WIND TUNNELS: DESIGN, CFD SIMULATION AND CONSTRUCTION.

Introduction.

“Designing is Art”. Inspired in “Starry Night”, Vincent van Gogh, 1889.?MOMA, New York.

It all started with a freehand sketch on some of the papers lying around (Figure 1): what if we built the biggest and most powerful wind tunnel in the entire University...? We already had several in operation, but we were convinced that a new atmospheric boundary layer wind tunnel (ABLWT) was the icing on the cake. And a phenomenon was unleashed that mixed equal parts illusion and difficulties, knowledge and mistakes, friends and companies, imagination and a few means...

With the work and collaboration of several professors of the Department of Energy, the Rectorate of the University of Oviedo (mainly through the Vice Rectorates of Infrastructure and Research) and the Administration Service of the Campus of Mieres, with funding from the Principality of Asturias (through FC- GRUPIN- IDI/2018/000205) that project is now a reality.

Figure 1: Preliminary designs.

Civil Aerodynamics (or Wind Engineering) [1,2] is an art. It is not only about deep wisdom and intuition, or genius, nor about feeling the flow (since it also relies on two other fundamental legs, the experimental [3-5] and the numerical simulation of the analytical equations of the fluid-structural phenomena and interactions), but it has a lot of that.

This paper lists the main components of the tunnel and compares the results obtained in the first tunnel calibration campaign with those predicted by theory or assumed by the Engineers. In fact, the initial design starts from assumed premises based on rules of good practice and the team's experience (more information at?www.antonionavarromanso.com). It has, therefore, nothing to do with the muses [HESIODUS: Theogony 77 and following], with music or with any other allegory, myth or metaphor of this kind... (well, maybe yes, who knows... what would Calliope and company say if they heard us talking like this?).

A good design, which ends up becoming art (is this possible?), handles equal parts experience and expertise, study and capacity, analysis, numerical simulation and experimentation (with the muses' permission), without any of these elements allowing, by itself, to disregard or belittle the others.

And the whole process culminates with the execution of the project and the confirmation that it works correctly.?Therefore, he distrusts those who only talk about music and do not compose; those who do not solve a problem or do not quantify their solutions. As we have gathered in conversations with several colleagues, world experts in the field, a good "Wind Engineer or Aerodynamicist" should be able to design, build and operate a wind tunnel made by himself.

General overview.

The new EVE50-ablWT (Figures 2 and 3) is?a subsonic wind tunnel for the study of aerodynamic problems in Civil Engineering (Wind Engineering)?and has been conceived as a closed circuit, with a closed test chamber and a sufficient length to develop a boundary layer in a controlled manner [6-7].

Structures such as bridges, buildings, solar panels, towers, wind turbines, cars, bicycles, etc. can be tested in it. The purpose of the whole design and construction is to ensure a very good quality air flow in the test chamber (where good quality means, basically, a uniform velocity and pressure distribution for the desired flow rates and with very low turbulence at the nozzle outlet). Additionally, it is possible to use the return and discharge pipes as secondary test chambers, of larger size and lower wind speed.

The main dimensions are 31.52 m long, 10.84 m wide, and variable height from 2.54 m in the standard section to 3.48 m in the backwater chamber.?The development of the length of its axis is more than 65 m.

Figure?2: Wind tunnel rendering.

Figure 3: View of the wind tunnel from the top of the curve.

To achieve this objective, a series of elements or parts into which the tunnel can be divided are necessary, elements that must have, in turn, a correct operation in terms of speeds and pressures; and that have each one of them a specific purpose. Figure 4 shows a plan of the installation.

In this first part (the second part will be called WIND TUNNELS: FLOW CHARACTERIZATION, and will include aspects related to the instrumentation [8,13]) all these elements are explained starting with their design [14] and function, showing a rendering of them prior to the execution, and, in parallel, the images once they are built.

The data of head losses, velocities and pressures have been measured at their mean values?(and, since the objective here is an initial confirmation of the goodness of the design, no high precision instruments have been used). Likewise,?the 3D numerical simulation by means of CFD?has been carried out with a not too refined mesh (1,000,000 elements), and k-?e?RNG turbulence model, with the intention of obtaining average steady state results quickly.

Figure?4: Figure 4: Wind tunnel plan.

The wind tunnel is built in wood (3 cm MDF and pine planks, about 20 tons) for reasons of economy and, also, for the versatility in the adaptation/substitution of any of the 61 sections or segments (plus 2 splice spools) that compose it. The structural part of the construction is not the subject of this article; all the elements are bolted together (approximately 10,000 screws), and different interior and exterior stiffeners have been built, as well as flaps and seals between modules, to guarantee their resistance and watertightness. The infrastructure rests on beams on the laboratory floor and has been anchored to the roof in those sections with a larger span. Figure 5 shows the elevations and floor plan of the tunnel:

Figure 5: Wind tunnel layout.

Fans.

Starting with the power plant (Figure 6),?the 4 fans, 1250 mm diameter and 12 blades, are installed in parallel (4x45 kW)?and provide the flow with enough energy (in the form of pressure jump) to move inside the tunnel. Their characteristic curves are (Figure 7):

Figure?6: Fans, 180 kW power.

Figure?7: Fan characteristic curve, and theoretical operating point at maximum flow rate.

Transition 1 and Straightener 1.

A small contraction (transition 1) to adapt the fan cross-section to the duct cross-section leads to the first straightener, which divides the cross-section into 9 channels and is responsible for partially dissipating the strong radial component of the flow at the outlet of the fans (Figure 8). To substantially improve the performance of the installation, a stator could be installed on the fans (currently in the design phase and to be carried out together with the rest of the final finishing touches to the infrastructure). Alternatively, or, in addition, diffusers could be designed at the outlet of each fan. The gain of each of these elements will be quantified at the end of the article.

Figure?8: Straightener 1 at the outlet of the fans.

Impulsion and Curve.

The drive conduit, approximately square in cross-section and?measuring 2.44 m high x 2.30 m wide?for construction reasons, leads to the first of the turns (Figure 9). The entire tunnel tube is finished off at its corners with an auberge at 45o and 15 cm on each side. This drive tube incorporates the first of the doors to access this part of the tunnel, and can be used as a secondary test chamber, given the correct distribution of average velocity across the width of the section. For this, it would be essential to have an additional honeycomb, and to reduce at least 27 times the gap of the straightener number 1, at its downstream end. In this case, a second entrance would have to be provided for inspection and maintenance of the fans.

Figure?9: Impulsion and curve.

It was decided to design the first loop in a circular shape, with a constant cross-section and equal to the discharge duct, instead of using the classic configuration of a straight section and 90o corners. In theory, some flow detachment was to be expected on the inside face of the bend. Measurements made so far seem to indicate that no separation occurs.

Figure?10: Flaps at the end of the curve.

Diffuser 1 and Straightener 2.

The curve flows into the first diffuser. Two flaps (see Figure 10) have been arranged in the last three and two segments of the curve, respectively, in order to correct the flow distribution in view of the certain flow detachment at the end of the curve and the beginning of the diffuser.

This diffuser is designed in straight line with a symmetrical enlargement in plan of angle?10o, and an expansion of the roof of about?8o. The parameter that conditions its aspect ratio and, therefore, its angle, is the length (the longer, the better). It incorporates at the beginning the second of the straighteners, which, as a continuation of the curve guides, distributes the flow homogeneously throughout the diffuser section, while partially mitigating the effect of the flow separation that undoubtedly occurs.

Figure?11: Straightener 2 at the diffuser 1.

After a first design of the flaps and this second straightener (Figure 11), CFD numerical simulations have been carried out with a detailed model (impulsion, curve, diffuser, settling chamber, nozzle and test chamber) to optimize the operation of these elements. Although the goal is to achieve a good quality flow in the test chamber, an enormous effort has been made in the design and construction of each part: in this case, the aim is to achieve a distribution of velocities and flow rates as uniform as possible, even before reaching the honeycomb. To this end, the length and position of the flaps, the straightener channels (and the deflectors at the outlet of the straightener, which will be discussed at the end of the article) have been analyzed parametrically and modified to achieve the necessary compromise between efficiency and economy of construction, minimizing flow detachments and head losses as much as possible. Another design alternative would be to entrust all flow redistribution to the settling chamber screens, at the cost of introducing a significant head loss (of the order of 100 Pa, in this case).

Settling Chamber.

The next element is the settling chamber, whose?dimensions of 3.38 x 4.74 m?have been taken to the maximum available in the space occupied by the tunnel. In it, the flow practically stops, and the pressure is maximum, thus starting the process of turbulence decrease.?

For this purpose, a?honeycomb?[15] is placed at the entrance of the tunnel (Figure 12), with?a mesh size of about 3.5 cm and a total length of 25 cm; this element (with a very high porosity and, therefore, a very low head loss) reduces the mainly transverse turbulence and the longitudinal one, channeling the flow through the channels.

Figure?12: Settling chamber and honeycomb

With a gap of a few centimeters, one or more screens can also be placed (depending on the needs), in order to reduce mainly the longitudinal turbulence, by means of a pressure jump (the porosity of this element is much lower than that of the honeycomb and produces a very significant flow uniformization).

The remaining turbulence scale that would need to be dissipated once the flow has passed through these elements would already be reduced by the available length just before the nozzle. Therefore, the length of the chamber must be at least as long as its wide.

It is important to note that the entire pressure chamber has flow distributed over the entire section (even without the need for screens) although this flow is of very low velocity (in a first approximation, it can even be considered 0). The settling chamber incorporates the two access doors to the tunnel and inside the ring.

Nozzle.

The nozzle is the element responsible for accelerating the flow and driving it into the test chamber and is shown in Figure 13. It must meet several requirements, controlled mainly through its shape and length.

The curve must link two very different sections (the ratio between the cross-sectional area of the backwater chamber and that of the test chamber is called contraction N, in this case?N=7.12), providing a uniform flow and very low turbulence, for which the variation of pressures along the development of the nozzle must be carefully calculated [16].

As indicated above, the preceding elements also play a crucial role in the quality of the flow that will finally be obtained at the nozzle outlet. In this case, a?logarithmic curve?has been chosen (actually, there are two, due to the slight dissymmetry of the settling chamber), with only one branch [17] since the total length of the wind tunnel had to be optimized.

Figure?13: Nozzle.

Test Chamber.

We now come to the essential element, which is the (main) test chamber, in which the models to be tested are placed. This?1.5 x 1.5 m square chamber?(Figure 14) has two clearly differentiated areas:

-???????in the first part adjacent to the nozzle, tests requiring uniform flow conditions (in this case,?the maximum velocity is over 45 m/s, about 165 km/h) and very low turbulence can be performed;

-???????in the second, after?10 m in length, problems requiring the development of an atmospheric boundary layer (a controlled flow with a variable velocity profile in height and with the appropriate turbulence spectrum and consistent in intensity and scale) can be studied. For the latter case, spires and roughness-generating elements are arranged on the chamber floor. The chamber has two removable frames, with a rotating table and two access windows.

The scale models of the prototype to be tested must, in general, fulfill the condition of not exceeding 10% blockage. Otherwise, the flow around the object would be significantly altered if the corrective measures are not carefully carried out [18].

Figure?14: Test chamber (main).

Diffuser 2 y Straightener 3.

Once the flow passes out of the test chamber, and to gradually reduce the velocity to minimize losses due to friction in the return duct and the detachment in the tunnel circuit turning, a diffuser is designed with the particularity that it is curved; this variation is materialized by increasing the distance between the walls and lowering the floor as the curve develops, until it meets the return duct cross-section.

Again, this solution has been preferred over the classic solution of a succession of diffusers linked with 90o corners, since it has a much finer behavior, despite increasing the difficulty of construction. In contrast to the first curve, in this one there will be flow detachment, mainly due to the diffuser opening and not so much to the radius of curvature (as we saw earlier). For this reason, a straightener or flap is designed that splits the diffuser in half in the vertical dimension (Figure 15), starting practically at the beginning of the diffuser.

Additionally, two deflectors are built at the end of diffuser 2 (see also Figure 20), in order to improve the uniform distribution of velocities at its outlet. The performance of this design is satisfactory, since part of the kinetic energy lost in the diffuser is recovered in the form of a percentage of the pressure in the return duct.

Figure?15: Diffuser 2, curved, and the middle flap.

Return and Transition 2.

Finally, the return tube of approximately square section, equal to the section of the impulsion duct, returns the flow to the fans, ending in a small diffuser (transition 2) to couple the section of the duct type to the section of the matrix of the 4 fans (Figure 16).

Figure?16: Return duct.

For safety reasons, a fence has been installed to prevent access to them, adding a metal net to protect them against the possible impact of any particle or element that may be released from the models located in the test sections.

As in the impulsion tube, it is possible to have another test chamber in this element (with?a speed of up to 17 m/s, about 60 km/h, as in the impulsion tube), for which a honeycomb would be needed again at the exit of the curved diffuser. Here, too, there are two doors or accesses to the tunnel.

Initial Calibration.

Regarding the initial calibration campaign,?the uniformity of the velocity measured?in the critical sections of the tunnel?has been above 95%?(considering that the measurements have been made with a portable windmill anemometer, it is expected to exceed 99% in the main test chamber during the characterization campaign). A first approximation of?the turbulence intensity?measured in the main test chamber?is 2%. Its measurement will be described in detail in the next part of the monograph.

Figure 17: CFD simulations, velocities (top) and turbulence intensity (bottom) through the horizontal midplane at the test chamber.

The numbers corresponding to the CFD simulation [19] (Figure 17) do not take into account the corrections made a posteriori in the flaps of both curves, nor the presence of straighteners, deflectors; in some of them, an equivalent head loss has been considered to simulate the honeycomb.

This simulation was carried out prior to the construction of the tunnel. The only modification to the original design that was necessary consisted of separating the fans from the outlet of diffuser 2. This reduced the head losses and improved the behavior of the pressure and velocity fields along the circuit.

The analytical results follow the methodology proposed by [20] adjusting them to our design by means of the hypotheses collected in [21].

Finally, the head losses have been measured with a digital manometer, at several points of the tunnel, confirming them for the straight sections with Colebrook's formulas. Detailed descriptions of the formulation can be found in [22-26].

It is indicated that these losses have been measured with two large openings (one in test chamber 1 and one in door 5) and without the finished tunnel sealing. The measured head losses (static pressure) and those calculated by analytical and numerical means are shown below (Table 1). And Figure 18 shows the distribution of the percentage of losses in each element of the tunnel.

Table 1: Head losses, comparison of different methods.

Figure 18: Percentage of static pressure loss in each element.

The differences observed are due to several factors, but in general, they demonstrate the correct design and more importantly,?a good performance of the wind tunnel.

The following graph in Figure 19 shows the total energy line, as well as the static and dynamic pressure in the circuit, where the correct operation of the diffuser 2 can be verified.

Figure 19: Energy and pressure lines measured.

Later modifications.

Once these measurements were made, it was decided to refine some parts of the tunnel. First, in order to recover the circumferential kinetic component of the flow at the outlet of the fans in the form of pressure (after discussing several possible solutions), it was decided to design a stator for each fan, with 13 fixed blades with an angle varying between 50 and 18 degrees, from hub to tip (Figure 20).

The energy recovery?(and therefore efficiency) is estimated at?17.8% (157 Pa)?and the?speed increase?in the test chamber?is 4%.

It is also possible to construct diffusers at the outlet of each fan, splitting the transition 1 section into 4 channels, which make the 4 air jets independent and appreciably limit the circumferential component of the impelled flow, significantly reducing energy losses.

This solution produces a recovery of about 207 Pa of pressure (23.5%), which can be translated into a?5.3% increase of velocity?in the test chamber. A final pressure gain could be produced by increasing the frequency of the drive above 50 Hz, without reaching the power limit of the motor (about 225 kW).

Figure 20: Stator designed for each fan.

Since these elements are not critical for the correct operation of the tunnel and can be executed at any time without interfering too much in the operation of the infrastructure, it has been decided to postpone their construction to accelerate the flow characterization campaign.

And secondly, after experimentally verifying that, although the settling chamber is completely filled with flow and the distribution of velocities in the test chamber has a uniformity in average values of more than 95% (in the first measurements made),?there is some dissymmetry in the distribution of pressures and flow rates.

Therefore, it was decided to modify the?straightener?of diffuser 1 (building one with more uniform inlet and outlet sections, since the flow does not detach), shortening the inlet?flaps?located on the curve (to obtain the most uniform flow distribution possible), and arranging three new?deflectors?at the outlet of the straightener 2 (so that the velocity distribution is also uniform throughout the entire section of the settling chamber). Due to the detachment of the flow on the inner faces when entering the diffuser after the bend, these modifications correct the situation significantly (Figure 21).

Figure 21: Deflectors and modifications on flaps and straighteners (diffuser 1, left and diffuser 2, right).

Several?detailed CFD numerical simulations?have also been performed (Figure 22), with several possibilities for these elements, until a compromise between flow quality, head loss minimization and constructive ease has been reached.

Figure 22: CFD Simulations detailing the?curve, diffuser 1 and settling chamber. Velocities, left and static pressure, right.

These modifications have substantially improved the speed distribution, at the cost of introducing a small additional load loss, over that initially foreseen. The maximum speed is also a function of the actual mechanical and aerodynamic performance of the fans. However, the speeds achieved in the test chambers are sufficient for the intended studies. In qualitative terms, the speed in the main chamber is sufficient to move a person, as it is close to free-fall speed (about 180 km/h).

Conclusions.

This first part, WIN TUNNELS: DESIGN, CFD SIMULATION AND CONSTRUCTION, develops the process followed?to design and build?the new EVE50-ablWT Wind Tunnel of the Energy Department of the University of Oviedo.

The?parts?of the infrastructure are listed (Figure 23), briefly explaining their?function.

The?hypotheses?and considerations that had to be made for the correct dimensioning of all the elements are also developed.

Finally, the measurements made during the tunnel start-up tests have confirmed the?goodness of the design, as well as all the decisions taken during the process.

A wind tunnel is a complex equipment, the design of which involves equal parts experience, calculation accuracy and economy. Note that numerical simulations by means of CFD have served only to finalize some details of its operation, but they should never be carried out without the solid theoretical knowledge required and should never be checked against experimental data.?Therefore, science and art go hand in hand in this discipline of wind engineering.

The second part of this monograph, WIND TUNNELS: FLOW CHARACTERIZATION, will address the process of flow calibration and characterization.

-------------------------------------------------------------------------------------------------------------

Currently, the Formula Windy Team is composed of:

Beatriz Bayón García, Olaya Gómez Carril, Daniel Fernández de la Cruz, Aitor Jiménez Fernández, and Professors of Hydraulic Engineering and Fluid Mechanics at University of Oviedo (Department of Energy).

Acknowledgements.

This infrastructure has been possible thanks to the collaboration of several Organizations, Companies and the work of Professors and Students of the University of Oviedo.

The funds come from the Projects that the Department of Energy develops with the companies mentioned below, and from funding calls of the University of Oviedo itself and the Department of Energy; and from the Project FC- GRUPIN- IDI/2018/000205 financed by the Principality of Asturias - Science, Technology and Innovation Plan-, co-financed with FEDER funds.

Among others, the Professors who have provided funding to the Project are Eduardo Blanco, Jorge Parrondo, Bruno Pereiras, Eduardo álvarez, Joaquín Fernández and Antonio Navarro.

Special mention should be made to Iván and Carlos ("Mari?o Construcciones") and all their team, José Antonio Dos Santos (Ogensa) and the students of Civil Engineering and Master of Civil Engineering of the Polytechnic School of Mieres.

References.

[1] MESEGUER, J., SANZ, A., PINDADO & S., FRANCHINI, S., ALONSO RODRIGO, G. (2003). Aerodinámica Civil: Efectos del viento en edificaciones y estructuras. 2a edición. Madrid: Garceta Grupo Editorial.

[2] MESEGUER, J. & SANZ, A. (2011). Aerodinámica básica. 2a edición. Madrid: Garceta Grupo Editorial.

[3] DOBYRN, C.M., ROBERTSON, L.E. & SEE, S.T.??(1982). When a designer should consider having a wind tunnel test conducted to estimate structural loads and Wind Tunnel Modeling for Civil Engineering Applications. T.A. Reinhold (editor). Cambridge University Press (Cambridge, U.K.). pg. 3-7.

[4] LIU, H. (1991). Wind Engineering. Prentice Hall. Englewood Cliffs (New Jersey, U.S.A.).

[5] FRANSOS, D., BRUNO, L. (2010). Edge degree-of sharpness and free-stream turbulence scale effects on the aerodynamics of a bridge deck. Journal of Wind Engineering and Industrial Aerodynamics. pg. 661-671.

[6] DYRBYE, C. & HANSEN, S.O. (1997). Wind Effects on Structures. John Wiley & Sons, Inc. New York (N.Y., U.S.A.).

[7] FARREL, C. & IYENGAR, A.K.S. (1999). Experiments on the wind tunnel simulation of atmospheric boundary layers. Journal of Wind Engineering and Industrial Aerodynamics, 79. pg. 11-35.

[8] BARLOW, J.B., RAE, W.H. & POPE, A. (1999). Low-Speed Wind Tunnel Testing. John Wiley & Sons, Inc. New York (N.Y., U.S.A.).

[9] CERMARK, J.E. (2003). Wind-tunnel development and trends in applications to civil engineering. Journal of Wind Engineering and Industrial Aerodynamics, 83. pg. 225-237.

[10] GORLIN, S.M. & SLEIZINGER, I.I. (1966). Wind Tunnel and Their Instrumentation. Israel Program for Science Translation. Jerusalem (Israel).

[11] IRWIN, H.P.A.H., COOPER, K.R. & GIRARD, R. (1979). Correction of distortion effects caused by tubing systems in measurements of fluctuating pressures. Journal of Wind Engineering and Industrial Aerodynamics, 5. pg. 93-107.

[12] GUMBEY, S.J. (1983 a). Tubing systems for pneumatic averaging of fluctuating pressures. Journal of Wind Engineering and Industrial Aerodynamics, 12. pg. 189-228.

[13] GUMBEY, S.J. (1983 b). A detailed design method for pneumatic tubing systems. Journal of Wind Engineering and Industrial Aerodynamics, 13. pg. 441-452.

[14] IDEL’CIK. I.E. (1969).?Memento des pertes de charge: Coefficients de pertes de charge singulières et de pertes de charge par frottement.?Eyrolles Editeur. Paris.

[15] PRANDTL, L. (1933). Attaching a Steady Stream in Wind Tunnels. NACA TM 726.

[16] METHA, R. D, & BRADSHAW, P. (1979). Design Rules for Small Low-Speed Wind Tunnels.?Aero. Journal (Royal Aeronautical Society), 73. 443 pg.

[17] RODRíGUEZ, M., FERNáNDEZ, J.M., GALDO, M., BLANCO MARIGORTA, E. & SANTOLARIA, C. (2013).?Novel design and experimental validation of a contraction nozzle for aerodynamic measurements in a subsonic wind tunnel. Journal of Wind Engineering and Industrial Aerodynamics, 118. pg 35-43.

[18] MASKELL, E. C. (1963). A theory of the blockage effects on bluff bodies and stalled wings in a closed wind tunnel. R. & M. 3400. November.

[19] LAROSE, G.L. & D’AUTEUIL, A. (2006) On the Reynolds number sensitivity of the aerodynamics of bluff bodies with sharp edges. Journal of Wind Engineering and Industrial Aerodynamics. pg. 365-376.

[20] AHMED, N.A. (2015). Wind Tunnel Designs and Their Diverse Engineering Applications, Edited by N. A. Ahmed, 2nd, ISBN-13: 978-953-51-1047-7.

[21] VALLINA PITA, N. & NAVARRO-MANSO, A. (2018). Dise?o y Construcción de un Nuevo Túnel Aerodinámico para Aplicaciones de Ingeniería Civil y de Capa Límite Terrestre, MEng. Thesis, Máster Universitario en Ingeniería de Caminos, Canales y Puertos, EPM – Universidad de Oviedo, 23-07-2018.

[22] SKREBKOV, G.P. & LOZHKIN, S. N. (1985).?Application of the LeShatelier principle to the calculation of a uniform flow in packed rods channels. Teplofiz. Vys. Temp.., vol. 23, no. 4. pg. 748-753.

[23] NIKURADSE, I. (1936). Laws governing turbulent flow in smooth tubes, in Probl. Turbulentnosti. M. A. Velikanov and N. G. Shveikovsky, Eds. ONTII Press. pg. 75-150.

[24] TEPLOV, A. V. (1955). On the hydraulic calculation of asbestos-cement tubes. Vodosnabzh. Sanit. Tekh, no 1. pg. 13-16.

[25] ALTSHUL, A. D. (1982). Hydraulic Resistance. Nedra Press. Moscow. 224 pg.

[26] NEKRASOV, B. B. (1954). Hydraulics. 290 pg.