DimensionalAnalysisandSimilarity

Dimensional analysisisveryusefulfor planning,presentation,andinterpretationofexperimentaldata.

As discussed previously, most practical fluid mechanics problems are too complex to solve analytically
andmustbetestedbyexperimentorapproximatedbycomputationalfluiddynamics(CFD).Thesedata
havemuchmoregeneralityiftheyareexpressedincompact,economicnondimensionalform.
Dimensionalanalysisisamethodforreducingthenumberandcomplexityofexperimentalvariablesthat
affectagivenphysicalphenomena.
Advantagesofdimensionalanalysis
1) Reduce the number of variables: Suppose the force F on a particular body shape immersed in a
streamoffluiddependsonlyonthebodylengthL,velocityV,fluiddensitypandviscosityp:
F = (I, I, p, p)
Ingeneral,ittakesabout10pointstodefineacurve.Tofindtheeffectsofeachparameterontheforce,
we need to perform 10X10X10X10=10
4
tests! However, using dimensional analysis, we can reduce the
parameterstoonlyone:
F
pI
2
I
2
= g_
pII
p
] = g(Rc)
That is, the nondimensional force is a function of the dimensionless parameter Reynolds number. So,
wedo not havetovaryL, V,p, pseparatelybutonlythegrouping pIIp.This can be donebyvarying
velocityVinthewindtunnelorwaterchannel.
2) Nondimensional equations: that will provide insight on controlling parameters and the nature of
theproblem.
3) Scaling laws: that allows testing models instead of expensive large fullscale prototypes. There are
rules for finding scaling laws or conditions of similarity. In our force example, if the similarity
conditionexists:
Rc
m
= Rc
p

Thenonecanwrite:
F
p
F
m
=
p
p
p
m
_
I
p
I
m
]
2
_
I
p
I
m
]
2

wheresubscriptsmandpindicatemodelandprototype,respectively.Thisequationisthescalinglaw:if
youmeasurethemodelforceatthemodelReynoldsnumber,theprototypeforceatthesameReynolds
numberequalsthemodelforcetimesthedensityratiotimesthevelocityratiosquaredtimesthelength
ratiosquared.

Principle of dimensional homogeneity: in a dimensionally homogenous relationship for a physical

process,eachtermwillhavethesamedimensions.Asanexample,considertheBernoullisequation:
p
p
+
I
2
2
+gz = const
Eachterm,includingtheconstant,hasdimensionsofvelocitysquared[L
2
T
2
].
M.BahramiFluidMechanics(S09)DimensionalAnalysisandSimilarity2

Variablesandscalingparameters
The variables are the things we wish to plot, the basic output of the experiment or theory. The
parametersarethosequantitieswhoseeffectsonthevariableswewishtoknow.Forexample,consider
therelationthatexpressesthedisplacementofafallingbody:
S = S
0
+I
0
t +
1
2
gt
2

To nondimensionalize our results, we need to know how many dimensions are contained among our
variables(S,t)andparameters(V
0
,S
0
,andg);inthiscaseonlytwo,length{L}andtime{T}:
{S}={S
0
}={L}{t}={T}{V
0
}={LT
1
}{g}={LT
2
}
Among our parameters, we therefore select two to be scaling parameters (or repeating parameters),
used to define dimensionless variables. For the fallingbody problem, we select any two of the three
parameterstobescalingparameters,thuswehavethreechoices:
Option1:ScalingparametersS
0
andV
0
;theeffectofgravityg.Usingthesescalingparameters,wedefine
nondimensionaldisplacementandtimeas:
S
-
=
S
S
0
t
-
=
I
0
t
S
0

Substituting these variables into the fallingbody equation, after simplifications, we find the non
dimensionalequation:
S
-
= 1 + t
-
+
1
2
ot
-2

where
o =
gS
0
I
0
2

There is a single dimensionless parameter o. Note that this plot cannot show the effect of S
0
and V
0

sincetheyarehiddenintheordinateandabscissa.
Option 2: Scaling parameters V
0
and g: the effect of initial displacement S
0
. The nondimensional
parameterswillbe:
S
--
=
Sg
I
0
2
t
--
=
tg
I
0

M.BahramiFluidMechanics(S09)DimensionalAnalysisandSimilarity3

Substituting these variables into the fallingbody equation, after simplifications, we find the non
dimensionalequation:
S
--
= o + t
--
+
1
2
t
--2

where
o =
gS
0
I
0
2

Thesamesingleparameterappearsagain.
Option3:ScalingparameterS
0
andg:theeffectofinitialspeedV
0
.Usingthescalingparameters(S
0
, g),
dimensionlessdisplacementandtimecanbefound:
S
---
=
S
S
0
t
---
= t _
g
S
0
]
12

Substituting these variables into the fallingbody equation, after simplifications, we find the non
dimensionalequation:
S
---
= 1 + [t
---
+
1
2
t
---2

where
[ =
1
o
=
I
0
gS
0

Note that, in all three options, the same parameters o appears but has a different meaning:
dimensionlessgravity,initialdisplacement,andinitialvelocity.So,ingeneral,onecanwrite:
S
i
= (t
i
, o)
M.BahramiFluidMechanics(S09)DimensionalAnalysisandSimilarity4

Theselectionofscalingparametersislefttotheuser,buttherearesomeguidelines:
1) Thescalingvariablesmustnotformadimensionlessgroupamongthemselves,butaddingonemore
variable will form a dimensionless quantity. For example, in the above fluid flow problem, test the
powersofp, I, onJ I:
p
u
I
b
I
c
= (HI
-3
)
u
(II
-1
)
b
(I)
c
= H
0
I
0
I
0
only i o = b = c = u
2) Donotselectoutputvariablesforyourscalingparameters.
3) If convenient, select popular, not obscure, scaling variables because they will appear in all of your
dimensionlessgroups.Forexample,selectdensitynotsurfacetension.
Note: the two following criteria must be satisfied before performing dimensional analysis: 1) the
proposed physical relation is dimensionally homogenous, and 2) all the relevant variables have been
includedintheproposedrelation.
ThePitheorem
The Buckingham theorem is a key theorem in dimensional analysis. This provides a method for
computingsetsofdimensionlessparametersfromthegivenvariables,eveniftheformoftheequation
isstillunknown.However,thechoiceofdimensionlessparametersisnotunique:Buckingham'stheorem
only provides a way of generating sets of dimensionless parameters, and will not choose the most
'physicallymeaningful'.
Let q
1
, q
2
, q
3
, q
n
be n dimensional variables that are physically relevant in a given problem and that
are interrelated by an (unknown) dimensionally homogeneous set of equations. These can be
expressedviaafunctionalrelationshipoftheform:
F(q
1
, q
2
, q
3
, q
n
) = u or q
1
= ( q
2
, q
3
, q
n
)
Ifkisthenumberoffundamentaldimensionsrequiredtodescribethenvariables,thentherewillbek
primary variables and the remaining j = k n variables can be expressed as n k dimensionless and
independent quantities or Pi groups
1
,
2
,
3
,
n-k
. The functional relationship can thus be
reducedtothemuchmorecompactform:
(
1
,
2
,
3
,
n-k
) = u or
1
= (
2
,
3
,
n-k
)
Note:thissetofnondimensionalparametersisnotunique.Theyarehoweverindependentandforma
completeset.
Application:
1)Clearlydefinetheproblemandthinkaboutwhichvariablesareimportant.Identifywhichisthemain
variableofinterest,i.e.,q
1
= ( q
2
, q
3
, q
n
).Itisimportanttophysicallythinkabouttheproblem.Are
thereanyconstrains,i.e.canIvaryallthesevariablesindependently?Forexample,weightofanobject
w = pIg (onlytwooftheseareindependent,unlessgisalsovariable).
M.BahramiFluidMechanics(S09)DimensionalAnalysisandSimilarity5

2)Expresseachofnvariablesintermsofitsfundamentaldimensions,{MLT}or{FLT}.Itisoftenuseful
to use one system to solve the problem, and then check that groups you obtain are dimensionless by
convertingtoothersystem.
3) Determine the number of Pi groups, j = n k variables, where k is the number of reference
dimensions and select k primary or repeating variables. Typically pick variables which characterize the
fluidproperties,flowgeometry,flowrate
4)Formjdimensionlessgroupsandcheckthattheyareallindeeddimensionless.
5) Express result in form
1
= (
2
,
3
,
n-k
) where
1
contains the quantity of interest and
interpretyourresultphysically!
6)Makesurethatyourgroupsareindeedindependent;i.e.canIvaryoneandkeepothersconstant?
7)Comparewithexperimentaldata!
Example1:
Usethepitheoremtofindthedimensionlessparametersinthedragproblem.
F = F(I, I, p, p)
Nondimenalizationofthebasicequations
Anotherusefulapplicationofpitheoremisapplicationtothebasic(governing)equations.Eventhough
these equations cannot be solved in general, they will reveal basic dimensionless parameters, such as
Reynoldsnumber,intheirproperformandproperposition,givingclueswhentheyarenegligible.
Considerincompressiblesteadyflowgoverningequations:
v. I = u
p
JI
Jt
= pg - vp + pv
2
I
Withboundaryconditions.TheseequationscontainthethreebasicdimensionsM,L,andT.Allvariables
p,V,x,y,z,andtcanbenondimensionalizedbyusingdensityandtworeferenceconstantsthatmight
becharacteristicoftheparticularfluidflow:
Referencevelocity=UReferencelength(orcharacteristicslength)=L
As an example, U can be the inlet or upstream velocity and L the diameter of a body immersed in the
stream.Wecandefinealldimensionlessvariables:
I
-
=
I
u
v
-
=
v
I

x
-
=
x
I
, y
-
=
y
I
, z
-
=
z
I
, R
-
=
R
I

M.BahramiFluidMechanics(S09)DimensionalAnalysisandSimilarity6

t
-
=
tu
I
, p
-
=
p + pgz
pu
2

Note that for the pressure, we considered the piezometeric pressure, assuming that z is up (Bernoulli
equation).Sincep, u, onJ Iareconstants,thederivativesinthegoverningequationscanallbehandled
indimensionlessform,forexample:
ou
ox
=
o(uu
-
)
o(Ix
-
)
=
u
I
ou
-
ox
-

Aftersubstitutionandsimplifications,weobtain:
7
-
. I
-
= u
JI
-
Jt
-
= -7
-
p
-
+
p
puI
v
-2
(I
-
)
Theboundaryconditionsshouldalsobenondimensionalizedinasimilarmanner.
NondimensionalParameters
The continuity equation contains no parameter; however, the momentum (NavierStokes) equation
revealstheReynoldsnumber:
Rc =
puI
p

TheReynoldsnumberisalwaysimportant,withorwithoutafreesurface,andcanbeneglectedonlyin
flowregionsawayfromhighvelocitygradients,e.g.awayfromsolidsurfaces,jetsorwakes.
Euler number (pressure coefficient) is rarely important unless the pressure drops low enough to cause
vaporformation(cavitation)inaliquid:
Eu =
p
pu
2

Froude number is the dominant effect in freesurface flows and it does not appear if there is no free
surface:
Fr =
u
2
gI

Machnumberhasastrongeffectoncompressibleflowpropertiesifitisgreaterthan0.3:
Ho =
u
o

whereaisthespeedofsoundinthefluid.
M.BahramiFluidMechanics(S09)DimensionalAnalysisandSimilarity7

Table 52, in the textbook, provides a comprehensive list of important dimensionless groups in fluid
mechanics.
Modelinganditspitfalls
Performing scale analysis is mathematically straightforward; however, there are several issues to
consider:
1) We took for granted the important variables for the phenomena. Selection of important variables is
notatrivialtaskandrequiresconsiderablejudgmentandexperience.Notethateachextrapigroupthat
isretainedincreasestheexpenseandeffortandlevelofcomplexityofthesolution.
2) Typically the Reynolds number of the model is too small (by a factor of 10 to 1000). As a result, we
endupestimatingthedesiredhighReynoldsnumberprototypedata,byextrapolatingthemodeldatain
low Re, as shown in Fig.1. This may result in high uncertainty in the analysis; but there is no other
practicalalternativeinhydraulicmodeltesting.

Fig.1:Reynoldsnumberextrapolation.
3)Afterselectingimportantparametersanddeterminingpigroups,weshouldseektoachievesimilarity
betweenthemodeltestedandtheprototypetobedesigned.Ingeneral:
Flow conditions for a model test are completely similar if all relevant dimensionless parameters have
thesamecorrespondingvaluesforthemodelandtheprototype.
Establishingcompletesimilarityishighlyunlikely.Therefore,weseekparticulartypesofsimilarity.
Geometricsimilarity:concernsthelengthdimension{L}andmustbeensuredbeforeanymodeltesting
canproceed.Ingeneralgeometricalsimilarityisestablishedwhen:
A model and prototype are geometrically similar if and only if all body dimensions in all 3 coordinates
havethesamelinearscaleratio.
M.BahramiFluidMechanics(S09)DimensionalAnalysisandSimilarity8

Thismeansalltheangles,flowdirections,andtheorientationsofmodelandprototypewithrespectto
surroundingsmustbeidentical.

Fig.2:Geometricsimilarityinmodeltesting.
Kinematic similarity: requires that the model and prototype have the same length scale ratio and the
sametimescaleratio;thusthevelocityscaleratiowillbethesameforboth.
Length scale equivalence simply implies geometric similarity, but time scale equivalence may require
additionaldynamicconsiderationssuchasequivalenceoftheReynoldsandMachnumbers.

Fig.3:FreesurfaceflowsarekinematicallysimilarwithlengthandtimescalesrelatedbytheFroude
number.
Frictionless flows with a free surface, see Fig. 3, are kinematically similar if their Froude numbers are
equal:
M.BahramiFluidMechanics(S09)DimensionalAnalysisandSimilarity9

Fr
m
=
I
m
2
gI
m
=
I
p
2
gI
p
= Fr
p

IfI
m
= oI
p
whereisadimensionlessratio,thevelocityscalebecomes:
I
m
I
p
= _
I
m
I
p
_
1
2
= o
1
2

I
m
I
p
=
I
m
I
m

I
p
I
p

= o
Dynamic similarity: exists when the model and the prototype have the same length scale ratio, time
scaleratio,andtheforcescale(massscale)ratio.
Dynamicsimilarityexists,simultaneouslywithkinematicsimilarity,ifthemodelandprototypeforceand
pressurecoefficientsareidentical.Thisisestablishedif:
1) For compressible flow, the model and prototype Reynolds number and Mach number and specific
heatratioarecorrespondinglyequal.
2)Forincompressibleflow
a)Withnofreesurface:modelandprototypeReynoldsnumbersareequal.
b) With a free surface: model and prototype Reynolds number, Froude number, and (if
necessary)Webernumberandcavitationnumberarecorrespondinglyequal.

Fig.4:dynamicsimilarityinsluicegateflow.Modelandprototypeyieldidenticalhomologousforce
polygonsiftheReynoldsandFroudenumbersarethesamecorrespondingly.