in two-dimensional turbulent boundary layers with the following exception: the angle of inclination relative to the wall can be much larger.