aerospace Article

Uncertainty Quantification of the Effects o

Article
ertainty Quantification Manufacturing
of the Effects
Uncertainty Quantification ofthe
of Small Deviations
Effects of Smallon Film Coolin
nufacturingManufacturing FilmACooling:
Deviations onDeviationsFan-Shaped Hole
on Film Cooling:
an-ShapedAHole
Fan-Shaped Hole Wei Shi, Pingting Chen, Xueying Li *, Jing Ren and Hongde Jiang
Department of Energy and Power Engineering, Tsinghua University, Beijing 10084, Ch
i, Pingting Chen,Wei
Xueying , Pingting
Shi Li Chen,
*, Jing Ren andXueying
HongdeLiJiang
*, Jing Ren and Hongde Jiang
shiwei15@mails.tsinghua.edu.cn (W.S.); cpt13@mails.tsinghua.edu.cn (P.C.); renj@tsin
Department of Energy and Power Engineering, Tsinghua University, Beijing
jianghd@mail.tsinghua.edu.cn 10084, China;
(H.J.)
ment of Energy and Power Engineering, Tsinghua University, Beijing 10084, China;
shiwei15@mails.tsinghua.edu.cn (W.S.); cpt13@mails.tsinghua.edu.cn
* Correspondence: (P.C.); renj@tsinghua.edu.cn
li_xy@tsinghua.edu.cn; (J.R.);
Tel.: +86-151-2000-4131
15@mails.tsinghua.edu.cn (W.S.); cpt13@mails.tsinghua.edu.cn (P.C.); renj@tsinghua.edu.cn (J.R.),
jianghd@mail.tsinghua.edu.cn (H.J.)
d@mail.tsinghua.edu.cn (H.J.) Received: 19 February 2019; Accepted: 12 April 2019; Published: date
* Correspondence: li_xy@tsinghua.edu.cn; Tel.: +86-151-2000-4131
respondence: li_xy@tsinghua.edu.cn; Tel.: +86-151-2000-4131




ed: 19 February 2019;Received:
Accepted:19
12February
April 2019; Published:
2019; date
Accepted: Abstract:
12 April The film19cooling
2019; Published: holes in the blade of

April 2019 modern gas turbines h
manufactured by laser drilling, Electric Discharge Machining (EDM), and Ad
act: The film cooling Abstract:
holes inThethefilm cooling
blade holes ingas
of modern (AM)
the in recent
blade
turbines ofhave years.
modern These manufacturing
gas turbines
commonly been have commonly processesbeen often result in small
manufactured by laser drilling, Electric such
Dischargeas conical
factured by laser drilling, Electric Discharge Machining (EDM), and Additive ManufacturingMachining angles, filleted
(EDM), and edges, and
Additive diameter
Manufacturing deviations of the hol
(AM)manufacturing
in recent years. These in recent years. These manufacturing
processes often result deviations smallongeometric
in processes theoften
distribution
result inofsmall
deviations, adiabatic coolingdeviations,
geometric effectiveness (η) values, the
such as conical
as conical angles, filleted edges, angles, filleted edges,
and diameter andcoefficient
deviations diameter (Cd), and
deviations
of the hole, which the characteristic
ofcan
the hole, to ofcan
lead which thelead
in-hole flow field. The current stud
to deviations
on the
ions on the distribution ofdistribution of adiabatic
adiabatic cooling cooling(η)
effectiveness fan-shaped
effectiveness
values, the (η)film
value cooling
values,
of thethe holes
value of the discharge coefficientvalues (L/D) equ
discharge with length-to-diameter
(Cd ), and theof
cient (Cd), and the characteristic characteristic
the in-hole flowof thefield. Theinvestigate
in-hole flow field.
current the employed
The
study effects of
current these
study manufacturing
flatemployed
plate flat platedeviations
fan-shaped on the distribution
aped film coolingfilm holescooling
with holes with length-to-diameter
length-to-diameter values of C d, and
(L/D) the (L/D)
values
equal characteristic
equal
to 3.5 andtoofsix
in-hole
3.5 and flow
to six tofield. An Uncertainty
investigate the Quantificatio
igate the effects of effects of these manufacturing
these manufacturing on the the
deviationsdeviations Polynomial
on of ηChaos
the distribution
distribution values,of η values,
Expansion
the value(PCE) model
the value of C was carried
d , and the out to quantify t
and the characteristiccharacteristic of in-hole
of in-hole flow flowUncertainty
field. An values of ηQuantification
field. An Uncertainty
Quantification and(UQ) Cd. The statistical
(UQ)using
analysis characteristics
analysis using the (mean
Polynomialvalues, standard devia
Chaos Expansion
olynomial Chaos Expansion (PCE) model(PCE)was model wasout
carried Probability
carried out to Density
to quantify quantify Function in(PDF)
the uncertainty
the uncertainty the values) of η and
in the values ofCηd and
were also calculated.
Cd . The characteristics
s of η and Cd. The statistical statistical characteristics
(mean values, (mean conical
values,
standard angle deviations
standard
deviation exert no
deviation
(Std) values, visible
(Std)
and changes
values, on the value of η. However, t
and Probability
Density
bility Density Function Function
(PDF) values)(PDF) values)
of η and of η and
Cd were alsoby d 6.2%
Ccalculated.
were when
also the
calculated.
The conical
results The
show angle
results
that changes
show that from
conical 0–0.5°.
angle The area averag
deviations
al angle deviations exert exertchanges
no visible no visible onchanges
the value onofthe value of η. the
effectiveness
η. However, (𝜂̿ ) decreases
However,
Cd valuethe by 3.4%,
decreases while the C d increases
Cd value decreases by 6.2% when by 15.2% with the f
2% when the conical angle angle
the conical changes fromfrom
changes 0–0.5°.
0–0.5 ◦
The existing
area
. The area alone. However,
averaged
averaged adiabatic
adiabatic the deviation
cooling value of 𝜂̿ decreases
effectiveness
veness (𝜂̿ ) decreases by 3.4%, while the C increases by 15.2% with the filleted edge deviation
by 3.4%, while the Cd increases by 15.2% with the filleted edge deviation existing alone. However,
d
Keywords: manufacturing deviations; film cooling; uncertainty quantification
ng alone. However, the the deviation
deviation value
value of of 𝜂̿ is 7.6%, and that of Cd is 25.7% with the filleted edge deviation and the
diameter deviation existing.
ords: manufacturing deviations; film cooling; uncertainty quantification; CFD
Keywords: manufacturing deviations; film cooling; uncertainty quantification; CFD
1. Introduction
Modern gas turbines are widely used for aircraft propulsion, land-based p
duction in other industrial applications. To achieve progressively higher efficiency valu
1. Introduction
odern gas turbines are widely used for aircraft propulsion,the averaged power
land-based turbinegeneration,
inlet temperature
and (TIT) is raised higher with each successi
gas turbines are widely usedthus
forincreasing
aircraft the
propulsion,cooling
r industrial applications. To achieve progressively higher efficiency values and higher power, power
Modern demand
land-based of the turbine blade.
generation, and Bunker
in [1] determine
raged turbine inletother industrial
temperature applications.
(TIT) withineach
To achieve
is raised higher half when the engine
progressively
successive temperature
higher increases
efficiency
generation, values byand
25 K, which
higher demonstrated the imp
power,
creasing the coolingthedemand
averaged of turbine inletblade.
the turbine temperature turbine
Bunker(TIT) cooling
is raised
[1] determined techniques.
higher
that with each
blade The
life film cooling
issuccessive
cut technique
engine generation, is one of the most ef
thus increasing the cooling demand of cooling
the turbine techniques
blade.
when the temperature increases by 25 K, which demonstrated the importance of improving Bunker applied
[1] to turbine
determined cooling.
that blade The
life is film
cut in cooling holes h
halfThe
cooling techniques. when thecooling
film temperature increases
technique by 25
is one manufactured
of K,
thewhich by laser
mostdemonstrated
effective drilling,
and the Electric of
importance
beneficial Discharge
improving Machining
turbine (EDM), and Ad
cooling
techniques applied to techniques.
turbine cooling. The film
The cooling (AM) in
technique
film cooling recent
holes is one
have years.
theThese
of commonly manufacturing
most effective processes
been and beneficial often introduce micro-
cooling
techniques
actured by laser drilling, applied
Electric to turbine
Discharge e.g.,
cooling.(EDM),
Machining conical
The filmand coolingangles, filleted
have commonly been manufactured by which may chan
holesManufacturing
Additive edges, and diameter deviations,
lasermanufacturing
n recent years. These drilling, Electricprocesses
DischargeoftenMachiningadiabatic
introduce(EDM), cooling
and Additive
micro-geometric effectiveness (η) values,
Manufacturing
deviations, (AM)the value years.
in recent of the discharge coe
These manufacturing processes often characteristic
introduce
nical angles, filleted edges, and diameter deviations, which may change of
micro-geometric the in-hole flow
deviations,
distribution of field.
e.g., Thus,
conical the
angles, influence
filleted of film cooling
edges, and
ic cooling effectiveness (η) diameter
values, the deviations,
value ofwhich deviations
the may on
thethe
changecoefficient
discharge heat(Ctransfer
distribution performance
of adiabatic
d), and the and
cooling aerodynamic characteristics
effectiveness
(η) values,
eristic of the in-hole the value
flow field. Thus, of the discharge
influence coefficient
of film cooling(Cd ), and
holethe characteristic of the in-hole flow field.
manufacturing
Thus, the
ons on the heat transfer influence ofand
performance filmaerodynamic
cooling hole manufacturing
Aerospace 2019, 6, deviations
characteristics x;must on the
be PEER
doi: FOR determined.heat transfer performance and
REVIEW www.m

2019, 6, x; doi: FOR PEER REVIEW www.mdpi.com/journal/aerospace

Aerospace 2019, 6, 46; doi:10.3390/aerospace6040046 www.mdpi.com/journal/aerospace
Aerospace 2019, 6, 46 2 of 17
Aerospace 2019, 6, x FOR PEER REVIEW 2 of 17

aerodynamic
These deviationscharacteristics must be determined.
can be determined quantitatively Theseby deviations
Uncertainty canQuantification
be determined(UQ) quantitatively
analysis,
by Uncertainty Quantification
which will be introduced later. (UQ) analysis, which will be introduced later.
This study
This studyfocuses
focuseson onthetheslight
slight geometry
geometry variations,
variations, such such as conical
as conical angles,angles, filleted
filleted edges, edges,
and
and diameter deviations of the hole, caused by manufacturing processes
diameter deviations of the hole, caused by manufacturing processes of fan-shaped film holes. Bunker of fan-shaped film holes.
Bunker
[2] and an[2] open
and an open[3]report
report from [3] from North
PRIMA PRIMA North America,
America, Inc. both Inc. both the
collected collected
geometry the geometry
deviation
data by laser drilling techniques. They both pointed out that the statistical data of a film of
deviation data by laser drilling techniques. They both pointed out that the statistical data holea film
are
hole are subjected to a Gaussian distribution even with the most advanced
subjected to a Gaussian distribution even with the most advanced laser drilling techniques (see laser drilling techniques
(see Figure
Figure 1a). Generally,
1a). Generally, the laser
the laser beam beam
energyenergy is subjected
is subjected to a to a Gaussian
Gaussian distribution
distribution [2,3].[2,3].
Thus, Thus,
the
the diameter of the parallel hole entrance is different from that of the exit. This
diameter of the parallel hole entrance is different from that of the exit. This will cause the film hole will cause the film hole
obtained through
obtained through laser
laser drilling
drillingor orEDM
EDMtotobecome
becometrumpet-like
trumpet-likeininshapeshape [4,5].
[3,5].InInaddition,
addition, thethe
fillet in
fillet
thethe
in manufacturing
manufacturing rootroot
is a common
is a common feature in theinreal
feature theprocess. Conical
real process. angle deviations
Conical always always
angle deviations exist in
holes manufactured by the laser drilling method, but can be eliminated by
exist in holes manufactured by the laser drilling method, but can be eliminated by EDM (see Figure EDM (see Figure 1b) or other
manufacturing methods. However, the processing space of the blade is extremely
1b) or other manufacturing methods. However, the processing space of the blade is extremely small, small, especially for
the blade with
especially double-wall
for the blade with cooling, to applycooling,
double-wall EDM, which makes
to apply the conical
EDM, which angle
makesdeviations
the conical stillangle
exist
in the present blade. Sridharan et al. [6] and Montomoli et al. [7] studied the
deviations still exist in the present blade. Sridharan et al. [6] and Montomoli et al. [7] studied the effect of the filleted edge
on theoffilm
effect thecooling
filleted performance
edge on the film using the advanced
cooling performancetripod holethe
using andadvanced
fan-shaped hole,hole
tripod respectively,
and fan-
while their conclusions were a little different. In addition, Cerantola et al.
shaped hole, respectively, while their conclusions were a little different. In addition, Cerantola [8] also studied the effect of
et al.
the fillet caused by AM on the cylindrical film hole.
[8] also studied the effect of the fillet caused by AM on the cylindrical film hole.

(a) (b)
Figure 1. Real
Real manufacturing
manufacturing film
film hole data. (a)
(a) Hole
Hole diameter
diameter deviation
deviation distribution
distribution [3]; (b) conical
angle deviations [3].

investigators evaluated
Some investigators evaluated the the effect
effect of film cooling hole geometrical deviations deviations on the heat
transfer performance and aerodynamic characteristics. Johnson et
performance and aerodynamic characteristics. Johnson et al. [9] studied al. [9] studied the impact theofimpact
geometric of
deviations deviations
geometric on the distribution of adiabatic cooling
on the distribution effectiveness
of adiabatic coolingvalues (η) of a row
effectiveness valuesof cylindrical
(η) of a row holesof
using both experimental
cylindrical holes using and bothCFD methods. These
experimental and cylindrical
CFD methods. holes These
were manufactured
cylindrical holes usingwerefour
different manufacturing
manufactured using four techniques.
differentThe results indicated
manufacturing that the manufacturing
techniques. technique can
The results indicated thathave
the
a noticeable effect,
manufacturing either negatively
technique can have by the impedance
a noticeable effect,of either
irregularities
negativelyor positively by creatingofa
by the impedance
diffusion-shaped
irregularities “cylindrical”
or positively hole. Cerantola
by creating and Birk [8]
a diffusion-shaped comparedhole.
“cylindrical” a sample laser-drilled
Cerantola and Birkhole [8]
against cylindrical,
compared a samplenozzled, diffusing,
laser-drilled hole and fileted
against holes assuming
cylindrical, nozzled, adiabatic
diffusing,walls
andusing
filetedtheholes
CFD
method. The
assuming diffusing
adiabatic wallshole wasthe
using found
CFD to deliverThe
method. the diffusing
best film hole coolingwasdue
foundto the lowest the
to deliver effluent
best
velocity and greatest amount of in-hole turbulent production; while the investigated
film cooling due to the lowest effluent velocity and greatest amount of in-hole turbulent production; laser-drilled hole
exhibited
while the similar cooling
investigated and discharge
laser-drilled performance
hole to the simplified
exhibited similar cooling and nozzled holes.performance
discharge S. Haydt et al. to [10]
the
examined the
simplified potential
nozzled holes.impact
S. Haydtof the manufacturing
et al. [10] examined defect on the film
the potential cooling
impact effectiveness
of the manufacturing for a
well-characterized
defect fan-shaped
on the film cooling hole known
effectiveness for a as the 7-7-7 hole. The
well-characterized meter and
fan-shaped holediffuser
known of as the
the holes
7-7-7
were The
hole. manufactured by EDMof
meter and diffuser inthe
two separated
holes steps, which led
were manufactured to an in
by EDM offset
twobetween
separated the meter
steps, and
which
diffuser. Several types and sizes of the offset were studied, and the offset was generally
led to an offset between the meter and diffuser. Several types and sizes of the offset were studied, detrimental to
the cooling
and performance,
the offset was generally which indicated
detrimental to that gas turbine
the cooling manufacturers
performance, which should
indicatedtrythat
to minimize
gas turbine or
eliminate the meter-diffuser
manufacturers should try to offset.
minimize or eliminate the meter-diffuser offset.
Currently, the application of Uncertainty Quantification (UQ) analysis in gas turbines has been
carried out by researchers due to the improvement of CFD capability [11]. UQ research on the CFD
methods aims to increase and maintain engine efficiency under variability [12]. The CFD Vision 2030
Aerospace 2019, 6, 46 3 of 17

Currently, the application of Uncertainty Quantification (UQ) analysis in gas turbines has been
Aerospace 2019, 6, x FOR PEER REVIEW 3 of 17
carried out by researchers due to the improvement of CFD capability [11]. UQ research on the CFD
methods aims to increase and maintain engine efficiency under variability [12]. The CFD Vision
Study [13] of NASA revealed that UQ will be an important direction of CFD in the future. Montomoli
2030 Study [13] of NASA revealed that UQ will be an important direction of CFD in the future.
et al. [7] studied the effect of filleted edge variations on the distributions of the η value and the Cd
Montomoli et al. [7] studied the effect of filleted edge variations on the distributions of the η value
values. Over 10% variations of the η and Cd values were found for the cases with the filleted edge
and the Cd values. Over 10% variations of the η and Cd values were found for the cases with the
from the cases without the filleted edge. Bunker [2] applied Monte Carlo simulations to consider the
filleted edge from the cases without the filleted edge. Bunker [2] applied Monte Carlo simulations to
effect of geometrical parameter and operational parameter variations on thermal boundary
consider the effect of geometrical parameter and operational parameter variations on thermal boundary
conditions for a typical highly-cooled turbine blade. The results showed that the factors determined
conditions for a typical highly-cooled turbine blade. The results showed that the factors determined by
by the manufacturing of film holes exerted the most important impact levels on the metal
the manufacturing of film holes exerted the most important impact levels on the metal temperature
temperature distribution and the aerodynamic performance of the cascade.
distribution and the aerodynamic performance of the cascade.
To extend our understanding of the influence of micro-manufacturing features on the heat
To extend our understanding of the influence of micro-manufacturing features on the heat
transfer performance and aerodynamic characteristics, the current study investigates the impact of
transfer performance and aerodynamic characteristics, the current study investigates the impact of
the small manufacturing deviations of a baseline 7-7-7 fan-shaped film hole on the distribution of
the small cooling
adiabatic manufacturing deviations
effectiveness of a baseline
(η) values 7-7-7 fan-shaped
and the value film hole
of the discharge on the(C
coefficient distribution
d) using CFD
of
adiabatic cooling effectiveness (η) values and the value of the discharge coefficient
methods. The detailed in-hole flow fields are also analyzed. To quantify the uncertainty (C ) using CFD
d of η and Cd
methods. The detailed in-hole flow fields are also analyzed. To quantify the uncertainty
values due to the manufacturing uncertainty, an UQ analysis is performed using the Polynomial of η and
C d values due to the manufacturing uncertainty, an UQ analysis is performed using the
Chaos Expansion (PCE) model. The statistical characteristics (mean values, Standard deviation (Std) Polynomial
Chaos Expansion
values, (PCE) model.
and Probability DensityThe statistical
Function characteristics
(PDF) values) of (mean
η and values,
Cd are Standard deviation
also calculated. All(Std)
the
values, and in
simulations Probability
this paper Density
used theFunction
commercial(PDF)
CFDvalues) of ηCCM+
code Star and Cwith
d arethe
also calculated.
realizable All the
two-layer k-
εsimulations
turbulencein this paper used the commercial CFD code Star CCM+ with the realizable two-layer k-ε
model.
turbulence model.
2. Computational Setup and Validation
2. Computational Setup and Validation
2.1.
2.1. Geometry
Geometry Models
Models
Small
Small manufacturing
manufacturing deviations
deviations were
were applied
applied to
to aa baseline
baseline fan-shaped
fan-shaped hole
hole geometry
geometry called
called the
the
7-7-7
7-7-7 fan-shaped hole [14,15] to study their effect on film cooling. This 7-7-7 fan-shaped hole
fan-shaped hole [14,15] to study their effect on film cooling. This 7-7-7 fan-shaped hole was
was
developed
developed by by Schroeder
Schroeder and
and Thole
Thole [14,15]
[14,15] as
as aa benchmark
benchmark shape
shape of
of aa hole.
hole. The
The detailed
detailed geometry
geometry ofof
the
the hole is shown in Figure 2. Two different baseline 7-7-7 fan-shaped holes with L/D equal to
hole is shown in Figure 2. Two different baseline 7-7-7 fan-shaped holes with L/D equal to 3.5
3.5 and
and
6.0
6.0 were
were employed
employed inin the
the current study to
current study to evaluate
evaluate the
the effect
effect of
of hole
hole length.
length.

Figure
Figure 2.
2. Shaped
Shaped hole
hole [14,15].
[14,15].

Three manufacturing deviations

deviations of
of the
the fan-shaped
fan-shaped hole,
hole, conical
conical angles
angles (θ),
(θ), filleted
filleted edges
edges (fillet),
(fillet),
and diameter deviations (ΔD),
(∆D), as shown in Figure 3, were studied. TheThe hole
hole diameter
diameter deviation
deviation only
only
occurred in the hole’s cylindrical section, which is always drilled using a low-cost manufacturing
process, e.g., laser drilling. The diffuser section of the hole does not always feature this size deviation
due to higher precision manufacturing processes.
Aerospace 2019, 6, 46 4 of 17

occurred in the hole’s cylindrical section, which is always drilled using a low-cost manufacturing
process, e.g., laser drilling. The diffuser section of the hole does not always feature this size deviation
Aerospace
Aerospace2019,
2019,6,6,xxFOR
FORPEER
PEERREVIEW
REVIEW 44 of
of 17
17
due to higher precision manufacturing processes.

Figure
Figure3.3.3.The
Figure The
The three
three
three manufacturing
manufacturing
manufacturing deviation
deviation
deviation factors
factorsfactors of
of the
the fan-shaped
of the fan-shapedfan-shaped hole
hole
hole conical conical
(θ), angles
conical
angles angles
filleted
(θ),
(θ),filleted
edges (fillet),edges
filleted edges (fillet),
(fillet),and
and diameter anddiameter deviations
deviations(D).
diameter(D).
deviations (D).

In
In order
order to make aa clear
to make clear distinction
distinction for the hole
for the hole geometric
geometric model
model with
with the
the small
small geometric
geometric
deviation,
deviation, 000 shows
deviation, shows the
shows the schematic
the schematic
schematic of of the
of the hole
the hole geometry
hole geometry with
geometry with hole
with hole diameter
hole diameter deviations
diameter deviations (Figure 4a),
deviations (Figure 4a),
conical
conicalangle
angle(Figure
(Figure4b),
4b),and filletededges
andfilleted edges(Figure
(Figure4c).
4c).

Figure
Figure4.
Figure 4.4.Schematic
Schematicof
Schematic ofthe
of thehole
the holegeometry:
hole geometry:(a)
geometry: (a)hole
(a) holegeometry
hole geometrywith
geometry withhole
with holediameter
hole diameterdeviations;
diameter deviations;(b)
deviations; (b)hole
(b) hole
hole
geometry
geometry with
withconical
conical angle;
angle;(c)
(c)hole
holegeometry
geometry with
with filleted
filletededges.
edges.
geometry with conical angle; (c) hole geometry with filleted edges.

The
Thehole
The holediameter
hole diameterdeviation
diameter deviationof
deviation ofthe
of thecircular
the circularhole
circular holesection
hole sectionresulted
section resultedin
resulted inaaasmall
in smallstep
small stepat
step atthe
at theinterface
the interface
interface
of
of the
of the cylindrical
the cylindrical and
cylindrical and expansion
and expansion sections.
expansion sections. The
sections. The hole
The hole diameter
hole diameter deviation
diameter deviation was
deviation was set
was set to
set to be
to be in
be in the
in the range
the range of
range of
of
±10%
±10% of
ofthe
the diameter
diameter of
of the
thebaseline
baseline geometry.
geometry. The
The diameter
diameter deviation
deviation was
was assumed
assumed
±10% of the diameter of the baseline geometry. The diameter deviation was assumed to be a Gaussian to
to be
be aaGaussian
Gaussian
22
distribution
distributionas
distribution as ∆D~N(7.75,
∆D~N(7.75,
as ∆D ∼ N 7.75, 0.775)2)[2,3].
0.775
0.775 [2,3].Five
[2,3]. Fivesamples
Five sampleswere
samples werechosen
were chosenfor
chosen forthe
for theCFD
the CFDsimulation
CFD simulation and
simulation andwere
and were
were
denoted
denotedas
denoted asthe
as thecases
the casesD1–D5
cases D1–D5(where
D1–D5 (whereD1,
(where D1,D2
D1, D2<<<mean
D2 meanvalue
mean valueof
value of7.75
of 7.75mm;
7.75 mm;D3
mm; D3is
D3 isequal
is equalto
equal to7.75
to 7.75mm;
7.75 mm;and
mm; andD4,
and D4,
D4,
D5 > mean value of 7.75
D5 > mean value of 7.75 mm).
D5 > mean value of 7.75 mm).
mm).
The
Theconical
The conicalangle
conical angle(Figure
angle (Figure4b)
(Figure 4b)of
4b) ofthe
of fan-shaped
thefan-shaped
the hole
fan-shapedhole deviation
holedeviation
deviationisisisdefined
defined
definedas as [3,5]:
as[3,5]:
[3,5]:

θθ==arctan((D
arctan((D--d)
d)//(2
(2HH))))
θ = arctan((D − d )/(2H ))
(1)
(1)
(1)
which
which isis determined
determined by by the
the hole
hole diameter
diameter deviation
deviation alongalong thethe hole
hole central
central axis.
axis. Based
Based onon the
the
maximum
maximumhole holediameter
diameterdeviation
deviationvalues,
values,the
theconical
conicalangle
angleof ofthe
thedrilling
drillinghole
holedeviation
deviationwaswas0.95°.
0.95°.
In
In the
the present
present study,
study, conical
conical angle
angle values
values of
of 0°,
0°, 0.25°,
0.25°, 0.5°,
0.5°, and
and 0.8°
0.8°were
were selected,
selected, and
and cases
cases with
with
these
these conical angle values are represented as C1, C2, C3, and C4. The filleted edge (Figure 4c) was
conical angle values are represented as C1, C2, C3, and C4. The filleted edge (Figure 4c) was
chosen
chosenas asr/D
r/D==0,0,2.5%,
2.5%,5%,
5%,and
and10%
10%bybyreferring
referringto toprevious
previousstudies
studies[7].
[7].

2.2.
2.2.Numerical
NumericalMethod
Methodand
andValidation
Validation
Aerospace 2019, 6, 46 5 of 17

which is determined by the hole diameter deviation along the hole central axis. Based on the maximum
hole diameter deviation values, the conical angle of the drilling hole deviation was 0.95◦ . In the present
study, conical angle values of 0◦ , 0.25◦ , 0.5◦ , and 0.8◦ were selected, and cases with these conical angle
values are represented as C1, C2, C3, and C4. The filleted edge (Figure 4c) was chosen as r/D = 0, 2.5%,
5%, and 10% by referring to previous studies [7].
Aerospace 2019, 6, x FOR PEER REVIEW 5 of 17
2.2. Numerical Method and Validation
To match the design size of the baseline 7-7-7 fan-shaped film hole, the diameter of the hole (D)
To match the design size of the baseline 7-7-7 fan-shaped film hole, the diameter of the hole (D)
was 7.75 mm. The width plenum (pitch-wise size) was 6D. The origin point was the middle of the
was 7.75 mm. The width plenum (pitch-wise size) was 6D. The origin point was the middle of the
trailing edge of the hole exit. The mainstream inlet was 10D upstream of the origin, and the
trailing edge of the hole exit. The mainstream inlet was 10D upstream of the origin, and the mainstream
mainstream outlet was 30D downstream of the origin. Other cases to be simulated will use the same
outlet was 30D downstream of the origin. Other cases to be simulated will use the same mainstream
mainstream domain and the same coolant domain, but different hole shapes.
domain and the same coolant domain, but different hole shapes.
Polyhedral meshes were generated using STAR-CCM+, as shown in Figure 5. The grid on the
Polyhedral meshes were generated using STAR-CCM+, as shown in Figure 5. The grid on the
bottom wall of the mainstream was refined. Prismatic grids were distributed near all the non-slip
bottom wall of the mainstream was refined. Prismatic grids were distributed near all the non-slip
walls to make y+ is close to 1. Three-dimensional steady viscous Reynolds-averaged Navier–Stokes
walls to make y+ is close to 1. Three-dimensional steady viscous Reynolds-averaged Navier–Stokes
equations were solved for the film cooling problems using the commercial CFD code STAR-CCM+.
equations were solved for the film cooling problems using the commercial CFD code STAR-CCM+.

Figure 5.
Figure 5. The
The schematic
schematic of
of the
the CFD
CFD domain,
domain, the
the polyhedral
polyhedral mesh
mesh with
with prismatic
prismatic grid
grid near
near the wall,
the wall,
and the
and the boundary
boundary conditions.
conditions.

The boundary
boundaryconditions
conditionswere determined
were to match
determined to an experiment
match case done by
an experiment caseSchroeder
done by et
al. [14]. Velocity
Schroeder and Velocity
et al. [14]. total temperature
and total were set at the
temperature mainstream
were set at theinlet. Static pressure
mainstream was pressure
inlet. Static imposed
at the
was outlet. A
imposed at 10
the°Coutlet. A 10 ◦in
difference C temperature
difference in was set between
temperature wasthe
set mainstream
between the and the coolant.
mainstream and The
the
mass flow
coolant. Therate of the
mass flowcoolant
rate ofwas imposed
the coolant on imposed
was the coolant
on inlet to match
the coolant thetoblowing
inlet match theratio (M) asratio
blowing one
because
(M) thebecause
as one 7-7-7 exhibited
the 7-7-7the largest the
exhibited filmlargest
coolingfilm
effectiveness values withvalues
cooling effectiveness the blowing
with theratio equal
blowing
to 1. All
ratio theto
equal walls were
1. All theset as adiabatic.
walls were set as The detailed The
adiabatic. boundary
detailedconditions
boundary areconditions
listed in Table 1.
are listed in
Table 1.
Table 1. Boundary conditions.

Parameter Range
Mainstream inlet velocity, U 20 m/s
Mainstream inlet temperature, T∞ 450 K
Mainstream Re(based on hole diameter) 5190
Turbulent intensity of mainstream, Tu∞ 5.4%
Turbulent intensity of coolant, Tuc 5%
Coolant inlet temperature, Tc 300 K
Blowing ratio, M 1.0
Density ratio, DR 1.5
 Aerospace 2019, 6, 46 6 of 17

Table 1. Boundary conditions.

Parameter Range
Mainstream inlet velocity, U 20 m/s
Mainstream inlet temperature, T∞ 450 K
Mainstream Re (based on hole diameter) 5190
Turbulent intensity of mainstream, Tu∞ 5.4%
Turbulent intensity of coolant, Tuc 5%
Coolant inlet temperature, Tc 300 K
Aerospace 2019, 6, x FOR PEER REVIEW Blowing ratio, M 1.0 6 of 17
Density ratio, DR 1.5
effects), the EB model, and the RST model, were employed to predict the adiabatic cooling
To reduce the impact
effectiveness distributions and to of epistemic
compare uncertainty, numerical
the adiabatic validation
cooling was performed
effectiveness against the
distributions with the
experiment done by Schroeder et al. [14]. Several different turbulence models, such as the standard k-ε
experimental data. Figure 6 plots the lateral averaged adiabatic cooling effectiveness distributions by
model, the realizable k-ε model, the V2F model (The V2f model is based on the standard k-ε model
experimentby and by CFD
adding with theanisotropy
the turbulence standard k-εnear-wall,
to the model, the realizable
it also contains thek-ε model,pressure-strain
non-local the V2F model, the EB
model, andeffects),
the RSTthe EBmodel. Figure
model, and the RST7model,
shows werethe adiabatic
employed cooling
to predict effectiveness
the adiabatic distributions. The
cooling effectiveness
distributions and to compare the adiabatic cooling effectiveness distributions with the experimental
lateral averaged η predicted by the standard k-ε model and the realizable k-ε model were in better
data. Figure 6 plots the lateral averaged adiabatic cooling effectiveness distributions by experiment and
agreement with
by CFD the experimental
with the standard k-εdata bythe
model, Schroeder etmodel,
realizable k-ε al. [14]thethan thosethe
V2F model, byEBother
model,turbulence
and the models.
ComparingRST themodel.
distribution of thethe
Figure 7 shows η on the wall
adiabatic between
cooling the experimental
effectiveness distributions. The data and
lateral the data
averaged η predicted
predicted
by the standard k-εbymodel andk-ε
the standard model
the and the realizable
realizable k-ε model
k-ε model, the were in better k-ε
realizable agreement with the
equations modeled the
experimental data by Schroeder et al. [14] than those by other turbulence models. Comparing the
current filmdistribution
cooling problem best;
of the η on the wallthus, thetherealizable
between experimental k-ε model
data and thewas
data applied inthe
predicted by thestandard
CFD simulations
in the current study.and the realizable k-ε model, the realizable k-ε equations modeled the current film cooling
k-ε model
problem
Therefore, due best;
tothus,
thethe realizable k-ε
agreement model wasthe
between applied
CFD in the CFD simulations
results in the current study.
and the experimental data with the
Therefore, due to the agreement between the CFD results and the experimental data with the
blowing ratio equal to 1 and the density ratio equal to 1.5, the geometric uncertainty analysis was
blowing ratio equal to 1 and the density ratio equal to 1.5, the geometric uncertainty analysis was
conducted with the with
conducted same thecoolant flowflow
same coolant condition.
condition.

Lateral averaged
Figure 6.averaged
Figure 6. Lateral adiabatic
adiabatic cooling effectiveness
cooling effectivenessdistributions by experiment
distributions by and by CFD withand by CFD
experiment
the standard k-ε model, the realizable k-ε model, the V2F model, the EB model, and the RST model.
with the standard k-ε model, the realizable k-ε model, the V2F model, the EB model, and the RST
model.
 Figure 6. Lateral averaged adiabatic cooling effectiveness distributions by experiment and by CFD
with 2019,
Aerospace the standard
6, 46 k-ε model, the realizable k-ε model, the V2F model, the EB model, and the RST
7 of 17
model.

Figure
Figure7.7.Adiabatic
Adiabaticcooling
coolingeffectiveness
effectiveness distributions byexperiment
distributions by experimentand
andbybyCFD
CFDwith
withthe
the standard
standard
k-ε
k-εmodel,
model,the
therealizable
realizablek-ε
k-ε model,
model, the V2F
V2F model,
model,the
theEB
EBmodel,
model,and
andthe
theRST
RST model.
model.

AAgrid
gridindependence
independencetest test was
was carried
carried out (see
(see Figure
Figure8)8)ininwhich
whichlateral
lateralaveraged
averaged film cooling
film cooling
effectiveness distributions in the streamwise direction with 0.2 million, 1.1 million, and
effectiveness distributions in the streamwise direction with 0.2 million, 1.1 million, and 2.5 million 2.5 million
gridpoints
grid pointswere
were compared
compared with
withone
oneanother
anotherand
andwith the the
with experimental data. data.
experimental With With
the 2.27-times grid
the 2.27-times
refinement (1.1 million–2.5 million), the lateral averaged cooling effectiveness value
grid refinement (1.1 million–2.5 million), the lateral averaged cooling effectiveness value changes changes were
invisible.
Aerospacewere
2019, ThePEER
6, x FOR
invisible. 1.1 million
The REVIEW
1.1 meshmesh
million method was therefore
method employed
was therefore in this in
employed study.
this study. 7 of 17

Figure 8.Figure
Grid independency test: test:
8. Grid independency lateral averaged
lateral averagedfilm
film cooling effectiveness
cooling effectiveness withwith the
the 0.2 0.2 million
million grid, grid,
1.1 million
1.1grid, and
million 2.5
grid, million
and grid;
2.5 million MM
grid; = 1.0
= 1.0 and
andDR
DR = 1.5.
1.5.

3. Uncertainty Quantification Methodology

3. Uncertainty Quantification Methodology
The UQ method used in this study was the non-intrusive Polynomial Chaos Expansion (PCE).
The This
UQmethod
method wasused
first in this study
introduced was the
by Wiener [16],non-intrusive Polynomial
using the Hermite Chaos Expansion
orthogonal polynomial to build (PCE).
the PCE
This method model.
was Xiu and Karniadakis
first introduced [17] improved
by Wiener it and
[16], using proposed
the Hermitea generalized
orthogonal polynomial chaosto build
polynomial
to deal with different distribution forms. This generalized PCE was employed in the current study.
the PCE model. Xiu and Karniadakis [17] improved it and proposed a generalized polynomial chaos
The mathematical basis of the PCE method is polynomial theory, which is equivalent to constructing
to deal with different distribution forms. This generalized PCE was employed in the current study.
a surrogate model for random variables, and uncertainty analysis was performed on the surrogate model.
The Strict
mathematical
mathematicalbasis
proofsofshowed
the PCE method
that there is polynomial
were corresponding theory,
optimal whichpolynomial
orthogonal is equivalent to
constructing
basesafor
surrogate model forforms
different distribution random
and variables,
converging atand uncertainty
exponential analysis
velocity [11]. was performed on the
surrogate model. Strict mathematical proofs showed that there were corresponding optimal
orthogonal polynomial bases for different distribution forms and converging at exponential velocity
[11].
The PCE model can be constructed by expanding the random output variable on the orthogonal
polynomial basis, as follows:
  i1
Y ( ) = a0 0 +  ai1 1 (i1 ( )) +  ai1i2 2 (i1 ( ), i2 ( ))
i =1 i =1 i =1
Aerospace 2019, 6, 46 8 of 17

The PCE model can be constructed by expanding the random output variable on the orthogonal
polynomial basis, as follows:

∞ i1
∞ P
Y ( θ ) = a 0 ψ0 + ai1 ψ1 (ξi1 (θ)) + ai1 i2 ψ2 (ξi1 (θ), ξi2 (θ))
P P
i1 =1 i1 = 1 i2 = 1
i1 P
∞ P i2 (2)
ai1 i2 i3 ψ3 (ξi1 (θ), ξi2 (θ), ξi3 (θ)) + . . .
P
+
i1 = 1 i2 = 1 i3 = 1

where θ represents the input random variable and ψ represents the orthogonal polynomial basis.
The above formula can also be abbreviated into a common format:
P
X
ŷ(ξ j (θ)) = ai ψi (ξ j (θ)) (3)
i=0

In the PCE model, there is a relationship between the unknown coefficient and the dimension of
the order and random variables.
Np = (d + p)!/(d!p!) (4)

in which d is the dimension of the random variable and p is the order of PCE truncating. Once the PCE is
built, the unknown coefficients of the PCE model can be solved with projection or projection-regression.
The Stochastic Response Surface Method (SRSM) was used to solve the unknown coefficients
in the PCE model [11,18]. The method was proposed by Dr. Isukapalli [19,20]. The coefficient was
evaluated from the oversampling point with the least-squares approach [11]. It is generally believed
that a sample with twice the number of unknown PCE coefficients can be used to obtain satisfactory
results. The use of an oversampling ratio of around 2 could improve the robustness of the method.
In addition, this treatment method also prevented some examples from getting a convergent solution.
Bringing the sample and the corresponding function output into the PCE model, you will get the
following determinant:

 ψ0 (ξ1 ) ψ1 ( ξ 1 ) ψNp (ξ1 )   y(ξ1 )

    
···  a0 
 ψ0 (ξ2 ) ψ1 ( ξ 2 ) ··· ψNp (ξ2 )  a1   y(ξ2 )
    

 =  (5)
    
 .. .. .. ..  .. .. 

 . . . .  .   . 

  
ψ0 ( ξ n s ) ψ0 (ξns ) ··· ψNp (ξns ) aNp y(ξns )
 

The above formula can be expressed as:

ψ∗a = Y (6)

To solve the PCE coefficients with the least-squares approach,

N N h 2
X X i
J (a) = ε2j = y(ξ j ) − ŷ(ξ j ) (7)
j=1 j=1

P
X
ŷ(ξ j ) = ai ψi (ξ j ) (8)
i=0

After obtaining the PCE coefficient, the PCE model was built. Then the random output variable
could be obtained. One of the methods was analytical. The mean and stand deviation of the random
output variable could be calculated by the formula,

µ= a0 (x) (9)
 Aerospace 2019, 6, 46 9 of 17

Np −1
X
2
σ = (a2i ψ2i ) (10)
i=1

Another method using the Monte Carlo (MC) method to solve the mean and stand deviation
based on the PCE model was also carried out in the current study. Then, the probability density of the
output variable can be obtained.

4. Results and Discussion

4.1. Film Cooling Effectiveness

4.1.1. Effect of the Conical Angle Deviation on Film Cooling Effectiveness

Figure 9 shows the laterally-averaged adiabatic film cooling effectiveness (η) with the drilled
conical angle equal to 0◦ , 0.25◦ , 0.5◦ , and 0.8◦ adapted from the two baseline 7-7-7 fan-shaped holes
with L/D = 3.5 and L/D = 6. The drilled conical angle deviation exerted no evident effect on the
value of η. This phenomenon was not consistent with that from an investigation with the baseline
cylindrical hole geometry by Wen [3]. It was found that the existence of a conical angle may result
in negative effects on the aerodynamic and heat transfer performance of the film cooling. However,
for the fan-shaped hole in the current study, the impact of the conical angle deviation on the film hole
heat transfer performance can be negligible. For the geometry adaption of the conical angle deviations
in the current study, only the cylindrical part of the fan-shaped hole as adapted, while the diffuser part
of the hole remained unchanged, which made the lateral expansion capacity of the in-hole cool air
remain the same, resulting in almost a constant local blowing ratio value and jet momentum value for
the holes with conical angle adaption. Therefore, η remained unchanged with conical angle adaption.
Aerospace 2019, 6, x FOR
As a result, thePEER
effectREVIEW
of the meter section offset on film holes’ heat transfer performance can be reduced 9 of 17
if the manufacturing accuracy of the diffuser section is high. However, the adaption of conical angle
of conical angleondeviation
deviation the film holeon the film
geometry can behole geometry
effective can be characteristic,
on the aerodynamic effective one.g.,the aerodynamic
discharge
coefficient (C ), which
characteristic, e.g., discharge
d will be discussed later in the current paper.
coefficient (Cd), which will be discussed later in the current paper.

Figure 9.Figure
Laterally-averaged
9. Laterally-averaged adiabatic
adiabatic film
filmcooling effectiveness
cooling effectiveness (η) (𝜂̅
for) the
fordrilled
the drilled
conical conical
angle angle
deviations (0◦ , 0.25◦ , 0.5◦ , and 0.8◦ ) for holes with L/D = 3.5 and L/D = 6.
deviations (0°, 0.25°, 0.5°, and 0.8°) for holes with L/D = 3.5 and L/D = 6.
Figure 10 shows the distribution of the η for the two fan-shaped hole case of L/D = 3.5 and
Figure
L/D 10
= 6shows
adaptedthe distribution
with conical angleofequal
the η to for
0.8◦the
. Thetwo fan-shaped
lateral diffusion ofhole
the case of for
cool air L/Dthe= case
3.5 and L/D
= 6 adapted L/D =conical
with with angle equal
6 was considerably to 0.8°.
stronger The
than thatlateral diffusion
of the case of =the
with L/D 3.5,cool airmade
which for the case with L/D
the lateral
coverage area larger, but the streamwise coverage length shorter for
= 6 was considerably stronger than that of the case with L/D = 3.5, which made the lateralthe case with L/D = 6 thancoverage
that for the case with L/D = 3.5. The reason can be attributed to the difference in the hole exit area.
area larger, but the streamwise coverage length shorter for the case with L/D = 6 than that for the case
With larger hole length-to-diameter ratio values, the hole exit area will be larger. According to a study
with L/D = 3.5. The reason can be attributed to the difference in the hole exit area. With larger hole
length-to-diameter ratio values, the hole exit area will be larger. According to a study by Haydt et al.
[21] on the effects of area ratio on the distributions of η for the baseline 7–7–7 fan-shaped hole, η
increases with the increase of the ratio of the hole exit area value to hole entrance area value.
 Figure 10 shows the distribution of the η for the two fan-shaped hole case of L/D = 3.5 and L/D
= 6 adapted with conical angle equal to 0.8°. The lateral diffusion of the cool air for the case with L/D
= 6 was considerably stronger than that of the case with L/D = 3.5, which made the lateral coverage
area larger,
Aerospace 2019,but
6, 46the streamwise coverage length shorter for the case with L/D = 6 than that for the10case of 17
with L/D = 3.5. The reason can be attributed to the difference in the hole exit area. With larger hole
length-to-diameter ratio values, the hole exit area will be larger. According to a study by Haydt et al.
by Haydt
[21] on theeteffects
al. [21]ofonarea
the effects of area
ratio on ratio on the distributions
the distributions of η for theofbaseline
η for the7–7–7
baseline 7–7–7 fan-shaped
fan-shaped hole, η
hole, η increases with the increase of the ratio of the hole exit area value to hole
increases with the increase of the ratio of the hole exit area value to hole entrance area value. entrance area value.
Therefore, it
Therefore, it can
can bebe aa reason
reason why
whythethehole
holewith L/D==66features
withL/D featuresbetter
bettercooling
coolingperformance
performance than that
than thatof
the hole with L/D = 3.5 even with conical angle
of the hole with L/D = 3.5 even with conical angle adaption. adaption.

Figure
Figure10.
10.Distribution
Distributionof theηηfor
ofthe forthe
thetwo
twofan-shaped
fan-shaped hole
hole case
case of L/D ==3.5
of L/D 3.5and L/D ==66adapted
andL/D adaptedwith
with
conical
conicalangle
angleequal
equalto ◦
to 0.8°.
0.8 .

4.1.2.Effect
4.1.2. Effectof
ofthe
theDiameter
Diameterand
andFilleted
Filleted Edge
Edge Deviations
Deviations on
on Film
Film Cooling
Cooling Effectiveness
Effectiveness
The filleted
The filleted edge
edge and and diameter
diameter deviation
deviation are are not
not independent
independent factors factors forfor thethe film
film cooling
cooling
effectiveness. The two factors interact with
effectiveness. The two factors interact with each other. each other.
=
Figure11
Figure 11presents
presentsthe thearea
areaaveraged
averagedadiabatic
adiabaticcooling
coolingeffectiveness
effectiveness (𝜂̿ ( η) withdiameter
) with diameterdeviations
deviations
of D1, D2, D3, D4, and D5 and with fillet deviations of 0, 2.5%, 5%,
of D1, D2, D3, D4, and D5 and with fillet deviations of 0, 2.5%, 5%, and 10%. The area averaged and 10%. The area averaged
calculation
calculation was done through a rectangle
was done through a rectangle of x/D of x/D = 0~30 and y/D = −3~3. For all the
= 0~30 and y/D = −3~3. For all the cases with10both cases with both
Aerospace 2019, 6, x FOR PEER REVIEW = of 17
diameter deviations
diameter deviations and and fillet
fillet deviations,
deviations,the theη𝜂̿value
valueinitially
initiallydecreased
decreased and then
and stabilized
then when the
=stabilized when
radius
the of the
radius of filleted
the edgeedge
filleted increased. Therefore, the effect of theoffillet
the adaption on η was 𝜂̿inwas a certain
baseline case. However, with increased.
filleted edgesTherefore, the effect
and diameter deviations, fillet
all adaption
the cases on with diameter in a
range for
certain the current
range for the study,
current and the smaller
study, and the the fillet, the
smaller the more
fillet,evident
the more theevident
change by the the filleted
change byedge.
the
deviations, no matter positive or negative, featured a decrease = in the 𝜂̿ value from the baseline case.
For the case with no diameter adaption, the change of η value due to the fillet edge can be up to 3.4%,
filleted
With a edge.
largerFor the case
radius with
of the no diameter
filleted edge, theadaption, the change
deviations of the of 𝜂̿ 𝜂̿value
= value duelarger.
were to the fillet
This edge
makes canit
while
be up for3.4%,
to the case
whilewith
for the
the D1
case diameter
with the adaption,
D1 the adaption,
diameter change ofthe η valueofdue
thechange the to𝜂̿ the filletdue
value edge
to can
the
important to understand the change in the flow field with the filleted edge adaption. The in-hole flow
be up to 7.6%.
fillet
field edge
with can be upfilleted
different to 7.6%.edge adaptions will be discussed later.
With the filleted edge radius equal to zero, the influence on the 𝜂̿ value by the diameter
deviation was small. The 𝜂̿ value decreased with the diameter of the hole increasing. Cases with
smaller diameters (D1,D2) than that of the baseline case acquired larger 𝜂̿ values than that of the

Figure 11. Area averagedadiabatic

adiabaticcooling
coolingeffectiveness
effectiveness = ) with diameter deviations of D1, D2, D3,
Figure 11. Area averaged ( η)(𝜂̿with diameter deviations of D1, D2, D3, D4,
D4,
and and D5 with
D5 and and with
fillet fillet deviations
deviations of 0, 2.5%,
of 0, 2.5%, 5%, and5%,10%
and(averaged
10% (averaged
throughthrough x/D =y/D
x/D = 0~30, 0~30, y/D =
= −3~3).
−3~3).
=
With the filleted edge radius equal to zero, the influence on the η value by the diameter deviation
4.2. Discharge =
Coefficient
was small. The η value decreased with the diameter of the hole increasing. Cases with smaller
=
diameters (D1,D2) than that of the baseline case acquired larger η values than that of the baseline case.
4.2.1. Effect
However, of the
with Conical
filleted Angles
edges on the Discharge
and diameter Coefficient
deviations, all the cases with diameter deviations, no matter
=
positive featured a decrease in the η
Figure 12 shows the discharge coefficient values (Cfrom
or negative, value d) forthe
thebaseline case. With
drilled conical a larger
angle radius(0°,
deviations of
=
the filleted
0.25°, edge,0.8°)
0.5°, and the deviations of the
for holes with L/Dη value wereL/D
= 3.5 and larger.
= 6 inThis
themakes it important
current boundary to understand
condition. the
Cd is a
measure of the aerodynamic performance of the film hole, which is defined as:
m
CD =
p 
( k +1)/ 2 k
2k   p ( k −1)/ k  (11)
ptc  m    tc  − 1 D 2
 ptc  ( k − 1) RTtc   pm  4
 
Aerospace 2019, 6, 46 11 of 17

change in the flow field with the filleted edge adaption. The in-hole flow field with different filleted
edge adaptions will be discussed later.

4.2. Discharge Coefficient

4.2.1. Effect of the Conical Angles on the Discharge Coefficient

Figure 12 shows the discharge coefficient values (Cd ) for the drilled conical angle deviations (0◦ ,
0.5◦ , and 0.8◦ ) for holes with L/D = 3.5 and L/D = 6 in the current boundary condition. Cd is a
0.25◦ ,
measure of the aerodynamic performance of the film hole, which is defined as:
.
m
CD = r (11)
 p (k+1)/2k  (k−1)/k
p

ptc m
ptc
2k
(k−1)RTtc
tc
pm − 1 π4 D2

in which Pm is the mainstream pressure and Ptc is the total pressure of the coolant.
Though the conical angle deviation did not change the η value much, it affected the film hole Cd
value visibly. With a conical angle, the meter section contracted along the coolant streamline, increasing
the blockage effect, thus changing the Cd value. With the conical angle deviation equal to 0.5◦ , the Cd
value was 6.2% less than the baseline case for the case of the hole with L/D = 6. With the angle
deviation equal to 1.0◦ , the change in the Cd value from the baseline case was up to 12.9%. For cases
with L/D = 3.5, the Cd value and the change of the Cd value were smaller than those of the case with
L/D = 6. It is interesting to note that the Cd value and the conical angle featured a linear relation.
With a smaller Cd value and an unchanged η value by conical angle adaption, the coolant supply
pressure must be increased to maintain the jet mass flow rate with the conical angle arising in the
drilling film hole, or the mass flow rate of the coolant will decrease with supply pressure unchanged.
It will change the distribution of pressure or the distribution of coolant mass flow rate in the blade,
thus resulting in the imbalance of the surface temperature of the blade. Therefore, the existence of the
conical angle should be considered in the initial design phase.
Aerospace 2019, 6, x FOR PEER REVIEW 11 of 17

Figure 12.
12. Discharge
Dischargecoefficient
coefficientvalues
values(C(Cd) for the drilled conical angle deviations (0°, 0.25°,
◦ 0.5°,
◦ and
◦
Figure d ) for the drilled conical angle deviations (0 , 0.25 , 0.5 ,
0.8°) for holes with L/D = 3.5 and L/D = 6.
and 0.8 ) for holes with L/D = 3.5 and L/D = 6.
◦

4.2.2. Effect of the Diameter and Filleted Edge Deviations

Deviations on
on the
the Discharge
Discharge Coefficient
Coefficient
Figure
Figure 13
13shows
showsdischarge
dischargecoefficient
coefficientvalues (Cd(C
values ) with
d) withdiameter deviations
diameter of D1,
deviations D2, D2,
of D1, D3, D4,
D3, and
D4,
D5
andandD5with
and fillet
with deviations of 0, 2.5%,
fillet deviations of 0,5%, and5%,
2.5%, 10%.andThe positive
10%. The deviation of the diameter
positive deviation of the increased
diameter
the Cd value,
increased theindicating
Cd value,that the discharge
indicating performance
that the discharge of the hole wasofimproved.
performance Withimproved.
the hole was larger flowWith
area
by the flow
larger position
areadiameter deviation,
by the position the blockage
diameter effectthe
deviation, of blockage
the hole was decreased,
effect of the holeandwas
thedecreased,
discharge
performance was improved.
and the discharge performance Thus,
was with negativeThus,
improved. diameter
withdeviation, the holes deviation,
negative diameter featured smaller Cd
the holes
values
featuredthan that of
smaller Cdthe baseline
values thancase.
that of the baseline case.
With the increase of the radius of the filled edge, the Cd value increased, though the 𝜂̅ value
decreased. For cases with the same diameter deviation values, the Cd value was almost linear to the
radius of the filled edge. The widespread existing of the filleted edges in the real blade or in the
experimental conditions can be a reason for the underestimation of the Cd value and the
overestimation of the 𝜂̅ value (Adami [22]) calculated by the CFD method compared with the
experimental data, as the micro-geometric features and deviations such as fillet and step are often
and D5 and with fillet deviations of 0, 2.5%, 5%, and 10%. The positive deviation of the diameter
increased the Cd value, indicating that the discharge performance of the hole was improved. With
larger flow area by the position diameter deviation, the blockage effect of the hole was decreased,
and the discharge performance was improved. Thus, with negative diameter deviation, the holes
featuredAerospace
smaller Cd6,values
2019, 46 than that of the baseline case. 12 of 17

With the increase of the radius of the filled edge, the Cd value increased, though the 𝜂̅ value
decreased. ForWith cases with the
the increase of same diameter
the radius deviation
of the filled values,
edge, the theincreased,
Cd value Cd value though the η value
was almost linear to the
decreased. For cases with the same diameter deviation values, the C value
radius of the filled edge. The widespread existing of the filletedd edges in the real blade was almost linear to theor in the
radius of the filled edge. The widespread existing of the filleted edges in the real blade or in the
experimental conditions can be a reason for the underestimation of the Cd value and the
experimental conditions can be a reason for the underestimation of the Cd value and the overestimation
overestimation of the 𝜂̅ value (Adami [22]) calculated by the CFD method compared with the
of the η value (Adami [22]) calculated by the CFD method compared with the experimental data, as the
experimental data, as the
micro-geometric micro-geometric
features features
and deviations such and
as fillet anddeviations such
step are often as fillet
simplified andand stepinare often
ignored
simplified and ignored
numerical in numerical simulations.
simulations.

Figure 13. Discharge

Figure coefficient
13. Discharge values
coefficient (C(C
values d)dwith
) withdiameter deviations
diameter deviations of D2,
of D1, D1,D3,
D2,D4,
D3, D4,
and D5 and
and D5 and
with fillet deviations of 0, 2.5%, 5%,
with fillet deviations of 0, 2.5%, 5%, and 10%. and 10%.

4.3. In-Hole Flow Field with Filleted Edge

4.3. In-Hole Flow Field with Filleted Edge
To understand the effect of the filleted edge on the heat transfer performance and discharge
To understandthe
performance, thein-hole
effectflow
of field
the with filleted
filleted edgeedgesonarethestudied
heatintransfer
the current paper. Figure 14
performance and is the
discharge
schematic of the in-hole planes (red lines) chosen to plot the velocity distributions, and Figure 15 shows
performance, the in-hole flow field with filleted edges are studied in the current paper. Figure 14 is
the velocity distributions on the chosen planes for the case without a filleted edge and the case with a
the schematic of the in-hole planes (red lines) chosen to plot the velocity distributions, and Figure 15
filleted edge of r/D = 10%. For the baseline case, the cool air came from the large plenum, and the
shows theflowvelocity distributions
accelerated and deflected onintothethechosen planes
hole. Thus, for the
a counter case without
rotating kidney vortexa filleted
(CRVP)edgeformed and in the case
with a filleted
Aerospaceedgeofofthe
2019,
the entrance 6, x r/D
FOR = 10%.
PEER
hole, REVIEWForinthe
resulting baseline
a lower case,
velocity zonethe cool
near the air came
trailing from
edge thehole.
of the large
With12plenum,
of
the17 and
expansion of the flow passage in the diffuser part of the hole,
the flow accelerated and deflected into the hole. Thus, a counter rotating kidney vortex (CRVP) the high velocity zone near the hole
leading edge
leading edgedecelerated
and the trailingalong edge were
the hole smaller
central axis,inand
thethediffuser
pressurepartgradients
of the hole than the
between thatleading
in the
formed in the entrance
cylinder part of of the
the hole,hole,
making resulting
the CRVP in move
a lowerup velocity
to the zone
center of near
the theand
hole trailing
spread edge of the hole.
laterally.
edge and the trailing edge were smaller in the diffuser part of the hole than that in the cylinder part of
With thetheexpansion
For the case
hole, making
of the
with the flow
theCRVP
filletedpassage
moveedge, inacceleration
thethe
up to
the diffuser
center of the
part
ofhole
the flowofinthe
and spreadthehole, theFor
hole inlet
laterally.
high
was velocity
themore
case gradual,
with the
zone near
the hole filleted
leading
and theedge, edge
kidneythedecelerated
vortex
acceleration ofalong
was substantially
the flow the hole
inweaker central
the hole than
inletthe axis,
was and
baseline
more the pressure
case
gradual, and
andattached gradients
the kidneyto the between
trailing
vortex was the
edge of the hole. Due to the small distance of the CRVP for the case with
substantially weaker than the baseline case and attached to the trailing edge of the hole. Due to thethe filleted edge, the flow
deflected
small againofinthe
distance theCRVP
interface of case
for the the meter
with the andfilleted
diffuser sections
edge, because
the flow of theagain
deflected backward wall. An
in the interface
Anti-Counter-Rotating Vortex (ACRV) formed there and was reinforced
of the meter and diffuser sections because of the backward wall. An Anti-Counter-Rotating Vortex along the streamlines, which
can be beneficial to the heat transfer performance, but detrimental to the
(ACRV) formed there and was reinforced along the streamlines, which can be beneficial to the heat discharge performance of
the hole.performance, but detrimental to the discharge performance of the hole.
transfer

Figure 14. Schematic of the in-hole planes (red lines) chosen to plot
plot the
the velocity
velocity distributions.
distributions.
Aerospace 2019, 6, 46 13 of 17
Figure 14. Schematic of the in-hole planes (red lines) chosen to plot the velocity distributions.

Figure 15. Velocity contour for different

different planes
planes in-hole
in-hole with
with fillet
fillet deviation only: the top is r/D =
deviation only: = 0%;
the bottom is r/D =
= 10%.

Figure
Figure 1616 shows
shows the the velocity
velocity andand vector
vector distributions
distributions on on the
the center
center plane
plane (Y (Y == 0) of the
0) of the hole
hole with
with
filleted
filleted edge
edge radius
radius equal
equal to to 0D,
0D, 2.5%D,
2.5%D, 5%D, 5%D, and and 10%D.
10%D. There
There was “jetting region”
was “jetting region” andand “separation
“separation
region” inside of
region” inside ofthe
theholes
holesfor forthe
thefour
four cases.
cases. There
There werewere
twotwo separation
separation zones zones
insideinside
of the ofhole.
the hole.
One
One was due to the jetting effect in the hole inlet. The other large separation
was due to the jetting effect in the hole inlet. The other large separation zone was in the diffuser zone was in the diffuser
part
part
of theofhole
thedue
holetodue to the laying
the laying back ofbackthe holeof the hole edge.
trailing trailing
At edge.
the regionAt the
nearregion
the hole near thethe
inlet, hole inlet,
coolant
the
wascoolant was compressed
compressed to move neartothe movewallnearof thetheleading
wall ofedge
the leading
becauseedge of thebecause of the jetting
high velocity high velocity
region
jetting
with a region
sharp edgewith aorsharp
a smalledge or a small
radius filleted radius
edge,filleted
making edge,
the making
separationthezoneseparation
delayed zone delayed
in-tube for
in-tube for these
these cases. However,cases.the However,
low-speed theregion
low-speed region for
was thicker wasthe thicker
sharpforedgethecase,
sharp edge case,
making making
the turbulent
the turbulent
mixing effect mixing
betweeneffect between
the jetting andthe jetting and
separation separation
regions stronger;regions
thus,stronger;
the velocitythus,atthethevelocity at
hole exit
the
nearhole
the exit nearedge
leading the leading edge than
was smaller was smaller
that in thethan thatwith
cases in the cases with
filleted edges. filleted edges. between
The mixing The mixing the
between the coolant and mainstream hot gas was weaker for the sharp-edge
coolant and mainstream hot gas was weaker for the sharp-edge case, improving the coolant coverage. case, improving the
coolant coverage.
The separation The separation
region produced region
by theproduced
diffuser partby the
fordiffuser
the cases part for the
with filletcases
edgewithwasfillet edgefrom
further was
further fromedge
the trailing the trailing edge and
and smaller smaller
in size, in size,
making themaking the high-speed
high-speed airflow inairflow in the
the jetting jettingdirectly
region region
directly spray out of the film hole without sufficient in-hole mixing.
spray out of the film hole without sufficient in-hole mixing. The velocity near the leading edge ofThe velocity near the leading
edge of hole
hole exit exitcases
for the for the cases
with with edge
filleted filleted
was edge wasthan
higher higherthatthan
of thethat of the baseline,
baseline, exertingexerting
a strongareverse
strong
reverse
impulseimpulse on the mainstream.
on the mainstream. As a result,As the
a result, thecoverage
coolant coolant coverage
for cases for
with cases
filletwith
was fillet
worse was
thanworse
that
than that for the sharp-edge
for the sharp-edge case. case.

4.4. UQ Analysis
According to the results in Sections 4.1–4.3, the small geometric deviations by manufacturing
process, such as conical angles, filleted edges, and diameter deviations of the hole, can change the heat
transfer and aerodynamic performance of the hole visibly, which makes the performance of the film
holes uncertain due to random geometry deviations. It is important to quantify this uncertainty based
on the random geometry deviations and consider it in the initial design phase.
UQ analysis aims to provide a framework to calculate the effects of the input parameter uncertainty
on the model output variables. This analysis is used to determine the effects of the uncertainty factors
(manufacturing tolerances or boundary conditions) on the cooling system and reduce the variability in
the final design. In the current study, the PCE model of the UQ analysis was constructed based on two
factors: the hole diameter and the fillet radii. SRSM was used to solve the PCE model coefficients in
this paper.
Figure 17 shows the 95% confidence interval distribution for the lateral averaged adiabatic cooling
effectiveness (η) value and the adiabatic cooling
 effectiveness
 (η) value along the centerline. The hole
diameter deviation distributed as ∆D ∼ N 7.75, 0.7752 , and the fillet radii r/D = 0, 2.5%, 5%, and 10%.
The range of the 95% confidence interval for the for the η value and the η value along the centerline
were very small relative to their values, which indicates that the η value is not sensitive to the hole
Aerospace 2019, 6, 46 14 of 17

diameter and fillet radii deviations. This is mainly because only the cylindrical part of the fan-shaped
hole was adapted, while the diffuser part of the hole remained almost unchanged. As previously
mentioned, the fillet radii affected the in-hole flow field greatly, which can have greater influence on
Cd value
theAerospace 2019,than thatPEER
6, x FOR the η value.
on REVIEW 13 of 17

Figure
Figure 16.16.Velocity
Velocitydistributions
distributions and
and vectors
vectors distributions
distributionson
onthe
theYY= =0 0plane with
plane thethe
with filleted edge
filleted edge
radius equal to 0D, 2.5%D, 5%D, and 10%D.
radius equal to 0D, 2.5%D, 5%D, and 10%D.
Aerospace 2019, 6, x FOR PEER REVIEW 14 of 17
4.4. UQ Analysis
According to the results in Sections 4.1, 4.2, and 4.3, the small geometric deviations by
manufacturing process, such as conical angles, filleted edges, and diameter deviations of the hole,
can change the heat transfer and aerodynamic performance of the hole visibly, which makes the
performance of the film holes uncertain due to random geometry deviations. It is important to
quantify this uncertainty based on the random geometry deviations and consider it in the initial
design phase.
UQ analysis aims to provide a framework to calculate the effects of the input parameter
uncertainty on the model output variables. This analysis is used to determine the effects of the
uncertainty factors (manufacturing tolerances or boundary conditions) on the cooling system and
reduce the variability in the final design. In the current study, the PCE model of the UQ analysis was
constructed based on two factors: the hole diameter and the fillet radii. SRSM was used to solve the
PCE model coefficients in this paper.
Figure17.17The
Figure shows 95% the 95% confidence
confidence interval interval
distributiondistribution
of the for theaveraged
lateral lateral averaged
adiabatic adiabatic
cooling
Figure 17. The 95% confidence interval distribution of the lateral averaged adiabatic cooling effectiveness
cooling effectiveness
effectiveness (𝜂̅ )
(𝜂̅
value
) value
and the
and the adiabatic
adiabatic cooling
cooling effectiveness
effectiveness (η) value
(η) value
along the
along the
centerline
centerline.
with the
(η) value and the adiabatic cooling  effectiveness (η) value along
2 ),the centerline with the hole diameter
The hole diameter deviation distributed as ∆D~N(7.75, 0.775
2) and the fillet radii r/D = 0, 2.5%, 5%,
hole diameter deviation distributed as ∆D~N(7.75, 0.775 and the fillet
deviation distributed as ∆D ∼ N 7.75, 0.775 and the fillet radii r/D = 0, 2.5%, 5%, 10%.
2 radii r/D = 0, 2.5%, 5%, 10%.
and 10%. The range of the 95% confidence interval for the for the 𝜂̅ value and the η value along the
centerline were very small relative to theirthe
distributions values, whichparameters
statistical indicates that the
of η: η value is not sensitive to
η: the
Figure 18 shows the distributions of the statistical parameters of mean value distribution
the hole diameter and fillet radii deviations. This is mainly because only the cylindrical part of the
and the standard
standard deviation
deviationvalue valuedistribution.
distribution.The Thedistribution
distribution of of
thethemeanmeanvalue η was
of ηofwas
value almost the
almost
fan-shaped hole was adapted, while the diffuser part of the hole remained almost unchanged. As
the same as that of the baseline case. However, there were visible deviations of the η distributions.
same as that of the baseline case. However, there were visible deviations of the η distributions. The
previously mentioned, the fillet radii affected the in-hole flow field greatly, which can have greater
deviation
The of ηof
deviation η was
was larger
largerin the near
in the hole
near region,
hole region, mainly
mainly near
nearthe theregion
regionwhere
wherethe theCRVP
CRVP existed,
existed,
influence on the Cd value than that on the η value.
which indicated that the adaption in the distribution of η η was mainly due to the exists of the ACRV
and the change in the structures
structures of of the
the CRVP,
CRVP,as aspreviously
previouslyinformed.
informed.
 Figure 18 shows the distributions of the statistical parameters of η: the mean value distribution
and the standard deviation value distribution. The distribution of the mean value of η was almost the
same as that of the baseline case. However, there were visible deviations of the η distributions. The
deviation of η was larger in the near hole region, mainly near the region where the CRVP existed,
which indicated
Aerospace 2019, 6, 46 that the adaption in the distribution of η was mainly due to the exists of the ACRV
15 of 17
and the change in the structures of the CRVP, as previously informed.

Figure
Figure 18.
18.Distributions
Distributionsof the statistical
of the parameters
statistical of η; top:
parameters of η;mean
top: value;
meanbottom:
value; standard
bottom: deviation
standard
value.
deviation value.
=
Figure 19 shows the PDF distribution of the area averaged adiabatic cooling effectiveness (𝜂̿ ( η))
=
and the value of Cdd.. The ThePDF
PDFof the 𝜂̿η value
ofthe value did
did not
not fit
fit for
for the
the regular
regular Gaussian
Gaussian distribution.
distribution. It was
distributed overa asmall
distributed over small range
range withwith the average
the average valuevalue of 0.1854.
of 0.1854. The unfitting
The unfitting for the deviation
for the standard standard
deviation
was becausewasthebecause
standard the standardofdeviation
deviation of the
the statistics wasstatistics
small. Thiswasalsosmall. This also
indicates that indicates that the
the hole diameter
=
hole diameter
and fillet radii and fillet radii
deviations haddeviations
little effecthad η effect
little
on the value.on the 𝜂̿ value.
However, the PDF distribution of the Cdd value value basically
basically still
still satisfied
satisfied the
the Gaussian
Gaussian distribution.
distribution.
variation of hole diameter satisfied the Gaussian distribution, and the
As mentioned previously, the variation
fillet radii satisfied the uniform
uniform distribution
distribution for the the PCE
PCE model.
model. The Cdd value was almost linear to
fillet radii (which can be found in Figure 13). 13). The coefficient of variation of Cdd was 0.074, which was
less than the coefficient of variation of hole diameter.diameter. All these factors resulted in the propagation
characteristics of the input uncertainty variables.
Aerospace 2019, 6, x FOR PEER REVIEW 15 of 17

Figure 19.
19. PDF distribution of
ofthe
thearea
areaaveraged
averagedadiabatic
adiabaticcooling
coolingeffectiveness
effectiveness= ) and the value
Figure PDF distribution ( η)(𝜂̿and the value of Cd .
of Cd.
5. Conclusions
5. Conclusions
This study investigates the effects of small geometric deviations on the values of η and Cd by
usingThis
the baseline 7-7-7 fan-shaped
study investigates hole.ofAsmall
the effects detailed flow field
geometric analysis on
deviations wasthe
carried out
values ofto
η determine
and Cd by
the impact mechanism of geometric deviation on film cooling. To quantify the uncertainty of
using the baseline 7-7-7 fan-shaped hole. A detailed flow field analysis was carried out to determine η and Cd
values due to
the impact the manufacturing
mechanism uncertainty,
of geometric deviationanon
UQfilm
analysis wasTo
cooling. implemented
quantify thewith PCE to obtain
uncertainty the
of η and
statistical characteristics
Cd values due (the mean uncertainty,
to the manufacturing value and StdanofUQη and Cd ), which
analysis were causedwith
was implemented by the
PCEgeometric
to obtain
deviations. Thecharacteristics
the statistical following conclusions
(the mean were obtained.
value and Std of η and Cd), which were caused by the
geometric deviations. The following conclusions were obtained.
• Conical angle deviation exerts no obvious effect on η. However, Cd decreases by 6.2% when the
• Conical angle deviation exerts no obvious effect on η. However, Cd decreases by 6.2% when the
◦ and increases to 12.9 when the angle is 1.0◦ .
conical angle deviation is 0.5
conical angle deviation is 0.5° and increases to 12.9 when the angle is 1.0°.
• The presence of diameter and fillet deviations produces a superposition effect. The area average
η decreases by 3.4%, and Cd increases to 15.2% with the fillet deviation existing alone. However,
the values are 7.6% and 25.7% when the two deviations exist.
• The decrease in Cd is mainly caused by the weakened streamwise vorticity of in-tube and the
blockage effect when the hole possesses a fillet. The velocity distribution of the hole exit exerts
an important impact on the change in η.
Aerospace 2019, 6, 46 16 of 17

• The presence of diameter and fillet deviations produces a superposition effect. The area average η
decreases by 3.4%, and Cd increases to 15.2% with the fillet deviation existing alone. However,
the values are 7.6% and 25.7% when the two deviations exist.
• The decrease in Cd is mainly caused by the weakened streamwise vorticity of in-tube and the
blockage effect when the hole possesses a fillet. The velocity distribution of the hole exit exerts an
important impact on the change in η.
• The UQ analysis shows that the 95% confidence interval of the centerline and laterally-averaged
η both are relatively small. The effects of cylindrical section deviations on η is limited, and the
fillet radii mainly affects the flow field in-hole. The result also shows that the PDF of area
average η does not satisfy the regular distribution, while the PDF of Cd basically still satisfies the
Gaussian distribution.

Author Contributions: Writing, original draft preparation, W.S.; proofread and revise, P.C.; supervision, X.L. and
H.J.; funding acquisition, J.R.
Funding: This study is supported by the National Natural Science Foundation of China (Project No. 51676106).
Acknowledgments: The authors appreciate Fenfen Xiong from Beijing Institute of Technology for the guidance
on the PCE method.
Conflicts of Interest: The authors declare no conflict of interest.

Nomenclature
EDM Electrical Discharge Machining
AM Additive Manufacturing
Cd discharge coefficient
UQ Uncertainty quantification
PCE Polynomial Chaos Expansion
Std Standard deviation
PDF Probability Density Function
SRSM Stochastic Response Surface Method
MC Monte Carlo
L Hole Length
D Hole Diameter
TIT Turbine Inlet Temperature
BR Blowing ratio
DR Density Ratio
r Fillet radius
EB Elliptic Blending model
Pm mainstream pressure
Ptc total pressure of the coolant
LE Leading Edge of hole
TE Trailing Edge of hole
CRVP Counter Rotating Vortex Pair
RST Reynolds Stress Turbulence
Greek symbols
η Film cooling effectiveness
θ Conical angle
Subscripts
w wall
m mainstream flow
c inlet of coolant duct
aw adiabatic wall
1 entrance
2 exit
Aerospace 2019, 6, 46 17 of 17

References
1. Bunker, R.S. Gas Turbine Cooling: Moving from Macro to Micro Cooling. ASME Conf. Proc. 2013, 2013, 94277.
2. Bunker, R.S. The Effects of Manufacturing Tolerances on Gas Turbine Cooling. J. Turbomach. 2009, 131, 041018.
[CrossRef]
3. PRIMA North America, Inc. Laser Drilling of Cylindrical and Shaped Holes. Available online: https:
//slideplayer.com/slide/6880655/ (accessed on 20 September 2017).
4. Wen, Z.; Pei, H.; Zhang, C.; Wang, B. Analysis of Surface Quality of Multi-film Cooling Holes in Nickel-based
Single Crystal Superalloy. Mater. Sci. Technol. 2016, 32, 1845–1854. [CrossRef]
5. Maina, M.R. Modeling and Control of Nd: YAG Laser Percussion Drilling of Nickel-Based Super Alloys.
Master’s Thesis, Jomo Kenyatta University of Agriculture and Technology, Juja, Kenya, 2015.
6. Ramesh, S.; Ramirez, D.G.; Ekkad, S.V.; Alvin, M.A. Analysis of film cooling performance of advanced tripod
hole geometries with and without manufacturing features. J. Heat Mass Transf. 2016, 94, 9–19. [CrossRef]
7. Montomoli, F.; Massini, M.; Salvadori, S.; Martelli, F. Geometrical Uncertainty and Film Cooling: Fillet radii.
J. Turbomach. 2012, 134, 011019. [CrossRef]
8. Cerantola, D.J.; Birk, A.M. Quantifying Blowing Ratio for Shaped Cooling Holes. J. Turbomach. 2018,
140, 021008. [CrossRef]
9. Johnson, P.L.; Ricklick, M.; Kapat, J.S.; Zuniga, H.; Brown, G. The Impact of Manufacturing Techniques on
Film Cooling Effectiveness. In Proceedings of the 49th AIAA/ASME/SAE/ASEE Joint Propulsion Conference,
San Jose, CA, USA, 15–17 July 2013.
10. Haydt, S.; Lynch, S.; Lewis, S. The Effect of a Meter-Diffuser Offset on Shaped Film Cooling Hole Adiabatic
Effectiveness. J. Turbomach. 2017, 139, 091012. [CrossRef]
11. Montomoli, F.; Carnevale, M.; D’Ammaro, A.; Massini, M.; Salvadori, S. Uncertainty Quantification in
Computational Fluid Dynamics and Aircraft Engines, 2nd ed.; Springer: Cham, Switzerland, 2018.
12. Seshadri, P.; Shahpar, S.; Parks, G. Aggressive Design in Turbomachinery. In Proceedings of the 7th Dresdner
Probabilistic Workshop, Dresden, Germany, 9 October 2014.
13. Slotnick, J.; Khodadoust, A.; Alonso, J.; Darmofal, D.; Gropp, W.; Lurie, E.; Mavriplis, D. CFD Vision 2030
Study: A Path to Revolutionary Computational Aerosciences; NASA/CR-2014-218178; NASA Langley Research
Center: Hampton, VA, USA, 2014.
14. Schroeder, R.P.; Thole, K.A. Adiabatic Effectiveness Measurements for a Baseline Shaped Film Cooling
Hole. In Proceedings of the ASME Turbo Expo 2014: Turbine Turbine Technical Conference and Exposition,
Düsseldorf, Germany, 16–20 July 2014. GT2014-25992.
15. Schroeder, R.P.; Thole, K.A. Effect of high freestream turbulence on flowfields of shaped film cooling holes.
J. Turbomach. 2016, 138, 091001. [CrossRef]
16. Wiener, N. The homogeneous chaos. Am. J. Math. 1938, 60, 897–936. [CrossRef]
17. Xiu, D.; Karniadakis, G.E. The Wiener—Askey polynomial chaos for stochastic differential equations. SIAM J.
Sci. Comput. 2002, 24, 619–644. [CrossRef]
18. Xiong, F.; Yang, S. Engineering Probability Uncertainty Analysis Method; Science Press: Beijing, China, 2015.
19. Isukapalli, S.S. Uncertainty Analysis of Transport-Transformation Models. Ph.D. Thesis, the State University
of New Jersey, New Brunswick, NJ, USA, 1999.
20. Isukapalli, S.S.; Roy, A.; Georgopoulos, P.G. Efficient sensitivity/uncertainty analysis using the combined
stochastic response surface method and automated differentiation: Application to environmental and
biological systems. Risk Anal. 2000, 20, 591–602. [CrossRef] [PubMed]
21. Haydt, S.; Lynch, S.; Lewis, S. The Effect of Area Ratio Change via Increased Hole Length for Shaped Film
Cooling Holes With Constant Expansion Angles. J. Turbomach. 2018, 140, 051002. [CrossRef]
22. Adami, P.; Martelli, F.; Montomoli, F.; Saumweber, C. Numerical Investigation of Internal Crossflow Film
Cooling. In Proceedings of the ASME Turbo Expo 2002: Power for Land, Sea, and Air, Amsterdam,
The Netherlands, 3–6 June 2002; pp. 51–63.

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).