COMPARISON OF TWO METHODS TO INCREASE TIP CLEARANCE AND ITS EFFECT ON PERFORMANCE OF TURBOCHARGER CENTRIFUGAL COMPRESSOR STAGE .doc

CHINESE JOURNAL OF MECHANICAL ENGINEERING* 60 Vol. 20, No. 4, 2007SYED Noman DanishCOMPARISON OF TWO METHODS TO INCREASE TIP CLEARANCE AND ITS EFFECT ON PERFORMANCE OF TURBOCHARGER CENTRIFUGAL COMPRESSOR STAGEMA ChaochenYANG Ce!, o WeiSchool of Mechanical anri Vehicu!?/LiigineerirtyBeijing Instate of Technology,Beijing 100081, ChinaAbstract: Tip clearance between th; b'le tip and casing of a centrifugal compressor can be varied through two methods: by changing the hijde height (MP or by cnanging the casing diameter (M2). Numerical simulations are carried out to c><v.p&ie Uiese two methods and their effect on the stage and impeller performance. The i.T,peiier and diffuser are connected through rotor stator boundary using frozen rotor approach. Overall stage performance and the flow configuration have been investigated for nine tip clearance levels from no gap to 1 mm. Impeller and difruser performances are also presented separately. It has been found that the overall and impeller performance are comparatively better for Ml below tip clearance of 0.5 mm whereas M2 is found advantageous above 0.5 mm of tip clearance. Both Ml and M2 show performance degradation with the increase in tip clearance. Two models have been proposed for the stage total pressure ratio and efficiency, which are found to be in agreement with experimental results. The impeller efficiency and the pressure ratio are found to be maximum at tip clearance of 0.1 mm for both the cases however minimum diffuser effectiveness is also observed at the same clearance level. Difruser effectiveness is found to be maximum at zero gap for both cases. As it is practically impossible to have zero gap for unshrouded impellers so it is concluded that the optimum thickness is 0.5 mm onwards for Ml and 0.5 mm for M2 in terms of difruser effectiveness. Mass averaged flow parameters, entropy, blade loading diagram and relative pressure fields are presented, showing the loss production within the impeller passage with tip clearance. Key words: Centrifugal impeller Diffuser Stage Tip clearance Numerical simulation Entropy0 INTRODUCTIONThe flow structures within the centrifugal compressors are considered amongst the most complicated and convoluted in all turbomachinery. In recent past paramount advancement in the performance of the centrifugal compressor has been made primarily because of the computer aided design and analysis techniques that are cautious combination of empirical correlations and extensive modeling of the flow physics.Unshrouded centrifugal compressors are mostly favored compared to shrouded compressors in order to avoid high stresses involved with the increased weight. As a result the leakage flow through the tip clearance between the blades and casing is inevitable that further complicates the flow and may depreciate the overall performance of the centrifugal compressor. There are two unique and evenly significant aspects of the tip clearance flows as suggested by DENTON, et alm. First, the reduction in the blade force and the second foremost aspect is the mixing of flow through the tip clearance gap with the flow between blades. The interaction of tip clearance flow field, blade vortex flow and leakage vortex flow generates an extremely complex flow structure.In recent past, a number of numerical and experimental investigations have been conducted'2"6' to investigate the effect of tip clearance in unshrouded compressors. DANISH, et al2), numerically investigated the effect of tip clearance on the performance and flow characteristics of a centrifugal impeller. Entropy fields and the secondary flow development were presented showing the loss production within the impeller passage. No optimum clearance was found for all simulated results except no gap level. USHA, et alt3', numerically predicted that the performance was degraded with the increase in tip clearance. GAO, et al'41, usedReceived June 22, 2006; received in revised form March 28, 2007; accepted April 12,2007their own computation fluid dynamics (CFD) code to investigate the effect of tip clearance on 3D viscous flow field and performance of NASA LSCC impeller with a vaneless diffuser. The study indicated that the location of the throughflow wake was influenced by the tip clearance and there probably exist an optimal clearance at which flow loss was minimum. Their simulations indicated that the optimum clearance was about 0.9 percent of the blade height. EUM, et al5, numerically studied six clearance levels. The effect was decomposed into inviscid and viscous components using one-dimensional model expressed in terms of the specific work reduction and the additional entropy generation. Both inviscid and viscous effects affected performance to similar extent, while efficiency drop was mainly influenced by viscous loss of the tip leakage flow. Performance reduction and efficiency drop due to tip clearance was proportional to the ratio of tip clearance to blade height. The proposed 1D model was found to be in close agreement with the experimental results. HONG, et al61, experimentally measured the discharge flow of a centrifugal compressor at six levels of tip clearance. The study found an optimum tip clearance ratio of 0.12 in terms of surge margin, however the overall performance degradation was found with the increase in tip clearance. They also concluded that the wake region was increased with tip clearance and the deficit of the relative total pressure governed the wake region therefore the loss was magnified. The slope of linear correlation between the impeller efficiency and the tip clearance was - 0.37. PAMPREEN?1 did an extensive experimental study for finding the effect of tip clearance on six different compressors and he observed that the average slope of the linear correlation between the fractional change in efficiency versus the tip clearance ratio was - 0.3.In order to study the effect of tip clearance, there are two methods to increase the tip clearance. In first method (Ml), the distance between hub and shroud is kept constant while blade height is reduced. In second method (M2), the blade height is kept constant but the shroud radius is increased. The present study uses both the methods to increase the tip clearance. The impeller andCHINESE JOURNAL OF MECHANICAL ENGINEERING61 diffuser are evaluated for nine tip clearance levels (0 mm, 0.1 mm, 0.2 mm, 0.3 mm, 0.4 mm, 0.5 mm, 0.6 mm, 0.8 mm and 1 mm) at various mass flow rates ranging between stall to choke conditions for both the methods. The simulations were executed and the results were predicted keeping in mind the existing impeller theories for secondary flow transport, jet-wake flow and internal diffusion. This paper is ordered as follows. First of all the description of studied compressor stage is provided. Computation method is then discussed in section 3 with a brief description of CFD software package. Results of CFD are then presented.1 DESCRIPTION OF COMPRESSOR STAGEFig. 1 Investigated centrifugal compressorTable Dimensions of the impellerParameterValueNumber of blades (full + splitter) N 7+7Leading edge hub angle aC)31Leading edge shroud angle aj(°)46Trailing edge hub angle &/(")-27Trailing edge shroud angle 9J(°)-33Rotor/stator interfaceAs shown in Fig. 1, the impeller has 7 full blades and 7 splitter blades with dimensions shown in Table and Fig. 2. Splitter blade leading edges are located at 30% of full blade chord. The exit diameter of the impeller is 90 mm and the nominal point tip speed is 377 m/s at 80 kr/min. A vane less diffuser (6.5 mm*26.75 mm) is connected aft the impeller.Fig. 2 Meridional view of the blade2 COMPUTATIONAL METHODSThe mesh was created using NUMECA's IGG/AutoGrid81. Briefly AutoGrid is an automatic meshing scheme for turbo-machinery configurations. It provides tools to generate automatically a turbo-machinery mesh and ensure an optimal control of orthogonality and mesh point clustering for a correct depiction of viscous effects in the boundary layer. The mesh generation is based on a conformal mapping between the 3D spatial system (xyz coordinates) and the cylindrical surfaces of the 2D blade-to-bladespace (dm/r-theta plane).A sixteen-block mesh was generated using AutoGrid. The four main blocks represent part of the blade-splitter-blade passage extending from inlet to outlet. These blocks are generated using H-type mesh with 9x45x85 points in each block. Two C-type skin mesh blocks (9x45x137 and 9x45x97) containing the pressure and suction surfaces of both full and splitter blades are generated in order to improve the orthogonality. Two blocks of 41x45x9 points each are generated downstream of the full blade and splitter blade. To capture the flow physics upstream of the blades, two mesh blocks of 945x9 and 9x45x33 points are generated before main and splitter blades respectively. Diffuser is split into two blocks containing 57x45xl7 points each. Rests of the four blocks (9x13x137, 9x13x65, 9x13x97 and 9x13x45) are generated automatically with fine resolution (9 to 17 nodes from 0.11 mm clearance) to explore the flow phenomenon inside tip clearance region. Consequently, a total number of mesh points is 411 768.In order to investigate the impeller and diffuser separately, a rotor/stator boundary is set in CFD model as shown in Fig. 2 and frozen rotor approach is used in order to impose the continuity of velocity components and pressure.The computational simulations of the impeller physics were executed using the NUMECA's EURANUSm flow solver. In brief, EURANUS is a multipurpose code for 2D and 3D flows in complex geometries. It solves the time-dependent Reynolds-averaged Navier-Stokes equations. Turbulence can be modelled by an algebraic Baldwin-Lomax, Spalart Allmaras or two-equation k-e models. In present study, Spalart Allmaras model is used. The spatial discretization is based on a finite volume approach allowing a fully conservative discretization. An explicit time discretization is applied through a multi-stage Runge-Kutta procedure.The computer running for the present study were executed on Pentium IV 2.8 GHz and with 512 MB of memory. Typically, for each computation, 600 iterations were enough to reduce the mass flow residual by two orders of magnitude except for few computations where the iteration were increased up to 700. Each computation for the finest grid generally requires 3 h.3 RESULTS AND DISCUSSION3.1 Overall performanceThe performance maps of the stage at various clearance levels are shown in Fig. 3. The experimental and CFD results are found to be in close agreement. Minor deviations are found primarily because of the redesign of diffuser. The behavior of the CFD curves seems to satisfy the designer's goal. The mass flow rates and clearance sizes were chosen for two reasons. Firstly to undergo a parametric survey of the effect of tip clearance on the impeller performance and secondly to find the possible optimum size of the tip-clearance, which is not the zero tip clearance as indicated by GAO, et al'4.Similar to the work of DANISH, et alpl, the characteristic curves imply a considerable reduction in performance. For Ml, at mass flow rate of 0.35 kg/s, a loss of about 15 percent in peak pressure rise is observed as the tip clearance is increased from 0 to 1 mm. Similarly for M2 the loss is about 13%. The peak efficiency has also been reduced by 10 percent for the same case for Ml and 8.5% for M2.Fig. 3 further reflects that in contrast to GAO, et al'*1 and in agreement to Refs. 2-3 and Ref. 6, there is no clearance level indicating better performance than zero tip clearance hence the optimum tip clearance level, other than zero, doesn't exist for both the methods and for the compressor stage under study. 3.1.1 Overall performance below 0.5 mm tip clearanceIn case of M2, the choke flow is found to be reduced because of the reduction in shroud diameter however, as compared to method 1, the surge point seems to shift towards low mass flow because the simulation at the mass flow of 0.3 kg/s gives a very stable convergence but the similar simulation for Ml diverges indicating an early surge.SYED Noman Danish, et al: Comparison of two methods to increase tip clearance and its effect 62 on performance of turbocharger centrifugal compressor stage2.5 2.4 2.3 2.2 2.1 2.01.9 1.8 1.7A-0.32 d/l(2)2.80.30kg/s.Ml,o)=-1.624 80.35kg/s, Ml, <u=-1.66890.40kg/s, Ml, <m=-1.68912.60.45kg/s. Ml, <d=-1.647 1. 0.30 kg/s, M2, <d=-1.608 42.42.22.01.80.35kg/s, M2,<d=-I.743 90.40kg/s, M2.<d=-1.669 1_._0.425 kg/s, M2,o=-0.777 41.61.40.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 Tip clearance ratio X Fig. 4 Stage total pressure ratio vs tip clearance ratio0.30kg/s, Ml, y0.280 50.35kg/s, Ml, y=-0.32230.40kg/s, Ml, r=-0.398 30.45 kg/s, Ml, y = -0.810 50.30 kg/s, M2, r = -0.366 00.850.80£ 0.75'5w 0.70.0.35kg/s, M2,y=-0.32l 6-.0.40kg/s, M2,y=-0.299 2- 0.425 kg/s, M2,y=-0.298 50.65 - -(1)+ Xo1.0 mmExperiment0.J00.250.450.300.350.40Mass flow rate v/(kg«s"') (a) Pressure ratio8479I74-i69" +0mm0.2 mmX0.4 mm64-00.6 mm01.0 mm59,0.250.300.350.400.450.50Mass flow rate v/(kg. s') (b) Efficiency Fig. 3 Characteristic curves of the stageAll pressure contours for M2 are found to be more steeper than those of Ml indicating more pressure drop at high mass flow rates and low choke flow. Similar trend is found with efficiency curves. 3.1.2 Overall performance above 0.5 mm tip clearanceAll pressure and efficiency contours for M2 are found to be less steeper than those of Ml indicating less pressure and efficiency drop at high mass flow rates. Increased efficiency at high mass flow rates for M2 also indicates the improvement in choke flow because of the increase in shroud diameter.3.2 Modeling of total pressure ratio of stageFig. 4 is the plot of total pressure ratio against the tip clearance ratio (A). Tip clearance ratio is defined as the ratio of axial tip clearance to the blade width at impeller exit. Curve fitting of the data reveals that all the characteristic curves have almost the identical slope which can be given asdnAX=-1.665Eq. (1) is valid both for Ml and M2 except for the mass flow rate of 0.425 kg/s (M2)3.3 Modeling of stage efficiencyFig. 5 is the plot of stage efficiency against the tip clearance ratio. Curve fitting of the data discloses that all the characteristic curves have almost the identical slope which can be given as0.60,0 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 Tip clearance rat