Hot-wire Array for Turbulence Research in Subgrid-scale Modeling

Modeling of turbulence for Large Eddy Simulations (LES) must be based on sound understanding of the relationships between large and small scales. In order to study flows with a large range of active length-scales, high Reynolds number experimental data can be used. Thermal anemometry provides the possibility of acquiring long records of turbulence data, which are necessary for evaluating statistically converged high-order moments or conditional averages. However, standard hot-wire probes provide, using Taylor’s hypothesis, linear sections through the threedimensional fields only. Thus, the separation between the small and large scales required for the analysis can only be made by one-dimensional filtering. For LES, three or two dimensional filtering is more directly relevant. In this paper, a new hot-wire sensor consisting of an array of 4 sub-miniature X-wires that allows an approximation to two-dimensional filtering of the velocity field is described. To analyze the data, we use Taylor’s hypothesis to perform stream-wise filtering, while cross-stream averaging over the probes provides the transverse filtering. The approximation of a box filter using only a few points in the crossstream direction causes aliasing errors. We present a spectral analysis showing that resulting second-order statistics of filtered velocity gradients and average sub-grid stresses are acceptably small. Initial results from measurements in the turbulent cylinder wake are presented. We compare unfiltered spectra with spectra of 2-D and 1-D filtered velocity components. Also, we present probability density functions of filtered velocity increments over distances r (with ∆ < r < `), emphasizing similarities and differAddress all correspondence to C. Meneveau. Email: [email protected] ences with the pdfs of unfiltered velocity increments. We conclude that such hot-wire arrays can provide experimental data that can be used both to guide subgrid-scale modeling and to validate LES code.