Spatial averaging of streamwise and spanwise velocity measurements in wall-bounded turbulence using ∨- and ×-probes

Abstract

The effect of finite dimensions of ∨- and ×-probes is investigated for the measurement of mean and variances of streamwise and spanwise velocities in wall-turbulence. The probes are numerically simulated using a Direct Numerical Simulation database of channel flow at a friction Reynolds number (Reτ) of 934 by varying the probe parameters, namely, the wire-lengths (l), the angle between the wires (θ) and the spacing between the wires (Δs). A single inclined wire is first studied to isolate the effect of l and θ. Analytical expressions for the variances of the streamwise and spanwise velocities are derived by applying a linear-box-type filter to the unfiltered velocity field for both ∨- and ×-probes (at θ = 45°, and arbitrary l and Δs). A similar expression for the streamwise variance in the case of a single inclined wire (for arbitrary l and θ) is also derived. These analytical expressions, supplemented with a model for the correlation over the wire-length, compare favourably with the numerical simulation results, and more importantly explain various trends that are observed in the variances with varying parameters. Close to the wall (where the errors are generally higher) the errors in spanwise variances of the ×-probes are much lower than the ∨-probes, owing to an ‘error-cancelling’ mechanism present in ×-probes due to the effect of l and Δs, as well as due to the procedure of recovering the velocities from two wires. The errors in the streamwise variances are comparable for both ∨- and ×-probes. On the other hand, mean velocities are measured with almost no error by the ∨-probe, whereas the ×-probe induces finite errors in mean velocities due to the fact that the two wires experience different mean velocities in ×-probes unlike ∨-probes. These results are explained using the corresponding analytical results, which also show that under the effect of a linear filter, measured variances depend only on the fluctuating velocities (more precisely on the two-point correlations) and the measured means depend only on the unfiltered mean. The various results are found to be in accordance with the experimental measurements carried out in a turbulent boundary layer at Reτ ≈ 5000. Finally, considering the physical positioning of wires in the ∨- and ×-probes combined with the above results suggests that ×-probes might be more suitable for stream–spanwise velocity measurements than the ∨-probes in conventional wind tunnels.