This Thesis has endorsed employing a novel indraft configuration for a severely spatially and financially constrained wind tunnel aimed at undergraduate and postgraduate aeronautical and automotive instruction.
The novel horseshoe indraft configuration employed may be considered to either bend a traditional open circuit or remove corners 3 and 4 from a traditional closed circuit. By connecting the inlet and exit to atmosphere the new configuration prevents pressure loading of the surrounding building; eliminates the problem of exhausting a jet within a laboratory; and eliminates costs associated with a heat exchanger. The modest budget (£350,000) is commensurate with the financial means of a University or small enterprise.
Aerodynamic performance data suggests future designers should not shy away from an indraft tunnel by default: Velocity uniformity in the working area of jet has been shown to vary by less than 0.3% of the mean in the presence of ambient gusts up to 11.5% of the test velocity. Lift and drag coefficients derived from a 27% scale Davis automotive model (5.9% frontal area blockage) repeated to 6 units (0.6%) and 2 units (0.2%) respectively in the presence of ambient gusts up to 13% of the test velocity. Axial turbulence intensity was measured to be in the region of 0.15% (negligible ambient gusts) and 0.35% (ambient gusts up to 16% of the test velocity). This data compares favourably to that for the significantly larger NASA Ames 80ft x 120ft open circuit wind tunnel. Maximum test section velocity has been shown to be in excess of the desired 40m/s. The test section boundary layer closely follows the profile for a 1/7th power law turbulent boundary layer, which suggests the contraction is free from separation.
This Thesis contributes to the body of knowledge by publishing performance data for a new type of wind tunnel configuration. It also augments existing design guidelines and rules of thumb by providing a complete reference point (including design flowcharts) for the design of comparable low speed wind tunnels. The Thesis offers the following specific conclusions and implications:
Screens: Whilst the inlet filer mesh is effective at damping ambient gusts it suffers the worst correlation to the governing equations (significant under prediction of loss), likely due to wire-wake coalescence. This highlights the importance of performing pipe rig tests for screens with open areas significantly less than 57%. Safety screen loss was under-predicted (assumed drag coefficient, CD of 1.0 due to treatment as isolated wires). Whilst measurements suggest a CD of ~1.25 designers are advised to conduct pipe rig tests.
Contraction: To allow pressure gradients to decay prior to the working section, it is advised that the parallel duct at end of the contraction be 1 hydraulic diameter rather than the 1 hydraulic radius proposed by the major texts.
Working section: To allow for model wake recovery (and hence reduce the effect of non-uniformity on the downstream diffuser), a working section length-to-diameter ratio of 2.5 is suggested rather than 2 proposed by the established texts. Additionally, the static ports of tunnel pitot-static should be at least 0.55 hydraulic diameters upstream of the model leading edge to position them away from the static pressure signature of the model.
Diffusers: Whilst the safety screen would ideally have to be removed to prove the hypothesis - it is suggested that turbulent mixing aft of the safety screen (located at the end of the working section) appears to offer a ~10% Cpr improvement to the first diffuser.
Corner cascades: Whilst the established texts focus on corner loss coefficient (KL) this Thesis has shown that KL should not be the sole metric used to select the space-to-chord ratio (s/c) of corner cascades. Uniformity far downstream of a test cascade has been shown to improve with more closely spaced vanes (s/c of 0.190 rather than 0.237) despite KL being similar. Improvements to inlet boundary layer quality have also been shown to reduce KL.
Fan: The fan static pressure rise was measured to be less than predicted due to smaller than expected leakage losses. A leakage loss of 2.5% is therefore proposed rather than the 10% suggested by the major texts.
A Doctoral Thesis. Submitted in partial fulfillment of the requirements for the award of Doctor of Philosophy of Loughborough University.