Stability Enhancement with Self-recirculating Injection in a Single-stage Axial Flow Compressor

Jichao Li Juan Du Feng Lin

Key Laboratory of Advanced Energy and Power, Institute of Engineering Thermophysics,

Chinese Academy of Sciences, P.R. China lijichao@iet.cn

Abstract: Self-recirculating injection that was demonstrated to delay the occurrence of stall is experimentally studied in a single-stage axial flow compressor. It bleeds air from the downstream duct of the stator blade row and injects the air as a wall jet upstream of the rotor blade row. The bleed ports, injection ports and recirculating channels are circumferentially discrete and occupy only 38% of the circumference. The external injection and outlet bleed air are selected for comparison with the self-recirculating injection. Results show that the self-recirculating injection has an equivalent stabilizing ability compared to the external injection with the maximum injected mass flow under self-recirculating injection condition. It can improve the stall margin by 13.67% with no efficiency penalty compared with smooth casing; the re-circulated mass flow ratio near stall is accordingly generated by 0.48%. However, the outlet bleed air basically has no contribution on the stall margin improvement. It indicates that the upstream injection in the self-recirculating injection plays an important role in terms of stability-enhancement. The details of flow field are captured using a collection of high frequency pressure transducers on the casing with circumferential and chord-wise spatial resolution. The detailed comparative analysis of the characteristic flow in terms of endwall flow in the tip gap indicates that self-recirculating injection can postpone the occurrence of stalling in the proposed compressor through weakening the unsteadiness of tip leakage flow, delaying the forward movement of the interface between the tip leakage and main stream flow, and sharply declining the circumferentially propagatingspeed induced by tip leakage flow that triggers the spike-type stall inception. This study may be helpful to guide the design of self-recirculating injection in actual application. Keywords: compressor; self-recirculating injection; stall margin improvement; tip leakage flow;circumferential propagation; stall

1. INTRODUCTION

As a type of flow instabilities in an aero-engine compressor, stall generally occurs before surge in axial flow compressor. To avoid the occurrence of stall, numerous passive and active stability-enhancing technologies have been proposed almost from the outset of the first gas turbine engines. Hathaway [1] has summarized a historical survey of passive endwall treatment designs. A typical form of endwall treatment involves extracting high-pressure air from a downstream location and re-injecting the air just upstream of the critical blade row in the same compressor [2]. Some researchers separately extract the upstream re-injecting and develop tip air injection, including active tip air injection [3-7] and steady tip air injection [8-10].

However, the external auxiliary compression that supplies the high-pressure injected air is still under question for the actual application, especially in the aero-engine. This embarrassing problem makes researchers return to the initial self-recirculating injection concept. See, for example, Freeman et al. [11] and Leinhos et al.[12] reported stall control efforts with tip air injection that the air is recirculated from the high-pressure stage and injected into the front stage in the actual aero-engine. Hathaway [13] proposed the self-recirculating casing treatment concept numerically in the National Aeronautics and Space Administration (NASA) Stage 67 that can provide a considerable benefit in stall range extension with no predicted loss in efficiency or total pressure rise capability.

* Corresponding author: Institute of Engineering Thermophysics, CAS, P.R. China, Email:lijichao@iet.cn

Strazisar et al. [14] experimentally investigated the endwall recirculation by redesigning and optimizing the injector in the NASA Sage 35. They demonstrated 2% of stall margin improvement (SMI) at the design speed and 6% of SMI at 70% of speed by injecting 0.9% of annulus flow at the stall point under both operating speeds by using the redesigned compact injectors. Recent year, Weichert et al. [15] proposed the self-regulating casing treatment that bleeds air over the rotor blade tip and injects in the upstream of the same rotor. They reported parameter studies that the reinjection ports are optimized for axial location, shape, and orientation, and 6.1% of SMI with 0.8% of efficiency penalty is achieved for less than 0.25% of the inlet mass flow recirculated near stall. Guinet et al. [16] simulated the recirculating tip blowing casing treatment in a transonic compressor by using URANS and analyzed its influence on the blockage of tip leakage flow preliminarily under different tip gaps. They also chose the axial slot to make a comparison and found that the axial slot is more effective in enhancing the compressor stability by suppressing the tip vortex and giving the blockage flow the possibility to evade from the main flow [17]. Grosvenor et al. [18] reported numerical studies of the recirculated flow tip injection in the NASA Stage 67 and demonstrated that the air injection can postpone the stall limit by influencing the shock and the tip leakage flow. All the aforementioned findings show that the self-recirculating injection as an effective stability-enhancing method has more potential for practical application than the separate tip air injection by using external air sources in the future [19]. However, there still exist some problems in the previous researches: 1) The SMI contributed by self-recirculating injection is limited, sometimes is unstable due to lack of understanding of its effect mechanism on the flow field; 2)

The design guideline is difficult to form that obtains an enough SMI with minimum efficiency penalty due to lack of structural optimization and sufficient experimental validation. Motivated by these two urgent problems for the actual application, the self-recirculating injection is experimentally studied through the elaborate design and optimization of the self-recirculating device.

Thus, this paper is organized as follows: after the brief

introduction of previous studies in Section 1. The test rig and measurement is described in Section 2. In Section 3, the design and optimization of the self-recirculating device is presented. Subsequently, the self-recirculating injection is experimentally studied in a single-stage axial flow compressor in Section 4. Section 5 shows the description of the effect mechanisms on the flow filed. Finally, the conclusion is summarized.

2. TEST RIG AND MEASUREMENT SETUP

An in-house low-speed axial flow compressor with a design speed of 2400 rpm was used for the experimental work in this study. The blade is designed to model a modern high-pressure compressor. The detailed parameters of this compressor are shown in Table 1.

Table 1. Key parameters of test compressor

Design speed (rpm) 2400 Rotor blade number 60 Outer diameter (mm) 500 Design Mass flow rate (kg/s) 2.9 Rotor tip chord (mm) 36.3 Rotor tip stagger angle (deg) 39.2 Hub–tip ratio 0.75 Tip clearance(mm) 0.8

The single-stage axial flow compressor is taken as an example in

Fig.1 to describe the arrangements of the injector, bleed port and the steady, and the unsteady measuring positions. A total of 15 discrete self-recirculating devices are equally distributed around the annulus and occupy only 38% of the circumferential area. The installation position of the injector is approximately 0.06 axial chord (Cax) away from the leading edge (LE). The bleed port is approximately 0.3 Cax away from the trailing edge (TE). The measurement points of the inlet and outlet static pressures are 6 Cax away from the LE and 3 Cax away from the TE, respectively. Six pressure transducers from 0.21 Cax upstream of the LE to the TE are placed in the casing. A row of sensors consist of 16 pressure transducers fitted at the axial location as 0.26 Cax; and evenly spaced circumferentially. The unsteady pressure in the tip gap is measured by using high-frequency Kulite pressure transducers. The sampling rate of the unsteady measurement is 100 kHz. Compared with the blade passing frequency (BPF), which is approximately 2.4 kHz at design speed, this sampling rate is sufficient for dynamic measurements. A Hall sensor installed on the shaft is also used to resolve the shaft revolution in dynamic measurements. This sensor provides a window-shaped signal at the fixed circumferential position in each revolution, which is used for the phase locking when processing the dynamic data. In addition, a steady Kiel-probe with an adjacent static pressure hole is installed in the self-recirculating device to measure the total pressure and static pressure. Because of the low rotational speed of the compressor, the density of the air is approximately constant that the re-circulated mass flow could be calculated using the total pressure and static pressure.

Rotating

Inflow

1

4

3

2LE

TE

RotorStator

Injector

Suction port

5

1. Inlet static pressure 2. Position of the injector 3. Pressure transducers 4. Position of the bleed port 5. Outlet static pressure

Figure1: Arrangement of the sensors and self-recirculating device

3. OPTIMIZATION OF SELF-RECIRCULATING DEVICE

Before the experimental performance test with the self-recirculating injection, the structure of the self-recirculating device has to be appropriately optimized. Previous results have demonstrated that the injected momentum ratio (injected momentum /incoming main stream momentum) plays an important role in terms of SMI (see Li et al. [7] and Kefalakis et al. [9]). In this study, the optimization guideline is based on the understanding of stabilizing mechanism with tip air injection. Li et al. [7] verified the existence of the demarcation ratio of injected momentum which is defined to distinguish the different trends in the curve of SMI versus injected momentum ratio as shown in Fig.2, they presented dual stability-enhancing mechanisms of the tip air injection and tried to provide a unified explanation for the different injected momentum ratios (relative to the mean stream momentum at the stall point), which is separated by the demarcation momentum ratio. Li demonstrated that when the injected momentum ratio is less than the demarcation ratio (defined as micro injection), it can only act on the tip leakage flow and thus provide a small SMI by weakening the self-induced unsteadiness of tip leakage flow (UTLF). By contrast, the macro injection with an injected momentum ratio more than the demarcation ratio can provide a larger SMI by suppressing the UTLF, acting on the main flow, decreasing the inlet angle of attack, and thus unloading the blade tip. Thus, the optimization guidelines are formed in the process of the structural optimization of the self-recirculating device. One is that the recirculated air should be injected into the tip clearance as much as possible for the upstream injector. The other is that the recirculated mass flow ratio should be sufficient to generate more injected momentum ratio for the downstream bleed port.

Injector

Pressure transducer

LE

TE

Rotor

Bleed port

Stator

Rotor

Self-recirculating device

Inflow

0.1 1 10 1000

10

20

30

40

50

S

tall

mar

gin

impr

ovem

ent(

%)

Injected Momentum/Free Stream Momentum(%)

Micro injection Macro injection

Tip air injection

Figure 2: SMI versus injected momentum ratios cited from Li et al.[7]

Experience indicates that the Coanda injector is recognized to

have the best wall-attachment effect (see Strasizer et al. [14]); thus, the injected air can fully act on the tip clearance flow to extend the stall margin (see Tong et al. [6] and Li et al. [7]). Therefore, the injector in this study is designed using the Coanda structure. Its structure and the corresponding numerically simulating results through computational fluid dynamic (Ansys Fluent software) are shown in Fig.3. The demand of the parameter setting in the Ansys Fluent software and its detailed introduction are ignored in this study. The detailed procedures for the numerical simulation were given by Geng et al. [20]. The velocity contour in Fig.3 indicates that most of injected air could affect the flow field in the tip clearance; it conforms to the design requirement. For the bleed port, in order to obtain sufficient extracted air considering the injected momentum ratio of upstream injection, the optimization of the shape of the suction port is crucial. Especially considering the future application in the actual multi-stages compressor that the gap between rotor and stator is very short, taking for an example our three-stage compressor, the stage gap is only 5mm. However, the optimization work for the suction port is not often reported in the open literature. Weichert et al. [15] investigated the geometric shape of the extraction hole, but indicated that the shape of extraction hole has an insignificant effect on the re-circulated mass flow. Previous results [21] validated that the bleed direction and the geometry of port shown in Fig.4 is better. It can extend the stall margin of an isolated-rotor axial flow compressor by 6.12% combined with the Coanda injector. This study will take full advantage of the optimized bleed port to do the further research.

Through the reasonable optimization and design for the

self-recirculating device, the effect of different parameters, including both pressure rise coefficient and efficiency, on the performance is the focus that needs to be investigated in the following parts. The actual bleed mass flow and momentum are also measured.

Figure 3: Structure of the injector

Figure 4: Structure of the bleed port

4. STEADY CHARACTERISTIC PARAMETERS The compressor performance is evaluated by using pressure rise and efficiency characteristics. The pressure rise coefficient, , is expressed as(P , − P , )/(0.5U ). The flow coefficient, Φ, is defined as V , /U . P , and P , are the ambient pressure and the outlet static pressure on the casing wall, respectively. V , is the inlet average velocity. U is the tangential velocity at the mid-span, and represents the environmental density. SMI is defined as the percent reduction in stalling flow coefficient under three cases of the stability-enhancing technology, namely, self-recirculating injection, tip air injection, and outlet bleed air relative to the smooth wall casing (SC), i.e.,(Φ −Φ )/Φ × 100%. The efficiency, , is calculated from the shaft torque and the change in total pressure across the stage as

follows: /(− 1)R T∗ P , /P , − 1 ( )/m/(2NM/60), where is the adiabatic index, R is the ideal gas constant, T∗ is the inlet total temperature, P , is the outlet total pressure, m is the inlet mass flow, N is the shaft speed, and M is the torque which is measured using a high-precision torque meter. Characteristics are measured by continuously recording the inlet and casing pressures as the compressor is slowly throttled into stall. The efficiency and pressure rise characteristics for each test case are repeated several times to ensure representative results. For each parameter that is studied, back-to-back tests are performed, and the smooth wall case is measured at the beginning and end of each set of measurements. In this way, results that are reliably comparative can be ensured.

The repetitive works have been conducted before the dynamic measurement to ensure the measuring accuracy of unsteady flow field. Fig.5 shows that the characteristic lines have a better repetition at the stall point, in which the measuring error of the steady transducer is less than 0.5%. The flow coefficient and the pressure rise coefficient at the stall point are 0.459 and 0.619 respectively.

On the basis of good repetition, three flow coefficient conditions

are selected to conduct the detailed measurements of the flow field and defined them as large-flow coefficient (Φ=0.55), low-flow coefficient (Φ=0.50), and near-stall point (Φ=0.48).

YX

Z

X Velocity

2520151050

A

C R

A C - Circumferential direction A - Axial direction R - Radial direction

LE

0.40 0.45 0.50 0.55 0.60

0.2

0.3

0.4

0.5

0.6

0.7

Exp_001 Exp_002 Exp_003

Stall point

deep

ly s

tall

(a) -Φ

0.40 0.45 0.50 0.55 0.60

0.6

0.7

0.8

0.9

1.0

Exp_001 Exp_002 Exp_003

de

sign

poi

nt

(b) -Φ

Figure 5: Characteristic line with SC

Characteristic curves with different cases

During the comparatively experimental test with different cases, the bleeding port under outlet bleed air is the same as that under self-recirculating injection. Nevertheless, for the external tip air injection, except the consistence of injector, the injected mass flow (injected momentum) must also keep consistent with the maximum bleeding mass flow under the self-recirculating injection that it is generated at the stall point. When carrying out the external tip air injection, the injected air is supplied from an external compression system, and the injected mass flow for each injector is measured by using an additional mass flow meter. Under the above preconditions, the following comparison of characteristic line is meaningful. The experimental results show that the maximum injected mass flow ratio with self-recirculating injection is approximately 0.48% (injected mass flow divided by design mass flow) at the stall point, and its corresponding injected momentum ratio is approximately 0.7%. Back to the original optimization guideline, the re-circulated momentum ratio should be located in the section of macro injection. Thus, the SMI as a function of injected momentum ratio that is obtained by using the external tip air injection should be given first as shown in Fig.6. One can observe that the injected momentum ratio under self-recirculating injection marked by the red solid arrow is located in the area of macro injection. It achieves the aim of the optimization design in terms of the recirculating structure compared to the experiment with no optimization proposed by Li et al. [22].

0.1 1 10

5

10

15

20

25

Recirculated momentum ratio

Macro injection

Sta

ll m

argi

n im

prov

emen

t(%

)

Injected momentum ratio(%)

Micro injection

Figure 6: SMI versus injected momentum ratio in the single-stage compressor

Fig.7 shows the comparative characteristic line of pressure rise and efficiency under self-recirculating injection compared with SC, outlet bleed air, and external tip air injection. The abscissa is flow coefficient Φ, and the ordinate represents the pressure rise coefficient and efficiency η. In Fig.7 (a), the flow coefficients at the stall point under the self-recirculating injection and external tip air injection are obviously less than that under SC, especially for self-recirculating injection; it can postpone the stalling to give rise to the best SMI. However, the outlet bleed air has no effect on the stall point basically. Based on the definition of SMI at the beginning of this section, the SMI with different cases is given in Table 2.

0.35 0.40 0.45 0.50 0.55 0.60 0.65

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Smooth casing Self-recirculating injection Injection Bleed Bleed with injection

Extension range

(a) Ψ-Ф

0.35 0.40 0.45 0.50 0.55 0.600.6

0.7

0.8

0.9

1.0

d

esi

gn

poin

t

Smooth casing Self-recirculating injection Tip air injection Outlet bleed air Considering bleed mass flow

(b) -Ф

Figure 7: Characteristic curve line in single-stage axial flow compressor

Table 2 The SMI under different cases

Case SC Self-recirculating injection

External tip air injection

Outlet bleed air

ΦNS 0.459 0.396 0.403 0.465

SMI 0 13.67% 12.3% -0.97%

In Table 2, the SMI of self-recirculating injection is approximately 13.67%. The outlet bleed air and tip air injection can obtain 12.3% and -0.97% of SMI, respectively. These data indicate that the self-recirculating injection can generate the best SMI among all the three cases. Besides its good stability-enhancing capability, the impact on the efficiency is also crucial for future application. Some researchers have indicated that the efficiency of a compressor is affected by a penalty with their self-designed self-regulating injection (see Weichert [15] and Guinet [16]). The effect on the efficiency with the proposed self-recirculating injection in this study should also be investigated.

In Fig.7 (b), the three different cases do not reduce the

compressor efficiency basically. In the definition of efficiency,m is calculated by using the inlet static pressure that is measured in the inlet duct upstream of the injector, as shown in Fig.1. Thus, the impact on m in the outlet duct downstream of rotor blade row under the separate tip air injection and outlet bleed is not considered. If the outlet bleed mass flow and consumable energy of external air source that supplies the injected mass flow are considered, the efficiency under outlet bleed air and external tip air injection will decrease rapidly. Owing to the difficult evaluation of the exact energy loss of external air source, the real efficiency line with external tip air injection that considers the external loss is not given in this paper. Unlike the external tip air injection, the bleed mass flow can be measured and calculated exactly. Its corresponding efficiency line is marked by the green circle shown in Fig.7 (b) and has a sharp drop compared with other efficiency lines. All the above-mentioned results show that the self-recirculating injection not only can enhance the stability but also can increase the efficiency of the compressor. However, the following question is why the self-recirculating injection can postpone the stall point so late. The outlet bleed air cannot delay the occurrence of stall in the proposed single-stage compressor; it doesn’t influence the flow field that only the tip air injection is selected to make a comparative analysis of the stability-enhancing mechanism in the following parts. For the tip air injection, judged by the demarcation momentum ratio in the same compressor proposed by Li [7], the injected momentum ratio in this study is macro injection, which both acts on the tip leakage flow and main flow. By contrast, for the self-recirculating injection, its influencing mechanism on the flow field is sorely lacking to explain the inner reason of its stability-enhancing capability experimentally. In this study, considering the difficult measurement in the gap between rotor blade and stator blade that it is less than 5mm, only the detailed measurements in the casing are performed under different cases. The analysis of the effect on the internal flow field is conducted in the following section.

5. STABILITY-ENHANCING MECHANISM OF SELF-RECIRCULATING INJECTION Trajectory of Tip leakage flow: Previous researches indicate that tip leakage flow plays an important role in the throttling process under a certain tip clearance (see Du et al. [23] and Vo et al. [24]). They found the forward movement of the interface between the tip leakage flow (TLF) and incoming main stream flow (MF) as the throttle is closed and the spillage of the interface will induce the spike-type stall inception. In order to express the interface, the velocity vector distribution, root mean square of the pressure (RMSP) contour, and

entropy distribution are adopted in the numerical simulation. However, only the dynamic pressure can be measured in the tip gap in the experiment, thus, the RMSP contour is selected to investigate the trajectory of the tip leakage flow in this study.

Fig.8 depicts the RMSP contour under different flow coefficients. The RMSP contours obtained by the first phase-locking time series of each sensor with the help of a key phasor, which is a Hall sensor that produces a pulse for each rotor revolution, the phase-averaged mean, and the RMS values can be calculated. In early research, Du’s numerical results [25] indicated that a high RMS can represent the oscillation of the TLF, and the trace of the high-RMS regions can reflect the TLF/MF interface similar to what the entropy contours can do. Based on these results, the RMSP contour at the flow coefficient of 0.48 in Fig.8 is considered an example to give a general description of the phase-locked unsteady feature of the tip leakage flow. The red region (high RMS) representing the high fluctuation indicates the unsteady tip leakage flow. The blue region (low RMS) near the blade LE represents the steady incoming main flow. Thus, an “interface” between these two flows is formed, which is marked as red dashed arrow in the plot. Comparison of the three plots in Fig.8 indicates that the red region only appears in the low-flow coefficient conditions (Φ=0.50), and the distance (marked by the double-headed arrow) from the interface in the pressure side to the LE (marked by the red dashed line) becomes shorter during the throttling. Therefore, the forward movement of the TLF/MF interface in the throttling process is also a typical feature in the test compressor. The interface even aligns to the LE at the near stall point (Φ=0.48). In addition, the high RMS (i.e., the red color region) can also represent a stronger TLF oscillation in the tip region. This TLF oscillation becomes stronger with a gradual decrease in flow coefficient, particularly the TLF oscillation located at the 26% Cax is strongest at the near stall point. But it lacks a strong persuasion only according to the shade of the red region. The power spectrum density (PSD) can generally be used to display the unsteady fluctuation of the tip leakage flow because the PSD can distinguish the frequency band distribution of different signals in all the time-resolved data along the blade chord [20-23].

(a) Φ=0.55

(b) Φ=0.50

(c) Φ=0.48

Figure 8. RMSP contours with SC

Fig.9 depicts the RMSP contours under different cases at the near-stall point with SC (Φ=0.48). Compared with the RMSP contour under SC; the TLF/MF interface marked by the red arrow under self-recirculating injection and external tip air injection moves rearward towards the blade TE, as shown by the distance from the red circle to the blade LE. From the analytical results for Fig.8, we can conclude that the self-recirculating injection and tip air injection can delay the forward movement of TLF/MF interface towards the blade LE to contribute to SMI. In addition to the effects on the TLF/MF interface with the stability-enhancing technology, the strength of the red region becomes weaker under the influence of self-recirculating injection and tip air injection than that under SC. The red region in the RMSP contour can represent the unsteady fluctuation of the tip leakage flow in the previous section. This phenomenon implies that the self-recirculating injection and tip air injection may influence the unsteady fluctuation of the tip leakage flow. In order to further verify this result, the frequency spectrum analysis on the basis of PSD method is presented to assess the time-resolved data along the blade chord under different cases of stability-enhancing technology.

(a) SC

（b）Self-recirculating injection

(c) Tip air injection

Figure 9: RMSP contours with different cases

Unsteadiness of tip leakage flow: Fig.10 plots the PSD distributions from the LE to the TE of the rotor blade at different flow coefficients under SC. The abscissa is the normalized frequency relative to the BPF, and the vertical ordinate represents the power spectrum magnitude (Pa2/Hz). Besides the BPF band with a normalized frequency of 1, there exist another two frequency bands, one band is located near 0.2BPF, and another band is located from 0.24BPF to 0.56BPF. In the throttling process, the change of the first band, whose is centered at around 0.2BPF, is not regular. However, the amplitude of the signature frequency band marked by the rectangular box presents an increasing trend with the throttling of the compressor. Tong [6] experimentally illustrated that the signature frequency band is related to unsteadiness of tip leakage flow (UTLF), which is centered at approximately 0.4BPF. According to the PSD profiles, the amplitude of the signature frequency band will arise

obviously near the stall point (Φ=0.50) compared with the large-flow coefficient (Φ=0.55). Especially for the position located from CH3 to CH4 in the figure, the obvious increasing trend of the amplitude in the throttling process implies that this region is more sensitive to the unsteady tip leakage flow than other positions. The changing process of the amplitude forms a good corresponding relationship with unsteady oscillation (high RMS), i.e., the gradual deepening in the red region during the throttling process in Fig.8. The results not only validate that the high-RMS region can represent the unsteady fluctuation of TLF but also can be used to distinguish the different influencing mechanisms on the unsteady fluctuating characteristic of TLF under different types of stability-enhancing technology further.

Figure 10: PSD distribution with different flow coefficients under SC

Fig.11 gives the comparative PSD profiles under the different cases of stability-enhancing technology at a flow coefficientof0.5, which is near the stall limit with SC. Both self-recirculating injection and tip air injection can decrease the amplitude of the signature frequency band compared with SC. Therefore, the application of self-recirculating injection and tip air injection can stabilize the compressor through weakening the self-induced UTLF in the tip gap region.

Figure 11: PSD distributions under different cases

Circumferential propagation: According to the above-mentioned PSD profiles, the amplitude of the signature frequency band obviously rise near the stall point (Φ=0.48) as compared with the large flow coefficient (Φ=0.55), especially for the location at 26% Cax. Therefore, the time-resolved pressure signals located at 26% Cax are used to perform the individual PSD analysis in the throttling process as shown in Fig.10. Note that there exist two signature frequency bands named as Band 1 and Band 2 in addition to the BPF.

0 0.5 10

2000

4000Axial location=-0.2073

0 0.5 10

2000

4000Axial location=-0.0516

0 0.5 10

2000

4000Axial location=0.1041

Am

plitu

de(P

a2 /Hz)

0 0.5 10

2000

4000Axial location=0.2600

0 0.5 10

2000

4000Axial location=0.4155

0 0.5 10

2000

4000Axial location=0.7269

Normalized Frequency(/BPF)

Large flow coeffcient Low flow coefficient Near stall point

0 0.5 10

2000

4000Axial location=-0.2073

0 0.5 10

2000

4000Axial location=-0.0516

0 0.5 10

2000

4000Axial location=0.1041

Am

plitu

de(P

a2 /Hz)

0 0.5 10

2000

4000Axial location=0.2600

0 0.5 10

2000

4000Axial location=0.4155

0 0.5 10

2000

4000Axial location=0.7269

Normalized Frequency(/BPF)

Smooth casing Self-recirculating injection Tip air injection

CH1

CH2

CH3

CH4

CH5

CH6

CH1

CH2

CH3

CH4

CH5

CH6

Bandis als0.56 relatithat tflow throttthat tornathe ctraceinducthereincep

Hspeedof thilot ofto theal. [2propadisturet al.low hal. [3[32],

Fsignacondslopethe boccurpropahighethe cat a seven

d 1 is verified toso described inBPF, was num

ive coordinate bthe proposed sicoefficient; its

tling process. Inthe forward s

adolike vorticesasing, these tors of casing stce the blockageby indicating ption in terms o

However, in cond is around 60%is propagating pf attention. Thee stall inception29] reported “roagation speed rbances are not [30]. März et hub-tip ratio an30]). Low press

although the

Fig.13 depicts tal with differentition of ten rev

e of the color blade tip speed. rs to propagatagating speed oer than the bladircumferential speed of arounabsent.

2

o reflect the UTn this study. Bamerically confirmby Deng [26] anignature freques amplitude win addition, Lin spilled TLF ans with a certain rnado-like vortitatic pressure ge/separate floa relationship f the flow struc

nsideration of %-70% of the bphenomenon ise propagation on has been repootating instabilto those in thet observed in val. [31] also sh

nd operating at ure disturbancespacing betwee

Figure12:

the circumferet frequency banvolutions beforband representsOnly the pressu

te below the bof the signal wide tip speed; itspropagation of

nd 60% of blad

00

1000

2000

3000

4000

Am

plit

ude(

Pa2 /H

z)

TLF in the prevand 2, which ismed to represend Geng [20]. T

ency band onlyll accordingly et al. [27] gavend the reversetip clearance. T

ices appear to bmeasurements.

ow in the probetween the T

cture.

stall inception blade tip speed [s another conceof the unsteadyrted in some litity (RI)” of sime present work

very similar exphowed similar “partial speed (

es are also obseen events was

PSD distributio

ential propagatinds, which is mre stall. The abs the propagatinure signal with blade tip speedith other freque

s relationship wf the stall inceptde tip speed is

0.2No

Filter 1

vious results [6 centered at ar

ent the UTLF inThe figure indi

y appears at theincrease during

e the stalling proe flow evolve To the observe

be spikes in the The vortices

oposed compreTLF and spike

whose propag[28], the origin

ern issue; it cauy TLF that is reteratures. Mailamilar frequencyk. However, simperiments by W“RI” in a fan wagain, see Wisl

erved by Inoue larger. Their g

on of the dynam

ion of the premeasured at the bsolute value ong speed relativ0.24 BPF-0.66d before stall. ency bands is m

with the formatiotion that propagextremely sma

0.4ormalized freq

Filter

Band1

] and round n the icates e low g the ocess

into rs on time will

essor, stall

gating nation sed a

elated ach et y and milar

Wisler with a ler et et al.

group

descthe Youthe pthe d

Iinveprocspeedistuconfpropflow(namreflefilteBPFFiltethe peachdiffestopthe chossign

mic pressure loc

ssure flow

of the ve to BPF The

much on of gates

all, or

0.6quency(/Blad

2

Band 2

cribed their dis“blue holes” w

ung et al. [33], wpressure signaldisturbance is n

In this study, thestigated againcessing the timeed of the unsurbances with ofirm whether opagation. Givenw, the signal omed as Band ects the self-iner will be designF and 0.7 BPF-er 2, and Filter processing signh other, when derence between

p band and the sstart of the secsen as 0.02BPF

nals with differe

cated at 26% Ca

0.8de passing freq

Bla

de p

assa

ge

0 20

2

4

Bla

de p

assa

ge

0 20

2

4

Filter 3

turbances as thwith a similar dowho defined the. Geng et al. [3

numerically rela

he circumferentin through dee-resolved presssteady TLF dother frequencyother disturbann that the zero-fof the frequenc1) and 0.50 Bduced UTLF aned piecewise:0.9 BPF. They 3 marked by th

nal with differedesigning the sn the frequencystart of the passcond stop bandF. Thus, it can aent frequency ba

ax in the throttli

(a) Filter with 0

(b) Filter with 0

1quency)

Time(Bla4 6 8

Time(Bla4 6 8

BPF

he “mild stall”.ominant frequee phenomenon

34] indicated thated to the TLF.

ial propagationesigning diffesure data to obs

during the throy bands should

nces also existsfrequency signacy band with

BPF-0.65 BPF as shown in Fi 0.08 BPF-0.2 are respectivel

he red dashed bnt frequency batructure of the

y at the edge ofs band, the freq

d and the end oavoid the overlands.

ing process

0.08BPF-0.2BPF

0.24BPF-0.66BPF

1.2

=0.60=0.58=0.56=0.55=0.53=0.51=0.50=0.48=0.46

ade passings)10 12 14

ade passings)10 12 14

. The propagatiency was report

as the irregularat the propagat.

n of the disturbaerent filters. serve the propagottling process

d also be analyzs the circumferal represents av0.24 BPF-0.50(named as Bag.12. The bandBPF, 0.24 BPF

ly named as Fibox. Considerinand cannot inflband-pass filte

f the start of thquency at the edof the pass banlap between adj

F

086531086

16 18 20

16 18 20

ion of ted by rity of tion of

ance is When gating s, the zed to rential verage 0 BPF and 2) d-pass F-0.66 lter 1,

ng that luence er, the

he first dge of d was

djacent

(c) Filter with 0.7BPF-0.9BPF

Figure 13: Circumferential propagation of pressure signal with different frequency bands

Fig.14 depicts the evolutionary process of the circumferential propagation of the unsteady tip leakage flow in the process of to be near stall; the contours are plotted based on the band-pass filter with 0.24 BPF-0.66 BPF. The previous results [34] indicated that the absolute value of the slope of the color band represents the relative propagating speed of the self-induced UTLF. For example, if the absolute value of the slope of the bright color band is 1, the propagating speed is equal to the blade tip speed. When the flow conditions (Φ=0.58) is far away from the stall limits, no circumferential propagation exists based on the slope of the color band (the absolute value of slope is 1). Only when the flow coefficient decreases to 0.53, the slight propagation begins to appear; the propagating phenomenon becomes increasingly evident with further throttling. Additionally, the red region with a negative slope will become more and more evenly spaced until the stall occurs as shown in Fig.14 (e). The propagating speed in Fig.10 (c) was found to be 43% of the blade tip speed (angular frequency is approximately 40Hz), and the propagating speed at the near stall point can rise to 47% of the blade tip speed shown in Fig.14 (d). Thus, the detailed investigations of the circumferential propagation induced by unsteady TLF before the emergence of stall inception present the relationship between the TLF and spike-type stall inception in terms of circumferential propagation.

(a) Ф=0.58

(b) Ф=0.55

(c) Ф=0.53

(d) Ф=0.50

(e) Ф=0.46

Figure 14：Circumferential propagation of TLF with different flow coefficients under SC

Fig.15 depicts the propagating speed of the disturbance in the

process to be near stall with different cases, including SC, external tip air injection, and self-recirculating injection. The abscissa is the flow coefficient Φ; the ordinate represents the propagating speed (relative to the blade tip speed). The data processing of the propagating speed is based on the evolutionary process of the circumferential propagation in Fig.14. The absolute value of the slope of the red band region can represent the propagating speed of the fluctuating disturbance induced by the tip leakage flow. The value of speed can be collected by calculating the slope. The evolutionary process of propagating speed under SC in Fig.15 is taken for an example. The circumferential propagation only occurs at the low flow coefficient (Φ=0.53 in the tested compressor). As the flow coefficient decreases from 0.53 to the stall point, the propagating speed accordingly increases from 0.41 to 0.61 that corresponds to the propagating speed of the spike-type stall inception. Compared with the SC, the propagating speed dominated by the self-induced TLF becomes slower under the condition of self-recirculating injection and external tip air injection. Base on the analytical results of the circumferentially propagating process under SC, the reduction in propagating speed implies that the flow field in the compressor shifts farther away from the stall point and the stability is improved. For example, the dashed box in Fig.15 shows the propagating speed with different cases at the near stall point under the SC (Φ=0.48); the relative reduction of the propagating speed as compared to SC with self-recirculating injection is more than that with external tip air injection, thereby causing the SMI to increase a little more with self-recirculating injection as revealed by the characteristic lines in Fig.7. Additionally, in the comparative tendency curve of the propagating speed in the throttling process under SC, self-recirculating injection and external injection in Fig.15, the propagating speed of the disturbance induced by the unsteady TLF at their respective near stall point is near 0.61. It implies that the stall characteristics under these different cases are the same and are all related to the tip leakage flow. This result reaches agreement with Li’s previous conclusion [7] that the micro injection and macro injection could not change the stall route of the compressor compared with SC.

0.40 0.42 0.44 0.46 0.48 0.50 0.52 0.540.30

0.35

0.40

0.45

0.50

0.55

0.60

0.65

Pro

paga

ting

spe

ed(/

blad

e ti

p sp

eed)

Smooth casing Self-recirculating injection Tip air injection

stall point

Figure 15: Circumferentially propagating speed with different cases

Time(Blade passings)

Bla

de p

assa

ge

0 2 4 6 8 10 12 14 16 18 200

2

4

Time(Blade passings)

Bla

de p

assa

ge

0 2 4 6 8 10 12 14 16 18 200

2

4

-100

0

100

200

Time(Blade passings)

Bla

de p

assa

ge

0 2 4 6 8 10 12 14 16 18 200

2

4

-100

0

100

200

Time(Blade passings)

Bla

de p

assa

ge

0 2 4 6 8 10 12 14 16 18 200

2

4

-100

0

100

200

Time(Blade passings)

Bla

de p

assa

ge

0 2 4 6 8 10 12 14 16 18 200

2

4

-100

0

100

200

Time(Blade passings)

Bla

de p

assa

ge

0 2 4 6 8 10 12 14 16 18 200

2

4

-100

0

100

200

In this section, the characteristic performance and stability-enhancing mechanism under self-recirculating injection installed in single-stage axial flow compressor is detailedly described. Self-recirculating injection can obtain SMI by 13.67% with no efficiency penalty compared with SC, external tip air injection, and outlet bleed air. It can postpone the occurrence of stall by suppressing the forward movement of the TLF/MF interface, weakening the unsteadiness of tip leakage flow, and slowing the circumferentially propagating speed of the disturbance originated from the unsteady tip leakage flow that induces the propagation of the spike-type stall inception.

6. CONCLUSION

A detailedly experimental study about the self-recirculating injection in a single-stage axial flow compressor is conducted. In this study, the steady response to pressure rise and efficiency, and the unsteady response to trajectory and fluctuation of tip leakage flow and circumferential propagation with self-recirculating injection compared with outlet bleed and external tip air injection are investigated. The following conclusions can be drawn:

1. The well-designed and optimized self-recirculating injection,

including the Coanda injector and bleed port, that injects air over the blade tip at the LE and bleeds the air from the casing downstream of the last blade row can improve the stall margin by 13.67% in single-stage compressor on the premise of no efficiency penalty. Due to the comprehensive advantages in terms of balancing the effect on stability and efficiency, this approach is better than the separate outlet bleed, external tip air injection with the same injected momentum ratio, and other traditional casing treatment that must consider the efficiency loss in application.

2. The unsteady response to the tip leakage flow demonstrates

that the self-recirculating injection can suppress the self-induced UTLF and postpone the forward movement of the TLF/MF interface similar to the separate external tip air injection. The influencing capability on the tip leakage flow under self-recirculating injection is one of the factors to delay the occurrence of the stall compared with SC.

3. The unsteady response to circumferentially propagating

characteristic induced by the tip leakage flow indicates that although the final circumferentially propagating speed of stall inception are basically the same (approximate 0.6 relative to the blade tip speed) under self-recirculating injection, external tip air injection, and SC, the self-recirculating injection can slow down the increasing speed of the circumferentially propagating speed in the process of to be near stall compared with SC. It is another factor to promote stall margin extension.

This study gives a transparent explanation about the inner

stability-enhancing mechanisms of the self-recirculating injection with no efficiency penalty in the single-stage axial compressor and answers the questions about how and why the self-recirculating injection can enhance stability clearly. The self-recirculating injection couples the advantages of the effects on the stability and efficiency, and it can promote a more valuable application prospect than the separate outlet bleed air and tip air injection. This study is helpful to understand the response mechanism of self-recirculating injection in the entire stalling process and to give some design guideline in stall instability control.

ACKNOWLEDGMENT

The authors are thankful for the supports of the National Science Foundation of China with project No.51306178 and No.51676183. The contributions of Prof. Jingyi Chen and Prof.

Chaoqun Nie for insightful discussion and the contributions of Mr. Le Liu in carrying out the experiment are also gratefully acknowledged.

REFERENCE

[1] Hathaway M.D, 2007, “Passive endwall treatments for enhancing stability” NASA Report TM-2007-214409.

[2] Wilde, G.L., “Improvements in or Relating to Gas Turbines,” British Patent (701,576), filed, June 28, 1950.

[3] Day, I. J., 1993, “Active Suppression of Rotating Stall and Surge in Axial Compressors”. ASME J.Turbomachinery, 1993, 115(3): 40-47.

[4] Weigl, H.J., Paduano, J.D., Frechette, L.G., et al., 1998. “Active Stabilization of Rotating Stall and Surge in a Transonic Single Stage Axial Compressor”. ASME J.Turbomachinery, 1998, 120(4): 625-636.

[5] C. M. Van Schalkwyk, J. D. Paduano, E. M. Greitzer, et al., 1998. “Active Stabilization of Axial Compressors with Circumferential Inlet Distortion”. ASME J.Turbomachinery, 1998, 120(3): 431–439.

[6] Tong, Z., Lin, F., Chen, J., et al., 2007. “The Self-Induced Unsteadiness of Tip Leakage Vortex and its Effect on Compressor Stall Inception,” ASME Paper,2007, GT2007-27010.

[7] Li, J., Lin, F., Tong, Z., et al., 2014. “The Dual Mechanisms and Implementations of Stability Enhancement with Discrete Tip Injection in Axial Flow Compressor”. ASME J.Turbomachinery, 2014, 137(3):031010-1-10.

[8] Suder, K.L., Hathaway, M.D., Thorp, S.A., et al., 2001. “Compressor Stability Enhancement Using Discrete Tip Injection”. ASME J. Turbomachinery, 2001, 123(1): 14-23.

[9] Kefalakis, M., and Papailiou, K.D., 2001. “Detailed Measurement on an Axial Compressor Stage with Application of Discrete Tip Injection for Increasing the Surge Margin”. ECT Paper, 2001-044.

[10] Nie, C., Xu, G., Cheng, X., et al., 2002, “Micro Air Injection and Its Unsteady Response in a Low-Speed Axial Compressor”. ASME J. Turbomachinery, 2002, 124(4): 572-579.

[11] Freeman, C., Wilson, A.G., Day, I.J., et al., 1998, “Experiments in Active Control of Stall on an Aeroengine Gas Turbine,” ASME J. Turbomachinery, 1998, 120(4): 637-647.

[12] Leinhos, D.C., Scheidler, S.G., Fottner, L., et al., 2002, “Experiments in Active Control of a Twin-Spool Turbofan Engine”. ASME Paper, 2002, GT2002-30002.

[13] Hathaway, M.D., 2002. “Self-recirculating Casing Treatment Concept for Enhance Compressor Performance”. ASME Paper, 2002, GT2002-30368.

[14] Strazisar, A.J., Bright, M.M., Thorp, S., et al., 2004. “Compressor Stall Control through Endwall Recirculation”. ASME Paper, 2004, GT2004 -54295.

[15] Stephanie Weichert, Ivor Day, Chris Freeman, 2011. “Self-regulation Casing Treatment for Axial Compressor Stability Enhancement”. ASME Paper, 2011, GT2011-46042.

[16] Guinet, C., Streit, J., Kau, H., et al., 2014. “Tip gap variation on a transonic rotor in the presence of tip blowing”. ASME Paper, 2014, GT2014-25042.

[17] André Inzenhofer, Cyril Guinet, Andreas Hupfer, Bernd Becker, Patrick Grothe, Frank Heinichen. “Difference in the Working Principle of Axial Slot and Tip Blowing Casing Treatments”. ASME Paper, GT2016-56966

[18] Grosvenor, A. D., Rixon, G. S.,Sailer, L. M., et al., 2015. “High Resolution RANS Nonlinear Harmonic Study of Stage 67 Tip Injection Physics”. ASME J. Turbomachinery, 2015, 137(5): 051005-1-13.

[19] I.J.Day, 2015. “ Stall, Surge and 75 years of research”. ASME

Paper, 2015, GT2015-44109. [20] Geng S., Zhang H., Chen J., Huang W., 2007. “Numerical

Study on the Unsteady Response of Tip Leakage Flow Unsteadiness to Discrete Micro Tip Injection in a Low-speed Isolated Compressor,” ASME Paper, 2007, GT2007-27729.

[21] Li J., Du J., Nan X., et al., 2016. “Coupling Stability-enhancing Mechanism with Compact Self-recirculating Injection in an Axial Flow Compressor”. Proc IMechE Part A: J. Power and Energy. 2016: 230(7) 696–708

[22] Li J., 2013. “Stability Enhancement Technology of Tip Air Injection in Axial Flow Compressor-Mechanism and Intelligent control”. 2013, Ph.D dissertation in Chinese Academy of Sciences.

[23] Du, J., Lin, F., Chen, J.Y., et al., 2013. “Flow Structures in the Tip Region for a Transonic Compressor Rotor”. ASME J. Turbomachinery, 2013, 135(1):013012-1.

[24] Vo, H.D., Tan, C.S. and Greitzer, E.M., 2008. “Criteria for Spike Initiated Stall”. ASME J. Turbomachinery, 2008, Vol. 130(1): 011023.

[25] Du J., Liu L., Nan X., et al., 2013, “The Dynamics of Prestall Process in an Axial Low-Speed Compressor with Single Circumferential Casing Groove”. ASME Paper, 2013, GT2013-95432.

[26] Deng X., Zhang H., Chen J., et al., 2005. “Unsteady Tip Clearance Flow in a Low-Speed Axial Compressor Rotor with Upstream and Downstream Stators”. ASME Paper, 2005, GT2005-68571

[27] Lin, F., Zhang, J. X., Chen, J. Y., et al., 2008, “Flow Structure of Short Length Scale Disturbance on an Axial Compressor”. AIAA J. Propul. Power, 2008, 24(6):1301–1308.

[28] Pullan, G., Young, A.M., Day, I.J., et al., 2012. “Origins and Structure of Spike-type Rotating Stall”. ASME Paper, 2012, GT2012-68707.

[29] Mailach, R., Lehmann, I., Vogeler, K., 2001. “Rotating Instabilities in an Axial Compressor Originating from the Fluctuating Blade Tip Vortex”. ASME J. Turbomachinery, 2001,123 (3): 453-463.

[30] Wisler, D. C., Beacher, B. F., Shin, H-W., 2002.“Effects of Loading and Clearance Variation on Tip Vortex and Endwall Blockage”. In Proceedings of the 9th International Symposium on Transport Phenomena and Dynamics of Rotating Machinery, 2002, Feb 10-14, Honolulu, Hawaii.

[31] Marz, J., Hah, C., Neise, W., 2002. “An Experimental and Numerical Investigation into the Mechanisms of Rotating Instability”. ASME J. Turbomachinery, 2002, 124(3): 367-374.

[32] Inoue, M., Kuroumoru, M., Tanino, T., et al., 2001. “Comparative Studies on Short and Long Length-Scale Stall Cell Propagating in an Axial Compressor Rotor”. ASME J.Turbomachinery, 2001, 123(1):24-30.

[33] Young, A., Day, I. J., Pullan, G., 2011. “Stall Warning by Blade Pressure Signature Analysis”. ASME Paper, 2011, GT2011-45850.

[34] Geng S., Zhang X., Li J., et al., 2013. “Evolution of Pressure Signature Dominated by Unsteady Tip Leakage Flow”. ASME Paper, 2013, GT2013-94523.

APPENDIX NOMENCLATURE

Ф Flow coefficient Ψ Pressure rise coefficient η Efficiency P , Inlet static pressure P , Outlet static pressure

P , Outlet total pressure T ∗ Inlet total temperature V , Inlet average velocity U Tangential velocity at the mid-span ρ Environmental density Adiabatic index R Ideal gas constant N Shaft speed M Torque m Inlet mass flow

Abbreviations BPF Blade passing frequency Cax Axial chord LE Leading edge MF Main flow NS Near stall point

PSD Power spectrum density RMS Root mean square SMI Stall margin improvement SC Smooth wall casing TE Trailing edge

TLF Tip leakage flow UTLF Unsteadiness of tip leakage flow