| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
||||
|---|---|---|---|---|
|
Most of the power required to run a closed-circuit tunnel is absorbed by losses in the diffuser, the test section and the corners. Axial fan efficiency can exceed 90 percent if the wall boundary layers at entry are not too thick. The contraction ratio is usually chosen so that total-pressure losses in the honeycomb and screens are acceptably small, and this usually means that losses in the third and fourth corners are small also. All losses are basically due to turbulent skin friction and therefore slowly decrease, in proportion, as the tunnel size (Reynolds number) increases.
The dimensionless parameter representing consumption of power delivered to the fan drive shaft, P, is the "power factor"
where U and A are the test-section velocity and cross-sectional area and the denominator is the rate of flow of kinetic energy, (1/2)U2 per unit mass, through the test section (i.e. K.E. per unit mass times mass flow rate A poor estimate of power factor obviously leads either to deficiency in maximum speed (the electric drive motor reaches maximum available voltage or current before reaching maximum speed) or to incomplete use of the available motor power (the converse). With a commutator motor it is simply necessary to alter the field excitation, and if the fan or blower is belt-driven (usual with commercial centrifugal blowers) the pulley ratio can be changed, in either case to ensure that the maximum power is extracted from the motor. If the fan or blower is so overloaded that its efficiency is low (more likely with an axial fan, due to blade stall) there is no simple cure.
The loss in the test section due to boundary-layer growth is easy to estimate from formulas for total drag of "flat plates" such as Cf = 0.074(UL/ |
Charts for estimating diffuser efficiency are given by Sovran and Klomp "Experimentally determined optimum geometries for rectilinear diffusers with rectangular, conical or annular cross-section", in "Fluid Mechanics of Internal Flow" (G. Sovran, ed.) Elsevier (1967), as the ratio of actual pressure rise to that estimated from one-dimensional inviscid theory for the given area ratio. These charts have been reproduced in several later textbooks. For a modern analytical treatment see Johnston "Review - Diffuser design and performance analysis by a unified integral methd", ASME J. Fluids Engg 120, 6 (1998). Diffuser losses are the biggest unknown in power-factor estimation: an underestimate may lead to overloading the fan or blower, with consequent low efficiency.
The loss in a 90 deg. corner with thin vanes corresponds to a pressure-loss coefficient between 0.12 (high Reynolds number) and 0.15 (low Reynolds number).
Centrifugal blower efficiency can be obtained from manufacturers' performance charts. Axial fan efficiency near the design point can be taken as 90 percent.
Pressure-loss coefficients for honeycombs are usually negligible compared to those for screens. For the latter, Wieghardt's data fit (see Screens and Honeycombs) can be used. If tests of screen or honeycomb samples are done in a small duct, it is essential to put a screen a short distance upstream of the sample to reduce the boundary-layer thickness: otherwise the reduction of displacement thickness as the boundary layer flows through the sample will decelerate the flow and lead to an underestimate of the pressure-loss coefficient.
Losses due to the boundary layers on the walls of the settling chamber and contraction are negligible compared to those in the test section.
 |
|||
= P/[(1/2)
U3A]
)-1/5 where Cf is the streamwise-average skin-friction coefficient and L is the test-section length. Since the test-section surfaces usually bear some excrescences or cavities, leading to extra drag, and since the boundary layer entering the test section from the contraction has a small but finite thickness, the calculated figure should be "rounded up". The losses due to the maximum likely drag of a model mounted in the test section require knowledge of the test program throughout the life of the tunnel, which is not normally available - just make a generous guess. (If a model is so large that its wake disturbs the flow in the diffuser significantly, it is probably too large for the tunnel: however in at least two recent experiments "oversize" models were used, in the expectation that any computations of the flow would use the tunnel walls as the boundaries of the domain of integration.