2.1 Flow facility and model
The experiments were performed in the water tunnel of the Laboratory for Aero and Hydrodynamics at Delft University of Technology (Fig.
1), which had a cross-section of 600 × 600 mm
2. The boundary layer was created over a flat Plexiglas plate with an elliptical leading edge placed vertically in the tunnel, where at the sides of the plate, the flow was bounded by the tunnel bottom wall and the free surface (i.e., the plate protrudes the air water interface). The zigzag trip was put 145 mm downstream of and parallel to the leading edge and had the following dimensions: the height was 1.6 mm from the wall, the width was 11 mm in the streamwise direction, and finally the pitch was 6.0 mm in the spanwise direction (Fig.
1). Furthermore, the zigzag top angle is 60 degrees. The trip is a tape that sticks directly onto the model surface and is available from Glasfaser Flugzeug-Service GmbH. Similar trips, mainly with a different height, are frequently used on gliders and in aerodynamic research (examples from Delft include Van Rooij and Timmer
2003 and Boermans
2006). The free-stream velocities
U
e
considered in this study were 0.21, 0.29, and 0.53 m/s with a free-stream turbulence intensity level below 0.5% in all cases (for additional details on the tunnel with model see also Schröder et al.
2008).
The x, y, z system of coordinates, and associated u, v, w velocity components, in this study is defined with respect to the transition trip, where y is the distance to the wall, z the coordinate along the trip, and x the distance from the trip along the wall. At the measurement location, the free-stream flow direction is tilted by approximately 5 degrees with respect to the trip resulting in a non-zero average spanwise velocity w (i.e., cross-flow). It is possible that this cross-flow is caused by the difference in boundary conditions at both sides of the plate (solid wall and free surface) combined with the pressure field near the elliptical leading edge, which also results in static surface waves locally. Note that the x-direction will nonetheless be referred to as the streamwise direction. Although the tilt was unintended, it does represent a more realistic situation occurring in practical applications on, for instance, actual wings or bluff bodies.
Below, the velocities and distances are made dimensionless using the free-stream velocity,
U
e
, and the undisturbed laminar boundary layer thickness,
δ
0
, at the trip location,
x
0
= 145 mm. Assuming Blasius laminar velocity profiles, the thickness
δ
0
is estimated by (White
1991):
$$ \delta_{0} = 5.0{\raise0.7ex\hbox{${x_{0} }$} \!\mathord{\left/ {\vphantom {{x_{0} } {\sqrt {Re_{x0} } }}}\right.\kern-\nulldelimiterspace} \!\lower0.7ex\hbox{${\sqrt {Re_{x0} } }$}}\quad {\text{with}}\quad Re_{x0} = \frac{{U_{e} x_{0} }}{\nu } $$
(1)
which results in
δ
0
= 4.2, 3.5 and 2.6 mm for the free-stream velocities in Table
1. Also shown in Table
1 is the estimated Reynolds number of the undisturbed boundary layer at the trip location,
Re
θ, 0
. These are well below the threshold for which a transitional or fully turbulent boundary layer may exist, i.e.,
Re
θ,
= 162 and 320, respectively (Preston
1958). Therefore, to cause early transition, not only a disturbance but also an added momentum loss is required. The surface-mounted roughness, in this case the zigzag strip, provides these both by means of the flow structures introduced in its wake and its overall drag.
Table 1
Undisturbed laminar boundary layer thickness, δ
0
, Reynolds number based on the momentum thickness, Re
θ,0
, at the trip position, Reynolds number based on the streamwise trip position with respect to the leading edge, Re
x0
, and the critical trip heights, k
cr
, according to three different methods for each free-stream velocity considered
0.21 | 4.2 | 116 | 3.0·104
| 1.6 | 2.0 | 4.0 |
0.29 | 3.5 | 136 | 4.2·104
| 1.2 | 1.6 | 2.9 |
0.53 | 2.6 | 184 | 7.7·104
| 0.8 | 1.0 | 1.6 |
The current trip can, furthermore, be compared against a few common engineering criteria available in the literature for forcing turbulent boundary layer flow (Table
1). First, Braslow and Knox (
1958) provide an empirical relation between the flow conditions in the undisturbed laminar boundary and the minimum roughness height,
k
cr
, needed for tripping it, which is based on extensive wind tunnel testing. In particular, the method employs a Reynolds number,
Re
k
, based on the flow velocity at the trip height, which has a critical value that depends on the roughness geometry (e.g., 2-D wires or isolated 3-D elements). The original work of Braslow and Knox contains a value of
Re
k
for 2-D tripping wires (
Re
k
= 300), but not for zigzag strips. Other authors (Van Rooij and Timmer
2003), however, have suggested that the latter are more efficient, and hence they have proposed a lower critical Reynolds number for zigzag trips (
Re
k
= 200). Due to this uncertainty, the resulting minimum trip heights for both critical Reynolds numbers are listed in Table
1. It can be seen that for
U
e
= 0.21 m/s the current trip is approximately at the design condition (slightly over- or under-tripping depending on the actual
Re
k
used). At higher free-stream velocities, the current forcing is somewhat stronger than necessary.
An alternative method for wires is mentioned in the book by White (
1991, p 386; after the work of Gibbings
1959). It simply considers the free-stream velocity (compared to the boundary layer velocity at the trip height before) when defining the critical Reynolds number for the roughness height. The resulting trip heights (Table
1) are clearly higher compared to the previous method, which is illustrative for the arbitrariness in defining and determining when transition would be effective.
The case with
U
e
= 0.21 m/s most closely resembles the design criteria for turbulence forcing according to Braslow and Knox (
1958); therefore, the focus in this paper will be on that condition. Moreover, the results from the other cases are qualitatively similar and quantitative differences mainly concern the downstream length over which the transition takes place (that is, shortening with increased forcing).
Finally, the spanwise wavelength of the zigzag strip (i.e., 6.0 mm) is taken to be in between the boundary layer thickness,
δ
0
, and 2.8
δ
0
(Table
1), which represent the most energetic spanwise wavelength in the outer layer of a turbulent boundary layer (e.g., Elsinga et al.
2010) and the spanwise wave length experiencing maximum spatial energy growth in a laminar boundary layer according to the work of Andersson et al. (
1999), respectively. These are the length scales expected to dominate the flow during and/or after transition, and the current trip acts within that range.
2.2 Tomographic-PIV setup
The tomographic system consisted of four cameras (LaVision Imager Pro X) with a 2,048 × 2,048 pixels image format and 14-bit gray-scale dynamic range, which were mounted with Scheimpflug adapters and lenses (Nikkor) with a f = 60 mm focal length and a f/16 aperture. The off-axis viewing angle was approximately 30 degrees in air, reducing to 22 degrees in water due to the changes in refractive index at the tunnel wall. Given the small lens aperture (and therefore large depth-of-field), no prisms were deemed necessary to correct for the effects due to refraction. The working fluid, which is water, was seeded with 56-μm polyamide tracers up to a concentration equivalent to particle image density of 0.03 particles/pixel. These particles were illuminated by a dual-cavity frequency-doubled 200 mJ/pulse Nd:YAG laser in a 7-mm-thick sheet touching the wall. The total measurement volume located directly behind the trip was 120 × 55 × 7 mm3 in the streamwise, spanwise, and wall-normal direction, respectively, which was imaged at a resolution of 18.5 pixels/mm. The recording rate was constant at 2 Hz, while the time separation between the light pulses was adjusted between runs to yield an approximate 20 pixels particle displacement in the free stream for the three velocities considered.
The instantaneous particle intensity distribution was reconstructed in 3-D space using the MART tomographic algorithm (Elsinga et al.
2006). Compared to the original images, the volume resolution was reduced to 15.7 voxels/mm in order to reduce the memory requirements for the computation of the tomographic volume reconstruction. Image pre-processing (that is, background subtraction and Gaussian smoothing using a 3 × 3 pixel filter length) and volume self-calibration (Wieneke
2008) were applied to improve the reconstruction.
The particle displacement field was obtained from these reconstructed volumes using an iterative cross-correlation technique with multigrid and window deformation (Scarano and Riethmuller
2000). The final cross-correlation volume size was 28 × 28 × 28 voxels corresponding to 1.8 × 1.8 × 1.8 mm
3, which resulted in 258 × 123 × 17 vectors per snapshot using 75% overlap of interrogation volumes between adjacent correlation positions. At each flow condition, a dataset consisting of 1,000 of such velocity snapshots was acquired.
The current spatial resolution is comparable to the trip height and emphasizes the larger scales of motion in the transition. However, owing to the limited Reynolds number at the trip location (Re
k
~ 200), much smaller energetic flow scales are not expected to be present.
The suitability of the tomographic-PIV technique for the study of coherent structures in wall-bounded flow has already been established previously (Elsinga
2008, Elsinga et al.
2010). The uncertainty of the particle displacement in these measurements has been estimated at approximately 0.2 pixels, which was found to apply here as well. The uncertainty was assessed based on an analysis of the divergence in the measured velocity fields (for details on this method we refer to the book by Adrian and Westerweel
2011).