experimental analysis of the unsteady, turbulent …
TRANSCRIPT
Paper ID: ETC2017-320 Proceedings of 12th European Conference on Turbomachinery Fluid dynamics & ThermodynamicsETC12, April 3-7, 2017; Stockholm, Sweden
EXPERIMENTAL ANALYSIS OF THE UNSTEADY,TURBULENT FLOW THROUGH THE FAN STAGE OF A
HIGH-BYPASS TURBOFAN IN WINDMILLING CONDITIONS
A. Thacker, N. García Rosa, G. Dufour.
?Université de Toulouse – Institut Supérieur de l’Aéronautique et de l’Espace(ISAE-SUPAERO)
10 avenue Edouard Belin, BP 54032, 31055 Toulouse Cedex 4, France
ABSTRACTA detailed study of the unsteady airflow through the fan stage of a bypass turbofan is pro-posed. Experiments are conducted in the turbofan test facility of ISAE, specially suitedto reproduce windmilling operation in an ambient ground setup. Hot-wire anemometryis used to obtain steady Mach number and flow angle profiles as well as to characterizethe unsteady, turbulent flow, at different stations across the fan stage (fan inlet, outlet androtor/stator interface). Phase averaging of unsteady signals and integral length scale esti-mations provide further evidence of the presence of a vortex shedding from the stator, aswell as the fact that it is correlated to the rotor blade passing frequency. Self-similarity ofmean Mach number, flow angle and turbulent intensity profiles is observed. Furthermore,turbulent scale profiles are found to scale well with blade passing period.
KEYWORDSTURBOFAN, WINDMILLING, HOT-WIRE ANEMOMETRY, TURBULENT SCALES
NOMENCLATURE
Latin lettersA2 Inlet cross section (m2)
~e Unit vector
h Distance to hub (m)
h/hmax Normalized distance to hub (-)
ir Rotor incidence (◦)
is Stator incidence (◦)
Iv Turbulence intensity
Lt Temporal turbulent length scale (s)
M Time averaged Mach number (-)
M2A Mach number at engine inlet (-)
βrotor Rotor leading edge metal angle (◦)
αstator Stator leading edge metal angle (◦)
Nfan Rotational speed (rpm)
Ntr Number of blade passings (-)
R Ideal gas constant for air (287.04 J/kg/K)
Rvv Auto-correlation function (-)
T Temperature (K)
Tr Blade passing period (s)
v Velocity magnitude (m/s)
Greek lettersα Absolute azimuthal angle (◦)
β Relative azimuthal angle (◦)
Φ2 Dimensionless mass flow parameter (-)
σv Standard deviation of v (m/s)
τ Correlation time lag (s)
Operators
OPEN ACCESSDownloaded from www.euroturbo.eu
1 Copyright c© by the Authors
· Phase average
· Time average
AcronymsBPF Blade passing frequency (Hz)
Subscripts
2 Relative to engine inlet conditions
i Relative to stagnation conditions
θ Relative to azimuthal component
r Relative to radial component
x Relative to axial component
INTRODUCTIONWhen an aircraft engine flames out during flight, the ram pressure at the fan inlet creates an
internal airflow that causes spool rotation, leading to windmilling operation. The prediction ofkey performance parameters in these conditions, such as the in-flight relight capability, the fanrotational speed, or the drag of an inoperative engine in flight, is crucial for engine designers.Studies on turbofan windmilling and, more generally, on sub-idle operation, focus on the wholeengine by resorting to either thermodynamic engine cycle modelling (Braig et al., 1999; Riegleret al., 2003; Fuksman and Sirica, 2012) or performance experiments (Wallner and Welna, 1951;Mishra et al., 2008). Although this global approach provides with essential information on en-gine performance, it brings little insight into the aerodynamics of the different turbomachineryelements. Yet a detailed understanding of the flow features inside the critical components isessential for a reliable prediction of the overall performance, especially in severe off-designoperating conditions. Available detailed experimental and numerical studies, conducted on acompressor cascade (Zachos et al., 2011) and on a contemporary high-bypass ratio turbofanin an altitude test cell (Prasad and Lord, 2010) indicate that, in windmilling operation, the fanstage operates under severe off-design (negative) angle-of-attack conditions, leading to the de-velopment of flow separation on both rotor and stator. These observations were confirmed bya previous experimental study on a small, geared turbofan, at ground level (García Rosa et al.,2015). These studies agree that the relative flow leaves the fan at nearly the metal angle, and thatthe bulk of the pressure loss occurs in the stator. A recent study (Dufour et al., 2015) has furthershown that a steady numerical approach fails to reproduce some important features of the flow.The authors conducted unsteady simulations on a 2D section of the fan stage. They show thatthe presence of flow separation in both the rotor and stator leads to a strong rotor/stator inter-action, in which the rotor wake triggers a periodic movement of the reattachment point at thetrailing edge of the stator, leading to vortex shedding. Due to the 2D nature of the simulatedsection, the results are not directly representative of the 3D flow, and the question is still openwhether the vortex shedding occurs synchronously with the rotor wake passage. This paperaims at developing a better understanding of the unsteady, turbulent flow features through thefan stage. At the same time, a first experimental database is constituted, for the validation ofsteady numerical simulations.
EXPERIMENTAL SETUPThe engine under consideration is an advanced, high bypass ratio, unmixed flow geared
turbofan developed by Price Induction. Figure 1 shows a schematic view of the engine layoutand instrumentation.
The engine is equipped with conventional instrumentation consisting of steady pressure andtemperature measurements as depicted in the upper part of Figure 1. A set of radial intrusionports completes this setup. For each station in the lower part of Figure 1, a number of ports
2
2 13
25 318
85
41 45
21
ConventionalInstrumentation
Total Pressure & Temperature Array
Wall Static Pressure Port
21A2R 22A 18 82A
Radial Probe Traverse System
SpecificInstrumentation
Figure 1: Engine conventional instrumentation, traverse positions and station nomencla-tures.
located at the different azimuthal positions allow the introduction of probes. Tests are carriedout on the ISAE turbofan test facility, which has been suited to simulate windmilling conditionsat ground level. A 75 kW centrifugal fan is used to blow air through the engine, up to 6 kg/smass flow rate and a local Mach number of 0.16 at the engine inlet. Fuel feed is naturallycut off. A more complete description of the setup as well as the windmilling operating rangeunder study can be found in García Rosa et al. (2015). The operating point is defined by thedimensionless flow parameter at engine inlet
Φ2 =m2√
R · Ti,2
pi,2 · A2. (1)
The values of rotational speed, and the resulting blade passing frequency (BPF) are recalled intable 1.
Φ2 (–) 0.09 0.11 0.13 0.15 0.18
Nfan (rpm) 1050 1330 1610 1900 2200BPF (Hz) 245 310 375 440 510
Table 1: Rotor parameters as a function of the tested windmilling operating points.
The flow velocity magnitude and azimuthal angle are measured across the fan stage byusing a dual hot wire probe. The anemometer is a Dantec 2D fiber-film-probe 55R52 connectedto Dantec conditioners operating in the constant temperature mode. Velocity and directionalcalibrations have been performed by exposing the probe to a set of known velocities and flowangles in the free jet of the Dantec calibration unit. During the experiments the probe outputvoltages are reduced into measured velocity and flow angle by applying a look-up table method,and temperature correction is applied according to Jorgensen (2002). The probes are radiallyintroduced in the flow, so that the plane of the wires coincides with the local azimuthal, blade-to-blade plane (~ex, ~eθ) depicted in figures 2.b and 5. Precision stepper motors allow the probesto traverse the flow radially, and rotate along their axis. The initial positioning consists in
3
2A 2R 22A21A
12345
12345
12345
145
858
14
58
14 5
8
14
B-10
B-09
B-08
B-02B-03
B-06
B-05
B-01 B-07
B-04
1
Probe Traverse
Pressure Array
2A 2R 22A21A
1
2
3
4
5
12
34
5
1
2
3
4
5
145
8
5
8
14
5
8
1
45
8
1
4
B-10
B-09
B-08
B-02
B-03
B-06
B-05
B-01 B-07
B-04
1
Probe Traverse
Pressure Array
2A 2R 22A21A
12345
12345
12345
145
858
14
58
14 5
8
14
B-10
B-09
B-08
B-02B-03
B-06
B-05
B-01 B-07
B-04
1
Probe Traverse
Pressure Array
(a)
Fib
er-
film
wit
hh
eavy
qu
art
zco
ati
ng
Nic
kel f
ilmde
posi
ted
on70
µm
diam
eter
quar
tzfib
er. O
vera
ll
leng
th3
mm
, sen
sitiv
ele
ngth
1.25
mm
. Cop
per
and
gold
plat
edat
the
ends
. Film
ispr
o-
tect
edby
quar
tzco
atin
gap
prox
. 2µm
inth
ickn
ess.
Sam
ead
vant
ages
over
wire
sens
ors
asfib
erfil
ms
with
thin
quar
tzco
atin
gbu
t hav
elo
wer
uppe
r
freq
uenc
ylim
itdu
eto
thei
rhe
avie
rqu
artz
coat
ing.
For
use
inco
nduc
ting
liqui
dsco
verin
gve
loci
ties
upto
10m
/s.
Go
ld-p
late
dw
ire
Pt-
plat
edtu
ngst
enw
ire, d
ia-
met
er5
µm, o
vera
llle
ngth
3m
m. S
ensi
tive
wire
leng
th
1.25
mm
. Cop
per
and
gold
plat
edat
the
ends
toa
dia-
met
erat
appr
ox. 1
5µm
.
Thi
sse
nsor
poss
esse
sth
esa
me
high
flow
sens
itivi
ty
and
wid
efr
eque
ncy
resp
onse
asth
em
inia
ture
wire
but f
eatu
res
cons
ider
ably
less
pron
g-to
-sen
sor
influ
ence
due
toits
gold
-pla
ted
wire
ends
. Pro
bes
with
gold
-pla
ted
wire
sens
ors
are
spec
ially
suite
dfo
r
use
inhi
gh-t
urbu
lenc
eflo
ws
and
cove
rve
loci
ties
up
toap
prox
. 300
m/s
inw
ell-f
ilter
edai
r.G
old-
plat
edw
irepr
obe:
55P
01
Fib
er-f
ilmpr
obes
:
Thi
nqu
artz
-coa
ting:
55R
01
Hea
vyqu
artz
-coa
ting:
55R
11G
old-
plat
edw
irepr
obe:
55P
02
Fib
er-f
ilmpr
obes
:
Thi
nqu
artz
-coa
ting:
55R
02
Hea
vyqu
artz
-coa
ting:
55R
12G
old-
plat
edw
irepr
obe:
55P
03
Fib
er-f
ilmpr
obes
:
Thi
nqu
artz
-coa
ting:
55R
03
Hea
vyqu
artz
-coa
ting:
55R
13G
old-
plat
edw
irepr
obe:
55P
04
Fib
er-f
ilmpr
obes
:
Thi
nqu
artz
-coa
ting:
55R
04
Hea
vyqu
artz
-coa
ting:
55R
14G
old-
plat
edw
irepr
obe:
55P
05
Fib
er-f
ilmpr
obes
:
Thi
nqu
artz
-coa
ting:
55R
05
Hea
vyqu
artz
-coa
ting:
55R
15
Fib
er-
film
wit
hth
in
qu
art
zco
ati
ng
Nic
kel f
ilmde
posi
ted
on70
µm
diam
eter
quar
tzfib
er. O
vera
ll
leng
th3
mm
, sen
sitiv
efil
m
leng
th1.
25m
m. C
oppe
ran
d
gold
plat
edat
the
ends
. Film
is
prot
ecte
dby
quar
tzco
atin
gap
prox
. 0.5
µmin
thic
k-
ness
. Fib
er-f
ilmse
nsor
spo
sses
shi
gher
robu
stne
ss
and
are
less
susc
eptib
leto
cont
amin
atio
nth
anw
ire
sens
ors
but h
ave
low
erfr
eque
ncy
limit.
Fib
er-f
ilmw
ithth
inqu
artz
coat
ings
are
for
use
in
gase
san
dno
n-co
nduc
ting
liqui
dsw
here
the
dem
and
for
robu
stne
ssel
imin
ates
the
use
ofw
irese
nsor
s.
May
beus
edat
ambi
ent t
empe
ratu
res
upto
150
C.
o
Pro
bes
for
hot-
wir
ean
emom
etry
Film
sen
so
rsw
ith
thin
qu
art
zco
ati
ng
Nic
kel f
ilms
depo
site
don
wed
ge-s
hape
d,co
nica
l or
flat
quar
tzro
ds. A
ppro
xim
ate
dim
ensi
onof
film
0.2
x1
mm
.
Film
ispr
otec
ted
byqu
artz
coat
ing
appr
ox. 0
.5µm
inth
ickn
ess.
Film
sens
ors
on
non-
cylin
dric
alsu
bstr
ates
are
very
robu
stin
com
-
paris
onw
ithbo
thw
irean
dfib
er-f
ilmse
nsor
sw
ithle
ss
tend
ency
tofo
ulin
cont
amin
ated
flow
s.
For
use
inga
ses
and
non-
cond
uctin
gliq
uids
at
mod
erat
ean
dhi
ghve
loci
ties
and
atam
bien
t tem
per-
atur
esup
to15
0C
.o
Film
sen
so
rsw
ith
heavy
qu
art
zco
ati
ng
Nic
kel f
ilms
depo
site
don
wed
ge-s
hape
d,co
nica
l and
flat
quar
tzro
ds. A
ppro
x.fil
m
dim
ensi
on0.
2x
1m
m. F
ilmis
prot
ecte
dby
quar
tzco
atin
g
appr
ox. 2
µmin
thic
knes
s.T
hese
sens
ors
poss
ess
the
sam
ech
arac
teris
tics
asfil
mse
nsor
sw
ithth
in
quar
tzco
atin
gbu
t hav
elo
wer
uppe
rfr
eque
ncy
limit
due
toth
eir
heav
ier
quar
tzco
atin
g.
For
mea
sure
men
tsin
cond
uctin
gliq
uids
atve
loci
ties
upto
appr
ox. 2
5m
/san
dat
ambi
ent t
empe
ratu
res
up
to15
0C
.o
Min
iatu
rew
ire
Pt-
plat
edtu
ngst
enw
ire, d
ia-
met
er5
µm, l
engt
h1.
25m
m.
The
sens
orpo
sses
ses
grea
t
flow
sens
itivi
tyan
dw
ide
freq
uenc
yre
spon
sein
turb
u-
lent
flow
.
Min
iatu
re-w
irepr
obes
are
mai
nly
inte
nded
for
use
in
low
-tur
bule
nce
flow
sof
tem
pera
ture
sup
to15
0C
and
cove
rve
loci
ties
upto
500
m/s
inpu
reai
r.o
Sen
so
rty
pes
Sin
gle
-sen
so
r,cylin
dri
cal
4m
msu
pp
ort
s
Gu
ide
tub
es
4m
mm
ou
nti
ng
tub
es
6m
msu
pp
ort
s
6m
mm
ou
nti
ng
tub
es
Sin
gle
-sen
so
r,n
on
-cylin
dri
cal
Min
iatu
rew
ire-p
robe
:55
P11
Min
iatu
rew
ire-p
robe
:55
P12
Wed
ge-s
hape
dpr
obe,
stra
ight
gene
ral-p
urpo
sety
peW
edge
-sha
ped
prob
e,90
sens
orpa
ralle
l to
prob
eax
is
o
Wed
ge-s
hape
dpr
obe,
90se
nsor
perp
endi
cula
rto
prob
eax
is
o
Con
ical
prob
e,st
raig
htge
nera
l-pur
pose
type
Flu
shm
ount
ing
prob
e
Cab
lele
ngth
appr
ox. 1
m
Min
iatu
rew
ire-p
robe
:55
P13
Min
iatu
rew
ire-p
robe
:55
P14
Min
iatu
rew
ire-p
robe
:55
P15
Str
aigh
t gen
eral
-pur
pose
type
45sl
antin
gse
nsor
type
o
90se
nsor
para
llel t
opr
obe
axis
o
90se
nsor
perp
endi
cula
r
topr
obe
axis
o
Bou
ndar
yla
yer
type
Gol
d-pl
ated
wire
prob
e:55
P51
Fib
er-f
ilmpr
obes
:
Thi
nqu
artz
-coa
ting:
55R
51
Hea
vyqu
artz
-coa
ting:
55R
61G
old-
plat
edw
irepr
obe:
55P
52
Fib
er-f
ilmpr
obes
:
Thi
nqu
artz
-coa
ting:
55R
52
Hea
vyqu
artz
-coa
ting:
55R
62G
old-
plat
edw
irepr
obe:
55P
53
Fib
er-f
ilmpr
obes
:
Thi
nqu
artz
-coa
ting:
55R
53
Hea
vyqu
artz
-coa
ting:
55R
63G
old-
plat
edw
irepr
obe:
55P
54
Fib
er-f
ilmpr
obes
:
Thi
nqu
artz
-coa
ting:
55R
54
Hea
vyqu
artz
-coa
ting:
55R
64
Gol
d-pl
ated
wire
prob
e:55
P91
Fib
er-f
ilmpr
obes
:
Thi
nqu
artz
-coa
ting:
55R
91
Tria
xial
prob
e,st
raig
htty
pe
Sen
sors
are
plac
edin
tria
xial
arra
yw
ith90
spac
ing.
o
55P
31R
esis
tanc
e-th
erm
omet
erpr
obe
Spe
cial
lysu
ited
for
mea
sure
men
tsof
high
-fre
quen
cyte
mpe
ratu
reflu
c-
tuat
ions
. The
sens
oris
a0.
4m
mlo
ng, 1
µmdi
a.pl
atin
umw
ire. T
hepr
obe
mat
ches
the
4m
msu
ppor
t and
mou
ntin
gtu
besy
stem
show
nab
ove.
55P
81Te
mpe
ratu
re-c
ompe
nsat
edpr
obe
For
mea
sure
men
tsof
mea
nflo
wve
loci
tyan
dflo
wflu
ctua
tions
inga
ses
at
vary
ing
tem
pera
ture
-w
ithte
mpe
ratu
reco
mpe
nsat
ion.
The
prob
em
atch
es
the
6m
msu
ppor
t and
mou
ntin
gtu
besy
stem
show
nab
ove.
Du
al-
sen
so
r,cylin
dri
cal
Oth
er
pro
bes
Tri
ple
-sen
so
r,cylin
dri
cal
Du
al-
sen
so
r,n
on
-cylin
dri
cal
Min
iatu
rew
ire-p
robe
:55
P61
Min
iatu
rew
ire-p
robe
:55
P62
Min
iatu
rew
ire-p
robe
:55
P63
Min
iatu
re w
ire-p
robe
:55
P64
Min
iatu
rew
ire-p
robe
:55
P71
V-ar
ray
prob
e, s
traig
ht ty
pe
Thi
nqu
artz
-coa
ting:
55R
71
Hea
vyqu
artz
-coa
ting:
55R
72P
aral
lel-a
rray
prob
e,90
,
sens
ors
perp
endi
cula
r
topr
obe
axis
o
X-a
rray
prob
e,se
nsor
plan
e
para
llel t
opr
obe
axis
X-a
rray
prob
e,se
nsor
plan
e
perp
endi
cula
rto
prob
eax
is
X-a
rray
prob
e,90
, sen
sor
plan
epa
ralle
l to
prob
eax
is
o
X-a
rray
prob
e,90
, sen
sor
plan
epe
rpen
dicu
lar
topr
obe
axis
o
55A
75H
igh-
tem
pera
ture
prob
e
Spe
cial
lysu
ited
for
mea
sure
men
tsin
hot g
ases
, up
to75
0C
.
Typi
cal a
pplic
atio
nsar
em
easu
rem
ents
onin
tern
alco
mbu
stio
nen
gine
san
dga
stu
rbin
es.
The
sens
oris
a2.
2m
mlo
ng, 1
0µm
dia.
Pt-
rhod
ium
wire
.
The
prob
een
dsin
a9
mm
thre
adpe
rmitt
ing
mou
ntin
gdi
rect
lyin
toa
guid
etu
be.
o
Thi
nqu
artz
-coa
ting:
55R
31
Hea
vyqu
artz
-coa
ting:
55R
32
Thi
nqu
artz
-coa
ting:
55R
33
Hea
vyqu
artz
-coa
ting:
55R
34
Thi
nqu
artz
-coa
ting:
55R
35
Hea
vyqu
artz
-coa
ting:
55R
36
Thi
nqu
artz
-coa
ting:
55R
41
Hea
vyqu
artz
-coa
ting:
55R
42
Thi
nqu
artz
-coa
ting:
55R
45
Hea
vyqu
artz
-coa
ting:
55R
46
Cab
lele
ngth
appr
ox. 1
m
55H
20
Cab
lele
ngth
appr
ox. 0
.8m
Cab
lele
ngth
appr
ox. 0
.75
m
55H
21(o
vera
llle
ngth
235
mm
)
55H
22(o
vera
llle
ngth
235
mm
)
Th
esh
ort
4m
msu
pp
ort
(Typ
e55H
20)
fits
into
bo
th
the
4m
mch
uck
an
dth
e4
mm
mo
un
tin
gtu
bes
sh
ow
nb
elo
w.
Th
esh
ort
6m
msu
pp
ort
(Typ
e55H
24)
fits
into
bo
th
the
6m
mch
uck
an
dth
e6
mm
mo
un
tin
gtu
bes
sh
ow
nb
elo
w.
55H
24
Cab
lele
ngth
appr
ox. 1
m
55H
25(o
vera
llle
ngth
235
mm
)
Cab
lele
ngth
appr
ox. 0
.8m
55H
26 (
over
all l
engt
h 23
5 m
m) C
able
leng
thap
prox
. 0.7
5m
55H
136
4m
mch
uck
55H
241
6m
mch
uck
55H
151
(ove
rall
leng
th21
5m
m)
55H
153
(ove
rall
leng
th21
5m
m)
55H137Guidetube,smooth,overalllength220mm.
55H139Guidetube,smooth,overalllength440mm.
55H
141
Gui
detu
beco
uplin
g
55H
530,
wat
ertig
ht6
mm
(ove
rall
leng
th21
5m
m)
55H
540,
wat
ertig
ht6
mm
(ove
rall
leng
th21
5m
m)
55H
238
(ove
rall
leng
th23
5m
m)
55H
240
(ove
rall
leng
th23
5m
m)
55H
27S
uppo
rt, s
trai
ght (
over
all l
engt
h22
6m
m)
Cab
lele
ngth
appr
ox. 0
.8m
55R
47G
lue-
onpr
obe
Sen
sor
0.1
x0.
9m
m
nick
elfil
mde
posi
ted
on
a0.
05m
mth
ick
poly
-
imid
efo
ilca
rryi
nga
0.5
µmqu
artz
coat
ing.
Pro
beis
glue
don
toth
e
wal
l,an
dca
ble
lead
ing
toth
ean
emom
eter
is
sold
ered
onto
the
copp
erle
ads.
www.dantecdynamics.com
~e1
~e1
~e2
~e2
~er
~er
Measurement plane
Probe axis
Prongs
Hot-Wires
~ex
~e�~�
�
(b)
Figure 2: Traverse azimutal positions (a) and hot-wire probe coordinate system (b).
placing the probe as close as possible to the hub (h/hmax = 0), and aligning the probe axis withthe engine axis. This is a delicate, manual operation that yields positioning uncertainties of1 mm in the radial direction and 1◦ in the azimuthal direction. Radial explorations have beencarried out at the fan inlet (station 2A, port B-08), between rotor and stator (station 2R, portB-01) and the fan outlet (station 21A, port B-06), as shown in Figure 2. The unsteady andturbulent components of the flow field are extracted using phase averaging procedure thanks tothe high frequency response of hot-wire probe. For the experiments performed, the samplingfrequency ranges from 50kHz at stations 2A and 2R, and up to 84kHz at station 21A. For eachmeasurement point, the sampling time is fixed to 50 rotor revolutions so that phase averagingis computed over 700 blade passing periods. The phase averaging is performed thanks to anoptical sensor capturing the rotor blade passings. The output signal of the optical encoder issampled at the same sampling rate as the hot-wire signals.
STEADY ANALYSISA steady analysis is performed by time averaging the velocity and flow azimuthal angle
at each location independently of the rotor blade passing period. This analysis specificallyhighlights the existence of flow separation on both the rotor and stator blades leading to thestagnation pressure loss that is typically found in windmilling conditions.
The time averaged Mach number profiles at stations 2A, 2R and 21A are shown in Figure 3.For each station, the results for the different windmilling operating points are displayed. Pro-files are non-dimensionalized with the engine inlet Mach number M2A, measured at mid-span(h/hmax = 0.5), in order to observe the inlet conditions dependency. Results are consistent withprevious work (García Rosa et al., 2015; Dufour et al., 2015). At engine inlet (Figure 3 (a)),the Mach number is uniform across the first 75% of the span, then decreases near the shroud.This is attributed to the large boundary layer that develops along the 4m long duct linking thewindmilling blower to the engine. Downstream of the rotor (Figure 3 (b)), the results show aflow acceleration compared to engine inlet 2A. A large part of this acceleration is due to stream-tube contraction, between 2A (upstream of the spinner) and the fan plane, and then across therotor. This effect is more pronounced close to the hub due to the meridional channel geometry(Figure 1). In windmilling, the inner sections of the fan operate in compressor mode, whilethe outer sections operate in turbine mode (Prasad and Lord, 2010; García Rosa et al., 2015).However, since the inlet flow is purely axial, the absolute velocity increases along the entire
4
Figure 3: Time-averaged Mach number profiles through the fan stage : (a) rotor inlet, (b)rotor/stator interface and (c) stator outlet.
span. This effect is more pronounced at the hub because the flow turning (in the absolute sense)is more important there due to the shape of the blade. Downstream of the stator (Figure 3 (c)),a strong velocity deficit is observed with a minimum Mach number located at h/hmax = 0.65.This corroborates the existence of another flow separation on the stator blades. According toDufour et al. (2015), this flow separation has a more complex topology than on the rotor bladeand takes the shape of a massive 3D separation. Indeed, the separation on the stator blade in-creases in size along the span and reaches a maximum at around 60% span, where the flowseparates on the leading edge without reattachment. A massive flow separation is thus observedat this position, leading to a high velocity deficit as observed in Figure 3.c. For the same testedengine in windmilling conditions, García Rosa et al. (2015) state that stagnation pressure lossesare encountered in the turbine-like mode zone along the fan stage, that is, for h/hmax > 0.4.More specifically, the massive flow separations that develop on stator blades, around 60% span,widely contribute to losses, which is consistent with the conclusions of Prasad and Lord (2010).Finally all the profiles presented in Figure 3 show a fairly good self-similarity with respect tothe range of tested inlet windmilling operating points. Indeed, the size of flow separations onboth the rotor and the stator as well as the strength of velocity deficit observed in the wakesare independent of the inlet flow parameter. The presence of flow separations on both the rotorand the stator blades can be explained by highly negative incidences in windmilling conditions.Time averaged profiles of incidence at the rotor and the stator inlets are presented in Figure 4.In both cases, the incidences are negative along the whole span, which is likely to generateseparation on concave surfaces of the rotor and stator blade. Additionally, it can be observedthat the incidence is less negative on the rotor than on the stator, which suggests that the flowseparation is more pronounced in the latter, as sketched in Figure 5. Remarkably, a local mini-mum of incidence is observed around h/hmax = 0.65, which matches the maximum of velocitydeficit in the stator wake as described in Figure 3 (c). This explains the larger size of the stator
5
separation around 60% span observed by Dufour et al. (2015). Finally, similarly to the Machnumber profiles, a fairly good self-similarity of incidence profiles is observed in the range ofinlet flow parameters being tested.
Figure 4: Incidence through the fan stage : (a) rotor inlet, (b) stator inlet. βrotor and αstator
are the metal angles of the rotor and stator leading edge, respectively.
UNSTEADY AND TURBULENT FLOW STRUCTURESThe existence of flow separation on both the rotor and stator in windmilling conditions
leads to complex unsteady interaction between the rotor and stator wakes. The recent numericalstudy of Dufour et al. (2015) has shown that the rotor wake triggers a periodic movement ofthe stator flow separation leading to vortex shedding. In order to confirm experimentally thismechanism, the hot-wire signals have been post-processed using phase averaging method assuggested by Fernández Oro et al. (2007). In this case, each time series of velocity v(r, t) (andflow angle α(r, t)) is decomposed into a periodic, deterministic component V(r, t), and a randomcomponent v′(r, t), as per
v(r, t) = V(r, t) + v′(r, t), (2)
where the deterministic component V(r, t) is obtained by phase averaging over the Ntr = 700rotor blade passing periods. This is defined by
V(r, t) =1
Ntr
Ntr∑n=0
v(r, t + n · Tr), (3)
where Tr is the blade passing period. Note that in order to get in-phase informations at allmeasurement locations, a reference time is defined and locked to a blade passing signal by
6
26
49
Nominal
Windmillin
g
u
Windmilling
Nominal
�e�
�ex
ROTOR STATOR
�
cx
c�
�
c
w
separation separation
Figure 5: Velocity triangle between rotor and stator.
triggering the data acquisition with the optical blade detector signal. Using this formulation,the unsteady component of the velocity V(r, t) can be decomposed into
V(r, t) = V(r) + V(r, t), (4)
where V(r) is the steady time averaged component and V(r, t) is the fluctuating component.From this formulation, the level of velocity fluctuations around the phase averaging can beexpressed as as unsteady turbulence intensity Iv(r, t) by
Iv(r, t) =
√v′2
V(r, t)=
1
V(r, t)
1Ntr
Ntr∑n=0
v′2(r, t + n · Tr)
1/2 (5)
Contours of the phase averaged unsteady component of Mach number M, as a function ofthe blade passing period Tr is plotted in Figure 6 for both the rotor and stator outlet. In Figure 6(a), the periodic emission of the rotor wake is clearly visible and is distinguished by periodicdiagonal stripes of low Mach number for h/hmax < 0.6. Moreover, for h/hmax > 0.6 a periodicpassing of strong velocity deficit in line with the diagonal stripes is observed.
Interestingly, this strong velocity deficits are located along the last 40% span, close to therotor blade tip where the separation bubble on the rotor blade roughly fills the last 75% of chord,as pointed out by Dufour et al. (2015) through numerical simulation. This suggests that the pe-riodic high-velocity deficit observed near the shroud may correspond to the periodic emissionof wakes generated by the flow separation on the rotor blades. Downstream of the stator (Fig-ure 6 (b)) another unsteady process is observed. Contours of unsteady Mach number displaylarge periodic velocity deficit patches being convected downstream of the stator blades. Thesepatches are located between h/hmax = 0.6 and h/hmax = 0.8, with maximum of velocity deficitsobserved at h/hmax ≈ 0.65, which corresponds to the position of the massive flow separation onthe stator blade, as concluded in steady analysis. In this case, the periodic signature of strongvelocity deficit downstream of the stator blades seems to be correlated to a periodic convectionof the stator flow separation wake in the form of a vortex shedding. Finally, the observed pe-riodicity of this process is connected to the rotor blade passing period which confirms that theunsteady wake of the stator is modulated by the rotor wake.
7
Figure 6: Contours of the phase averaged Mach number phase locked with the rotor bladepassing for Φ2 = 0.18 : (a) rotor outlet (2R), (b) stator outlet (21A).
The rotor/stator wakes modulation has been explained numerically by Dufour et al. (2015),stating that the stator vortex shedding is triggered by a periodic variation of the stator incidence,under the periodic passing of the rotor wakes. Experimental results presented in Figure 7 con-firm this description. Figure 7.a shows the phase averaged unsteady components of the absoluteflow angle α at the stator inlet and Figure 7.c presents its associated time series at the specificheight h/hmax = 0.65, where vortex shedding from stator is observed. It is observed that thepassing of the rotor blade wake generates a periodic variation of the absolute flow angle. Morespecifically, for h/hmax > 0.4, negative values of flow angle are observed with high level of peri-odic fluctuations at h/hmax = 0.65, certainly associated with the existence of flow separation onthe rotor blades. At h/hmax = 0.65, the time averaged component of the flow angle α displays asevere off-design value leading to flow separation on the stator blade as suggested in the steadyanalysis (is = −42o). However, the unsteady component of flow angle α fluctuates from highlyto weakly negative values leading to a periodic fluctuation of the stator incidence. This confirmsthe findings of Dufour et al. (2015) that this unsteady mechanism triggers a periodic movementof the reattachment point at the trailing edge of the stator, leading to vortex shedding.
The phase averaged unsteady component of turbulent intensity Iv downstream of the stator ispresented in Figure 7.b while its associated time series at h/hmax = 0.65 is shown in Figure 7.d.At stator outlet, contours of Iv display periodic patterns of high turbulence levels (up to 21%)in phase with the strong velocity deficit described in Figure 6.b and located at h/hmax = 0.65.Consequently, it can be said that vortex shedding from the stator is connected to the highly tur-bulent patches periodically convected downstream. By considering that the centers of convectedlarge scale structures are associated to maximum of Iv at h/hmax = 0.65, it can be observed thatfrequency signature of the stator vortex shedding matches the rotor blade passing frequency assuggested by Dufour et al. (2015).
Thanks to phase averaging procedure, turbulent velocity fluctuations are obtained by sub-tracting the unsteady component from velocity time series. Therefore, the steady turbulenceintensity is assessed using
Iv(r) =
√v′2(r)
V(r), (6)
8
Figure 7: Contours of the phase averaged absolute flow angle and turbulence intensity andphase averaged time series for Φ2 = 0.18 : (a) contours of α at rotor outlet (2R),(b) contours of Iv at stator outlet (21A), (c) time series of α at h/hmax = 0.65 atstation (2R), (d) time series of Iv at h/hmax = 0.65 at station (21A).
and the turbulent time-scale Lt is obtained from the auto-correlation function,
Rvv(r, τ) =v′(r, t) · v′(r, t + τ)
v′2, (7)
where τ is the correlation time lag. Lt is then assessed by integrating Rvv(r, τ) until the first zerocrossing.
Profiles of turbulence intensity Iv are shown in Figure 8 at the different stations 2A, 2R, 21Aalong the fan stage. Turbulence intensity profiles show that the level of turbulence is globallyunchanged across the fan stage, along the first 20% of the blade span (Iv ' 3%). Turbulenceis uniform at the inlet (2A) for h/hmax < 0.7, despite a higher level observed near the shroud,caused by the development of the upstream boundary layer. Downstream of the rotor (2R)increasing values, up to 12%, are observed in the last 50% of the span. This is caused bythe rotor flow separation which generates a shear layer downstream of the rotor, producingturbulence. At stator outlet (21A) substantial levels of turbulence (up to 18%) are observed ath/hmax = 0.65, which is consistent with the vortex shedding mechanism previously described.
Figure 9 presents the profiles of the Tr/Lt ratio across the fan stage, for the different inletwindmilling conditions. The ratio Tr/Lt gives the number of turbulent scales being convectedduring a blade passing period. The results demonstrate a fairly good self-similarity, which in-dicates that turbulent timescales across the fan stage scale well with the blade passing period in
9
Figure 8: Turbulence intensity profiles across the fan stage.
Figure 9: Turbulent time scale profiles across the fan stage.
10
windmilling conditions. Moreover, turbulent time scale profiles are consistent with the varia-tion of turbulence intensity. Indeed, the number of turbulent scales increases in the regions ofhigh turbulence intensity, especially in the wake of both the rotor and the stator. Interestingly,the Tr/Lt profiles downstream of the stator display a sharp increase for h/hmax > 0.3, in accor-dance with the presence of the stator flow separation. However, the smallest timescales (ie. themaximum of Tr/Lt) are observed at h/hmax = 0.4, while the maximum of turbulence intensityis found at h/hmax = 0.65. These results are not still completely understood and need furtherinvestigations.
CONCLUSIONAn experimental investigation of the unsteady, turbulent flow through the fan stage of a by-
pass turbofan in windmilling condition has been performed using hot-wire anemometry. Thesteady analysis of the flow corroborates the existence of flow separations on both the stator andthe rotor, independently of the inlet flow parameter. The massive flow separation on the statorblades is found to be triggered by highly negative incidence. The unsteady characterization ofthe flow field, through phase averaging, allows the observation of vortex shedding downstreamthe stator. Results demonstrate that the periodic variation of stator incidence modulates thevortex shedding downstream the stator, with a frequency signature linked to the blade pass-ing frequency. Flow separations through the fan stage generate higher level of turbulence andsmaller turbulent length scales. Finally, vortex shedding downstream of the stator is captured,in the form of a periodic advection of highly turbulent patches.
ACKNOWLEDGEMENTThe authors would like to thank Price Induction for providing technical advice and engine
data.
REFERENCESBraig, W., Schulte, H., and Riegler, C. (1999). Comparative analysis of the windmilling perfor-
mance of turbojet and turbofan engines. Journal of Propulsion and Power, 15(2):326–333.
Dufour, G., García Rosa, N., and Duplaa, S. (2015). Validation and flow structure analysis in aturbofan stage at windmill. IMechE Journal of Power and Energy.
Fernández Oro, J. M., Argüelles Díaz, K. M., Santolaria Morros, C., and Blanco Marigorta,E. (2007). On the structure of turbulence in a low-speed axial fan with inlet guide vanes.Experimental Thermal and Fluid Science, 32(1):316–331.
Fuksman, I. and Sirica, S. (2012). Real-Time Execution of a High Fidelity Aero-Thermodynamic Turbofan Engine Simulation. Journal of Engineering for Gas Turbines andPower, 134(5):054501.
García Rosa, N., Dufour, G., Barènes, R., and Lavergne, G. (2015). Experimental analysis of theglobal performance and the flow through a high-bypass turbofan in windmilling conditions.Journal of Turbomachinery, 137(5).
Jorgensen, F. E. (2002). How to measure turbulence with hot-wire anemometers, a practicalguide. Dantec Dynamics.
11
Mishra, R. K., Gouda, G., and Vedaprakash, B. S. (2008). Relight Envelope of a Military GasTurbine Engine: An Experimental Study. ASME Conference Proceedings, 2008(43116):55–60.
Prasad, D. and Lord, W. K. (2010). Internal losses and flow behavior of a turbofan stage atwindmill. Journal of Turbomachinery, 132(3):031007–1–10.
Riegler, C., Bauer, M., and Schulte, H. (2003). Validation of a mixed flow turbofan performancemodel in the sub-idle operating range. In ASME Conference Proceedings, number GT2003-38223, pages 83–90. ASME.
Wallner, E. E. and Welna, H. J. (1951). Generalization of turbojet and turbine-propeller en-gine performance in windmilling condition. Technical Report NACA-RM-E51J23, NationalAdvisory Committee for Aeronautics.
Zachos, P. K., Grech, N., Charnley, B., Pachidis, V., and Singh, R. (2011). Experimental andnumerical investigation of a compressor cascade at highly negative incidence. EngineeringApplications of Computational Fluid Mechanics, 5(1):26–36.
12