Control of Corner Separation by Optim ized Slot Configuration in a Linear Compressor Cascade

Jinjing SunYangweiLiu Lipeng Lu/SchoolofEnergy and Power Engineering,Beihang UniversityXavier OTTAVY/Laboratoire deMecanique des Fluidesetd’Acoustique,Ecole Centrale de Lyon

Control of Corner Separation by Optim ized Slot Configuration in a Linear Compressor Cascade

Jinjing Sun＊YangweiLiu Lipeng Lu/SchoolofEnergy and Power Engineering,Beihang University
Xavier OTTAVY/Laboratoire deMecanique des Fluidesetd’Acoustique,Ecole Centrale de Lyon

：为控制压气机叶片吸力面与端壁处所形成的三维角区分离，本文在叶根处引入了从压力面到吸力面的槽道来控制角区分离流动。基于利用优化算法对叶根处槽道几何的优化结果，本文通过数值模拟和实验来分析槽道在不同进口攻角下对角区分离的控制效果。本文通过对比原始和开槽叶栅中总压损失、静压升和三维角区分离特征表明叶根处引入的槽道对角区分离有明显的控制效果。

compressor；cornerseparation；slot bleeding；passivecontrol

Nomenclature

Ca=axialchord

β1'=design inflow angle

i=incidenceangle

z=Spanwise direction

RSM=Reynolds StressModel

MRSM=Modified RSMModel

p=static pressure

p0=totalpressure

1=inlet

2=local

LMFA=Laboratoire de Mecanique des Fluidesetd’Acoustique

Introduction

Flow in axial fans,especially in off-design conditions,is very complicated. The flow separationswould cause deleterious consequences such as a decrease in the efficiency and an increase in the losses and blockage,which even might cause rotating stalland surge in the fan.

Three-dimensional corner separation which occurs at the junction of the suction surface of the blade and the end-wall(hub or casing)is one of the high loss flow structures in compressor systems[1].As the trend of compressor designs is towards higher loadings,with less stages and larger operating ranges,the adverse pressure gradientsmight be of higher importance and the boundary layers separation might be more severe.Thus the corner separation strengthened by these two elements would highly restrict the efficiency and stability of the compressor.Therefore,control of corner separation is one of the keys to improve compressor performance[2].

In the previouswork of Mu X.etal.[3],a newmethod with slotat the rootof the blade has been proposed to control corner separation in a linear compressor cascade. This paper further optimizes the slot design at the rootof the blade in a linear compressor cascade to control the corner separation.The benefit achieved by the flow controlmethod allows the compressor to be designed with fewer blades(increasing the loading per blade)which indicates a reduction of theweight and/or an improvement in the efficiency and theoperating range.

Nume rical Approach

This work was carried out with a low speed linear cascade.The cascade consists of 13 NACA65-009 blades.The notation used in describing this subsonic compressor cascade isshown in Figure1.

For this NACA blade cascade the optimum (minimum losses from the NACA correlations)angle of incidence is about 0.18°,which corresponds to an inlet flow angleβ1'equaled to 54.31°.For thisstudy,an incidence angle of 4°was chosen in order to have a large enough corner separation.This corresponds obviously to an off-design operating point,with an increase of the loading.The parameters of the compressor cascade are summarized in Table 1.The Reynolds number of 3.82×105with the dimension of the blade leads to an inflow velocity of 40 m/s.The experimentwas carried outby Ma and Zamboninietal. at LMFA[4-6].For the numerical simulations of the flow in the cascade with and without flow control, RANSmodels in the commercialsoftware FLUENTwas used in thiswork.

Table1 Geometry and aerodynam ic parameters

In order to uncouple the corner separation effects from what is happening at mid-span and also to decrease the influence of the other side of the blade, the span height in this study is very large compared to standard compressor blades(leading then to a high value of the aspect ratio of 2.47).Therefore,in order to reduce the computational time,one half of the blade passage(from the end-wall tomid-span of the blade)is simulated and a symmetry condition is employed at mid-span.HOH mesh type is generated by the commercial software package AutoGrid5TM.Fig.2 shows the 2D mesh and computation domain in the cascade.The inlet plane is positioned at1.59c upstream of the blade leading edge.Considering of the outgoing perturbations and spurious reflections,the computation domain of the outletsection was setone chord extended downstream of the experiment outlet section.The first mesh distanceat the solidwall isset to aboutΔy=1× 10-5m,whichmeans thatΔy+＜1 both for the blade and the end-wall and the grid expansion ratio is less than 1.2,which satisfies the requirement of the turbulence models used in this work.For this computational domain,81 grid points are given in the span-wise direction.Caseswith different grid number in themain flow have beencalculated to insure the grid convergence.Details in the flow field and total pressure loss coefficients at the blade trailing edge(presented here in Fig.3)show that non-significant differen-ces are found when the grid number reaches 1.2 million. Therefore,the total points numberof themesh has been chosen around1.2million in thisstudy.

In order to setup the correct inlet boundary conditions,the turbulent boundary layer that develops on the end-wall has to be taken into account,as it plays amajor role in the corner separation.With this goal,a RANS calculation of a turbulent boundary layer that develops on a flat plate has been achieved.The inlet boundary layer profiles of velocity as well as the turbulence quantities have been extracted at the location where the simulated displacement thicknessδ1of the boundary layer meets the experiment results of Ma[4].The extracted boundary layer profiles have been imposed at the inletsection with the inlet flow angle as the inlet flow boundary conditions.A uniform standard atmospheric static pressure is employed at the outlet section.Both the blade surface and the end-wall are set as non-slip adiabatic walls.Periodic conditions are employed on the two pitch-wiseboundariesof themesh. The pressure-velocity coupling is performed using the SIMPLE algorithm.Second-order spatial interpolation is used for the convection terms and a second order central difference scheme is used for the diffusion terms.

Turbulence Model

As it is known there is no universal turbulence modelwhich is suitable for all kinds of flow conditions. This study concerns the optimization of a flow control design.Simulations such as LES(Large Eddy Simulation)can not be used for optimization as the needed CPU costwould be not reasonable.On theother hand,for the corner separation simulation in the case of this cascade,RANS simulations have show nover-predictions of the separation[12].The reason for the turbulence modelmisalignment is thatmost of its calibration is performed in equilibrium flows,while the turbulence in the corner separation is in non-equilibrium state.As a consequence the model cannot capture the complete corner separation characteristics.

In order to predict the flow field accurately with a RANS approach,turbulence modelmodifications have been conducted with the goalof correctly predicting the separated flow in the case of the original blade(without slot).The modification of the RSM model is not the scope of this paper,butsome indications are given here after to justify theuseof thisapproach.

The exact transport equations for the transport of

Wang et al.[13]proposed that the prediction of the corner separation can be improved by modifying the turbulence transport nature.For the work of Yan et al. [14],decreasing the valueof the production term in theω equation leads to better results for the simulation of the corner separation.Based on this concept,dissipation modelmodification in RSMmodelwas conducted in this work.The dissipation tensorεijismodelled as:

The scalar dissipation rate,ε,is computed with a model transportequation as:

whereσε=1.0,Cε1=1.44,Cε2=1.92 are constant which can bemodified in the solver.Judging from the simulations results with different values for those constants,better results can be reached decreasing the constant parameter Cε1.Thosemodifications have been adjusted using the results of the experiment and the LES achieved by Gao et al.[15].Those modifications have been optimized for the case with no slot and an incident angle of 4°.With a modified value of the constant Cε1=1.008 the corner separation flow is very satisfactory.This constant has been kept the same for all the simulations in this study,i.e.with orwithout the control of the flow using slot,and for all the incidence angles.

In the next sections results will be discussed usingsome coefficients.The static pressure coefficient is defined in equation.（4）.

The total pressure loss coefficient is defined in equation.（5）.

Quantification of the passage end-wall blockage employed in thispaper isdefined as[16]:

whereρis the local density,umis the velocity component normal to A,A is the azimuthal cross-section.ρeand Ueare the edge density and velocity of the defect region,is the area of the test plane of the blade passage.Note that the density is constant in this configuration.

Fig.4 shows the total pressure loss coefficient plotted for the experiment,LES result,the original RSM and themodified RSM models.The z/h=0 is for the location of the endwall and z/h=0.5 corresponds to the mid-span.The vertical orange strip at the left of each plot represents the loss generated by the wake of the blade.The maximum of the loss is created in the corner between the suction side of the blade and the end-wall.Note thateven withoutany corner separation it will remain some loss at the endwall which is generated by the turbulent boundary layer that develops on the end-wall upstream of the cascade.In the first left plot,the black points are for the location of the measurements with a 5-hole pressure probe. Taking the experimental results as reference,the Modified RSM approach gives a better prediction for the total pressure loss compared to the original RSM model.The size of the separation is not toomuch over estimated and the total pressure gradient is better calculated.Not presented here,the static pressure distributions on the suction and pressure sides of the blade are in good agreementwith the experiment.This leads then to a much better distribution of the total pressure loss mass weighted averaged along the pitch-wise direction,which is plotted in Fig.5.Such resultsvalidate the choiceof the constant.

Definition of the slot

Slot Con figuration Arrangem ent

The slot geometry with curves is illustrated in Fig. 6.Five geometric parameters have been given in order to define the geometry of the slot configuration:the inlet position of the slot S1,the outlet position of the slot S2, the inlet length of the slot L1,the outlet length of the slot L2and the height of the slot H.This should be noted that,the inlet and outlet position of the slot are defined at themiddle position of line AC and BD.Line AB is a S shape governed by a cubic equation with 4 constants which are defined by the position of the points A and B,and the tangents to the blade pressure surface and suction surface at points A and B respectively.Line CD is an arc which is defined by the position of the points C and D with also the tangent to the suction surface at point D.The curves are then tangent to the blade suction and pressure sides at pointA,B and D.The positions of points A and C are determined by the parameters S1and L1,the positions of points B and D are determinedby the parameters S2and L2(see Fig.6).

In the optimization process,the outlet position of the slot on the suction surface S2is set initially at the location of the onset of the corner separation close to the end-wall,in the reference case without a slot.For the reference slot case(initial slot configuration for the optimization)calculated for an incident angle of 4°,S2is 55%Ca and L2is 27%Ca.In order to take advantage of the pressure difference between the pressure and suction surfaces,so as to obtain an accelerated flow getting out of the slot on the suction side,the inlet position of the slot is recommended to tend towards the leading edge.

It should be pointed out that to avoid simulation errors induced by differentmeshes between the original case and the slotted cases,themesh of the slotted cases outside of the blade remains the same(as the original case)but the mesh in the slot is added and generated with the unstructured mesh.For the reference slotted blade,31 nodes are distributed for the slot height and for lines AB and AC,201 nodes are distributed,which results in a total number of about 10000 cells tomesh the slot.

Param eterization Formu lation for SlotCon figuration

The parametrization technology is based on the calculation of the derivatives of the aerodynamic steady flow-filed by the Navier-Stokes equationswith respect to the design parameters and on the computation of a Taylor expansion of the solution to higher order derivatives.The calculation is performed at one particular configuration thatwill be treated as the base line configuration and in general the selection of the base line case is essential to the convergence of the optimization.Using the flow-field derivatives and the base line configuration,it is then possible to extra polate the flow-field for any configuration using an appropriate reconstructionmethod.

The Navier-Stokes equations can be written as a simple symbolic form:

F is the vector resulting from viscousand convective fluxes over any arbitrary volume of fluid.q is the flux vector expressing mass,momentum and energy conservation with respect to the conservative variables（ρ,ρV,ρE）,and also the turbulentvariablesκandω.

First differentiating of equation（8）with repect to the variables p can bewritten as:

where q（1）is the firstorder variation of q.Denoting the Jacobianmatrix:

and

Equation（9）can bewritten as:

The high order derivatives q（n）with respect to the parameters variations can be built by a multi-parameters Taylorexpansion ofequation（10）:

Finally the new flow field corresponding to the modified parameters vector can be reconstructed in the manner:

根据女性盆底功能障碍的特点及临床表现，其可归属于中医学“阴脱”、“ 遗溺”等病证范畴，其病因病机主要是由于年高体衰，或妊娠、产次过多，损伤脏腑功能，致脾肾亏虚，脾主肌肉，肾主摄纳，脾肾亏虚则摄纳之力减弱，致冲任不固而下脱，导致阴脱、遗溺等病证的发生。因此在治疗本病时，总以健脾补肾、调理冲任为主。本研究选择的会阴穴是任脉要穴，也是冲任及督脉之交会穴，具有统摄气血运行、维持阴阳平衡的功效。温针灸是一种针刺、艾灸的有机结合治疗方法，用其刺激会阴穴可发挥大补元气、升举清阳、固脱止遗的作用。综上，采用温针灸治疗女性盆底功能障碍效果理想，值得推广。

The linear systems solved in equation（11）are only based on the Jacobian matrix G,which is already available in classical CFD solver.

The whole procedure is implemented by the commercial software FLUENTTMcoupled with the parameterized code developed in Turbo’Opty at ECL [7].More detailed introduction of the parameterized code can be referred in[8-11].

The reference value of the slotconfiguration and the range for the optimization for each parameter are listed in Table2.

Table2 Parameter ranges for the slotcon figuration

Results and Discussion

Three objectives are defined to evaluate the performanceof the cascade.

·The minimization of the total pressure loss coefficientatblade trailing edge.

·The maximization of static pressure rise coefficient.

·Theminimization of the blockage B in the blade passage.

Based on the configuration with 4°incidence angle the slot geometry parameters have been optimized. These parameters,presented in Table 3,have then been chosen to calculate the cascade performance for the whole incidenceangles range[-2°，7°].

Table3 Optimalgeometricalparameters for the slotat them inimum Yp

Table 4 shows the comparison of the total pressure lossat 4°incidence angle between the original cascade and the so-called slotted cascade.Seen from the table, cascade performance has been improved by the slot in the blades.More details of the flow field in the slotted cascadewillbeanalyzed below.

Table4 Optimaldesign results at4°incidence ang le

Experiments have been carried out in order to validate those results obtained with numerical simulations.Five blades at the middle of the cascade have been replaced by 5 new blades designed with a slot at both extremities.It has been checked that the new blades have been correctly manufactured and inserted in the cascade.Tape has been used to seal the slots on both the pressure and suction sides,so as to obtain a similar profile to the original blades and to restore the original performance of the cascade.A five-hole pressure probe was used to measure the outlet flow.The relative uncertainty of the downstream pressure measurements was about 2%in total pressure coefficient.Measurements have been performed and compared to the experiment conducted by Zamboniniat 4°incidence angle[6].The total pressure loss coefficients are plotted at 0.363 downstream of the trailing edge in Fig.8.A good agreement is observed between the reference cascade and the modified cascade with taped blades.For the configuration no significantdifference can be noticed in thewake and in the corner separated flow regions,which indicates that the setup of the slotted blades is proper and does not modify the flow field.The experiment results of slotted blades with tape are chosen as the original cascade performance for the followinganalysis.

The contours of total pressure loss measured and simulated at the blade trailing edge at4°incidence are presented in Fig.9 for the original and slotted blades. Simulation results fit very wellwith the experiment for both the originaland slotted cascades.The high loss region near the end-wall is significantly improved by the effects of the slot located at the root of the blades. The phenomenon can be also observed in the pitch-wise-averaged total pressure loss coefficient in Fig.10.This should be noticed that the height of the slot is 0.02m,which corresponds to 13.3%of the blade chord.Therefore,the effects produced by the slot and the induced jetat the rootof the blade improve the flow in the cascade notonly locally butalso until 20%of the span height.

The slot generates some changes in the losses because it reduces the corner separation.That leads obviously to some changes in the pressure distribution around the blades.Detailed comparisons of the static pressure coefficient around the blade are plotted at mid-span and close to the end-wall in Fig.11.The experimental results are represented by the black symbols.The black and redsolid lines are for the LES and RANS results respectively.Note that the EXP,LESand RANS results are obtained for the original cascade (without slot).The blue solid line stands for the RANS results obtained for the slotted blades.Concerning the original configuration,the corner separation can be seen close to the endwall at 5.4%of the span-height（Fig.11(a)).It is characterized by constant values of Cp on the suction side in the region of the trailing edge; this characteristic plateau is obviously notobservable at mid-span where the flow is completely attached（Fig. 11（b））.

The distribution for the bladeswith the slotdoesnot introduce any significant difference at mid-span. However,the blade loading is significantly improved at 5.4%of the span height.The plateau at the rear of the suction surface disappears,meaning that the separation has almost been suppressed.The global blade loading is even slightly improved.Two singularities appear on the distribution for the slotted blade;they are induced by the sharp geometry of the slotat point B and C.At point B,the flow getting out of the slot with higher momentum produces a sudden decrease in the static pressure coefficient.Note that it should be paid attention to the geometry at point B during the manufacturing of the slotted blades,as the S curve AB is tangent to the suction surface at this point,leading to an extremely thin bladeatpoint B.

The static pressure rise coefficienton the endwall is presented in Fig.12 for the original and slotted cascades at 4°incidence angle.The static pressure rise has been improved to a large extent,especially at the trailing edge of the suction side,where Cp increased from 0.2～0.3 to 0.3～0.4 compared with the original case.The static pressure rise in the slot is consistentwith themain flow.

Another part of the loss sourceis mainly for the mixing of the fluid in the slot and themainflow which can be seen in Fig.13.

Flow jets which generated by the pressure gradient from the blade pressure surface to the suction surface can accelerate the lowmomentum fluid in the boundary layeron the suction surface and the end-wall.Note that themass flow rate of the slot is about0.85%of the inlet mass flow rate.The separation is suppressed because the flow out from the slot prevents the secondary flow from interacting with the suction surface boundary layer near the end-wall（Fig.13(b)).So theaccumulation of the low energy fluid at the corner decreases and the flow capacity is also enhanced.Not only the total pressure loss reduces but also the deflection of the flow is improved,that leads toenlarge theoperating rangeof the vane.

Nevertheless,the presence of the slot jet in the flow can have some disadvantages.It is noticing that at low incidence angles,where the corner separation keeps small without slot,the presence of the slot creates a blockage in the passage and generates extra loss induced by its mixing with the flow in the blade passage.This iswhy in the design process of the slot it has to be checked that these negative effects keep small for the low incidence angles.Experiment and numerical simulations have been conducted to validate the effectiveness of the slot for the range of the available incidence angle,from i=2°to 7°with the slot configuration optimized at 4°incidence.The pitch-wise averaged total pressure loss coefficients are plotted in Fig.14 and Fig.15 for i=2°and i=7° respectively.In the cases without slot,the coreloss region located at the corner,formed by the suction side of the blade and the endwall,becomes stronger when increasing the incidence angle from 2°to 7°.This is directly linked to the development of the corner separation.At the incidence of 2°,the corner separation remains small,so the benefit induced by the slots is not so obvious but at least the slot does not introduce some extra loss.At 7°the control of the separation is very efficient.This indicates that after eliminating the high loss region generated by the corner separation,themain loss that remains in the passage is the profile loss(located in the wake of the blade)and the loss induced by theboundary layeron theend-wall. In Fig.15,the RANS results obviously over predict the size of the separation and the loss that itgenerates.For larger separations the MRSM model,modified for the case of i=4°which produces a reasonable separation, does not capture accurately the physics but give the good trends,as it can be seen when comparing the caseswith andwithoutslot.

Conclusions

Experimental and numerical simulations have been conducted to study the effects on the corner separation of the slotat the rootof the blade in a linear compressor cascade.Optimization is also conducted for the optimal control result for the corner separation.The results are summarized as follows:

（a）At low incidence angles,lower than 4°,the slotused to control the flow does not produce significant effects.The distributions of the total pressure loss vary slightly which indicates that the main total pressure loss remaining in the cascade are induced by（i）the blade profile itself,（ii）the boundary layer on the end-wall and（iii）the flow in the slot and itsmixing with the low momentum flow on the suction side of the blade.

（b）At large incidence angles,higher than 4°, experimentaland numerical results have shown that the slot has a large beneficial impact.It can decrease the total pressure loss,the passage blockage caused by the corner separation and improve the deviation of the flow.

The goal of this study is to prove that slots can be effective to reduce losses and blockage induced by the corner separation in a compressor cascade when the loading of the blade is increased.Those results can be extended to the case of a real compressor,keeping in mind that in that case the design and optimization of the blade and the slot should obviously be carried out at the same time.

As a conclusion,the controlof the corner separation using slots as described in this paper is an efficientway to reduce the lossand enlarge theoperating range.

Acknow ledgments

This work is supported by the National Natural Science Foundation of China(No.51676007,No. 51376001,No.51420105008).Jinjing Sun was supported by the China Scholarship council(CSC).

[1]W isler D C.Loss reduction in axial-flow compressors through low-speedmodel test-ing[J].Journal of engineering for gas turbines and power,1985,107(2):354-363.

[2]Joslyn H D,Dring R P.Axial compressor statoraerodynamics[J]. Journal of Engineering for Gas Turbines and Power,1985,107(2): 485-492.

[3]Mu X,Lu LP.ControlofCorner Separation by Slotat the Rootof the Blade[J].Gas Turbine Experimentand Reasearch,2007,20(3):28-33. (in Chinese)

[4]Wei M A,Ottavy X,Lipeng L U,et al.Experimental study of corner stall in a linear compressor cascade[J].Chinese Journal of Aeronautics,2011,24(3):235-242.

[5]Ma W,Ottavy X,Lu L,et al.Experimental investigations of corner stall in a linear com-pressor cascade[C].ASME 2011 Turbo Expo:Turbine Technical Conference and Exposition.American Society ofMechanical Engineers,2011:39-51.

[6]Zambonini G,Ottavy X.Unsteady pressure investigations of corner separated flow in a linear compressor cascade[C].ASME Turbo Expo 2015:Turbine Technical Conference and Exposition.American Society of Mechanical Engineers,2015:ASME Paper No. GT2015-42073.

[7]Aubert S,Smati L,FerrandP.Numerical Analysis of Oscillating Shock-Boundary Layer Interaction[C].Proceedings of the 21st International Symposium on Shock Waves,Great Kepple Island, Australia,1997.

[8]Moreau S,Aubert S,Grondin G,et al.Op-timization of a Fan Blade Cascade Using the Parametric Flow Solver Turb’Opty[C].ASME 2006 2nd Joint US-European Fluids Engi-neering Summer Meeting Collocated With the 14th International Conference on Nuclear En-gineering.American Society of Mechanical Engineers,2006: 345-350.

[9]Aubert S,Ferrand P,Pacull F,et al.Fast CFD for Shape And Flow Parameterization With Metamodels built on High-Order Deriv-atives.Application to Fast Design[C].27th International congress of The AeronauticalSciences France,ICASPaper.2010,2:2.

[10]Soulat L,Ferrand P,Moreau S,et al.Efficient optimisation procedure for design problems in fluid mechanics[J].Computers& Fluids,2013,82:73-86.

[11]Buisson M,Ferrand P,Soulat L,et al.Optimal design of an automotive fan using the Turb’Opty meta-model[J].Computers& Flu-ids,2013,80:207-213.

[12]Gao F,Ma W,Boudet J,et al.Numerical analysis of three-dimensional corner separation in a linear compressor cascade[C]. ASME Turbo Expo 2013:Turbine Technical Confe-rence and Exposition.American Society of Mechanical Engineers,2013:ASME Paper No.GT2013-95581.

[13]DanhuaW,Lipeng L,Qiushi L.Im-provement on SA model for predicting corner separation based on turbulence transport na-ture[C]. 45th AIAA/ASME/SAE/ASEE Joint Propulsion Conference&Exhibit (August),2009.

[14]Liu Y W,Yan H,Fang L,et al.Modified k-ωmodel using kinematic vorticity for corner separation in compressor cascades[J]. ScienceChina Technological Sciences,2016,59(5):795-806.

[15]Gao F,Ma W,ZamboniniG,etal.Large-eddy simulation of 3D corner separa-tion in a linear compressor cascade[J].Physics of Fluids (1994-present),2015,27(8):085105.

[16]Khalid SA,Khalsa A S,Waitz IA.Endwall blockage in axial compressors[J].Journalof Turbomachi-nery,1999,121:499-509.

声明

为适应我国信息化建设，拓宽本刊及作者知识信息交流渠道，本刊已被CNKI中国期刊全文数据库、万方数据——数字化期刊群、中文科技期刊数据库及本刊的电子媒体《中国风机技术网》收录，其作者著作权使用费与本刊稿酬一次性给付。如有作者不同意文章被收录，请在来稿时向本刊声明，本刊将做适当处理。

本刊

压气机；角区分离；叶根开槽；被动控制

TH453;TK05

A

1006-8155（2016）06-0009-10

10.16492/j.fjjs.2016.06.0056

*Jinjing Sun/Laboratoire de Mecanique des Fluidesetd’Acoustique,EcoleCentralede Lyon Received date：2016-03-19 Beijing 100191

Abstract:In order to control the separation which occurs in the corner formed by the suction side of a compressor blade and the end-wall,aslotconnecting thepressureside to the suction side can be used.Based on the optim ization of theslotdesign atthe rootof the linear compressor cascade blades,experiments and numerical simulations have been conducted to evaluate the effects of the slot at several incidence angles.Total pressure loss coefficient,averaged static pressure rise coefficientand feature of the 3D separation in the cascade passages at 4°incidence angle have been compared for the cases w ith and w ithout slots.The results show an obvious improvement when using such a slot for the controlofcornerseparation.