Several different kinds of rig are in common use, but the simplest example is the wind tunnel, for passing a uniform, steady stream of air over a stationary model in the "test section". (Water tunnels are similar in principle, but many water-flow facilities have a free liquid surface and are called "water channels": providing that the Froude number is low so that surface deformation can be neglected, fluid-dynamic design is much the same as for wind tunnels but construction and practical details differ.) Most other fluid-flow test rigs such as ducts have at least some of the features of wind tunnels. The number of published descriptions of wind tunnels is unsatisfactorily small. It is inevitable that unsuccessful tunnels are not often publicized, but tunnel design would advance much more rapidly if they were.
Estimates of power consumption are discussed in a separate section.
In aerospace terminology, a "low speed" rig is one whose test Mach number (ratio of flow speed to speed of sound) is small enough for compressibility effects to be generally negligible (say Mach number less than 0.3, speed less than 100 m/s, difference in density between stagnation and static conditions less than 5 percent). A "high speed" rig has a test Mach number near or above unity: "transonic" implies that even if the test Mach number is less than unity, Mach numbers greater than unity appear somewhere in the flow over the model; "supersonic" implies a test Mach number significantly greater than unity; and "hypersonic" implies a test Mach number greater than about 5. Various short-duration facilities like shock tunnels are used for hypersonic flow, but will not be discussed here: see G. Briassulis et al., Exptl. Thermal and Fluid Sci. 13, 430 (1996) for details of a recently-constructed shock tunnel.
Typical Wind Tunnels
The different types of wind tunnel were briefly described in the Introduction. Here are examples and more detailed comments. (The Figures are .gif files: the largest is 55kB.)
Fig. 1 shows the NASA Ames Research Center's 80 ft x 120 ft open-circuit "suckdown" tunnel, with six axial-flow fans at exit (shared, in a unique arrangement, with the 40 ft x 80 ft. tunnel; when the 80 x 120 is in use, air flows from the entry at top right of the figure to the exhaust louvers at bottom left), the "return path" being the Earth's atmosphere. Basically-similar suckdown tunnels, some with diffusers (see Fig. 4) between the test section and the (usually single) fan, are found in all sizes, but most commonly with test sections less than about 2 ft. square. Wind tunnels are usually referred to by the cross-sectional size of the test section (minor x major dimension in the United States, major x minor in many other countries): the 80 x 120 is the world's largest.
Fig. 2 shows a small open-circuit "blower" tunnel with a centrifugal fan (blower) at entry. Unless fitted with a diffuser to reduce drafts, a blower tunnel more than about 2 ft. square is unwelcome in an open laboratory.
Fig. 3 shows another blower tunnel, with the (almost closed!) circuit folded over to fit in an 8 ata. pressure vessel. The tunnel walls only bear the ordinary aerodynamic pressure difference and are lightly built. This is far cheaper than a conventional pressure tunnel with load-bearing walls, but the size is constrained by the largest available, or affordable, pressure vessel. Again, the diffuser is intended to reduce drafts rather than power consumption.
Fig. 4 shows a conventional closed-circuit low-speed tunnel, the most popular arrangement for tunnels with test sections more than about 2 ft. square. This Figure is intended for general reference: the main components are labeled with links to the pages describing them. The diffuser is divided into two sections, one upstream of the fan and one downstream, for reasons described below. A tunnel with this layout and a 80 x 120 ft test section would be almost half a mile long!
Fig. 5a and Fig. 5b shows two NASA pressure tunnels. Their most noticeable feature is the use of 180 deg. "racecourse" bends instead of 90 deg. corners, to reduce bending loads on the pressure-bearing structure. The 10-atm. Langley tunnel has a first diffuser with an unusually large area ratio, over 5:1, and a heat exchanger just before the screens. The first diffuser of the Ames tunnel also has a large area ratio, approx. 4:1: the wide-angle diffuser before the settling chamber has about the largest acceptable area ratio for a diffuser with no internal screens.
Fig. 6 shows the convergent-divergent nozzle and test section of a typical supersonic tunnel. A "blowdown" tunnel would be connected to an upstream pressure vessel, via a conventional settling chamber and contraction, and a "suckdown" tunnel would be connected to a downstream vacuum (strictly, "low pressure") vessel, probably via a subsonic diffuser to reduce total-pressure losses in the jet entering the vacuum vessel. Several types of short-duration subsonic test rigs have been devised to run from pressure vessels. The Ludwieg tube (search for the name in References to High-Speed Tunnels) uses a very long tubular pressure vessel. When a diaphragm upstream of the test section is ruptured, an expansion wave travels "upstream" along the tube. The tube length is chosen so that the wave does not return to the test section, after reflection from the far end of the tube, within the required steady operating time. In the case of a closed-circuit, continuous-running tunnel, most of the circuit is subsonic, and the only basic differences are that most supersonic tunnels need multi-stage fans (compressors) because of the extra total-pressure losses due to shock waves at the end of the supersonic region, and heat exchangers, to remove the heat equivalent of the work done by the compressor.
If the tunnel discharges its flow to the atmosphere (or laboratory, or pressure vessel, or other unprepared environment) it is called an open-circuit tunnel. If the cross-section of the return path is many times that of the test section, it is hoped that the discharged air will lose most of its mean vorticity, unsteadiness and turbulence before it is re-ingested. The example shown in Fig. 2 does not have a diffuser downstream of the test section, but many other open-circuit tunnels do: in the case of suckdown tunnels the fan is usually placed at the end of the diffuser. As the kinetic energy of the air discharged from the diffuser is usually only a small percentage of the kinetic energy of the air in the test section, the power required by an open-circuit tunnel may be less than that required by a closed-circuit tunnel of the same aerodynamic design, because no power is wasted in the drag of the corner vanes.
A popular configuration for small tunnels is the blower type (Fig. 2), with the impeller (usually a centrifugal blower) at entry, and -- usually -- with no exit diffuser because power consumption is not very important. This allows any type of test section to be fitted without problems of matching to the diffuser. Centrifugal blowers are preferred to axial fans mainly because they will run efficiently, and generate acceptably steady flow, over a wider range of load -- i.e. a wider range of test-section configurations.
Open-circuit tunnels which take in air from the atmosphere or the laboratory are sensitive to draughts -- the NASA Ames 80 x 120 (Fig. 1) points into the prevailing northwest wind. Centrifugal blowers seem to attenuate most entry disturbances, but they are very sensitive to fluctuating swirl (axial vorticity) in the inlet, which changes the rate of rotation of the blades relative to the airflow. However, blower tunnels are often fitted with commercial air filter panels covering a large box connected to the blower intake, and the filter extracts airflow irregularities as well as dust. It is more difficult to fit a filter to a suckdown tunnel because the air entering the intake must have roughly uniform total pressure over the cross section or the "turbulence management" devices (screens and honeycombs) will not be able to deliver adequately uniform total pressure to the test section.
Fig. 4 shows a low-speed (100 m/s) closed-circuit wind tunnel of the closed-circuit type, in which the same air is recirculated. The stream is turned, usually in four steps of nominally 90 deg. each, by rows or "cascades" of closely spaced vanes. There is always a small vent, called a "breather", somewhere in the circuit so that the internal pressure does not increase as the air heats up during the run: it is usual to have a slot around the perimeter at the downstream end of the test section, so that the latter is close to atmospheric pressure to reduce the effect of leaks through the holes usually cut in the walls for model support struts, etc.. If the slot mechanically disconnects the test section from the diffuser it may be useful as a vibration isolator. The remainder of the tunnel is above atmospheric pressure (by almost the full test-section dynamic pressure in the case of the settling chamber) and the flow through any leaks is outward. The compensating inflow through the breather is (i) bad for diffuser performance but (ii) easy to detect by releasing smoke just outside the breather.
Sometimes the settling chamber is vented to atmosphere instead, to relieve the structural load on this large section: the test section static pressure is then below atmospheric, with the disadvantage that unless special care is taken air will rush into the tunnel through any holes made for model mountings, cables etc. The British Royal Aerospace Establishment (now Defence Research Agency) 4 ft x 3 ft tunnel Fig. 4) can be run with a vent either in the test section or between the third and fourth corners. In the latter case the observation chamber enclosing the test section is sealed.
Some older tunnels have "open jet" test sections, with a floor but no walls or roof: the idea is attributed to M Eiffel (he of the Tower). This simplifies access to the test model, but the turbulent mixing layer that bounds the flow entrains air from the laboratory, and this air has to be spilled from the intake or "collector" at the start of the diffuser. Collectors almost always require tedious trial-and-error development to avoid unsteady flow, and even oscillations due to acoustic feedback to the flow leaving the contraction. This configuration should be avoided for normal use. A disadvantage not thought of by Eiffel is the difficulty of using laser Doppler velocimetry (LDV/LDA) or particle image velocimetry (PIV/PTV): the seed particles or droplets escape into the laboratory.
Completely sealed closed-circuit tunnels, which can be pressurized or partly evacuated, can be used to economize on power. The power required to drive a tunnel is proportional to the rate of flow of kinetic energy in the test section, which is equal to (1/2)U2, the kinetic energy per unit volume, times the rate of volume flow UA, where A is the cross-sectional area. The Reynolds number of a model in the tunnel is proportional to U times a typicallinear dimension, say A1/2.
If we consider two tunnels of the same size, one using atmospheric air and the other air at n times atmospheric pressure, we see that the power required by the latter to produce a given Reynolds number is n-2 times the power required by the former, since is proportional to absolute pressure. The first pressure tunnel was the NASA Langley Research Center's Variable Density Tunnel, now retired. It has an annular return circuit whose outer wall is the pressure vessel (built by the nearby Newport News Shipbuilding and Dry Dock Co., with the sections beautifully riveted, rather than welded, together: it appears to have been acceptably airtight). Stanford University has a more-or-less conventional low-speed blower tunnel (Fig. 3) fitted inside a standard pressure vessel. Most other pressure tunnels are of conventional closed-circuit design, with the tunnel walls bearing the full pressure difference. High-speed and low-speed examples exist.
"Low density" tunnels are designed to run at low pressures, to simulate conditions at high altitude where the continuum approximation breaks down. Alternatively, a high-speed tunnel can be run at a given Mach number at reduced power by reducing the pressure and thus the density: the Reynolds number is reduced also, but is usually of secondary importance to Mach number.
A relative of the pressure tunnel is the "cryogenic" tunnel, which runs on very cold gas, again to increase the density but also to decrease the viscosity , thus increasing the Reynolds number on both counts. Near the boiling point, gas properties, especially the ratio of specific heats, behave oddly. Pressure increases the boiling point so a high-pressure cryo. tunnel is unlikely to be a good buy.
Modern pressure tunnels and cryo. tunnels usually have sealing doors upstream and downstream of the test section, so that the latter can be returned to room temperature and pressure (for work on the model) without disturbing the gas in the return circuit. Cooling or pressurizing takes a lot of power.
Several small tunnels have been made to run on heavy gases, notably SF6, to reduce the speed of sound (and thus the power consumption at given Mach number) as well as to increase the Reynolds number. Again the disadvantage is that the ratio of specific heats is no longer 1.4 so compressible-flow results may not be valid for air flows.
Most companies or universities concerned with high-speed flows have supersonic tunnels, often of small size. High-pressure or low-density tunnels, cryogenic tunnels and tunnels using heavy gases are more expensive and are mainly found as national facilities.
Transonic and Supersonic Tunnels
There is no fundamental difference between low-speed tunnels and tunnels designed to produce a transonic or supersonic flow, since the flow speed is in the low subsonic range in all parts of a high-speed tunnel except near the test section. Supersonic tunnels have a convergent-divergent nozzle ahead of the test section, in which the flow is accelerated to sonic speed at the "throat" and reaches the required supersonic speed at the end of the diverging portion (actually, within the rhombus shown in Fig. 6).
One-dimensional compressible-flow theory shows that the Mach number of the flow depends only on the ratio of the cross-sectional area at the nozzle exit to that at the throat, and not on the power input to the tunnel as long as the latter is enough to produce sonic speed at the nozzle throat. The power required to bring the tunnel up to supersonic speed is greater than that required for continuous running at low supersonic Mach numbers, because when the flow through the throat is just subsonic it decelerates again downstream of the throat, the boundary layers separate from the diverging walls. Also, as the test-section flow accelerates past the speed of sound a shock wave travels through the test section, again causing boundary-layer separation and large total-pressure losses.
A "blow-down" tunnel (usually a high-speed tunnel) runs intermittently by discharge from a compressed-air reservoir: high-speed "suck-down" tunnels taking air at atmospheric pressure and discharging into a vacuum vessel are rarer, partly for the practical reason that water vapor in humid atmospheric air may condense as the absolute pressure decreases with increasing Mach number. Compressed-air systems usually have silica-gel dryers.
At Mach numbers near unity the required nozzle area ratio is very nearly one, the rate of change of Mach number with respect to area becomes very large, and the use of solid-wall convergent-divergent nozzles become impractical: even if a uniform flow could be obtained, the presence of a model in the test section would form a second throat in which the Mach number would necessarily pass through unity, with gross disturbances to the flow round the model. This phenomenon is known as choking, and for many years prevented tunnel tests being made at all between Mach numbers of about 0.8 and 1.2. The problem was solved by fitting transonic tunnels with a converging nozzle, similar to the contraction of a low-speed tunnel, leading directly into a test section with slotted or perforated walls surrounded by a plenum chamber. Since the Mach number is a unique function of the ratio of total pressure to static pressure and the total pressure is given, any desired Mach number can in principle be obtained by sucking air out of the plenum chamber until the static pressure in the plenum chamber is equal to the desired static pressure in the test section.
First considering the case where the desired Mach number slightly exceeds unity, we see that the pressure at the exit of the convergent nozzle leading to the test section (where the Mach number cannot exceed unity) will be above the plenum chamber pressure, so that air will flow out of the test section and produce the effect of a diverging nozzle, until at a point further downstream the pressure will be equal to the pressure in the plenum chamber and no further outflow will occur. A nominally-uniform flow at the required Mach number will continue to the end of the test section. If a model is now installed, any shock wave generated by it will be reflected from the solid part of the wall as another shock wave and from the apertures in the wall, which form a constant-pressure boundary, as an expansion wave. The open-area ratio of the wall can be chosen so that the two types of reflection cancel out.
Secondly, if the required Mach number is just below unity the flow from the nozzle can be set to that Mach number, as in a low-speed tunnel, and the insertion of a model merely produces a compensating flow through the walls into the plenum chamber, with return flow further downstream (and upstream): an almost constant pressure is maintained on the wall, so that the free-stream Mach number remains almost constant as required.