Flow Measurements in a Fishtail Diffuser With Strong Curvature

Abstract

The paper presents detailed measurements of the incompressible flow development in a large-scale 90 deg curved diffuser with strong curvature and significant streamwise variation in cross-sectional aspect ratio. The flow path approximates the so-called fishtail diffuser utilized on small gas turbine engines for the transition between the centrifugal impeller and the combustion chamber. Two variations of the inlet flow, differing in boundary layer thickness and turbulence intensity, are considered. Measurements consist of three components of velocity, static pressure and total pressure distributions at several cross-sectional planes throughout the diffusing bend. The development and mutual interaction of multiple pairs of streamwise vortices, redistribution of the streamwise flow under the influence of these vortices, the resultant streamwise variations in mass-averaged total-pressure and static pressure, and the effect of inlet conditions on these aspects of the flow are examined. The strengths of the vortical structures are found to be sensitive to the inlet flow conditions, with the inlet flow comprising a thinner boundary layer and lower turbulence intensity yielding stronger secondary flows. For both inlet conditions a pair of streamwise vortices develop rapidly within the bend, reaching their peak strength at about 30 deg into the bend. The development of a second pair of vortices commences downstream of this location and continues for the remainder of the bend. Little evidence of the first vortex pair remains at the exit of the diffusing bend. The mass-averaged total pressure loss is found to be insensitive to the range of inlet-flow variations considered herein. However, the rate of generation of this loss along the length of the diffusing bend differs between the two test cases. For the case with the thinner inlet boundary layer, stronger secondary flows result in larger distortion of the streamwise velocity field. Consequently, the static pressure recovery is somewhat lower for this test case. The difference between the streamwise distributions of measured and ideal static pressure is found to be primarily due to total pressure loss in the case of the thick inlet boundary layer. For the thin inlet boundary layer case, however, total pressure loss and flow distortion are observed to influence the pressure recovery by comparable amounts.