Abstract

INTRODUCTION The modest angle of divergence between the walls of conventional diffusers is an acute embarrasment to the Gas Turbine designer. Several diffusers can be needed in the engine and their accumulative length contributes significantly to the overall engine length and weight. In addition, mechanical complications arise due to the distances imposed by the diffusers between the compressor spools and their driving turbines. It is common knowledge, however, that if the divergence angle in the diffusers is increased, then the flow will separate from the walls giving rise to a pressure loss, instability of flow, and a marked deterioration of any component which may be located immediately downstream. A potential solution to the diffuser problem was reported in (1). This described a diffuser which can achieve greater pressure recoveries than those obtainable from conventional diffusers, and in only one-third of the length. A tubular arrangement of such a diffuser is shown in The rate of extraction of this vented air, or bleed, was reduced to a Copyright © 1980 by ASME The bleed rate quantity coinciding with the top of the step was termed "the minimum bleed requirement, B' ". This is expressed as either a fraction or.as a percentage of the mainstream flow and was found to be strongly dependent on the area ratio of the step A 2 /AI . A correlation based on experimental data taken from small scale models was given as follows: Ihere d, the equivalent diameter at inlet (mm) was f definedas __ 4 (cross sectional area) e vortex controlled perimeter ... This relationship was used as a basis for figure 3 which shows that the minimum bleed requirement increases quite rapidly with area ratio. Assuming that the relationship can be extrapolated up to the larger scale diffusers used in gas turbine engines, then a typical bleed requirement would be of the order of 5%. Such quantities may sometimes be acceptable as, for instance, when drawn from the precombustor diffuser and then used for turbine cooling, the drop in pressure between the two sections providing the necessary motivation. Similar convenient usage of the bleed air, taken from diffusers at other locations in the engine cannot always be envisaged, however, and the discriminate dumping of such quantities of air would significantly increase the engine's fuel consumption. The purpose of the present work was to investigate a means of reducing the quantity of bleed-off with the object of making the vortex controlled diffuser applicable to most locations in the gas turbine engine and installation. In order to fully exploit the vortex controlled diffuser it is firstly necessary to consider the mechanism by which it operates. A completely satisfactory explanation has not yet been devised but a useful hypothesis was proposed in reference I This used the model sketched here as figure 4. Due to the application of external suction, the static pressure inside the vortex chamber is below that in the mainstream. As a result, the stream 'a' which is being drawn into the vortex, experiences a considerable acceleration. On the other hand, the stream 'b', which flows down the diffuser is flowing to a region of greater static pressure and will therefore decelerate. A shearing action, produced by the velocity differential between the streams, then results in the creation of an extremely turbulent layer. This has been confirmed by Sutherland, (2), who measured turbulence intensities in the order of 50% near the aperture above the vortex chamber and also in a layer flowing beyond the fence and which eventually reattaches to the downstream wall. An indication that this turbulent layer can inhibit flow separation from a wide-angled conventional diffuser, located immediately behind the vortex diffuser, was first given in reference (3). Here, a description is given of a highly efficient but relatively short length diffuser, located in the Rolls-Royce compressor test facility at Bristol, U.K. The concept of hybrid diffuser being evaluated in the present study is illustrated in The turbulent layer generated by the step is then used to inhibit flow separation from a relatively wide angled conventional diffuser which contains a much greater part of the overall increase in area. Of particular interest was the level of anticipated compromise that would be involved between pressure recovery, overall length and bleed-off. Scope of Experiments Due to their exploratory nature, the tests were confined to the tubular/conical configuration shown in figure 5. They were also limited to one inlet condition with a nominal inlet Mach number of 0.25, chosen so that the flow could be treated as incompressible and yet be at a sufficiently high Reynolds number as to be considered insensitive to any small variation in it. The Reynolds number based on inlet diameter was 8.4 x 10 5 while that based on the approach distance between the contoured contraction piece and the vortex aperture was 14.5 x 10 5 . This latter value ensured a turbulent boundary layer condition in the critical region near the vortex. The diffuser inlet diameter was fixed at 152.4 mm but the other parameters illustrated in D, L GEOMETRIC PARAMETERS FIGURE 6 Three sets of diffusing cones were manufactured from sheet metal, each covering a range of included divergence angle, 6, from 10 up to 30 degrees. Details of these cones, together with their associated total diffusing length ratio, L/D 1 are given in table 1. In addition, a simple parallel pipe was provided which could revert the diffuser back to a wholly vortex controlled arrangement similar to that shown in figure 1. This gave an overall area ratio, A3 /Al , of 2.0:1 with a length ratio, L/D I of 1.025:1. figure 7, air was supplied from a contoured transition piece having a contraction ratio of 4:1. Variation of the fence axial gap 'x' was obtained by relaxing the nip on the '0' ring seal and sliding the entire vortex chamber assembly along the major axis. The concentricity of the fence and follow-on conical diffuser was guaranteed by using fitted locating pins which penetrated into the vortex chamber assembly In turn, the accurate concentricity of this assembly was controlled by the machined discs which could slide along the outer surface of the inlet duct. Numerous small holes were drilled uniformly around the machined disc located inside the assembly. These were to ensure that the bleed-off was extracted evenly from around the circumference of the vortex. Sharp edged orifice plates, installed according to British Standards 1042 were used to measure the airflow rate approaching the diffuser and also that flowing in the bleed-off duct. These measurements were made to an estimated accuracy of t 12%. Performance of the diffuser was assessed by the rise in static pressure between diffuser inlet and exit. The inlet pressure was measured using three static wall tappings, located at a distance of one duct diameter upstream of the vortex chamber aperture. This location was chosen so as to avoid the strong pressure gradients to be found nearer the commencement of diffusion. The pressure at diffuser exit was simply assumed equal to the test cell ambient pressure. Further static pressure measurements were taken from three tappings, located in the wall of the vortex chamber. Readings from these were used to assess the pressure loss experienced by the bleed air. The mean static pressures, both at diffuser inlet and in the vortex chamber, were estimated to have an error of less than t 0.25% during steady flow conditions. Performance Parameters These were specifically chosen to permit a direct comparison with the more conventional type of diffuser. Diffuser length is non-dimensionalised by dividing by inlet duct diameter, and static pressure recovery is presented as a static pressure coefficient, C p defined as: The quantity of bleed-off, B, like the minimum bleed requirement, B', is expressed as a percentage of the flow rate at diffuser inlet. In many applications when a bled diffuser is used it is necessary to take full account of the losses entailed by bleeding air off. Unfortunately there is no universal term which can be used to satisfactorily provide this assessment. It depends on whether the bled air is to be simply dumped overboard or whether it is to be re-introduced back into the flow stream. In this latter case, consideration must be made as to where the air is to be reintroduced, whether at diffuser inlet or exit, and to any losses that may be involved during the pumping process.. Furthermore, consideration should also be given to the effects that poor diffusion might have on any component, such as a heat exchanger, combustor or compressor that may be located just downstream. To enable assessment to be aimed at a particular application the pressure loss experienced by the bled air is given by the coefficient of vortex chamber depression, defined as: Test Procedure The supply air flow was adjusted to give a nominal Mach number of 0.25 at diffuser inlet. This was obtained before the introduction of suction to the vortex chamber; when suction was applied, however, there was a slight increase in Mach number produced by the improvement to diffusion but no attempts were made to readjust this. Static pressure measurements were recorded at the zero bleed condition and at various bleed rates of up to about 5% of the total flow. From these the values of C were calculated and then plotted against percentage bleed to give control curves similar to that shown as figure 8. Most of the control curves derived from the tests can be obtained from reference (4). Interpolated values of C p , corresponding to bleed rates of 0%, 1%, 2% and 3% were transferred from the control curves to the results tables given here in the appendix. Due to the limited movement of the vortex chamber assembly however, the smaller angles could not be achieved with the two larger fences. When this situation arose then the smallest angle obtainable was substituted. Details of the angles used are given in table 2 (appendix), together with an identity code which is also applicable to the tables giving the diffuser performance. RESULTS AND DISCUSSION (a) Wholly Vortex Controlled Diffuser Tests were conducted on this arrangement to provide a comparison with the earlier work, reported in [1), and to establish a datum for assessing the hybrid diffuser arrangements. The curve presented in The value of C coinciding with the top of the step is subjective, but is clearly in the order of 0.63. The ideal value of C for a diffuser of this area ratio, given by C p= 1-(AR) -2 , is 0.75 and so the pideal diffuser has achieved an effectiveness of 84%. Inserting this quantity into equation It is therefore reasonable to accept that there is a good agreement between present work and the earlier series of tests reported in reference (1). Additional data derived from using all fourteen of the available fence configurations are presented in table 3. A noteworthy feature shown by the table is that at a bleed rate of 3% the highest values of C were obtained by using both the widest available fence radial gap, y = 5.46 mm, and the largest fence A direct comparison between figures 9 and 10 clearly indicates that the arrangement using the smallest vortex controlled step (A2/A 1 = 1.2:1), produces the highest pressure recovery. Both hybrid arrangements, however, outperform the wholly vortex controlled diffuser when constrained to the same length (L/D1 = 1.025) and rate of bleed-off. For instance, at 3% bleed-off the hybrid diffuser produced a C p of 0.75 compared to a corresponding value of 0.65 from the vortex controlled diffuser, and at lower bleed rates the improvement is even more evident. • FIGURE 10 Alternatively, the benefits to be obtained by the new diffuser can take the form either of a reduction in bleed rate or a reduction in length, or some compromise between the two. This statement can be substantiated by making reference to figure 9. This shows that in order to produce a C p of 0.65, within a non-dimensional length of 1.025, it is only necessary to extract a bleed flow of 1% compared to the corresponding quantity of 3% required by the wholly vortex controlled arrangement. Furthermore, if the length of the hybrid diffuser is halved (ie. L/D1 = 0.5), then at a bleed-off of only 2% it still produces a higher pressure recovery than the standard vortex controlled diffuser did when operating with 3% bleed-off. Probably the most important discovery to be made during these tests was that the hybrid diffuser can produce a high level of pressure recovery even without the application of bleed. This important fact is most pronounced in Hybrid Diffusers of Overall Aera Ratio 2.5:1. As mentioned previously these were only tested with a vortex controlled step of area ratio 1.2:1. The data, as extracted from control curves, are presented in table 6, and summarised in figure 11. A comparison with figure 9 indicates that the performance at this larger area ratio only generally exceeded that of the smaller hybrid diffuser for non-dimensional lengths of greater than 2.0. However, a superior performance could still be obtained below lengths of 2.0, but only at the higher rates of bleed-off. For instance, at a non-dimensional length of 1.0 the larger arrangement gave greater pressure recoveries once the bleed rates had exceeded 2%. Another interesting comparison is that the new arrangement produced a C of nearly 0.8 within a non-dimensional length of 1.0, albeit at a bleed rate of 3%; a similar pressure recovery from a diffusing cone would have required a non-dimensional length of 9.5 and an area ratio of 4.2:1 (ref At zero bleed-off, the new arrangement again outperformed the equivalent conventional conical diffuser (AR = 2.5:1), but only after their nondimensional lengths had exceeded 1.0. For lengths of greater than 1.5, the improvement in pressure recovery again was particularly significant. Comparison Between Zero Bleed Hybrid Diffusers and Optimum Conical Diffusers. In most practical applications, and in the gas turbine in particular, the major constraint on diffusion is that imposed by length. Accordingly a diffuser is required with an area ratio which is optimised to produce the maximum pressure recovery within the specified length, such a device is referred to in the literature as a C p* diffuser. Reference 5 was once again the source of standard data for the values of C p * and corresponding area ratios that are used in Pressure loss experienced by the bleed air The value of vortex chamber depression, Vc , was found to be sensibly independent of geometric features downstream of the fence such as the cone angle, 0, the overall area ratio, and the size of the vortex controlled step. In general the value of V c was found to increase with bleed flow rate and with increases in the radial fence gap, y, and the subtended angle, 0. Values of V c , covering the whole range of bled diffuser experiments are presented in On breakdown of the vortex, the bleed flow will originate from down-stream of the fence and will still produce a significant region of low pressure as it accelerates over the tip of the fence see /TURBULENT SHEATH VORTEX CONTROLLED DIFFUSER OPERATING AT LOW BLEED FIGURE 14 At zero bleed, it is likely that the above mechanism is maintained because of a 'jet pump action', produced by the mainstream flow in passing over the vortex chamber aperture, as sketched in HYBRID DIFFUSER (ZERO BLEED) i1 With suction, it can match the pressure recovery of the extremely efficient vortex controlled diffuser when constrained to the same length. The hybrid diffuser, however was found to require only one-third the rate of bleed-off and so is more suitable for a wide range of applications. If a comparable rate of bleed-off is employed, then the new diffuser offers considerably higher pressure recoveries which are obtainable within a remarkably short length; for instance a value of C of near 0.80 was obtained within a non-dimensional P length, L/D 1 of 1.10 at a bleed-off rate of 3% of the mainstream flow. Without suction being applied the diffuser is less efficient than the vortex controlled diffuser operating with suction. Even so, it produced pressure recoveries which were 25% higher than those obtainable from the optimum 5. E.S.D.U. design of conventional divergent duct (C* diffusers) P having the same length. Alternatively it can match the pressure recovery of the conventional diffuser but then only requires half of the diffusing length. More data is required in order to optimise the design of the hybrid diffuser, to determine its performance over a wide range of operating conditions, and to establish a design proceedure. APPENDIX ACKNOWLEDGEMENTS