Experimental Methods for Engineers

ME505M
Lecture #2: Similarity Principles
and Dimensional Analysis
Recap
● 7 standardized units, others like density, pressure etc. are derived quantities
● Measurement scale continuous, eg. length/temperature/pressure scale
○ Counting is NOT measurement. With care, counting can be error free.
○ All measurements subjected to one or more sources of error, some make it
large while others make it small. Reading scale just one of them.
○ Errors in measurement tend to obscure truth or to mislead experimenter.
● When we perform measurements we do not get whole number of standardized
unit: we get fractional part of standardized unit.
○ Measurement results in a rational number with units.
○ Need suitable units to make measurements (eg. how many time standard unit
meter contained in a measured distance?)
Recap
● Why standardize units? Without this standardization, scientists from different
countries would be greatly handicapped in exchange of scientific information
○ Remember, all scientific discoveries are built one on top of another
● Goal of experimenter in error analysis: Identify all sources of error and keep them
as small as possible. Note: still error is NOT zero!,
○ Need to quantify error, thus information on sources of error indispensable.
● We cannot measure everything, only a small randomly chosen ‘sample’ is
subjected to measurement. Eg. measure rainfall in area, steel bar from mill, water
impurity on a lake or river, blood sugar test etc.
Choosing ‘right method of measurement’ and ‘unit needed’ depends on
magnitude of thermophysical property measured (i.e. or magnitude of
measurand) and accessibility of object.
Does Small Differences in Measurement Matter?
● I hear you telling, common small differences unimportant as gold article will give
acceptable service even if slightly below 14K. It is going to shine same!
● But small differences may be very important for a number of reason
● If one variety of rice yields just 1% more grain than another variety, this
difference may be unimportant to a small farmer. But added up for whole of India
this small difference would mean millions of rice tonnes to feed hungry.
● Small difference has tremendous scientific consequences. Illustrated with story
○ Our atmosphere is ~80% nitrogen. Chemists can remove O2, CO2, and moisture. At
one time residual gas was believed to consist solely of N2. There is an interesting
chemical, ammonium nitrite, NH4NO2, which can be prepared in a very pure form.
When heated, it decomposes to give N2 and H2O. Pure N2 obtained from air or from
NH4NO2 should have identical chemical and physical properties.
Importance of Small Difference: Fascinating Example
● In 1890, Lord Rayleigh, undertook a study in which he compared N2 obtained
from air with N2 released by heating NH4NO2. He wanted to compare densities of
the two gases; that is, their weights per unit of volume. He did this by filling a
bulb of carefully determined volume with each gas in turn under standard
conditions: sea level pressure at 0° C.
● Weight of bulb when full minus its weight when N2 was exhausted gave weight of
N2. One measurement of weight of atmospheric N2 gave 2.31001 g. Another
measurement on nitrogen from NH4NO2 gave 2.29849 g.
○ Difference, 0.01152 g, is very small. Lord Rayleigh was faced with a
problem: was difference a measurement error or was there a real difference
in densities? On the basis of existing chemical knowledge there should have
been NO difference in densities.
Importance of Small Difference: Fascinating Example
● Several additional measurements were made with each gas (why?), and Lord
Rayleigh concluded that his data were convincing evidence that observed small
difference in densities was in excess of experimental errors of measurement
and therefore actually existed.
● There now arose intriguing scientific problem of finding a reason for observed
difference in density. Further study finally led him to believe that N2 from air
contained some hitherto unknown gas or gases that were heavier than N2, and
which had not been removed by means to remove other known gases.
● Proceeding on this assumption, he soon isolated argon. Then followed discovery
of whole family of rare gases, existence of which had not even been
suspected. Small difference in densities, carefully evaluated as not accidental,
led to a scientific discovery of major importance.
Measurements Share Certain Common Properties
● Tremendous efforts are made to improve techniques (or methods) of making
measurements: who knows what other exciting discoveries still lie hidden behind
small differences!
● Only when we know what are sources of error in our measurements can we set
proper tolerances, evaluate small differences, and estimate accuracy of our
measurements of physical constants.
● Study of measurements has shown that there are certain properties common to
all measurements: certain mathematical laws apply to all measurements
regardless of what it is that is measured.
● When scientist make measurement, they get numerical result: They trust it and
use it only if measurement apparatus and technique are adequate.
● Scientific measurements need to be performed precisely and accurately
Scientific problem that require precise readings
● When you read your body temperature with thermometer, it can easily be read to
an extra decimal place - as its scale is expanded over readings made for room
temperatures in which we are generally satisfied with nearest whole number.
(PS: mercury may end at a mark or anywhere in between two adjacent marks in
both cases)
● Here is an example of a scientific problem that requires precise temperature
readings. Suppose you collect some rain water and determine that its freezing
point is 0°C. Now take 1 litre of rain water sample & add ~28 g of table sugar.
● Place a portion of this solution in a freezing brine mixture of ice and table salt,
stirring it all the while. Ice will not begin to appear until temperature has dropped
to a little more than -17.5°C., original sample turned to ice at 0°C. (Brine is
anti-icing and deicing agent. Prevents ice from sticking onto surface.)
Freezing Point Depression
● From this simple experiment you can see that freezing points can be used to
determine whether or not a solvent is pure.
● Depressions of freezing point produced by dissolving substances in solvents
have long been a subject of study/research. In these studies, temperatures are
usually read to at least one thousandth of a degree, with special thermometers
(can you guess how they work?).
○ Position of Hg estimated to a tenth of interval between marks 0.01°C apart.
○ Very exact temperature measurements taken just as liquid begins to freeze
can be used to detect presence of minute amounts of impurity.
● In scientific work knack of estimating tenths of a division on scales and
instrument dials becomes almost automatic, it comes with practise!
● With digital instruments: must be calibrated to required decimal place precision.
Similarity Principles and Dimensional Analysis
● An experiment that is carried out in a lab under reference conditions will invariably
be a scaled version of prototype for which a mathematical model is to be
developed.
● Data obtained from scaled experiments will be useful for prototype only if certain
similarity principles/conditions are satisfied. These principles developed for systems
operating in deterministic conditions (not stochastic, eg. boundary-layer transition):
measured data can then be linearly scaled from experiment to real-life process.
Similarity Principles and Dimensional Analysis
Principles of similarity that relate experimental setup with prototype are stated as:
● Geometric similarity should be observed. It requires that experiments preserve shape of
prototype, while linear dimensions be scaled in proportion.
● Dynamic similarity should be enforced. It states that ratio of relevant forces in prototype
be preserved in model as well. This leads to matching of certain dimensionless
parameters.
● Kinematic similarity should be realized. It requires that constant flux lines (streamlines,
for example) in experiment match those in prototype.
○ This requirement is most appropriate for steady state patterns since constant flux
lines in unsteady problems may not be physically meaningful.
● If geometric and dynamic similarity are realized, kinematic similarity will be automatically
satisfied.
Dimensional Homogeneity
● There are seven primary dimensions. All other dimensions can be formed by
combinations of these. The seven primary dimensions are, along with their symbols:

● The primary dimensions of variables in an experiment or analysis can be used to our

advantage – to reduce the required amount of effort.
● Law of dimensional homogeneity: “Every additive term in an equation must have he
same dimensions.” Example: The total energy (E) of a system is composed of internal
energy (U), kinetic energy (KE), and potential energy (PE), i.e., E = U + KE + PE.
● Law of dimensional homogeneity is basis for useful technique of dimensional analysis.
Dimensional Analysis and Method of Repeating Variables
● Dimensional analysis is a simple, powerful tool that is useful in all disciplines (but
unfortunately, is usually taught only in fluid mechanics).
● Goal of dimensional analysis is to reduce number of independent variables in a problem.
We accomplish this by converting all dimensional variables into nondimensional
parameters, called “Pi’s” or Π(upper case Greek letter pi).
● Method of Repeating Variables has 6 steps:
Example
● Given: Drag force FD on a car is function of 4 variables: air
velocity V, air density ⍴, air viscosity μ, & frontal area A of car.
● Express this relationship in terms of non-dimensional variables.
Solution: Follow 6 steps of method of repeating variables.
● Step 1: We list and count the variables. FD = f(V,⍴,μ,A). There
are 5 variables, so n = 5.
● Step 2: List primary dimensions of each variable: {FD} = {mL/t2}, {V} = {L/t}, {⍴} = {m/L3}, {μ} =
{m/Lt}, and {A} = {L2}
● Step 3: We set reduction j. As a first guess, we set j equal to number of primary dimensions
represented in problem (i.e. mass (m), length (L), and time (t)) so set j = 3. [PS: If this does
not work, we decrease j by 1 & start over from here.] We expect k = n – j = 5 – 3 = 2 Πs.
● Step 4: We choose j repeating parameters. Since j = 3, we choose V, ⍴, and A. (PS: This is
trickiest part of process.) How do we know which parameters to pick as repeating
parameters? Here are some guidelines or “rules” about picking repeating parameters:
Guidelines or Rules for Picking Repeating Parameters
● Never pick dependent variable (the one on the left) – in this case, we cannot pick FD.
● Ensure that repeating variables cannot by themselves form a dimensionless group.
● All primary dimensions in the problem must be represented by j repeating parameters.
● Do not pick variables that are already dimensionless (e.g., angles).
● Do not pick two variables with same dimensions or with dimensions that are powers of each
other: e.g., cannot pick both length {L} and area {L2} – in terms of dimensional analysis, these
are really same, since they both represent only a length.
● Pick “common” variables, since repeating variables end up appearing in more than one Π.
(That’s why we call them “repeating variables” in first place!)
● Whenever possible, choose simple variables (e.g., pick {L} instead of {mL2t/T} if appropriate.)
Step 5
● Generate the Πs
● Π1 = dependent Pi = FDVa⍴bAc.[i.e. dependent Pi contains dependent variable on left side]

● We “manipulate” by multiplying by a constant (this is perfectly okay in dimensional analysis,

since a constant has no dimensions, we are not changing the nondimensional nature of the Π).
● We manipulate in order to get agreement with the commonly accepted and published
dimensionless parameter called the drag coefficient.
Step 5
● Similarly, we generate the second Π by using the same three repeating variables,

●
Step 6
● We write the final functional relationship between the Πs. In general, this relationship is of the form
Π1 = function(Π2, Π3,....Πk) for k Πs. In this example, k = 2, so our final result is

● What have we achieved? Well, let’s compare the number of parameters in original problem with that
of reduced (nondimensionalized) problem:
○ Original problem: FD = f(V,⍴,μ,A)....5 parameters (drag force is function of 4 variables).
○ Reduced problem: CD = f(Re)...2 parameters (drag coefficient is function of 1 nondimensional
parameter, namely Reynolds number).
○ We have reduced number of independent parameters by 3, i.e., from 4 to 1!
○ Saves us significant time & money on experiments &/or in CFD since we need to vary only 1
parameter (Re) instead of 4 parameters (V,⍴,μ,A). This is power of dimensional analysis.
○ In some simple dimensional analysis problems, k = 1 (only one nondimensional parameter is formed).
Step 6, Π1= function(nothing), makes sense only if Π1=constant, found experimentally.
Summary
● Statements of principles of similarity can now be stated as

● If conditions of similarity are satisfied then dimensionless drag coefficient recorded in

experiments will be equal to one for prototype at every Reynolds number.