JTM041015

High-Fidelity Numerical Analysis
of Per-Rev-Type Inlet Distortion

Transfer in Multistage Fans—Part
II: Entire Component Simulation
Jixian Yao
GE Global Research,
One Research Circle,
and Investigation
Niskayuna, NY 12309
Part I of this paper validated the ability of the unsteady Reynolds-Averaged Navier-
Steven E. Gorrell Stokes (RANS) solver PTURBO to accurately simulate distortion transfer and generation
Department of Mechanical Engineering, through selected blade rows of two multistage fans. In this part, unsteady RANS calcu-
Brigham Young University, lations were successfully applied to predict the 1/rev inlet total pressure distortion trans-
435 CTB, Provo, UT 84602 fer in the entirety of two differently designed multistage fans. This paper demonstrates
that high-fidelity computational fluid dynamics (CFD) can be used early in the design
Aspi R. Wadia process for verification purposes before hardware is built and can be used to reduce the
GE Aviation, number of distortion tests, hence reducing engine development cost. The unsteady RANS
30 Merchant Street, P20, code PTURBO demonstrated remarkable agreement with the data, accurately capturing
Cincinnati, OH 45215 both the magnitude and the profile of total pressure and total temperature measurements.
Detailed analysis of the flow physics identified from the CFD results has led to a thor-
ough understanding of the total temperature distortion generation and transfer mecha-
nism, especially for the spatial phase difference of total pressure and total temperature
profiles. The analysis illustrates that the static parameters are more revealing than their
stagnation counterpart and that pressure and temperature rise are more revealing while
the pressure and temperature ratio could be misleading. The last stage is effectively
throttled by the inlet distortion even though the overall engine throttle remains un-
changed. The total temperature distortion generally grows as flow passes through the fan
stages. 关DOI: 10.1115/1.3148479兴
1 Introduction resulting in a very successful validation, Part II builds on the

initial high-fidelity CFD capability to model the entirety 共three
Unsteady Reynolds-Averaged Navier-Stokes 共RANS兲 calcula-
stages plus IGV兲 of two multistage fans subject to 1/rev inlet
tions were successfully applied to predict temperature distortion distortion.
generation and transfer of total pressure distortion in response to Part II is focused on a higher fidelity understanding of distor-
1/rev inlet total pressure distortion 关1兴. Validation with total pres- tion transfer and fan response besides the large-scale simulations.
sure and total temperature experimental data showed that simula- A full-range validation is carried out for the multistage CFD pre-
tions predicted the proper magnitude and phase of distortion trans- diction, and the simulations are analyzed to understand the phys-
fer and generation in the front block of one multistage fan and the ics of temperature distortion generation and transfer. Accurate pre-
rear block of another fan. Each of these fans consists of an inlet diction of total temperature distortion is important as it affects the
guide vane 共IGV兲 and three stages, with engine test data to vali- corrected speed for downstream rotors and flow capacity of sta-
date the CFD procedure. tors. The level of induced temperature distortion entering the core
As the capability of parallel computing and processor speed has compressor is also an important factor since the temperature dis-
steadily increased, numerical simulations play an increasingly im- tortion is usually amplified as the fan attenuates the pressure dis-
portant role in discovering the physics of unsteady flows in turbo- tortion. Analysis of the simulations from Part I 关1兴 suggested that
machinery. High performance computing allows full annulus static temperature dominated the total temperature phasing. This
simulations to be obtained in an acceptable time frame to be of alluded to the examination of the distortion profiles of both static
benefit to the engine design process. pressure and density in Part I. Further investigation from the fan
Part I presented the motivation for this research and thoroughly response perspective is carried out in this paper for the multistage
reviewed the history and models used to understand and predict environment for the underlying mechanism of distortion transfer.
the impact of inlet distortion on stall margin. This high-fidelity The paper is divided into sections discussing the computational
CFD effort is unique in its attempt to evaluate the ability to model domain and boundary conditions, the full-range validation against
inlet distortion at a much higher fidelity level and to establish a engine test data, the mechanism of total temperature distortion
capable prediction tool for distortion transfer and fan response generation, and the distortion transfer analysis.
characterization in multistage turbomachinery. With the CFD
methodology and computational requirements defined in Part I
2 Computational Domains and Boundary Conditions
This section describes the overall configuration of the two se-
Contributed by the International Gas Turbine Institute of ASME for publication in lected fans, the computational domains, and the boundary condi-
the JOURNAL OF TURBOMACHINERY. Manuscript received September 9, 2008; final
manuscript received January 28, 2009; published online May 6, 2010. Review con-
tions used for the simulations. The flow solver chosen was the
ducted by David Wisler. Paper presented at the ASME Turbo Expo 2008: Land, Sea unsteady RANS code, PTURBO 共see Part I兲, developed by Dr. Jen-
and Air 共GT2008兲, Berlin, Germany, June 9–13, 2008. ping Chen with support from NASA, Industry, and the DoD. Its
Journal of Turbomachinery Copyright © 2010 by ASME OCTOBER 2010, Vol. 132 / 041015-1
Downloaded 28 May 2010 to 128.113.26.88. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
Fig. 1 Computational domain and grid of the first multistage
fan „not drawn to scale…
solution algorithm is an implicit finite volume solver that incor-

porates Newton subiterations and a block-Jacobi relaxation
scheme at each physical time step. It has a third-order accurate
monotonic upstream-centered schemes for conservation laws
共MUSCL兲-type spatial discretization scheme coupled with a
second-order accurate temporal discretization, which helps to im-
prove flow resolution and minimize dispersion error. Turbulence
modeling is accomplished via the NASA/CMOTT k-epsilon
model specifically developed for turbomachinery flows. This code
was modified by GE to accept distortion boundary conditions at
inlet and exit boundaries. With predetermined load balancing, this
flow solver has demonstrated high scalability and parallel effi-
ciency.
2.1 First Multistage Fan. The first multistage fan consists of
three stages with an IGV, as shown in Fig. 1. The computational
domain consists of all seven blade rows, full annulus for each
blade row. The blade counts and grid sizes for each passage of all
the blade rows are listed in Table 1.
The IGV and the stator rows were properly clocked to match
engine test setup. Boundary conditions were obtained from the
engine inlet distortion test. The inlet total pressure was a station-
ary 共in absolute frame of reference兲 1/rev sinusoidal distribution
in the fan circumference. There were no circumferential distor-
tions of total temperature and inflow swirl. The peak-to-peak total
pressure distortion at inlet was about 20% of the circumferential
mean. Figure 2共a兲 shows the absolute total pressure distribution at
inlet. The inlet boundary conditions were exactly the same as that
used in Part I of this paper.
The exit boundary condition downstream of Stator-3 was more
difficult to specify. The total pressure distortion was expected to
be attenuated at fan exit but the static pressure there was not
completely distortion-free. The engine test only had readings at
hub and casing, with none in-between. The static pressure profile
at this location was obtained from GE’s data matched through-
flow solution and was scaled to match the measurements from the
hub and casing. The resulting exit static pressure distribution is
Table 1 Blade counts and grid sizes of the first multistage fan,
201Ã 106 grid cells total „including halo cells…
Blade row Blade count Mesh size/passage 共␪ ⫻ z ⫻ r兲
IGV 13 137⫻ 145⫻ 81

Rotor-1 24 97⫻ 131⫻ 81
Stator-1 62 49⫻ 105⫻ 81
Rotor-2 46 65⫻ 109⫻ 81
Fig. 2 Boundary conditions of the first multistage fan: 1/rev
Stator-2 94 41⫻ 101⫻ 81
total pressure distortion at inlet „a… and the static pressure dis-
Rotor-3 50 65⫻ 109⫻ 81
Stator-3 96 41⫻ 121⫻ 81 tribution at Stator-3 exit „„b… and „c……. †„Pt − P̄t… / P̄t‡% is plotted
in „a…, and †„Ps − P̄s… / P̄s‡% is plotted in „b… and „c….
041015-2 / Vol. 132, OCTOBER 2010 Transactions of the ASME
validating against another fan similar to the first one.
The entire second fan model is shown in Fig. 3. The computa-
tional domain consists of all seven blade rows, full annulus for
each blade row. The blade counts and grid sizes for each passage
of all the blade rows are listed in Table 2. The IGV and the stator
rows were properly clocked to match engine test setup. The rotor
rows were properly clocked as well to the specification of the
engine test. The engine test used a distortion screen to produce the
designed total pressure distortion pattern. The inlet boundary con-
ditions were similar to that used for the first fan. A stationary 共in
absolute frame of reference兲 1/rev sinusoidal distortion of total
pressure was specified. Total temperature and flow angles were
uniformly distributed in the circumference, but the radial profiles
were retained. The peak-to-peak total pressure distortion at inlet
was about 35% of the circumferential mean, which was signifi-
Fig. 3 Computational domain and grid of the second multi- cantly higher than the inlet distortion level of the first fan. Figure
stage fan „not drawn to scale… 4共a兲 shows the absolute total pressure distribution at inlet.
The exit boundary condition was determined using the same
method described for the first fan. The exit location was set in-
shown in Fig. 2共b兲. This exit boundary condition corresponds to between the fan OGV and the leading edge of the splitter. The
the operating condition at about 97% corrected speed on the op- static pressure profile at this location was obtained from the data-
erating line. Physical rotating speed is about 8788 rpm. matched through-flow solution and was scaled to match the mea-
An additional near-stall operating condition for this fan was surements from the hub and casing. This exit boundary condition
also simulated at the same speed. At this condition, additional corresponds to the operating condition at about 100% corrected
measurement of static pressure at the engine splitter leading edge speed on the operating line. The physical rotating speed of this fan
was used for the specification of the back pressure profile. The is about 12,000 rpm. The resulting exit boundary condition distri-
near-stall back pressure distribution is shown in Fig. 2共c兲. bution is shown in Fig. 4共b兲.
2.2 Second Multistage Fan. The second multistage fan also 2.3 Nonuniform Boundary Condition Specification. The
consists of three stages with an IGV. This fan has a higher overall distortion screens for both fans were designed to produce sinu-
pressure ratio and has a different design philosophy than the first soidal total pressure distortions, so a sinusoidal function was fitted
fan. Validation for both fans, therefore, would enhance our confi- through the inflow total pressure data using Eq. 共1兲 at various
dence toward the numerical prediction capability, in contrast to immersions. As a result, the radial profiles of the circumferentially
averaged total pressure, the amplitude, and the phase angle of the
total pressure were specified to the CFD flow solver. The CFD
Table 2 Blade counts and grid sizes of the second multistage flow solver then interpolates the boundary conditions onto the
fan, 313Ã 106 grid cells total „including halo cells… computational mesh at inlet to the fans. The same method was
used for the back pressure specification.
Blade row Blade count Mesh size/passage 共␪ ⫻ z ⫻ r兲
Pt = Ptave · 共1 + amp_Pt cos共phase_Pt + ␪兲兲共1兲
IGV 17 97⫻ 169⫻ 81
Rotor-1 28 85⫻ 129⫻ 81
Stator-1 68 49⫻ 113⫻ 81 3 Validation
Rotor-2 42 65⫻ 127⫻ 81
Stator-2 108 49⫻ 113⫻ 81 This section provides the full validation of the numerical simu-
Rotor-3 50 65⫻ 127⫻ 81 lations to both multistage fans. The validation is a fundamental
Stator-3 118 49⫻ 169⫻ 81 and foremost effort for this research, and the focus is to validate
Fig. 4 Boundary conditions of the second multistage fan: 1/rev total pressure distortion at inlet „a… and the static pressure
distribution at Stator-3 exit „b…. †„Pt − P̄t… / P̄t‡% is plotted in „a… and †„Ps − P̄s… / P̄s‡% is plotted in „b….
Journal of Turbomachinery OCTOBER 2010, Vol. 132 / 041015-3
Fig. 5 Comparison of total pressure and total temperature profiles at about 10% immersion; first fan. Lines with rapid
oscillation are the time instantaneous solution. Overlaid smoother lines are time-averaged solution. Solid symbols are
measured data.
the general unsteady RANS capability instead of a particular flow distortion screen located further upstream of the IGV. Compari-
solver. No parameters relating to numerical algorithms and turbu- sons of the CFD results against data are included in this section
lence models were varied to gain a better match to data. Even for all available total pressure and total temperature data. The
though the distortion transfer does have its requirement for CFD CFD results are time averaged from the time-accurate calculations
procedures, these procedures can be met by any well-written flow covering one rotor revolution which had 15,360 physical time
solvers in the unsteady RANS flow regime. For the flow solver steps. The time-average process is a simple trapezoidal averaging
information and initial validation, please refer to Part I of this of time-resolved total pressure and total temperature.
paper 关1兴. Figures 5–9 compare the total pressure and total temperature at
3.1 Validation of the First Multistage Fan. Total pressure all five immersions. Time-averaged CFD results are shown with
and total temperature measurements were available at various ra- overlay of one time instantaneous solution to illustrate unsteady
dial locations at the leading edge of all three vane rows and at the variations. Symbols represent steady-state data points. Not all
measured data.
measured data.
measured quantities were available at all immersions, and at cer- sectors of the fan were more sensitive to upstream conditions than
tain immersions, only one probe had valid readings. At certain other sectors of the fan when distortion was present. This sensi-
immersions and certain circumferential locations, the measure- tivity will need further attention. The fan inlet total pressure and
ments showed scattering of data. This might be due to the fact that total temperature are also given in these plots as a reference.
the distortion screen needed to be rotated to generate whole-wheel Overall, the total pressure distortion transfer is predicted very
readings for a limited number of probes on the vanes. Two vanes well. No particular immersion had overwhelmingly better or
per row were instrumented for this fan, and 13 distortion screen worse comparisons. The total temperature distortion generation by
positions were used. It could also be an indication that certain Rotor-1 was predicted. The measurement showed that the total
measured data.
measured data.
temperature profile 共in circumferential direction兲 had a “phase circumferential phase兲 for different screen positions. However, at
lag” 共about 90 deg in the direction of rotation兲 from the corre- the near-stall condition, this assumption is less valid than at nor-
sponding total pressure profile. This “phase” difference and its mal operating conditions, hence only one screen position with
subsequent transfer to the downstream stages were accurately cap- which the data were taken. Figures 10–12 are the comparisons of
tured by the CFD simulations. total pressure and total temperature profiles at about 30%, 50%,
This fan was further throttled to a near-stall condition, where and 70% immersions. The data match well to the numerical simu-
limited measurement was taken. This was due to the fact that only lation results, thus providing confidence in the use of the simula-
two stator vanes per row were instrumented with Pt and Tt tion results to analyze the relevant flow physics.
probes, and the circumferential resolution was achieved via dis-
tortion screen rotation. The underlining assumption for this com- 3.2 Validation of the Second Multistage Fan. Similar to the
mon setup is that the flow fields are the same 共only differ in first multistage fan, total pressure and total temperature measure-
Fig. 10 Comparison of total pressure and total temperature profiles at about 30% immersion. First fan, near-stall condition.
ments were available at various radial locations at the leading are less data scattering for this fan as compared with the data of
edge of all three vanes. The fidelity of the measurement is about the first fan. More total pressure and total temperature probes
the same as that for the first fan for total pressure. Total tempera- were installed; hence less rotation of the distortion screen was
ture measurements at certain locations had less resolution than the necessary to reach the same resolution as of the first fan. In gen-
first fan due to invalid readings from certain probes. Six vanes per eral, the quality of experimental data was better for the second fan
stator row were instrumented, and three distortion screen rotations compared with the first fan. For the numerical simulations, how-
were used. Comparisons of the CFD results against data are in- ever, there is no need to mimic the way the engine measurement
cluded in this section for all available total pressure and total was done. The computational domain includes all the blade pas-
temperature data. The CFD results are time averaged from the sages of all the blade rows, with the inlet flow condition mapped
time-accurate calculations covering one rotor revolution, which to the entire inlet surface of the fan.
has 7350 physical time steps. Overall, the total pressure distortion transfer was predicted very
Figures 13–17 present comparisons of total pressure and total well. No particular immersion had overwhelmingly better or
temperature at all five measured stator vane leading edge immer- worse comparisons. Similar to the first fan, the total temperature
sions. Comparisons are also presented at the fan exit. The plots for distortion generation by Rotor-1 is evident and is matched by
the fan exit are stacked on top of the main plots for clarity. There numerical simulation. The total temperature profile at Rotor-1 exit
Fig. 13 Comparison of total pressure and total temperature profiles at about 10% immersion; second fan. Lines with rapid
measured data.
measured data.
measured data.
exhibits a 90 deg “phase lag” from the total pressure profile. The the total temperature distortion profiles, with a large phase differ-
generation and subsequent transfer of the total temperature distor- ence by the first stage, progresses through the fan with rather
tion, as well as the total pressure distortion propagation, were small shifts 共further lags兲 by the second and the last stage.
closely matched by the numerical simulation. This observation was introduced in Part I of this paper by the
The validation for both multistage fans is considered excellent, authors 关1兴. That discussion was from the equation-of-state per-
establishing confidence in high-fidelity CFD capability to predict spective, which highlighted the role of density variation. The den-
distortion transfer and, in general, the overall work output of each sity role is often neglected by distortion transfer models, which
fan stage. The validation is significant for the following reasons. assume incompressible flow. Part I suggested that the spatial
First, it establishes the foundation for both higher-fidelity inte- phase of the stagnation parameters was dictated by the respective
grated inlet/fan simulations and lower-fidelity reduced-order mod- static parameters. With the complete fan domain modeled in this
eling of distortion transfer. Second, it shows the value of bringing paper, this concept is re-examined to better understand the mecha-
unsteady CFD into earlier stages of the engine design process to nism. Figure 18 shows the static pressure and static temperature
assess distortion handling capability before the hardware is built. profiles throughout the first fan. The purpose of examining the
A consequence is more accurate predictions of pressure and tem- static parameters is to comprehend the role of velocity distortion
perature profiles entering the core compressor. It can also be used since the stagnation parameters have an overall “masquerading”
as a virtual test tool to study distortion management schemes. It effect and may not reveal the underlying mechanism completely.
can guide engine distortion characterization to achieve better test There are two major findings. First, the static pressure profile
methodology and to reduce the number of tests that need to be shifts every time it passes through a blade row, either a rotor or a
performed. Third, the validation is not tied to one particular CFD stator row. The total pressure profiles, however, do not show this
flow solver; it is essentially an assessment of the general capabil- “zigzag” shifting pattern. Second, the static temperature profiles
ity of unsteady RANS. The CFD requirement can be met by any closely mimic the total temperature profiles, suggesting that the
well-written CFD flow solver of compressible unsteady RANS. static temperature profile is dominating the total temperature
phase. This leaves the velocity distortion to a nonimportant role in
4 Spatial Phase of Total Pressure and Total Tempera- terms of phasing. It is worth noting that the profiles of these
parameters are not in strict sinusoidal form anymore. In the posi-
ture Profiles tive pressure gradient sector 共0–180 deg兲, it has a fuller 共convex兲
A notable feature observed in both data and CFD results was shape than a sinusoidal form, while at the negative pressure gra-
that the total temperature profile had about 90 deg phase differ- dient sector 共180–360 deg兲, it has a more concaved form than a
ence in the circumferential direction. Compared to the high peak sinusoidal function. On the other hand, it is important to point out
of the total pressure profiles, the high peak of the total temperature that the density has distortion as well 共see Fig. 19兲. The velocity
profiles lags behind by about 90 deg. The total pressure high peak profile has its own propagation behavior, with the largest distor-
is leading in the direction of rotation. For the distortion propagation at IGV inlet and much less distortion level at all stator inlets.
tion, the total pressure profiles largely stays in phase as it Mazzawy 关2兴 attributed the total temperature phase difference
progresses through the fan 共small phase shifts are observed兲, and to angular particle displacement, and was able to incorporate the
measured data.
particle displacement in a multisegment parallel compressor rise of the same rotor, this in turn leads to the generation of the
model to match data significantly better. From the Lagrangian observed total temperature distortion generation by Rotor-1 共ve-
perspective, while the fluid particles are transported by the rotor locity profile at Rotor-1 exit is relatively flat, see Fig. 19兲, and
blades, the properties of the particles 共i.e., pressure and tempera- explains the phase difference of the resultant total temperature
ture兲 are altered as they pass though the blade passage. A particle profile. Note that the Rotor-1 inlet static temperature profile is
at rotor inlet with the highest temperature of the incoming flow affected by the local velocity distortion due to the uniform total
may not be at the high peak of the temperature profile downstream temperature profile at fan inlet. This finding is also explained by
of the rotor at the displaced circumferential location. With a uni- Gibb’s equation, Tds = dh − dp / ␳. The entropy change in this equa-
form Rotor-1 inlet total temperature, the particle displacement tion may include flow state change between inlet and exit among
analogy does not explain the total temperature distortion genera- other factors.
tion by Rotor-1 with a 90 deg phase difference. Using the average Figures 20 and 21 reveal that pressure and temperature distor-
axial velocity component at the midspan in the low PTA 共low tions are generated and transferred quite differently by each indi-
axial velocity兲 section, the estimated particle displacement using vidual stage. Across the first stage, the peak pressure rise is not at
⌬␪ = 共b / u兲␻ is 52 deg 共positive in the direction of rotation兲, where the same location of the high peak of the incoming pressure pro-
b is the axial chord length, u is the axial velocity, and ␻ is the file, rather it lags quite a distance from the high peak of the in-
angular wheel speed. This angle is then offset by the Rotor-1 coming pressure profile. Combined with the incoming temperature
blade stagger angle, and the net angular displacement is about 9 共which is influenced by the velocity distortion since the inlet total
deg ahead of the particle location at the Rotor-1 inlet. This does temperature is constant兲, the temperature profile high peak rests at
not agree with total pressure and total temperature profiles from about 90 deg in the fan circumference. Similar explanation can be
either data or CFD results 共total pressure high peak is leading the found for the low peak of the temperature profile. The second
total temperature high peak in the direction of rotation兲, although stage, however, has a relatively flat pressure rise profile. It shifts
the particle angular displacement certainly exists in the flow. whatever the incoming profile of pressure to a higher level by
It is important to note that the apparent phase shift is not a applying relatively uniform pressure rise. There are other factors,
literal shift. It is the result of pressure and temperature rise caused i.e., inlet swirl distortion, that might drive small changes in the
by the blade rows operating at inlet and discharge conditions that pressure and temperature profiles for the middle stage. The distor-
vary in the circumferential direction. To uncover the mechanism, tion level at the second stage 共discussed in Sec. 5兲 may appear to
both the static pressure and static temperature rise of each stage be lower because of the increased overall pressure level at the
are examined and shown in Figs. 20 and 21 for both multistage second stage, but the amount of distortion remains largely un-
fans. In each of the figures, there are three individual plots that are changed. The last stage 共Rotor-3兲 has a pressure rise profile almost
stacked up for the clarity to show the profiles for each stage. Static in the mirror image of the incoming pressure profile. This is due to
pressure rise is normalized by density and C p to arrive at the same the Rotor-3 discharge pressure profile that is relatively flat.
level of static temperature rise. These two figures, showing the Rotor-3 is thus “forced” to attenuate the distortion, which has
pressure and temperature rise of both fans, clearly illustrate the been amplified by the upstream stages. The relatively flat dis-
correlation between the pressure rise and the temperature rise. The charge pressure profile is determined due to large axial space in-
temperature rise through Rotor-1 is directly related to the pressure between the fan exit and the splitter of bypass and core. As a
measured data.
result, Rotor-3 passages are effectively throttled as they pass difference varies significantly. The dominating factors and their
through the pressure distortion at inlet. The stability margin of roles for the PTA / TTA phase difference remain as a research
Rotor-3 is therefore consumed even though the overall engine topic. Furthermore, whether and how this phase difference affects
throttle remains constant. This stage 共rotor兲 is more likely to be the stability margin is a question that needs to be answered.
the limiting stage in the multistage environment, and needs to
have plenty room in the stall margin by design.
It needs to be noted that the total pressure and total temperature 5 Distortion Transfer and Generation
phase difference is not a universal phenomenon, as it can be seen Detailed discussion is presented in this section for total pressure
from the above analysis. Among some other GE fans, the phase distortion transfer, the generation of total temperature distortion,
Fig. 18 Static pressure and static temperature profiles of the first fan at about 50% immersion
Fig. 19 Density and absolute velocity profiles of the first fan at about 50% immersion
and the swirl distortion. The distortion levels for both fans are TTmax − TTavg
plotted using the same scale for comparison purposes. The trans- TT _ distortion levellocal = TTDLlocal =
TTavg
fer of the total pressure distortion can be initially observed from
the validation section above. To highlight the distortion transfer,
5.1 First Multistage Fan. The stage-by-stage distortion level
the ARP-1420 standard was used for defining distortion levels and
was applied to various immersions of the fan to examine the trans- at various immersions is shown in Figs. 22–24. In those plots,
fer behavior. The definitions are recited as follows: distortion levels for both the stagnation and static parameters are
included. For total pressure distortion, at all the immersions, the
PTavg − PTmin inlet distortion is amplified by the first stage and is then slightly
PT _ distortion levellocal = PTDLlocal =
PTavg attenuated by the second stage. The distortion level by the second
Fig. 20 Static pressure and static temperature rise across the first fan at
50% immersion
Fig. 21 Static pressure and static temperature rise across the second fan
at 50% immersion
stage is still significantly higher than the inlet distortion level. section.
This elevated distortion level is significantly attenuated by the last Total temperature distortion is generated when the flow passes
stage, for all the immersions, though the tip section shows more through the first stage. As the flow progresses downstream, more
attenuation than at the hub section. More attenuation at the tip distortion is generated. At 50% immersion, the total temperature
section is due to steeper characteristics at the tip than at the hub distortion level exhibits a linear trend of growth. The significant
Fig. 22 Comparison of total pressure and total temperature distortion levels at about 10% immersion;
first fan
first fan
attenuation of the total pressure distortion by the last 共third兲 stage ation兲 and low temperature distortion generation.
does not seem to cause the generation of total temperature distor- The static pressure distortion largely follows the trend of the
tion to be above the linear trend. However, this is not the case near total pressure distortion. The most significant amplification is seen
the tip section. At 10% immersion, the total temperature distortion by the first stage, where the static pressure distortion level is more
level at the last stage does seem to respond to the rapid attenuation than doubled from the inlet. This comes with the fact that the total
of total pressure distortion by the last stage. A surprise comes pressure ratio of the first stage is the highest among the three
from the hub section at about 90% immersion, where there is a stages, whereas the static pressure rise across the first stage is the
similarly rapid attenuation of the total pressure distortion by the least among the three stages.
last stage but to a lesser degree as compared with the other im-
mersions. From the first stage to the second, the total temperature 5.2 Second Multistage Fan. The stage-by-stage distortion
distortion responds with a 50% increase 共in distortion level兲 to a level at various immersions is shown in Figs. 25–27. Distortion
mild total pressure distortion attenuation by the middle 共second兲 levels for both the stagnation and static parameters are included in
stage. However, from the second stage to the third, the tempera- these plots. Compared to the first fan, the following observations
ture distortion level decreases in response to a rather significant can be made.
attenuation of pressure distortion by the third stage. This reveals
the nonlinear nature of distortion generation and transfer, and a 1. The inlet distortion level is significantly higher than that of
certain degree of separation of correlation between the total pres- the first fan; as a result, the distortion levels at each stage are
sure distortion and total temperature distortion. The common be- significantly higher.
lief that more PTA distortion attenuation leads to more TTA dis- 2. The general behavior of the distortion transfer is similar to
tortion generation is thus conditionally true. Examining the stage/ that of the first fan. However, some differences are observed.
blading design philosophy together with the CFD results could At 10% immersion, the total pressure distortion level at
lead to designs with low pressure distortion transfer 共more attenu- Stator-2 inlet is about the same as the distortion level at the
first fan
second fan
fan inlet, while for the first fan, the distortion level at different kinds of swirl distortions being observed from the nu-
Stator-2 inlet is still higher. This is true for other immersions merical simulations of the two multistage fans. The first kind is
as well, that the total pressure distortion level at Stator-2 the induced swirl upstream to the fan inlet. This is due to the
inlet is at or below the distortion level at the fan inlet. This imbalance of the static pressure in the circumference caused by
could be due to the higher overall pressure level at Stator-2 the total pressure distortion. This imbalance of the static pressure
inlet generated by a higher pressure ratio of the Rotor-2.
gives rise to a secondary flow that goes from high pressure sector
3. For total temperature distortion levels, at 10% immersion,
no growth is found above the linear trend by the last stage as to low pressure sector. Near the leading edge of the IGV, the
is it true for the first fan. At 90% immersion, the total tem- induced swirl distortion reaches ⫾10 deg. The IGV removes this
perature distortion level does not dip below the distortion induced swirl distortion significantly, illustrating the important
level at the Stator-2, as it is the case for the first fan. role of the IGV. The remnant induced swirl distortion is absorbed
by the Rotor-1. Details of the swirl distortion were reported by the
These differences in distortion transfer do not really underscore authors in 2007 关3兴. The other kind of swirl distortion is generated
a significant change in transfer behavior, even though the two fans by the rotor blades in response to the distorted inflow, be it total
were designed with quite different philosophies. pressure distortion alone or combined total pressure and total tem-
5.3 Generation of Swirl Distortion. Due to the fan response perature distortions. This kind of swirl distortion can be observed
to the inlet pressure distortion, not only the total temperature dis- downstream of Rotor-1 and all the way across the fan. It has
tortion is generated, but also swirl distortions are generated as different spatial phase than the induced swirl at the fan inlet. Sig-
well. Figure 28 presents the absolute swirl profiles of the first fan nificant amount of swirl distortion is generated by the last stage
at stator leading edges with special attention to the IGV. The sec- rotor, which is “forced” to attenuate the elevated total pressure
ond fan has very similar swirl distortion profiles. There are two
second fan
second fan
distortion at Rotor-3 inlet. The swirl distortion by the last stage fidelity CFD can be brought in early design process for veri-
rotor is effectively removed by the fan OGV 共Stator-3, not fication purposes before hardware is built and can be used to
shown兲. guide the distortion characterization in distortion tests, hence
reducing engine development cost.
6 Conclusions 2. Even with inlet total pressure distortion alone, both fans see
Unsteady RANS calculations were successfully applied to pre- the distortion of total temperature and swirl, generated in the
dict the 1/rev inlet total pressure distortion transfer in the entirety multistage environment. Therefore, this validation is not just
of two differently designed multistage fans. This is a significant for the inlet total pressure distortion but also for three major
advance from the numerical experiments reported in the Part I of distortion types.
this paper. The following conclusions were drawn from the analy- 3. Detailed analysis of the CFD results has led to a thorough
sis of the numerical results and the comparisons of CFD to engine understanding of the total temperature distortion generation
test data. and transfer mechanism, especially for the spatial phase dif-
ference of total pressure and total temperature profiles. This
1. High-fidelity CFD is further demonstrated to be able to pre- illustrates that the static parameters are more revealing than
dict the distortion transfer accurately for multistage fans. their stagnation counterpart and that pressure and tempera-
The work split among the stages is accurately predicted as ture rise are more revealing while the pressure and tempera-
well. This demonstration validates and verifies the general ture ratio could be misleading.
capability of the unsteady RANS flow solvers. It lays a foun- 4. The total pressure distortion transfer behavior is analyzed at
dation for higher-fidelity integrated inlet/fan simulations and various immersions and in a stage-by-stage fashion. The first
serves as a resource for reduced-order modeling. high- stage amplifies the inlet distortion, while the second stage
carries it along. The distortion level to the last stage is either
at or higher than the distortion level at the fan inlet. The last
stage, subjecting to a near uniform discharge pressure pro-
file, significantly attenuates the total pressure distortion.
5. The last stage is effectively “throttled” by the inlet distortion
even though the overall engine throttle remains unchanged.
This illustrates how the last stage rotor stability margin is
consumed, which makes it the limiting stage for the overall
stability of the fan.
6. In general, the total temperature distortion level grows as
flow passes through the fan stages. The temperature distor-
tion at fan exit concerns the downstream core compressor
and combustor as well.
7 Future Work
In addition to the distortion simulations presented in this paper,
simulations with clean inlet flow have been performed as well
under the same operating conditions as reported in this paper.
Analysis of fan response and the impact of inlet distortion to
stability margin are being carried out, and will be reported in the
future.
Acknowledgment
Fig. 28 Induced swirl at inlet and swirl distortion transfer of The authors wish to thank the support of the DoD High Perfor-
the first fan at 50% immersion mance Computing Modernization Program Office and the Aero-
nautical System Center Major Shared Resource Center for the b ⫽ blade chord length
Challenge Award that provided the high performance computing u ⫽ axial velocity of the fluid
resources. Without their support, this work is not possible as it ␳ ⫽ density of the fluid
requires large number of processors and large data memory. The s ⫽ entropy
authors also want to thank Jenping Chen, now at Ohio State Uni- h ⫽ enthalpy
versity, for his support of PTURBO; Peter Szucs and Peter Wood, Cp ⫽ fluid specific heat at constant pressure
both of GE Aviation, for their support of this research and many ␻ ⫽ engine wheel speed
insightful discussions. Ravi Ravindranath of NAVAIR is recog-
nized for providing early motivation to embark on this project. Subscripts
Mike Macrorie and Joe Capozzi from GE Aviation and Professor avg, ave ⫽ time average or engine circumferential average
Garth Hobson from the Naval Postgraduate School also assisted min ⫽ minimum value in engine circumference
with the simulations. The authors are grateful for the funding sup- max ⫽ maximum value in engine circumference
port provided by the Advanced Virtual Engine Test Cell local ⫽ local value at a specific immersion
共AVETeC兲. Finally, they thank the General Electric Co. and the
Propulsion Directorate management for supporting the research
and allowing the publication of this paper. References
关1兴 Yao, J., Gorrell, S. E., and Wadia, A. R., 2010, “High-Fidelity Numerical
Analysis of Per-Rev-Type Inlet Distortion Transfer in Multistage Fans—Part I:
Nomenclature Simulations With Selected Blade Rows,” ASME J. Turbomach., 132, p.
PTA , PT , Pt ⫽ absolute total pressure 041014.
PS , Ps , p ⫽ static pressure 关2兴 Mazzawy, R., 1977, “Multiple Segment Parallel Compressor Model for Cir-
TTA , TT , Tt ⫽ absolute total temperature cumferential Flow Distortion,” ASME J. Eng. Power, 99, pp. 288–296.
关3兴 Yao, J., Gorrell, S. E., and Wadia, A. R., 2007. “A Time-Accurate CFD Analy-
TS , Ts ⫽ static temperature sis of Inlet Distortion Induced Swirl in Multistage Fans,” AIAA Paper No.
␪,r,z ⫽ engine tangential, radial, and axial coordinates 2007-5059.

JTM041015

Uploaded by

Copyright:

Available Formats

JTM041015

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

JTM041015

Uploaded by

Copyright:

Available Formats

High-Fidelity Numerical Analysis

of Per-Rev-Type Inlet Distortion

1 Introduction resulting in a very successful validation, Part II builds on the

solution algorithm is an implicit finite volume solver that incor-

Blade row Blade count Mesh size/passage 共␪ ⫻ z ⫻ r兲

IGV 13 137⫻ 145⫻ 81

041015-2 / Vol. 132, OCTOBER 2010 Transactions of the ASME

Journal of Turbomachinery OCTOBER 2010, Vol. 132 / 041015-3

041015-4 / Vol. 132, OCTOBER 2010 Transactions of the ASME

Journal of Turbomachinery OCTOBER 2010, Vol. 132 / 041015-5

041015-6 / Vol. 132, OCTOBER 2010 Transactions of the ASME

Journal of Turbomachinery OCTOBER 2010, Vol. 132 / 041015-7

041015-8 / Vol. 132, OCTOBER 2010 Transactions of the ASME

Journal of Turbomachinery OCTOBER 2010, Vol. 132 / 041015-9

041015-10 / Vol. 132, OCTOBER 2010 Transactions of the ASME

Journal of Turbomachinery OCTOBER 2010, Vol. 132 / 041015-11

041015-12 / Vol. 132, OCTOBER 2010 Transactions of the ASME

Journal of Turbomachinery OCTOBER 2010, Vol. 132 / 041015-13

041015-14 / Vol. 132, OCTOBER 2010 Transactions of the ASME

Journal of Turbomachinery OCTOBER 2010, Vol. 132 / 041015-15

041015-16 / Vol. 132, OCTOBER 2010 Transactions of the ASME

Journal of Turbomachinery OCTOBER 2010, Vol. 132 / 041015-17

You might also like