13.

04LECTURENOTES
HYDROFOILSANDPROPELLERS

JustinE.Kerwin
January2001

Contents
1 TWODIMENSIONAL FOILTHEORY 1
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Foil Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Conformal Mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3.1 History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3.2 Potential FlowAround A Circle . . . . . . . . . . . . . . . . . . . 9
1.3.3 Conformal Mapping forDummies . . . . . . . . . . . . . . . . . . 13
1.3.4 TheKarman-TretzMapping Function . . . . . . . . . . . . . . . 15
1.3.5 TheKuttaCondition . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.3.6 PressureDistributions . . . . . . . . . . . . . . . . . . . . . . . . 19
1.3.7 Lift and Drag . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.4 LinearizedTheoryfor a2Dimensional FoilSection . . . . . . . . . . . . 26

c Justin E.Kerwin2001

Web document updated March9

i
1.4.1 Problem Formulation . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.4.2 Vortex andSourceDistributions. . . . . . . . . . . . . . . . . . . 27
1.4.3 Glauerts Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
1.4.4 ExampleThe Flat Plate . . . . . . . . . . . . . . . . . . . . . . . 35
1.4.5 ExampleThe ParabolicMeanLine . . . . . . . . . . . . . . . . . 36
1.4.6 TheDesignofMeanLines-TheNACAa-Series. . . . . . . . . . . 37
1.4.7 LinearizedPressure Coecient. . . . . . . . . . . . . . . . . . . . 40
1.4.8 ComparisonofPressureDistributions . . . . . . . . . . . . . . . . 41
1.4.9 SolutionoftheLinearizedThicknessProblem . . . . . . . . . . . 42
1.4.10 TheEllipticalThickness Form . . . . . . . . . . . . . . . . . . . . 43
1.4.11 The Parabolic Thickness Form . . . . . . . . . . . . . . . . . . . . 44
1.4.12 Superposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
1.4.13 Lighthills Rule . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
1.5 2-D VortexLattice Theory . . . . . . . . . . . . . . . . . . . . . . . . . . 52
1.5.1 Constant Spacing . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
1.5.2 Cosine Spacing . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
1.5.3 Convertingfrom
n
to(x) . . . . . . . . . . . . . . . . . . . . . 55
1.5.4 Drag andLeading EdgeSuction . . . . . . . . . . . . . . . . . . . 56
1.5.5 Adding Foil Thickness to VLM . . . . . . . . . . . . . . . . . . . 62
1.5.6 The Cavitation BucketDiagram. . . . . . . . . . . . . . . . . . . 66
2 LIFTINGSURFACES 74
2.1 Introductory Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
2.2 TheStrengthofthe Free VortexSheetinthe Wake . . . . . . . . . . . . 78
ii
2.3 The velocity induced byathree-dimensionalvortexline . . . . . . . . . . 81
2.4 Velocity Induced by aStraight VortexSegment . . . . . . . . . . . . . . 84
2.5 LinearizedLifting-SurfaceTheoryfor aPlanar Foil . . . . . . . . . . . . 87
2.5.1 FormulationoftheLinearizedProblem . . . . . . . . . . . . . . . 87
2.5.2 TheLinearizedBoundaryCondition . . . . . . . . . . . . . . . . . 89
2.5.3 Determining the Velocity. . . . . . . . . . . . . . . . . . . . . . . 90
2.5.4 RelatingtheBoundand Free Vorticity . . . . . . . . . . . . . . . 91
2.6 Lift and Drag . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
2.7 Lifting Line Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
2.7.1 Glauerts Method . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
2.7.2 VortexLatticeSolutionforthePlanarLiftingLine . . . . . . . . 104
2.7.3 ThePrandtl Lifting Line Equation . . . . . . . . . . . . . . . . . 115
2.8 LiftingSurfaceResults . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
2.8.1 Exact Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
2.8.2 VortexLatticeSolutionoftheLinearizedPlanar Foil . . . . . . . 122
3 PROPELLERS 133
3.1 Inow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
3.2 Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
3.3 Actuator Disk 140 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4 Propeller Lifting Line Theory . . . . . . . . . . . . . . . . . . . . . . . . 150
3.4.1 TheActuatorDiskas a ParticularLiftingLine . . . . . . . . . . . 157
3.5 OptimumCirculationDistributions . . . . . . . . . . . . . . . . . . . . . 161
3.5.1 AssigningThe WakePitch Angle
w
. . . . . . . . . . . . . . . . 165
iii
3.5.2 PropertiesofConstantPitch Helical VortexSheets . . . . . . . . 166
3.5.3 TheCirculationReduction Factor . . . . . . . . . . . . . . . . . . 169
3.5.4 Application of theGoldstein Factor . . . . . . . . . . . . . . . . . 172
3.6 LiftingLineTheoryforArbitraryCirculationDistributions . . . . . . . . 175
3.6.1 LerbsInduction FactorMethod . . . . . . . . . . . . . . . . . . . 175
3.7 Propeller VortexLatticeLiftingLineTheory . . . . . . . . . . . . . . . . 178
3.7.1 Hub eects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
3.7.2 The VortexLatticeActuator Disk. . . . . . . . . . . . . . . . . . 185
3.7.3 Hub and TipUnloading . . . . . . . . . . . . . . . . . . . . . . . 185
4 COMPUTERCODELISTINGS 194
5 APPENDIX 218
5.1 Derivation ofGlauerts Integral . . . . . . . . . . . . . . . . . . . . . . . 219
List of Figures
1 Illustration ofnotationforfoilsection geometry. . . . . . . . . . . . . . . 3
2 SampleoftabulatedgeometryandowdataforanNACAmeanlineand
thicknessform. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3 Anexampleofatrailingedgemodicationusedtoreducesinging. This
particularprocedureisfrequentlyusedforU.S.Navyandcommercialap-
plications. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
iv
4 Anexampleofacompletegeometricaldescriptionofafoilsection(includ-
inganti-singingtrailingedgemodications)usingafourthorderuniform
B-spline. ThesymbolsconnectedwithdashedliesrepresenttheB-spline
control polygonwhich completelydenes the shape of thefoil. The result-
ingfoil surface evaluated from the B-splineis shownas the continuous
curve. Theuppercurvesshowanenlargementoftheleadingandtrailing
edgeregions.Thecompletefoilisshowninthelowercurve.. . . . . . . . 8
5 Flow around a circle with zero circulation. The center of the circle is
locatedatx = .3, y = 0.4. Thecircle passesthroughx = a = 1.0. The
owangleofattackis10degrees. . . . . . . . . . . . . . . . . . . . . . . 10
6 Flowaroundacirclewithcirculation. Thecenterofthecircleislocated
atx= .3, y = 0.4. Thecircle passesthroughx= a= 1.0. Notethatthe
rearstagnation pointhasmovedtox = a.. . . . . . . . . . . . . . . . . . 12
7 FlowaroundaKarman-Tretzfoilderivedfromtheowaroundacircle
shown ingure 6with a specied tailangle of = 25 degrees.. . . . . . . 14
8 Flownearthetrailingedge. Thegureontheleftisforzerocirculation.
Notetheowaroundthesharptrailingedgeandthepresenceofastagna-
tion pointontheuppersurface. Thegureontherightshowstheresult
ofadjustingthecirculationtoprovidesmoothowatthetrailingedge.. . 17
9 Earlyowvisualizationphotographshowingthedevelopmentofastarting
vortex. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
10 Streamlinesandpressurecontoursforathin,highlycamberedsectionat
zeroangleofattack. Thissectionissymmetricalabout mid-chord, and
thereforehassharpleadingandtrailingedges. Asexpected,thepressure
contoursshowlowpressureontheuppersurface(green)andhighpressure
onthelowersurface(blue). . . . . . . . . . . . . . . . . . . . . . . . . . 21
11 Thisisthesamesectionas before,butatanangleofattackof10degrees.
Theow patternisnolongersymmetrical,withhighvelocitiesandhence
lowpressures(red)neartheleadingedge. . . . . . . . . . . . . . . . . . . 22
12 Closeupviewoftheowneartheleadingedgeatanangleofattackof10
degrees. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
13 Verticaldistributionofthe u velocity at themid-chord of a constant strength
vortex panelofstrength = 1. . . . . . . . . . . . . . . . . . . . . . . . 28
v
14 Illustrationofthecirculation pathused toshowthatthejumpinuvelocity
isequaltothevortexsheetstrength,. . . . . . . . . . . . . . . . . . . . 29
15 Horizontaldistributionofthev velocityalongaconstantstrengthvortex
panelofstrength = 1.. . . . . . . . . . . . . . . . . . . . . . . . . . . 30
16 Horizontaldistributionofthev velocityalongaconstantstrengthsource
panelofstrength = 1.. . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
17 Horizontaldistributionofthev velocityalongaconstantstrengthvortex
panelofstrength = 1.. . . . . . . . . . . . . . . . . . . . . . . . . . . 36
18 EnlargementofgureshowingthedierencebetweenanNACAa = 1.0
and parabolicmeanlinenearthe leading edge. . . . . . . . . . . . . . . . 42
19 Shapeandvelocitydistributionforellipticaland parabolicthicknessforms
fromlineartheory. Thethickness/chordratio,t
o
/c = 0.1. Thevertical
scaleofthethicknessformplotshas beenenlarged forclarity.. . . . . . . 43
20 Comparisonofsurfacevelocitydistributionsforanellipticalthicknessform
witht
o
/c = 0.1 andt
o
/c = 0.2 obtainedfromanexactsolutionandfrom
lineartheory. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
21 Local representation ofthe leading edge region of afoil bya parabolawith
matchingcurvatureatx = 0. Thisissometimesreferredtoasanoscu-
lating parabola.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
22 Surfacevelocitydistribution near the leading edge of asemi-innite parabola. 49
23 Vortex lattice approximationof the vortex sheets representing a marine
propeller.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
24 Arrangementofvortex positions,x
v
,and control point positions,x
c
. The
vorticesareplottedasledcircles, andthecontrol pointsareshownas
opentriangles.Thenumberof panels,N = 8. . . . . . . . . . . . . . . . 54
25 Comparisonof exact solution and vortex lattice method for a at plate
using 10and 20 panels.Thevortexsheetstrengthandtotallift coecient
isexact. Increasingthenumberof panelsimprovestheresolutioninthe
representationof(x). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
vi
26 Comparison of exact solution and vortex lattice method for a parabolic
meanlineusing10and20 panels.Thevortexsheetstrengthandtotallift
coecientisexact. Increasingthenumberof panelsimprovestheresolu-
tionintherepresentationof(x). . . . . . . . . . . . . . . . . . . . . . . 57
27 Comparisonofexactsolutionandvortexlatticemethod foranNACAa =
.8 meanlineusing10and20 panels. Thevortexsheetstrengthandtotal
liftcoecientisnotexact,butveryclosetotheanalyticresult.Theerror
in is visible near the leading edge, where VLM cannot deal with the
logarithmicsingularityinslopeofthemeanline. . . . . . . . . . . . . . 58
28 Vectordiagramofforcecomponentsonaatplate.Forclarity,theangle
ofattack,,has beendrawnatanunrealisticallyhighvalueof30degrees 60
29 Suction parameterC(x) foraatplatecomputedwith8and64vortex
elementsforunitangleofattack, . . . . . . . . . . . . . . . . . . . . . 61
30 Suction parameter C(x) for a at plate, parabolic and NACA a = .8 mean
line at unit lift coecient, computed with 32 vortex elements . . . . . . . 62
31 Comparisonofsourcelatticeandexactconformalmappingcalculationsof
thepressuredistributionaroundasymmetricalKarman-Tretzfoil. The
foilwasgeneratedwithx
c
= 0.1,y
c
= 0.0 and = 5 degrees. Source
lattice resultsaregiven for 20 panels(symbols) and 50 panels (continuous
curve). TheScherer/RiegelsversionofLighthillsleadingedgecorrection
has beenapplied. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
32 Vortex lattice approximationof the vortex sheets representing a marine
propeller.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
33 PressuredistributionsforacamberedKarmanTretzsectionattwodif-
ferentanglesofattack. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
34 Variationofpressurecoecientwithangleofattackatseveralxed chord-
wiselocationsfor asymmetricalKarmanTretzsection. Thenumbers
indicatetheapproximatechorwdiselocations,in percentofchord fromthe
leadingedge. Thedashedcurvesareforthecorresponding pointsonthe
lowersurface. Sincethefoilissymmetrical,thecurvesfor pointsonthe
lowersurface are the mirror image of the corresponding points on the upper
surface.Thefoilwasgenerated withx
c
= 0.1, y
c
= 0.0, = 10 degrees. . 68
35 ThisisthesamedataasisshowninFigure34,exceptthatitisplotted
for180chordwise positionsonthefoil. Notethat awelldenedenvelope
curveisnowapparent. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
vii
36 ThisisthesamepresentationofdataasshowninFigure35,butforthe
cambered Karman Tretzfoil computed inFigure33. Theenvelopecurve
isnolongersymmetrical,andshowstwodistinctknucklesatanglesof
attackofapproximately+3 and2 degrees.Themapping parametersare
x
c
= .05, y
c
= 0.1, = 10 degrees. . . . . . . . . . . . . . . . . . . . . . 70
37 Hereisthesametypeofdatapresentation,butforanearlyogivalfoilsec-
tion.Themapping parameters inthis case arex
c
= 0.01, y
c
= 0.1, = 20
degrees.Theregionwithintheenvelopecurvehasnownarrowedconsider-
ablyincomparisontoFigure36.. . . . . . . . . . . . . . . . . . . . . . . 71
38 SamplefamilyofbucketdiagramsforNACA-66thicknessformswithNACA
a = .8 meanlines,allwithacamberratiooff
o
/c = 0.02. Thesection
thickness/chordratios(labeledas onthechart)rangefromt
o
/c = 0.02
tot
o
/c = 0.20. ThedataisfromT.Brockett,MinimumPressureEn-
velopesforModiedNACA-66SectionswithNACAa = .8 Camberand
BuShips Type I and TypeIISections,DTMB Report1780, February1966. 72
39 DesignchartforoptimumNACA-66sectionsfromBrockett.Thelefthand
plotshowsC
P
(min) alongthex axisversust
o
/c alongthey axis. The
linesonthegraphareforconstantcamberratio,f
o
/c rangingfromzero
to0.06.Therighthandgraphshowstheresultingwidth,indegrees,ofthe
cavitationfreerangewithinthebucket. . . . . . . . . . . . . . . . . . . . 73
40 Aliftingsurface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
41 Velocitydiagraminthetangentplane. . . . . . . . . . . . . . . . . . . . . 76
42 Relating to VelocityDierence. . . . . . . . . . . . . . . . . . . . . . . 77
43 Velocityandvortexsheetstrengthforthespecialcasesoftwo-dimensional
owandfreevortexow.. . . . . . . . . . . . . . . . . . . . . . . . . . . 78
44 GeneralcaseBoundandfreevorticityispresent. . . . . . . . . . . . . 79
45 Circulation pathusedtodeterminethestrengthofthefreevorticityinthe
wake.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
46 Notationforvelocity,V

at pointP(x, y, z) induced by avolumedistribution

ofvorticity(, , ) contained involumeV. . . . . . . . . . . . . . . . . 81
viii
47 Developmentofavortexline. Ontheleftisavolumedistributionofvor-
ticity. Inthemiddle,thevolume hs been putthrough a pastamachine
toformanoodlewithcrosssectionareada. Ontheright,thenoodlehas
beenturnedintoangelshair,withzerocrosssectionalareaandinnite
vorticity,butwiththetotalcirculationkeptxed.. . . . . . . . . . . . . . 83
withthex axiscoincidentwiththevortex,andtheeld point,P located
onthey axis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
48 Notationforastraightlinevortexsegmentusingalocalcoordinatesystem
49 Normalized velocity,2yw(x, y)/ induced byastraightvortexsegment. . 86
50 Notationforaplanarhydrofoil. . . . . . . . . . . . . . . . . . . . . . . . 87
51 Cutthroughfoilsectionatxed spanwiselocation,y. . . . . . . . . . . . 88
52 Circulationcontourstogetfreevorticityonthefoil. . . . . . . . . . . . . 91
53 Controlvolumeformomentumanalysisforlift. . . . . . . . . . . . . . . 93
54 Controlvolumeforkineticenergyfardownstream.. . . . . . . . . . . . . 96
55 Concentration of boundvorticityalong aliftingline.. . . . . . . . . . . . 97
56 Interpretationofliftanddragintermsoflocalowataliftingline. . . . 98
57 PlotofrstfourtermsofGlauertscirculationseries.. . . . . . . . . . . 101
58 Plotofvelocityinduced byrstfourtermsofGlauertscirculationseries. 101
59 Notationfor avortexlatticeliftingline.Inthis case,thereare 8 uniformly
spaced panels,with a1/4 panelinsetateachend. . . . . . . . . . . . . . 105
60 Spanwisedistributionofvelocityinducedbyavortexlattice. Thespacing
isuniform,withten panelsand25%tipinset.Duetosymmetry,onlyhalf
thespanisshown.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
liftinglinewitha
1
= 1.0. Thesolutionwasobtainedwith 8 panels,using
uniformspacingwithzerotipinset. . . . . . . . . . . . . . . . . . . . . . 109
61 Comparisonof vortex lattice and exact results for an elliptically loaded
liftinglinewitha
1
= 1.0. Thesolutionwasobtainedwith 8 panels,using
uniform spacingwith 25 % tipinset. . . . . . . . . . . . . . . . . . . . . 110
62 Comparisonof vortex lattice and exact results for an elliptically loaded
ix
63 Comparisonof vortex lattice and exact results for an elliptically loaded
liftinglinewitha
1
= 1.0. Thesolutionwasobtainedwith 8 panels,using
cosine spacingwith central control points.. . . . . . . . . . . . . . . . . . 110
64 Comparisonof vortex lattice and exact results for an elliptically loaded
liftinglinewitha
1
= 1.0.Thesolutionwasobtained with64 panels,using
cosine spacingwith central control points.. . . . . . . . . . . . . . . . . . 111
65 Comparisonof vortex lattice and exact results for an elliptically loaded
liftinglinewitha
1
= 1.0. Thesolutionwasobtainedwith 8 panels,using
cosine spacingwith cosine control points. . . . . . . . . . . . . . . . . . . 111
66 Comparisonofvortexlatticeandexactresultsforatip-unloaded liftingline
witha
1
= 1.0 anda
3
= 0.2. Thesolutionwasobtainedwith 8 panels,
using cosine spacingwith cosine control points.. . . . . . . . . . . . . . . 113
67 Comparisonofvortexlatticeandexactresultsforatip-unloaded liftingline
witha
1
= 1.0 anda
3
= 0.2. Thesolutionwasobtainedwith32 panels,
using cosine spacingwith cosine control points.. . . . . . . . . . . . . . . 114
68 Liftslope,dC
L
/d,ofanellipticwingasafunctionofaspectratio,A.
(from VanDyke1975) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
69 Eectofplanformshapeonspanwisedistributionofcirculationobtained
fromPrandtlsliftinglineequation. Thefoilsallhaveanaspectratioof
= 4,andareatunitangleofattack.Equation146wasused with M = 32,
whichismorethanenoughforaconverged solution. . . . . . . . . . . . . 119
70 Notationforavortexlatticesolutionforarectangularfoil. . . . . . . . . 122
71 VortexlatticegridforarectangularfoilwithaspectratioA = 2. Inthis
example,there are 32 spanwise and 16chordwise panels. The plot on the
upperrightisanenlargementofthestarboard tipnearthetrailingedge. . 124
72 Convergenceofvortexlatticecalculationforrectangularfoilwithaspect
ratio1.0. TabulatedvaluesofdC
L
/d. Eachrowshowsconvergencewith
numberofchordwisevortices.Eachcolumnshowsconvergencewithnum-
berof spanwise panels. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
73 Vortexlatticegridforacircularfoilwithan 8 8 grid. . . . . . . . . . . 125
74 Vortexlatticegridforacircularfoilwith64spanwiseand32chordwise
panels. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
x
75 Enlargementofthetipregionofthevortexlatticegridforacircularfoil
with 64 spanwiseand 32chordwise panels. . . . . . . . . . . . . . . . . . 128
76 Vortexlatticegridforaswept,taperedfoil. Therootchordisc
r
/s = 0.5
andthetipchordisc
t
= 0.2. Theleadingedgeissweptback45 degrees.
Thegridconsistsof16 spanwiseand8 chordwise panels. One particular
horseshoeelementishighlighted.. . . . . . . . . . . . . . . . . . . . . . . 129
77 Vortexlatticegridforaswept,un-tapered foil.Therootchord is c
r
/s= 0.2
andthetipchordisc
t
= 0.2. Theleadingedgeissweptback45 degrees.
Thegridconsistsof16 spanwiseand8 chordwisae panels.. . . . . . . . . 130
78 Vortexlatticegridforaswept,un-tapered foil.Therootchord is c
r
/s= 0.2
andthetipchord isc
t
= 0.2.Theleadingedgeissweptforward45 degrees.
Thegridconsistsof16 spanwiseand8 chordwisae panels.. . . . . . . . . 130
79 Theeectofsweeponthespanwisecirculationdistribution.. . . . . . . . 131
80 Typicalnominalaxialwakeeldforasingle-screwcontainership . . . . . 135
81 Propellercoordinatesystemandvelocitynotation. . . . . . . . . . . . . . 137
82 Controlvolumeforactuatordiskmomentumcalculation.Thestreamtube
contractionhas beenexaggerated forclarity. . . . . . . . . . . . . . . . . 143
83 Resultsofnumericalcalculationofslipstreamradiusandvelocityeldin
theplaneofthediskandfardownstream.ThethrustcoecientisC
T
= 2.0.145
84 Ultimate slipstream radius as a function of thrust coecient, C
T
from
Eq.183 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
85 Streamtubesnearthetipofanactuatordiskinstaticthrust,fromSchmidt
andSparenberg. Notethatthetipstreamtube(labeled1.0) initiallygoes
upstream. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
86 Controlvolumeforactuatordiskenergy balance . . . . . . . . . . . . . . 148
87 Eciency as a function of thrust coecient for the general case of an
actuatordiskwithswirl.ThecurveforJ = 0 correspondstoEq.190. . . 149
88 Illustration ofthe concept of alifting-line propelleras alimit ofvanishing
chord length.Theradialdistributionofbladecirculation,(r) remainsthe
same,sothatthestrengthofthetrailingvortexsheet,
f
(r) isunchanged. 150
xi
89 Velocityandforcediagramat a particular radial positionon aliftingline. 151
90 Velocity induced on a lifting line at radius r
c
by a set of semi-innite
helicalvorticesoriginatingatr
v
= 1.0. Thenumberofbladesinthiscase
isZ = 5. Resultsareshownforpitchangles
w
= 10,20,30,40,50,60
degrees. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
91 Eectofbladenumberonthevelocityinducedonaliftinglineatradius
r
c
byasetofsemi-innitehelicalvorticesoriginatingatr
v
= 1.0. The
pitch angle is
w
= 30 degrees. Results are also shownfor an innite
numberofbladesfromEquations206-207. Thetotalcirculation,Z, is
keptconstantasthebladenumberisvaried,andmatchesthevalueused
forthevebladed propellershowninFigure90. . . . . . . . . . . . . . . 156
92 Axialinductionfactorsfora5bladedpropellerderivedfromFigure90.
Theenlargedplotshowsthelocal behaviornearr
c
/r
v
= 1. Theanalytical
limitofi
a
= cos
w
isplottedassquaresymbolsonthegraph. . . . . . . 158
93 Eciencyversusadvancecoecientforavebladedpropellerwithopti-
mumradialdistributionofcirculationinuniformow. Resultsaregiven
forinviscidow,andforviscousowwithsectionalLift/Dragratiosof25
and50.TheactuatordiskresultisshownasthesymbolplottedatJ
s
= 0.0. 164
94 Induced velocitiesresolved intocomponentsnormaltoandalongthehelical
surface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
95 Circulation path relating circulation aroundblades to circumferential mean
tangentialvelocity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
96 Prandtls simplied representation of the ow induced by helicalvortex sheets.171
97 Example of Circulation Reduction Factors . . . . . . . . . . . . . . . . . 172
98 KramerDiagramforIdealPropellerEciency . . . . . . . . . . . . . . . 174
99 Geometric representation of the Glauert cosine transformation . . . . . . 176
100 Illustration of the image of a 2-D pointvortex in a circle of radiusr
h
.The
vortexisatradiusr,whiletheimageisatradiusr
i
= r
h
2
/r. Ifthetwo
vorticeshaveequalandoppositestrengths,thenormal(radial)component
of thevelocityinduced by the pair ofvortices cancels at all points on the
circleofradiusr
h
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
xii
101 SampleinputdataleforPVL.Thislewasusedtogeneratetheresults
plottedingure102. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
102 Liftinglineresultsfora5bladedpropellerobtainedwiththePVLcode.
Inthisexample,there isnohub,and a Lerbsoptimumcirculationdistri-
butionhasbeenselected. C
T
= 1.0 C
P
= 1.3432 K
T
= 0.2513 K
Q
=
0.0430 V
a
/V
s
= 0.8526 = 63.47%.. . . . . . . . . . . . . . . . . . . . . 186
103 Liftingline resultsfor a 5 bladed propellerobtained withthePVLcode.In
thisexample, there is an image hub, and a Lerbs optimum circulation dis-
tributionhas beenselected. C
T
= 1.0 C
P
= 1.3744 K
T
= 0.2513 K
Q
=
0.0440 V
a
/V
s
= 0.8526 = 62.03%. Theeciencyhasbeenreduced
slightlyduetohubvortexdrag.. . . . . . . . . . . . . . . . . . . . . . . . 187
104 Propelleroperatingalone.Asubstantialcavitatinghubvortexisevident.. 188
105 Pre-swirlstatoroperatingalone.A substantialhubvortexisagainevident.
Thesignofthisvortexisoppositefromtheoneshowningure104.. . . 188
106 Propellerandstator operatingtogether.Thehubvortexhas been canceled. 189
107 Liftinglineresultsfora5bladedpropellerobtainedwiththePVLcode.
Inthisexample,there isanimagehub,and a Lerbsoptimumcirculation
distributionhasbeenmodiedtounloadthetip,usingHT = 1.0. C
T
=
1.0 C
P
= 1.4391 K
T
= 0.2513 K
Q
= 0.0461 V
a
/V
s
= 0.8526 =
59.24%. Theeciency has beenfurther reduced due to tip unloading.Note
theverydierentshapeoftheaxialinducedvelocitydistribution. . . . . . 190
108 Liftinglineresultsfora5bladedpropellerobtainedwiththePVLcode.
In this example, there is an image hub, and a Lerbs optimum circula-
tiondistributionhasbeenmodiedtounloadthehub, usingHR = 1.0.
C
T
= 1.0 C
P
= 1.3442 K
T
= 0.2513 K
Q
= 0.0431 V
a
/V
s
= 0.8526 =
63.43%.Theeciencyhasactuallyimproved,sincethereducedhubload-
ingreducesthehubvortexdrag. . . . . . . . . . . . . . . . . . . . . . . . 191
109 Liftinglineresultsfora5bladedpropellerobtainedwiththePVLcode.
Inthisexample,there isanimagehub,and a Lerbsoptimumcirculation
distributionhas been specied. Inaddition,theidealizedcounterrotating
propelleroptionhas beenselected,sothattherearenotangentialinduced
velocities.Notethatthecirculationnearthehubhas been greatlyincreased.
C
T
= 1.0 C
P
= 1.2532 K
T
= 0.2513 K
Q
= 0.0401 V
a
/V
s
= 0.8526 =
68.03%. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
xiii
110 Liftinglineresultsfora25bladedpropellerobtainedwiththePVLcode.
Theinducedvelocitiescorrespondtothoseofaninnitebladedpropeller,
andthetangentialinduced velocitieshave been canceled.Viscous draghas
beensettozero. This,therefore,correspondstoanactuatordisk. C
T
=
1.0 C
P
= 1.2071 K
T
= 0.2513 K
Q
= 0.0386 V
a
/V
s
= 1.000 = 82.84%.
Thecirculationisaconstant,withavalueofG= 0.005093andtheaxial
inducedvelocityisu

/V
a
= 0.20711,whichagreesexactlywithactuator
a
disktheory. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
.
xiv
1 TWO DIMENSIONAL FOIL THEORY
1
1.1 Introduction
Wewillbeginourexaminationofhydrofoilandpropellerowsbylookingattheow
aroundtwo-dimensionalfoilsections. Itisimportanttorecognizeattheoutsetthata
two-dimensionalowisanidealization. Flowsaroundmarinepropellers,sailboat keels
orcontrolsurfacesareinherentlythree-dimensional. Moreover,itisevenimpossibleto
create a truly two-dimensional ow in a wind or water tunnel. While the foil model
may be perfectly placed between the walls of thetunneltestsection, interaction between
the tunnel wall boundary layers and the foil generate three-dimensional features that
disturbthetwo-dimensionality oftheoweld. Reliableexperimentalmeasurementsof
two-dimensionalfoilsectionsthereforerequirecarefulattentiontotheissueofavoiding
unwantedthree-dimensionaleects.
Ofcourse, two-dimensional owscan be modeledtheoretically, andare much easier to
dealwiththanthree-dimensionalows.Moreover,thefundamentalmechanismforcreat-
ingliftaswellasmuch ofthemethodologyfordesigningoptimumfoilsectionshapescan
beexplainedbytwo-dimensionalconcepts. Designmethodsforairplanewings,marine
propellers,andeverythingin betweenrelyheavilyontheuseofsystematicfoilsection
data. But,itisimportanttorecognizethatonecannotsimplypiecetogetherathree-
dimensional wing or propeller in astrip-wisemannerfrom asequence of two-dimensional
foilsectionsandexpecttogetanaccurateanswer. Wewillseelaterwhy thisistrue,and
how twoandthree-dimensionalowscan beproperlycombined.
A surprisingly large number of methods exist for predicting the ow around foil
sections,anditisimportanttounderstandtheiradvantagesanddisadvantages. They
can be characterized inthefollowingthree ways,
1. AnalyticalorNumerical
2. Potential Flow (inviscid) methods, Fully Viscous Methods or Coupled Potential
Flow/BoundaryLayermethods
3. Exact,Linearized,or PartiallyLinearizedmethods.
Notallcombinationsofthesethreecharacteristicsare possible. Forexample, fully
viscous ows (except inafew trivialcases) must besolved numerically. Perhaps one
could construct a three-dimensional graph showing all the possible combinations, but
this will not be attempted here! In this chapter, we will start with the method of
conformalmapping,whichcaneasily beidentiedas beinganalytical,inviscidandexact.
Wewillthenlookat inviscid,lineartheory,which iscaneither beanalyticalornumerical.
Theprincipalattributeoftheinviscid,linear,numericalmethodisthatcan bereadily
extendedtothree-dimensionalows.
2
Thiswill befollowed byabrief look atsomecorrections tolineartheory,afterwhich
we will look at panel methods, which can be categorized as numerical, inviscid, and
exact.Wewillthenlookatcoupled potentialow/boundarylayermethods,whichcan be
characterizedasanumerical,exactmethod
1
Finally,wewilltakeabrieflookatresults
obtained by a ReynoldsAveragedNavier-Stokes(RANS)code, which is fully viscous,
numerical,andexact
2
.
1.2 FoilGeometry
Figure1: Illustration ofnotationforfoilsection geometry.
1
Well, moreorless. Boundarylayertheory involveslinearizingassumptionsthat the boundarylayer
is thin, but the coupledmethod makesno assumptions that the foil is thin.
2
Here we go again! The foil geometry is exact, but the turbulence models employed in RANS codes
are approximations.
3
Beforewestartwiththedevelopment ofmethodstoobtaintheowaroundafoil,
wewillrstintroducetheterminologyusedtodenefoilsectiongeometry. Asshownin
gure1,good foilsectionsaregenerallyslender,withasharp(ornearlysharp)trailing
edge,andaroundedleadingedge.Thebaselineforfoilgeometryisalineconnectingthe
trailingedgetothe pointofmaximumcurvatureattheleadingedge,andthisisshown
asthedashedlineinthegure. Thisisknownasthenose-tailline,anditslengthisthe
chord,c ofthefoil.
Theparticularcoordinatesystemnotationusedtodescribeafoilvarieswidelyde-
pendingonapplication,andone musttherefore becarefulwhenreadingdierenttexts
orresearch reports.Itisnaturaltousex, y asthecoordinateaxesfor a two-dimensional
ow, particularly if one is using the complex variable z = x+iy. The nose-tail line
isgenerallyplacedonthex axis,butinsomeapplicationsthex axisistakento bein
thedirectionoftheonsetow,inwhichcasethenose-taillineisinclinedatanangleof
attack,withrespecttothexaxis. Positivexcan beeitherorientedintheupstreamor
downstreamdirection,butweshallusethedownstreamconventionhere.
Forthree-dimensionalplanarfoils,itiscommontoorient the ycoordinate inthespan
wisedirection. Inthiscase,thefoilsectionordinateswill beinthez direction. Finally,
in the case of propeller blades, a specialcurvilinear coordinate system must be adopted,
andwewillintroducethislater.
Asshown ingure 1, afoilsectioncan bethought ofasthecombination of amean
line, f(x) withmaximumvalue f
o
and asymmetricalthicknessform, t(x), withmaximum
valuet
o
. Thethicknessformisaddedatrightanglestothemeanline,sothat pointson
theupperandlowersurfacesofthefoilwillhavecoordinates,
x
u
= x
t(x)
sin
2
y
u
= f(x) +
t(x)
cos
2
x
l
= x+
t(x)
sin
2
y
l
= f(x)
t(x)
cos (1)
2
where=arctan(df/dx) istheslopeofthemeanlineat point x.
Thequantityf
o
/ciscalledthecamber ratio, and in asimilarmanner, t
o
/ciscalledthe
thicknessratio. Ithasbeencommonpracticetodevelopfoilshapesby scalinggeneric
mean line and thickness forms to their desired values, and combining then by using
equation 1toobtain thegeometry of thefoilsurface. A majorsourceofmeanline and
4
thickness form data wascreated by the NACA (now NASA)in the1930s and1940s
3
For
example,Figure2showssampletabulationsofthegeometryoftheNACAMeanLine
a=0.8and the NACA 65A010Basic Thickness Form. Note that the tabulated mean
line has acamber ratiot
o
/c = 0.0679, while the thickness formhas athickness ratio
t
o
/c= 0.10. Includedinthetablesissomecomputedvelocity andpressuredatathatwe
will
.4
.8
1.2
1.6
0 .2 .4 .6 .8 1.0
x/c

V
( )
2
NACA 65A 010 NACA =0.8
mean line
.2 .4 .6 .8 1.0 0
.2
0
1.0
2.0
x/c
y
e
c
P
R
0
c
1
= .10 Upper surface
10 Lower surface
0
x
(per cent c)
y
(per cent c) (/V)
2
/V a/V
/V = PR/
4
PR dyc/dx
Yc
(per cent c)
x
(per cent c)
cl
i
=1.0 ; = 1.54
0
cm
c/4
= - 0.202
0.5
0.75
1.25
2.5
5.0
7.5
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85
90
95
100 0.021
0.604
1.188
1.771
2.352
2.912
3.432
3.899
4.304
4.632
4.863
4.983
4.995
4.912
4.742
4.483
4.127
3.658
3.040
2.650
2.182
1.623
1.183
0.928
0.765 0.897
0.948
1.010
1.089
1.148
1.176
1.194
1.218
1.234
1.247
1.257
1.265
1.272
1.277
1.271
1.241
1.208
1.172
1.133
1.091
1.047 1.023
1.045
1.064
1.083
1.099
1.114
1.127
1.130
1.128
1.125
1.121
1.117
1.111
1.104
1.093
1.084
1.071
1.044
1.005
0.974
0.947
0.999 0.999
0.949 0.974
0.945 0.029
0.045
0.058
0.071
0.084
0.097
0.111
0.127
0.144
0.161
0.178
0.198
0.222
0.249
0.281
0.323
0.382
0.478
0.559
0.679
0.936
1.303
1.619
1.878
2.987
0 0
0.5 0.287 0.48535
0.75 0.404 0.44925
1.25 0.616 0.40359
2.5 1.077 0.34104
5.0 1.841 0.27718
7.5 2.483 0.23868
10 3.043 0.21050
15 3.985 0.16892
20 4.748 0.13734
25 5.367 0.11101
30 5.863 0.08775
35 6.248 0.06634
40 6.528 0.04601
45 6.709 0.02613
50 6.790 0.00620
55 6.770 - 0.01433
60 6.644 - 0.03611
65 6.405 - 0.06010
70
75
80
85
90
95
100 0
1.163
2.435
3.683
4.771
5.514 - 0.12311
- 0.18412
- 0.23921
- 0.25583
- 0.24904
- 0.20385
Data for NACA Mean Line = 0.8
NACA 65A010 Basic Thickness Form
L.E. radius: 0.639 per cent c
T.E. radius: 0.023 per cent c
0.833
1.111 0.278
0.556
0.278
0
0.208
0.139
0.069
0
6.037 - 0.08790
0.893
0 0 0
0 0 0
Sample of tabulated geometry and flow data for an NACA mean line and thickness form.
refer to later.
5
Figure by MIT OCW. Adapted from Abbott and von Doenhoff, 1959.
An important geometrical characteristic of a foil is its leading edge radius, r
L
, as
showningure1.Whilethisquantity is,inprinciple,containedinthethicknessfunction
t(x),extractinganaccuratevaluefromsparselytabulateddataisrisky. Itistherefore
providedexplicitlyintheNACAtablesforexample,theNACA65A010hasaleading
edgeradiusof0.639 percent ofthechord. If youwishtoscalethisthickness formto
anothervalue,alloftheordinatesaresimplyscaledlinearly. However,theleadingedge
radiusscaleswiththesquareofthethicknessofthefoil,sothat afteen percentthick
section of thesameform would have aleading edgeradius of 1.44 percent ofthe chord.
We can show why this is true by considering an example where we wish to generate
thicknessform(2) by linearlyscalingalltheordinatesofthicknessform(1),
t
(2)
(x) = t
(1)
(x)
[t
o
/c]
(2)
(2)
[t
o
/c]
(1)
forallvaluesofx. Then,thederivativesdt/dx andd
2
t/dx
2
willalsoscalelinearlywith
thickness/chordratio. Now,attheleadingedge,theradiusofcurvature,r
L
is

3/2
1 +
dt
dx
r
L
=lim (3)
x0
d
2
t
dx
2
evaluatedattheleadingedge,whichwewilllocateatx=0. Sincetheslopedt/dxgoes
toinnityat aroundedleadingedge,equation 3 becomes

3
dt
2
r
L
=lim
dx
= (const)
t
o
(4)
x0
d
2
t
c
dx
2
which conrmstheresultstatedearlier.
Someattentionmustalso begiventothedetailsofthetrailingedgegeometry. As
wewillsee,theuniquesolutionfortheowaroundafoilsectionoperatinginaninviscid
uidrequiresthatthetrailingedge besharp.However,practicalissuesofmanufacturing
andstrengthmakesharptrailingedgesimpractical. Insomecases,foilsarebuiltwith
asquare(butrelativelythin)trailingedge,asindicatedingure2,althoughtheseare
sometimesrounded. Anadditionalpracticalproblemfrequentlyarisesinthecaseoffoil
sectionsformarinepropellers. Organized vortexsheddingfrombluntorroundedtrailing
edgesmayoccuratfrequencieswhich coincidewithvibratorymodesofthebladetrailing
edgeregion. Whenthishappens,strongdiscreteacousticaltonesaregenerated,which
arecommonlyreferredtoassinging.Thisproblemcansometimes becured bymodifying
thetrailingedgegeometryinsuchawayastoforceowseparationontheuppersurface
ofthefoilslightlyupstreamofthetrailingedge.
6
Anexampleofananti-singingtrailingedgemodicationisshowningure3. Itis
important tonotethatthenose-taillineofthemodiedsectionnolongerpassesthrough
thetrailing edge, sothat theconvenientdecomposition of thegeometry into ameanline
andthicknessformissomewhatdisrupted.
Figure3:Anexampleofatrailingedgemodicationused toreducesinging.Thispartic-
ularprocedureisfrequentlyusedforU.S.Navyandcommercialapplications.
Courtesy of U.S. Navy. Used with permission.
Theprocedureforconstructingfoilgeometrydescribedsofarisbasedontraditional
manualdraftingpracticeswhichdatebackatleasttotheearly1900s. Deningcurves
b
y sparse pointdata,withtheadditionalrequirement offairing into a speciedradius
of curvature leaves a lot of room for interpretation and error. In the present world
of CAD software and numerically controlled machines, foil surfaces and ultimately
three-dimensional propeller blades, hubs and llets are best described in terms
of
standardized geometric entities such as Non-Uniform Rational B-Splines(NURBS)
curvesandsurfaces. Asanexample,gure4showsaB-splinerepresentationofafoil
section with proportions typical ofcurrent marine propeller. In this case, the foil, together
withitssurfacecurvatureandnormal vector, isuniquelydened byaset of 16 (x, y
)
coordinates representing thevertices ofthe B-spline control polygon. This isallthat
isneededtointroducetheshapeintoacomputationaluiddynamicscode,construct
a
model,orconstructthefullsizeobject.
7
Figure4: Anexampleofacompletegeometricaldescriptionofafoilsection(including
anti-singingtrailingedgemodications)usingafourthorderuniformB-spline.Thesym-
bolsconnectedwithdashedliesrepresenttheB-splinecontrol polygonwhichcompletely
denestheshapeofthefoil. TheresultingfoilsurfaceevaluatedfromtheB-splineis
shownasthecontinuouscurve. Theuppercurvesshowanenlargementoftheleading
andtrailingedgeregions.Thecompletefoilisshowninthelowercurve.
8
1.3 ConformalMapping
1.3.1 History
Theinitialdevelopmentoftheeldofairfoiltheorytookplaceintheearly1900s,long
beforetheinventionofthecomputer. Obtaininganaccuratesolutionfortheowaround
suchacomplexshapeasafoilsection,evenintwo-dimensions, wasthereforeaformi-
dabletask. Fortunately, oneanalyticaltechnique, known asthemethodofconformal
mapping,wasknownatthattime,andprovidedameansofdeterminingtheexactin-
viscidowaroundalimitedclassoffoilsectionshapes. Thistechniquewasrstapplied
by Joukowski [ ] in1914,andthesetoffoilgeometriescreated by themappingfunction
whichhedevelopedbearshisname. Amoregeneralmappingfunction,whichincludes
theJoukowskimapping as a specialcase, wasthen introduced byKarman and Tretz
[ ]. Whileotherseveralinvestigatorsintroduceddierentmappingfunctions,thenext
signicantdevelopment was by Theodorsen [ ],whodevelopedanapproximateanalyti-
cal/numericaltechniqueforobtainingthemappingfunctionforafoilsectionofarbitrary
shape. Theodorsens work was thebasis for the development of anextensivesystematic
series offoilsections published by theNationalAdvisory Commission onAeronautics
(NACA) inthelate1930sand1940s [ ]. Theold NACAsectionresultsweredone, of
necessity, byacombination ofgraphical andhandcomputation. An improvedconformal
mappingmethod ofcomputingtheowaroundarbitrarysections,suitableforimplemen-
tationonadigitalcomputer, wasdevelopedby
4
whofound, notsurprisingly, thatin
accuraciesexistedintheearlierNACAdata. Brockettsworkledtothedevelopmentof
foilsectiondesignchartswhichareusedforpropellerdesignatthepresenttime.
Thetheoreticalbasisforthemethod ofconformalmappingisgiveninmostadvanced
calculustexts [ Hildebrand?],soonlytheessentialhighlightswill bedevelopedhere. One
startswiththeknownsolutiontoasimpleprobleminthiscasetheowofauniform
streampast a circle. The circle is then mapped into some geometry thatresembles a foil
section, and if you follow the rules carefully, the ow around the circle will betransformed
insuchawayastorepresentthecorrectsolutionforthemappedfoilsection.
1.3.2 PotentialFlowAround ACircle
Let us startwith the ow around acircle. We know thatin atwo-dimensional ideal
ow, the superposition of auniform freestream and a dipole (whose axis isoriented in
opposition to thedirection of thefreestream)willresult in adividingstreamlinewhose
formiscircular. Wealsoknowthatthisisnotthemostgeneralsolutiontotheproblem,
because wecanadditionally superimpose the owcreated by a pointvortex ofarbitrary
4
DTMB Report 1780,1966
9

strengthlocatedatthecenterofthecircle. Thesolutionisthereforenotunique,butthis
problem will beaddressedlater when we look at theresulting owaround afoil.
Tofacilitatethesubsequentmappingprocess,wewillwritedownthesolutionfora
circleofradiusr
c
whosecenter is located at anarbitrary point (x
c
, y
c
) inthe xyplane,
asshowninFigure5. Thecirclewill berequiredtointersectthe positivex axisatthe
point x =a, sothat theradius of thecircle must be,
r
c
= (x
c
+ a)
2
+ y
2
(5)
c
Figure5: Flowaroundacirclewithzerocirculation. Thecenterofthecircleislocated
atx=.3, y = 0.4. Thecircle passesthroughx=a= 1.0. Theowangleofattackis
10degrees.
Wewillseelaterthatinordertoobtain physicallyplausiblefoilshapes, the point
x=a musteither beinteriorofthecircleorlieonits boundary. Thissimplyrequires
10

thatx
c
0. Finally, theuniformfreestreamvelocitywillbeofspeedU andwillbe
inclinedatanangle withrespecttothex axis.
Withthesedenitions,the velocity components (u, v) inthe x andy directionsare

2
u(x, y) = Ucos() U
r
c
cos(2)

sin()
r 2r
r
c

v(x, y) = Usin() U sin(2) + cos() (6)

r 2r
wherer and are polarcoordinateswithoriginatthecenterofthecircle,sothat
x = x
c
+ rcos()
y = y
c
+ rsin() (7)
Notethat wearefollowing astrictright-handedcoordinatesystem,sothat positive
angles and positivetangential velocitiesare in acounter clockwisedirection. Avortex of
positivestrength,,thereforeinducesavelocity whichisinthe negativex directionon
thetopofthecircleandapositive x directionatthe bottom.
Figure 5 shows theresult in the specialcasewhere thecirculation, , has been set
tozero,andtheresultingowpatternisclearlysymmetricalaboutalineinclinedatthe
angle ofattackwhich in thiscase wasselected to be tendegrees. If,instead, we set the
circulation equal to a value of =7.778695,theowpatternshowningure 6 results.
Clearly,theow isnolongersymmetrical,andthe twostagnation pointsonthecircle
have bothmoveddown. Theangularcoordinatesofthestagnation pointsonthecircle
can beobtaineddirectlyfromequation 6 by settingr=r
c
andsolvingforthetangential
componentofthevelocity,
u
t
= vcos() usin()

= 2Usin() +
2r
c
(8)
Ifwesetu
t
= 0inequation 8anddenotetheangularcoordinatesofthestagnation
pointsas
s
,weobtain

sin(
s
) =
4r
c
U
(9)
11

Figure6: Flowaroundacirclewithcirculation. Thecenterofthecircleislocatedat

x =.3, y = 0.4. Thecircle passesthroughx =a = 1.0. Notethattherearstagnation
pointhasmoved tox=a.
Fortheexampleshowningure6,substitutingr
c
= (1.3
2
+0.4
2
) = 1.3602,=
7.778695,U = 1.0 and =10degreesintoequation9,weobtain
sin(
s
) = 0.45510 :
s
=17.1deg, 142.9deg (10)
Inthis specialcase, weseethat we havecarefullyselected insucha wayas to move
therearstagnation pointexactlytothe pointaonthex axis,since
s
=,where
=arcsin
y
c
(11)
r
c
12
1.3.3 ConformalMappingforDummies
Conformalmapping is ausefultechniqueforsolving two-dimensionalidealuidproblems
because of the analogy between the properties of an analytic function of a complex
variableandthegoverningequationsofauid. Weknowthattheowofanidealuid
intwodimensionscanberepresentedeitherbyascalarfunction(x, y) knownasthe
velocity potential,or byascalarfunction(x, y) knownasthe streamfunction. To be a
legitimateidealuidow, both mustsatisfyLaplacesequation. Theuid velocitiescan
then beobtainedfromeither,asfollows,

u = =
x y

v =
y
=
x
(12)
Nowletussupposethatthe physicalx, ycoordinates oftheuidowaretherealand
imaginarypartsofacomplexvariablez=x+ iy. Wecanconstructacomplex potential
(z) by assigning thereal part to be the velocity potential and theimaginary part to be
thestreamfunction,
(z) = (x, y) + i(x, y) (13)
Since the real and imaginary parts of each satisfy Laplaces equation, is an
analyticfunction
5
. Inaddition,thederivativeof hastheconvenientproperty of being
theconjugateoftherealuidvelocity,u+iv.Aneasyway toshowthisistocompute
d/dz by takingtheincrementdz inthex direction,
d
= = + i
dz x x x
= uiv (14)
wherethesecondlineofequation14followsdirectlyfromequation12. If youarenot
happy withthisapproach,trytakingtheincrementdz intheiy direction,and youwill
get theidenticalresult. This has to betrue,since isanalytic anditsderivative must
therefore beunique.
5
Remember,ananalyticfunctionisonethatissinglevaluedandwhosederivativeisuniquelydened,
i.e. the valueof its derivativeis independent ofthe path takentoobtain the limiting valueof /z
13
Wenowintroduceamappingfunction(z),withrealpart andimaginarypart.
Wecaninterpretthe z planeandthe graphicallyastwodierentmaps. Forexample,
if the z plane is the representation of the ow around acircle (shown in gures 5or
6),theneachpairofx, y coordinatesonthesurfaceofthecircle,oronanyoneofthe
owstreamlines, willmaptoacorresponding point, inthe plane, depending on
theparticularmappingfunction(z). Thisideamaymakemoresenseif youtakean
advanced lookatgure 7. The fancy lookingfoilshape was, indeed, mapped froma
circle.
Figure 7:FlowaroundaKarman-Tretzfoilderived fromtheowaroundacircleshown
ingure 6with a specied tailangle of = 25 degrees.
Whileitiseasytoconrmthatthecirclehas beenmappedintoamoreusefulfoil
shape,howdoweknowthattheuidvelocitiesandstreamlinesinthe planearevalid?
Theanswer isthat if (z) andthemappingfunction (z) are bothanalytic,then () is
alsoanalytic. Itthereforerepresents a valid 2-Duid ow, but it may notnecessarily be
onethatwewant. However,ifthedividingstreamlineproducesashapethatweaccept,
14

thentheonlyremainingowproperty thatweneedtoverifyiswhetherornottheow
atlargedistances fromthefoilapproaches auniformstreamofspeedU andangleof
attack.Thelatterisensuredifthemappingfunctionisconstructedinsuchaway that
=z inthelimitasz goestoinnity.
Finally,thecomplexvelocity inthe planecansimply beobtainedfromthecomplex
velocity inthez plane,
dz z
[uiv]

=
d
=
d
=
[uiv]
(15)
d
d d
dz dz
Even though we introduced theconcept of the complex potential, ,wedontactually
needit. Fromequation15,allweneedtogetthevelocityeldaroundthefoilisthe
velocity aroundthecircleandthederivativeofthemappingfunction. And,ofcourse,we
needthemappingfunctionitselftondthelocationoftheactual pointinthe plane
wherethis velocityoccurs.
Performingcomplex operationshas beengreatlyfacilitated by the availability ofcom-
puterlanguagesthatunderstandhowtodoit. Inparticular,complexarithmeticisbuilt
intotheFortran language. A listingofaFortran90 computercodecalledMAPSLispro-
vided in the lastsection of thesenotes.This code performs all of the operationsdescribed
in thissection, and mayserve as auseful guide inunderstanding the process.
1.3.4 TheKarman-TretzMappingFunction
The Karman-Tretz transformation maps a point z to a point using the following
relationship

a (z+ a)

+(za)

=
(z+ a)

(za)

(16)
where anda aregiven realconstants, whose purpose we will discover shortly. The
derivativeofthemappingfunction,whichwewillneedtotransformthevelocitiesfrom
thez planetothe planecan beobtaineddirectlyfrom16
d
2
(za)
1
(z+ a)
1
dz
= 4
2
a
[(z+ a)

(za)

]
2
(17)
Wecanseeimmediatelyfromequation16that= 1themappingfunctionreduces
to =z,sothisproducesanexactphotocopy oftheoriginalow! Notealso,thatwhen
z=a, =a.Sincewewant tostretch outthecircle,usefulvaluesofwilltherefore
begreaterthat 1.0.
15
Finally, fromequation17,thederivativeofthemappingfunctioniszerowhenz =a.
These are called critical pointsin themapping function, meaning thatstrange things
arelikelytohappenthere. Mostdicultconceptsofhighermathematicscanbestbe
understood by observing the behavior ofsmall bugs. Suppose a bug is walking along the
perimeterofthecircleinthe z plane,startingatsome pointz belowthe pointa. The
bugsfriendstarts walkingalongthe perimeter ofthefoil inthe planestartingatthe
mapped point(z). Themagnitudeanddirectionofthemovementofthesecondbug
isrelatedtothatoftherstbug by thederivativeofthemappingfunction. Ifd/dz is
non-zero,therelativeprogressof bothbugswill besmoothandcontinuous. Butwhen
therstbuggetstothe pointa,thesecondbugstopsdeadinitstracks,whiletherst
bugcontinuessmoothly. After pointa,thederivativeofthemappingfunctionchanges
sign,sothesecondbugreversesitsdirection. Thus,asharpcornerisproduced,asis
evidentfromgure7.
Theincludedangleofthecorner(ortailangleinthiscase)dependsonthewayin
whichd/dzapproacheszero.Whilewewillnotproveithere,thetailangle (indegrees)
andtheexponentinthemappingfunctionaresimplyrelated,

= 2 =180(2) (18)
180
so thatthe tailangle corresponding to = 1.86111is 25degrees, which is the value
speciedforthefoilshown ingure 7.Notethat if= 2 inequation18theresultingtail
angleiszero,i.e. acuspedtrailingedgeresults. Inthatcase,themappingfunctionin
equation 16reduces to a much simpler form whichcan berecognized as themorefamiliar
Joukowskitransformation,
a
2
=z+ (19)
z
Finally,if=1,thetailangleis =180degrees,orinotherwords,thesharpcorner
hasdisappeared. Sincewesawearlierthat =1resultsinnochangetotheoriginal
circle, thisresultisexpected. Thusweseethatthe permissiblerangeof is between
(1,2).Infact,sincepracticalfoilsectionshavetailanglesthataregenerallylessthan30
degrees,thecorrespondingrangeof isroughlyfrom(1.8,2.0).
Ifthecirclepassesoutsideofz =a,thereisnosharpleadingedge. Ontheother
hand, we canconstruct a foil with a sharp leading and trailing edge by placing the center
ofthecircleontheimaginaryaxis,sothatacirclepassingthroughz =a willalsopass
throughz =a. Thisisshowningure10. Inthiscase,theupperandlowercontours
ofthefoilcanbeshown toconsist ofcirculararcs. Inthelimitofsmall camberand
thickness,these becomethesameasparabolicarcs.
16
1.3.5 TheKuttaCondition
We can see from equation 6 that the solution for the potential ow around a circle
is not unique, but contains an arbitrary value of the circulation, . If we were only
interestedinthisparticularow,it would belogicaltoconclude,fromsymmetry,that
theonly physicallyrational valueforthecirculationwould bezero. Ontheotherhand,
ifthecylinder wererotatingaboutitsaxis, viscous forcesactinginarealuidmight
beexpectedtoinduceacirculationinthedirectionofrotation. Thisactuallyhappens
inthecaseofexposedpropellershaftswhichareinclinedrelativetotheinow. Inthis
case, a transverse force called the magnuseectwill be present. Similarly, if uid is
ejectedthroughjetsorientedtangenttothesurfaceofthecylinder,acirculationcanalso
beinduced. However,thesearenotofinterestinthepresentdiscussion,wheretheow
aroundacircleissimplyanarticialmeansofdevelopingtheowaroundarealisticfoil
shape.
Figure8:Flownearthetrailingedge.Thegureontheleftisforzerocirculation.Note
theowaroundthesharptrailingedgeandthepresenceofastagnation pointonthe
uppersurface. Thegureontherightshowstheresultofadjustingthecirculationto
providesmoothowatthetrailingedge.
Figure 8 shows the local ownear thetrailing edge for theKarman-Tretzfoil shown
ingure5. Theowintheleftgureshowswhathappenswhenthecirculationaround
thecircleissettozero.Theowontherightgureshowsthecasewherethecirculation
is adjusted to produce a stagnation point at the point a on the x axis, as shown in
gure6.Intheformercase,thereisowaroundasharpcorner,which fromequation15
willresult ininnite velocities atthat pointsince d/dz iszero. Onthe otherhand,
17
theowintherighthandgureseemstoleavethetrailingedgesmoothly. Ifweagain
examineequation15,weseethattheexpressionforthevelocity isindeterminate,with
both numerator and denominator vanishing at z = a. It can be shown from a local
expansionofthenumeratoranddenominatorintheneighborhood ofz=athatthereis
actually astagnation pointthereprovidedthatthetailangle > 0. Ifthetrailingedge
iscusped ( =o),thevelocity isnite,withavalueequaltothecomponentoftheinow
which istangenttothedirectionofthetrailingedge.
Kuttas hypothesis was that in a real uid, the ow pattern shown in the left of
gure 8 is physically impossible,andthatthecirculationwilladjustitselfuntiltheow
leavesthetrailingedgesmoothly.Hisconclusionwasbased,inpart,onaverysimplebut
clever experimentcarried out by L.Prandtl in theKaiserWilhelmInstitute in G ottingen
around 1910. A model foil section was set up vertically, protruding through the free
surfaceofasmalltank. Finealuminumdustwassprinkledonthefreesurface,andthe
modelwasstartedupfromrest. Theresultingowpatternwasthenphotographed,as
showningure 9fromanearlytext
6
. Thephotographclearlyshowstheformationof
avortexatthetrailingedgewhichisthenshedintotheow. SinceKelvinstheorem
statesthatthetotalcirculationmustremainunchanged,avortexofequalbutopposite
signdevelopsaroundthefoil. Thus,theadjustmentofcirculationisnotarbitrary,but
isdirectlyrelatedtotheinitialformationofvortexinthevicinity ofthesharptrailing
edge. Whilethisprocessisinitiatedbyuidviscosity,oncethevortexhasbeenshed,
theowaroundthefoilactsasthoughitisessentiallyinviscid.
ThisbasisforsettingthecirculationisknownastheKutta condition, andisuniversally
appliedwheninviscidowtheoryisusedtosolve both twoandthreedimensionallifting
problems. However, it is important to keep in mind that the Kutta condition is an
idealizationofanextremelycomplexrealuidproblem. Itworksamazinglywellmuch
ofthetime,butitisnotanexactsolutiontotheproblem. Wewillseelaterhowgood it
reallyis!
In the case of the present conformal mapping method of solution, we simply set
the position ofthe rearstagnation pointto
s
= . The required circulation, from
equation 9 is,
= 4r
c
Usin(+ ) (20)
6
L. Prandtl and O.G. Tietjens, Applied Hydro and Aerodynamics, 1934. Dover edition published in
18
1957
Figure9:Earlyowvisualizationphotographshowingthedevelopmentofastartingvor-
tex.
1.3.6 PressureDistributions
Thedistribution ofpressure on the upper and lowersurfaces of a hydrofoil is of interest
inthe determination ofliftanddragforces, cavitation inception, andin thestudy of
boundarylayer behavior. Thepressureeldintheneighborhood ofthefoilisofinterest
instudying the interaction between multiple foils, and in the interaction betweenfoils and
adjacent boundaries. Thepressureatanarbitrary pointcan berelatedtothepressure
at a pointfarupstreamfromBernoullisequation,
1 1
p

+
2
U
2
=p+
2
q
2
where q is themagnitude of thetotaluid velocityat the point inquestion,

q u
2
+ v
2
and (u, v) arethecomponentsofuidvelocity obtainedfromequation15. Thequantity
p

is thepressure farupstream, taken at thesame hydrostatic level. A non-dimensional

pressurecoecientcan beformed by dividingthedierence betweenthelocalandup-
streampressure by theupstreamdynamicpressure,
19
pp

2
C
P

1
= 1
U
2
U
2
Note that at a stagnation point, q = 0, so that the pressure coecient becomes
C
P
= 1.0. A pressure coecient of zero indicates that the local velocity is equal in
magnitudetothefreestreamvelocity,U,while anegativepressurecoecientimplies a
local velocity which exceedsfreestream.Whilethis is theuniversallyacceptedconvention
fordeningthenon-dimensionalpressure,many authorsplotthenegativeofthepressure
coecient. Inthatcase, astagnation pointwill beplottedwith a value ofC
P
=1.0.
20
Figure10: Streamlinesandpressurecontoursforathin,highlycamberedsectionatzero
angleofattack. Thissectionissymmetricalaboutmid-chord,andthereforehassharp
leadingandtrailingedges. Asexpected,thepressurecontoursshowlowpressureonthe
uppersurface(green)andhighpressureonthelowersurface(blue).
21
Figure11: Thisisthesamesectionasbefore,butatanangleofattackof10degrees.
Theow patternisnolongersymmetrical,withhighvelocitiesandhencelow pressures
(red)neartheleadingedge.
22
Figure12: Closeupviewoftheowneartheleadingedgeatanangleofattackof10
degrees.
23
1.3.7 LiftandDrag
Determining the overall lift and drag on a two-dimensional foilsection in inviscid ow is
incrediblysimple. Theforce(perunitofspan)directedatrightanglestotheoncoming
owofspeedU istermedliftandcan beshownto be
L=U (21)
whiletheforceactinginthedirectionoftheoncomingowistermeddragiszero. Equa-
tion21isknownasKutta-Joukowsks Law
7
.
Wecaneasily verifythatequation 21 iscorrectfortheowaround acircle by inte-
gratingthey andx componentsofthepressureactingonitssurface. Withoutlossof
generality,letusassumethatthecircleiscenteredattheorigin,andthattheangleof
attack iszero. In this case, the velocityon thesurface of the circle, from equation 8 is,

u
t
=2Usin+
2r
c
(22)
As before, wecanwrite downthepressurefromBernoullisequation,
pp

=
1
2

U
2
u
2
t

(23)
andtheliftistheintegralofthey componentofthepressurearoundthecircle,

2
L= (pp

) sin r
c
d (24)
0
By substitutingequations22and23intoequation24,andrecognizingthatonlytheterm
containingsin
2
survivestheintegration,onecanreadilyrecoverequation21.
Inasimilarway,wecanwritedowntheintegralfordrag,

2
D= (pp

) cos r
c
d (25)
0
andshowthatalltermsarezero.
Wecouldnowresorttofuzzymathandarguethatequation21mustapplytoany
foilshape. Theargument isthatwecouldhavecalculatedtheliftforceonthecirclefrom
anapplication of themomentumtheoremaround acontrol volumeconsisting of acircular
pathatsomelargeradiusr >> r
c
. Theresultmustbethesameastheoneobtained
frompressureintegrationaroundthefoil. Butifthisistrue,theresultmustalsoapply
toany foilshape,sincetheconformalmappingfunctionusedtocreateitrequiresthat
the oweldaround thecircle andaround thefoil become thesame at large values ofr.
7
The negative sign in the equation is a consequence of choosing the positive direction for x to be
downstreamand using a right-handed conventionfor positive
24
MappingSolutionsfor FoilsofArbitraryShape
Closedformmappingfunctionsareobviouslylimitedinthetypesofshapeswhichthey
canproduce. Whilesomefurtherextensions totheKarman-Tretzmappingfunction
weredeveloped,thisapproachwaslargelyabandoned by the1930s. Then, in1931, T.
Theodorsen published amethodby which one could startwith the foilgeometry and
develop the mapping function that would map it back to a circle
8
. This was done
by assuming a series expansion for the mapping function and solving numerically for
a nite number of terms in the series. The method was therefore approximate, and
extremelytimeconsuminginthepre-computerera. Nevertheless,extensiveapplication
ofthismethodledtothedevelopmentoftheNACAseriesofwingsections, including
thesamplefoilsectionshowningure2.
An improved version of Theodorsens method,suitable for implementation on a digital
computer, was developed by T. Brockett in 1966
9
. He found, not surprisingly, that
inaccuraciesexistedinthetabulatedgeometryandpressuredistributionsforsomeofthe
earlierNACAdata. BrocketsmodiedNACA-66thicknessformwasdevelopedatthat
time,andhas beenusedextensivelyforpropellersections.
Bythemid1970s,conformalmappingsolutionshadgivenwaytopanelmethods,
whichwewill bediscussinglater. Thishappenedforthreereasons,
1. Conformalmappingmethodscannot beextendedtothree-dimensionalow,while
panelmethodscan.
2. Bothmethods involve numerical approximation when applied tofoils ofa given
geometry,andimplementationandconvergencecheckingismorestraightforward
withapanelmethod.
3. Panelmethodscan beextended toincludeviscous boundary layereects.
8
Theodore Theodorsen,Theory of Wing Sections of Arbitrary Shape, NACA Rept. No. 383,1931
9
TerryBrocket,Minimum Pressure Envelopes for Modied NACA-66 Sections with NACA a=-.8 Cam-
ber and Buships Type I and Type II Sections,DTMB Report 1780,1966
25
1.4 LinearizedTheoryfora2DimensionalFoilSection
1.4.1 Problem Formulation
Inthissectionwewillreviewtheclassicallinearizedtheoryfor2-Dfoilsininviscidow.
Theproblemwill besimplied by makingtheassumptionsthatthethicknessandcamber
ofthefoilsectionissmallandthattheangleofattackisalsosmall. Theoweldwill
beconsideredas the superposition of auniformoncoming ow of speedU andangleof
attack and aperturbationvelocityeld caused by thepresence ofthefoil. Wewilluse
thesymbolsu, vtodenotethe perturbation velocity,sothatthetotaluid velocity inthe
xdirectionwill beUcos+u, whilethecomponentinthe ydirectionwill beUsin+v.
Theexactkinematic boundarycondition isthattheresultantuid velocitymust be
tangenttothefoilon boththeupperandlowersurface,
dy
u
Usin+ v
dx
=
Ucos+ u
on y=y
u
dy
l
=
Usin+ v
on u=y
l
(26)
dx Ucos+ u
However, since we are lookingforthe linearized solution, three simplications can be
made. Firstofall,since issmall, cos 1andsin . Butifthecamberand
thickness of thefoil isalsosmall, the perturbation velocitiescan be expected to besmall
compared tothe inow
10
. Finally, since the slopeofthe mean line, , is alsosmall,
thecoordinatesoftheupperandlowersurfacesofthefoilshowninequation 1will be
approximately,
y
u
(x) f(x) +
t(x)
2
y
l
(x) f(x)
t(x)
(27)
2
Introducingtheseapproximationsintoequation26,weobtainthefollowing,
dy
u
=
df(x)
+
1 dt(x)
=
U+ v
+
(x)
ony= 0
dx dx 2 dx U
dy
l
=
df(x)

1 dt(x)
=
U+ v

(x)
ony= 0 (28)
dx dx 2 dx U
Notethatthe boundaryconditionisappliedontheliney= 0 ratherthanontheactual
foilsurface,which isconsistentwiththelinearizingassumptionsmadesofar.Thisresult
10
Actually this assumption is not uniformly valid, since the perturbation velocity will not be small
in the case of the ow around a sharp leading edge, nor is it small close to the stagnation point at a
rounded leadingedge. We will see later that linear theory will be locally invalid in those regions.
26
can bederived in amoreformal way by carefullyexpanding thegeometry and oweld
intermsofasmallparameter,butthisisalotofwork,andisunnecessarytoobtainthe
correctlinearresult. Thenotationv+andv

meansthat the perturbation velocity is to

beevaluatedjustaboveandjust belowthex axis. Now,ifwetakehalfofthesumand
halfofthedierenceofthetwoequationsabove,weobtain,
df(x)
= +
[v
+
(x) + v

(x)]
dx 2U
dt(x)
=
[v
+
(x) v

(x)]
(29)
dx U
We now see that the linearized foil problem has been conveniently decomposed into
two parts. The mean value of the vertical perturbation velocity along the x axis is
determined by theslopeofthecamberdistributionf(x) andtheangleofattack, .The
jumpinverticalvelocityacrossthexaxisisdirectlyrelatedtotheslopeofthethickness
distribution,t(x). This is the key to thesolution of theproblem,since wecangenerate
thedesired even and odd behavior ofv(x) by distributingvorticesandsourcesalongthe
x axisbetween theleadingandtrailingedgeofthefoil, aswillbeshown inthenext
section.
1.4.2 VortexandSourceDistributions
The velocity eld of a pointvortex ofstrength located at a point onthex axisis,
u(x, y) =

(x
y
)
2
+ y
2
2
x
v(x, y) =
2(x)
2
+ y
2
(30)
whilethecorresponding velocity eldfor a pointsource ofstrengthS is,
S x
u(x, y) =
2(x)
2
+ y
2
v(x, y) =
2
S
(x
y
)
2
+ y
2
(31)
Wenextdeneavortexsheetasacontinuous distributionofvorticeswithstrength
perunitlength. Thevelocityeldofavortexsheetdistributedbetweenx =c/2 to
x= +c/2 will be,

c/2
u(x, y) =

2
1
c/2
(x

(
)
2
)y
+ y
2
d
v(x, y) =
2
1

c/2
(

x
(

)(

x
)
2

y
)
2
d (32)
c/2
27


Itisinstructivetolookatthevelocity eldinthespecialcaseswherethevortexstrength
isconstantover the interval. Thisresultwillalso beusefullaterwhen we look at panel
methods. Inthiscase comesoutsidetheintegral,andequation32can beintegrated
analytically,givingtheresult,
11
u(x, y) =
2

tan
1
x
y
c/2
tan
1
x+
y
c/2
(xc/2)
2
+ y
2
v(x, y) =
4
ln
(x+ c/2)
2
+ y
2
(33)
Figure13: Verticaldistributionoftheu velocityatthemid-chordofaconstantstrength
vortex panelofstrength =1.
Figure 13showsthe velocityeldobtainedfromequation 33for pointsalongthey
axis in thecasewhere the vortexsheetstrength has been set to=1.Notethat ajump
inhorizontalvelocity existsacrossthesheet,andthatthevalueofthevelocity jumpis
equaltothestrengthofthesheet. Thisfundamentalproperty ofavortexsheetfollows
directlyfromanapplicationofStokestheoremtoasmallcirculationcontourspanning
thesheet,asshowningure14.
dx = u

dx+0 +u
+
(dx) +0
= u

u
+
(34)
11
See, for example, J.Katz and A. Plotkin,Low-Speed Aerodynamics, From Wing Theory to Panel
Methods,McGraw Hill, 1991
28


Figure14:Illustrationofthecirculation pathusedtoshowthatthejumpinu velocityis
equaltothevortexsheetstrength,.
Even though gure 13 wascomputed for a uniformdistribution of(x) between x
1
and
x
2
,thelocal behavioroftheu component of velocity close to the vortexsheet would be
thesameforanycontinouslyvaryingdistribution. Ontheotherhand,thev component
ofvelocity dependson(x),butiscontinuousacrossthesheet. Figure15showsthev
componentofvelocity alongthex axis,againforthecasewhere =1.
Wecandevelopsimilarexpressionsforthevelocity eldofauniformstrengthsource
sheet. If we let thestrength of thesourcesheet be perunitlength,thevelocity eldof
asourcesheetextendingfromx=c/2 to x=c/2 will be,

c/2
u(x, y) =
1 ()(x)
d
2
c/2
(x)
2
+ y
2
1

c/2
()y
d (35) v(x, y) =
2
c/2
(x)
2
+ y
2
Again, if we specifythatthesourcestrength isconstant,equation 35can be integrated,
sogivetheresult,

ln
(xc/2)
2
+ y
2
u(x, y) =
4 (x+ c/2)
2
+ y
2
v(x, y) =
2

tan
1
x
y
c/2
tan
1
x+
y
c/2
(36)
29
Figure15:Horizontaldistributionofthev velocityalong a constantstrengthvortex panel
ofstrength =1.
Figure16showsthev componentofthevelocityobtainedfromequation36evaluated
just above andjustbelow the x axis foravalue of = 1. The jump inthe vertical
velocity isequaltothevalueofthesourcesheetstrength,whichfollowsdirectlyfroma
considerationofmassconservation.
Returningtoequation29,wenowseethat,withintheassumptionsoflineartheory,
a foilcan berepresented byadistribution ofsourcesand vortices along thex axis. The
strengthofthesourcedistribution,(x) isknowndirectlyfromtheslopeofthethickness
distribution,
dt
(x) = U (37)
dx
whilethevortexsheetdistributionmustsatisfytherelationship,

c/2
df(x)
=
1
c
()
d (38)
dx 2U x
c/2
Thesymbolcsuperimposed ontheintegralsignisthereforareason. Thiswillbe
explainedinthenextsection.
Thisdecomposition offoilgeometry,velocity eldsandsingularity distributionshas
revealed a very important result.According tolinear theory, the vortexsheetdistribution,
andhencethetotalcirculation,isunaected by foilthickness,sinceitdependsonlyon
themeanlineshapeandtheangleofattack. Thismeansthattheliftofafoilsectionis
30
Figure16:Horizontaldistributionofthev velocityalong a constantstrengthsource panel
ofstrength= 1.
unaected by itsthickness. Now,theexactconformalmappingproceduredevelopedin
theprevioussectionshowsthatliftincreaseswithfoilthickness,butonlyslightly. So,
there is no contradiction, sincelinear theory is only supposed to be valid for small values
ofthickness. Wewillseelaterthatviscouseectstendtoreducetheamountofliftthat
afoilproducesasthicknessisincreased. So,insomesense,lineartheoryismoreexact
thanexacttheory! Wewillreturntothisfascinatingtalelater.
To complete the formulation of the linear problem, we must introduce the Kutta
condition. Since the jump in velocity between the upper and lowersurface of thefoil is
directlyrelatedtothevortexsheetstrength,itissucient tospecifythatgamma(c/2) =
0. Ifthis werenottrue,there would beowaroundthesharptrailingedge.
1.4.3 GlauertsTheory
Inthissection,wewilldeveloptherelationshipbetweentheshapeofameanlineand
its bound vortex distribution following the approach of Glauert
12
. A distribution of
boundcirculation(x) overthechordinducesavelocityeldv(x) whichmustsatisfy
thelinearized boundaryconditiondevelopedearlierinequation38
12
H. Glauert, The Elements of Aerofoil and Airscrew Theory, Cambridge UniversityPress,1926
31

df
v(x) = U (39)
dx
Glauertassumedthattheunknowncirculation(x) could beapproximated by aseries
in atransformed xcoordinate, x,
c
x= cos( x) (40)
2
Note thatattheleading edge, x = c/2, x = 0, while atthe trailingedge, x =
c/2, x =. Thevalueof xatthemid-chordis/2. Theserieshasthefollowingform,

1 +cos( x)

( x) = 2U a
0
sin( x)
+ a
n
sin(nx) (41)
n=1
All terms in equation 41 vanish at the trailing edge in order to satisfy the Kutta
condition. Sincethesinetermsalso vanishattheleadingedge,theywillnot beableto
generate aninnite velocity whichmay bepresentthere. Therstterm in theseries has
therefore beenincludedtoprovideforthissingular behaviorattheleadingedge. This
rsttermisactuallythesolutionforaatplateatunitangleofattack obtainedfromthe
Joukowskitransformation,afterintroducingtheapproximationthatsin =. Itgoes
withoutsayingthatithelpstoknowtheanswer beforestartingtosolvetheproblem!
Withtheseriesforthecirculationdened,wecannowcalculatethetotalliftforce
onthesectionfromKuttaJoukowskislaw,
L = U

c/2
= U (x)dx
c/2
= 2U
2

a
0
1 +cos( x)
+

a
n
sin(nx)

dx
dx (42)
0
sin( x)
n=1
dx
Introducingtheexpressionforthederivative,
dx c
= sin( x) (43)
dx 2
32

andnotingthattheintegralofsin(nx) sin( x) overtheinterval (0, ) iszerofor n > 1, we

obtainthenalresult,
_ _
_

_
L = cU
2
_
a
0
(1 +cos( x))dx + a
n
sin(nx) sin( x)dx
_
0
n=1
0
= cU
2

a
0
+

a
1

(44)
2
Equation 44 can be expressed in nondimensional form in terms of the usual lift
coecient,
L
C
L
=
1
2
U
2
c
= 2a
0
+ a
1
(45)
Wewillnextdevelopanexpressionforthedistributionofverticalvelocity, v, overthe
chordinduced by the bound vortices,
1

c/2
()
v(x) = c d (46)
2 x
c/2
Notethattheintegralinequation46issingular,sincetheintegrandgoestoinnity
whenx =. ThisistermedaCauchyPrincipalValueintegral,whichmeansitisnot
one that youcan simply look up in thetables, and thecsymbolcenteredontheintegral
signisputtheretoserveasawarningsign. Thenextstepistore-writeequation46in
termsofthetransformedx coordinate,andtointroducetheseriesforthecirculation,
v( x) 1

a
0
(1 +cos

) +

a
n
sin(n

) sin(

)
= c
n=1
d

(47)
U
cos xcos

0
Wewillnextintroducethefollowingtrigonometricidentities inordertoputequa-
tion47intoaformsuitableforintegration,
1
sin(n) sin() = [cos((n1)) cos((n+ 1))]
2
1
cos(n) sin() = [sin((n+1)) sin((n1))] (48)
2
33

which thengivestheresult,

v( x)
=
1

a
0
(1 +cos

) +
2
1

a
n
cos((n1)

) cos((n+1)

)
(49)
n=1
U
cos xcos

0
Withthesubstitutionoftherstidentity above,theintegralofeach termoftheseries
hastheform,

cosn

)
I
n
( x) = c
cos xcos

(50)
0
A Cauchy principal value integral isobtained by takingthelimitas approacheszero
of,
_ _

c f(

)d

=lim

_ f(

)d

+ f(

)d

_ (51)
0
0 0

+
Asshownintheappendix,GlauertshowedthatI
0
= 0andI
1
=. Usingtrigono-
metricidentities,hethendevelopedarecursionformulaexpressingI
n
intermsofI
n1
andIn2. Thesolutionofthisrecursionformulaproducedthegeneralresult,
I
n
( x) =
sin(nx)
(52)
sin x
Substitutingequation52intoequation50,andmakinguseofthesecondtrigonometric
identity inequation48,thenalexpressionforthevelocitycomesoutinanamazingly
simpleform,
v( x)
=a
0
+

a
n
cos(nx) (53)
U
n=1
Solving equation 39 for df/dx and substituting equation 53 for v, we obtain the
desiredrelationship betweentheshapeofthemeanlineandtheseriescoecientsforthe
chordwisedistributionofthe boundcirculation,
34

df

=a
0
+ a
n
cos(nx) (54)
dx
n=1
The resulting expression looks line the Fourier cosine series representation for the
functiondf/dx. Henceequation54canbeinverted bytheusualmethodofharmonic
analysistogivetheresult
1

df
a
0
= dx
dx
0
2

df
a
n
= cos(nx)dx (55)
dx
0
Aparticularlyimportantresultisobtainedby solvingequation55fortheangleof
attack forwhich thea
0
coecientvanishes,
1

df

ideal
dx (56)
dx
0
Thisisknownastheidealangleofattack,andisparticularlyimportantin hydrofoil
andpropellerdesignsinceitrelatestocavitationinceptionattheleadingedge. Forany
shapeofmeanline,oneangleofattack existsforwhich thevelocity isniteattheleading
edge. Fromthesymmetryofequation56,weseethattheidealangleofattack iszerofor
anymeanlinewhich issymmetricalaboutthemidchord.
1.4.4 ExampleTheFlatPlate
Foraatplateatangleofattack wecanseeimmediatelyfromtheGlauertresultsthat
a
0
= anda
n
= 0 for n > 0. Theliftcoecientisthenfoundto beC
L
= 2andthe
boundcirculationdistribution overthechordis

1 +cos x

( x) = 2U
sin x
Thisresult,togetherwithsomeothercasesthatwewilldealwithnext,areplottedin
gure 17. In this gure, all of themean lines have beenscaled to produce a lift coecient
ofC
L
= 1.0. Inthecaseofaatplate, theangleofattackhasthereforebeensetto
= 1/(2) radians.
35

Figure17:Horizontaldistributionofthev velocityalong a constantstrengthvortex panel
ofstrength =1.
1.4.5 ExampleTheParabolicMeanLine
Theequationofaparabolicmeanlinewithmaximumcamberf
0
is
f(x) = f
o
1 (
2x
)
2
c
sothattheslopeis
df 8f
o
x
=
dx c
2
butsincex=c/2 cos x,theslopecan bewrittenas
df f
o
= 4 cos x
dx c
36
We cantherefore again solve for theGlauert coecients of thecirculation very easily,
1

df
a
0
= dx = 0
0 dx
2 4f
0

f
0
a
1
= cos xdx = 4
c 0 c
a
n
= 0 for n >1
Theliftcoecientisthengiven by theexpression
f
0
C
L
= 2a
0
+ a
1
= 2+4
c
andthecirculationdistribution becomes,

1 +cos x

f
0
( x) = 2U 8U sin x
sin x c
Thesolution for the parabolic camber linethereforeconsists of the sum of two parts
a lift and circulation distribution proportional to the angle of attack and a lift and
circulationdistributionproportionaltothecamberratio.Thisistrueforanymeanline,
except that in the general case the lift due to angle of attack is proportional to the
dierencebetweentheangleofattackandtheidealangleofattack. Thelatteriszero
fortheparabolicmeanlineduetoitssymmetryaboutthemidchord. Theresultplotted
ingure 17 is for a parabolicmeanline operatingwith alift coecient ofC
L
= 1.0 at
itsdealangleofattackwhichiszero.
1.4.6 TheDesignofMeanLines-TheNACAa-Series
From a cavitation point of view, the ideal camber line is one which produces aconstant
pressuredierenceoverthechord. Inthisway,axedamountofliftisgeneratedwith
theminimumreductioninlocalpressure. Sincethelocalpressurejumpisdirectlypro-
portional to the bound vortexstrength,such acamberline has aconstant circulation over
the chord. Unfortunately,this type ofcamberline doesnot perform up to expectations,
sincetheabruptchangeincirculationatthetrailingedgeproducesanadversepressure
gradientwhich separates theboundarylayer. Onemusttherefore beless greedy, and
acceptaloaddistribution which isconstantuptosome percentageofthechord, and
thenallowthecirculationtodecreaselinearlytozeroatthetrailingedge. Aseriesof
suchmeanlines wasdeveloped by the NACA
13
,andisknownasthea-series,wherethe
parameteradenotesthefractionofthechordoverwhichthecirculationisconstant.
13
Abbott andVon Doenho,Theory of Wing Sections, Dover1959
37
TheoriginalNACAdevelopmentofthesemeanlines,whichdatesbackto1939,wasto
achievelaminar ow wingsections. Theuse ofthesemeanlines in hydrofoil and propeller
applicationstodelaycavitationinceptionwasalaterdevelopment.
These shapescould, inprinciple, bedeveloped fromtheformulasdevelopedinthe
precedingsection by expanding thedesiredcirculationdistribution in a sineseries. How-
ever,thesineseriesapproximationtoasquarewaveconvergesveryslowly,sothatalarge
number ofterms would berequired. It istherefore better toderive the shapedirectly for
thisspecialfamilyofshapes.
Wewilldothe a= 1.0 camberlinehere,sincewealreadyhavetheexpressionforthe
velocity induced by a constant strength vortexdistributionfromequation33.Thegeneral
caseinvolvesthecombinationofaconstantandalinearlyvaryingvortexdistribution,
andwewillsimplyprovidethenalresult.
Ifwesety= 0 inequation33,shifttheoriginsothatthefoilgoesfrom(0, c) rather
thanfromc/2, c/2 inordertomatch theoriginalNACAconvention,weobtain,
14

v(x) = [log(1x/c) log(x/c)] (57)

2
Weknowthatthetotalcirculation isrelatedtotheliftcoecient
2
C
L
=
Uc
and in the specialcase ofuniform bound vortexstrength over the chord,
= c
sothat
C
L
=
U 2
Thenondimensionalverticalinducedvelocityoverthechordistherefore
v(x)
=
C
L
[log(1x/c) log(x/c)] (58)
U 4
Thelinearized boundaryconditionis
df x(x)
=
dx U
14
Toavoidtheembarrassmentofanegativelogarithm,Ireplaced(xc/2)
2
with(c/2x)
2
inderiving
the equation
38

sothat
f(x)

x
v() d
=
c 0 U c
f(x)
=
C
L
[(1x/c) log(1 x/c) + x/clog(x/c)]
c 4
Forexample,themaximumcamber,which occursatthemid chord,x=c/2,is
f
0
C
L
= (1/2 log(1/2) +1/2 log(1/2))= 0.05516C
L
c 4
Note thatthe slope ofthe mean line is logarithmically innite atthe leading and
trailingedges,which isexpectedsincetheinducedverticalvelocities,fromequation57are
innitethere. This is obviously anon-physicalsituationwhich needs to betreatedwith
somesuspicion! Asindicated at the beginning of thissection, maintaininguniform vortex
sheetstrengthrightuptothetrailingedgecertainlyviolatesthespiritoftheKutta
condition,andcan beexpectedtoresultinowseparation.
Thecorrespondingequations for the shape of the generalseries ofmean lines are much
morecomplicated,andwillnot bederivedhere. However,theyarestilllogarithmicin
form,asindicated below,
f(x)
=
C
L

1
(a
x
)
2
log |a
x
|
1
(1
x
)
2
log(1
x
)
c 2(a+1) 1 a 2 c c 2 c c
1 x 1 x x x x
)
2
+ (1 (a )
2
log( ) + gh
4 c 4 c c c c
where
1

1 1 1

log a ) + g=
1 a
a
2
(
2 4 4
1

1 1

h= (1 a)
2
log(1a) (1 a)
2
+ g
1 a 2 4
ExceptfortheNACAa= 1.0 meanline,thisseriesofmeanlinesisnotsymmetrical
about the mid chord. The ideal angles of attack are therefore non-zero, and may be
foundfromthefollowingequation,
1

df C
L
h

ideal
= dx =
0 dx 2(a+1)
39

Experience has shown that the bestcompromise between maximumextent ofconstant

circulation, and avoidance of boundary layerseparationcorresponds to a choice ofa= 0.8.
Thetabulatedcharacteristicsofthemeanline,takenfromTheoryofWingSectionsare
giveningure2.
1.4.7 LinearizedPressureCoecient
Thedistribution ofpressure on the upper and lowersurfaces of a hydrofoil is of interest
bothin thedetermination ofcavitation inception andin the study ofboundary layer
behavior. We saw in the preceding section on conformal mapping methods that the
pressure atanarbitrary pointcan berelatedtothepressure at a pointfarupstream
fromBernoullisequation,
p

+
1
U
2
=p+
1
q
2
2 2
where q is themagnitude of thetotaluid velocityat the point inquestion,
q (U + u)
2
+ v
2
and p

is the pressure far upstream, taken at the same hydrostatic level. A non-
dimensional pressure coecient can be formed by dividing the dierence between the
localandupstreampressure by theupstreamdynamicpressure,
pp

2
C
P
= 1
1
U
2
U
2
Sincethedisturbancevelocities (u, v) areassumedto besmallcomparedwiththefree
streamvelocity inlineartheory,

2
q
= 1 +2
u
+(
u
)
2
+(
v
)
2
1 +2
u
U U U U U
sothatthepressurecoecientcan beapproximated by
u
C
P
2
U
Thisisknownasthelinearized pressure coecient,which isvalidonlywherethedistur-
bance velocitiesaresmallcomparedtofreestream. Inparticular,at astagnation point
where q = 0the exact pressure coecient becomes one, while the linearized pressure
coecient givesanerroneousvalue of two!
For alinearized two-dimensional hydrofoilwithoutthickness,the ucomponentofthe
disturbance velocity at points just above and below the foil is u = /2. Thus, the
40
linearized pressure coecient and the local vortex sheet strength are directly related,
with

C
P
=
U
ontheuppersurface,and

C
P
= +
U
onthelowersurface.
Cavitation inception can be investigated by comparing the minimum value of the
pressure coecienton the foilsurface to the value of the cavitation index,
p

p
v
=
1
U
2
2
where p
v
is the vapor pressure of the uid at the operating temperature of the foil.
Comparing the denitions of andC
P
, itis evident thatifC
P
> , then p < p
v
.
Suppose that a foil is operating at a xed angle ofattack at a value of the cavitation index
sucientlyhightoinsurethatthepressureiswellabovethevaporpressureeverywhere.
It istherefore safe toassume that no cavitation will be present at this stage. Nowreduce
the cavitation number, either by reducing p

orincreasing U. The point on the foil

surfacewiththeminimumpressurecoecient, (C
P
)
min
willreachvaporpressurerst,
andthiswilloccurwhen(C
P
)
min
=. Atthis pointequilibrium canexist between
liquidandvapor,sothatinprincipleuidcanevaporatetoformacavity.
Thephysicsofthisprocessisactuallyverycomplicated, anditturnsoutthatthe
actualpressure at whicha cavity forms may be below the vaporpressure, and will depend
on thepresence of cavitation nuclei in theuid.These may be microscopicfree airbubbles
orimpuritiesintheuidoronthesurfaceofthefoil.Ifthereisanabundanceoffreeair
bubbles,asisgenerallythecaseneartheseasurface,cavitationwilloccuratapressure
veryclosetovaporpressure. Ontheotherhand,underlaboratoryconditionsinwhich
the water may be too pure, cavitationmaynotstart until thepressure issubstantially
below vaporpressure. This wasresponsibleforerroneouscavitationinceptionpredictions
in the past, before the importance of air contentwasunderstood.
1.4.8 ComparisonofPressureDistributions
Sincethevortexsheetstrength(x)/U andthelinearizedpressurecoecientisequiv-
alent,wenowhaveallthenecessaryequationstocomparetheshapeandpressuredis-
tributionsforaatplate,aparaboliccamberline,anNACAa= 1.0 meanlineandan
NACAa = 0.8meanline. Wewillcomparethemataliftcoecientofone, withall
threemeanlinesoperatingattheiridealanglesofattack. Figure17showstheshape
(includingangleofattack)ofthefoursectionsinquestion. Notethattheslopeofthe
41
atplateandparabolicmeanlineisthesameatthethreequarterchord,whichisan
interestingresultthatwewillcomebacktolater. Itisalsoevidentthattheslopeofthe
NACAa = 0.8meanlineisalsoaboutthesameatthethreequarterchord,andthat
thecombinationofidealangleofattack andmeanlineslopemakestheback halfofthe
parabolicandNACAa= 0.8 meanlineslookaboutthesame.
TheNACAa = 1.0meanlinelooksstrange,sinceitlooksmoreorlessthesameas
the parabolicmean line, but with much less camber, yet it is supposed to have thesame
liftcoecient. Thelogarithmicformofthelattermakes adierence, andwecansee
that in the blow up of therst ten percent of the chord shown in Figure 18. Even at this
largescale,however,thereisnoevidenceofthelogarithmicallyinniteslopeattheend.
Asindicatedearlier,liftpredictedfortheNACAa= 1.0 isnotachievedinarealuid,
sothatourrstimpressiongainedfromFigure18istosomeextentcorrect.
Figure18: Enlargementofgureshowingthedierence betweenanNACAa= 1.0and
parabolicmeanlineneartheleadingedge.
1.4.9 SolutionoftheLinearizedThicknessProblem
Wewillnow turntothesolutionofthethicknessproblem.Equation37givesusthesource
strength,(x) directlyintermsoftheslopeofthethicknessform,whileequation35gives
usthe velocityatany point (x, y). Combiningtheseequations,andsettingy = 0gives
us theequation for thedistribution ofhorizontal perturbation velocity due to thickness,
42

Figure19:Shapeandvelocitydistributionfor ellipticaland parabolicthicknessforms from

lineartheory. Thethickness/chordratio,t
o
/c = 0.1. Theverticalscaleofthethickness
formplotshas beenenlarged forclarity.
u 1

c/2
dt/d
U
=
2
c
c/2 x
d (59)
1.4.10 TheEllipticalThickness Form
Thethicknessdistributionforanellipticalsectionis

2x

2
t(x) = t
0
1
c
wheretheoriginistakenatthemidchord,sothattheleadingedgeisatx=c/2 and
thetrailingedgeisatx=c/2. Transformingthe chordwise variableas before,
c
x= cos( x)
2
thethicknessfunction becomes
t(x) = t
0
1 cos
2
( x) = t
0
sin( x)
43

anditsslopeis
dt dx cosx
=t
0
cos( x) =t
0
dx dx
2
c
sinx
The integral for the velocity then becomes,
_ _
cos(

)
u 1

t
0
c/2 sin(

)
c
= c _ _
U 2 0
c
(cos( x) cos(

))
2
sin(

)d
2
u t
0

cos(

) t
0
U
=
c
c
0 cos( x) cos(

))
d=
c
Linear theorytherefore yields the very simpleresult in thiscase that thenon-dimensional
horizontaldisturbancevelocity,u/U,isconstant overthechord,withavalueequaltothe
thickness/chordratiooftheellipticalsection. Itturnsoutthatthisresultisexactatthe
mid-chord,andverynearlycorrectovermostofthechord. However,lineartheoryhas
aseriousawinthatnostagnation pointresultsattheleadingandtrailingedge. Of
course,theassumptionofsmallslopesisnotvalidattheends,sothatthebreakdownof
lineartheoryintheseregionsisinevitable.
1.4.11 The ParabolicThickness Form
Theparabolicthicknessformhasthesameshapeasaparabolicmeanline,exceptthat
itissymmetricalabouty =0. Thisissometimesreferredtoasabi-convexfoil. The
shapeofthisthicknessform,anditsslopeare,

2x

2
t(x) = t
0
1
c
dt 8t
0
x
=
dx c
2
andthehorizontaldisturbancevelocity is
8t
0

u
=
1

c
c/2
c
2
d
U 2 c/2x
4t
0

c/2

= c d
c
2
c/2 x
TheaboveCauchy principalvalueintegralisoneofaseriesofsuchintegralswhose
44

evaluationisgiven by VanDyke
15
,

1 + x

c d=xlog 2
1 x 1 x
which,iftransformedtotheinterval (c/2, c/2)yieldstheresult,
u 4t
0
c/2 +x
U
=
c
2
xlog
c/2 x
c
Inthiscasethevelocityislogarithmicallyinniteattheleadingandtrailingedge,
sothatlineartheoryfailsonceagaintoproduce astagnation point! However,thelog-
arithmicsingularityisverylocal,sothattheresultisquiteaccurateovermostofthe
chord.Thefollowingtablegivesthehorizontaldisturbancevelocity forathickness/chord
ratioofonetenth. Thisresultisalsoplottedingure19,togetherwiththeresultforan
ellipticalthicknessform.
|x/c| u/U |x/c| u/U
0.498 -0.267 0.460 -0.059
0.497 -0.240 0.450 -0.041
0.496 -0.221 0.400 0.015
0.495 -0.206 0.350 0.050
0.494 -0.194 0.300 0.074
0.493 -0.184 0.250 0.092
0.492 -0.175 0.200 0.106
0.491 -0.167 0.150 0.116
0.490 -0.159 0.100 0.122
0.480 -0.111 0.050 0.126
0.470 -0.081 0.000 0.127
Note that themaximumvelocity occurs at the mid-chord and has a valueu/U = 0.127.
Anellipticalthicknessformwithathickness/chordratioofonetenthwouldhaveavalue
u/U = 0.100and wouldtherefore have a lower value of (C
P
)
min
,andwouldtherefore
be betterfromthe pointofviewofcavitationinception.
1.4.12 Superposition
We can combine mean lines and thickness forms to produce a wide range of section
shapes. The linearized pressure coecient can be determined simply by adding the
15
Van Dyke,M.D.,Second-Order Subsonic Airfoil Theory and its Practical application, NACA TN
3390,March1955
45
pressure coecients due to camber at ideal angle of attack, at plate loading due to
departurefromidealangle ofattack,andthickness. Forexample, by adding aparabolic
meanlinewithacamberratiooff
0
/c= 0.05toparabolicthicknessformwiththickness
ratiot
0
/c = 0.10, we obtain asection with aat bottomand parabolictop. This is
knownasanogival
16
section,whichwascommonlyusedforshippropellersinthepast,
andisstillusedformany quantity producedpropellersforsmallvessels.
At theidealangleofattack ofthemeanline,which iszerointhiscase,thedisturbance
velocity duetocamberatthemid chordwill be
u 1 f
0
= = 4 sin( x) = 4 0.051.0 = 0.200
U 2 U c
ontheuppersurface,and0.200onthelowersurface. Thevelocity duetothickness,we
seefromtheabovetable,will beu/U = 0.127on boththeupperandthelowersurface.
Hence,ontheuppersurface,
u
U
= 0.327 C
P
=.654
whileonthelowersurface,
u
U
=.073 C
P
= 0.146
16
Actuallytheuppercontourofanogivalsectionisacirculararc,butthisapproachesaparabolicarc
as the thickness becomes small.
46
1.4.13 LighthillsRule
Wesawintheprecedingsections thatlineartheorycannotpredictthelocalbehavior
of the ownear aroundleading edge becausetheassumption ofsmall slopes isclearly
violated. While this does not aect the overall lift, any attempt to predict pressure
distributions(andcavitationinception)neartheleadingedgewillclearlyfail. However,
since the problem is local, arelatively simplecorrection tolinear theory can be used to
overcomethisdiculty. Thisproblemwasrstsolvedby Lighthill
17
in1951. Amore
recent mathematicaltreatment of this problemmay be found in VanDyke
18
.Animproved
formulationofLighthillsmethodhas beenrecentlyintroduced by Scherer
19
,whoalso
citedearlier work by Brockett in1965,whodiscovered a1942publication (inGerman)
by F.W.Riegels.Thederivationpresentedhereisbased,inpart,onclassnotesprepared
by RobertJ. VanHoutenin1982.
Figure20: Comparisonofsurfacevelocitydistributionsforanellipticalthicknessform
witht
o
/c= 0.1andt
o
/c= 0.2obtained fromanexactsolutionandfromlineartheory.
Figure20showsthevelocity distributionneartheleadingedgeofanellipticalthick-
ness form obtained both by linear and exact theory. Linear theory gives the correct
answeratthemid-chord,regardlessofthicknessratio,butfailstopredictthestagnation
17
Lighthill,M.J.,A New ApproachtoThin Aerofoil TheoryAero. Quart 3,193-210
18
VanDyke,M.,Perturbation Methods in Fluid Mechanics,Academic Press,1964
19
Otto Scherer,Second Order Wing TheoryThe analytical Method used in Computer Code BBF,
publication pending
47
point at the leading edge. On the other hand, as the thickness ratio is reduced, the
regionofdiscrepancy between exactandlineartheory becomesmorelocal. Ifthefoil
isthin, lineartheorycanbeexpectedtoprovidethecorrectglobalresult, butthatit
must besupplemented bya localsolution inorder to becorrect at theleading edge. The
techniqueofcombiningaglobalandlocalowsolutionisknownformallyasthemethod
ofmatched asymptoticexpansions.However,wewillfollowamoreinformalpathhere.
Figure21: Local representation ofthe leading edge region of afoil by a parabolawith
matchingcurvatureatx = 0.This issometimes referredto as anosculating parabola.
Wesawearlierthattheleadingedgeradiusofafoil, r
L
,scaleswiththesquareofthe
thickness/chordratio.Ifweareconcernedwiththelocalowintheleadingedgeregion,
themaximumthicknessofthefoil occursat a pointwhichisfar awayfromtheregion
ofinterest. Infact,ifweconsultourresidentsmallbugasfarassheisconcerned,the
foilextendstoinnity inthex direction. Therelevantlengthscaleforthelocalproblem
isthereforetheleadingedgeradius. Asshowningure21,ashapewhichdoesthisisa
parabola(turnedsideways). Wecanndtheequationforthedesiredparabolaeasily by
startingwiththeequationofacircleofradiusr
L
withcenteronthex axisatadistance
r
L
back fromtheleadingedge,andexaminingthelimitforx r
y
p
2
+(x r
L
)
2
=r
L
2
y
p
(x) =

2r
L
x x
2

2r
L
x x =
y
p
2
(60)
2r
L
Thevelocity distributiononthesurfaceofaparabolainauniformstreamU
i
can be
found by conformalmappingto be
20
20
The procedure is to start with the potential solution for the ow approaching an innite at wall
48

x
q
p
=U
i
x+ r
L
/2
(61)
andthisisplottedingure22.WeusedU
i
inequation61ratherthanthefoilfreestream
velocity,U, becausethelocalleadingedgeowisreallyburiedintheglobaloweld.
Theonlyremainingtask,therefore,istoassignthepropervaluetoU
i
.
Figure22:Surfacevelocitydistributionnearthe leading edge of asemi-innite parabola.
Letusdeneu
t
(x) as the perturbation velocity due tothicknessobtainedfromlinear
theory. Thetotalsurfacevelocity accordingtolineartheoryisthenq(x) U + u
t
(x).
Inthelimitofx c,thelineartheoryresult becomesq(x) U + u
t
(0). Ontheother
hand,inthelimitofx r
L
thelocalleadingedgesolution becomesq(x) U
i
. Thus,
thefreestreaminthelocalleadingedgesolutionmustapproachU
i
=U + u
t
(0),and
thecompleteexpressionforthesurface velocity then becomes,
x
q(x) = [U+ u
t
(x)]
x+ r
L
/2
(62)
TheLighthillcorrectioncan beextendedtoincludetheeectsofcamberandangle
ofattack. Ifwedeneu
c
(x) as the perturbation velocity due tocamber atideal angle of
attack,thelineartheoryresultinthiscasecan bewrittenas
(sometimes called cornerow) and mappingthe at wall into a parabola. The derivationwill not be
presented here
49

cx
q(x) = U + u
t
(x) + u
c
(x) + U(
ideal
) (63)
x
Multiplyingequation63 by thesamefactorrepresentingthelocalleadingedgeow
givestheresult,
q(x)
= 1 +
u
t
(x)
+
u
c
(x) x cx
(64)
U U U x+ r
L
/2
+(
ideal
)
x+ r
L
/2
Notethatsquarerootinnityinequation63hasnowbeencanceled, andthatthe
velocityattheleadingedgehasthenitevalue
q(0)
= (
ideal
)
2
=
C
L
C
L
ideal
2
(65)
U r
L
/c 2 r
L
/c
The velocity predicted from equation 65 can be compared with NACA tabulated
resultsobtainedfromTheodorsensconformalmappingmethod for a variety offoil types.
This isdone in thetable below, and it isclearthat the Lighthillcorrectionworks very
well.
Velocityq/U attheleadingedgeforvariousthicknessformsatunitliftcoecient
Section Type
NACA16-0006
NACA16-0012
NACA16-0021
NACA63A006
NACA0006
r
L
/c
0.00176
0.00703
0.02156
0.00265
0.00400
Lighthill
5.37
2.68
1.53
4.37
3.56
Theodorsen
5.47
2.62
1.57
4.56
3.99
Inadditiontocorrectingthevelocity rightattheleadingedge,wecanuseLighthills
ruletomodifythevelocityandpressuredistributionfromlineartheoryoverthewhole
forward part of the foil. However, if we were to apply equation 64 to anelliptical thickness
form, we wouldndthattheresult would beworseatthemidchord. Forexample,we
knowthattheexactvalueofthesurfacevelocityatx/c= 0.5 foranellipticalthickness
form with a thickness/chord ratio of 20 percent is q/U = 1.2, and that results in a
pressurecoecient ofC
P
=0.44.Wewouldalsogetthesameresultwithlineartheory.
However,ifweapplyequation64,wewouldget,
50

2
r
L
= 0.5
t
o
= 0.02

q/U = 1 +
t
o
0.5
= 1.177
c 0.5 + r
l
/2
C
P
= 0.385 (66)
Assuggested by Scherer, a variant of theoriginal Lighthill formula solves this problem,
andcan bederived asfollows. If we take thederivativeequation 60,
dy
=

r
L
(67)
dx 2x
andformthequantity

1 x
=
r
L
(68)
1 +(
dy
)
2 x+
2
dx
weseethatwerecoverthesamemultiplicative factorasshownbeforeinequation61.
However,foractualfoilleadingedgeshapes,theslopewilldierfromthatofaparabola
asxincreases, andnaturally goes tozero at the point ofmaximumthickness. Theresult
is exact for an ellipse, and appears to give better results for typical foil shapes. We
willshowthislaterwhenwelookatnumericalmethodsforsolvingtheproblemofan
arbitrarytwo-dimensionalfoil.
51
1.5 2-D VortexLatticeTheory
Intheprecedingsections,wesawhowthesolutionforthelinearizedproblemofatwo
dimensionalfoilsectioncouldbesolved. Inparticular,analyticsolutionscouldbede-
rived for a variety of mean lines and thickness forms. However, this is generally not
possibleforthreedimensionalfoils. Asaresult,onemustresorttonumericalmethods
forsolvingsuch problems. Itthereforemakessensetorstdevelopnumericalsolutionsto
the two-dimensionalproblem,where wecan check theresultsagainstknownexactan-
swers. Oneveryeectiveway ofsolvingthreedimensionalliftingproblemsistorepresent
thecontinuous vortexsheet byalattice ofconcentrated,straightline, vortexelements.
Figure23isanexampleofavortexlatticerepresentationofamarinepropeller.
Figure23: Vortexlatticeapproximationofthevortexsheetsrepresentingamarinepro-
peller.
Todevelopavortexlatticesolutiontothetwo-dimensionalfoilproblem,wesimply
dividethe chord intoN panels,concentratethecontinuousvortexsheet(x) intodiscrete
point vortices
n
somewherewithineachpanel,andcomputetheverticalvelocityv(x)
by summing theinuence of the pointvortices. If weare given the shape of themean
line, then the integral equation forthe bound vortex sheet strength becomes aset of
linearequationsforthediscretevortexstrengths.
52
1.5.1 ConstantSpacing
Theearliest vortexlatticemethod (VLM) was due to Faulkner
21
,whodividedthechord
intoequallyspacedpanels,andlocatedthevortexatthequarterchordofeachpanel,
and computed the velocity at control points located at the three quarter chord. We
saw in gure 17that the slope (and hence the vertical velocity) at the three quarter
chordofaatplateandaparabolicmeanlinewasthesame. Thisfactprobablyledto
Faulknersdiscoverythattheoptimumlocationofthecontrol pointineachpanelwas
atthissamelocation. Onecaneasilyverifynumericallythatthequarterchord/three
quarter chordarrangement of vorticesandcontrol points givestheexacttotalcirculation
forany numberofpanelsforaatplateandaparabolicmeanline.However,thedetailed
chordwisedistributionofcirculationisnotexactforthesecases,butgetsmoreaccurate
withincreasingnumbersofpanels. Curiously,thestrengthoftherstpanelisalways
eleven percenttoolow!
ItisinterestingtonotethatnoexplicitKuttaconditionisemployedwiththismethod.
Aslongasthecontrol pointsareplaceddownstreamofthe vortices,the vortexlattice
method will converge to the desired solution in which the circulation vanishes at the
trailingedge.
A formal proof ofthese properties of theconstantspacing VLM is given by James
22
.
1.5.2 CosineSpacing
Whiletheconstantspacingscheme works well, betterarrangementemployscosine spac-
ing. This algorithm was proved to be rigorous independently by Lan
23
and Stark
24
.
Anotherproofcan beobtained by rstsolvingfortheowaround acircleatanangle
ofattack using pointvortices andcontrol pointsdistributed atuniformangular intervals
aroundthecircumference. Itcan beshownthatthe local vortexsheetstrength,obtained
by dividingthe point vortexstrengths by the intervalarclength, isexact. Mappingthe
circletothex axisusingtheJoukowskitransformationagainleadstothesamecosine
distributionforcontrol pointsand vortices.
Werstdeneanauxiliaryangularvariable x such that
21
Falkner,V.M., The Solution of Lifting-Plane Problems by Vortex Lattice Theory,Aeronautical
ResearchCouncilR&M No.2591,1947
22
James,R.M., On the Remarkable Accuracy of the Vortex Lattice Method, Comput. Methods
Appl. Mech. Eng. Vol.1, 1972
23
Lan,C.E.,AQuasi-VortexLatticeMethodinThinWingTheory,J.Aircraft,Vol11,No. 9,Septem-
ber, 1974
24
Stark,V.J.E.,AGeneralizedQuadratureFormulaforCauchyIntegrals,AIAAJournalVol9,No.9,
1970
53


Figure 24: Arrangementofvortexpositions, x
v
, andcontrol point positions, x
c
. The
vorticesareplottedasledcircles,andthecontrol pointsareshownasopentriangles.
Thenumberof panels,N = 8.
c
x= (1 cos x) (69)
2
sothat x= 0 correspondstotheleadingedgeand x= correspondstothetrailingedge
x=c.This isthesameasthe change of variables introduced by Glauert,exceptthatthe
xcoordinatehas beenshifted byc/2 toplacetheleadingedgeat x=0. Wenextdivide
thechordintoNequalintervalsof x withcommonintervalx =/N. Pointvortices

n
arelocatedatthemid-pointsofeach xinterval,andcontrol pointsarelocatedatthe
downstream boundary ofeach x interval,
c
1 cos
(n1/2)
x
v
(n) =
2 N
x
c
(n) =
2 N
c
1 cos
n
(70)
Notethatwiththisspacingalgorithmthelastcontrol pointisatthetrailingedge,
x=c.
Thevelocity inducedatthen

th control pointissimply,
54

N
1

m
df
v
n
=
2
m=1
x
v
(m) x
c
(n)
=U
dx
(71)
wherethelastequality in 71 is astatement ofthe boundaryconditiondevelopedearlier
Equation71,writtenforeachoftheN control points,representsasetofsimultaneous
equationsfortheunknown pointvortexstrengths
n
.
1.5.3 Convertingfrom
n
to(x)
Convertingeach pointvortexstrengthto vortexsheetstrengthrequiresdivision byan
appropriateintervalx
n
. Sincethecommonintervalinthe xvariableisx
n
=/N,the
corresponding interval inthe physicalcoordinate is
dx
x
n
= x
dx
c
= sin x
2

N
x(cx)
= (72)
N
wherewehaveconvertedfrom x tox inthelastequation by usingequation69. Thus,
thevortexsheetstrengthsare
(x
n
) =
N
n
(73)
x
n
(cx
n
)
ThisprocedureisillustratedinthecomputercodeVLM2Dwhich islistedintheCodes
sectionofthenotes. Thisprogramsolvesforthechordwisedistributionofcirculationfor
aatplate,aparabolicmeanlineandanNACAa=.8 meanline,allwithaprescribed
camberandangleofattack.
As with constant spacing, the total circulation is exact for the at plate and the
parabolaforanynumberofpanels(including1). Thedistributionofcirculationisexact
fortheatplateforonepanelorgreater. Itisexactforaparabolawithtwoelements
orgreater.
ThetotalliftandthedistributionofcirculationisnotexactfortheNACAa =.8
meanlineforanynumberofpanels,buttheresultswithtenpanelsareaccurateenough
55
formostpurposes. Withtenpanels thetotalcirculationis 0.6 percentlow, whilefor
twenty panelsitis0.2 percentlow. Onecouldnot possiblymanufactureamodeland
testitwithsucientaccuracytodetectthissmalldiscrepancy!
ThefollowingguresshowtheresultsobtainedwithVLM2Dforthecasesjustde-
scribed.
Figure25: Comparisonofexactsolutionandvortexlatticemethodforaatplateusing
10and 20 panels.Thevortexsheetstrengthandtotallift coecient isexact.Increasing
thenumber of panels improves the resolution in the representation of(x).
1.5.4 DragandLeadingEdgeSuction
We knowthat the drag of a two-dimensionalfoil iszero in theabsence ofviscousforces.
However,anticipatingthat vortexlatticemethodswill beusedforthree-dimensionalows
whereinduceddragisnotzero,itisusefultolookattheproblemofcomputingdragfor
two-dimensionalfoils.
56

Figure26: Comparison ofexactsolutionandvortexlatticemethodfor a parabolicmean

lineusing 10and 20 panels. Thevortexsheetstrengthandtotallift coecient isexact.
Increasing thenumber of panels improves the resolution in the representation of(x).
Considerthesimplecaseofaatplateatangleofattack. Theexactsolutionfor
thevortexsheetstrengthis
cx
(x) = 2Usin (74)
x
Thetotalvelocity ontheuppersurfaceoftheplateisUcos (x)/2, whilethe
correspondingvelocityonthelowersurfaceisUcos+(x)/2.Integratingtheresulting
pressurejumpoverthechordoftheplateresultsinaforcenormaltotheplate,
1
F
N
=
2
U
2
c2sincos (75)
57
Figure27: ComparisonofexactsolutionandvortexlatticemethodforanNACAa=.8
meanlineusing 10and 20 panels. Thevortexsheetstrengthandtotallift coecient is
notexact,butveryclosetotheanalyticresult.Theerrorin isvisibleneartheleading
edge,whereVLMcannotdealwiththelogarithmicsingularityinslopeofthemeanline.
Converting F
N
to aforcecoecient by dividing by
1
U
2
c,andresolvingtheforce
intocomponentsnormaltoandinthedirectionoftheundisturbedowgives,
C
N
= 2sincos
C
L

= 2sincos
2

C

= 2sin
2
cos (76)
D
Obviously there is something wrong since the dragshould be zero, and we have labeled
C
L
andC
D
inequation 76withasterisksas a warningthattheseareapparentvalues.The
missingingredient istheleadingedgesuctionwhichcan bethought ofasthelimitingcase
ofaninnitevelocity (andhenceinnitelylowpressure)actingontheinnitelysmall
frontalareaoftheplate.
2
58

Weknowthatthecirculation,andhencelocalvelocity,goestoinnity attheleading
edgeas,
c
(0) 2Usin (77)
x
sothatthestrengthoftheleadingedgesingularity,whichwewillcallC,is,
C =lim (x) x/c = 2Usin (78)
x0
Aformalanalysisinvolvingtheintegrationofpressurearoundtheleadingedgeofa
foilinthelimitingcaseofvanishingleadingedgeradiusyieldstheresultthatthethereis
anitesuctionforcewhichisproportionaltothesquareoftheleadingedgesingularity,
C. However, wecan avoidthismathematicalextravaganza by imposing therequirement
that,fromKuttaJoukouwskislaw,thedrag must bezero. Asillustratedingure28,
thenecessaryleadingedgesuctionforce,F
S
must be,
F
s
=
1
U
2
c2sin
2
=cC
2
/4
2
C
S
= 2sin
2
=

2
(79)
2 U
whichactsinthenegativex direction. Ifweresolvethesuctionforceintocomponents
atrightanglesto,andinthedirectionoftheinowandaddthesetothecorresponding
componentsofthenormalforcefromequation76,weobtain,
C
L
= C
N
cos+ C
S
sin= 2sin
C
D
= C
N
sinC
S
cos= 0 (80)
Equation80conrmsthatthenetdragisnowzero,butalso,thattheliftforcenow
agreeswiththe value given by KuttaJoukowskis law.
In a numericalsolution we only knowdiscrete values of theconcentrated bound vor-
ticesandthereforedonotknowthevalueofC. However,wecanapproximateitintwo
dierentwayseither by extrapolating the bound vortexsheetstrength insucha wayas
toobtainthesingularity parameterC, or by using amethod developed by Lan, inwhich
C isrelatedtotheupwashcomputedattheleadingedge.
59

Figure28: Vectordiagramofforcecomponentsonaatplate. Forclarity,theangleof

attack,,has beendrawnatanunrealisticallyhighvalueof30degrees
Fortherstmethod,wedenea suction parameterC(x) as
C(x)
N(x
v
)
=(x
v
)
x(cx)
(81)
c c
2
Comparingequations81and78,weseethatinthelimitasx0,C(x) approaches
theleading-edgesingularity parameterC. Forthespecialcaseofaatplate,combining
equations74and81givestheresultforunitangleofattack andinowspeed,
C(x) = 2(cx) C(0) C = 2 (82)
Figure 29 provides a check on the numericalresults produced byVLM2D.Thediscrete
valuesof
n
areconvertedtoC(x) usingequation81andplottedfor N = 8 and N =64.
The results for both discretizations are exactly on the line (at least to 5 signicant
digits). Thisdemonstratesthatonecouldobtainanumericalvaluefortheleadingedge
singularity parameter,C, by linearextrapolationofC(x).
Figure30provides acomparison ofC(x) foraatplate, parabolicmeanline and
NACAa= 0.8 meanline,allwithunitliftcoecient. Theresults werecomputedusing
60

Figure29:Suction parameterC(x) for a at plate computed with 8 and 64vortexelements

forunitangleofattack,
64 vortex elements. Note that there is no leading edge suction, i.e. C = 0, for the
parabolic and a = 0.8 mean lines since they are at their ideal angle of attack. The
numericalresultsareexactfortheparabola,butslightlyoforthea = 0.8meanline.
However,theslighterrorisnotvisibleonthegraph.
Thesecondmethodmakesuseofacuriousresultderivedby Lanwhichstatesthat
theleadingedgesingularity parameterC isequaltothetotal vertical velocity (upwash)
computed atthe leadingedgedivided by thenumber ofelements, N. Theproofwill
not be givenhere,but numerical validation is verysimple. Movingthecontrol point in
equation71totheleadingedgeandaddingtheverticalcomponentoftheinowgives
1
N

m
v
le
(n) = Usin+
2 x
v
(m)
(83)
m=1
Numerical evaluation ofequation 83, usingdiscretevalues of
n
,recoverstheexact
valueofC fortheatplateandparabolicmeanline.
Beforeleaving thetopicofleadingedge suction, itisworthwhile toconsider what
happensinthecaseofarealuid,orinthecaseofanidealuidwithcavitationpresent.
Clearly, ifthereisasharpleadingedge,theleadingedgesuctionforcecannotmaterialize,
61
Figure30: SuctionparameterC(x) foraatplate,parabolicandNACA a =.8mean
line at unit lift coecient, computed with 32 vortexelements
sotheapparentdraggiveninequation76isreal,andadverselyaectstheeciencyof
thefoilsection. Ifthefoilhas aroundedleadingedge,butonethat is poorlydesignedor
manufactured,someorallofthetheoreticalleadingedgesuctionmay belost. Finally,
ifthesectioniscavitatingattheleadingedge,thelocallowpressureattheleadingedge
will belimited byvaporpressure, thusreducing (or eveneliminating) theleading edge
suction.
1.5.5 Adding FoilThicknesstoVLM
Thecontinuousdistributionofsourcestrength,(x),which representsfoilthicknesscan
also beapproximated byaset ofconcentrated pointsourcesofstrengthS
n
. Itmakes
sensetolocatethe pointsourcesinthesame positions,x
v
asthediscretevortices. To
maintainthesamemassux,thestrengthofthen

thpoint source must be,

xc(n)
S
n
= ()d
xc(n1)

xc(n)
dt
= U d
xc(n1) d
62

= U t
(
n) t
(n1)
(84)
Figure31:Comparisonofsourcelatticeandexactconformalmappingcalculationsofthe
pressuredistributionaroundasymmetricalKarman-Tretzfoil. Thefoilwasgenerated
withx
c
=0.1,y
c
= 0.0 and = 5 degrees.Sourcelattice results are given for 20 panels
(symbols)and50 panels (continuouscurve). TheScherer/RiegelsversionofLighthills
leading edge correctionhas beenapplied.
where t
n
t(x
c
(n)). The discrete source strengths are therefore simply proportional
tothedierence infoilthickness betweenthe twocontrol point positionssurrounding
thesource. Theoneexceptionistherstsource,whosestrengthisproportionaltothe
dierence betweenthethicknessattherstcontrol point(whichisdownstreamofthe
source)andthethicknessatleadingedge,whichiszero. Thesingularintegralgivenin
equation 59 for thehorizontal velocity induced by thesources thenreduces to the simple
summation,
N
u
n
1

S
m
U
=
2 x
c
(n) x
v
(m)
(85)
m=1
The velocity induces by thickness cantherefore be easily added to the velocity due to
thevortexsheetinordertoobtainanapproximatefoilpressuredistribution. Inaddition,
63
the Lighthill leading edgecorrection can be applied to the total computed velocity prior
tocomputingthepressure. ThisprocedureiscontainedintheprogramVLM-2D,which
is listed in the appendix. The program can accommodate either parabolic or NACA
a =.8meanlineswithDTMBmodiedNACA66thicknessforms. Inordertoenable
comparison with exact conformal mapping solutions, the program can also import a
symmetricalKarman Tretzthicknessformgenerated byMAPSL.
The latter feature was used to generate gures 31 and 32. In the rst gure, a
symmetrical section with athickness/chord ratioofabout 12 percent was set at zero
angle of attack in order to test only the source lattice feature of the program. The
resultsobtainedwith20and50panelsarequiteclosetoeach other,althoughitisevident
thatthe50panelresultisslightlyclosertotheexactconformalmappingsolution. The
Scherer/RiegelsversionoftheLighthillleadingedgecorrection,asgiveninequation68
has beenapplied,and does aremarkably good job ofpredictingthepressuredistribution
neartheleadingedge.
Figure 32 shows thesamefoil at an angle ofattack of 5degrees, sothat now both
sourcesandvorticesarepresent. Theresultsobtainedwith20panelslookverysimilarto
theexactresults,butclearlyindicatelesslift. Thisisnotsurprising,sincelineartheory
doesnotaccountforthecoupling betweenliftandthickness. Inthiscase,theconformal
mappingsolutionindicatesthatthisparticularfoilsectiongeneratesaboutten percent
moreliftthanaatplateatthesameangleofattack. Iftheconformalmappingsolution
isrun at areducedangle ofattack ( actually 4.53degreesratherthan 5.00degrees),the
resultingpressuredistributionisshownasthedashedlineinthegure.Thelatteragrees
almost perfectlywiththelineartheoryresult,andtheLighthillcorrectioncapturesthe
nitepressure peakneartheleadingedge very well.
64
Figure32: Vortexlatticeapproximationofthevortexsheetsrepresentingamarinepro-
peller.
65
1.5.6 TheCavitationBucketDiagram
Figure33:Pressuredistributionsfor a cambered Karman Tretzsectionattwodierent
anglesofattack.
Theprecedingsectionshavedescribedavariety ofmethodsfordeterminingthepres-
suredistributionsaroundfoils.Theycantherefore beusedtodeterminecavitationincep-
tion by nding the minimumpressure on the foil, and comparing it with vaporpressure.
In non-dimensional terms, this is the same as equating the negativeof the minimum
pressurecoecient,[C
P
]
min
withthecavitationnumber,. Thiscan beillustrated by
gure33, which shows thepressure distribution aroundafoilsection attwo dierent
angles ofattack. Atan angle ofattack ofonedegree, theminimumpressureoccursat
a pointapproximately35%ofthechordfromtheleadingedgewherethepressurecoef-
cientisC
P
0.8. Atanangle ofattack of 4degrees,theminimumpressure occurs
practically at the leading edge, with apressure coecient ofC
P
2.0.
66

But, toobtain this kind ofinformation, we do not need the entirepressuredistribution

onthefoiljusttheminimumvalueattheparticularangleofattackinquestion. We
willnowdescribeaconvenientway ofpresentingthisinformation,calledthecavitation
bucketdiagram.
Thesurfacevelocity,q, at a xedpoint onthefoilvarieswithangleofattackas
q(x) = a(x) cos + b(x) sin (86)
wherea, bdepend on the shape of thefoil and the position along the chord oneither the
upperorlowersurface. Thereasonthatthisistruecan beseen bothfromlineartheory
andfromanexactconformalmappingsolution. Thelineartheoryresult
25
isdirectly
evidentfromequation63,where
cx
a(x) = U + u
t
(x) + u
c
(x) U
ideal
x
cx
b(x) = U (87)
x
Theexactconformalmappingresultcan beobtained bycombiningequations 8and
20. Thesurfacevelocityonthemappingcircleis
[q]
z
= 2U[sin(+ ) +sin(+ )]
= 2U[sincos+cos sin+sin cos+cos cos]
= 2U[(sin+sin ) cos +(cos+cos ) sin] (88)
Sincethe velocityonthefoil, [q]

, isobtainedfromthe velocityonthecircle by dividing

by thederivativeofthemappingfunction(which isobviouslyindependent of ), theform
ofequation88isunchanged,andthecombinationofallthefactorsthataremultiplied
by cos and sin yields the coecients a, b in equation 86. Thus, regardless of the
complexityof the shape of the foil, the dependence of thesurfacevelocityon atany
point onthefoilisverysmooth,andispracticallyalinearfunctionforsmallvaluesof
theangleofattack.
Thisisillustratedingure34forasymmetricalKarman-Tretzfoil. Note,inpar-
ticular,that the variation inpressure coecient right at theleading edge, (x/c=0),is
approximatelyparabolicinform. Thisfollowsdirectlyfromequation65,whichshows
thatthetotalvelocity atx= 0 iszeroattheidealangleofattack,andincreaseslinearly
with
ideal
at arate which is inversely proportional to thesquare root of theleading
edgeradius. Thepressurecoecientisthereforequadraticin.
25
Keepingin mind that asfar as linear theory is concerned,cos = 1.0 andsin =.
67
Figure34:Variationofpressurecoecientwithangleofattackatseveralxed chordwise
locationsfor asymmetricalKarman Tretzsection. Thenumbersindicatetheapproxi-
matechorwdise locations, in percent ofchord fromthe leading edge. Thedashedcurves
areforthe corresponding pointsonthelowersurface. Sincethefoil issymmetrical,the
curvesfor pointsonthelowersurfacearethemirrorimage ofthe corresponding points
ontheuppersurface.Thefoilwasgeneratedwithx
c
=0.1, y
c
= 0.0, = 10 degrees.
Itisalsoevidentfromgure34thatthepressurecoecientincreasesrapidlywith
angleofattackfor pointsneartheleadingedge,andisrelativelyinsensitivetoangleof
attack for pointsnearthetrailingedge. Thisfollowsdirectlyfromequation87.
If wereconstructgure 34with alarge number of pointsalongthefoilsurface,the
individual C
P
() lines forman intricate spider-web patternwhich reveals adistinct
envelopecurve,asillustratedingure35. Thisenvelopecurve
26
actuallycontainsall
theinformationthatweneedtodeterminecavitationinception. Forexample,suppose
thatthis foilwere tobe operatedatanangleofattack of, say, ve degrees. We can
then see immediately from the graph that [C
P
]
min
1.7. Thus, cavitation can be
26
Evidently someone thought this looked like a bucket turned sideways. This particular one looks
more like aspade to me, but we will stick to tradition here!
68
Figure35:ThisisthesamedataasisshowninFigure34,exceptthatitisplotted for180
chordwise positions on thefoil.Notethat a well dened envelopecurve isnow apparent.
expectedtooccur
27
whenthecavitationnumber = 1.7. Ontheotherhand,suppose
thecavitationnumberisxedatavalueof= 2.0.Then,itisevidentfromthebucket
diagramthat thefoilwill be expected to operatefree of cavitation over arange of angles
ofattack between5.8 degrees.
Figure 36 shows theresults of a similarcalculation for a camberedfoil.The bucket now
showsthreedistinctregions. Thetopofthebucketindicatestheinceptionofcavitation
near the leading edge on the upper surface, and the bottom of the bucket indicates
inceptionneartheleadingedgeofthelowersurface. Theleftsideofthebucketindicates
cavitationinceptionontheuppersurfacesomewhereinthemid-chordregion.
Thenextexampleshowstheresultsfora(nearly)ogivalsection,whichhasasmall
but nite leading edge radius. Due to the rapidincrease in velocity with angle of attack at
theleadingedge,thebucketisextremelynarrow,indicatingavarysmallcavitation-free
rangeofangleofattack.
27
Assuming that the uid containssucient cavitation nuclei forinception tooccurwhen p =p
v
.
69
Figure 36: Thisisthe samepresentation ofdata as showninFigure 35, but for the
camberedKarmanTretzfoilcomputedinFigure33. Theenvelopecurveisnolonger
symmetrical,andshowstwodistinctknucklesatanglesofattackofapproximately+3
and2degrees.Themapping parametersarex
c
=.05, y
c
= 0.1, = 10 degrees.
Practical design chartscan beconstructed by omitting thespider-webs shown in
theprecedinggures,andplottingthebucketsforsystematicallyvaryingfoilgeometries.
Anexample ofthis type of chartdeveloped by Brockett isshown ingure38.Thisclearly
showsthatathinfoil(withasmallleadingedgeradius)hasaverynarrowbucket,but
canoperateataverylowcavitationnumber(andhencehighvehiclespeed)atitsideal
angleofattack. Conversely,thickfoilshavewidebuckets,butcannotoperateataslow
acavitationnumber.
Forany specied operatingcavitation number,anoptimumfoilthicknessratioexists
which results in cavitation free operation at itsideal angle ofattack, yet also provides for
themaximum possibletoleranceforangle ofattackvariations. This type ofdiagram
wasdevised by Brockettandisshowningure39.
The use of this diagram can be shown bya specicexample. Suppose that we need to
nd the optimumfoilsection to operate at alift coecient ofC
L
= 0.295atacavitation
70
Figure37:Hereisthesametypeofdatapresentation,butforanearlyogivalfoilsection.
Themapping parameters inthis casearex
c
=0.01, y
c
= 0.1, = 20 degrees.Theregion
withintheenvelopecurvehasnownarrowedconsiderablyincomparisontoFigure36.
numberof = 0.5,usinganNACA-66thicknessformandanNACAa=.8meanline.
Therststepistondthecamberratiosuch thatthefoilwillgenerateitsdesignedlift
at its ideal angle ofattack. We know from the table in gure 2 that the camberratio of
anNACAa =.8meanlinewithunitliftcoecientisf
o
/c =0679andtheidealangle
ofattackis 1.54degrees. Scalingthis to alift coecient ofC
L
= 0.295givesacamber
ratiooff
o
/c= 0.02
28
andanangleofattackof 0.45degrees. Enteringtheleft portion
ofthechartingure39atthisvalueofcamberratio,andatavalueof[C
P
]
min
= 0.5
givesavalueofthickness/chordratiooft
o
/c = 0.10. Therighthandchartthenshows
that the width of the bucket is about three degrees. We can thenconrm this result by
enteringgure38atavalueof = 0.45degreesand[C
P
]
min
= 0.5,andweseethat
the operating pointfalls perfectlywithinthebucketcurvefort
o
/c
28
Yes, this is a remarkable coincidence that the camber ratio happened to comes out to match the
chart in gure38.
71
Figure38: SamplefamilyofbucketdiagramsforNACA-66thicknessformswithNACA
a =.8meanlines,allwithacamberratiooff
o
/c = 0.02. Thesectionthickness/chord
ratios(labeledas onthechart)rangefromt
o
/c = 0.02tot
o
/c = 0.20. Thedatais
fromT.Brockett,MinimumPressureEnvelopesforModiedNACA-66Sectionswith
NACAa=.8CamberandBuShips Type Iand TypeIISections,DTMB Report1780,
February1966.
72
Figure39: DesignchartforoptimumNACA-66sectionsfromBrockett. Thelefthand
plotshowsC
P
(min) alongthex axisversust
o
/c alongthey axis. Thelinesonthe
graph arefor constant camberratio, f
o
/c ranging fromzero to 0.06. The right hand
graphshowstheresultingwidth,indegrees,ofthecavitationfreerangewithinthebucket.
73
2 LIFTING SURFACES
74
2.1 IntroductoryConcepts
We saw in the preceding section that a two-dimensional foil could be represented in
lineartheorybyavortexandsourcedistribution,andthattheliftonthesectionwas
dueentirelytotheformer. Wealsosawthatlineartheoryyieldedsurprisinglyaccurate
resultsincomparisontoexacttheory,particularlyforfoilsectionsthatwererelatively
thin.Inthissection,wewillthereforeextendtheconceptofathin,lifting2-Dfoilsection
tothreedimensions. Suchanidealizationistermedaliftingsurface.
Wewillstart by considering asurface of vanishingthickness,butotherwisearbitrary
shape, as illustrated in Figure 40. We will further assume that this lifting surface is
placedinasteady,irrotationaloweld,andthattheuid,asinthecaseof2-Dow,
may beregardedasincompressibleandinviscid.
Figure40: Aliftingsurface.
The uidvelocities atanarbitrary pointP onthe liftingsurface must, ofcourse,
betangenttothesurface. However,the velocitiesatcorresponding pointsontheupper
andlowersidesofthesurfaceneednot beequalineithermagnitudeordirection. The
boundaryconditionsimplyrequiresthatthey becoplanar.
LetusdenotethevelocityontheuppersurfaceasV

u
andthevelocityonthelower
surfaceasV

l
.These can be viewed in a plane tangent to thesurface at point P,as shown
ingure41,andrepresentedasthevectorsumofameanvelocity,V

m
,andadierence
velocity,V

d
. Therefore,
V

u
=V

m
+ V

d
75

Figure41: Velocitydiagraminthetangentplane.
V

l
=V

m
V

d
(89)
where
V

m
=
1
(V

u
+ V

l
)
2
V

d
=
1
(V

u
V

l
) (90)
2
Anon-zerodierencevelocity impliesthepresenceofavortexsheetwhosestrength
atthe pointP is
29
= 2 n V

d
(91)
andwhosedirection isnormal to the planeformed by thesurfacenormal vector,n,and
thedierencevelocityvector,V

d
. Equation 91 may be veriedsimply by calculating the
circulationaroundasmallcontour,asillustratedingure42.
29
The quantity is frequently referredto asthe vorticity,even though it is really the vortexsheet
strength. Vorticity,strictlyspeaking,isthe curlofthe velocityvector,andavortexsheetisthe limit of
a thin layerof uid containingvorticityasthe thickness ofthe layergoes to zeroandthe vorticitygoes
to innity.
76
Figure42: Relating to VelocityDierence.
Wenowdene as the angle between themean ow and the vorticityvector at point
P, andproceedtocalculatethepressurejumpacrosstheliftingsurfaceusingBernoullis
equation,
1
p=p
l
p
u
= (V
2
V
l
2
) (92)
2
u
Asimpleapplicationofthelawofcosinesrelatestheupperandlowervelocitiestothe
meananddierencevelocitiesasfollows,
V
l
2
=V
m
2
+ V
d
2
2V
m
V
d
cos(/2 )
V
2
=V
2
+ V
d
2
2V
m
V
d
cos(/2 +) (93)
u m
whichmay then becombinedwithequations91and92togivetheresult
p=V
m
sin (94)
Thusweseethatthepressurejumpacrossaliftingsurfacevanishesifthevorticity
isalignedwiththemeanow. Ontheotherhand,if =/2equation94reducedto
Kutta-Joukowskis law for the force on a vortex in two dimensional ow. These two
specialcasesareillustratedusingthecurrentnotationingure43.
In themore general case, it is convenient toresolve the vector intocomponentsalong
andnormaltothemeanow. Theformeriscustomarilydenedasthefreevorticity (or
freevortexsheetstrength),
f
, whilethelatter isknownasthe bound vorticity,
b
.Hence
equation94may berestatedas
p=V
m

b
Thedistributionof overaliftingsurfaceofprescribedshapeisuniquelydetermined
by thefollowingfourrequirements:
77
Figure43:Velocityandvortexsheetstrengthforthespecialcasesoftwo-dimensionalow
andfreevortexow.
The boundaryconditionV

n = 0 must besatisedatall pointsonthesurface.

TheKuttacondition,which requiresthatthere benoowaroundthetrailingedge,
must beimposed.
Kelvins theorem must be imposed in determining the relationship between the
boundandfreevorticity.
Thetrailing vortex wake must beforcefree. Therefore it mustconsist only offree
vorticity,which by denitionisalignedwiththelocalmeanow.
Therst tworequirementsareequallyapplicableto twoandthreedimensionalows,
andarethereforefamiliar. Thelasttworequirementsresultinafundamentaldierence
between twoandthreedimensionalow. Inthelattercase,the bound vorticity can vary
overthespanandthisrequiresthatfreevorticitydeveloptoprovideforcontinuityof
vorticityasrequiredbyKelvinstheorem. Thelatterwillexistnotonlyonthelifting
surface,butmustproceedintothewake,asshowninthenextsection. Thisfreevor-
tex wake mustdeform inorder toremainforcefree,sothatitsnal positionmust be
determinedaspartofthesolutiontotheproblem.
2.2 TheStrengthofthe Free VortexSheetinthe Wake
Therelationship betweenthe bound vorticityontheliftingsurfaceandthefree vorticity
inthe wakecan be derivedfromKelvinstheoremwhich statesthatthecirculationaround
aclosedcontourcontained in asimplyconnectedregion of potentialow iszero. This
78

Figure44: GeneralcaseBoundandfreevorticityispresent.
isshowningure45.Theselectedclosedcontourisrstshownoutinthemiddleofthe
uid,whereitobviouslymusthavezerocirculation. Wenowcarefullymovethecontour
over the lifting surface and wake in such a way as to avoid cutting through it. The
circulationaroundthedisplacedcontourmustthereforeremainzero. Wemustnowlook
inmoredetailathowthecontourisplaced.Firstdeneacurvilinears
1
coordinatewhich
beginsat a point s
L
attheleadingedge,andiseverywheretangenttothemeanvelocity
vectorV

m
. Thispassesthetrailingedgeatsome points
T
andproceedsintothewake.
Nextdeneans
2
coordinate with an origin at anarbitrary point in the wake, which lies
inthewakeandisorthogonaltothes
1
coordinate. Lets
W
betheintersectionofthe s
1
ands
2
coordinates. Thus,s
L
ands
T
arefunctionsoftheparameters
W
. Finally,dene
s
E
asthes
2
coordinate of a point justoutside the wake.
Wenowseethattheselectedcirculationcontourstartsjustupstreamofs
L
,passes
overthe hydrofoil to a point just aboves
T
, follows down into the wake keeping just above
thes
1
coordinate,makes aleftturnatthe points
W
,andproceedsoverthewaketothe
point s
E
.Thereturn trip issimilar,exceptthat thecontourremains below the wake and
the hydrofoiluntilitreachesthestarting pointats
L
. Thecontourthereforeconsistsof
twoalmostclosed loops,connected bya pair ofparallelcurves. Thecirculationaround
theforwardloop,fromStokestheorem,is
s
T

1
(s
W
) =
b
(s
1
; s
W
)ds
1
s
L
79

Figure45: Circulation pathusedtodeterminethestrengthofthefreevorticityinthe

wake.
Thecirculationaround the portion of thecontourconsisting of the twoparallelconnecting
paths iszero,since it is everywheretangent to the vorticityvector. Therefore, to keep
thetotalcirculationzero,
s
E

2
(s
W
) =
f
(s
2
)ds
2
=
1
(95)
s
W
where
2
is the circulation around the second loop in the wake. The nal result is
obtained by dierentiatingequation95withrespecttos
W
,
s
E
d
1
d

ds
W
=
ds
W
s
W

f
(s
2
)ds
2
=
f
(s
w
) (96)
Thusweseethatthestrengthofthefreevorticity isrelatedtothespanwisederivative
ofthe boundcirculationaroundthe hydrofoil.However,thedierentialds
2
must betaken
inthewake, notonthehydrofoil. Ifthefreevortexlinesweretomovestraightback
(whichwill beassumedsubsequentlyinlinearized hydrofoiltheory),thenthespanwise
increment in the wake and on the hydrofoil would be the same, and this distinction
would beunimportant. However,inthemoregeneralcaseinwhichdeformationofthe
freevortexwakeisallowed,equation96isanexactresult.
80

2.3 The velocity induced byathree-dimensionalvortexline

Asarststepinthesolutionoftheliftingsurfaceproblemjustformulated,weneedto
beabletocomputethe velocity eldinduced by threedimensionalvortexsheets. Since
thesesheets can be thought of as being composed of elementary vortex lines, we can rst
determinetheirindividualvelocity eld,andthenobtainthe velocity induced byentire
vortexsheet by integration.Theexpressionforthe velocity induced byanarbitrarythree
dimensionalvortexlineisknownasBiot-Savartslaw.
Wewillstart thederivation by considering a ow inwhich the vorticity isconned to
a volumeV withintheuid. WewishtondthevelocityV

at ageneral pointP(x, y, z)
asillustratedingure46
Figure46: Notationforvelocity,V

at pointP(x, y, z) inducedbyavolumedistribution
ofvorticity(, , ) contained involumeV.
AnelementofvorticitywithinV isshownatthe pointQ(, , ),andthedistance
vectorr totheeld pointP is
r =

i(x) +

j(y) +

k(z)
r= (x)
2
+(y)
2
+(z)
2
Anyvectoreldwhosedivergenceiszeroeverywhere,andwhosecurlisnon-zeroin
a portion of the eld can beexpressed as the curl of a vectoreldwhose divergence is
zero.
30
Interpretingtheformerasthevelocity,V

,andthelatterasthe vector potential

A

,
V

=A

= 0 (97)
Thevorticityis = V

. Introducing97andusingaparticularvectoridentity,we
canexpress the vorticity interms of the vector potentialA

= A) = (

A = V (

A) ( )

30
See, forexample,Newman,MarineHydrodynamicsp115.
81

SincethedivergenceofA

iszero,thisreducesto

2

(98) = A
Thesolution ofequation 98,which is a vectorform of Poissonsequation, is,
1 dV
A

= (99)
4 V r
At this point wecan verifydirectly that the divergence ofA

iszero,
1 1 dV ndS
A

= =
r 4 S r 4 V
Thelast integralcan beseen to bezero as aconsequence ofKelvinstheoremwhich
requiresthatvortexlinesnotendintheuid.
We can now roll outthe volume into along, thin vortex tube, which inthe limit
becomesavortexline. Asshowningure47, dl isadierentialelementoflengthalong
the vortex line ands is a unit vector tangent to the vortex. The volume integral in
equation 99then becomes aline integral,
sdl
A

=
4 r
andthe velocitiesmay then beobtained by takingthecurl ofA

.
s s r
V

=A

= dl= dl (100)
r
3
4 4 r
Thelatterformofequation100isknownasBiot-Savartslaw. Thelaststepcan be
veried byworkingouteach component. Forexample,the xcomponent is,

j

k

s
=

(
s
z
)

(
s
y

) + . . . (101) =
x y z
sy
y z r r r
sx sz
r r r
Remembering that r is a function of (x, y, z) buts is not, the x component then
becomes,
s
z

(
1
) s
y

(
1
) = s
z
(y)
+ s
y
(z)
(102)
y r z r r
3
r
3
82

Figure47:Developmentofavortexline.Ontheleftisavolumedistributionofvorticity
. Inthemiddle,thevolume hs been putthrough a pastamachine toform anoodlewith
crosssectionareada. Ontheright,thenoodlehasbeenturnedintoangelshair,with
zerocrosssectionalareaandinnitevorticity,butwiththetotalcirculationkeptxed.
Goingthroughthesameoperationwiththexcomponentofthelastformofequa-
tion100,

k
s r 1
=

i
s
y
(z)
r

3
s
z
(y)
(103) =
r
3
r
3
s
x
s
y
s
z
x y z
weseethatthisisthesameasequation102thusverifyingequation100.
Insummary,the velocity eld of aconcentrated vortexline ofstrength is,
s r
V

dl (104) =
4 r
3
whichcan bewrittenoutincomponentformasfollows,
s
y
(z) s
z
(y)
dl u=
4 r
3
s
z
(x) s
x
(z)
dl v=
4 r
3
s
x
(y)
r

3
s
y
(x)
dl (105) w=
4
Equation 105 isparticularlyuseful inderivingexpressions for the velocity induced by
particularshapesofvortexlines,aswewillseelater.
83

Somecommentsshould bemade about thelimits of the integral inBiot-Savarts law.
FromKelvins theorem, avortex linemust have constantstrength andcannotend in
theinterioroftheuid. Ifwearereallysolvingforthevelocityeldofaconcentrated
vortex, then theintegral must betaken over aclosed path. However, we started out
this section with theobservation thatavortex sheet couldbe considered tobemade
up ofindividual vortexlamentsrepresenting bound and free vorticity. In this case, the
strengthsofthevortexlamentscanvaryalongtheirlength,providedthatthevariation
of bound and free vorticity is always set in sucha way thatKelvinstheorem issatised.
The velocity induced byonecomponent ofthe vortexdistributioncanstill beobtained
fromBiot-Savartslaw,butthestrengthofthevortexinequation104willhavetobe
movedinsidetheintegral,andthecontourofintegrationwillnotnecessarily beclosed.
2.4 Velocity Induced byaStraight VortexSegment
Thevelocity eldofastraightvortexsegmentservesasasimpleillustrationoftheap-
plicationofBiot-Savartslaw. However,itisalsoaveryusefulresultsincethenumerical
solution ofmorecomplicatedgeometriescan beobtained by discretizingthe vortexsheet
intoalattice ofstraightconcentrated vortex elements. The result tobederived here
thereforeservesastheinuencefunctionforsuchacomputationalscheme.
Wecansimplifytheanalysis bychoosing acoordinatesystemsuchthatthe vortex
segment coincides withthe xaxisandtheeld point, P,lies onthe y axis, asshown
ingure 48. Thisisnotreallyarestriction, sinceonecanalwaysmakeacoordinate
transformation to do this, and the resulting velocityvector can then betransformed back
totheoriginalglobalcoordinatesystem.
Thevortexextendsalongthex axisfromx
1
tox
2
. Inthiscase
s = (1,0,0) and r= (, y,0)
andfromequation104wecanseeimmediatelythatu=v =0,sothatweonlyneedto
developtheexpressionforw,
w(y) =
4

x
x
1
2
(
2
+
yd
y
2
)
3/2
=
4y

+ y
2

x
2
=
4y

x
b
2

x
c
1

(106)
x
1
Whereb = x
2
2
+ y
2
andc = x
2
1
+ y
2
. Thisresultcanalso beexpressedintermsof
thetwoangles
1
and
2
whichareillustratedingure48,

w(y) =
4y
[cos
2
+cos
1
]
84
Figure48: Notationforastraightlinevortexsegmentusingalocalcoordinatesystem
withthex axis coincidentwiththevortex,andtheeld point,P locatedonthey axis.
Twolimitingcasesareofparticularinterest. Foraninnitelylongvortex,
1
=
2
= 0
sothat

w(y) =
2y
which is the correct result for a two dimensional vortex. For a semi innite vortex
extendingfromx= 0 toinnity,
1
=/2 and
2
= 0 sothatthevelocity ishalfthatof
aninnite vortex. Thisresult isuseful inliftinglinetheorywhichwewill be looking at
later.
Thecompletevelocity eldisshowningure49foravortexextendingalongthex
axisfrom1 to +1. The variation in velocity with y is shown for severalxed values of
x.Inthisplot,the velocity has beennon-dimensionalized by thefactor /2ysothatthe
resultscan be interpreted as theratio of the velocity induced by the vortexsegment to
85
thatinduced byaninnite vortex of thesamestrength located thesamedistance away.
Therefore, for|x| < 1theresultapproaches a value ofunityas ybecomessmall. For
|x| >1 theresultapproacheszeroforsmall y. Forlarge y distances away fromthe vortex
segment, thenon-dimensional velocity becomes independent of x and decays as 1/y.The
dimensional velocity thereforedecaysas 1/y
2
.
Figure49: Normalized velocity,2yw(x, y)/ induced byastraightvortexsegment.
86
2.5 LinearizedLifting-SurfaceTheoryfor aPlanar Foil
2.5.1 FormulationoftheLinearizedProblem
Wewillnowconsiderthecaseofahydrofoilofzerothickness, whosesurfaceliesvery
nearlyinaplanewhichisalignedwiththeoncomingow. Suchasurface,forexample,
mightbeexactlyatbutinclinedatasmallangleofattackwithrespecttotheow.
Moregenerally, however,thesurfacemay havesomearbitrarydistributionofcamberand
angleofattack. Theonlyrestrictionsarethattheresultingdeviationofthesurfacefrom
thereferenceplane besmallandthattheslopesofthesurface beeverywheresmall.
Figure 50illustratestheparticularnotationforthisproblem. A Cartesiancoordinate
system isorientedwiththe positivex axisinthedirectionofauniformonsetowof
magnitude U. The y axis normal to U, and the (x, y) plane serves as the reference
surface. Thetipsofthe hydrofoilarelocatedaty=s/2,sothatitstotalspaniss.
Figure50: Notationforaplanarhydrofoil.
Theoutline,orplanform,ofthe hydrofoilisdescribed by thecurvesx =x
L
(y) and
x=x
T
(y),denotingrespectivelytheprojectionoftheleadingandtrailingedgesonthe
(xy) plane.
87
A sectionofthesurfaceatspanwise position y isshowningure51. Thenotationis
thesame as for a two-dimensional foil, asillustrated in gure 1, except that the angle of
attack,(y),andtheverticaldisplacementofthenose-tailline,z
0
(y) must beconsidered
aspartofthethree-dimensionalgeometry. Inaddition,theverticalcoordinateisnow
z,ratherthany. Theangleofattack ofthenosetaillinerelativetotheoncomingow
is
(
y) andthedistributionofcamberisdenotedby thefunction f(x, y). Thecamber
ismeasuredinadirectionnormaltothenosetailline,anditsmaximumvalueatany
spanwisestationisf
0
(y). When weaddthicknesslater, itwillsimilarly bedenoted by
thefunctiont(x, y) withamaximumvalue t
0
(y) ateach spanwise location.Thethickness
functionwillthenbeaddedatrightanglestothecamberline,witht/2projectingon
eachside. Finally,thechordlengthc(y) isthelengthfromtheleadingtothetrailing
edgemeasuredalongthenosetailline.
Figure51: Cutthroughfoilsectionatxed spanwiselocation,y.
Iftheangleofattack issmall,wecanassumethatcos 1 sothatthechordlength
isessentiallythesameasitsprojectiononthe (xy) plane,
c(y) x
T
(y) x
L
(y)
Similarly, since the angle of attack has beenassumed to be small, we can write thez
coordinate of a pointon thesurface asfollows,
z
s
(x, y) z
0
(y) (y)x+ f(x, y)
88

wherez
0
(y) istheelevationofthenosetaillineat x= 0.
2.5.2 TheLinearizedBoundaryCondition
Theexact boundarycondition isthatthenormalcomponent ofthetotaluid velocity
vanishatall pointsonthe hydrofoilsurface,
V

n = 0 on z =z
s
(x, y)
If the slopes of thesurface are small, the unitnormal vector can be approximated as
n
z
s
+
z
s
+

k(1)
x y
Theuid velocitycan beexpressedasthesum oftheoncomingowU andadisturbance
velocity eldwithcomponents (u, v, w).Thetotalvelocity istherefore
V

=(U+ u) +v +

kw
anditsdotproductwiththenormalvectoris
z
s
z
s
V

n (U + u)
x
v
y
+ w (107)
Ifthecamberandangleofattackissmall,wecanexpectthatthedisturbanceve-
locitieswill besmallcomparedwiththeoncomingow. Wecanthereforeeliminatethe
productsofsmallquantities inequation107toobtainthenalformofthelinearized
boundarycondition,
w U
z
s
on z = 0 (108)
x
Notethatthe boundarycondition issatisedonthereferenceplanez = 0 ratherthan
ontheactualsurface,whichisthesameapproximationaswasmadeforlinearized2D
theory.
89
Itisimportanttonotethatequation108doesnotinvolvetheslopeofthesurface
in the spanwisedirection, andactually looks just like the boundarycondition for two
dimensionalow. Thisisnottheresultofany assumptionthatthespanwiseslopesare
smallerthanthechordwiseones,butfollowsfromtheassumptionthatthepredominant
velocity isinthechordwisedirection.
2.5.3 Determiningthe Velocity
Thenextstepistodeterminetheverticalcomponentofthedisturbancevelocity induced
by the boundandfree vorticesrepresentingthe hydrofoilandits wake. Ifthedisturbance
velocities are small, we can assume that the mean inow is equal in magnitude and
direction to the oncoming ow. This means, in particular, that the bound vorticity
will beorientedinthey direction,andthatthefree vorticitywill beoriented inthex
direction. The vortexsheetsareprojectedontothe (xy) plane,andthetrailingvortex
wakewillconsistonlyofsemiinnitestraightvorticesextendingdownstreaminthex
directionfromthetrailingedge.
It is important toremark atthis pointthatthisassumptionregarding the wake does
notconformtorealityevenfor hydrofoilswhich havesmallanglesofattack andcamber.
Theplanarfreevortexsheetwhichweareassumingisactuallyunstable,anditsoutboard
edgestend toroll up andformconcentrated tip vortices.Whilethis is amatter ofconcern
inmany applications (asweshallseesubsequentlyinlookingatmarinepropellers)it
isfortunatelynotofgreatimportancetotheproblemathand. Thereasonisthatthe
velocity induced at points on theliftingsurface byanelement offree vorticity in the
distantwake is bothsmall andinsensitive to itsprecise position.
Anexpression forthevelocityinducedata point (x, y,0)ontheprojectionofthe
hydrofoilsurface by the bound vorticity
b
can beobtaineddirectlyfromequation105,
w
(b)
(x, y,0) =
1

s/2

c
x
T
()

b
(, )(x)
dd (109)
4 s/2 x
L
() r
3
Here and aredummyvariablescorrespondingtoxandy.Theintegraloverthespan
isinthedirectionoftheelementaryboundvortexlines, asinequation104,whilethe
integrationinthechordwisedirectionofthecontinuousvortexsheetisequivalenttothe
totalcirculationofaconcentratedvortexline. Sincethedenominatorinequation109
vanisheswhenx= andy=,theintegralissingular. Thesingularity isoftheCauchy
principalvaluetype,whichwesawearlierinthesolutionofthe2Dproblem.
Theexactdistancevector,r,fromanelementofvorticityonthehydrofoilsurface
locatedat (, ,
s
) toanother pointonthesurface (x, y, z
s
) wherewewanttocalculate
thevelocity hasamagnitude,
90

r= (x)
2
+(y)
2
+(z
s

s
)
2
butsincelineartheoryprojectseverythingontothe (xy) plane,thisreducesto,
r (x)
2
+(y)
2
The velocity induced by thefree vorticitycan be developedfromequation 104 in the
sameway,givingtheresult,
w
(f)
(x, y,0) =
1

s/2

f
(, )(y)
dd (110)
4 s/2 x
L
() r
3
Notethatinthiscasethattheupperlimitofthe integrationisnotthetrailingedge,
butextendstoinnity downstream.
2.5.4 RelatingtheBoundand Free Vorticity
Therelationship betweenthe boundandfree vorticitycan beobtainedusingthesame
approachaswasusedinsection2.2. However,thingsarenowsimpler,sincethes
1
and
s
2
coordinates in gure 45are now just thexandycoordinates.If we chose acirculation
pathasshowningure52,wherethe twoconnectedloopsalmosttouch the hydrofoilat
the point (x, y),itisevidentfromStokestheoremthat,
Figure52: Circulationcontourstogetfreevorticityonthefoil.

s/2

b
(, y)d
f
(x, )d= 0 (111)
x
L
(y) y
91
Dierentiatingequation111withrespectto y, keepinginmindthatthelowerlimitof
therst integral is afunction of y,weobtain anexpression for thefree vorticityat any
pointwithinthe hydrofoil,
dx
L

b
(, y)

f
(x, y) =
b
(x
L
, y)
dy

x
L
y
d (112)
Therstterm inequation112can be interpretedasthe bound vorticity whichruns
into theleading edge beingturnedninety degrees to becomefree vorticity. Forexample,
ifahydrofoilhadconstantboundvortexstrengthoveritssurface,thesecondtermin
equation 112 would be zero, and the onlysource of free vorticityon the hydrofoil would
beduetotherstterm.
Ifthe pointx movesdownstreamofthetrailingedge,equation111 becomes,

x
T
(y)

s/2

b
(, y)d
f
(x, )d= 0 (113)
x
L
(y) y
which looksalmostthesame,exceptthattheupperlimitoftherstintegralisnowalso
afunction of y. Dierentiationwith respect to y thereforeresults inoneadditionalterm,

f
(y) =
b
(x
L
, y)
dx
L

x
T
(y)

b
(, y)
d
b
(x
T
, y)
dx
T
=
d
(114)
dy x
L
(y) y dy dy
Thus, the free vorticity in the wake is independent of x and depends only on the
spanwise derivative of the total bound circulation around the hydrofoil. This result
agreeswiththemoregeneralresultgiveninequation96.
The formulation of the linearized planar hydrofoil problem is now complete. The
boundaryconditiongiveninequation108relatesthegeometryandangleofattackof
thehydrofoil tothevertical induced velocity. The latteris related tothe boundand
free vorticity by equations 109 and 110. Finally, the free vorticity is related to the
bound vorticity by equations112and114. Thus,therelationship betweenthe hydrofoil
geometryanditsboundvortexdistributionisestablished. Finally, thedistributionof
pressurejump ( which iscommonlyreferredtoastheloaddistributionacrossthelifting
surface isdirectlyrelated to the bound vorticity fromequation 94.
If we are given the geometry and angle of attack,we can solve for the loaddistribution
using the aboveequations. Since theunknown bound vortexdistribution appearsinside
anintegralinthiscase,wemustsolveanintegralequation. Thisisknownastheanalysis
problem.
92
Ontheotherhand, ifwe aregiven theloaddistribution, thequantities inside the
integrals are known, and we can obtain the slope of the hydrofoil section directly by
integration.Onemore integration isthenrequired to get theactual shape of the hydrofoil
fromtheslopeofitssurface. Thisisknownasthedesign problem.
While this is simple enough in principle, the problem is complicated by the fact
that the integrals in either case are singular, and no closed form solution exists even
forthislinearizedproblem. Onemustthereforeresorttonumericalprocedures,ortoa
combinationofanalyticalandnumericaltechniquestosolvetheproblem. Beforeweget
intothis, wewilldevelopexpressionsforthe overallliftanddragforceson a hydrofoil,
whichcan bedoneanalytically.
2.6 LiftandDrag
Theresultantforce perunit span on asection at spanwise location ycan beresolved
intocomponentsF
z
(y) inthezdirectionand F
x
(y) inthexdirection. Theforceinthez
direction, being at right angles to thedirection of theoncoming ow, as we sawearlier in
looking at two-dimensional ow, istermedlift, whiletheforceinthexdirectionistermed
drag. Inarealuid,thedragforceconsistsofacontributionduetoviscousstresseson
thesurface of the hydrofoil, and acontribution due to thepresence of thetrailing vortex
system. Thelatteristermedinduced dragandistheonlycomponentofdragconsidered
inourpresentinviscid analysis. Inthecaseoftwo- dimensionalow, thereisnofree
vorticity,andhence,noinduceddrag.
Figure53: Controlvolumeformomentumanalysisforlift.
93

Wecanderivetheforceactingon asection by applyingthemomentumtheoremto
acontrolvolumeofinnitesimalspanwiseextentyextendingtoinnity inthex andz
directions,asshowningure53. Theforceinthez directionis
F
z
(y)y= w(V

n)dS (115)
S
wheretheintegralistakenoverallsurfacesofthecontrolvolume.
Thecontributionsofthetop, bottomandfrontsurfacestoequation115can beseen
to be zero, since the velocity atlarge distances decays ata faster rate than the area
increases. Thecontributionofthesidesisalsozeroduetothefactthatw isaneven
function ofzwhileV

n isodd.Thisleavestheuxofmomentumthroughtheaftsurface.
Sincethis is farfrom the hydrofoil, the velocity induced by the bound vorticity goes to
zero. Theonlyinducedvelocity isthatduetothefreevortices,which hasnocomponent
inthex direction. Thus,themomentumuxis

F
z
(y)y=U w(, y, z)dzy (116)

Since weareinnitely far downstream, the velocity induced by thefree vortices ap-
pearsasthatduetoasheetofvorticesofinniteextentinthe x direction,
1

s/2
y
w(, y, z) =
2 s/2

[(y)
2
+ z
2
]
d (117)
Combiningequations116and117,andreversingtheorderofintegrationgivesthe
result
U

s/2

y
F
z
(y) =
2 s/2 [(y)
2
+ z
2
]
dzd (118)
Theintegralinequation118is
y
dz=tan
1
z
(119)
[(y)
2
+ z
2
] y
which issimpleenough,exceptthat we have to becareful in evaluating thelimits. As
z theinversetangent becomes/2dependingonthesignofz andy . The
94

safe way is tobreak up the spanwise integral into two intervals, depending on the sign of
y,
tan
1
z
y
tan
1
z
y

d (120)

s/2
U
y

F
z
(y) = d+
2 s/2 2 y

which leadsustothenalresult
U U
F
z
(y) = [(y) (s/2)] [(s/2) (y)] =U(y) (121)
2 2
Thus,thetotalliftforceonasectionisthesameasthatwhichwouldresultifthe
distribution of boundcirculationoverthe chord wereconcentrated in asingle vortex of
strength (y).
We cannot use the same control volume to determine the drag, since x directed
momentumisconvectedacrossthesidesofthecontrolvolume, andwewouldneedto
knowmoreaboutthedetailsoftheowtocalculateit. However,wecandeterminethe
total induced drag by equating the work done by the drag force when advancing the
hydrofoilaunitdistancetotheincreaseinkineticenergyintheuid. Forthispurpose
wecanmakeuseofGreensformula

E = dS (122)
2 S n
to determine the kinetic energy, E in the uid region bounded by the surface S. In
equation122,isthe velocity potentialandn isaunitnormalvectordirectedoutwards
fromthecontrolvolume,asillustratedingure54.
Thecontributiontotheintegralin122fromallthesurfacesexceptforthosecut by
the free vortex wake is zero as the outer boundaries move toinnity. On the inner surface,
thenormalderivative of the velocity potential isw(, y, 0)ontheupper portion,and
w(, y, 0) on the lower portion.The jump in potential (
u

l
) isequaltothecirculation
(y) aroundthe hydrofoilatspanwise position y.
Thekineticenergyimpartedtotheuidasthefoiladvancesaunitdistanceinthex
directionisthetotalinduceddrag,F
x
(total). This,inturn,can beequatedtothetotal
kineticenergy between twoplanesfar downstreamseparated by unitx distance,
95
Figure54: Controlvolumeforkineticenergyfardownstream.

s/2

s/2
F
x
(total) =
2 s/2

l
w(, y, 0)dy+
2 s/2

u
(w(, y, 0))(dy)

s/2
=
2 s/2
(
l

u
)w(, y, 0)dy

s/2
=
2 s/2
(y)w(, y, 0)dy (123)
Wecanrelatethistothevelocityeldnearthehydrofoilasfollows. Supposethat
thetotal boundcirculation wereconcentratedon asingle vortexlinecoincidentwiththe
y axis,asshowningure55. Thevelocityw(0, y, 0)induced by thefree vortices would
behalfthevalueinducedatinnity, asshownearlier. Deningadownwashvelocity
w

(y) w(0, y, 0)
equation123 becomes

s/2
F
x
(total) =
s/2
(y)w

(y)dy (124)
The total induced drag force is therefore the same as that which would result if
theresultantforceoneachspanwisesectionwerenormaltotheinducedinowvelocity,
96
Figure55: Concentration of boundvorticityalong aliftingline.
V

(y),asshowningure56. HereV

(y) istheresultantof U andw (y),andF(y) is

theresultantoftheliftF
z
(y) andtheinduceddrag F
x
(y).
Asimple interpretationofthisresult followsfromaconsiderationoftheoweld
seen byasmall bugtravelingontheconcentratedvortex atspanwise positiony. The
owappearstobetwo-dimensionaltothebug, butwithaninowrepresented byV

ratherthanU. Hencetheforce,fromKuttaJoukowskis lawisatrightanglestothe
localow,andthereforehascomponentsin boththeliftanddragdirections.
Thereisonedangerinthisinterpretation. Itwouldappearfromequation124that
wehavedeterminedthespanwisedistributionofinduceddrag, yetGreensformulafor
kineticenergyisaglobalresult. Theanswertothisparadoxisthattheintegrandof124
onlyrepresents thelocalforceifthefoilisreallyastraightliftingline, whichallowed
ustorelatethedownwashtothevalueofw atinnity. Thisspanwisedistributionof
induceddragisalsoreasonablycorrectforstraight,highaspectratiofoils.Butifthefoil
isswept,theactualspanwisedistributionofinduceddragcanbecompletelydierent,
yetthetotaldragwill becorrectlypredicted by equation124.
Thus,thespanwisedistributionoflift,andthetotalinduceddragofa hydrofoilis
a function only of its spanwise distribution of circulation. The planform of the foil,
anditschordwisedistributionofcirculationhasnoaectonthesequantities. Thisisa
97
Figure56: Interpretationofliftanddragintermsoflocalowataliftingline.
directconsequenceofthelinearizingassumptionsmade,particularlywithregardtothe
positionofthefreevortexwake. Nevertheless,predictionsofliftanddragmadeonthis
basisaregenerallyincloseagreementwithmeasurementsexceptincaseswhereextreme
localdeformation of thefree vortexsheet occurs.
Lifting line theory is therefore extremely useful preliminarydesigntoolforfoils of
arbitraryplanform,sinceitwilltellushowmuchwehavetopay indragforaprescribed
amountoflift.Inordertodothis,weneedtoknowhowtocalculatew

(y) asafunction
ofthecirculationdistribution(y).Readthenextsectiontondout!
98

2.7 LiftingLineTheory
2.7.1 GlauertsMethod
Wewillnowdevelopanexpressionforthevelocity w

(y) inducedon aliftingline by the

freevortexsheetshedfromanarbitrarydistributionofcirculation(y) overthespan.
WewillfollowtheclassicalmethodofGlauert
31
. Itwill beevidentthattheapproach
is very similar tothemethod used by Glauerttond thevelocity induced by atwo-
dimensionalvortexdistribution. Oncethisisdone,expressionsfortheliftandinduced
dragcan beobtainedfromtheresultsoftheprecedingsection.
Werstdeneanewspanwisevariable y which isrelated to the physical coordinate
y asfollows,
s
y= cos( y) (125)
2
sothat y = 0wheny =s/2and y = wheny =s/2. Thespanwisedistributionof
circulation isassumed to berepresented by thefollowing sineseries in y,

(y) = 2Us a
n
sin(ny) (126)
n=1
which has the property that = 0 at the tips for anyvalues of the coecientsa
n
. The
free vortexstrength is thenobtained by dierentiating 126 with respect toy,

f
(y) =
d
dy

=
d
d

y dy
dy

1
= 2Us na
n
cos(ny)
s
sin( y)
n=1 2
4U

=
sin( y)
na
n
cos(ny) (127)
n=1
The velocity w

(y) can now be expressed as an integral over the span, keeping in

mind thatthe velocity induced bythe semiinnite freevortices issimply halfofthe
twodimensionalvalue,
31
Elements of Aerofoil and Airscrew Theory, Cambridge UniversityPress,1926
99

s/2

f
()
w

(y) =
4

s/
c
2
y
d (128)
Introducingequation127intoequation128anddeningadummy spanwisevariable
asinequation125,
s
= cos( ) (129)
2
wecanobtainthenalresultforw

(y),

4U

na
n
cos(n)
s
sin( )d
w

(y) =
4
1

0
c
sin
n=1

2
s
{cos ycos
2
}

n=1
na
n
cos(n)
= c d
cos ycos
0

na
n
sin(ny)
= U
n=1
sin y
(130)
Thelast step inequation 130 makesuse ofGlauerts integral,whichwe saw before in
thesolutionofthetwo-dimensionalfoilproblem,
I
n
( y) =

cos(n)
d =
sin(ny)
(131)
cos ycos sin y
0
Figure57showstherstfourtermsoftheGlauertseriesforthespanwisedistribution
ofcirculation,whilegure58showstheresultinginducedvelocity. Notethattheleading
termproducesaconstantvelocity of1overthespan,whilethehighertermsproduce
progressivelylargerandmoreoscillatoryvelocitydistributions. Notethatthevelocity
giveninequation130isindeterminateatthetips,wheresin y=0,butthatitcaneasily
beevaluated,givingtheresultthat w

/U =n
2
a
n
.Thefactthatthevelocity atthetips
inducedbyeachtermintheseriesgrowsquadraticallywithn hasimportantpractical
consequenceswhichwewilldiscusslater.
Atthis pointthetotalforce inthe zdirectioncan befound,
100
Figure57: PlotofrstfourtermsofGlauertscirculationseries.
Figure58: Plotofvelocityinduced byrstfourtermsofGlauertscirculationseries.
101

s/2

s/2
F

z
= F
z
(y)dy=U (y)dy
s/2 s/2

U
2
s
2
a
n
sin(ny) sin ydy =
0
n=1
=

2
U
2
s
2
a
1
(132)
which isseentodependonlyontheleadingtermintheassumedseriesforthespanwise
distributionofcirculation. Theremainingtermsservetoredistributetheliftoverthe
span,butdonotaectthetotal.
Thetotalinduceddragforcecannow becomputedfromtheformulaobtainedinthe
previoussection,

s/2
F

x
= w

(y)(y)dy (133)
s/2
whichcan beaccomplished by substitutingequations126and130 into133. Sincethis
involvestheproductoftwoseries,twosummationindicesarerequired. Notingthatall
butthediagonaltermsintheproductofthetwoseriesvanishuponintegration(0, ),
thestepsnecessarytoobtainthenalresultareasfollows,
_

_
_

n=1
U na
n
sin(ny)

2Us a
m
sin(my)
s
F

x
=
= U
2
s
2
sin ydy

_
sin y 2
m=1
0

0
n=1 m=1
na
n
sin(ny) a
m
sin(my)dy

2
n=1
It is instructive to extract the leading coecient in the circulation series, and to
expressitintermsofthetotalliftforceF

z
from132,
=

U
2
s
2 2
(134)
n
na
102

x
=

U
2
s
2
a
2
1 +

n
a
n
2
1
n=2
a
1
2F

z
2

a
n

2
=
U
2
s
2
1 +
n=2
n
a
1
(135)
Thus,weseethatthereciprocaloftheterminbracketsisaformofeciencywhich is
maximizedwhena
n
= 0 for n > 1. Thepresenceofhighertermsinthecirculationseries
does not change the lift, but increases the drag. The optimum spanwise distribution
of circulation is therefore one in which the lift is distributed as sin y, or in physical
coordinates,asanellipse.
For a xed spanwise distribution of lift, equation 135 shows that induced drag is
directlyproportionaltothesquareofthetotallift,inverselyproportionaltothesquare
of the speed, and inversely proportional to thesquare of the span.
This result is frequently presented in terms of lift and drag coecients based on
planformarea. Thisrequiresthe introduction of anon-dimensionalparameterA called
aspect ratio,which istheratioofthespansquaredtothearea,S ofthe hydrofoil,
s
2
A = (136)
S
Deningthetotalliftandinduceddragcoecientsas,
C

L
=
1
F

z
U
2
S
2
C

D
=
F

x
(137)
1
U
2
S
2
thenon-dimensionalformofequation135 becomes,

2
C

D
=
C

L
2

1 +

a
n

(138)
A
n=2
a
1
While equation 138 is more concise, it can lead to the erroneous conclusion that
increasingaspectratioalwaysreducesinduceddrag. Itdoesreduceinduceddragifthe
103

increaseinaspectratioisachieved by increasingthespan. However,ifitisachieved by

keepingthespanxedandreducingthechord,equation135showsthatthedragisthe
same. Theconfusioniscaused by thefactthatiftheareaisreduced,theliftcoecient
must beincreasedinordertoobtainthesamelift. Therefore,inthiscase boththelift
and drag coecientsincrease, but thedimensionalvalue of the dragremains thesame.
2.7.2 VortexLatticeSolutionforthePlanarLiftingLine
WesawintheprevioussectionthatGlauertsanalyticalsolutionforthe twodimensional
foilcould be replicated with highprecision by a discrete vortex lattice method. Itis
therefore reasonable to expect that the same success can be achieved with a vortex
latticeliftinglinemethod. In bothcases,themotivationisobviouslynottosolvethese
particular problems, but to tune up the vortex lattice technique so that it can be
appliedtomorecomplicatedproblemsforwhich thereisnoanalyticalsolution.
As illustrated in gure 59, the span of the lifting line is divided into M panels,
whichmayor may not be equally spaced, and whichmay be inset a givendistance from
each tip
32
. The continuous distribution of circulation over the span is considered to
bereplacedbyasteppeddistributionwhichisconstantwithineachpanel. Thevalue
ofthecirculationineachpanelisequaltothevalueofthecontinuousdistributionat
someselected value ofthe ycoordinatewithineachpanel. Theinduced velocity will be
computedatasetofcontrol points. UsingthenotationoftheFORTRAN90program
which contained in appendix, the coordinate ofthe control pointin the nth panel is
y
c
(n),andthecorrespondingcirculationis(n).
Since the circulation is piecewise constant, the free vortex sheet is replaced by a
set of concentrated vortex lines shed from each panel boundary, with strength equal
tothedierenceinboundvortex strengthacrosstheboundary. This isequivalent to
replacing the continuous vortex distribution with a set of discrete horseshoe vortices,
eachconsisting of a bound vortex segment, and twoconcentrated tip vortices. The y
coordinate ofthepanel boundaries,whicharethenthecoordinates ofthefree vortices,
will bedenotedy
v
(n). IfthereareM panels,therewill beM +1 freevortices.
The velocity eld ofthisdiscreteset ofconcentrated vorticescan becomputed very
easilyat pointsontheliftingline,sincethesingularintegralencounteredinthecontinuous
caseisreplaced by thesummation
M
w

(y
c
(n)) w
n

=
m
w
n,m
(139)
m=1
32
The optimum tip inset is not atall obviousat this point, but will be addressed later
104
Figure59: Notationforavortexlatticeliftingline. Inthiscase,thereare8uniformly
spaced panels,with a1/4 panelinsetateachend.
wherew
n,m
isthe velocity inducedatthecontrol pointy
c
(n) by aunithorseshoevortex
surroundingthe pointy
c
(m). Since the bound vortexsegment of thehorseshoe does not
induceanyvelocityontheliftinglineitself,theinuencefunctionw
n,m
consistsofthe
contributionoftwosemi-innitetrailingvorticesofoppositesign,
1 1
w
n,m
=
4(y
v
(m) y
c
(n))

4(y
v
(m+1) y
c
(n))
(140)
However, it isclear that theresulting velocitywill not beaccurate for all values of
y. In particular, the velocity will become as y
c
is moved past any ofthe vortex
coordinates y
v
. Nevertheless, our intuition says that the result might be accurate at
pointsthataremoreorlessmidway betweenthe vortices.
Thisisillustratedingure60. Thisshowsthedistributionofinducedvelocityw

(y)
foranelliptically loadedliftinglineusing tenequally spacedpanels inset onequarter
panelfromeach tip. The velocity has beencomputed at a large number of pointswithin
eachpanel,andonecanclearlyseethevelocity tendingto neareachofthepanel
boundaries. The velocity can obviously not be calculated exactly on the panel boundaries,
so what is shown in the graph is asequence of straight linesconnecting theclosest points
computedoneach side.
105

Figure60: Spanwisedistributionofvelocityinducedbyavortexlattice. Thespacingis

uniform, with ten panelsand 25%tip inset. Due to symmetry, only halfthe spanis
shown.
Alsoshowninthegureistheexactsolutionfortheinducedvelocity,whichinthis
caseissimplyaconstantvaluew

(y)/U =1. Thenumericalsolutiondoesnotlookat

all like this, but if you look closely, you can see that the numerical values arecorrect at
the mid points ofeach of the intervals. We wouldtherefore get the rightanswer if we
chosethemidpointsofeach intervalasthecontrol points.
Theliftandinduceddragcannow bewrittenassumsoftheelementaryliftanddrag
forcesoneach panel,
M
F

z
=U (n)[y
v
(n+1) y
v
(n)] (141)
n=1
M
F

x
= w

(n)(n)[y
v
(n+1) y
v
(n)] (142)
n=1
Equation 139 can represent the solution to two dierent types of problems. The
rstisthedesignproblem,wherethecirculationdistribution(y),andhencethetotal
lift, is given. We can use equation 139directly toevaluate w

, and we can then use

equation 142 to obtain the induced drag. We will also see later that the downwash
106
velocity is an important ingredient in establishing the spanwise distribution of angle
attackrequired toachieve thedesign circulation. The second istheanalysisproblem
wherewearegiventhespanwisedistributionofdownwash,w

, andwewishtodetermine
thecirculationdistribution. Ifwewritedownequation139forM dierentcontrol points
y
c
(1) . . . y
c
(n) . . . y
c
(M) we obtain a set of simultaneous equations where w
n,m
is the
coecientmatrix,w

is the right hand side, and is the unknown. Once is found, we
canobtain boththeliftandthedragfromequations141and142.
Theremainingquestionishowtodeterminetheoptimumarrangementofvortexand
control points. Whilemuchtheoreticalworkinthisareahas beendone,rightnowwe
willuse acutand tryapproach. This isfacilitated byasimpleFORTRAN90program
calledHVLL,whichcalculates boththeexactandthe numericalvaluesoftheinduced
velocity, totalliftandtotalinduceddragforacirculationdistribution denedbyany
specied number ofGlauert coecients a
j
. The designproblem is rst exercised by
calculating the numericalapproximation to the downwashinduced by the speciedcir-
culationdistribution. Theanalysisoption isthenexercised by calculating the numerical
approximationtothecirculationstartingfromtheexactdownwashassociatedwiththe
originally speciedcirculationdistribution. In bothcases, thetotal lift andinduced drag
canbecomputedandcomparedwiththeexactvalues. Thus, theaccuracyofagiven
latticearrangement andtheconvergenceofthemethod withincreasingnumbersofpanels
can bestudied.
Thesimplestarrangement consistsofequallyspacedpanelswithnotipinset,andwith
thecontrol pointsatthemid pointofeach panel. Thisschemewill bedemonstratedfor
thesimplecaseofellipticalloading,wheretheexactdownwashisaconstant. Theresults
for M = 8 panels is shown in gure 61. Here we see that the predicted circulation
distributionhasthecorrectshape,butisuniformlytoohigh. Thenumericalresultfor
thedownwashisquitegood inthemiddleofthespan,butgetsworseatthetips.
Thefollowingtableshowstheeectofnumberofpanelsonthecomputedforces. For
example, ifthecirculation is specied,theerror inpredictedlift is 1.3%with 8panels,
andreduces to0.1%with 64panels. But, the errorindragis much greater, ranging
from10.1%with 8panelsto 1.3%with64panels. Ontheotherhand,ifthedownwash
is specied, thecomputedlift and drag is inerrorby about thesameamount,ranging
from12.5%with 8 panelsto 1.6%with64panels.
While this might notseem too bad, it is easy to get much betterresultswithout any
extracomputingeort. Theproblemwiththetippanelisthatthestrengthofthefree
vortexsheetinthecontinuouscasehasasquarerootsingularity atthetips,which isnot
approximatedwellinthepresentarrangement.
Figure 62 showswhat happens if the tippanelsareinset byonequarter of a panel
width. Nowtheinduced velocity inthe tippanel is much better ( butstillnotas good
107
as for the rest of the panels ), and theerror in forces is around one percent for 8 panels,
and 0.1 percent orlessfor 64panels. Onecanexploretheresult of changingthe tipinset
andndvalueswhichwilleithermaketheinducedvelocityatthetip,thetotallift,or
theinduced dragcorrect. However, no single value will be best for allthree. Thus, a tip
inset of one quarter panel isconsidered to be the best. A proof that a quarter panel inset
iscorrectintheanalogoussituationofthesquarerootsingularityattheleadingedgeof
a twodimensionalatplate waspublished by James
33
.
Another possiblespacingarrangement ismotivated by the change in variablesused
by Glauertin thesolutionoftheliftingline problem. Wesaw thatthisarrangement
worked very wellforthe2-Dproblem. Inthiscasethe vorticesandcontrol pointsare
spacedequallyintheangularcoordinate y. Thisarrangementiscalledcosinespacing,
andtheequationsfory
v
(n) and y
c
(n) can befoundinthecode. Inthiscasenotipinset
isrequired. Proofsthatthisarrangementiscorrectmay befoundinLan
34
andStark
35
.
Figure63showswhathappenswhencosinespacingisusedwith 8 panels,butwhere
thecontrol pointsare locatedmidway betweenthe vortices,as isthecasewithconstant
spacing. Figure64isanillustrationofhowafastcomputercanmakeupforacertain
amount of humanstupidity.Using 64panels,thepredictedcirculation looksquite good
althoughitisstillalittlehigh. Thedownwashisagainaccurateoveralotofthemid
span,buttheresultsatthetipsareaevenmoreofadisaster. Increasingthenumberof
elements localizes theproblem, but thecomputed values at the tiparestill way o. In
spiteofthis,thetotalforcesseemto beconverging,withanerrorofaround two percent
with64elements.
Fortunatelythisisnottherealcosinespacing,andisincludedasacautionarytale
fornumericalhackers. Inrealcosinespacing, thecontrol pointsaremappedwiththe
samecosinetransformationasthevortices. Theyarethereforenotinthemiddlesofthe
intervals,butarebiasedtowardsthetips. Figure65resultsshowsthatthisarrangement
isextremely accurate, even with8panels. Note, inparticular, thatthe liftanddrag
obtainedfromthecirculationfound by specifyingw

isexact,andthattheratioofdrag
toliftsquared,F
x
/F
2
isexactforanynumberofpanels.
z
Alloftheexamplesconsideredsofarareforellipticalloading. Theremainingtwo
guresshow theresults ofaddinganadditionalcoecient withavalueofa
3
= 0.2
totheGlauertseriesforthecirculation. Thisunloadsthetips(whichmaybedesired
todelay tipvortexcavitationinception),producinglargeupwardinducedvelocitiesin
33
James,R.M.,Onthe RemarkableAccuracyof the Vortex Lattice Method,Comput. Methods Appl.
Mech. Eng. Vol.1,1972
34
Lan,C.E.,A Quasi-Vortex Lattice Method in Thin Wing Theory,J.Aircraft,Vol 11, No.9, Septem-
ber,1974
35
Stark,V.J.E.,A Generalized Quadrature Formula for Cauchy Integrals,AIAA Journal Vol 9, No.
9,1970
108
Figure61: Comparisonofvortexlatticeandexactresultsforanellipticallyloaded lifting
linewitha
1
= 1.0. Thesolutionwasobtained with 8 panels,usinguniform spacingwith
zerotipinset.
the tip region, and increase the induced drag. Figure 66 shows the results obtained
usinggoodcosinespacingwith 8 panels. Theresultsareextremelyclosetotheexact
value,althoughasmalldiscrepancyisvisibleinthegraph. Figure67showsthesame
casecalculatedwith32panels. Theresultsnowappearto berightontopoftheexact
results. Inaddition, theincreasednumber of panels provides much betterresolution of
the behaviorofthecirculationanddownwashnearthetips.
ConstantSpacing- ZeroTipInset
PercentErrorsin VortexLatticePredictions
ForF
z
,F
x
andF
x
/(F
z
)
2
Panels Given(y) Givenw

(y)
8 1.3
16 0.5
32 0.2
64 0.1
-7.6
-4.2
-2.3
-1.2
-10.1
-5.1
-2.6
-1.3
12.5
6.3
3.1
1.6
12.5
6.3
3.1
1.6
-11.1
-5.9
-3.0
-1.5
109
Figure62: Comparisonofvortexlatticeandexactresultsforanellipticallyloaded lifting
linewitha
1
= 1.0. Thesolutionwasobtained with 8 panels,usinguniform spacingwith
25 %tipinset.
Figure63: Comparisonofvortexlatticeandexactresultsforanellipticallyloaded lifting
linewitha
1
= 1.0. Thesolutionwasobtainedwith 8 panels,using cosine spacingwith
central control points.
110
Figure64: Comparisonofvortexlatticeandexactresultsforanellipticallyloaded lifting
linewitha
1
= 1.0. Thesolutionwasobtainedwith 64 panels,using cosine spacingwith
central control points.
Figure65: Comparisonofvortexlatticeandexactresultsforanellipticallyloaded lifting
linewitha
1
= 1.0. Thesolutionwasobtainedwith 8 panels,using cosine spacingwith
cosine control points.
111
ConstantSpacing- 25 % TipInset
PercentErrorsin VortexLatticePredictions
ForF
z
,F
x
andF
x
/(F
z
)
2
Panels Given(y) Givenw

(y)
8 -1.0
16 -0.4
32 -0.1
64 0.0
-1.6
-0.6
-0.2
-0.1
0.4
0.1
0.0
0.0
-0.3
-0.1
0.0
0.0
-0.3 0.3
-0.1 0.1
0.0 0.0
0.0 0.0
CosineSpacing- CentralControl Points
PercentErrorsin VortexLatticePredictions
ForF
z
,F
x
andF
x
/(F
z
)
2
Panels Given(y) Givenw

(y)
8 1.1
16 0.3
32 0.1
64 0.0
-12.1
-6.9
-3.6
-1.9
-14.0
-7.4
-3.8
-1.9
15.2
7.7
3.9
1.9
15.2 -13.2
7.7 -7.1
3.9 -3.7
1.9 -1.9
CosineSpacing- CosineControl Points
PercentErrorsin VortexLatticePredictions
ForF
z
,F
x
andF
x
/(F
z
)
2
Panels Given(y) Givenw

(y)
8 -0.6
16 -0.2
32 0.0
64 0.0
-1.3
-0.3
-0.1
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
112
Figure66: Comparisonofvortexlatticeandexactresultsforatip-unloadedliftingline
witha
1
= 1.0 anda
3
= 0.2. Thesolutionwasobtainedwith 8 panels,using cosine
spacingwith cosine control points.
113
Figure67: Comparisonofvortexlatticeandexactresultsforatip-unloadedliftingline
witha
1
= 1.0anda
3
=0.2. Thesolutionwasobtainedwith32 panels,usingcosine
spacingwith cosine control points.
114
2.7.3 ThePrandtlLiftingLineEquation
Our discussion of lifting line theory so far has addressed the question of relating the
spanwisedistributionofcirculationtothedownwash,liftandinduceddrag.Inaddition,
wehavefoundthespanwisedistributionofcirculationwhich minimizestheinduceddrag.
Lifting line theory, by itself, does not provide anyway ofdetermining the lift gener-
ated byaparticular foil shape,since thedetails of the ow over theactualsurfaceare
completelylostintheidealizationofaliftingline. Whiletheliftingsurfaceequations
developedearlierwillprovidethemeanstosolvethisproblem,asimpleralternativeexists
iftheaspectratioofthefoil(theratioofthespantothemeanchord)ishigh.
This idea was originated by Prandtl, who reasoned thatif the aspect ratiois suf-
cientlyhigh,thefoilsectionatagivenspanwise positionactsasthoughitwereina
two-dimensionalow(rememberthenear-sightedbug),butwiththeinowvelocity al-
teredbythedownwash velocityobtainedfromliftinglinetheory. Thesolutiontothe
problemofanalyzingtheowaroundagivenfoilthenrequiresthesolutionoftwocou-
pledproblems; a local two-dimensional problem at each spanwise position, and a global
three-dimensionalliftinglineproblem. Thisidea wasformalizedmany yearslater by the
theoryofmatchedasymptoticexpansionswherethesolutiontothewingproblemcould
befoundintermsofanexpansionininversepowersoftheaspectratio. Thematched
asymptoticsolutionmaybefoundinVanDyke
36
,butwewillonlypresentPrandtls
originalmethod here.
Firstrecallthatthesectionalliftcoecient,C
L
is,
C
L
(y) =
F
z
(y)
=
2(y)
(143)
1
U
2
c(y) Uc(y)
2
where c(y) is the local chord as illustrated in gure 50. To keep things as simple as
possibleforthemoment,letusassume thatthefoilsectionshavenocamber. Thenif
theow were two-dimensional,theliftcoecientatspanwise positiony would be
C
L
(y) =
2(y)
= 2(y) (144)
Uc(y)
FollowingPrandtls theory, equation 144 can be modied toaccount forthree-dimensional
eects by reducingtheangleofattack by theinducedangle,
36
Van Dyke,M.,Perturbation Methods in Fluid Mechanics, ParabolicPress,1975
115

C
L
(y) =
2(y)
= 2[(y)
i
(y)] = 2[(y) + w

(y)/U] (145)
Uc(y)
Boththecirculation andthedownwashw

can beexpressedintermsofthecoe-
cientsinGlauertsexpansion,fromequations126and130. Equation145then becomes

4s

na
n
sin(ny)
C
L
(y) =
c( y)
a
n
sin(ny) = 2 ( y)
sin y
(146)
n=1 n=1
Equation146 mustholdforany spanwise position yalongthefoil.Given a distribution
ofchordlengthc( y) andangleofattack (y),wecanndtherstM coecientsinthe
Glauertexpansionforthecirculation by satisfyingequation146atM spanwise positions.
Thesolutionwillpresumably becomemoreaccurateasM isincreased.
Anotheralternativeistogoback toequation145anduseavortexlatticemethod to
solvefordiscretevaluesofthecirculation. Sinceequation139givesusthedownwash,
w

atthen

th panelasasummationovertheM panels,weobtainthefollowingsetof
simultaneousequationsfortheM unknown vortexstrengths
n
,
M
2
n

Uc
n
= 2
n
+
m=1

m
w
n,m
n= 1, . . . M (147)
Bothmethodsworkwell,butareapproximations,sincetheresultsdependeitheron
thenumberoftermsretainedintheGlauertseries,oronthenumberofpanelsusedin
thevortexlattice.
Anexactsolution toequation 146can beobtained by inspection in the specialcase
thatthechorddistributionisellipticalandtheangleofattackisconstant. Ifwedene
c
0
asthechordlengthatthemid-span,wecanwritethechordlengthdistributionas

2y

2
c(y) = c
0
1 (148)
s
which hasaprojectedareaS =c
0
s/4 andanaspectratio A =s
2
/S = 4s/(c
0
).Before
introducingthis chordlengthdistribution in146 we musttransform it intothe yvariable
usingequation36,
c(y) = c
0
sin y (149)
116

Equation146then becomes

sin(ny)

na
n
sin(ny)
C
L
(y) = A a
n
sin y
= 2
sin y
(150)
n=1 n=1
butthisequalitycanonlyholdifthecirculationdistributioniselliptical,i.e.ifa
n
= 0 for
n > 1. Inthiscase, the locallift coecient,C
L
(y) andthetotalliftcoecient C

L
=Aa
1
areequal,andequation150reducesto
C
L
(y) = C

L
=
2
(151)
(1 +2/A)
Thisremarkablysimpleformulacapturestheessentialroleofaspectratiocontrolling
therateofchangeofliftwithangleofattack. Astheaspectratioapproachesinnity,
thelift slopeapproaches the two-dimensionalvalue of 2. Astheaspectratio becomes
small,theliftslopeapproacheszero. Thisresultisplottedingure68,togetherwith
accuratenumericalresultsobtainedfromliftingsurfacetheoryandwithresultsobtained
fromthetheoryofmatchedasymptoticexpansions. Thisgurecomesfrom VanDyke,
andalso appears in chapter 5 ofNewmansMarineHydrodynamics.
AnamazingattributeofPrandtlstheoryasappliedtoanellipticalwingishowwellit
worksevenforlowaspectratios. Ofcourse,if youlookcloselyatthegure, youcansee
thatPrandtlstheoryalways over-predictsthelift,andthatthe percenterrorincreases
withdecreasingaspectratio. Anotherimportantobservationisthat,evenatanaspect
ratioofA = 8, the lift slope issubstantially below the two-dimensionalvalue of 2.
Thethreecurveslabeled2ndapprox,3rdapproxandmodied3rdapproxarease-
quenceofsolutionsobtainedfromthetheoryofmatchedasymptoticexpansions. The
rstoftheselooksalmostlikePrandtlsresult,namely
C
L
(y) = C

L
= (2) (1 2/A) (152)
which is a little more accurate for high aspect ratios, but falls apart for low aspect
ratiosNotethatitpredictsthatafoilwithanaspectratiooftwowillhavezerolift
atallanglesofattack!. Thehigherordermatchedasymptoticapproximationsremain
accurateforprogressivelylowervaluesofaspectratio.
Figure69showstheapplicationofPrandtlsequationtodeterminetheeectofplan-
formtaperoncirculationdistribution.Asexpected,thecirculationnearthetipsdecreases
117
(andthecirculationattherootincreases)astheratiooftipchordtorootchord,c
t
/c
r
,
isdecreased.
BeforeleavingourdiscussionofPrandtlsliftinglineequation,letusconsiderwhat
happensifthefoilsectionshavecamber. Sincethecoupling betweenthree-dimensional
lifting-line theory and a local two-dimensional ow is based on the total lift at each
spanwise section, it doesnt matter whether the lift is generated by angle of attack,
camber, orsomecombination of the two. Wecanthereforegeneralizeequation 144 by
includingthetwo-dimensionalangleofzeroliftofthelocalsection,
0L
(y)
2(y)
C
L
(y) =
Uc(y)
= 2[(y)
0L
(y)] (153)
For asectionwith positivecamber, the angle ofzerolift isgenerallynegative, thus
increasing the lift in accordance with 153. All we have to do is replace (y) with
(y)
0L
(y) inequations145and146totreatthegeneralcaseofcamberedsections. It
isalsoeasytoincluderealuideects, by replacingthe two-dimensionalliftslopeof 2
andthetheoreticalangleofzeroliftwithexperimentallydeterminedvalues. Inthisway,
theresults of two-dimensional experimentscan beapplied tothree-dimensional ows
118
Slender-wing
theory
Prandtl's 2
nd
approx. (9.10)
2
nd
approx. (9.1b)
Modified 3
rd
approx. (N.30)
Germain
(N.31)
3
rd
approx, (N.29)
A =
6
4
2
0
A
0 2 4 6 8
Lifting-surface theory
Medan
Jordan (1971)
Lift slope, dC
L
/d, of an elliptic wing as a function of aspect ratio.
d
dC
L
Figure by MIT OCW. Adapted from van Dyke, 1975.
Figure69:Eectofplanformshapeonspanwisedistributionofcirculationobtained from
Prandtlsliftinglineequation.Thefoilsallhaveanaspectratioof = 4,andareatunit
angleofattack. Equation146wasusedwithM = 32,whichismorethanenoughfora
converged solution.
119
providedthattheaspectratioishigh.
Forthe specialcase ofanellipticalfoil,thespanwisedistribution ofcirculationwill be
elliptical if(y)
0L
(y) isconstant overthespan.Thiscan beachieved,forexample, by
having boththeangle ofattack andthezeroliftangleconstantoverthespan,or bysome
combinationofthetwowhosedierenceisconstant. Intheformercase,thespanwise
distributionofliftwillremainellipticaliftheangleofattack oftheentirefoilischanged
byaconstantamount(say duetosomedierentoperatingcondition). However,inthe
lattercase,aconstantincrementinangleofattack willintroduceaspanwisevariationin
thequantity(y)
0L
(y).Inthatcase,ellipticalloadingwillonly begeneratedatone
particularangleofattack.
120
2.8 LiftingSurfaceResults
2.8.1 ExactResults
Thesolutionofthelinearizedproblemofaplanarfoilinvolvesthesolutionofasingular
integralequationwhosemainingredientsaregiveninequations109,110and112. We
would expect that an analytical solution could be found in the simple case of a rec-
tangularplanformandwithzero camber, yetthis isunfortunately notthe case. E.O
Tuck
37
developedhighlyaccuratenumericalsolutionsforthiscasebyacombinedan-
alytical/numerical approach which involved an extrapolationof the error obtained by
dierent levels ofdiscretization. Inparticular, Tuck found that the lift slope of a square
(aspectratioA = 1.0)foilis
C
L
= 1.460227 (154)

with acondence of about 7 gures. Obviously this degree ofaccuracy is of no practical

value,butitisimportanttohaveexactsolutionsforspeciccasestotesttheaccuracy
of numericalmethods. Forexample, if youareexamining theconvergence of a numerical
methodas afunction of paneldensity, you might bemisled if theexact valuethat you
areaimingforisevenslightlyo.
A large number of investigators have published values for the lift slope of a at,
circular wing (a ying manhole cover) over the time period from around 1938-1974.
Their valuesrangefrom 1.7596to 1.8144,withseveralagreeingon a value of 1.790.None
oftheseareclosedformanalyticsolutions,andsomeofthedierencescan beattributed
toinsucientnumbersoftermsusedinseriesexpansions. Butin1986,Hauptmanand
Miloh
38
obtainedanexactsolutionbasedonaseriesexpansionofellipsoidalharmonics.
Inparticular,theywereabletoderivethefollowingsimpleequationfortheliftslopeof
acircularwing,
C
L
32

=
8 +
2
= 1.790750 (155)
and alsoobtained a somewhatmorecomplicatedequation for the lift slope of any elliptical
planform.
37
E.O.Tuck,SomeAccurateSolutionsoftheLiftingSurfaceIntegralEquation,J.AustralianMath-
ematicalSociety(Ser. B), 1991
38
A. Hauptman and T. Hiloh, On The Exact Solution of the Linearized Lifting Surface Problem of
an Elliptic Wing,JournalofMechanics and Applied Mathematics, Vol. 39, Pt.1, 1996
121
Figure70: Notationforavortexlatticesolutionforarectangularfoil.
To my knowledge, no other exact solutions exist. However, these two results are
extremelyvaluableinvalidatingthevortexlatticemethod,thatwewillexploreinthe
nextsection.
2.8.2 VortexLatticeSolutionoftheLinearizedPlanar Foil
We would obviously not have gone to all the trouble of developing the vortex lattice
solution for the two-dimensional foil and for the planar lifting line if wehadnt anticipated
puttingthese twotogethertosolvetheliftingsurfaceproblem. Thiscan bedone very
simplyinthecaseofarectangularfoil,asshowningure70.
As in two-dimensional ow, the chord of the foil is divided into N cosine spaced
panels, with concentrated vortex lines and control points spaced in accordance with
equation70.Similarly,thespanofthefoilisdividedintoM cosinespacedpanelsaswas
doneinthedevelopmentofvortexlatticeliftinglinetheory. Eachconcentratedbound
122

vortexelementofstrength
nm
iscombinedwithapairoffreevorticesofstrength
extendingdownstreamtoinnity,thusformingahorseshoevortexelement. Ifweleti, j
denotethe chordwiseandspanwiseindices of acontrol point,thenthe vertical velocity
at aparticularcontrol pointwill be,

N M
w
ij
=
nm
w
nm,ij
(156)
n=1m=1
wherew
nm,ij
isthehorseshoeinuencefunction, which isdened as he vertical velocity at
the (i, j)
th
control pointinduced byahorseshoe vortexelement ofunitstrengthcentered
atthe (n, m)
th
grid point. Eachhorseshoe consists ofthreestraightvortex segments,
whose velocity at any control point position can beobtained from equation 106 developed
earlier.
Fromthelinearized boundarycondition,
w z
s
=
U x
wecanformasetofN M simultaneous equationsfortheN M unknown vortex
strengths
nm
. Oncethecirculationstrengthsareknown,wecanobtainthespanwise
distributionandtotalvalueofliftusingthesameequationsasinvortexlatticelifting
linetheory.
Figure71showsthevortexlatticearrangementforafoilwithaspectratioA = 2.0
usingarelativelynegrid, with32spanwise and16chordwise panels. Notethatthe
panelsneartheleadingandtrailingedgesatthetipareextremelysmall.
ThefollowingtableshowsthecomputedliftslopeforafoilwithaspectratioA = 1.0
using a systematically rened grid ranging from N = 1,128 panels in the chordwise
directionandM = 4,128panelsinthespanwisedirection. Thenestgridinthestudy
requiredthesolution of aset of 8,192simultaneousequations!
39
Notethatthisresult
agreestosevensignicantgureswith Tucksresult. Butnotealso,that a verycoarse
4 8 grid produces aresultwhich is inerror by only 0.07 percent.
39
Advantage was taken of port-starboard symmetry. Otherwise there would have been 16, 384equa-
tions.
123
Figure 71: Vortexlatticegridforarectangularfoilwithaspectratio A = 2. Inthis
example,there are 32 spanwise and 16chordwise panels. The plot on the upper right is
anenlargementofthestarboard tipnearthetrailingedge.
M/N 1 4 8 16 32 64 128
4 1.428988 1.458702 1.459285 1.459351 1.459357 1.459357 1.459357
8 1.428994 1.459262 1.460010 1.460097 1.460105 1.460106 1.460106
16 1.428994 1.459306 1.460085 1.460196 1.460209 1.460210 1.460210
32 1.428994 1.459309 1.460091 1.460206 1.460222 1.460224 1.460224
1.428994 1.459309 1.460091 1.460207 1.460224 1.460226 1.460226
1.428994 1.459309 1.460091 1.460207 1.460224 1.460226 1.460227
Figure72: Convergenceofvortexlatticecalculationforrectangularfoilwithaspectratio
1.0.Tabulated valuesofdC
L
/d.Eachrowshowsconvergencewithnumberofchordwise
vortices.Each columnshows convergencewithnumber of spanwise panels.
124
64
128
Figure73: Vortexlatticegridforacircularfoilwithan8 8grid.
Thenextstepincomplexityistodevelopavortexlatticeschemeforfoilswithar-
bitrary planform, including ones with curved leading andtrailingedges. The natural
modelproblemforthiscaseisthecircularfoil,sinceweknowtheexactsolution.
Early vortexlatticeschemesmaintained aCartesiangrid,with bound vortexelements
orientedinthey direction,andtrailingvortexelementsorientedinthexdirection. This
meantthatasthe chordlength changedwithspanwise position,the vortexlatticegrid
had to have abrupt steps between each chordwise panel. A better arrangement is to
adaptthegridtotheplanform,asshowningure73. Thevortexlinesthatfollowthe
generalshapeoftheleadingandtrailingedgesarenolongernecessarilyintheydirection,
andarethereforenotnecessarilyboundvortices.
Wewillthereforecallthemspanwisevortexlines,andwillhaveto becarefulwhen
computingforces. FromStokestheorem,thetotalcirculationaroundthefoilatapar-
ticularspanwiselocationy
c
isthesumofthecirculationsofthespanwiseelements,no
matterhowtheyareinclined. Supposewelet betheinclinationofaparticularspan-
wisevortexwithrespecttothey direction, theforce perunitlengthonthevortexis
125
Ucos(). But,thelengthofthespanwisevortexisy/cos(),wherey isthewidth
ofthepanelinthespanwisedirection. So,theforce perunitspanisstillequaltoU,
regardlessoftheinclinationofthevortex,. Inaddition,Kelvinstheoremissatisedif
thevortexsystemisbuiltfromhorseshoeelements,originatingfromaspanwisevortex
element.
Theonlytrickindealingwithacircularplanformistoprovideforanitechordat
thetipinordertopreventhavingallthespanwisevorticescomingtogetheratasingle
point. Thiscanbedonerationallyby solvingforthetipchordsuchthattheareaof
theapproximatequadrilateraltippanelisequaltotheareaoftheactualcirculararc
segment. Thisisshowningure73. Figure74showsamuchnergrid,with64panels
overthespanand32panelsoverthechord. Figure75showsanenlargementofthetip
region,wherethenite,equalareatipchordisbarelyvisible.
Thecomputedliftcoecientsforthese twocasesareC
L
= 1.782forthe 8 8grid
andC
L
= 1.790forthe6432grid.Recallthattheexactsolutionforaatcircularfoil
isC
L
= 1.791,soitisclearthatthevortexlatticemethodworkswellfornon-rectangular
planforms.
126
Figure74:Vortexlattice grid for a circular foil with 64 spanwise and 32 chordwise panels.
127
Figure75:Enlargementofthetipregionofthevortexlatticegridforacircularfoilwith
64 spanwiseand 32chordwise panels.
128
Figure76: Vortexlatticegridforaswept,taperedfoil.Therootchordisc
r
/s= 0.5and
thetipchord isc
t
= 0.2.The leading edge isswept back45degrees.Thegridconsistsof
16spanwiseand8chordwise panels.One particularhorseshoeelementishighlighted.
Estensionofthevortexlatticemethodtosweptandtaperedplanformsisrelatively
simple. Figure76showsthevortexlatticegridforataperedfoilwhoseleadingedgeis
swept back 45degrees. Oneparticularhorseshoevortex element ishighlighted inthe
gure. Theinuencefunctionforthiselementagainconsistsofthecontributionofthree
straightvortexsegments.
Figure79showstheeectofsweeponthespanwisedistributionofcirculation,(y)
foraseriesofthreefoilswithzerocamberandconstantangleofattack. Thefoilshave
a constant chord length of c = 0.2s, and therefore have an aspect ratio of A = 5.0.
Thefoilwithzerosweephasan(almost)ellipticalcirculationdistribution. Theswept
backfoil,asshowningure77,hassubstantiallyincreasedcirculationatthetip,and
decreasedcirculationattheroot. Ontheotherhand, thefoilwithforwardsweep, as
showningure78,hasreducedcirculationatthetipsandincreasedcirculationatthe
root.Thismeans that if one wanted to have a swept back foil with anelliticaldistribution
ofcirculation over the span, the angle of attack of the tip sections would have to be
129
Figure77: Vortexlatticegridforaswept,un-taperedfoil. Therootchordisc
r
/s= 0.2
andthetipchord isc
t
= 0.2.The leading edge isswept back45degrees.Thegridconsists
of16spanwiseand8 chordwisae panels.
Figure78: Vortexlatticegridforaswept,un-taperedfoil. Therootchordisc
r
/s= 0.2
andthetipchordisc
t
= 0.2. Theleadingedgeissweptforward45degrees. Thegrid
consistsof16spanwiseand8chordwisae panels.
130
Figure79: Theeectofsweeponthespanwisecirculationdistribution.
131
reduced,comparedwiththoseattheroot. Theconversewouldbetrueforafoilwith
forwardsweep.
132
3 PROPELLERS
133

3.1 Inow
Propellers generally operate in the very complex ow eld that exists at the stern of
aship. This owmay behighly turbulent andspatiallynon-uniform. Toaddtothe
diculty,thisowmay bealteredsignicantly by thepresenceofthepropeller. While
the velocity distribution in the plane of the propeller ispartially due to the potential ow
aroundtheship,itsorigin islargelyviscous. Unlike apurely potentialow,the vorticity
inthisoweldinteractswithitselfasitisaccelerated bytheactionofthepropeller.
Thismeans that thetotal velocityat a pointnear the propeller is not simply thelinear
superposition ofthe inow (inthe absence ofthe propeller) and the velocity induced
by the propeller, but includes an additional interactive component. A full numerical
simulationofthecombinedowproblemrequiresmassivecomputationalresources,and
the validity of theoutcome islimited by presentempirical modeling ofturbulence. It is
thereforeapracticalnecessitytoemployasimplerowmodelformostpropellerdesign
andanalysisapplications.
Ratherthanincludingthiscomplexinteractioninthesolutionofthepropellerow
problem, it istraditionallyassumedthat the speciedinow is aneectiveinowwhich
isdened in a coordinate systemxed on the ship asthetotaltime-averagedvelocityin
the presence ofthe propellerminusthetime-average potentialowvelocityeldinduced
bythepropelleritself. Thenominalinowisdenedastheowthat would bepresence
intheabsenceofthepropeller. Ifthereisnovorticity intheinoweld,thisdenition
reduces to theusualresultthat thetotal velocity is thelinear superposition of theinow
in theabsence of the propeller, and the velocity induced by the propeller. Thus, in this
case, the nominalandeective inows arethesame. Wewillcome back latertothe
questionofhowoneactuallydeterminesthiselusiveeectiveinow. Fornow,wewill
focusonthepropelleritself,andassumethattheinowisgiven.
Inordertorepresentagiveninoweld,werstdeneaship-xedcylindricalcoor-
dinatesystem,withthex axiscoincidentwiththeaxisofrotationofthepropeller. The
origin ofthecoordinate is intheplaneofthepropeller, which servesasthereference point
forallaxialdimensionsofthepropellerbladesurfaces. Theradialcoordinateisdenoted
byr,andtheangularcoordinate by
o
,which ismeasuredinaclockwise(right-handed)
sense when looking downstream with
o
= 0 being at 12oclock. Since the inow is
periodic in
o
, thethreecomponentsofthetime-averagedvelocity intheship-xedcylin-
dricalcoordinatesystem ,V
a
, V
r
, V
t
can beexpressedas a Fourierseries. Theharmonic
coecientsforeach component,A
n
, B
n
arefunctions of position in themeridional plane,
(x, r),

V
a
(x, r,
o
) = A
a
0
(x, r) + A
a
(x, r) cos(n
o
) + B
a
(x, r) sin(n
o
)
n n
n=1 n=1
134



V
r
(x, r,
o
) = A
r
0
(x, r) + A
r
(x, r) cos(n
o
) + B
r
(x, r) sin(n
o
)
n n
n=1 n=1

V
t
(x, r,
o
) = A
t
0
(x, r) + A
a
(x, r) cos(n
o
) + B
t
(x, r) sin(n
o
) (157)
n n
n=1 n=1
Figure80: Typicalnominalaxialwakeeldforasingle-screwcontainership
Figure80shows atypicalnominalinoweldasmeasuredinatowingtank. Iso-
velocity contours are shown on the left, while a circumferential cut through the ow
eld at one radius is shown on the right, together with a bar graph showing the A
a
n
coecients. Sincetheoweldissymmetricalabouttheshipscenterplaneinthiscase,
theB
n
coecientsarezero. Notethattheinowharmoniccoecientsdecreaserapidly
withincreasingharmonicnumber. Whileitwouldseemfromgure80thatonlytherst
fewharmonicsareofimportance,wewillseelaterthattheapparentlyinvisiblehigher
harmonicsareactuallyresponsibleforproducingunsteadypropellershaftand bearing
forces.
A similardiagramcould bemade for aseries ofaxial positions over the extent of
thepropeller. However,inmostcaseswewouldndthatthevariationininowvelocity
would beslight. Itisthereforecustomarytoassumethattheinoweldisindependent
ofx,andthattheinowstreamtubesarethereforecylindrical. To beconsistentwith
thisassumption,conservationofmassthenrequiresthatthecircumferentialmeanradial
inowvelocitybeconsideredtobezero. However, tangentialinowvelocitiesmaybe
present.
135
Wenextintroduceanothercylindricalcoordinatesystemwhich rotateswiththepro-
peller. Thex andr coordinatesarethesameas before,but, representstheangular
coordinate of anarbitrary pointrelative to theangular coordinate of areference point
onthekeybladeoftherotatingpropeller. Ifapropellerisrotatingwithangularvelocity
inacounter-clockwisedirectionwhenlookingdownstream
40
, therelationship between
thexedandrotatingcoordinatesystemis,

o
= t (158)
Iftheinowisnon-uniformwithrespectto
o
intheship-xedcoordinatesystem,
anobserverrotatingwiththepropellerwillseeatime-varyingvelocity. Undernormal
operatingconditions,theresponseofapropellerbladetoeach harmonicoftheinowis
essentiallylinear. Thus,eachinowharmonicgivesrisetoanunsteadybladeforceata
frequencyrelatedtoitsharmonicnumber. Thesteady,ortime-averageforce,istherefore
determinedalmostentirely by thezerothharmonic. Oneshould becautionedthatthis
assumptionmaynot be valid ifmassiveamounts ofbladeowseparationorcavitation
ispresent.
Todeterminesteadypropellerforces,wehavethereforesimpliedtheproblemcon-
siderably. Wenowsimplyhaveagivenradialdistributionofaxialandtangentialeective
inow velocity,
V
A
(r) = A
a
0
(0, r) V
T
(r) = A
t
0
(0, r) (159)
Thesevelocitiesaregenerallynon-dimensionalizedwithrespecttotheshipspeed,V
S
.
3.2 Notation
Figure81showsaright-handedpropellerplacedintherotatingcoordinatesystemde-
scribedintheprecedingsection,togetherwiththeinowV
A
(r), V
T
(r).
ThepropellerhasamaximumradiusR(ordiameterD) andislocatedinthevicinity
oftheoriginofthecoordinatesystem
41
.
40
This is commonly referred to as a right-handed propeller, which rotates clockwise when looking
upstream. This is generally the assumed direction of rotation when developing equations for propeller
ow.
41
Variations in placement of the propeller in the x direction will simply add a constant to the rake,
while variations in the angular placement of the y axis will add a constant to the skew, as shown in
Section ?.
136
Figure81: Propellercoordinatesystemandvelocitynotation.
ThepropellerhasZ identicalbladeswhicharesymmetricallyplacedon a hub,which
inturn, isattachedtoashaft. Thehubandshaftcanbethoughtofasanarbitrary
axisymmetricbodyalignedwiththex axis,butitisfrequentlyidealizedasacylinder
of radius r
H
, as shown in gure 81 or ignored entirely in preliminary hydrodynamic
analyzes.
Thepropellerinducesavelocity eldwithCartesiancomponents (u, v, w) orcylindri-
calcomponents (u
a
, u
r
, u
t
) intheaxial,radialandcircumferentialdirections,respectively.
Thetotal velocity eld isthen the superposition of the propeller advance speed,rotational
speedandinducedvelocity eldandhascomponents
V
A
(r) + u
a
, u
r
, r + V
T
(r) + u
t
(160)
intheaxial,radialandcircumferentialdirections.
ThepropellerproducesathrustforceT inthenegativex direction,andabsorbsa
torqueQ aboutthex axis, witha positivevaluefollowingaright-handedconvention.
Thesecan benon-dimensionalizedeither on thebasis of ship speed,V
S
,
T
C
Ts
=
1
V
S
2
(R
2
)
2
137
C
Qs
=
1
2
V
S
2
Q
(R
3
)
(161)
oronavolumetricmeaninow,
T
C
Ta
=
1
V

A
2
(R
2
)
2
C
Qa
=
1
2
V

A
2
Q
(R
3
)
(162)
wherethevolumetricmeaninowis,
2

R
V

A
=
(R
2
r
h
2
) r
h
rV
A
(r)dr (163)
Alternatively, theforces may benon-dimensionalized with respect to anominalrota-
tional velocity,nD,wheren=
2

isthe number ofpropellerrevolutions persecond,

T
K
T
=
n
2
D
4
K
Q
=
Q
(164)
n
2
D
5
The kinematics ofthe ow depends onlyon theratioofthe inow and rotational
velocities,andthisiscustomarilyexpressedasanAdvanceCoecient J
S
,
J
S
=
V
S
(165)
nD
or
V

A
J
A
= (166)
nD
By introducing Eq. 165 into Eq. 161, weseethat the twosystems offorce coecients
arerelatedasfollows;
138
C
Ts
=

8 K
J
S
2
T
=
16K
Q
(167) C
Qs
J
2
S
withasimilarrelationshipforC
Ta
andC
Qa
.
Inmostcases,theK
T
, K
Q
systemispreferred,sincethepropellerrotationalspeed,n,
can bepreciselymeasured bothinthelaboratoryandonshipboard,andthecoecients
remainniteinthecaseofstaticthrust,whenV
A
=0. However, aswewillseefrom
thenextsection,theC
T
, C
Q
systemhasamoredirectrelationshiptopropellereciency.
Thus, bothsystemsareusefuland havetherefore beenretained overthe years.
139

3.3 ActuatorDisk
We will rst consider the simplest possible idealization of a propeller the so-called
ActuatorDisk, which wasrst introduced by Rankineand Froude.The physicalpropeller
isreplaced bya permeabledisk ofradius R,with vanishingthickness inthex direction.
Thediskintroducesauniformjumpintotalpressure,p
t
,totheuidpassingthrough
the disk, which tends to accelerate the uid in the positive axial direction and thus
resultsinathrustforceinthenegativexdirection. Notangentialvelocities,orswirl, are
introduced by thedisk,andas aconsequence oftheprinciple ofconservation ofangular
momentum,thereisnotorquesupplied.
It isdicult, atthis point, torelatethisamazingdevice to areal propeller. however,
wewillultimatelyseethattheactuatordiskisreallythelimit, inanidealuid, ofa
propeller with an innite number ofblades, zero chord length, andinnite rotational
speed. But right now, we will develop expressions forthethrust andeciency ofan
actuatordiskbasedonconservationofmomentumandenergy.
Thedevice isassumed to be operating in an unbounded, invisciduid,with auni-
forminow velocity, V
A
anduniform staticpressure, p
0
farupstream. Since theow
isaxisymmetricandwithoutswirl,weareleftwithaxialandradialvelocities,u
a
(x, r),
u
r
(x, r) asafunctionofx andr only.
Sincetheowisinviscid,thetotalpressureinaccordancewithBernoullisequation
is constant along any streamline, except for those that pass through the disk, where
a total pressure rise of p
t
occurs. Far downstream of the disk, we can expect that
owquantitieswill beindependentofx,withthetipstreamlineachievingsomelimiting
radiusR
w
. Theaxial perturbationvelocitywill beu
w
(r) for r < R
w
,andwill bezero
forr > R
w
. Theradial velocity will bezeroforallradii,andthestaticpressurewill be
independentofradius,withavalueequalingtheupstreamvalueofp
0
.
42
Since no uid iscreated within the disk, axial and radial velocities are continuous.The
increase intotalpressure p
t
isthereforefelt entirely as anincrease instaticpressure, p.
WritingBernoullisequation between a pointfarupstreamand a pointfar downstream
alonganystreamtubewhich passesthroughthedisk,
p
0
+
1
2
V
A
2
+p = p
0
+
1
2
(V
A
+ u
w
(r))
2
p = u
w
(r) V
A
+
u
w
(r)
(168)
2
42
In the absence of tangential velocities, a radial pressure gradient would imply a curvature to the
ur
streamlines. But,fardownstreamthestreamlinecurvaturemustbezerosince
u
x
a
=0andu
r
=
x
= 0.
140



Since p isindependentofradius,Eq.168yieldstheimportantresultthatthe per-
turbationvelocity intheslipstreamfardownstreamisindependentofradius,
u
w
(r) = u
w
=const (169)
Thetotalthrust,T,ontheactuatordiskcan bewrittendownimmediatelyas
T =R
2
p=R
2
u
w
V
A
+
u
2
w
(170)
Introducing thedenition ofthrustcoecient fromEq. 161, we obtainacompact
non-dimensionalformofEq. 170which onlyinvolvestheratiooftheslipstreamvelocity
totheadvancespeed,
T u
w

1 u
w

C
T

1
R
2
V
A
2
= 2
V
A
1 +
2 V
A
(171)
2
Wecanalsoobtainanindependentexpressionforthethrustbasedonconservation
ofmomentum,andthiswillenableustoobtaininformationaboutthevelocity eldat
thedisk. Thegeneralvectorformofthemomentumequationis,
F

+ pndS = V

(V

n)dS (172)
S S
which statesthattheuxinmomentumoftheuidpassingthroughanarbitrarycontrol
volume is equal to thesum of thepressureforces acting on the boundary of the volume
andthetotal bodyforce,F

,actingwithinthe volume. Inthiscase,theresultant body

forceisthetotalthrust,T, which actsinthenegative xdirection.Thus,thexcomponent
ofthemomentumequationcan bewrittenas,
T = pn
x
dS+ (V
A
+ u
a
)n
x
dS (173)
S S
Wewilldisposeofthepressureintegralrst,sinceitfortunatelywillturnoutto be
zero! Referringtogure82,wechooseacontrolvolumewhoseoutersurfacecorresponds
tothestreamsurfacewhich passesthroughthetipofthedisk,andwhoseupstreamand
downstream boundaries aresuciently far from the disk for the velocity andpressure to
havereachedtheirlimiting values. Thus,attheupstream boundary,thecontrol volume
radiusisR
u
,thepressureisp
o
,andthevelocityisV
A
. Atthedownstreamboundary,
141
theradiusisR
w
,thepressureisagainp
o
andthevelocity isV
A
+ u
w
. Conservationof
massrequiresthat
R
2
V
A
=R
2
(V
A
+ u
w
) (174)
u w
The netpressureforce acting on the two ends isthereforep
o
(R
2
R
2
) inthe positive
u w
xdirection.Determiningthexcomponent ofthepressureforceactingontheoutersurface
presentsaproblemsincewedonotknowthedetailsofitsshapeorpressuredistribution.
To overcomethisproblem, wewillexamineanothercontrol volumewhoseinner boundary

matchestheouter boundary ofthepresentcontrol volume. Theouter boundarywill be a

streamsurfacewhoseupstreamanddownstreamradii,R

, R
d
arelargeenoughforthe
u
velocitytohavereturnedtothefreestreamvalueV
A
. Sincethenewcontrolvolumelies
outside the propellerslipstream, the velocity at both theupstream and downstreamfaces
areV
A
. Conservationofmassrequiresthatthetwofaceshaveequalarea,andsincethe
pressuresareequalthereisnonetpressureforce. Again,conservationofmassrequires
that

d
)
2
u

(R
sothatthexcomponentofthepressureforceactingontheouter boundaryisp
o
(R
2

u
R
2
) inthenegative x direction. Sincethereisnomomentumuxoutof,andno body
w
forcewithintheoutercontrol volume,thenet integral ofpressure must bezero. Hence
thepressureforceontheinnerboundaryoftheoutercontrolvolumeisp
o
(R
2
R
2
)
u w
inthe positivex direction,whichmust beequalandoppositetotheforceontheouter
surface of the inner control volume. Now we see that this force justbalances thedierence
inpressureforceontheendsofthecontrolvolume,thusprovingtheassertionthatthe
netpressureforceiszero.
Weneedto be alittlecarefulindevelopingtheexpressionforthemomentumux,
since the velocities in the plane of the disk cannot be assumed to be independent of
radius. Werst introduce thefollowingnotation for the perturbation velocity at the disk
)
2
(R =R
2
R
2
(175)
u d

u
a
(r) u
a
(0, r)

u
r
(r) u
r
(0, r) (176)
whichwillalso beusedlater indevelopingpropellerlifting-linetheory. Nowconsidera
dierentialstreamtubeofradiusdratthedisk. Themassowthroughthestreamtube
is
142
Figure 82: Controlvolumeforactuatordiskmomentumcalculation. Thestreamtube
contractionhas beenexaggerated forclarity.
dm
= 2r(V
A
+ u

(r))dr (177)
dt
a
andthetotalthrustis

R
T = 2u
w
(V
A
+ u

(r)) rdr
a
0

R
= R
2
u
w
V
A
+2u
w
u

(r)rdr
a
0
= R
2
u
w
(V
A
+ u

) (178)
a
where

2

R

u
a
=
R
2
u
a
(r)rdr (179)
0
isthemassaveragedaxial perturbationvelocityatthedisk. ComparingEq.s170and
178, weseethat themass-averagedaxial perturbation velocity at thedisk isexactlyhalf
of the perturbation velocity far downstream,
143

u

=
u
w
(180)
a
2
Ifweintroduceequation180intoequation171,weobtainaveryusefulexpression
thatrelatestheaxial velocityatthedisktothethrustcoecient,

V
u
A
a

=
1 +
2
1 + C
T
(181)
Forsmall values ofthrustcoecient, C
T
<<1,equation181 becomes,
V
u
A

a

C
4
T
(182)
Mosttextbookpresentationsofactuatordisktheory, forsimplicity,donotdistinguish
betweenaveragedowquantitiesandtheiractualradialdistribution,andthereforeleave
theimpressionthatthe perturbationvelocityatthediskisindependentofradius. In
fact,detailedcomputationsshowthattheaxial perturbation velocity isnearlyconstant
overmostofthediskbutdecreasessignicantlyastheradiusapproachestheradiusofthe
disk. Thiscan beseenfromgure 83which showstheresults of a numericalcalculation
for the entire oweld for athrust coecient ofC
T
= 2.0 A detailedanalysisoftheow
right at the tip shows that the velocity eld has alogarithmicsingularity at this point, so
thattheactualowintheimmediateneighborhood ofthetipisactuallyquitecomplex.
Fortunately,forlowandmoderatevaluesofthrustcoecient,thisregionisverylocal
andisinvisibleinanumericalsimulation.
We can now write down an exact expression for the radius of the slipstream far
downstream,R
w
by applyingconservationofmass,
R
2
(V
A
+
1
u
w
) = R
2
(V
A
+ u
w
)
w
2
R
w
uw
1 +
1
2
=
V
A
(183)
R 1 +
uw
V
A

1 + 1 + C
T
=
2 1 + C
T
Thus,weseethatasthethrustloadingincreases,theultimateslipstreamradius,R
w
,
144
Figure83: Resultsofnumericalcalculationofslipstreamradiusandvelocityeldinthe
planeofthediskandfardownstream.ThethrustcoecientisC
T
= 2.0.
decreases. Inthelimitofstaticthrust
43
,whentheadvancespeediszero,thethrust
coecient becomes innite, and the slipstream radius reaches an asymptotic limit of

R
w
= 1/ 2, asshowningure84
Inthiscase,SchmidtandSparenberg
44
haveshownthatthetipstreamtubehasa
logarithmicspiralbehaviorintheimmediatevicinityofthetip,passesupstream, and
thendoubles back to pass through the disk at aradius of approximately 0.97R, asshown
ingure85. Furthermore,streamtubesjustoutsidecrossthedisk twice. Fluidparticles
traversingthesestreamtubesexperiencenonettotalpressurerise,sothattheirpressure
andvelocity fardownstreamreturntofreestreamvalues. Inspiteofthiscomplexow
behaviornearthetip,Eq.183isstilltheoreticallyexact. However,oneshouldremember
that anexactsolution to theactuator disk problem does not imply that it is a physically
realizableow.
The eciency of the actuator disk can be dened as the ratio of the useful work
done by thedevice to thetotalenergysupplied to it. As shown in gure 86, in axed
coordinatesystem, the disk will be moving in the negativexdirectionwithspeedV
A
. In a
timeincrementt,thediskmovesx=V
A
tinthenegativeX direction,andtheoutput
43
This is sometimes referred to as bollard pull, signifying that the ship is being held stationary at a
dock while power is applied to the propeller. It is also a condition encountered initially when a ship
starts from rest.
44
G.H. Schmidt and J.A. Sparenberg, On the Edge Singularity of an Actuator Disk with Large
Constant Normal Load,Journalof Ship Research,June1977,Volume 21, Number 2.
145
Figure84:Ultimateslipstreamradiusasafunctionofthrustcoecient,C
T
fromEq.183
workwill beW =Tx.At thesametime,thetotalamount ofkineticenergyimpartedto
theuidwill beincreased bysomeamountwhichwewillcallE. Conservationofenergy
thenrequiresthatthetotalenergyinput be (Tx+ E). Theeciencywillthen be
Tx T
=
Tx+ E
=
T + E/x
(184)
We now need to obtain an expression for E. During the time interval t, a uid
particleintheslipstreamfardownstreamwillhavemovedadistance
x
w
= (V
A
+ u
w
)t=
(V
A
+ u
w
)
x (185)
V
A
relativetothedisk. Theincreaseinkineticenergyimpartedtotheuidisthereforethe
kineticenergycontainedinacylindricalvolumeofuidofradiusR
w
andaxiallength
x
w
,

V
A
+ u
w

E = u
2
R
2
x (186)
2
w w
V
A
IntroducingEq.183,weobtain,
146
E 1
= u
2
R
2
V
A
+ u
w
/2
(187)
x 2
w
V
A
But,wealsoknowfromEq.178that
T =R
2
u
w
(V
A
+ u
w
/2) (188)
sothatEq.187can bewrittenexplicitlyintermsofthrust,
E 1

u
w

x
=
2 V
A
T (189)
whichwecanintroduceinEq.184toobtainanexpressionfortheeciency,
Tx 2
=
Tx+ E
=
1 +

1 + C
T
(190)
Eq. 190 is exact, in spite of the fact that the simple momentum/energy analysis
presented here cannotquantify thecomplex localow near thedisk. This important
equationshowshoweciency reduceswithincreasingthrustcoecient, andserves as
anupperboundoneciencyforrealpropulsors. Thisresultisplottedingure87,
togetherwiththeresultsforthemoregeneralcasewhichwewillconsiderlater.
147
1.0
0.99
0.9
1.0
0.8 0.7
0.6 0.5
0.4
0.3
0.2
0.1
-1.0 -0.5 0.0 0.5 1.0
y
(a)
x
Stream tubes near the tip of an actuator disk in static thrust. Note that the tip
streamtube (labeled 1.0) initially goes upstream. Adapted from Schmidt and
Sparenburg.
Figure by MIT OCW. Adapted from Schmidt and Sparenberg, 1977.
Figure86: Controlvolumeforactuatordiskenergy balance
Theresultsdevelopedsofarareforthesimplest possiblecase ofanactuatordiskwith
uniformtotalpressurerisewithnoadditionofswirl. Whileonecantreatthegeneral
case,wewilldeferthisuntilwehavedevelopedthefundamentalsofliftinglinetheory.
Wewillthen beabletoshowthattheactuatordiskcan berecoveredas a speciallimit
ofliftinglinetheory.
148
Figure87:Eciencyasafunctionofthrustcoecientforthegeneralcaseofanactuator
diskwithswirl.ThecurveforJ = 0 correspondstoEq.190.
149
3.4 PropellerLiftingLineTheory
We will now take one stepcloser to the real world and develop a lifting linerepresentation
of a propeller. As shown in gure 88, each propeller blade can beconsidered as a lifting
surfacewithsomedistribution of boundandfree vortexsheetstrength. Wethenconsider
the limitingcase of vanishing chord length, which isrepresented in the right hand portion
of the gure. As in thecase of the planar foil, the bound vortexsheetreduces to a single
concentratedvortexofstrength(r) oneachblade. Sinceallbladeswillhavethesame
circulation distribution in circumferentially uniform ow, we can select one blade (or
liftingline)anddesignateitasthekeyblade.
Figure88: Illustrationoftheconceptofalifting-linepropellerasalimitofvanishing
chord length.Theradialdistributionofbladecirculation,(r) remainsthesame,sothat
thestrengthofthetrailingvortexsheet,
f
(r) isunchanged.
Thestrengthofthefreevorticityinthewakemaybefoundfromtherelationships
developed in section 2.2. In this case, the s
2
curvilinear coordinate, as illustrated in
gure 88, will bemore or lesshelical in form, but may also contract (as we saw with the
actuatordisk)asitprogressesdownstream. However, ifwemaketheassumptionthat
thefreevortexlinesareconvecteddownstreamwithaconstantradius,thefreevortex
sheetstrength,asinthecaseoftheplanarliftingline,can beobtaineddirectlyfromthe
derivativeofthespanwise(inthiscase,radial)distributionofcirculation,
150

f
(r) = (191)
dr
Wecandevelopexpressions fortheforcesactingatradiusr onthekeyliftingline
fromalocalapplicationofKutta-Joukowskislaw. Figure89,showsacombinedvelocity
andforcediagram. Theaxialandtangentialinducedvelocitiesduetothehelicalfree
vortexsystem,u

(r),u

(r) combinewiththeeectiveinowvelocity componentsV

A
(r),
a t
V
T
(r) andthepropellerrotationalspeed rtoproduceaninowvelocityV

orientedat
anangle
i
withrespecttotheplaneofrotation,
Figure89: Velocityandforcediagramat a particular radial positionon aliftingline.
V

(r) = (V
A
(r) + u

(r))
2
+(r+ V
T
(r) + u

t
(r))
2
; (192)
a
V
A
(r) + u

(r)
a

i
(r) =tan
1
r+ V
T
(r) + u

t
(r)
. (193)
Thequantity
i
isthereforeanalogoustotheinducedangleinwinglifting-linetheory.
Theforce perunitradiusonthe vortex,F(r) istherefore
F(r) = V

(r)(r), (194)
151
andisdirectedatrightanglestoV

.
It isrelatively simple toinclude theeect ofviscous drag at this stage by adding a
forceF
v
(r) actinginadirectionparallelto V

. Thisforcemay beestimatedonthebasis
ofanexperimentally determined, ortheoretically calculated two-dimensional sectional
dragcoecientC
Dv
(r). Thismeans,ofcourse,thatthesectionchordlengthsc(r) must
be specied. Theviscousdragforcewillthen be
1
F
v
(r) = (V

(r))
2
c(r)C
Dv
(r). (195)
2
Theseforcescanthen beresolvedintocomponentsintheaxialandtangentialdirec-
tion,integratedovertheradiusandsummedoverthenumberofbladestoproducethe
totalpropellerthrustandtorque.

T =Z
r
h
V

cos
i

2
(V

)
2
cC
Dv
sin
i
dr; (196)

Q=Z
r
h
V

sin
i
+
2
(V

)
2
cC
Dv
cos
i
rdr. (197)
NotethatV

cos
i
is simply thetotal tangential velocity acting at the lifting line, (V
t
+
r+ u

),andthatV

sin
i
istheaxial velocity, (V
a
+ u

).
t a
The Velocity Induced by Helical Vortices
Thevelocity inducedatradiusr
c
onthekeyliftingline byaset ofZ unitstrengthhelical
vorticesshedatradiusr
v
can beexpressedasanintegralusingthelawofBiot-Savartas
developedinequation104.
Z
1

r
v
[r
v
r
c
cos(+
k
)]
u
a
(r
c
, r
v
) =
4 0 [(r
v
tan
w
)
2
+ r
v
2
+ r
c
2
2r
v
r
c
cos(+
k
)]
3/2
d (198)
k=1
Z

u
t
(r
c
, r
v
) =
1

r
v
tan
w
[(r
c
r
v
cos(+
k
)) (r
v
sin(+
k
)]
d (199)
4
k=1
0 [(r
v
tan
w
)
2
+ r
v
2
+ r
c
2
2r
v
r
c
cos(+
k
)]
3/2
Intheaboveexpressions,
w
isthepitch angleofthehelixatr
v
. Accordingtolinear
theory,
w
=,butwewillleave
w
unspecied atthemomentinordertofacilitate
152
subsequent renements to the theory. The variable of integration, , is the angular
coordinate of ageneral pointonthehelixshedfromthe keyblade. Thecorresponding
angularcoordinateof a pointonthek

th bladeisfoundby addingthebladeindexing
angle,
k1

k
= 2 k= 1, . . . , Z (200)
Z
The totalinduced velocityon the lifting line can now beobtained by integrating the
contributionsofthehelicalvorticesovertheradius,
u

(r
c
) =

c
R

(r
v
)
u
a
(r
c
, r
v
)dr
v
a
r
h
r
u

(r
c
) =

c
R

(r
v
)
u
t
(r
c
, r
v
)dr
v
(201)
t
r
h
r
However,Equations198-199cannot beevaluatedanalytically,sothatone mustresort
tosomeformofnumericalsolution. Ontheotherhand,ifthelimitsofintegrationare
changedfrom 0, to,,theywouldthenrepresentthevelocitiesinducedalonga
radialline byaset ofhelical vorticesextending toinnity in bothdirections. This would
be equivalent to the velocity inducedinnitely far downstream by thefree vorticity shed
fromtheoriginalliftingline. SincetheintegrandsinEq.198-199areevenwithrespect
tothevariableofintegration, ,
45
thevelocitiesinnitelyfardownstreamaredouble
theirvaluesattheliftingline.
Nowimagine ahelicoidalcoordinatesystem,withonecoordinatealongthehelix,one
coordinateradial, and thethird coordinatenormal to therst two. Far downstream, the
owwill beindependent ofthehelicalcoordinate. Thus,theow is two-dimensional in
terms of the tworemaininghelicoidal coordinates.
The potentialproblemforthis type ofow wassolved in dependently by Kawadaand
Lerbs, and aderivation may befound in Appendix 1 ofLerbs paper
46
.
Theresulting potentialcan beexpressedintermsofinnitesumsofmodiedBessel
functions. Whiledirect evaluation ofthesefunctions might be as timeconsuming as nu-
mericalintegrationoftheBiot-Savartlawresult,fortunatelyhighlyaccurateasymptotic
45
Providedthattheorderofsummationoverthebladeindex,k,isreversedwhen isreplacedby.
46
Lerbs,H.W.,Moderately Loaded Propellerswith a Finite Number of Blades and an Arbitrary Dis-
tribution ofCirculation,Trans. SNAME,Vol. 60,1952
153


formulasforthesumsofBesselfunctionsexist. Thisenabled Wrench
47
todevelopthe
followingclosedformapproximationstotheinducedvelocities;
Forr
c
< r
v
:
Z
u
a
(r
c
) =
4r
c
(y2Zr
v
F
1
)
Z
2
u
t
(r
c
) =
2r
c
y
0
F
1
(202)
Forr
c
> r
v
:
Z
2
u
a
(r
c
) =
2r
c
yy
0
F
2
Z
u
t
(r
c
) =
4r
c
(1 +2Zy
0
F
2
) (203)
where
1

1 + y
2

0.25
1 1

9y
0
2
+2 3y
2
2

F
1

2Zy
0
1 + y
0
2
{
U
1
1
+
24Z (1 + y
0
2
)
1.5
+
(1 + y
2
)
1.5
1
ln 1 + }
U
1
1
1

1 + y
2

0.25
1 1

9y
0
2
+2 3y
2
2

F
2

2Zy
0
1 + y
0
2
{
U 1

24Z (1 + y
0
2
)
1.5
+
(1 + y
2
)
1.5
1
ln 1 + } (204)
U 1
and
_

_
Z
_
y
0
( 1 + y
2
1)

_
U =
_
y(

1 + y
0
2
1)
exp 1 + y
2
1 + y
0
2
_
r
c
y =
r
v
tan
w
1
y
0
=
tan
w
(205)
Figure90showstheresultsofEquations202and203foraparticularcase. Asone
wouldexpect,thevelocity tendstoasthecontrol point radiusapproachesthe vortex
47
Wrench,J.W.,The Calculationof PropellerInduction Factors,,DTMBReport 1116,Feb. 1957.
154
Figure90: Velocityinducedonaliftinglineatradiusr
c
byasetofsemi-innitehelical
vorticesoriginatingatr
v
= 1.0.ThenumberofbladesinthiscaseisZ = 5.Resultsare
shownforpitchangles
w
= 10,20,30,40,50,60degrees.
radius. Whenthecontrol point is wellinsidethe vortex,theaxial velocity approaches a
constant value,which increaseswithdecreasingpitch angleofthehelix.Whenthecontrol
pointisoutside the vortex, theaxial velocity approaches zero rapidlywith increasing
radius,andisrelativelyinsensitivetopitch angle.
Thereverseistruewiththetangentialvelocity. Insidethevortex,thetangentialve-
locity approacheszerowithdecreasingradius.Outsidethevortex,thetangentialvelocity
appearstoreduceslowlywithincreasingradius. Thetangentialvelocity isrelativelyin-
sensitivetopitch angle,exceptintheimmediatevicinity ofthevortex.Thisisincontrast
to theaxial velocity,which isextremelysensitive to pitch anglewhen thecontrol point
isinsidethevortex.
Thelimit ofinnite number ofblades is ofparticular importance,sincethiswill yield
anaxisymmetricowthatwecanrelatetotheactuatordisk. Asthebladenumberis
increased,thequantityU inEquation204 becomesinnite,andas aresult,F
1
andF
2
bothapproach zero. Equations202and203thenreducetothesimpleexpressions,
155
Forr
c
< r
v
:
u
a
(r
c
) =
Z
4r
v
tan
w
u
t
(r
c
) = 0 (206)
Forr
c
> r
v
:
u
a
(r
c
) = 0
u
t
(r
c
) =
Z
4r
c
(207)
Figure91:Eectofbladenumberonthevelocityinducedonaliftinglineatradiusr
c
by
asetofsemi-innitehelicalvorticesoriginatingatr
v
= 1.0.Thepitchangleis
w
= 30
degrees.ResultsarealsoshownforaninnitenumberofbladesfromEquations206-207.
Thetotalcirculation,Z,iskeptconstantasthebladenumberisvaried,andmatches
thevalueusedforthevebladed propellershowninFigure90
Thesingular behavior of theinduced velocities issimilar tothat of astraightvortex,
where weknowthatthe velocity behavesas 1/(r
c
r
v
). Itisthereforeusefultofactor
outthesingularpart,leavingaregularfunctionthatdependsonthegeometryofthe
156
helix. Lerbsdenedaninductionfactoras theratio of the velocity induced byZ helical
vortices to the velocity induced byasemi-innitestraightvortex of thesamestrength,
i
a
(r
v
, r
c
)
u
a
(r
v
, r
c
)
1
4(rcrv)
i
t
(r
v
, r
c
)
u
t
(r
1
v
, r
c
)
(208)
4(rcrv)
Astheradiusofthevortex, r
v
,approachestheradiusofthecontrol point, r
c
, the
velocity induced by thehelical vorticeswillapproach the valueinduced byasemi-innite
vortexoriented in adirectiontangent to thehelix at itsstarting pointon theliftingline.
Therefore,asr
c
r
v
,wendthat
u
a
(r
v
, r
c
)
cos
i
4(r
c
r
v
)
sin
i
u
t
(r
v
, r
c
)
4(r
c
r
v
)
(209)
Comparing209with208weseethatinthelimitasr
c
r
v
,theaxialandtangential
inductionfactors become,respectively,cos
i
andsin
i
.Therefore,theinductionfactors
remainniteasthevortex pointandcontrol pointcoincide, whiletheactualvelocity
tendstoinnity. Thereasonfortheminussigninthedenitionoftheaxialinduction
factorisstrictlyfortheconvenienceofmakingtheinductionfactors positive.
Withthehelicalvortexinuence functionsknown, wearenowreadytotacklethe
evaluationofthesingularintegralsfortheinducedvelocitiesgiveninEq. 201.However,
before we do this, we willrevisitactuator disk theory this time from the point of view
ofaliftinglinevortexmodel.
3.4.1 TheActuatorDiskasa ParticularLiftingLine
It turns out that we can construct an actuator disk, using the concepts of propeller
liftinglinetheory. Considerthelimitingcaseofaninnitebladedliftinglinewithzero
hubradius,anduniformcirculation (r) = overtheradius. Inthelimitingprocess,
we will keep the product of the number ofblades, and the circulation per blade, Z
constant. Furthermore, since we saw that an actuator disk generates zero tangential
inducedvelocity (andhenceabsorbszeroshafttorque),wewillneedtoadoptoneofthe
followingmodels,
157
Figure92: Axialinductionfactorsfor a 5bladed propellerderived fromFigure90. The
enlarged plotshowsthelocal behaviornearr
c
/r
v
= 1.Theanalyticallimitofi
a
=cos
w
isplottedassquaresymbolsonthegraph.
1. Wecanassumethattherotationalspeed, ishigh, sothatu

<< r. Thisis
t
equivalenttotakingthelimitofzeroadvancecoecient,J
A
.
2. Wecanconstructanidealcounterrotatingpropeller by superimposing twoiden-
tical,innitebladedliftinglinesrotatinginoppositedirections. Inthiscase,the
axialinducedvelocitiesfromeachliftinglinecomponentwilladd,whilethetan-
gentialcomponentwillcancel. ThequantityZwillbedividedequallybetween
the twocomponents,butthesignof will beopposite. Withthecounterrotating
model, the specication of advance coecient isarbitrary.
Wewillnextshowthateithermodelwillrecovertheresultsobtainedpreviously by
momentum/energyconsiderations. Be begin by writing downtherelationship between
158
thrustandcirculation,usingequation196.Settingtheviscousdragtozero,andmaking
theassumptionthatu

<<r,
t

T =
0
V

cos
i
Zdr
0
rZdr=
2
ZR
2
(210)
whichwecanalsoexpressas athrustcoecient,

ZR
2
Z
C
T
=
1
2
= (211)
V
2
R
2
V
2
2
a a
Wenextobtainanexpressionfortheaxialinducedvelocity, u

. Sincethecirculation
a
isconstant,therewill be aconcentratedhelical tip vortex ofstrength shedfromeach
blade,andfromequation206,thesewillinduceaconstantaxial velocity,

Z
a
u =
4Rtan
w
(R)
(212)
andzerotangential velocity. But,therewillalso be aconcentratedhubvortexofstrength
shedfromeach bladeat r= 0. Fromequation 207thesewillinducezeroaxial velocity.
Thetangentialvelocityinducedbythehubvortexwillnotbezero(unlessweusethe
counterrotatingoption),but we haveassumed,nomatterwhat,that itwill benegligible.
Ifwemakethelinearizingassumptionthatthetangentofthepitchangleofthetip
vortices,
w
isequaltothethatoftheundisturbedow,tan,wecaneliminateZ in
equation212using211
u

a
=

V
a
2
C
T

=
C
T
(213)
V
a
4R V
a
V
a
4
ThisagreeswiththeactuatordiskresultinthelimitofvanishingC
T
whereu

/V
a
<<
a
1,asderivedpreviouslyinequation182. However,wecanobtainamoreaccurateresult
ifwerecognizethatthepitchangleofthetipvorticesisnottheundisturbedangle,
butthat,atleastinitially,istheangle
i
. Since
V
a
+ u

a
tan
i
= (214)
R
wecanrewriteequation212as
159

Z ZR
a
u =
4Rtan
i
(R)
=
2R(V
a
+ u

)
(215)
a
2R(V
a
+ u

)u

=ZR (216)
a a
andeliminateZ
Z =
V
a
2
C
T
(217)
2
toobtaintheresult
(V
a
+ u

)u

=
V
a
2
C
T
(218)
a a
4
Finally,wecanputequation218instandardquadraticformandsolveitforu

/V
a
a
(u

)
2
+ u

V
a

V
a
2
C
T
= 0 (219)
a a
4

u

a
1 + 1 + C
T
V
a
=
2
(220)
whichisexactlytheactuatordiskresultgiveninequation181. Thisisremarkable,in
away,sincetheactuatordiskresultincludesthecontractionoftheslipstream(butis
unaware of tip vortices), while the present result models the tip vortices as constant
radiushelicallieswithconstant(althoughsuitablyadjusted)pitchangle. Wewillsee
laterhowthisrelatestotheso-calledmoderatelyloaded theoryofpropellers.
Wecannoweasilyderivetheexpressionforeciency. Theinput poweris

Q= r(V
a
+ u )Zdr= Z(V
a
+ u )R
2
(221)
0
a
2
a
whiletheoutput poweris

TV
a
= V
a
ZR
2
(222)
2
160
Sotheeciency, istheratioofthetwo,
1 2
=
1 +
u
V

a
a
=
1 +

1 + C
T
(223)
Thisisalsotheexactactuatordiskresultgiveninequation190.
3.5 OptimumCirculationDistributions
Wewouldliketoobtaintheradialdistributionofcirculation,(r),which willminimize
the torque, Q, for a prescribed thrust, T. The diameter, advance coecient, blade
number, andeective inow arespecied.
48
While otherconsiderations, such as the
inceptionoftipvortexcavitation,may requireustodepartfromthisoptimum,thisis
generally a logical starting point in the design process. This problem can be solved using
themethod ofcalculusofvariations,andonecannddetailedaccountsofthisapproach
inYim[],Coney[]andBreslin[]. Whiletheresultingequationscan besolvedrapidlyon
currentcomputers,theyarenon-linear,intricateinappearance,andprovidelittledirect
physicalinsight.
Ontheotherhand,anearlierapproach developed by Betz in1919for apropeller in
uniforminow,V
S
, and laterextended by Lerbs in 1952, isrelatively simple to derive and
is physically intuitative. Suppose we have adistribution ofcirculation which is optimum,
andwhich generatesthedesiredthrust. Nowsupposethat we perturbthiscirculation by
addinganincrementofcirculation oversomeincrementofradiusratsomearbitrary
radiallocationr. Thiswillresultinincrementalchangesinthrustandtorque,T and
Q. Theeciencyofthisprocessistheratiooftheoutputtoinput power,

TV
S
=
Q
(224)
If (r) istrulyoptimum,then

must beindependentofradius. Otherwise,circu-
lationcould bedecreasedat aradiuswhere

is low,andincreased byacorresponding
amountatadierentradiuswhere

is high. Theresult would be areduction intorque
foraxedthrust,thuscontradictingtheassertionthat(r) isoptimum. Sofar,thisis
essentiallyaphysicallybasedstatementofthevariationalprinciple.
However, if we attempted to express Eq. 224 in terms of circulation and induced
48
Alternatively,wemaywishto maximizethrustforaprescribedtorque. In eithercase,theresulting
propellerwill havethe highest possible eciency.
161
velocity inuencecoecients,theresultingexpressions would befarfromsimple. This is
because theincrement ofcirculation introduced at aparticularradius, r, notonly changes
theforcelocally,butalterstheforceovertheentireradiusasaresultofitsalteringof
theinduced velocity distribution. Betz overcamethisobstacle by employing aprinciple
developedbyMunkthatstatesthatthetotalforceonaliftingsurfaceisunchangedif
anelement of boundcirculation isdisplaced in astreamwisedirection. Munkstheorem
followsfromtheprinciplethattheforceon aliftingsurfacecan beobtainedsolelyfrom
afar-eldmomentum/energyanalysis(aswedidfortheplanarliftingsurface),andthat
thefar-eldowdependsonlyonthestrengthofthetrailingvorticity. This,inturn,is
unaected byastreamwisedisplacement of the bound vorticity.
Betzthereforeadded theincrement of bound vorticity far downstream, so thatthere
would be no interaction between the addedcirculation and the ow at the propeller lifting
line. On theother hand, the localforce acting on theaddedelement of boundcirculation
mustincludetheeectofthedoublingoftheinducedvelocitiesfardownstream,
F
x
(r) = [r+2u

(r)]
t
F
t
(r) = [V
S
+2u

(r)] . (225)
a
Ifweassumethatu

t
(r) r andthatu

a
(r) V
S
wecan performsomealgebraic
manipulations,
TV
S

r+2u

(r) +([u

(r)]
2
)/(r)
t t
Q) V
S V
S
+2u
a

(r) +([u

(r)]
2
)/(V
S
)
=
V
S

(r)
2
+2ru
a

t
+(u

t
)
2

=
V
S

r+ u
t

2
. (226)
r V
S
2
+2V
S
u

+(u

)
2
r V
S
+ u
a

a a
If we put equation 226 together with equations 193 and 77 theeciency then becomes
(TV
S
)

r+ u

(r)

2
V
S

2
=

tan(r)

2
. (227)
t
(Q) V
S
+ u

(r) r tan
i
(r)
a
Forapropeller with optimum radialload distribution equation 227must be inde-
pendent of radius, otherwise a redistribution of circulation could increase the overall
eciency. Wethereforehavetheresult
tan(r)
tan
i
(r)
=constant. (228)
162
Thisresult was rst published by Betz in 1919 and is now known as the Betz Con-
dition.
Forthisconditiontoholdtheinducedinowattheliftinglinemustformaconstant
pitchhelicoidalsurface. Inother words, ifourfriendly bugandone ofhiscousins both
hadtheirfeetstucktothissurface,astheyspinaroundandgodownstreamtheywould
notthinktheyweremovinginrelationtoeach other. Betzconditionisthecorrectlinear
result for the inviscid, open water case. The unknown constant in equation 228 is a
function of the desired thrust. To nd that thrust the lifting line velocities must be
determined. Therstsolutionforthecirculationdistributionwhichsatises theBetz
condition was developed by Prandtl, using an approximatemethod toobtain theinduced
velocity. Thiswasfollowedbyanexactsolution,whichwaspublishedbyGoldsteinin
1929.
Similar arguments can lead one to the following result for non-uniform or wake-
adaptedaxialinow. Inthiscase
tan(r)
=

1 w
x
(r). (229)
tan
i
(r)
HeretheaxialinowV
a
(r) isgiven by
V
a
(r) = [1 w
x
(r)] V
S
, (230)
andisanunknown constant. Equation229waspublished in1952byLerbsandis
knownasLerbsCriterion.
163
Figure93: Eciencyversusadvancecoecientforavebladed propellerwithoptimum
radialdistributionofcirculationinuniformow.Resultsaregivenforinviscidow,and
forviscousowwithsectionalLift/Dragratiosof25and50.Theactuatordiskresultis
shownasthesymbolplottedatJ
s
= 0.0.
164
3.5.1 AssigningThe WakePitchAngle
w
Itisevidentfromgure89thatthepitch angleofthelocalowattheliftingline,
i
, is
greaterthanthepitch angleoftheundisturbedow,. Thisdierence becomesgreater
asthepropellerloading,andhenceinducedvelocities, increases. Thismeansthatthe
linearizingassumptionthatthefreevorticesareshedonhelicalsurfacesofpitch angle
becomeslessvalidforincreasedpropellerloadings.
A much betterassumption,which carriesthename moderatelyloaded propellertheory,
istorequirethatthefreevorticesfollowhelicalpathswithpitchangle
i
ratherthan
. This is still not exact, since both the pitch and radius of eachvortexlament should
continuouslydeforminorderto bealignedwiththelocalow. However,theassumption
ofaconstantradiusandpitchhelicalvortexwithpitchangle
i
hasthemeritof being
moreorlesscorrectintheimmediatevicinity oftheliftinglinewithoutundulycompli-
catingtheproblem.Inaddition,wesawthatthisassumptionrecoveredtheactuatordisk
resultexactlyinthelimitofinnitenumberofblades.
Oneshouldalsonotethatthecomputationofthehelicalvortexinuencefunctions
isunchanged. Wealreadycancalculatethemforarbitraryvaluesofthepitch angle
w
.
Wecanthereforejustaseasilysubstitute
i
for inequations198-200orinequation205
whencalculatingtheinuencefunctions.However,theproblemisnowmorecomplicated
sincetheequationsfortheinducedvelocitiesnowinvolve
i
,which,inturn,dependson
theinduced velocities. Aniterativesolution isthereforerequired,androbustandecient
methodsofaccomplishingthiswill bediscussedsubsequently.
Theactualgeometryofthetrailingvortexsheetsisextremelycomplex. Aswesaw
withtheactuatordisk,massconservationrequiresthatthestreamtubescontract. But
inaddition,theedgesofthevortexsheetroll-upinamannersimilartothatobserved
forawing,sothattheirultimateradiusisactuallylessthanthecircumferentialmean
valuederivedfrommassconservation. Inanidealuid,therolling-upprocessdevelops
into a tightly wound spiral, but in a real uid viscous stresses causes the individual
turns of the spiral to blend into a concentrated vortex core, or tip vortex. While
thisprocesscan bemodeled,thecomputationalburdenisextremelylargeandisclearly
inappropriateatthelifting-linedesignstage. Infact,theprecisedetailsofthedeformed
trailingvortexwakearenotcriticalindeterminingtheowatthepropellerblades,so
thatevenlifting-surfaceandpanelmethod calculationscangenerally bemadeeitherwith
themoderatelyloadedwakemodelorwithasemi-empiricalmodelwhichintroduces
thegrosscharacteristicsofthecontractionandrollupprocess
49
49
Greeley,D.S. and Kerwin, J.E.,Numerical Methods for Propeller Design and Analysis in
Steady Flow,Trans SNAME(90),1982.
165

3.5.2 PropertiesofConstantPitchHelical VortexSheets

We saw in Section 3.5 that for an optimum propeller in uniform ow, rtan
i
(r) =
constant. Therefore,whetherweuselineartheoryormoderatelyloadedtheory,thefree
vortex wake will lie on aconstant pitch helicoidalsurface. In this case, it can be proved
that theresultantinduced velocity isnormal to thehelix. Wewill provethis by showing
thatthecomponentalongthehelixiszero.
Figure94: Inducedvelocitiesresolvedintocomponentsnormaltoandalongthehelical
surface.
Asshowningure94,theaxialandtangentialcomponentoftheinducedvelocity
atr
c
can beresolved into componentsnormal to and along thehelix. Since we have to
distinguish betweenthepitchangleatthecontrol point,r
c
,andthevortex,r
v
,wewill
haveto becarefultolabelthemtan (r
c
) tan
c
andtan(r
v
) tan
v
,respectively.
Sincethehelicalsurfacehasconstantpitch,
r
c
tan
c
=r
v
tan
v
(231)
Theinduced velocity along thehelix will be identied as u
s
(r
c
, r
v
) andmay befound
fromtheaxialandtangentialcomponentsusingtherelationship
u
s
(r
c
, r
v
) = u
a
(r
c
, r
v
) sin
c
+ u
t
(r
c
, r
v
) cos
c
(232)
Multiplyingequation198 by sin(r
c
) andequation199 by cos(r
c
) weobtain
1
Z

F
u
s
(r
v
, r
c
) = d (233)
4
k=1
0 D
3
166

wheretosavewritingwehavetemporarilydesignatedthenumeratoranddenominator
oftheintegrandas
D= (r
v
tan
v
)
2
+ r
2
+ r
2
2r
v
r
c
cos(+
k
) (234)
v c
and
F = r
v
2
sin
c
r
v
r
c
cos(+
k
) sin
c
+ r
v
r
c
tan
v
cos
c
r
2
cos(+
k
) tan
v
cos
c
r
2
sin(+
k
) tan
v
cos
c
(235)
v v
Equation235canbesimpliedby replacingr
v
tan
v
withr
c
tan
c
andby making
useofthedenitionofD,
F =sin
c
D
2
(r
v
tan
v
)
2
r
v
r
c
sin(+
k
) (236)
Substituting236forF in233givestheresult,
sin
c

(r
2
tan
2
+ r
v
r
c
sin(+
k
))

u
s
(r
v
, r
c
) =
4
k=1
0 D

0
v v
D
3
d (237)
Sofaritdoesntlooklikewearegainingmuch ground. However,thesecondintegral
inequation237can besimplied by notingthat

1

=
r
v
2
tan
v
2
+ r
v
r
c
sin(+
k
)
(238)
D D
3
sothatequation237maynow bewritten

sin
c
Z

1


u
s
(r
v
, r
c
) =
4
k=1
0 D
+
0

D
d (239)
Thesecondintegralin239can beintegrated by parts,givingtheresult
167

+

D

d
D
(240)

Z
4 0 D
sin
c
d
u
s
(r
v
, r
c
) =
0
k=1 0
The rst andthird terms in 240cancel, and themiddle term is zero atthe lower
limitand 1/r
v
tan
v
attheupperlimit. Sincetheintegralisindependentoftheblade
index,thesum overthe number ofbladessimply multipliestheresult byZ. Thenal
expressionforthevelocity alongthehelixisthen
u
s
(r
v
, r
c
) =
Z cos
c
(241)
4r
c
Afterallthisanalysis,alittleassurancethatequation241istheright answerwouldnt
hurt. Thefollowingtableliststheaxialandtangentialinducedvelocity inuencefunc-
tionsforavebladedpropelleratr
c
= 0.65. Thevorticesareshedonaconstantpitch
helix with r
v
tan(r
v
) = 0.25. The component of induced velocity along the helix is
thencomputedfromequation232andlistedinthelastcolumn. Thevelocity isindeed
constant,andthevalueagreeswithequation241.
r
v
0.200
0.300
0.400
0.500
0.600
0.700
0.800
0.900
1.000
tan(r
v
)
1.25000
0.83333
0.62500
0.50000
0.41667
0.35714
0.31250
0.27778
0.25000
u

a
-.00002
-.00037
-.00484
-.05407
-.77324
2.47548
1.66977
1.60142
1.59285
u

t
0.61214
0.61228
0.61399
0.63293
0.90953
-.33997
-.03009
-.00380
-.00050
u

s
0.57133
0.57133
0.57133
0.57133
0.57133
0.57133
0.57133
0.57133
0.57133
Westarted out by saying that we would prove that theresultantinduced velocitywas
normaltothehelix. Theremainingstepistorecognizethatthetotalvelocityatradius
r
c
isobtained by integrating the product of theinuencefunctions and thefree vortex
strengthovertheradius,

R
d
u

(r
c
) =
r
h
u
s
(r
v
, r
c
)
dr
dr
s

R
d
= u
s
(r
c
)
r
h
dr
dr
168
= u
s
(r
c
) |(r)|
R
r
h
= 0 (242)
Anotherway tolookatthisistoconsiderthatthevortexsheetismadeupofdiscrete
horseshoevortexelements,each consistingofapairofhelicalvorticesofequalmagnitude
andoppositesign. Thesumofthispairofvorticeswillthereforeinducezerovelocity
alongthehelix. Ahorseshoevortexthereforeinduces avelocitywhichisnormaltoa
helicalsurfacewhosepitch matchesitsown. Thesummedeectofallthehorseshoes,or
theintegratedeectofthewholevortexsheet,thereforeinducesavelocity normaltothe
helicalsurface.
3.5.3 TheCirculationReduction Factor
Sinceanoptimumpropellerinuniformowhasaconstantpitchfreevortexsheet,the
resultantinducedvelocity,asshowninAppendix2,isnormaltoahelixofpitch angle
i
.
Thereforeifweknowthecirculationandeithertheaxialortangentialinducedvelocity,
wewillhavealltheinformationweneedtocompletethevelocity/forcediagramshown
ingure89. Sincecirculationandtangentialvelocityarecloselyrelated,itmakessense
tochosethelatter.
Figure95: Circulation pathrelatingcirculationaroundbladestocircumferentialmean
tangentialvelocity.
Thecircumferentialmeantangential velocity,u

(r) can berelatedtothecirculation
tm
applyingKelvinstheoremtothepathillustratedingure95,
Z(r) +4ru

(r) = 0
tm
169
4ru

(r) =
Z
tm
(243)
Forapropellerwithalargenumberofblades,thecircumferentialmeantangential
velocity and the local velocity at thelifting line would be about thesame, and this would
provideatleastaroughestimateofthedesiredsolution. However,foranitenumberof
blades,thecircumferentialmeanvelocityisdierentfromthelocalvalueattheblade,
andinparticular,issubstantiallylessnearthebladetip. Theratio
tm
(r)
u
u

t
(
(
r
r
)
)
(244)
iscalledtheCirculation Reduction Factorwhichdependsonthenumberofblades,the
pitch of the helical vortex wake and the non-dimensional radius, r/R. Clearly, once
isknown, theproblemoftheoptimum propellerinuniformowcanbereduced to
asequenceofsimpletrigonometricevaluationsateachradius. Inparticular,sincethe
resultantinducedvelocity isnormalto
i
,thetangentialvelocityattheliftinglineis
u
V

s
=
sin
i
sin (
sin

i
)
(245)
t
andanexpressionforthecirculationcan befound bycombiningEquations243and245,
(r) 2
r
(r) sin
i
sin (
i
)
G(r)
2RV
s
=
R
Z sin
(246)
However,determiningisfarfromsimple. Therstattempt wasmade by Prandtl in
192?
50
usingtheingeniousanalogy betweentherealowinduced by helical vorticesand
theowofauniformverticalfree-streampastaninnitearray ofsemi-inniteatplates,
asillustratedinFigure96. Faroutboardofthetipsofthearray ofplates,thevertical
velocityisuniform,andequaltothefreestreamvalue. Farinboardofthetipsofthe
plates, the ow must bezero. Near the tips,some of the freestreamleaks in between
theplates, andthis isanalogous to thereduction inmeantangential velocity between an
array ofhelical vortexsheets. Thisproblemcan besolved by classicalconformalmapping
techniques,andtheresultingapproximateexpressionforthecirculationreductionfactor
(which becameknownasthePrandtltipfactor) is,
50
Prandtl,l L,Application ofModern Hydrodynamicsto Aeronautics,NACA Report 116
170
_

_
2 Rr
1 +tan
2

w
(R)
(r) =

cos
1
exp
_
2 R tan
w
(R)
_
(247)
Figure96:Prandtlssimpliedrepresentationoftheowinduced byhelicalvortexsheets.
Prandtlsworkwasfollowedin1929 byGoldstein
51
whosolvedtheexactproblem
of the potentialeld of aninnitehelicalsurfacetranslatingwithuniformaxial velocity.
AswithLerbssolutionforindividualvortexlaments,aseparationofvariablessolution
can befoundinhelicoidalcoordinates. Thesolutionis periodicincoordinatenormalto
thevortexsheetsandcanthusbeexpandedinatrigonometricseries. Thesolutionin
theradialdirectionwasfound to beexpressibleas aseries ofLommelfunctions,which
arerelatedtoBesselfunctions. Goldsteinselegantsolutiontothisproblemisnotforthe
mathematicallytimid,but boththeoreticiansandpracticaldesignersshouldappreciate
thetremendouscontributiontoshipandaircraftpropellerdesignthatthisrepresents.
WhilesomenumericalresultswereincludedinGoldsteinsoriginalpaper,morecom-
pleteandaccuratecomputations weresubsequentlycarriedoutwhencomputers became
available. TachmindjiandMilam
52
publishedgraphsandtablesoftheGoldsteinfactor
in1956,and Wrench
53
computedhigh-precisionvaluesin19??.
Theprecedingresults were all for theidealizedcase of a propellerwithzero hubradius.
Tachmindji
54
extendedGoldsteinstheorytoincorporateaninnitecylindricalhubof
arbitraryradius.Finally, circulationreductionfactorscanalways bedeterminedafterthe
factfromageneral-purposenumericallifting-linecomputationusingEquation246where
boththecirculationandinducedvelocitiesarecomputed. Ifonerunssuch acode(which
will be described later in this Chapter) for the special case of an optimum propeller
inuniformow,Goldsteinsresultshouldberecovered. Anexampleofthecirculation
reductionfactorisgiveninFigure97,wheretheGoldsteinfactor(actuallyobtainedfrom
51
Goldstein,S.,On the vortextheory ofscrew propellers,Proc. RoyalSoc. 123A, 1929
52
Tachmindji and Milam, The calculation of Goldstein factors for three, four, ve and six bladed
propellers,DTMB Report 1051,1956.
53
Wrench, ???
54
Tachmindji, SomethinglikeGoldstein Factorsfor Finite Hubs
171
anumericallifting-linecode)iscomparedwiththePrandtltipfactor,Equation247.The
simplePrandtlapproximationissurprisinglyaccurateattheouterradiiforrelativelylow
pitchangles. Attheinnerradii,thePrandtlresultapproachesavalueofunity, while
theGoldsteinreductionfactorcanbesubstantiallygreater,particularlyatlargepitch
angles.
Figure97: ExampleofCirculationReduction Factors
3.5.4 ApplicationoftheGoldstein Factor
Ifwestartwithagivenspeed,V
S
,diameter,D andrevolutions persecond,n,wecan
calculatetheadvancecoecientJ
S
=V
S
/(nD) andthepitchangleoftheundisturbed
ow,(r) atselectedradii between the hub and the tip. For an optimum propeller in
uniformow,theBetzconditionrequiresthat
i
(r) beaconstant multipleof(r).The
inverseofthismultiplieristheidealeciency(theeciencyinanidealuidwithno
viscousdrag)ofthepropeller. Suppose,forthemoment,thatwearbitrarilypickavalue
forthismultiplier. Wecanthenconstructavelocity diagramsimilartotheoneshown
ingure89foreachoftheselectedradii. WecanalsoreadthevaluesoftheGoldstein
factor,, from thediagramcorresponding to the number of propellerblades,i.e.gure 97
forasixbladedpropeller.
172
Wecanthencalculate(r) fromequation246,andwecanalsocalculatethethrust
andtorquefromequations196and197. Iftheresulting thrustisnotwhatwewant,
the processwill have to be repeated a number oftimeswithdierenttrial values of the
multiplier.
Thiskind of aniterative process would have beenimpractical inpre-computer days.
Toavoidthis, Kramer,in1939,madesystematiccalculationsfordierentnumbersof
blades,advancecoecientsandidealeciencies. Foreachcombination,Kramerfound
thecirculationdistributionandtheresultingideal
55
thrustcoecient,C
Ti
.
The resultistheKramerdiagram, which isreproduceinFigure98
56
. Onecould
enterthisdiagramwiththedesiredthrustcoecientandadvancecoecient,andnd
theidealeciency. Thedesignercouldthen proceed asdescribed before to look up the
Goldsteinfactors,andcomputethecirculation.
Thisgeneralprocedureformedthebasisforhightechshipandairplanepropeller
design inthepre-computer era. However, itwassoonrecognized thatship propellers
generallyoperateinradiallynonuniforminow,andthatcavitationandvibrationcon-
siderationsmay forcethedesignertodepartfromtheBetz/Lerbsoptimumradialdistri-
butionofcirculation. WhiletheapplicationoftheGoldsteinfactor(andtheassociated
assumption that the total induced velocity is normal to
i
) is invalid in this case, a
pragmaticsolution was to applythem anyway and hope for the best. The development
oftheoreticaldesignmethodsbasedonthisassumptionislargelydueto VanManen
57
andtoEckhartandMorgan
58
. Thesemethodsprovedto bequiteeectiveandrepre-
sentedavastimprovementovertheexistingalternativeofselectingpropellerssolelyfrom
experimentallybasedpropellercharts.
Ofcourse, manual design procedures based on theGoldsteinfactors have been largely
replaced by computerizedmethodswhichcantreatthegeneralcaseofarbitraryinow
velocity andnon-optimumcirculationdistributions. TheGoldsteinfactorandtheKramer
chart, however, remainsasveryconciseandinstructive pictureofthekeyelementsof
propellerow.
AsanexampleoftheuseoftheKramerdiagram,supposethatwewantedaquick
estimateoftheideal(inviscid)eciencyofavebladedpropelleroperatinginuniform
ow atan advance coecient of J
s
= 1.0with a thrust coecient ofC
T
= 1.0. We
55
The thrust coecient in the absenceof viscousdragforces
56
Kramer, K.N.,Induzierte Wirkungsgrade von Best-Luftschrauben endlicher
Blattzahl,Luftfahrtforshung, July 1938, Vol 15. An English Translation appears in NACA TM
884,January1939
57
VanManen,J.D.,FundamentalsofShipResistanceandPropulsion,InternationalShipbuilding
Progress,(4),1957
58
Eckhart,M.K.and Morgan,W.B,APropeller Design Method, Trans. SNAME, Vol 63,1955.
173
saw from gure 93 that the eciency should be very nearly 70%. We can enter the
Kramerdiagramatadvancecoecientof =J/ = 1.0/ = 0.318. Wenextslideup
thediagonallineuntilwereachthehorizontallinecorrespondingtoZ = 5blades,and
then move vertically upward until wereach thehorizontalline at athrust coecient of
C
T
= 1.0. Ifwediditright,wewillberightonthe70%eciencyline. Wecanalso
conrm theactuator diskeciency byentering theKramer diagram at the left end of the
scale,correspondingtoanadvancecoecient0f= 0.001 0. MovinguptoC
T
= 1.0,
weseethattheeciencyis= 0.83,which againagreeswithgure93.
Figure98: KramerDiagramforIdealPropellerEciency
174

3.6 LiftingLineTheoryforArbitraryCirculationDistributions
3.6.1 LerbsInduction FactorMethod
Amajoradvancecamein1952,withtheintroduction byLerbs
59
ofarigorouslifting
linetheorywhichcouldhandlethegeneralcaseofapropellerinradiallynon-uniform
inowwithanarbitrarydistributionofcirculation. WewillnotdevelopLerbstheoryin
completedetailhere,butwilldescribetheessentialideas.
SinceBetzcriterion,equation228,andLerbscriterion229fortheoptimumcirculation
distributionareexpressed interms of thedistribution of hydrodynamic pitch angle,
i
(r),
theproblemwhich will besolved is todetermine thedistribution ofcirculation, (r),for
a prescribed
i
(r). The latter may either be the optimum distribution derived from
228 or 229, or may be asuitably modieddistribution of the hydrodynamic pitch angle
designedtounloadthetipor hubforcavitationreasons. Forthemoment,wewillnot
worryabouthowwecametospecify
i
(r).
Wealso know the ship speed, V
S
, thepropellerrotational speed , andthedistribution
ofeectiveaxialinowvelocityV
a
(r).. Fromthevelocity diagramingure89weknow
that
rtan
i
=rtan+ u

u
t

tan
i
(248)
a
whichcan bewrittenas,
u

a
u

t
r
V
S

V
S
tan
i
=
V
S
(tan
i
tan) (249)
butsinceV
a
/r=tan wecaneliminaterandobtainthenalresult
u

a

u

t
tan
i
=
V
a
tan
i
1 (250)
V
S
V
S
V
S
tan
Theright handsideofequation250isknown. Theinducedvelocitiesonthelefthand
sidearerelatedtothecirculation(r) by theCauchy principalvalueintegralsoverthe
radiusgivenin201. Wethereforehaveasingularintegralequationforthecirculation.
Lerbs developed a procedure for solving this equation by extending the method developed
59
Lerbs, H.W., Moderately Loaded Propellers with a Finite Number of Blades and an Arbitrary
Distribution of Circulation,Trans. SNAME, vol.60,1952.
175

byGlauert
60
foralifting-linewing. ThekeystepsofGlauertstheoryiscontainedin
theAppendix.
Therststep wastotransformthe physicalradialcoordinate,r toanangularcoor-
dinate, r, which takes on a value ofzero at the hub and atthetip,

R+ r
h
2r

r =cos
1
(251)
Rr
h
Figure99: GeometricrepresentationoftheGlauertcosinetransformation.
This cosine transformation is shown graphically in Figure 99. Next, the radial
distribution of circulation is approximated by a nite sine series in the transformed
variable r,
G( r)
( r)
=
J
a
j
sin(jr) (252)
2RV
S
j=1
withtheunknowncoecients, a
j
, tobedetermined. Themaximumnumberofterms
60
Glauert,H.,TheElementsofAerofoilandAirscrewTheory,CambridgeUniversityPress,1959
176
retainedinthesineseries,J,can beselected to provide thedesired level ofaccuracy. A
sineseries was chosen toensure that thecirculation would vanish at the hub and tip for
anychoice ofthecoecientsa
j
. It isclearon physicalgroundsthatthecirculation must
gotozeroatthetip(exceptinthecaseofaductedpropellerwithvanishingtipgap)but
this is not necessarily true at the hub.Lerbs argued that the close proximity of the blades
atthe hub juncture would tend to equalize the pressures onboth sides of the blades
and thus reduce the circulation tozero. In fact, subsequent calculations with lifting-
surfaceandpanelmethodshavedemonstratedthatthecirculationmay benon-zeroat
the hub. However,since the local behavior at the hub haslittleeect on overall propeller
performance,Lerbsidealizationofthehublesspropellerisnotaseriousshortcoming
ofthetheory. Wewillconsidertheeectofthe hubonpropeller performanceinmore
detailinasubsequentsection.
IntroducingthedenitionoftheinductionfactorsfromEquation208,andconverting
tothenon-dimensionalcirculation,G,wecanwriteequation201as
u

=
1

i
a
(r
v
, r
c
)

j
J
=1
ja
j
cos(jr
v
)
dr
v
a
V
S
1 r
h
/R 0 cos r
v
cos r
c
u

=
1

i
t
(r
v
, r
c
)

J
j=1
ja
j
cos(jr
v
)
d (253)
t
V
S
1 r
h
/R 0 cos r
v
cos r
c
r
v
Thisis beginningtolookmorelikeEquation130inGlauertsliftinglinetheory,ex-
ceptforthepresenceoftheinductionfactors.InordertomakeuseofGlauertsintegral,
Equation131,Lerbsnextapproximatedtheinductionfactorsatagivenvalueofr
c
by
a Fouriercosineseriesin r
v
. Thisis permissible,sincethesingularity has beenremoved
fromtheinductionfactors,sothatwhatremainsisawellbehavedfunction,asshown
in Figure 92. The coecients in the Fourier expansion were obtained by evaluating
Equations203-206atJ positionsovertheradius,andusingstandardharmonicanalysis
techniques toobtain the Fourier coecients. The numerator in Equation 253 thencon-
sistsofsumsofcosineswhosecoecientsarecombinationsoftheunknowncirculation
seriesandtheknowncoecientsofthe Fourierexpansionoftheinductionfactors. This
can then be integrated analytically term by term using Equation 130. The resulting
setofalgebraicequationsmaythen besolvedfortheunknowncirculationcoecients,
a
j
. Whilethissounds(andis)intricate,theresultingalgorithmisextremelyrobustand
computationallyecient. Theresultingprocedureisagood exampleofasemi-analytic
method,inwhich thesingularintegralsovertheradiusarecarriedoutanalytically,while
thecirculationandtheinductionfactorsareapproximated by serieswith anite number
ofterms.
Oncethecoecientsa
j
ofthecirculationarefound,thecirculation,thrustandtorque
can becomputed. If thethrust (ortorque) is not equal to thedesiredvalue, the input
177
tan
i
(r) can beadjusteduntilthedesired value isobtained. Thematching ofthe spec-
ied thrust is accomplished by rst computing two initial guesses, and then applying
Newtonsmethod to nd the multiplier to be applied to the rst approximation to tan
i
which produces in thecorrectthrust (towithin a speciedtolerance). Generallythree
orfourNewtoniterationsaresucienttoachieve athrustcoecienttowithin apracti-
caltolerance. Anexampleofthisprocedurewill beshowntogetherwithcorresponding
resultsobtained bya vortexlatticemethod.
3.7 Propeller VortexLatticeLiftingLineTheory
178

A vortex lattice solution to the propeller lifting line problem is conceptually very
similartothesolutionoftheplanarliftinglineproblem. Thespanofthekeybladeis
dividedintoM panelsextendingfromr=r
h
tor=R,asshowningure88.Theradial
distribution of (bound)circulation, (r), isapproximated byaset ofM vortexelements
ofconstantstrength
m
extendingfromr
v
(m) to r
v
(m+ 1). Adiscretetrailing(free)
vortex line is shed at each of the panel boundaries, with a strength equal to thedierence
instrengths oftheadjacent bound vortices. However,aswithplanarliftinglinetheory, it
ismoreconvenienttoconsiderthatthevortexsystemisbuiltfromasetofM horseshoe
elements,each consisting of a bound vortexsegment ofstrength
m
and twofree vortex
linesofstrength
m
. Butinaddition,each horseshoeelementactuallyrepresentsaset
ofZ identicalelementsofequalstrength,oneoriginatingfromeach blade.
Each set ofhorseshoe vortexelements,ofunitstrength, inducesanaxialandtangential
velocityat a speciedcontrol point,r
c
(n) onthe keyblade. Thecontributionofthe two
free vortices can be found from equations 202 and 203, or if one wishes to solve the
problemofaninnitebladedpropeller,fromequations206and207. Thecontribution
ofthe boundvortexelementofthesetofhorseshoevorticesiszero,providedthatthe
liftingline isradialandthattheblades haveuniformangularspacing. Clearlythe bound
vortexonthe keyliftinglineinduceszero velocity anywherealongthatline.Bound vortex
elementsonanotherblademayinduceavelocityonthekeyblade, buttheirsummed
eectwillcancelduetosymmetry. Thetotalinducedvelocityatcontrol pointr
c
(n) is
therefore,
M
u
a

(r
c
(n)) =
m
u
a
(n, m)
m=1
M
u

(r
c
(n)) =
m
u
t
(n, m) (254)
t
m=1
whereu
a
(n, m) and u
t
(n, m) arethehorseshoeinuencefunctions.
Aswithplanarfoilliftinglinetheory, it is best tousecosinespacing for the vortex
andcontrol points. Deningh= 0.5(Rr
h
) and =/(2M),theircoordinatesare,
r
v
(m) = r
h
+ h[1 cos(2(m1))]
r
c
(n) = r
h
+ h[1 cos(2n1))]
(255)
Thereare a number of possibleapproaches tosolving for thecirculation.Themethod
usedinPVLissimilartoLerbsmethodexcept,ofcourse,thatavortexlatticeisused
179

insteadofLerbsextensionofGlauertssineseriesmethod. Theobjectivewill betond
the circulation distribution (and hence, forces) for a propeller with a specied thrust
coecient,C
T
.
Figure101showsasampleinputdataleforPVL.Therstentryisthenumber
of panels, which in this case is M = 32. The last part of the data le consists of a
tabulationofthechord/diameterratio,c/D,theviscousdragcoecient,C
d
,theaxial
inow velocity,V
a
,andthetangentialinowvelocity,V
t
, at auser-speciedset ofnon-
dimensionalradii,r/R startingwiththehubandendingwiththetip. Thenumberof
inputradiiisarbitrary, andisgiveninthe5thentryinthetableinthiscase11.With
thisinformation,equation255isusedtocomputethevortexlatticegrid,andtheitems
tabulated at the inputradiiare interpolated to the vortexlattice grid byacubic spline
procedure.
Inparticular,theinowangle atany radiuscan becomputedfrom,
tan(r) =
V
+
a
(
V
r
t
)
(r)
(256)
r
Js
wheretheadvancecoecient,J
s
isgiveninthe7thentryintheinputleinthiscase
J
s
= 0.8. A plot offor thisparticularcase may be found in the lower right of gure 102.
Thenextstepistospecifyatrialvalueof
i
(r). Thiscaneitherbetheoptimum
distribution, as given by equation 228 or 229, or may bealtered tounload the hub or tip,
whichwewilldiscusslater. Ineithercase,tan
i
isobtainedfromtan intermsofan
unknownmultiplicativeconstant. Weknowthatforanoptimumpropellerinuniform,
inviscidow,themultiplierissimplytheinverseoftheidealeciency. Sincethethrust
coecient is specied, wecaneasilycomputetheeciency ofanactuatordisk. Wewill
therefore use 90%of theeciency of theactuator disk as a rst guess inestablishing
i
(r).
WewillthensolveforthecirculationandforC
T
usingthisvalue,andtheniteratively
adjustthemultiplieruntilthedesiredC
T
isachieved. This,ofcourse,isdoneinternally
intheprogramusingNewtonsmethod, whichconvergesveryrapidly. Themaximum
numberofiterationsisgivenasthesecondentryintheleinthiscase10. Usually,
convergencetosixsignicant guresisachievedinthreeorfouriterations.Theconverged
valuefor
i
forthisexampleisalsoshowninthe bottomrightofgure102.
With
i
(r) known,theinducedvelocitiesmustsatisfythesamekinematicrelationship
asisequation250inLerbsmethod,
u

a

u

t
tan
i
=
V
a
tan
i
1 (257)
V
S
V
S
V
S
tan
and bycombiningequation 254 withequation 257, weobtain thefollowing set of simul-
taneousequationsfortheunknowncirculationvalues,
n
,
180

M

V
a
(n) tan
i
(n)
m=1
[ u
a
(n, m) u
t
(n, m) tan
i
(n)]
m
=
V
S
tan(n)
1 n= 1, . . . M (258)
Oncethesimultaneousequationsfor aresolved,equations195and196can beused
toevaluatetheforces,includingtheeectsofviscousdrag(basedonthedragcoecient
valuestabulatedintheinput). Theintegralsovertheradiusfortheforcesaresimply
replacedby sumsover thevortex lattice. Thesolved valueofthecirculationandthe
inducedvelocitiesforthisexampleisshownontheleftsideofgure102.
Figure100: Illustration oftheimage of a2-D pointvortex in acircle of radiusr
h
. The
vortexisatradiusr,whiletheimageisatradiusr
i
=r
h
2
/r.Ifthetwovorticeshaveequal
and oppositestrengths,thenormal (radial) component ofthevelocityinduced bythe pair
ofvortices cancels at all points on the circle of radiusr
h
. .
3.7.1 Hubeects
So far, we have ignored thepresence of the hub, except for the fact that the lifting lines
representingthebladesstartatthehubradius,r
h
.Thisisequivalenttoassumingthat
theblades have afree tip at bothends,andthatsomemagicalmagneticforce ispresent
to keepthemrotating about theshaftaxis. Inthiscase, thecirculation goes tozero at
both the hub and the tip, as shown in gure 102. Inreality,a propeller hub is present,
181
Sample PVL input file: Propeller in idealized 1/7 power law wake field
32 : NUMBER OF VORTEX PANELS OVER THE RADIUS
10 : MAXIMUM ITERATIONS IN WAKE ALIGNMENT
0 : HUB IMAGE FLAG: 1=YES, 0=NO
0.25 : HUB VORTEX RADIUS/HUB RADIUS
11 : NUMBER OF INPUT RADII
5 : NUMBER OF BLADES
0.8 : ADVANCE COEFFICIENT, J, BASED ON SHIP SPEED
1.000 : DESIRED THRUST COEFFICIENT, CT
0.000 : HUB UNLOADING FACTOR: 0.0=OPTIMUM (NO UNLOADING)
0.000 : TIP UNLOADING FACTOR 1.0=REDUCED LOADING
1.000 : CRP SWIRL CANCELLATION FACTOR: 1.0=NO CANCELATION
r/R c/D Cd Va/Vs Vt/Vs
0.20000 0.17400 0.00800 0.71969 0.00000
0.25000 0.19700 0.00800 0.74300 0.00000
0.30000 0.22900 0.00800 0.76260 0.00000
0.40000 0.27500 0.00800 0.79460 0.00000
0.50000 0.31200 0.00800 0.82034 0.00000
0.60000 0.33700 0.00800 0.84198 0.00000
0.70000 0.34700 0.00800 0.86073 0.00000
0.80000 0.33400 0.00800 0.87731 0.00000
0.90000 0.28000 0.00800 0.89218 0.00000
0.95000 0.24000 0.00800 0.89911 0.00000
1.00000 0.00200 0.00800 0.90572 0.00000
Figure101: SampleinputdataleforPVL.Thislewasusedtogeneratetheresults
plottedingure102
182
whichiseitheranintegralpartofthesolidpropellercastingorisaseparateunitthat
containsthecontrollablepitchmechanism. Inthelattercase,each bladeangeis bolted
to aspindle in the hub.
Whilethehubgeometrycanbeincludedinadetailedpanelmethodcalculationof
theowaroundthepropeller,asimplerapproximationisappropriateatthelifting-line
designstage.It isknownthat in two-dimensionalow,the velocity eld of a pointvortex
situatedatradiusr outsideacircleofradiusr
h
can befound by superimposingtheow
eldofavortexinanunboundeduidwiththevelocity eldofanimagevortexsituated
withinthecircle. Asshowningure100,iftheimagevortexislocatedataradius
r
i
=
r
h
2
(259)
r
the total velocity normal to the circle can easily be shown to bezero.As the radius of the
hubisincreased,weobtainthelimitingcaseofavortexoutsideofaninnitewall,andin
thiscasetheimagevortexisobviouslyatthesamedistanceinsidethewall. Ontheother
hand, asthevortexradiusisincreased(orthehubradiusisdecreased), equation259
showsthattheimagevortexmovestotheaxis.
No simple image exists for a helical vortex, but numerical calculations show that
equation259worksamazinglywell. Resultsobtainedwiththesimpleimagemodelagree
almostexactlywithcomplete potentialowsolutionpublished byTachmindji
61
. Itis
thereforeasimplemattertosupplement eachhelicalhorseshoevortexwithitsimage
inside the hub. If the m

th horseshoe vortex extends from r

v
(m) to r
v
(m +1), its
imagewillextendfromr
h
2
/r
v
(m) to r
h
2
/r
v
(m+ 1). Thevelocityinducedbytheimage
horseshoe vortex can be combined with the inuence function for the real external
horseshoeelementinequation258sothatnoadditionalunknownsareintroducedinto
theproblem.
Figure103showstheeectofincludingthehubimage. Thecirculationnowhasa
nite valueatthe hub,andthederivativeof (r) withrespectto r iszeroatthe hub.
Thecirculationovertheouterpartofthebladeisessentiallythesameasforthezero
hubcase.
Thepresence ofnitecirculation at the hub introduces anewproblem. Thiscircula-
tion isessentiallytransferred intotheinside ofthe hubviathe hubimage,but it must
eventually beshedintotheowatthedownstreamterminationofthe hub
62
Thisforms
aconcentratedhubvortex,which isveryvisiblewhenitscorepressureissucientlylow
61
A.J. Tachmindji,Potential Problem of the Optimum Propeller with a Finite Number of Blades,
Journalof Ship Research,December, 1958
62
The only exception is in an experimental facility where the propeller is driven by a downstream
shaft. Most early propeller tunnels drove the propeller from downstream, so that the presence of hub
vortexcavitation wasnot observed.
183
for it to cavitate. This can be shown from thesequence of threephotographs reproduced
ingures104,105and106. Allthreephotographsweretakenatthesametunnelow
speedandpressure. Therstphotographshows apropeller operating by itself,whilethe
secondphotographshowsapre-swirlstator(designedtooperatewiththepropeller)op-
erating by itself. Finally,thethirdphotographshowsthepropellerandstator operating
together.
Thepresence of aconcentrated hub vortexalsoeectsthenetpropellerthrust. The
low pressure region in (and near) the core of the vortex acts on the aft side of the
hub,thuscreatingadrag. Itwouldseemsimpletointegratethepressureeldarounda
concentrated vortex over the projected downstreamface of the hub inorder toobtain the
drag. However, theunfortunateresult is that the drag isinnite! Toobtain a physically
realisticresult,the hub vortex must bemodeledasonewith aniteviscouscoreradius.
UsingtheclassicalRankinevortexmodel,inwhichthevelocity eldwithinthecoreis
simply one ofsolid bodyrotation, andoutside ofthecore is a potential vortex, M-H
Wang
63
computed the hub vortex drag as afunction of theratio of vortexcoreradius
to hubradius. Wangfoundthattheresultingpressureforceactingonthe downstream
endofthe hubcould beapproximatedasfollows,
F
h
=
16

ln
r
r
h
o
+3

(Z
o
)
2
(260)
wherer
o
is thecoreradius of the hub vortex and
o
isthecirculationatthebladeroot.
The hub vortexcoreradius must beestimated,butfortunately,duetothelogarithmic
natureofequation260,itsprecisevalueisnotcritical.
The optimumdistribution ofcirculation, ifbased on aninnitely long hub (andhence
zero hub drag), willtherefore not produce the mostecient propeller.This isparticularly
trueforpropellerswithlargehub/diameterratios. Itisthereforebettertoreducethe
circulationatthe hub,asshowningure107.
However, the problemof hub drag can be overcome with a multiple blade rowpropulsor
eitheracounterrotatingpropelleroracombinationofarotorandasetofxedstator
blades. Ineithercase,thenetcirculationatthe hubcan bedesignedto bezero,sothat
therewill beno hub vortexandno hub vortexdrag.
Figure 109 shows an optimum counter rotating propeller, in which the tangential
inducedvelocitiesfromeach bladerowexactlycancel. Theoptimumcirculationdistrib-
ution shows a large value at the hub. There is no hub drag inthiscase,sincethere is no
netcirculationatthe hub.
63
M-H Wang, Hub eects in Propeller Design and Analysis,PhD Thesis, Department of Ocean
Engineering,MIT, May1985
184
3.7.2 The VortexLatticeActuatorDisk
WecancreateanactuatordiskinPVLby settingthe hubradiusto a verysmall(but
non-zerovalue)andspecifyinganinnitenumberofblades,withzeroviscousdrag.This
isaccomplished by specifying anite number ofbladesexceeding20. Figure110shows
asampleresultfor athrustcoecient ofC
T
= 1.0. Thenumberofbladesisenteredas
25, so that the outputcirculation perblade isstillnite. However, theinduced velocity
inuence coecientsare for aninnitebladed propeller, not a 25bladed propeller. Tan-
gential velocity cancelationhas been specied. Thecomputedaxialinduced velocity and
circulationisconstantovertheradius,andagreesexactlywiththeactuatordiskvalue.
Inthiscase,theresultisindependentoftheinputadvancecoecient.
Wecanalsorunthecodewithouttangentialvelocity cancelation,butwithaninnite
number of blades and zero drag. The results then depend on advance coecient, as
shown in gure 93. Finally, an innite bladed propeller can be run with a specied
viscouslift/dragratio,andtheseresultsareshownonthesamegure. Notethatthere
isnowanoptimumadvancecoecient,which dependsonthelift/dragratio.
3.7.3 HubandTipUnloading
The user can easily adjust the degree ofhub and tip unloading in the PVLcode by
entering values inthe 9

th and10

th positionsintheinputle. Theunloadingfactors,
H
r
andH
t
aredenedasthefractionalamountthatthedierence betweentheoptimum
valuesoftan
i
andtan arereduced. Forexample,ifH
r
=0,tan
i
tan atthe hub
isretainedatitsoptimumvaluefromBetz/Lerbscriterion. IfH
r
= 1.0,tan
i
tan
atthehubissettozero,andthevaluesuptothemidspanofthebladeareblended
parabolicallytotheoptimumvalue. Thesameprocedureappliestothetip. Sincethis
adjustment ismade beforetan
i
isscaledtoproducethedesiredthrust,thenaloutcome
at the hub and tip may not be exactly as specied. However,one can easily adjustH
r
and
H
t
toproducethedesiredresult. Notefromthelowerrighthandplotsingures107and
108 howthedierence between
i
and issmoothlyreduced,comparedtotheoptimum
caseshowningure103.
185
Figure 102: Liftinglineresults fora5bladedpropellerobtainedwiththe PVLcode.
Inthisexample,there isnohub,and a Lerbsoptimumcirculationdistributionhas been
selected. C
T
= 1.0 C
P
= 1.3432 K
T
= 0.2513 K
Q
= 0.0430 V
a
/V
s
= 0.8526 =
63.47%.
186
Figure103: Liftingline resultsfor a 5bladed propellerobtainedwiththePVLcode. In
thisexample,thereisanimagehub,andaLerbsoptimumcirculationdistributionhas
beenselected. C
T
= 1.0 C
P
= 1.3744 K
T
= 0.2513 K
Q
= 0.0440 V
a
/V
s
= 0.8526 =
62.03%. Theeciency has been reducedslightly due to hubvortex drag.
187
Figure104: Propelleroperatingalone.Asubstantialcavitatinghubvortexisevident.
Figure105: Pre-swirlstatoroperatingalone. Asubstantialhubvortexisagainevident.
Thesignofthisvortexisoppositefromtheoneshowningure104.
188
Figure106: Propellerandstator operatingtogether.Thehubvortexhas been canceled.
189
Figure 107: Liftinglineresults fora5bladedpropellerobtainedwiththe PVLcode.
Inthisexample, thereisanimagehub, andaLerbsoptimumcirculationdistribution
hasbeenmodiedtounloadthetip,usingHT = 1.0. C
T
= 1.0 C
P
= 1.4391 K
T
=
0.2513 K
Q
= 0.0461 V
a
/V
s
= 0.8526 = 59.24%. Theeciencyhasbeenfurther
reducedduetotipunloading. Notetheverydierentshapeoftheaxialinducedvelocity
distribution.
190
Figure108: Liftingline resultsfor a 5bladed propellerobtainedwiththePVLcode. In
thisexample,thereisanimagehub,andaLerbsoptimumcirculationdistributionhas
beenmodiedto unloadthe hub, using HR = 1.0. C
T
= 1.0 C
P
= 1.3442 K
T
=
0.2513 K
Q
= 0.0431 V
a
/V
s
= 0.8526 = 63.43%.Theeciencyhasactuallyimproved,
sincethereduced hubloadingreducesthehubvortexdrag.
191
Figure 109: Liftinglineresults fora5bladedpropellerobtainedwiththe PVLcode.
Inthisexample, thereisanimagehub, andaLerbsoptimumcirculationdistribution
hasbeenspecied. Inaddition,theidealizedcounterrotatingpropelleroptionhasbeen
selected,sothattherearenotangentialinduced velocities.Notethatthecirculationnear
the hub has been greatly increased. C
T
= 1.0 C
P
= 1.2532 K
T
= 0.2513 K
Q
=
0.0401 V
a
/V
s
= 0.8526 = 68.03%.
192
Figure110: Liftinglineresultsfora25bladedpropellerobtainedwiththePVLcode.
Theinducedvelocitiescorrespondtothoseofaninnitebladedpropeller,andthetan-
gentialinducedvelocitieshavebeencanceled. Viscousdraghasbeensettozero. This,
therefore,correspondstoanactuatordisk. C
T
= 1.0 C
P
= 1.2071 K
T
= 0.2513 K
Q
=
0.0386 V
a
/V
s
= 1.000 = 82.84%. Thecirculation is a constant, with avalue of
G = 0.005093andtheaxialinducedvelocityisu

/V
a
= 0.20711,whichagreesexactly
a
withactuatordisktheory.
193
4 COMPUTER CODE LISTINGS
194
PROGRAM MAPSL
!-------------------------------------------------------------------------------
! This program uses the Karman-Trefftz mapping function to compute the flow
! field and pressure distribution around a foil section. The geometry of the
! foil is determined by specifying the position of the center of the circle,
! (XC,YC) and the trailing edge angle (TAU). The circulation is obtained by
! satisfying the Kutta condition, but the user may override this value to
! explore the effect on the local trailing edge flow. The output consists
! of two TECPLOT formatted files. MAPSL.PLT is used to generate a field plot
! showing the streamlines and pressure field. PLOTCP.PLT is used to generate
! an x-y plot of the presure distribution on the foil surface.
! WRITTEN BY: Justin E. Kerwin December, 2000
!-------------------------------------------------------------------------------
! Declare the variables-----------------------------------------------------
IMPLICIT NONE
CHARACTER*1 YESNO
REAL, PARAMETER :: PI=3.1415927E00, RAD=1.7453293E-02
REAL :: X,Y,THETA,RSQ,U,V,R,XC,YC,ALPHA_DEG,ALPHA,RC,BETA,RCSQ,G,GAMMA, &
LAMBDA,TAU_DEG,XI,ETA,X_CIRCLE,Y_CIRCLE,FL2,U_FOIL,V_FOIL, &
THETA1,THETA2,DTHETA,RMAX,DR,CP,XLE,XTE,CHORD,CL
REAL, ALLOCATABLE, DIMENSION(:) :: XI_FOIL,CP_FOIL
COMPLEX :: Z,ZP1,ZM1,ZETA,W_CIRCLE,DZETA_DZ,W_FOIL
INTEGER :: N,M,NTHETA,NRADIAL
! Input the variables defining foil geometry and angle of attack -----------
WRITE(*,(A)) ENTER XC,YC ! Position of center of circle---------
READ(*,*) XC,YC ! Note: XC must be <= 0 -------------
WRITE(*,(A)) ENTER TRAILING EDGE ANGLE (DEG)
READ(*,*) TAU_DEG
LAMBDA=2.0-TAU_DEG/180.0 ! The exponent in the K-T mapping function --
FL2=4.0*LAMBDA**2 ! Pre-comute a mapping function constant-----
WRITE(*,(A)) ENTER ALPHA (DEG) ! Note: Enter 180 to get a circle---
READ(*,*) ALPHA_DEG ! User selected angle of attack.-------------
ALPHA=ALPHA_DEG*RAD
RCSQ=(1.0-XC)**2+YC**2 ! Radius of circle passing through (1,0) ----
RC=SQRT(RCSQ)
BETA=ATAN(YC/(1.0-XC))
! Set circulation from Kutta condition, or override as desired -------------
GAMMA=-4.0*PI*RC*SIN(BETA+ALPHA) ! Circulation required to satisfy Kutta--
WRITE(*,( GAMMA FOR KUTTA CONDITION=,F12.7)) GAMMA
WRITE(*,(A)) ENTER "Y" TO ACCEPT, "N" TO OVERRIDE
READ(*,(A)) YESNO
IF(YESNO/=Y.AND.YESNO/=y) THEN
WRITE(*,(A)) ENTER DESIRED VALUE
READ(*,*) GAMMA
END IF
G=GAMMA/(2.0*PI)
!----- Define grid parameters --------------------------------------------------
195
WRITE(*,(A)) ENTER THETA1,THETA2(DEG) ! Angular extent of grid---
READ(*,*) THETA1,THETA2
WRITE(*,(A)) ENTER DTHETA(DEG) ! Angular resolution-------
READ(*,*) DTHETA
NTHETA=NINT((THETA2+THETA1)/DTHETA)+1 ! Number of radial grid lines
WRITE(*,(A)) ENTER MAXIMUM GRID RADIUS, RMAX ! Radial extent of
READ(*,*) RMAX ! the grid-----------
WRITE(*,(A)) ENTER RADIAL SPACING, DR
READ(*,*) DR ! Radial resolution -------
NRADIAL=NINT((RMAX-RC)/DR)+1
! Alocate the arrays and open output files ---------------------------------
ALLOCATE(XI_FOIL(NTHETA),CP_FOIL(NTHETA)) ! Arrays for foil Cp ------
OPEN(1,FILE=MAPSL.PLT,STATUS=UNKNOWN,FORM=FORMATTED) ! Output files
WRITE(1,(A)) VARIABLES = X,Y,U,V,CP
OPEN(2,FILE=PLOTCP.PLT,STATUS=UNKNOWN,FORM=FORMATTED)
WRITE(2,(A)) VARIABLES = X,CP
!-----Generate velocity and presure field --------------------------------------
WRITE(1,( ZONE I=,I4, J=,I4)) NTHETA,NRADIAL
DO M=1,NRADIAL
R=RC+REAL(M-1)*DR ! Grid radial coordinate in the Z (circle) plane---
RSQ=R**2
DO N=1,NTHETA
THETA=RAD*(THETA1-REAL(N-1)*DTHETA) ! Grid angular coordinate---
X=XC+R*COS(THETA) ! Convert to cartesian------
Y=YC+R*SIN(THETA)
!-----------Compute the velocity field around the circle------------------------
U=COS(ALPHA)-RCSQ*COS(2.0*THETA-ALPHA)/RSQ-G*SIN(THETA)/R
V=SIN(ALPHA)-RCSQ*SIN(2.0*THETA-ALPHA)/RSQ+G*COS(THETA)/R
!-----------Express the field point position and velocity in complex form-------
Z=CMPLX(X,Y) ! Form complex number (X+iY)--------
W_CIRCLE=CMPLX(U,-V) ! Complex velocity in Z plane-------
!-----------Use the Karman-Trefftz transformation to map points to ZETA plane---
ZP1=(Z+1.0)**LAMBDA
ZM1=(Z-1.0)**LAMBDA
ZETA=LAMBDA*(ZP1+ZM1)/(ZP1-ZM1)
XI=REAL(ZETA) ! Transformed X coordinate----------
ETA=AIMAG(ZETA) ! Transformed Y coordinate ---------
!-----------Compute the derivative of the mapping function; Transform velocities
DZETA_DZ=FL2*((Z-1.0)**(LAMBDA-1.0)*(Z+1.0)**(LAMBDA-1.0))/ &
(ZP1-ZM1)**2
IF(ABS(DZETA_DZ)>0.0001) THEN
W_FOIL=W_CIRCLE/DZETA_DZ ! Complex velocity in foil plane-------
196
ELSE
W_FOIL=0.0 ! Avoids divide by zero at trailing edge
END IF
U_FOIL=REAL(W_FOIL)
V_FOIL=-AIMAG(W_FOIL)
!-----------Compute the presure coefficient, CP, and output to plotting file----
CP=ABS(W_FOIL)**2-1.0
WRITE(1,(5F12.5)) XI,ETA,U_FOIL,V_FOIL,CP
IF(M==1) THEN ! Save presure on foil surface for later use---
XI_FOIL(N)=XI
CP_FOIL(N)=CP
END IF
END DO
END DO
!-----Scale the chordwise coordinate to (0,1) and output foil pressure dist.----
XLE=10.0
XTE=-10.0
DO N=1,NTHETA
XLE=MIN(XLE,XI_FOIL(N))
XTE=MAX(XTE,XI_FOIL(N))
END DO
CHORD=XTE-XLE
DO N=1,NTHETA
XI_FOIL(N)=(XI_FOIL(N)-XLE)/CHORD
END DO
WRITE(2,(2F10.5)) (XI_FOIL(N),CP_FOIL(N),N=1,NTHETA)
CLOSE(2)
CL=-2.0*GAMMA/CHORD
WRITE(*,( LIFT COEFFICIENT=,F8.4)) CL
!-----Generate coordinates of circle -------------------------------------------
WRITE(1,(A)) ZONE I= 361, J=1
DO M=1,361
THETA=-BETA+REAL(M-1)*RAD
X_CIRCLE=XC+RC*COS(THETA)
Y_CIRCLE=YC+RC*SIN(THETA)
Z=CMPLX(X_CIRCLE,Y_CIRCLE)
ZP1=(Z+1.0)**LAMBDA
ZM1=(Z-1.0)**LAMBDA
ZETA=LAMBDA*(ZP1+ZM1)/(ZP1-ZM1)
XI=REAL(ZETA)
ETA=AIMAG(ZETA)
WRITE(1,(5F10.5)) XI,ETA,0.0,0.0,0.0
END DO
CLOSE(1)
STOP
197
END PROGRAM MAPSL
198
! Last change: JEK 3 Mar 2001 3:50 pm
!-------------------------------------------------------------------------------
! Vortex/Source Lattice Program for 2-D Foils!
!-------------------------------------------------------------------------------
PROGRAM VLM2D
USE DUCKMOD ! Module for spline interpolation and sim.eqn. solver -----
USE GEOM ! Module for mean line and thickness form geometry library-
!-------------------------------------------------------------------------------
! Declare the variables
IMPLICIT NONE
REAL, DIMENSION(:), ALLOCATABLE :: XV,XC,XT,YT,DYDX,DX,B,GAMMA,G,GEXACT, &
F,UT,UTVP,CPU,CPL
REAL, DIMENSION(:,:), ALLOCATABLE :: A,CUBIC
REAL :: DELC,TOP,CL,ALPHA,SUMG,CLNUM,TOC,RLE,QU,QL,FLH
REAL, PARAMETER :: PI=3.141592653589793E00, HALF=0.5E00, RAD=PI/180.0, &
ZERO=0.0E00, ONE=1.0E00, TWO=2.0E00, ESL=0.0, ESR=0.0
INTEGER :: MC,N,M,IERR,MLTYPE,NTHICK
INTEGER, DIMENSION(:), ALLOCATABLE :: IPIVOT
!-----Compute vortex and control point positions and weight functions-----------
WRITE(*,(A)) Enter number of panels
READ(*,*) MC ! MC can be as large as your computer memory permits --------
ALLOCATE (XV(MC),XC(MC),A(MC,MC),DX(MC),B(MC),GAMMA(MC),IPIVOT(MC),G(MC), &
GEXACT(MC),F(MC),XT(MC+1),YT(MC+1),DYDX(MC),UT(MC),UTVP(MC), &
CUBIC(MC-1,5),CPU(MC+1),CPL(MC+1))
DELC=PI/REAL(MC)
DO N=1,MC
XV(N)=HALF*(ONE-COS((N-HALF)*DELC)) ! Vortex/source positions --------
XC(N)=HALF*(ONE-COS(N*DELC)) ! Control point positions --------
DX(N)=PI*SQRT(XV(N)*(ONE-XV(N)))/FLOAT(MC) ! Vortex weight factors ----
END DO
!-----Compute influence coefficient matrix A(N,M) and invert -------------------
TOP=ONE/(TWO*PI)
DO N=1,MC
DO M=1,MC
A(N,M)=TOP/(XV(M)-XC(N))
END DO
END DO
CALL FACTOR(A,IPIVOT,IERR) ! FACTOR does LU decomposition of matrix A ---
WRITE(*,( RETURN CODE FROM SUBROUTINE FACTOR (1 MEANS OK)=,I3)) IERR
!-----Solve for GAMMA(X) FOR NACA A=.8 or PARABOLIC meal line-------------------
WRITE(*,(A)) Enter ideal lift coefficient....
READ(*,*) CL
WRITE(*,(A)) Enter Alpha-Alpha(ideal) (deg).....
READ(*,*) ALPHA
WRITE(*,(A)) ENTER 1 FOR NACA A=.8, 2 FOR PARABOLA
READ(*,*) MLTYPE
199
IF (MLTYPE==1) CALL AEIGHT(XV,XC,B,F,GEXACT)
IF (MLTYPE==2) CALL PARABL(XV,XC,B,F,GEXACT)
DO N=1,MC
B(N)=CL*B(N)-ALPHA*RAD ! RHS of equation. Slope of mean line -alpha ---
F(N)=CL*F(N) ! Mean line camber at desired ideal CL ---------
END DO
CALL SUBST(A,B,GAMMA,IPIVOT) ! Obtain solution of A matrix with RHS B ----
!-----Sum circulation over chord and convert to vortex sheet strength-----------
! Note: These vortex strengths are the NEGATIVE of a strict RH rule notation
SUMG=ZERO
DO N=1,MC
SUMG=SUMG+GAMMA(N)
G(N)=GAMMA(N)/DX(N)
END DO
CLNUM=TWO*SUMG
WRITE(*,( Computed total lift coefficient=,F8.4)) CLNUM
OPEN(1,FILE=VLM2D.PLT,STATUS=UNKNOWN,FORM=FORMATTED)
WRITE(1,(A)) ZONE T=" Camber"
WRITE(1,(2F10.5)) (XC(N),F(N),N=1,MC)
WRITE(1,(A)) ZONE T=" -Point Vortex Strengths"
WRITE(1,(2F10.5)) (XV(N),GAMMA(N),N=1,MC)
WRITE(1,(A)) ZONE T=" -Vortex Sheet Strengths"
WRITE(1,(2F10.5)) (XV(N),G(N),N=1,MC)
!-----Velocity due to thickness------------------------------------------------
XT(1)=0.0
DO N=1,MC
XT(N+1)=XC(N)
END DO
WRITE(*,(A)) Enter 1 for NACA66, 2 for Karman-Trefftz...
READ(*,*) NTHICK
IF(NTHICK==1)Then
WRITE(*,(A)) Enter thickness/chord ratio...
READ(*,*) TOC
CALL NACA66(TOC,RLE,XT,YT,XV,DYDX)
ELSE
CALL KARGEO(TOC,RLE,XT,YT,XV,DYDX)
END IF
WRITE(*,( T/C=,F10.5, R_L/C=,F10.5)) TOC,RLE
WRITE(1,(A)) ZONE T=" Thickness"
WRITE(1,(2F10.5)) (XT(N),YT(N),N=1,MC+1)
!-----Compute the velocity induced by the thickness sources --------------------
DO N=1,MC
UT(N)=ZERO
DO M=1,MC
UT(N)=UT(N)+TOP*(YT(M+1)-YT(M))/(XC(N)-XV(M))
END DO
200
END DO
WRITE(1,(A)) ZONE T=" UT at control points"
WRITE(1,(2F10.5)) (XC(N),UT(N),N=1,MC)
!-----Interpolate thickness velocity to vortex points --------------------------
CALL UGLYDK(1,1,XC,UT,0.0,0.0,CUBIC)
CALL EVALDK(XV,UTVP,CUBIC)
WRITE(1,(A)) ZONE T=" UT at vortex points"
WRITE(1,(2F10.5)) (XV(N),UTVP(N),N=1,MC)
!-----Compute surface velocities:First get value at leading edge----------------
QU=ALPHA*RAD*SQRT(TWO/RLE)
CPU(1)=QU**2-ONE
CPL(1)=CPU(1)
!-----Next get remaining values over the chord----------------------------------
DO N=1,MC
IF(DYDX(N)>0.0) THEN
FLH=1.0/SQRT(1.0+DYDX(N)**2) ! Scherer/Riegels modified Lighthill--
ELSE
FLH=1.0 ! No leading edge correction beyond point of max thickness
END IF
QU=(ONE+UT(N)+HALF*G(N))*FLH
CPU(N+1)=QU**2-ONE
QL=(ONE+UT(N)-HALF*G(N))*FLH
CPL(N+1)=QL**2-ONE
END DO
WRITE(1,(A)) ZONE T= -Cp upper surface"
WRITE(1,(2F10.5)) ZERO,CPU(1)
WRITE(1,(2F10.5)) (XV(N),CPU(N+1),N=1,MC)
WRITE(1,(A)) ZONE T=" -Cp lower surface"
WRITE(1,(2F10.5)) ZERO,CPL(1)
WRITE(1,(2F10.5)) (XV(N),CPL(N+1),N=1,MC)
CLOSE(1)
STOP
END PROGRAM VLM2D
201
! Last change: JEK 3 Mar 2001 1:42 pm
MODULE GEOM
INTERFACE AEIGHT
SUBROUTINE AEIGHT(XV,XC,B,F,GEXACT)
REAL, DIMENSION(:), INTENT(IN) :: XV,XC
REAL, DIMENSION(:), INTENT(OUT) :: B,F,GEXACT
END SUBROUTINE AEIGHT
END INTERFACE
INTERFACE PARABL
SUBROUTINE PARABL(XV,XC,B,F,GEXACT)
REAL, DIMENSION(:), INTENT(IN) :: XV,XC
REAL, DIMENSION(:), INTENT(OUT) :: B,F,GEXACT
END SUBROUTINE PARABL
END INTERFACE
INTERFACE NACA66
SUBROUTINE NACA66(THK,RLE,XT,YT,XV,DYDX)
REAL, INTENT(IN) :: THK ! Section thickness/chord ratio
REAL, INTENT(OUT) :: RLE ! Leading edge radius/chord
REAL, DIMENSION(:), INTENT(IN) :: XT,XV ! Chordwise positions
REAL, DIMENSION(:), INTENT(OUT) :: YT,DYDX ! thickness and slope
END SUBROUTINE NACA66
END INTERFACE
INTERFACE KARGEO
SUBROUTINE KARGEO(THK,RLE,XT,YT,XV,DYDX)
REAL, INTENT(OUT) :: THK ! Section thickness/chord ratio
REAL, INTENT(OUT) :: RLE ! Leading edge radius/chord
REAL, DIMENSION(:), INTENT(IN) :: XT,XV ! Chordwise positions
REAL, DIMENSION(:), INTENT(OUT) :: YT,DYDX ! Thickness and slope
END SUBROUTINE KARGEO
END INTERFACE
END MODULE GEOM
!-------------------------------------------------------------------------------
SUBROUTINE AEIGHT(XV,XC,B,F,GEXACT)
IMPLICIT NONE
REAL, DIMENSION(:), INTENT(IN) :: XV,XC
REAL, DIMENSION(:), INTENT(OUT) :: B,F,GEXACT
REAL, PARAMETER :: PI=3.141592653589793E00, ONE=1.0E00, TWO=2.0E00, &
A=0.8E00, HALF=0.5E00, QUART=0.25E00
REAL :: G,H,AIDEAL,CONST,G1,C1,CA,P,R,S,T
INTEGER :: N,MC
MC=SIZE(XV)
G=-(A**2*(HALF*ALOG(A)-QUART)+QUART)/(ONE-A)
202
H=(HALF*(ONE-A)**2*ALOG(ONE-A)-QUART*(ONE-A)**2)/(ONE-A)+G
AIDEAL=-H/(TWO*PI*(A+ONE))
CONST=TWO*PI*(A+ONE)
G1=ONE/(ONE+A)
DO N=1,MC
C1=MAX(ONE-XV(N),1.0E-06)
CA=A-XV(N)
IF(ABS(CA)<1.0E-06) CA=CA+1.0E-05
P=HALF*CA**2*LOG(ABS(CA))-HALF*C1**2*LOG(C1)+QUART*(C1**2-CA**2)
F(N)=(P/(ONE-A)-XV(N)*LOG(XV(N))+G-H*XV(N))/CONST+C1*AIDEAL
IF(XV(N)<=A) THEN
GEXACT(N)=G1
ELSE
GEXACT(N)=G1*(ONE-XV(N))/(ONE-A)
END IF
END DO
DO N=1,MC
C1=MAX(ONE-XC(N),1.0E-06)
CA=A-XC(N)
IF(ABS(CA)<1.0E-06) CA=CA+1.0E-05
R=-(A-XC(N))*LOG(ABS(CA))-HALF*CA+C1*LOG(C1)+HALF*C1
S=-HALF*C1+HALF*CA
T=-LOG(XC(N))-ONE-H
B(N)=((R+S)/(ONE-A)+T)/CONST-AIDEAL
END DO
RETURN
END
SUBROUTINE PARABL(XV,XC,B,F,GEXACT)
IMPLICIT NONE
REAL, DIMENSION(:), INTENT(IN) :: XV,XC
REAL, DIMENSION(:), INTENT(OUT) :: B,F,GEXACT
REAL, PARAMETER :: PI=3.141592653589793E00
INTEGER :: N,MC
MC=SIZE(XV)
DO N=1,MC
B(N)=(1.0 -2.0*XC(N))/PI
F(N)=XV(N)*(1.0-XV(N))/PI
GEXACT(N)=4.0*SQRT(XV(N)*(1.0-XV(N)))/PI
END DO
RETURN
END
SUBROUTINE NACA66(THK,RLE,XT,YT,XV,DYDX)
USE DUCKMOD
203
IMPLICIT NONE
REAL, INTENT(IN) :: THK ! Section thickness/chord ratio
REAL, INTENT(OUT) :: RLE ! Leading edge radius/chord
REAL, DIMENSION(:), INTENT(IN) :: XT,XV ! Arrays of chordwise positions
REAL, DIMENSION(:), INTENT(OUT) :: YT,DYDX ! Section thickness and slope
!----- Database of thickness (T66) versus position (PC) for the DTMB MOD NACA66-
REAL, DIMENSION(18), PARAMETER :: PC=(/0.0000,0.0100,0.0250,0.0500,
0.1000,0.2000,0.3000,0.4000,
&
&
0.4500,0.5000,0.6000,0.7000, &
0.8000,0.9000,0.9500,0.9750, &
0.9900,1.0000/)
REAL, DIMENSION(18), PARAMETER :: T66=(/0.0000,0.1870,0.2932,0.4132,
0.5814,0.8000,0.9274,0.9904,
&
&
1.0000,0.9917,0.9256,0.7934, &
0.5950,0.3306,0.1736,0.0888, &
0.0360,0.0000/)
REAL, PARAMETER :: RLE_CONST=0.448
REAL :: TRLE,XSQ,YSPLN,DY,D2YDX
REAL, DIMENSION(18) :: PSQ
REAL, DIMENSION(17,5) :: CUBIC
INTEGER :: NT,N
NT=SIZE(XT)
RLE=RLE_CONST*THK**2
!-----Square root stretched coordinate for spline interpolation ----------------
PSQ=SQRT(PC)
!-----Get spline coefficients of thickness with LE radius prescribed -----------
TRLE=2.0*SQRT(2.0*RLE_CONST)
CALL UGLYDK(2,1,PSQ,T66,TRLE,0.0,CUBIC)
!-----Evaluate spline for thickness at XT positions-----------------------------
DO N=1,NT
XSQ=SQRT(XT(N))
CALL EVALDK(XSQ,YSPLN,CUBIC)
YT(N)=THK*YSPLN
END DO
!-----Evaluate derivative of spline to get DY/DX = 0.5*DT/DX at XV(N)-----------
DO N=1,NT-1
XSQ=SQRT(XV(N))
CALL DRIVDK(XSQ,DY,D2YDX,CUBIC) ! Derivative Dy/D(xsq) ----------------
DYDX(N)=DY*THK/(4.0*XSQ) ! Transform to Dy/Dx ----------------
END DO
RETURN
END SUBROUTINE NACA66
SUBROUTINE KARGEO(THK,RLE,XT,YT,XV,DYDX)
204
!-------------------------------------------------------------------------------
! Evaluates thickness and its derivative for a Karman Trefftz thickness
! form from a MAPSL.PLT file.
!-------------------------------------------------------------------------------
USE DUCKMOD ! Access module for spline interpolation subroutines
IMPLICIT NONE
! Declaring the arguments---------------------------------------------------
REAL, INTENT(OUT) :: THK ! Section thickness/chord ratio
REAL, INTENT(OUT) :: RLE ! Leading edge radius/chord
REAL, DIMENSION(:), INTENT(IN) :: XT,XV ! Arrays of chordwise positions
REAL, DIMENSION(:), INTENT(OUT) :: YT,DYDX ! Section thickness and slope
! Declaring and alocating the local variables-------------------------------
CHARACTER*64 :: FNAME,LINES
REAL :: XSQ,YSPLN,XLE,CHORD,DY,D2YDX
REAL, DIMENSION(:), ALLOCATABLE :: PSQ,XIN,YIN,YTK
REAL, DIMENSION(:,:), ALLOCATABLE :: CUBIC
INTEGER :: NT,N
INTEGER, PARAMETER:: NPI=181 ! No.of points from TE to LE from MAPSL -----
NT=SIZE(XT)
ALLOCATE(PSQ(NPI),XIN(NPI),YIN(NPI),YTK(NPI),CUBIC(NPI-1,5))
! Read in MAPSL.PLT file to obtain coordinates of the K-T section-----------
WRITE(*,(A)) ENTER NAME OF MAPSL OUTPUT FILE....
READ(*,(A)) FNAME
OPEN(2,FILE=FNAME,STATUS=OLD,FORM=FORMATTED)
READ(2,(A)) LINES ! Flush through first to lines to get past first----
READ(2,(A)) LINES ! occurence of the character string "ZONE" ---------
DO
READ(2,(A)) LINES
IF(LINES(2:2)==Z) EXIT ! Look for second occurence of "ZONE" -------
END DO
DO N=1,NPI ! Here is the data that we want -----------------------------
READ(2,*) XIN(N),YIN(N)
END DO
CLOSE(2)
XLE=XIN(NPI) ! Find the x coordinates of the LE and TE to scale the data-
CHORD=XIN(1)-XIN(NPI)
THK=0.0
DO N=1,NPI
PSQ(N)=SQRT((XIN(NPI-N+1)-XLE)/CHORD) ! Trans. coordinate for spline -
YTK(N)=2.0*YIN(NPI-N+1)/CHORD ! Scaled to give t/c=2y/c --------------
THK=MAX(THK,YTK(N)) ! Searching for the maximum thickness/chord ratio--
END DO
CALL UGLYDK(1,1,PSQ,YTK,0.0,0.0,CUBIC) ! Get spline coefficients ------
!-----Evaluate spline for total thickness, YT(N), at XT(N) ---------------------
DO N=1,NT
205
XSQ=SQRT(XT(N))
CALL EVALDK(XSQ,YSPLN,CUBIC)
YT(N)=YSPLN
END DO
!-----Evaluate derivative of spline to get DY/DX = 0.5*DT/DX at XV(N)-----------
DO N=1,NT-1
XSQ=SQRT(XV(N))
CALL DRIVDK(XSQ,DY,D2YDX,CUBIC) ! Derivative Dy/D(xsq) ----------------
DYDX(N)=DY/(4.0*XSQ) ! Transform to Dy/Dx ----------------
END DO
CALL DRIVDK(0.0,DY,D2YDX,CUBIC) ! Derivative Dy/D(xsq) at leading edge----
RLE=0.125*DY**2 ! Leading edge radius obtained from square of deriv. --
RETURN
END SUBROUTINE KARGEO
206
! Last change: JEK 20 Mar 2001 1:39 pm
PROGRAM HVLL
!-------------------------------------------------------------------------------
! HYDROFOIL VORTEX LATTICE LIFTING LINE PROGRAM
! Written by: Justin E. Kerwin, March 2001
!-------------------------------------------------------------------------------
USE DUCKMOD ! Contains interface to simultaneous equations routines ------
IMPLICIT NONE
REAL, PARAMETER :: PI=3.1415927E00
REAL, DIMENSION(:), ALLOCATABLE :: A,YV,YC,GAMMA,WE,WN,GN
REAL, DIMENSION(:,:), ALLOCATABLE :: WNM
INTEGER, DIMENSION(:), ALLOCATABLE :: IPIVOT
REAL :: DELTIP,DEL,ANG,S1,AJSN,FZE,FXE,FRE,FZN,FXN,FRN,FZGN,FXGN,FRGN, &
PERLG,PERLW,PERDG,PERDW,PEREG,PEREW
INTEGER :: MT,KCOS,KMP,NG,N,J,IERR,M
!-------------------------- Inputs ---------------------------------------------
WRITE(*,(A)) Enter number of panels ....
READ(*,*) MT
WRITE(*,(A)) Enter 1 for const. spacing, 2 for cosine...
READ(*,*) KCOS
IF(KCOS==1) THEN
WRITE(*,(A)) Enter tip vortex inset/panel width...
READ(*,*) DELTIP
KMP=2
END IF
IF(KCOS==2) THEN
WRITE(*,(A)) Enter 1 for cos control pts, 2 for midpt...
READ(*,*) KMP
END IF
WRITE(*,(A)) Enter number of Gamma coefficients...
READ(*,*) NG
ALLOCATE(A(NG))
WRITE(*,(A)) Enter Gamma coefficients....
READ(*,*) A
!----------------- Generate the vortex lattice ---------------------------------
ALLOCATE(YV(MT+1),YC(MT),GAMMA(MT),WE(MT),WNM(MT,MT),WN(MT),GN(MT), &
IPIVOT(MT))
IF(KCOS==1) DEL=1.0/(REAL(MT)+2.0*DELTIP) ! Panel spacing----------------
IF(KCOS==2) DEL=PI/REAL(2*MT)
DO N=1,MT+1 ! Get free vortex positions, YV(N) -------------------------
IF(KCOS==1) YV(N)=-0.5+(DELTIP+REAL(N-1))*DEL
IF(KCOS==2) YV(N)=-0.5*COS(2.0*REAL(N-1)*DEL)
END DO
DO N=1,MT ! Get control point positions, YC(N) -------------------------
IF(KCOS==1.OR.KMP==2) THEN
YC(N)=0.5*(YV(N)+YV(N+1))
ANG=ACOS(-2.0*YC(N))
ELSE
207
ANG=REAL(2*N-1)*DEL
YC(N)=-0.5*COS(ANG)
END IF
! Compute exact circulation, GAMMA and exact downwash, WE, from Glauert coeffs--
S1=SIN(ANG)
GAMMA(N)=2.0*A(1)*S1
WE(N)=-A(1)
IF(NG>=2) THEN
DO J=2,NG
AJSN=A(J)*SIN(REAL(J)*ANG)
GAMMA(N)=GAMMA(N)+2.0*AJSN
WE(N)=WE(N)-REAL(J)*AJSN/S1
END DO
END IF
END DO
! Compute the exact lift, drag and drag/lift**2 from Glauert -------------------
FZE=0.5*PI*A(1)
FXE=0.5*PI*A(1)**2
IF(NG>=2) THEN
DO J=2,NG
FXE=FXE+0.5*PI*REAL(J)*A(J)**2
END DO
END IF
FRE=FXE/FZE**2
WRITE(*,( FZE,FXE,FRE... ,3F10.5)) FZE,FXE,FRE
!-----Compute vortex lattice horseshoe influence functions WNM(N,M) and the ----
!-----resulting downwash, WN(N) based on the exact circulation, GAMMA ----------
FZN=0.0
FXN=0.0
DO N=1,MT
WN(N)=0.0
DO M=1,MT
WNM(N,M)=1.0/(4.0*PI*(YV(M)-YC(N)))-1.0/(4.0*PI*(YV(M+1)-YC(N)))
WN(N)=WN(N)+GAMMA(M)*WNM(N,M)
END DO
FZN=FZN+GAMMA(N)*(YV(N+1)-YV(N)) ! Numerical lift force ---------
FXN=FXN-WN(N)*GAMMA(N)*(YV(N+1)-YV(N)) ! Numerical drag force ---------
END DO
FRN=FXN/FZN**2
!-----Solve for the vortex lattice circulation, GN, based on the exact downwash,
!----- WE, using the simultaneous equation solver FACTOR and SUBST -------------
CALL FACTOR(WNM,IPIVOT,IERR)
WRITE(*,( RETURN CODE FROM FACTOR (1=OK)=,I4)) IERR
CALL SUBST(WNM,WE,GN,IPIVOT)
WRITE(*,(6F10.5)) (YV(N),YC(N),GAMMA(N),WE(N),WN(N),GN(N),N=1,MT)
WRITE(*,(F10.5)) YV(MT+1)
208
!-----Get the forces based on the vortex lattice circulation and exact downwash-
FZGN=0.0
FXGN=0.0
DO N=1,MT
FZGN=FZGN+GN(N)*(YV(N+1)-YV(N))
FXGN=FXGN-WE(N)*GN(N)*(YV(N+1)-YV(N))
END DO
FRGN=FXGN/FZGN**2
!-----Compute the percent errors in forces obtained by both methods ------------
PERLG=100.0*(FZGN-FZE)/FZE
PERLW=100.0*(FZN-FZE)/FZE
PERDG=100.0*(FXGN-FXE)/FXE
PERDW=100.0*(FXN-FXE)/FXE
PEREG=100.0*(FRGN-FRE)/FRE
PEREW=100.0*(FRN-FRE)/FRE
!-----Output TECPLOT file HVLL.PLT containing spanwise distributions for plotting
OPEN(1,FILE=HVLL.PLT,STATUS=UNKNOWN,FORM=FORMATTED)
WRITE(1,(A)) VARIABLES=YC,W-EXACT,W-NUM,G-EXACT,G-NUM
WRITE(1,(5F10.5)) (YC(N),WE(N),WN(N),GAMMA(N),GN(N),N=1,MT)
CLOSE(1)
!-----Output a LATEX file containing the percent errors in forces. Note that this
!----- file appends current results to all previous data stored in HVLL.TEX ----
OPEN(2,FILE=HVLL.TEX,STATUS=UNKNOWN,FORM=FORMATTED,POSITION=APPEND)
WRITE(2,(I8,6F8.1)) MT,PERLW,PERDW,PEREW,PERLG,PERDG,PEREG
CLOSE(2)
STOP
END PROGRAM
209
! Last change: JEK 30 Apr 2001 5:48 pm
PROGRAM PVL
!-------------------------------------------------------------------------------
! Propeller Vortex Lattice Lifting Line Code
! COPYRIGHT (C) 2001 JUSTIN E. KERWIN -------------------------------------------
USE PVLMOD
USE DUCKMOD
!------------------------- Declare the Variables -------------------------------
IMPLICIT NONE
CHARACTER*36 :: FNAME,LABEL
CHARACTER*72 :: TITLE
INTEGER :: MT,NX,ITER,NBLADE,N,M,KTRY,IERR,IHUB
REAL :: KT,KQ,DEL,HRR,RCWG,RM,DTANB,EDISK,ADVCO,CTDES,HR,HT,CRP,WAKE, &
CQ,CP,EFFY,HRF,CTH,RHV
REAL, PARAMETER :: PI=3.1415927E00, TOL=0.000005, R2D=57.29578E00
DOUBLE PRECISION :: TANBIW,RCW,RVW,UAIF,UTIF
REAL, DIMENSION(:), ALLOCATABLE :: XR,XCHD,XCD,XVA,XVT,XRC,RV,TANBV,RC, &
TANBC,VAV,VTV,VAC,VTC,TANBIV,TANBIC, &
UAW,UTW,B,G,UASTAR,UTSTAR,T,CT, &
TANBXV,TANBXC,VBAV,VBAC,CD,CDC
REAL, DIMENSION(:,:), ALLOCATABLE :: CHCUB,CDCUB,VACUB,VTCUB,UAHIF, &
UTHIF,A
!------------------ Start reading the input data -------------------------------
WRITE(*,(A)) ENTER INPUT FILE NAME.....
READ(*,(A)) FNAME
OPEN(2,FILE=FNAME,STATUS=OLD,FORM=FORMATTED)
READ(2,(A)) TITLE ! Title describing data file
READ(2,*) MT ! Number of vortex lattice panels
READ(2,*) ITER ! Number of iterations to align wake
READ(2,*) IHUB ! Hub image flag. IHUB=0 : No hub image, IHUB=1 : Image hub
READ(2,*) RHV ! Hub vortex radius/Hub radius. Only used if IHUB=1
READ(2,*) NX ! Number of radii used to specify the input data
!------------- Allocate all the arrays before reading rest of input-------------
ALLOCATE ( XR(NX),XCHD(NX),XCD(NX),XVA(NX),XVT(NX),XRC(NX) )
ALLOCATE ( CHCUB(NX-1,5),CDCUB(NX-1,5),VACUB(NX-1,5),VTCUB(NX-1,5) )
ALLOCATE ( RV(MT+1),TANBV(MT+1),RC(MT),TANBC(MT),VAV(MT+1),VTV(MT+1), &
VAC(MT),VTC(MT),TANBIV(MT+1),TANBIC(MT),UAW(MT+1),UTW(MT+1), &
UAHIF(MT,MT),UTHIF(MT,MT),A(MT,MT),B(MT),G(MT),UASTAR(MT), &
UTSTAR(MT),T(ITER),CT(ITER),TANBXV(MT+1),TANBXC(MT), &
VBAV(MT+1),VBAC(MT),CD(MT),CDC(MT) )
!----------------All arrays allocated. read in rest of input data --------------
READ(2,*) NBLADE ! Number of blades
READ(2,*) ADVCO ! Advance coefficient based on ship speed
READ(2,*) CTDES ! Desires thrust coefficient CT (based on ship speed)
READ(2,*) HR ! Unloading ratio at hub
READ(2,*) HT ! Unloading ratio at tip
READ(2,*) CRP ! Tangential velocity cancellation factor
READ(2,(A)) LABEL ! Alphanumeric label for output
210
!-------------------------------------------------------------------------------
! XR=Input radii r/R, XCHD=Input chord length c/D, XCD=Input viscous drag
! coefficient, Cd or Lift/Drag ratio, XVA,XVT=Input axial and tangential
! velocities, Va/Vs, V_t/Vs
!-------------------------------------------------------------------------------
READ(2,*) (XR(N),XCHD(N),XCD(N),XVA(N),XVT(N),N=1,NX)
CLOSE(2)
!-----Compute volumetric mean inflow velocity ratio VA/VS ----------------------
WAKE=VOLWK(XR,XVA)
!-----Spline chord over radius using square root stretched coordinates----------
XRC(:)=1.0-SQRT(1.0-XR(:))
CALL UGLYDK(0,0,XRC,XCHD,0.0,0.0,CHCUB)
!-----Spline Drag Coefficient Cd, Inflow Vx, Vt using radial coordinate directly
CALL UGLYDK(0,0,XR,XCD,0.0,0.0,CDCUB)
CALL UGLYDK(0,0,XR,XVA,0.0,0.0,VACUB)
CALL UGLYDK(0,0,XR,XVT,0.0,0.0,VTCUB)
!-----Compute cosine spaced vortex radii and get Va,Vt,tanB,Vt*tanB/Va----------
DEL=PI/(2.0*REAL(MT))
HRR=0.5*(XR(NX)-XR(1))
DO M=1,MT+1
RV(M)=XR(1)+HRR*(1.0-COS(REAL(2*(M-1))*DEL))
CALL EVALDK(RV(M),VAV(M),VACUB)
CALL EVALDK(RV(M),VTV(M),VTCUB)
TANBV(M)=VAV(M)/((PI*RV(M)/ADVCO)+VTV(M))
VBAV(M)=VTV(M)*TANBV(M)/VAV(M)
END DO
!-----Cosine spaced control point radii: Evaluate c/D,Va,Vt,tanB,Cd,Vt*tanB/Va -
DO M=1,MT
RC(M)=XR(1)+HRR*(1.0-COS(REAL(2*M-1)*DEL))
RCWG=1.0-SQRT(1.0-RC(M))
CALL EVALDK(RCWG,CDC(M),CHCUB)
CALL EVALDK(RC(M),VAC(M),VACUB)
CALL EVALDK(RC(M),VTC(M),VTCUB)
TANBC(M)=VAC(M)/((PI*RC(M)/ADVCO)+VTC(M))
CALL EVALDK(RC(M),CD(M),CDCUB)
VBAC(M)=VTC(M)*TANBC(M)/VAC(M)
END DO
!-----First estimate of tanBi based on 90 percent of actuator disk efficiency --
EDISK=1.8/(1.0+SQRT(1.0+CTDES/WAKE**2))
TANBXV(:)=TANBV(:)*SQRT(WAKE/(VAV(:)-VBAV(:)))/EDISK ! Lerbs optimum-----
TANBXC(:)=TANBC(:)*SQRT(WAKE/(VAC(:)-VBAC(:)))/EDISK
!-----Unload hub and tip as specified by input HR and HT -----------------------
RM=0.5*(XR(1)+XR(NX)) ! Mid-radius. Unloading is quadratic, starting here
DO M=1,MT+1
211
IF(RV(M).LT.RM) THEN
HRF=HR
ELSE
HRF=HT
END IF
DTANB=HRF*(TANBXV(M)-TANBV(M))*((RV(M)-RM)/(XR(1)-RM))**2
TANBXV(M)=TANBXV(M)-DTANB
END DO
DO M=1,MT
IF(RC(M).LT.RM) THEN
HRF=HR
ELSE
HRF=HT
END IF
DTANB=HRF*(TANBXC(M)-TANBC(M))*((RC(M)-RM)/(XR(1)-RM))**2
TANBXC(M)=TANBXC(M)-DTANB
END DO
!-------------------------------------------------------------------------------
! Iterations to scale tanBi to get desired value of thrust coefficient
!-------------------------------------------------------------------------------
DO KTRY=1,ITER
IF(KTRY.EQ.1) THEN
T(KTRY)=1.0 ! T(KTRY) is the scale factor to apply to tanBi
ELSE IF(KTRY.EQ.2) THEN
T(KTRY)=1.0+(CTDES-CT(1))/(5.0*CTDES) ! Guess for second iteration
ELSE IF(KTRY.GT.2) THEN
T(KTRY)=T(KTRY-1)+(T(KTRY-1)-T(KTRY-2))*(CTDES-CT(KTRY-1))/ &
(CT(KTRY-1)-CT(KTRY-2)) ! Secant method for remaining iters
END IF
TANBIV(:)=T(KTRY)*TANBXV(:) ! Scale tanBi at the vortex radii
TANBIC(:)=T(KTRY)*TANBXC(:) ! Scale tanbi at the control points
!-------------------------------------------------------------------------------
! Compute axial and tangential horseshoe influence coefficients !
!-------------------------------------------------------------------------------
DO M=1,MT
RCW=RC(M)
DO N=1,MT+1
!--------------Induction of trailing vortices shed at RV(N)---------------------
TANBIW=TANBIV(N)
RVW=RV(N)
CALL WRENCH(NBLADE,TANBIW,RCW,RVW,UAIF,UTIF)
UAW(N)=-UAIF/(2.0*(RC(M)-RV(N)))
UTIF=UTIF*CRP ! Note if CRP=0, the tangential velocity is zero--
UTW(N)=UTIF/(2.0*(RC(M)-RV(N)))
!--------------Induction of corresponding hub-image trailing vortices (if any)--
212
IF(IHUB/=0) THEN
RVW=XR(1)**2/RV(N)
TANBIW=TANBIV(1)*RV(1)/RVW
CALL WRENCH(NBLADE,TANBIW,RCW,RVW,UAIF,UTIF)
UAW(N)=UAW(N)+UAIF/(2.0*(RC(M)-RVW))
UTIF=UTIF*CRP
UTW(N)=UTW(N)-UTIF/(2.0*(RC(M)-RVW))
END IF
END DO
!-----------Final step in building influence functions--------------------------
DO N=1,MT
UAHIF(M,N)=UAW(N+1)-UAW(N)
UTHIF(M,N)=UTW(N+1)-UTW(N)
END DO
END DO
!-------------------------------------------------------------------------------
! Solve simultaneous equations for circulation strengths G(M) !
!-------------------------------------------------------------------------------
DO M=1,MT
B(M)=VAC(M)*((TANBIC(M)/TANBC(M))-1.0) ! Right-hand side
DO N=1,MT
A(M,N)=UAHIF(M,N)-UTHIF(M,N)*TANBIC(M) ! Coefficient matrix
END DO
END DO
CALL SIMEQN(A,B,G,IERR) ! Simultaneous equation solver
IF(IERR/=0) EXIT ! Error return for singular matrix
!-------------------------------------------------------------------------------
! Evaluate the induced velocities from the circulation GM) !
!-------------------------------------------------------------------------------
DO M=1,MT
UASTAR(M)=0.0
UTSTAR(M)=0.0
DO N=1,MT
UASTAR(M)=UASTAR(M)+G(N)*UAHIF(M,N)
UTSTAR(M)=UTSTAR(M)+G(N)*UTHIF(M,N)
END DO
END DO
!-------------------------------------------------------------------------------
! Compute the forces and test if Ct has converged to desired value !
!-------------------------------------------------------------------------------
CALL FORCES(NBLADE,MT,ADVCO,WAKE,RV,RC,TANBC,UASTAR,UTSTAR,VAC, &
CDC,CD,G,CT(KTRY),CQ,CP,KT,KQ,EFFY,RHV,CTH,IHUB)
WRITE(*,(I5, CT=,F10.5, DESIRED VALUE=,F10.5)) KTRY, &
CT(KTRY),CTDES
IF(ABS(CT(KTRY)-CTDES)<TOL) EXIT
213
END DO
!-----Stop run if matrix is sigular---------------------------------------------
IF(IERR/=0) THEN
WRITE(*,(A)) MATRIX SINGULAR. RUN TERMINATED.....
STOP
ELSE
!-------------------------------------------------------------------------------
! Output results to Tecplot file !
!-------------------------------------------------------------------------------
WRITE(*,(// EFFICIENCY =,F8.4)) EFFY
WRITE(*,( Kt, Kq,F8.4,F8.5)) KT,KQ
WRITE(*,( HUB DRAG COEFFICIENT Cth=,F8.4)) CTH
OPEN(1,FILE=APLOT.PLT,STATUS=UNKNOWN,FORM=FORMATTED)
WRITE(1,(A)) VARIABLES="R","G","VA","VT","UA","UT","BETA","BETAI",&
&"CDC","CD"
WRITE(1,( TEXT X=0.5, Y=0.50, T="
CT(KTRY)
WRITE(1,( TEXT X=0.5, Y=0.46, T="
CP
Ct=,F8.4, "))
Cp=,F8.4, "))
&
&
WRITE(1,( TEXT X=0.5, Y=0.42, T="
KT
Kt=,F8.4, ")) &
WRITE(1,( TEXT X=0.5, Y=0.38, T="
KQ
Kq=,F8.4, ")) &
WRITE(1,( TEXT X=0.5, Y=0.34, T=" Va/Vs=,F8.4, "))
WAKE
&
WRITE(1,( TEXT X=0.5, Y=0.30, T="
EFFY
E=,F8.4, ")) &
WRITE(1,( TEXT X=0.5, Y=0.26, T=",A, " ))
TITLE
&
WRITE(1,(F10.5,F10.6,4F10.5,2F10.3,2F10.5)) (RC(M),G(M),VAC(M),
VTC(M),UASTAR(M),UTSTAR(M),R2D*ATAN(TANBC(M)),
R2D*ATAN(TANBIC(M)),CDC(M),CD(M),M=1,MT)
CLOSE(1)
END IF
&
&
STOP
END PROGRAM PVL
214
! Last change: JE 24 Apr 2001 11:41 am
!*************** COPYRIGHT (C) 2001 JUSTIN E. KERWIN ******************
SUBROUTINE FORCES(NBLADE,MCP,ADVCO,WAKE,RV,RC,TANBC,UASTAR,UTSTAR, &
VA,CHORD,CD,G,CT,CQ,CP,KT,KQ,EFFY,RHV,CTH,IHUB)
IMPLICIT NONE
!--------------------- Declare the arguments -----------------------------------
INTEGER, INTENT(IN) :: NBLADE,MCP,IHUB
REAL, INTENT(IN) :: ADVCO,WAKE,RHV
REAL, DIMENSION(:), INTENT(IN) :: RV,RC,TANBC,UASTAR,UTSTAR,VA,CHORD,CD,G
REAL, INTENT(OUT) :: CT,CQ,CP,KT,KQ,EFFY,CTH
!--------------------- Declare the local variables -----------------------------
REAL, PARAMETER :: PI=3.1415927E00, TWO=2.0E00, FOUR=4.0E00, EIGHT=8.0E00
REAL :: DR,VSTAR,VTSTAR,VASTAR,VSTRSQ,DVISC,FKJ
INTEGER :: M
LOGICAL :: CD_LD
CD_LD=.TRUE. ! Default: Input CD interpreted as viscous drag coefficient
IF(CD(1)>1.0) CD_LD=.FALSE. ! CD(1)>1 signals that input is L/D ----------
CT=0.0
CQ=0.0
DO M=1,MCP
DR=RV(M+1)-RV(M)
VTSTAR=VA(M)/TANBC(M)+UTSTAR(M)
VASTAR=VA(M)+UASTAR(M)
VSTRSQ=VTSTAR**2+VASTAR**2
VSTAR=SQRT(VSTRSQ)
IF(CD_LD) THEN ! Interpret CD as viscous drag coefficient, Cd---------
DVISC=(VSTRSQ*CHORD(M)*CD(M))/(TWO*PI)
ELSE ! Interpret CD as the lift/drag ratio L/D -------------
FKJ=VSTAR*G(M)
DVISC=FKJ/CD(M)
END IF
CT=CT+(VTSTAR*G(M)-DVISC*VASTAR/VSTAR)*DR
CQ=CQ+(VASTAR*G(M)+DVISC*VTSTAR/VSTAR)*RC(M)*DR
END DO
IF(IHUB/=0) THEN ! Add hub vortex drag if hub image is present ----------
CTH=0.5*(LOG(1.0/RHV)+3.0)*(REAL(NBLADE)*G(1))**2
ELSE
CTH=0.0
END IF
CT=CT*FOUR*REAL(NBLADE)-CTH
CQ=CQ*TWO*REAL(NBLADE)
CP=CQ*TWO*PI/ADVCO
KT=CT*ADVCO**2*PI/EIGHT
KQ=CQ*ADVCO**2*PI/EIGHT
EFFY=CT*WAKE/CP
RETURN
215
END
REAL FUNCTION VOLWK(XR,XVA)
USE DUCKMOD
IMPLICIT NONE
REAL :: YDX
INTEGER :: NX,N
REAL, DIMENSION(:), INTENT(IN) :: XR,XVA
REAL, DIMENSION(:), ALLOCATABLE :: Y
REAL, DIMENSION(:,:), ALLOCATABLE :: VWCUB
NX=SIZE(XR)
ALLOCATE ( Y(NX),VWCUB(NX-1,5) )
Y(:)=XR(:)*XVA(:)
CALL UGLYDK(0,0,XR,Y,0.0,0.0,VWCUB)
CALL INTDK1(XR(1),XR(NX),YDX,VWCUB)
VOLWK=2.0*YDX/(1.0-XR(1)**2)
DEALLOCATE (Y,VWCUB)
RETURN
END FUNCTION VOLWK
216
! Last change: JEK
MODULE PVLMOD
INTERFACE
28 Mar 99 5:04 pm
SUBROUTINE SIMEQN(A,B,X,IERR)
REAL, DIMENSION(:,:), INTENT(IN) :: A ! Coefficient matrix
REAL, DIMENSION(:), INTENT(IN) :: B ! Right hand side vector
REAL, DIMENSION(:), INTENT(OUT) :: X ! Solution vector
INTEGER, INTENT(OUT) :: IERR ! Error flag
END SUBROUTINE SIMEQN
SUBROUTINE WRENCH(NB,TANB,RC,RV,UA,UT)
INTEGER, INTENT(IN) :: NB
DOUBLE PRECISION, INTENT(IN) :: TANB,RC,RV
DOUBLE PRECISION, INTENT(OUT) :: UA,UT
END SUBROUTINE WRENCH
SUBROUTINE FORCES(NBLADE,MCP,ADVCO,WAKE,RV,RC,TANBC,UASTAR,UTSTAR,
VA,CHORD,CD,G,CT,CQ,CP,KT,KQ,EFFY,RHV,CTH,IHUB)
INTEGER, INTENT(IN) :: NBLADE,MCP,IHUB
REAL, INTENT(IN) :: ADVCO,WAKE,RHV
REAL, DIMENSION(:), INTENT(IN) :: RV,RC,TANBC,UASTAR,UTSTAR,VA,
CHORD,CD,G
REAL, INTENT(OUT) :: CT,CQ,CP,KT,KQ,EFFY,CTH
END SUBROUTINE FORCES
&
&
REAL FUNCTION VOLWK(XR,XVA)
REAL, DIMENSION(:), INTENT(IN) :: XR,XVA
END FUNCTION VOLWK
END INTERFACE
END MODULE PVLMOD
217
5 APPENDIX
218
5.1 DerivationofGlauertsIntegral
Thefollowingderivationhas beenextractedfromGlauerts book
64
,withthenotation
changedtoconformtothepresent lecturenotes.Westartwithequation51,which denes
thelimitingprocessoftheCauchy principalvalueintegral,
_ _

c f(

)d

=lim_

f(

)d

+ f(

)d

(261)
0
0 0

+
ConsideringrstthevalueofI
0
,
_ _

=
_
1
log
sin
1
2
( x+

)
_
0 cos xcos

sin x
sin
1
( x

)
2
0
1 1 1
=
_
sin x
{log sin( x
2
)
_

log sin
2
}

1 sin
1
( x+

+ cos xcos

=
_
sin x
log
sin
1
2
2
( x

)
_

+
1 1 1
= {log sin log sin( x+ )} (262)
sin x 2 2
andhence
=lim 0

1
log
sin( x
2
1
)

= 0 (263) I
0
sin x sin( x+
1
2
)
Then

cos

I
1
= d

0 cos

cos x

cos x

= 1 + d

0 cos

cos x
= + I
0
cos x
= (264)
64
H. Glauert, The Elements of Aerofoil and Airscrew Theory, Cambridge UniversityPress
219
andmoregenerallyifn > 1,

cos(n+1)

+cos(n1)

I
n+1
+ I
n1
=
0 cos xicos x
d

2 cos

cosn

= d

0 cos xicos x

2 cos

cosn

= 2 cos n

+ d

0 cos xicos x
= 2 cos xI
n
(265)
Thesolutionofthisrecurrenceformula
I
n+1
2 cos xI
n
+ I
n1
= 0 (266)
withtheinitialconditionsI
0
= 0 and I
1
=,leadstothenalresult,

I
n
( x) = c
cos(n

)
d

=
sinnx
(267)
cos

cos x
sin x
0
Finally,sincethedenominatorinequation50contains

and x inthereverseorder,
thesignofthenalresultinequation52isalsoreversed.
220