ICMAR 2014 NUMERICAL AND EXPERIMENTAL REALIZATION OF AN INFINITE-SWEPT-WING BOUNDARY-LAYER FLOW IN A WIND TUNNEL

Transcription

1 ICMAR 2014 NUMERICAL AND EXPERIMENTAL REALIZATION OF AN INFINITE-SWEPT-WING BOUNDARY-LAYER FLOW IN A WIND TUNNEL A. Hanifi 1,2, S. Hein 3, D.G. Romano 4, M. Minervino 4, W. Würz 5, V.I. Borodulin 6, A.V. Ivanov 6, Y.S. Kachanov 6 1 Linné Flow Centre, KTH Mechanics, SE Stockholm, Sweden 2 Swedish Defence Research Agency, FOI, SE , Stockholm, Sweden 3 DLR - Institute of Aerodynamics and Flow Technology, Göttingen, Germany 4 Piaggio Aero Industries S.p.A., 34, Via Campi Flegrei, Pozzuoli (NA), Italy 5 Institut für Aerodynamik u. Gasdynamik, D Stuttgart, Germany 6 Institute of Theoretical and Applied Mechanics SB RAS, , Novosibirsk, Russia. Background Transition to turbulence in boundary layer flows is a process involving several steps: receptivity to ambient disturbances, linear and non-linear disturbance growth, non-linear breakdown and turbulent spot or wedge formation and growth, and finally the establishment of the fully turbulent boundary layer. Ambient disturbances may be in the form of free stream turbulence or sound, but also vibrations of the surface as well as surface roughness may have a distinct role in setting up the initial conditions for further disturbance development. The details of the transition process depend on if the base flow is two- or three-dimensional, whether the flow is subjected to a pressure gradient (favorable or adverse) or not, and if the surface has a curvature (convex or concave) or is flat. Although many steps in the process of these different cases are fairly well understood, the receptivity is in all cases the least researched as well as least understood mechanism, despite the fact that it defines the starting point for the further development of the disturbance field. The presented work is a part of activities within the European project RECEPT 1. The project includes three different receptivity experiments Experiment A : Linear-receptivity experiments on excitation of cross-flow (CF) waves by fully controlled unsteady free-stream vortices due to scattering on localized controlled surface roughness (waviness). Experiment B : Linear-receptivity experiments on excitation of 3D (in general) Tollmien Schlichting (TS) waves by fully controlled unsteady free-stream vortices due to scattering on localized controlled surface roughness (waviness). Experiment C : Investigation of nonlinear thresholds of surface-roughness receptivity mechanism leading to a jump from an evolutionary scenario of transition (i.e. through development of CF-or/and TS-modes) to an abrupt scenario of transition (i.e. the one attached to the roughness element) depending on the roughness size and the free-stream turbulence level. In this paper we describe design of the models and the experimental set-up, as well as flow quality measurements through comparison with numerical simulations. Results of receptivity measurements are presented in two other paper presented at ICMAR 2014 (see Refs. 2, 3). Wing-model geometry The receptivity experiments will be performed at two different angles of attack. A negative one where the cross-flow (CF) perturbations are dominant and a positive one with Tollmien Schlichting (TS) waves being the major source of boundary-layer instability. A number of different profiles were Current address C.I.R.A., SNC via Maiorise, Capua (CE), Italy A.Hanifi, S. Hein, D.G.Romano, M.Minervino, W.Würz, V.I. Borodulin, A.V. Ivanov, Y.S.Kachanov,

2 Section No. 2: Small Hall of Scientific House considered in preliminary phase of experiment design. Based on the physical constraints, the chord length of the model (measured normal to leading edge) was chosen to be 0.8m and the sweep angle to be 35 o. Based on the stability analysis, a NACA wing profile was chosen. This profile has the desired stability characteristics at both of the two different angles of attack. Further, to delay occurrence of the laminar separation bubble on the lower side of the airfoil at negative angle of attack, the lower-side shape was significantly modified (see Fig. 1). At 5 o angle of attack the boundary layer at upper surface, due to negative pressure gradient, is only unstable for CF waves. However, the amplification of these perturbations is low which makes the contamination of results by uncontrolled perturbation less critical. N-factors of stationary CF vortices at two different inflow velocities are given in Fig. 1. Detailed of stability analysis can be found in Ref. 1. The chord length is large in comparison with the test section dimensions which required contoured test section walls (top and bottom) in order to ensure good spanwise homogeneity of the flow over the model, i.e. the flow should resemble that of an airfoil of infinite span. The procedure of designing the contoured sidewalls is discussed in following section. Model manufacturing For the manufacturing process the wind tunnel model was built as a 2D airfoil section and then cut on both sides under an angle of 35 to provide the necessary sweep angle. The model was manufactured in NC-milled negative molds. The material for those molds is SIKA SLABS M610, a polyurethane foam with a very high density. A special milling technique with an inclined face cutter provides a surface roughness smaller than mm. This roughness is smoothed by carefully sanding until all milling tracks disappear. Then epoxy resin is used to remove the porosity of the surface. The mold is prepared with release agent and then UP-coating resin (Larit T-35) is sprayed into the mold. The skin of the models is built as a symmetrical carbon-fiber/glass-fiber/carbon-fiber sandwich with 5mm wall thickness. To reach the desired smoothness, it is finished with wet sanding (grit No.1200) and polished (Rot-Weiß-Paste). The surface roughness was measured with a high precision surface measurement device (DIAVITE DH-5). The outcome of the roughness measurement can be quite sensitive to the settings of the device, mainly if a mixture of roughness and waviness is present. The discrimination between roughness and waviness is done by a predefined high pass filter implemented in the measurement device. This filter was set to the longest possible distance, which corresponds to 2.5 mm (roughness elements with dimensions larger than this value are treated as waviness). Two scans were conducted with measurement 2 Fig. 1. Geometry and potential velocity distribution of the original and modified NACA airfoil (2D, AoA = 5 ). Nonlocal N-factors for stationary CF modes on upper side of modified NACA airfoil at AoA= 5 (PSE computations). Middle: U inf =10 m/s (spanwise wavenumber = [ ] 1/m). Right: U inf = 15 m/s. (spanwise wavenumber = [ ] 1/m).

3 ICMAR 2014 length set to 15mm, which is the maximum. Mean roughness values of 0.98 m were obtained. The airfoil is equipped with arrays of static pressure taps for measurements and adjustment of the chordwise pressure distributions. Parts of surfaces are covered with graphite to make them electroconductive. Design of the contoured sidewalls In order to satisfy the infinite swept-wing approximation, it is necessary to design the sidewalls of the wind tunnel to have the streamlines turned in the correct direction (i.e. to have the isobars parallel to model s leading edge). To this aim, a design procedure based on the following steps has been defined Generation of a parametric CAD model for the swept-mounted wing installed in the wind tunnel with flat walls. CFD mesh generation associated to the CAD model. CFD flow simulations for the infinite span WT model, replacing solid sidewalls with periodic boundaries. Extraction of typical streamlines shape from the CFD solution back to the CAD environment. CAD design of the contoured sidewalls based on the reference streamline. CFD mesh generation for the contoured WT configuration. Flow simulation for the contoured WT configuration. A fully parametric CAD model developed in CATIA v5 R19 is used to design the sidewalls. The starting swept-mounted airfoil is obtained through a best fitting of about 600 points defining the airfoil shape. In particular, a 5th order NURBS with 19 control points has been used. shows a view of the parametric CAD model. The model is mounted vertically, and positioned at mid-length of the test section. The rotation axis for angle of attack modification is parallel to the leading edge. Two different angles of attack (AoA) have been considered: 5 and CFD evaluations have been performed using CFD++, the commercial CFD code used by Piaggio Aero Industries, which is based on the finite volume method. A 2 nd order spatial discretization has been used with a multigrid technique (4 W-cycles on 20 levels) to accelerate the convergence. The boundary layer has been integrated directly to the wall around all solid boundaries. The 3-equations Goldberg s κ-ε-rt turbulence model has been applied 6. Both unstructured and structured meshes have been used and in both cases the y + was of order of unity around all solid adiabatic viscous walls. Sea level ISA conditions have been considered to define the values of the ambient temperature and pressure. A WT inlet mass flow rate equal to kg/s has been imposed. Transition has been imposed at 70% of chord length on the upper side and at 10% on the lower side. Fig. 2. Overview of the CAD model (left) and CFD mesh (right). 3

4 Section No. 2: Small Hall of Scientific House The ideal design would be obtained by designing a 3D surface enveloping all the streamlines normal to the model. However, to simplify the manufacturing of sidewalls, only one reference streamline for the upper and another one for the lower side of the model have been considered and should provide approximate satisfaction of the sweep condition in the vicinity of the airfoil surface. Besides, only a limited length of the WT sidewalls has been modified (about 2m). The above reported simplification causes a reduction of the parallelism of isobars to the leading edge, which is not preserved close to the intersection of the model to the wind tunnel walls. However, this is not a big issue, since to perform the present WT tests it is enough that the isobars are parallel to the leading edge in the central area of the model itself. This condition is respected by both the designs carried out at AoA of 5 and In particular, in the first design the isobars look perfectly parallel on the upper side, but not on the lower, while in the second design isobars are parallel on the lower side but not on the whole upper side (see Fig. 3). The preliminary design was based on RANS simulations including the test section and the boundary layers developing on the wind tunnel walls. In principle the design was based on the assumption that the shape of the wind-tunnel walls should resemble the streamlines in the potential flow for a spanwise infinite wing. However, two streamlines (one going to the suction side and one to the pressure side of the airfoil) next to each other far upstream would separate in the spanwise direction as they approach the wing. This would mean that walls had to have a step along the stagnation streamline that was not deemed to be satisfactory since it may create streamwise vorticity. Thus, the wall was made continuous in front of the wing. This gave rise to a small separation region near the wall but it was deemed to be acceptable. However at the trailing edge a rather large step from the pressure to the suction side was necessary in order to follow the correct shape of streamlines. From flow visualizations this step did not seem to generate any significant disturbances in form of streamwise vorticity. The true shape of the streamlines and, hence, the sidewalls is, of course, three-dimensional. However, it is difficult to manufacture and use such walls in an experiment. Therefore, the sidewalls were made two-dimensional, i.e. uniform in the wall-normal direction and corresponded a b c d Fig. 3. Isobars on the (a) lower model side for a rotation of 5, (b) upper model side for a rotation of 5, (c) lower model side for a rotation of +1.5, (d) upper model side for a rotation of

5 ICMAR 2014 Fig. 4. Wall contours for the 1.5 o angle of attack case. Flow is from right to left. to the streamlines calculated for the potential flow just outside the boundary-layer edge. This approach turned out to be very successful. The geometry for the 1.5 o angle of attack is shown in Fig. 4. The total vertical height of the flow region after the initial contraction was about 65 cm. Wind tunnel facility and measurement equipment The experimental studies are performed in the MTL (Minimum Turbulence Level) wind tunnel at KTH, Stockholm 5. The tunnel has a test chamber with a cross section of 1.2 m wide and 0.8 m high, with a total length of 7 m. The disturbance levels in terms of turbulence and noise are extremely low (Tu 0.02% at 10 m/s) and the tunnel is therefore an excellent facility for stability and receptivity research. In addition the tunnel has excellent constancy characteristics, both in terms of speed and temperature. To guarantee the latter, the wind tunnel is equipped with a heat exchanger that allows keeping the temperature constant within 0.1 o C even for measurements stretching for several days. An accurate traversing system has been designed and manufactured to make detailed boundarylayer measurements available. The accuracy of the hot-wire probe positioning has been measured by means of a non-contact laser optical displacement system (Micro-Epsilon ILD ) showing an accuracy of 2 microns. Model and traversing system are illustrated in Fig. 5. Fig. 5. Model installed in WT (left), traversing system Komarik with its controller (middle), Komarik installed for boundary-layer measurements (right). Results of measurements To avoid flow separation on the wing, experiments were carried out by imposing transition through a transition strip on the model at 70% on the upper side and at 10% on lower side. Four main types of the base-flow measurements have been performed: (i) pressure-tap measurements, (ii) x-hot-wire two-component potential-flow velocity measurements at relatively large distances from the airfoil surface, (iii) hot-wire and wake-wire two-component velocity measurements just above the boundary-layer edge, and (iv) one-component hot-wire measurements inside the boundary layer. The most detailed measurements at conditions of experiments A were carried out at the incident flow velocity of 10 m/s, while at conditions of experiments C, some of measurements were performed also at higher free-stream speeds. It was found that in absence of 5

6 Section No. 2: Small Hall of Scientific House turbulence-generating grids the potential-flow structure is speed independent. Despite the rather small distance of model to wind tunnel walls as compared to the chord of the wing profile, the measurements showed an excellent spanwise homogeneity as well as agreement with the calculated flow field. In Fig. 6, the measured and computed pressure coefficients for the case of AoA = 1.5 o are plotted. It shows a very good agreement between these data. Similar good agreement was found for AoA = 5 o. Fig. 6. Location of pressure taps (left) and comparison of measured and computed pressure distributions at the 1.5 o angle of attack case. Curves are from the RANS calculations and the symbols from the experiments. The measurements of the base-flow around the model have been performed using cross hotwire probes for two components of the potential-flow velocity vector field. For AoA = 1.5, measurements were made for three (x, z)-planes oriented vertical and parallel to the incident flow velocity, see Fig. 7. To compare the measured and computed flow field the CFD results were extracted at similar positions. Due to small differences in the magnitude of inflow velocity, both the measured and computed fields are normalized with their corresponding free-stream values. The results are shown in Fig. 8. As can be seen there, a close agreement between the measured and the computed values is found. 6 Fig. 7. Extend and location of planes for the free-stream velocity measurements. In particular, the spanwise variation of the streamwise velocity component does not exceed 1% at each fixed chordwise coordinate. Moreover, the spanwise potential-flow velocity component (parallel to the airfoil leading edge) is found to be practically independent of both the chordwise and the spanwise coordinate. The constancy of this velocity component is the inherent property of infinite-span swept wings and is often directly attributed to the satisfaction of the sweep condition. Measurements of potential-flow streamlines were performed by means of one-component hotwire measurements and wake-wire technique developed previously in ITAM. Here, three tungsten wirers (Ø 50 mm) mounted vertically on the airfoil (around x/c = 15%). The wake of these wires follow the potential streamlines. By detecting the spanwise location of these wakes at different streamwise stations one can find direction of the potential-flow streamlines. Distributions of streamwise mean velocity, measured outside the boundary-layer edge at various chordwise locations, are shown in Fig. 9 (left). The spanwise uniformity of the base flow and clear laminar wakes past three thing gilded tungsten wires is demonstrated in Fig. 9 (right). Here, the shapes of

7 Us/Use, UHW/UHWe Us, U [m/ ICMAR 2014 Fig. 8. Comparison of measured (left) and computed (right) streamwise velocity distributions at three different z y planes above the wing for the 1.5 o angle of attack case. The data is normalized with the incoming flow velocity. The plotted region corresponds to the planes marked in Fig. 7. Fig. 9. Streamlines measurements by Wake-Wire Technique (AoA = 5 o ). the measured potential-flow streamlines are presented in along with the corresponding average shape. It is seen that the streamlines are practically the same at all studied spanwise coordinates. This result supports the conclusion on the spanwise uniformity of the base flow and satisfaction of the sweep condition in the region of main stability and receptivity measurements. Detailed measurements of the flow inside the boundary layer have been performed using a single hot-wire prob. These data then have been compared to numerically obtained boundary-layer profiles. The latter are results of boundary-layer computations based on the pressure distribution taken from RANS computations of wing model installed in the wind tunnel. The measured and computed data for the AoA = 5 o case are compared in Fig. 10. As it is seen there, the experimental and theoretical profiles agree very well with each other. The experimental results do not display any d1, d2 [mm], H Directly d 1HW (directly measured measured) Calculated d 1Theory (calculated) Series2 d 2HW (directly measured) Series1 d 2Theory (calculated) Series6 H HW (directly measured) Series3 H Theory (calculated) x c - x csource [mm] Us/Use U/Ue y n [mm] Theoretical results: U s -profiles versus U-profiles (no any difference) x c x csourse = 490 [mm] Fig. 10. Comparison of measured and computed boundary-layer characteristics and normalised velocity profile (AoA = 5 o ) y n /d 1, y HW /d 1HW Experiment (z = 0 mm) Experiment (z = 60 mm) Experiment (z = -60 mm) Theory 7

8 Section No. 2: Small Hall of Scientific House spanwise dependence of their shape, supporting again the result on the observed spanwise uniformity of the base flow inside the boundary layer in the region of main stability and receptivity measurements. Summary In order to perform a set of receptivity experiments, an appropriate airfoil has been chosen and modified to make sure the boundary has the required characteristics. To insure infinite-swept wing approximation contoured sidewalls of WT have been designed. Further, an advance traversing system has been designed and manufactured which gives possibility of accurate measurements inside the boundary layer. The accuracy of the flow measurements has been checked through detailed comparison with the numerical simulations. All tests show good base-flow quality and spanwise homogeneity that is essential for the receptivity investigations. Results of receptivity measurements are presented in two other paper presented at ICMAR 2014 (see Refs. 2, 3). Acknowledgement The work reported here has been carried out within the EU-funded research project RECEPT (Grant Agreement no. ACPO-GA ). The experimental work includes significant contribution by H. Alfredsson, N. Tillmark, R. Örlü, S. Imayama and J. Vernet, their efforts are highly appreciated. REFERENCES 1. EU project RECEPT: RECEPTivity and amplitude-based transition prediction. URL: 2. Borodulin V.I., Ivanov A.V., Kachanov Y.S., Mischenko D.A. Experimental investigation of characteristics of steady and unsteady crossflow-instability modes developing in a 35-degree swept-airfoil boundary layer // Int. Conf. Methods of Aerophysical Research. June 30 July 6, 2014, Novosibirsk, Russia: Proc. / Ed V.M. Fomin. Novosibirsk: Inst. Theor. and Appl. Mech. SB RAS, URL: /users/libr/elib/confer/ ICMAR/2014/pdf/ Borodulin254.pdf 3. Borodulin V.I., Ivanov A.V., Kachanov Y.S. Experimental approach to investigation of excitation of crossflowinstability modes in a swept-wing boundary layer at scattering of freestream vortices on surface nonuniformities // Int. Conf. Methods of Aerophysical Research. June 30 July 6, 2014, Novosibirsk, Russia: Proc. / Ed. V.M. Fomin. Novosibirsk: Inst. Theor. and Appl. Mech. SB RAS, URL: ICMAR/ 2014/pdf/Borodulin256.pdf 4. Hanifi A., Hein S. Linear Stability characteristics of the boundary layer on the modified NACA airfoil: RECEPT project Technical Report TR D1.2, Lindgren B., Johansson A.V. Evaluation of the Flow Quality in the MTL Wind Tunnel. TRITA-MEK 2002:13. Technical Reports from Royal Institutes of Technology Department of Mechanics, SE Stockholm, Sweden. 6. Metacomp Technologies Inc., CFD++User Manual Version 8.1 P URL: 8