Blade Blocking Effects On Large Scale Isotropic Grid Turbulence



INTRODUCTION

This research project is concerned with the prediction of turbulence quantities for large-scale turbulence passing through a blade row. Such predictions are needed to enable accurate computations of the noise and vibration produced in, for example, shrouded submarine propulsors. Figure 1 shows schematically the flow through a marine propulsion pump. Large scale turbulence emanating from the vehicle's boundary layer encounters a row of blades, such as a rotor or stator, as it enters the propulsor system. The turbulence then flows past the blades, which distort it and add their own wake turbulence into the flow field. This combined flow then impinges on a second set of blades (e.g. the propeller). The interaction of the turbulence with both sets of blades will generate vibration and noise. To predict this noise it is necessary to have an estimate of the two-point space-time correlation of the turbulence as it enters the second (noise-producing) blade row.


Figure 1. Schematic of flow through a shrouded marine propulsor


Since the evolution of the two-point correlations through the blade row is required, RANS solutions for this part of the flow are of limited use. While such correlations could be obtained from LES of DNS, most realistic configurations are far too complex and of too high a Reynolds number to be modeled in this way. A viable, perhaps the only viable, approach to this problem is to use rapid distortion theory (RDT). Rapid distortion theory, in which the linearized equations of motion for the convection of the turbulence are solved, can, in principle, provide predictions of the alteration to the two-point correlation function in the first blade row.

Unfortunately, the ability of RDT to make such predictions has never been properly established. Previous studies of turbulence/flat plate interactions suggest that RDT could provide accurate predictions (at least of Reynolds stresses) for at least 1 integral lengthscale downstream of the leading edge and perhaps further (though evidence here is mixed). This appears sufficient in the current application (in which turbulence lengthscales on the order of the chord are envisaged). However, it is unknown whether this conclusion extends to space-time correlations, or to cascades where surface blocking effects by adjacent blades may combine, be staggered and will be coupled with curvature and acceleration or deceleration of the flow. At the same time, fairly sophisticated computational schemes based on RDT have been developed to predict turbulence interaction with cascades. However, these schemes are all designed to predict unsteady loading, pressures and acoustic radiation from a blade row and have not yet been applied to the prediction of turbulent space-time correlations.

In order to validate the use of RDT for this application we are performing experiments which are being compared to the calculations of Dr. Glegg at Florida Atlantic University.


EXPERIMENTAL APPROACH

In order to create the large-scale turbulence required for this application we decided to build a Makita-type active turbulence grid (figure 2) for the Virginia Tech Stability Wind Tunnel. The test section of this facility is particularly well suited to the generation of large-scale turbulence. The section width and height of 1.83m are sufficient to accommodate very large integral scales without compromising homogeneity. The section length, plus a short length of contraction to improve isotropy, allows over 9m of distance over which grid turbulence could be allowed to develop.

The active grid consists of rectangular array diamond-shaped vanes attached to horizontal and vertical rows of rotating bars. The rotation of each bar is controlled using a separate stepper motor. The stepper motors operate independently and, under computer control, change direction randomly. Using this grid we were able to generate homogeneous isotropic turbulence with integral scales between 8" and 23" with Taylor Reynolds numbers ranging from 400 to 1250. This is, to the best of our knowledge, some of the highest homogeneous isotropic turbulent flows ever generated in a wind tunnel

The cascade configurations, shown in figure 3, were constructed from flat 0.25"-thick aluminum plates. Circular nose sections at the plate leading edges were used to minimize any instantaneous leading edge separation associated with turbulent upwash. Sharp trailing edge sections were added to eliminate any vortex shedding. These idealized blade sections are not unrealistic models of the thin blades normally used in marine applications, and their geometric simplicity will simplify the tasks of interpreting experimental results and performing RDT predictions. The blades span the full 1.83m height of the test section and are placed under tension to prevent vibration and to maintain an accurate blade separation. The large aspect ratio of the blades ensures a two-dimensional flow, and thus greatly simplify the tasks of measurement, analysis and interpretation. Two configurations are being investigated- one with the blades in an unstaggered cascade as well as a 35 degree stagger configuration. The blade spacing is 10.4" which is on the same order as the chord (12.9") as well as the oncoming integral scale (11").

Figure 2. Active Turbulence Grid Figure 3. Cascade in unstaggered configuration.


Three-component velocity measurements were made in front, throughout and behind the two cascade configurations using subminature 4-sensor hot-wire probes, utilizing a direct angle calibration scheme. Probes were operated using Dantec 56C17 anemometers, optimized to give a matched frequency response greater than 25 kHz. The hot-wire system provides continuous signals of the velocity vector at the probe location. Measurements were made by sampling these signals at 25.6 kHz over one hundred 2 second long records.


INITIAL RESULTS

Initial Results of turbulence stresses and power spectra can be found in the following conference papers:

Last Updated: 07/31/2004