The impeller stall inception in a high-pressure ratio centrifugal compressor with bleed slots was investigated by the experiments and unsteady numerical analysis. This study reported that the flow instability of the internal flow was first initiated around the leading edge of the splitter blades and finally resulted in tip-leakage flow instability at the leading edge of the main blades. The impeller stall consisted of a stall cell and rotated about 70% of the impeller rotational speed. Furthermore, a pressure disturbance occurs at the splitter blade leading edge before the main blade leading edge. The blockage within the passage of the suction side at the leading edge of the splitter blade was first increased because the leading-edge separation on the suction surface of the splitter blade was enlarged. The blockage within the passage of the pressure side was increased after that of the suction side stopped increasing. Therefore, the blockage development of both pressure and suction passage at the leading edge of the splitter blade was the main cause of the stall inception. Then, a longitudinal vortex was formed near the bleed slots and blocked the mainstream near tip side. Subsequently, the reverse flow region in the inducer was developed and extended to the impeller inlet. Finally, the flow angle to the main blade increased, the spillage occurred at a leading edge of a main blade, and the rotating stall was generated. The cause of impeller instability was the separation of the suction surface of splitter blade and the formation of a stagnation region on the pressure surface at the splitter leading edge.

The unsteady flow mechanism of rotating instability (RI) in a 1.5-stage axial compressor is investigated through experimental and numerical analyses. The total pressure increase characteristics of the rotor blades show that a kink point occurred in a case involving a wide clearance, whereas the RI occurred below the kink point. The RI appears in the power spectrum as a gentle hump whose frequency is approximately 20%–40% of the blade-passing frequency at the blade tips. The mode of the RI disturbance increases as the flow rate decreases, whereas the propagation velocity remains unchanged. This causes the RI frequency to increase. The double-phase-locked averaging method is used to visualize the flow field inside the rotor blade when the RI occurred. The results show that as the flow rate decreases, the spacing of the pressure patterns inside the rotor blade decreases and the number of RI modes increases. Results of numerical analysis show that the RI is formed by the repeated collision and separation of the tip leakage vortex onto the pressure surface of the adjacent blade. The fluctuation in the inflow angle of the vortex entering the adjacent blade changes the blade-tip loading. The vortex fluctuations propagate in the circumferential direction.

Unsteady pre-stall behavior in a centrifugal compressor with a vaned diffuser was investigated by experimental and numerical analysis. The pre-stall disturbances occurred at a slightly higher flow coefficient at the stall point in the diffuser region. Five disturbances occurred in the circumferential direction, and each rotated at approximately 1.7%N at this flow coefficient. Numerical analysis showed that five stall cells rotated at approximately 2.0%N within the diffuser passage. To understand this pre-stall phenomenon, we focused on the rotation mechanism and initiation process of the five-cell rotating stalls. Each of the five-cell stalls was found to rotate by the following mechanism. When the preceding low-velocity region moved to an adjacent passage, the high-velocity region was circumferentially pushed by the low-velocity area and reached the following passage. The incoming flow collided with the back flow around the throat area, and the flow bent at the diffuser inlet of the passage. Consequently, the incidence angle toward the adjacent passage increased, and a separation was induced at the leading edge of the succeeding diffuser vane. Subsequently, the mass flow rate of the succeeding passage started to decrease. These phenomena occurred sequentially, causing the five-cell stalls to rotate. Five stationary low-velocity regions that did not rotate were observed before the initiation of the five-cell rotating stalls. When the outlet mass flow rate decreased, a one-cell rotating stall appeared within the diffuser passage. It provided a low-energy fluid to the diffuser passages where the low-velocity regions existed. Subsequently, five low-lowvelocity regions were clearly formed, which started rotating according to the rotating mechanism explained above.

This paper presents the stall inception of fishtail pipe diffuser investigated by both unsteady and steady numerical analyses. Fishtail pipe diffuser decelerates the air flow from impeller and induces the flow to the combustor by changing the flow direction from radial to axial. Each unsteady numerical simulation with full annulus model and steady simulation with 1 passage model was performed to investigate the stall inception of the fishtail pipe diffuser attached on the transonic centrifugal compressor stage by commercial code for the design speed, where the flow at the diffuser inlet is supersonic. The numerical simulations were well validated by experimental results. The characteristics of speed-lines and stall points by the numerical simulations matched the surge line of the experimental result. In near stall condition, the separation flow was observed on the suction side wall at the bend section of diffuser. The separation extended upstream by further increase of back pressure of diffuser. When the separation affected the boundary layer on throat wall, diffuser stall occurred. To confirm the effect of the bend, the numerical analysis of the straight shape diffuser was also performed. Straight diffuser was confirmed more stable than the diffuser with bend. From these observations, it was concluded that stall inception was the suction side boundary layer separation promoted by the secondary flow due to the bend.

The internal flow structure and loss generation mechanism of a single-stage compressor during windmilling were investigated through experimental and computational analyses. The windmilling state occurs when the air flowing through an unlit engine drives the compressor rotor blades, similar to a turbine. This study focused on the effect of tip clearance size on the internal flow during windmilling operations. Therefore, detached eddy simulations were conducted at three tip clearance conditions: without clearance, design clearance, and wide clearance. Consequently, the total pressure loss decreased when the size of the tip clearance was expanded under windmilling conditions. In the windmilling state, separation on the pressure side due to the high negative incidence was generated, which was the main reason for the total pressure loss. When the size of the clearance increased, the size of the separation decreased because of the tip leakage flow. According to the detailed numerical results for the windmilling state, the leakage flow held the separation to the blade surface near the tip area. In addition, the tip leakage flow formed a blockage in the midchord, so the tip side of the separation vortex was moved to the midchord and formed a significant loss region. Thus, the tip leakage flow reduced the loss region, and the overall loss was reduced.

The interaction mechanism of impeller and diffuser stall in a centrifugal compressor with a vaneless diffuser was investigated by experimental and computational analyses. This study focuses on the effect of impeller stall on the diffuser stall behavior. Impeller stall rotated at 58% of the impeller rotational speed was generated inside the impeller. Two-cell diffuser stalls (with each of the cells rotating at 25%-30% of the impeller rotational speed) were generated inside the diffuser. The diffuser stall fluctuations were observed at 180° from the cutoff. The magnitudes of the diffuser stall fluctuations gradually increased near the volute tongue. The diffuser stall fluctuations were generated near both the shroud and hub sides. Finally, the diffuser stall cell vanishes when it passes the cutoff because mass flow recovery occurs. The numerical results revealed that boundary layer separation occurred near the hub side at 45°-90° of the diffuser because of the circumferential adverse pressure gradient. Subsequently, the low-velocity region discharged from the impeller caused by impeller stall merged into the boundary layer separation, which was generated near the hub side at 45°- 90°. Diffuser stall was initiated by the hub-side boundary layer separations, which were caused by the impeller stall. The diffuser stall cell was then further developed by the boundary layer separation accumulation and the induced low-velocity area. The boundary separation was further developed by merging the wake from the impeller stall passage.

The rotating stall in a centrifugal compressor with a vaneless diffuser was investigated both experimentally and numerically with focus on the effect of the internal flow field within the impeller on the diffuser stall. Through numerical analysis, the boundary layer separation at the impeller outlet was found to play an important role in the expansion and rotation processes of the diffuser stall. In particular, the expanded boundary layer separation near the hub side at the outlet of the main blade (M.B.) suction surface passage was considered to be the main cause of the expansion and rotation processes. A longitudinal vortex existed at the throat of the M.B. passage, and the mass flow rate in the M.B. passage was significantly reduced by the blockage effect. In addition, the longitudinal vortex induced the rolling up flow near the hub side at the impeller exit. Thus, the boundary layer separation expanded.

Abstract

The effect of diffuser width on the noise level of a centrifugal compressor equipped with vaneless diffusers was investigated through experiments and numerical techniques. In this regard, two diffusers with different widths were installed in the compressor. In case the wider diffuser was used, there was the bandwidth (800 to 900 Hz) noise and pressure fluctuation which was closed to 1/2BPF (700 Hz). This fluctuation occurred because the leakage vortices were of two different frequencies. Type 1 vortices are the ones that were discharged from the main blade and covered Passage 1. Type 2 vortices had the same length as Type 1, but they were discharged from the splitter blade. Type 3 vortices covered both Passages 1 and 2; these were discharged from the main blade. Type 3 vortices were generated predominantly at 270°, and their frequency was approximately 700 Hz at 1/2 BPF frequency. The former two types of vortices occurred occasionally, and their frequency was approximately 1400 Hz. Type 3 vortices grew with the wider diffuser because of the higher pressure-recovery, which caused the serous loading to the impeller blades. The broadband noise fluctuation occurred by the coexistence of the two frequency vortices.

The rotating mechanism of diffuser stall in a centrifugal compressor with a vaneless diffuser is investigated via experimental and computational analyses. Diffuser stall is generated as the mass flow rate decreases, and it rotates at 25%-30% of the impeller rotational speed. First, a diffuser stall cell emerges at 180° from the cutoff by the hub-side boundary layer separation. Subsequently, the diffuser stall cell develops further owing to boundary layer separation accumulation and an induced low-velocity area. The low-velocity region forms a blockage across the diffuser passage span. The diffuser stall cell expands owing to the boundary layer separations that occurred on the shroud and hub wall by turns. Finally, the diffuser stall cell vanishes when it passes the cutoff because mass flow recovery occurred. Furthermore, the static pressure ahead of the rotating stall decreases because of the merging of the impeller discharge flow and the reverse flow from the casing. Accordingly, a reverse flow occurred owing to the evolution of the separation vortex at the diffuser exit. In addition, the flow angle decreases by the merging of the impeller discharge flow and reverse flow from the casing. Therefore, boundary layer separations start occurring on the shroud and hub wall ahead of the stall cell. The rotating mechanism of the diffuser stall is induced by the reverse flow development and a decrease in the flow angle ahead of the stall cell.

The generation mechanism of a diffuser stall in a centrifugal compressor with a vaneless diffuser was investigated by experimental and computational analyses. The diffuser stall generated as the mass flow rate decreased. The diffuser stall cell rotated at 25-30 % of the impeller rotational speed, with diffuser stall fluctuations observed at 180° from the cutoff. The diffuser stall fluctuation magnitude gradually increased near the cutoff. According to the CFD analysis, the mass flow fluctuations at the diffuser exit showed a low mass flow region, rotating at approximately 25% of the impeller rotational speed. They began at 180° from the cutoff and developed as this region approached the cutoff. Therefore, the diffuser stall could be simulated by CFD analysis. First, the diffuser stall cell originated at 180° from the cutoff by interaction with boundary separation and impeller discharge vortex. Then, the diffuser stall cell further developed by boundary separation accumulation and the induced low velocity area The low velocity region formed a blockage across the diffuser passage span. The diffuser stall cell expanded due to boundary separation caused by a positive flow angle. Finally, the diffuser stall cell vanished when it passed the cutoff, because mass flow recovery occurred.

<title>Abstract</title>
The unsteady diffuser stall behavior in a centrifugal compressor with a vaneless diffuser was investigated by experimental and computational analyses. The diffuser stall generated as the mass flow rate decreased. The diffuser stall cell rotated at 25–30% of the impeller rotational speed, with diffuser stall fluctuations observed at 180° from the cutoff. The diffuser stall fluctuation magnitude gradually increased near the cutoff. Based on diffuser inlet velocity measurements, the diffuser stall fluctuations generated near both the shroud and hub sides, and the diffuser stall appeared at 180° and 240° from the cutoff. According to the CFD analysis, the mass flow fluctuations at the diffuser exit showed a low mass flow region, rotating at approximately 25% of the impeller rotational speed. They began at 180° from the cutoff and developed as this region approached the cutoff. Therefore, the diffuser stall could be simulated by CFD analysis. First, the diffuser stall cell originated at 180° from the cutoff by interaction with boundary separation and impeller discharge vortex. Then, the diffuser stall cell further developed by boundary separation accumulation and the induced low velocity area, located at the stall cell center. The low velocity region formed a blockage across the diffuser passage span. The diffuser stall cell expanded in the impeller rotational direction due to boundary separation caused by a positive flow angle. Finally, the diffuser stall cell vanished when it passed the cutoff, because mass flow recovery occurred.

In this study, we aim to elucidate the effect of a forward-swept rotor on the stall margin and flow field at the distorted inflow conditions. The rig in this research is a low-speed, single-stage axial compressor, which has two types of rotor blades: The radially stacked blade (Radial) and the forward-swept blade (Sweep). The distortion screen that circumferentially generates distorted inflow is located upstream of the rotor. The stall margin of Sweep was found to be larger than that of Radial. Sweep was considered to improve the flow fields at the distorted inflow conditions. From the results of the study, it was observed that Sweep suppressed the circumferential expansion of the high-load regions and the spike-type disturbances generated at the distorted sector. Therefore, Sweep enlarged the stall margin more than to Radial.

The work characteristics and loss-generation mechanism of a single-stage axial flow compressor in windmilling operation were investigated via experiments and computational fluid dynamics analyses. The windmilling state occurs when air flowing through an unlit engine drives the compressor rotor blades, similar to a turbine. This phenomenon applies mostly to aircraft engines, where it is caused by ram pressure. When the inlet flow coefficient is gradually increased in the design, the rotor blades gradually enters the windmilling operation from the tip toward the hub. This research has focused on two windmilling operations: free windmilling (FW) and highly loaded (HL) windmilling. In the case of FW, the net work performed by the rotor blades to the fluid is canceled out (zero), and the rotor is in an idle state. In the HL windmilling condition, the work performed to the rotor blades by the fluid increases, the compressor acts as a turbine, and power is generated. According to the detailed numerical results, the total-pressure loss under the free and HL windmilling conditions was mainly caused by three flow structures: (1) tip leakage flow from the suction surface (SS) to the pressure surface (PS) near the leading edge and that from the PS to the SS near the trailing edge; (2) the interaction of leading-edge separation vortices due to the highly negative incidence and the rotor leading-edge vortex; and (3) the boundary- layer separation near the hub wall. Surface-pressure measurement on a rotating rotor blade revealed that the distribution of the rotor operating mode existed not only in the spanwise direction but also in the chordwise direction under the windmilling operations. The turbine mode region was observed near the leading edge, while the compressor mode region was observed near the trailing edge, even in the HL windmilling condition. Therefore, the driving force of the windmilling was dominated not by the area of the turbine mode on the rotor surface but by the strength of the operating mode, i.e., the static-pressure difference between the SS and PS on the rotor. Finally, the unsteady flow field within blade-to-blades passages was investigated via an unsteady detached eddy simulation, and the differences in the loss-generation mechanism between the FW and HL windmilling conditions were examined.

<title>Abstract</title>
The relationship between the growth of the stall cell and variation in the surge behavior was experimentally investigated. The aim of this study was to reveal the effect of the stall cell on the surge behavior from the viewpoint of the inner flow structure. In the experiment, the unsteady compressor characteristics during the surge and rotating stall were obtained by using a precision pressure transducer and a one-dimensional single hotwire anemometer. Under the coexisting states of surge and rotating stall, various surge behaviors were observed by throttling the mass flow rate. When the flow rate was set such that the surge behavior switched, an irregular surge was observed. During the irregular cycle, two different cycles were selected randomly corresponding to the stall behavior. When the amplitude of the plenum pressure is relatively large among the measurement results, the absolute value of the time-change rate in the flow coefficient and the static pressure-rise coefficient tend to be high. This shows that the flow field during stable operation near the peak point of the unsteady characteristics changes rapidly. In this case, an auto-correlation function of the wall-pressure fluctuation data showed that the stall inception of the compressor was induced earlier in the large cycle compared with the case of the top cycle. When studying the growth of the stall cell during the stalling process of the large cycle, the wall-pressure fluctuation data showed that the stall cell rapidly grew by gathering more than one spike-type disturbance into one stall cell. In this case, the stall cell fully expanded along the circumferential direction and developed into a deep stall. Therefore, the key factors that determine the surge behavior are the sudden change in the flow field near the peak point of the unsteady characteristics and the rapid growth in the stall cell during the stalling process.

<title>Abstract</title>
The transient process of the rotating stall development in a centrifugal compressor with a vaned diffuser was investigated by experimental and numerical analyses. Previous studies show that a diffuser stall triggers a stage stall, which rotates through rotor and stator passages. The vortex evolution at the diffuser throat represents the key factor in diffuser stall development. The developed diffuser stall cell blocked the impeller exit, causing an impeller passage stall. This paper focused on two aspects regarding the transient process of the diffuser stall development. The first aspect is the process by which the vortex at the diffuser throat near the hub side, develops in the circumferential direction. Secondly, we investigated the mechanism of the diffuser stall expansion into impeller passages. The transient analysis of the diffuser stall development was conducted experimentally and numerically by closing the throttle valve abruptly. The hub side blockage was initiated near the cutoff by the strong adverse pressure gradient in the diffuser throat area. Therefore, the key factor in the diffuser stall evolution was the development of a throat blockage near the cutoff, obtained from both experimental and computational fluid dynamics results. Furthermore, the transient stall cell blocked the impeller passages and induced a hub side blockage at the throat of the impeller passages and the impeller leading edge separation. The pressure surface separation of the impeller at the trailing edge had a great impact on the development of the stall cell within impeller passages.

This paper describes in detailed flow field in a centrifugal compressor with a vaned diffuser at off design point. Especially, we conducted both the experimental and numerical analysis in order to investigate the evolution process of a diffuser stall. At the stall point, the diffuser stall was initiated and rotated near the shroud side in the vaneless space. Furthermore, the diffuser stall was developed to a stage stall cell, as the mass flow was decreased. The developed stall cell was rotated within both the impeller and diffuser passages. The evolution process of the diffuser stall had three stall forms. First, the diffuser stall was rotating near the shroud side. Then, the diffuser stall shifted to the hub side and moved into the impeller passages. Finally, a stage stall was generated. From computational fluid dynamics (CFD) analysis, a tornado-type vortex was generated first, near the hub side of the diffuser leading edge, when the diffuser stall was shifted to the hub side. Next, a throat area blockage was formed near the hub side because of the boundary layer separation in the vaneless space. Finally, the blockage within the diffuser passages expanded to the impeller passages and developed into a stage stall. From the pressure measurements along the impeller and diffuser passages, the magnitude of pressure fluctuation on the casing wall of the diffuser throat area also suddenly increased when the diffuser stall shifted to the hub side. Therefore, the evolution area of the diffuser stall was caused by the evolution of the blockage near the throat area of the diffuser passage.

Interaction between surge behavior and internal flow field under coexisting phenomena of surge and rotating stall was experimentally investigated. In the experiment, the tank pressure of the compressor during surge was measured to detect the effect of the back-pressure fluctuation on the change in the internal flow field. Furthermore, the rotating stall in the compressor was investigated to define the influence of an unsteady internal flow field change on the surge behavior. From the tank pressure measurements, the amplitude of the tank pressure fluctuation was found to vary depending on the cycle. A larger maximal value for the tank pressure fluctuation led to a higher flow rate where the stall inception occurred. This difference in the flow rate indicated that the stall was induced by a severe adverse pressure gradient in the compressor. Then, the absolute rate of change in the flow coefficient was increased by both a large decrease in the compressor back pressure and performance degradation from stalling. In a case where the rate of decline in the flow rate was large, the scale of the stall cell developed up to a deep stall according to the movement of the operating point. Thus, a large trajectory for the surge cycle was selected, where the unsteady operating point went through the deep stall region. This development in the scale of the stall cell suggested to be influenced by the instability of the inner flow field caused by the rapid change in the flow rate.

The evolution process of a diffuser rotating stall in a centrifugal compressor with a vaned diffuser was investigated by experimental and numerical analyses. From velocity measurements, it was found that the diffuser stall propagated near the shroud side in the vaneless space. As the mass flow decreased, a stage stall rotated within both the impeller and diffuser passages, instead of a diffuser stall. The evolution process of the diffuser stall had three stall forms. First, the diffuser stall, which was rotating on the shroud side, shifted to the hub side. Then, the diffuser stall moved into the impeller passages and evolved to a stage stall. From computational fluid dynamics (CFD) analysis, a tornado-type vortex was generated first, near the hub side of the diffuser leading edge, when the diffuser stall was shifted to the hub side. Next, a throat area blockage was formed near the hub side because of the boundary layer separation in the vaneless space. Finally, the blockage within the diffuser passages expanded to the impeller passages and developed into a stage stall. From the pressure measurements along the impeller and diffuser passages, the magnitude of pressure fluctuation on the casing wall of the diffuser throat area also suddenly increased when the diffuser stall shifted to the hub side. Therefore, the evolution area of the diffuser stall was caused by the evolution of the blockage near the throat area of the diffuser passage.

In this study, the unsteady behavior of a diffuser rotating stall in a centrifugal compressor with a vaned diffuser was investigated through experiments and numerical analyses. From the casing static pressure measurements, it was determined that the diffuser stall propagated at 25% of impeller rotational speed in the vaneless space. The numerical results revealed the presence of a typical vortical structure on the diffuser’s leading edge. Under partial flow condition, a tornado-type vortex was generated on the diffuser’s leading edge. Furthermore, a longitudinal vortex at the shroud/suction surface corner (i.e., leading edge vortex (LEV)) was induced by the rolling-up flow on the diffuser suction surface. As the velocity was decreased, the development of the tornado-type vortex and LEV forms a substantial flow blockage within the diffuser passages. Furthermore, the diffuser stall cell was caused by the systematic vortical structure which consisted of the tornado-type vortex, LEV, and vortex in the throat area of diffuser passages. In addition to this, the developed LEV interacted with the next diffuser leading edge and formed the throat area blockage with the passage of time. Then, the tornado-type vortex and LEV developed by the throat area blockage and diffuser stall cell, which was caused by the systematic vortical structure, propagated to the succeeding diffuser vane. Therefore, the diffuser stall in the centrifugal compressor was caused by the evolution of the tornado-type vortex and LEV.

The transition process from a diffuser rotating stall to a stage stall in a centrifugal compressor with a vaned diffuser was investigated by experimental and numerical analyses. From the velocity measurements, it was found that the rotating stall existed on the shroud side of the diffuser passage in the off-design flow condition. The numerical results revealed the typical vortical structure of the diffuser stall. The diffuser stall cell was caused by the systematic vortical structure which consisted of the tornado-type vortex, the longitudinal vortex at the shroud/suction surface corner (i.e., leading edge vortex (LEV)), and the vortex in the throat area of the diffuser passages. Furthermore, the stage stall, which rotated within both the impeller and diffuser passages, occurred instead of the diffuser stall as the mass flow rate was decreased. According to the velocity measurements at the diffuser inlet, the diffuser stall which rotated on the shroud side was shifted to the hub side. Then, the diffuser stall moved into the impeller passages and formed the stage stall. Therefore, the stage stall was caused by the development of the diffuser stall, which transferred from the shroud side to the hub side in the vaneless space and expanded to the impeller passages.

The characteristics of a rotating stall of an impeller and diffuser and the evolution of a vortex generated at the diffuser leading-edge (i.e., the leading-edge vortex (LEV)) in a centrifugal compressor were investigated by experiments and numerical analysis. The results of the experiments revealed that both the impeller and diffuser rotating stalls occurred at 55 and 25 Hz during off-design flow operation. For both, stall cells existed only on the shroud side of the flow passages, which is very close to the source location of the LEV. According to the CFD results, the LEV is made up of multiple vortices. The LEV is a combination of a separated vortex near the leading-edge and a longitudinal vortex generated by the extended tip-leakage flow from the impeller. Therefore, the LEV is generated by the accumulation of vorticity caused by the velocity gradient of the impeller discharge flow. In partial-flow operation, the spanwise extent and the position of the LEV origin are temporarily transmuted. The LEV develops with a drop in the velocity in the diffuser passage and forms a significant blockage within the diffuser passage. Therefore, the LEV may be regarded as being one of the causes of a diffuser stall in a centrifugal compressor.

The characteristics of a diffuser rotating stall and the evolution of a vortex generated on the diffuser leading edge (i.e., leading-edge vortex (LEV)) in a centrifugal compressor were investigated using experiments and numerical analyses. The experimental results showed that both impeller and diffuser rotating stalls occurred at 55 and 25 Hz during off-design flow operation. Both the stall cells existed only on the shroud side of the flow passages, which is in close proximity to the source location of the LEV. The numerical results showed that the LEV is a combination of a separated vortex near the leading edge and the extended tip-leakage flow from the impeller. In the partial flow operation, the LEV develops as the velocity decreases in the diffuser passages and forms a huge flow blockage within the diffuser passages. Therefore, the LEV may be considered to be one of the causes of diffuser stall in the centrifugal compressor.

The experimental and CFD analysis were conducted to investigate the relationship between stall and leading-edge vortex (LEV) in a centrifugal compressor with vaned diffuser. The LEV is distinct from the separating vortex of the diffuser leading-edge and passage vortex of the diffuser. The LEV is made up of two longitudinal vortices. It is produced by the accumulation of vortices caused by the velocity gradient of the impeller-discharge flow. According to the experimental results, both the impeller and diffuser rotating stalls occurred at 55 and 25 Hz during off-design flow operation. Both stall cells were existed only on the shroud side of the flow passages, which is very close to the source location of the LEV. Additionally, the intensity and scale of the diffuser stall fluctuation are much larger than those of the impeller stall fluctuation. Therefore, the unsteady behavior of the LEV may play an important role in the inception of the rotating stall. According to the CFD results, the size of the LEV doesn't change, and the LEV is comparatively stable in the designed flow operation. On the other hand, the LEV develops and forms a huge flow blockage within the diffuser passages during off-design operation. Therefore, the LEV may be considered to be one of the causes of the diffuser stall in the centrifugal compressor.

Two examples of the use of vortex control to reduce noise and enhance the stable operating range of a centrifugal compressor are presented in this paper. In the case of high-flow operation of a centrifugal compressor with a vaned diffuser, a discrete frequency noise induced by interaction between the impeller-discharge flow and the diffuser vane, which appears most notably in the power spectra of the radiated noise, can be reduced using a tapered diffuser vane (TDV) without affecting the performance of the compressor. Twin longitudinal vortices produced by leakage flow passing through the tapered portion of the diffuser vane induce secondary flow in the direction of the blade surface and prevent flow separation from the leading edge of the diffuser. The use of a TDV can effectively reduce both the discrete frequency noise generated by the interaction between the impeller-discharge flow and the diffuser surface and the broadband turbulent noise component. In the case of low-flow operation:a leading-edge vortex (LEV) that forms on the shroud side of the suction surface near the leading edge of the diffuser increases significantly in size and blocks flow in the diffuser passage. The formation of an LEV may adversely affect the performance of the compressor and may cause the diffuser to stall. Using a one-side tapered diffuser vane to suppress the evolution of an LEV, the stable operating range of the compressor can be increased by more than 12 percent, and the pressure-rise characteristics of the compressor can be improved. The results of a supplementary examination of the structure and unsteady behavior of LEVs, conducted by means of detailed numerical simulations, are also presented.

The effects of the diffuser vane geometries on the compressor performance and noise characteristics of a centrifugal compressor equipped with vaned diffusers were investigated by experiments and numerical techniques. Because we were focusing attention on the geometries of the diffuser vane's leading edge, diffuser vanes with various leading edge geometries were installed in a vaned diffuser. A tapered diffuser vane with the tapered portion near the leading-edge of the diffuser's hub-side could remarkably reduce both the discrete frequency noise level and broadband noise level. In particular, a hub-side tapered diffuser vane with a taper on only the hub-side could suppress the development of the leading edge vortex (LEV) near the shroud side of the diffuser vane and effectively enhanced the compressor performance.

Experiments and numerical analyses were used to investigate the unsteady behavior of a vortex generated on the leading-edge of a diffuser (i.e., leading-edge vortex (LEV)) and the diffuser stall inception in a centrifugal compressor equipped with vaned diffusers. The LEV is distinct from the separation vortex of the diffuser's leading edge and passage vortex of the diffuser; it is generated by the accumulation of vortices caused by the velocity gradient of the impeller discharge flow. The LEV increases with decreasing velocity in the diffuser passage and forms a huge flow blockage within the diffuser passage. Therefore, the LEV may help cause the diffuser stall inception in the centrifugal compressor. A diffuser vane, that was tapered only on the hub side was designed and used in the experiment. The results of the computational fluid dynamics analysis and experiments showed that the tapered diffuser vane can suppress LEV evolution during off-design operations. Therefore, the tapered diffuser vane may control the diffuser stall inception in a centrifugal compressor by suppressing LEV evolution.