Horseshoe vortex formation around cylinders attached to an end-wall

摘要

Experiments have been performed to study the behaviour of the end wall boundary layers for anumber of wall-mounted cylinders. Surface flow visualisation using oil smear has been combinedwith a detailed measurement programme using two-component PIV.One aim of the investigation is to examine the possibility that the secondary flow losses generated bythe short support struts that are a common feature in turbo-machines can be minimised by carefulchoice of the cross-section of the cylinder. This idea was originally suggested by a theoretical analysisHawthorne (1951) and was studied experimentally by Williams (1970). The intention is to re-examinethe problem using the more sophisticated measurement techniques now available. A further objectiveof the investigation is to produce a data bank of high quality experimental data for the validation ofcomputational flow models.Experiments have been performed in two different wind tunnels. The first of these has a rectangularcross-section which is 4.68 wide × 1.56 m high and offers a maximum velocity of around 20 m/s.Circular cylinders of diameter 50, 100, 300 mm were fixed across the 1.56 m height of the workingsection, enabling Reynolds number values in the range 8.2×10 4 to 4.9×10 5 to be examined. The flowpatterns around the bottom end of the cylinder were then visualised using the oil streak method. Aftersome refinement, this method has been found to yield extremely good information on the time-averagedsurface flow patterns. In the second wind tunnel, with a rectangular cross-section 0.53 mhigh times 0.46 m wide and offering a maximum velocity of 10 ms -1 , cylinders with a variety ofdifferent cross-sectional shapes (circular, aerofoil and lenticular) were located across the height of theworking section. Two-component PIV was then applied to study the flow behaviour, particularemphasis being placed upon the region of the flow upstream of the cylinder and close to the end-wall.The third velocity component was then calculated by applying continuity. However, noting theinaccuracies present when two-component PIV is used in vortical flows, a limited series ofmeasurements were also obtained using a three-component PIV system. The two sets of data will becompared.