Fluid Mechanics II

by Dr. Ahmed Mongy

Assistant professor
Mechanical power engineering
1
Tanta University
 CHAPTER (5)

Dimensional analysis and

Similarity
 Introduction
❑ Before CFD (computers), there were two approaches that we take to study fluid dynamics,
one analytical and the other experimental. The analytical approach involves solving the
integral momentum, continuity and boundary layer equations. However, this technique is
limited to simple geometries such as smooth flat plate. In complicated geometries or
complex flows, analytical solution is too difficult to be obtained, and thus we rely on
experimentation. On the other hand, the experimental or empirical approach involves
performing measurements under laboratory conditions.
❑ With the development of high-speed computers, a third method, the numerical approach
(CFD), has become available by solving the governing equations numerically. However, we
still resort to testing and experimentation, but only in the final stages of design.

3
 ❑ Then, the results which obtained experimentally or numerically are typically reported as data
points and curves. Through dimensional analysis, (1) such data can be generalized over
other problems by scaling, and (2) the number of experimental variables can be reduced
by grouping them in dimensionless form. we do not change our data; we are changing the
way we present it. As a result,
❑ Example (1), suppose that the drag 𝑭 acting on a particular body immersed in a fluid flow
depends on the body length 𝑳, flow velocity 𝑽, fluid density 𝝆, and fluid viscosity 𝝁; such that:
𝐹 = 𝑔 𝐿, 𝑉, 𝜌, 𝜇
➢ Where the function 𝒈 is found experimentally or numerically.
• With dimensional analysis, we can reduce this equation to the equivalent form:

𝐹 𝜌𝑉𝐿
= 𝑔 𝐶𝐷 = 𝑔 𝑅𝑒
𝜌 𝑉 2 𝐿2 𝜇

• That is, the drag coefficient 𝑪𝑫 is function of the dimensionless Reynolds number 𝑹𝒆.
• Generally, if it takes 10 points to define a curve, we have to run the experiment for 10
different lengths to find the effect of body length. For each length 𝑳 we need 10 values of 𝑽,
10 values of 𝝆, and 10 values of 𝝁, making a grand total of 10,000 experiments. However,
with dimensional analysis, we need to run the experiment for a total of 10 experiments only
to find the equivalent relation.
4
 • Furthermore, the dimensional analysis can provide generality (similarity) between a small
model and a large-scale prototype by scaling laws. In the previous example, we can
measure the drag on a small model to predict the drag on the full-scale prototype car by
using scaling laws. However, there are some rules shall be existed, at first, to make the
scaling laws valid. These rules are so-called similarity conditions. In this simple example,
similarity is achieved if the Reynolds number is the same for the model and prototype
because the function 𝒈 requires the drag coefficient to be the same also:
𝑅𝑒𝑚 = 𝑅𝑒𝑝 𝐶𝐷𝑚 = 𝐶𝐷𝑝

➢ Where subscripts 𝒎 and 𝒑 mean model and prototype, respectively.

• From the definition of force coefficient, this means that
𝐹𝑚 𝐹𝑝 𝐹𝑚 𝜌𝑚 𝑉𝑚2 𝐿2𝑚
= =
𝜌𝑚 𝑉𝑚2 𝐿2𝑚 𝜌𝑝 𝑉𝑝2 𝐿2𝑝 𝐹𝑝 𝜌𝑝 𝑉𝑝2 𝐿2𝑝

➢ Which is the scaling law between the model and the prototype.

5 Which is the model, and which is the prototype?

 Dimensional homogeneity
• The idea behind dimensional analysis is that any dimensionally homogeneous equation can
be written in a nondimensional form.
• An equation is said to be dimensional homogeneous if each of its additive terms have the
same dimensions. Example (2), consider Bernoulli’s equation for incompressible flow as:

1
𝑃 + 𝜌𝑉 2 + 𝜌𝑔𝑍 = 𝑃𝑜
2
• Each term, including the constant 𝑷𝒐 , has dimensions of pressure, or {𝐌 𝐋−𝟏 𝐓 −𝟐 }.
• We can classify each term of the dimensionally homogeneous equation into variables and
parameters. Variables are the quantities that vary during a given case and have
dimensions such as 𝑷, 𝑽 and 𝒁. Parameters are the quantities that vary from case to case
but are kept constant during a given case and have dimensions such as 𝝆, 𝒈 and 𝑷𝒐 . The
parameters can be, further, subdivided into scaling or repeating or scaling parameters and
basic parameters.
• Now, in order to nondimensionalize our equation, we need, first, to know how many
fundamental dimensions are contained among our variables and parameters. In our case,
we have only three, mass {𝐌}, length {𝐋} and time {𝐓}.
6
 • We check each term to verify this:
𝑃 = 𝑃𝑜 = {M L−1 T −2 }
𝜌 = {M L−3 }
{𝑉} = {L T −1 }
𝑔 = {L T −2 }
𝑍 = {L}
• Among our parameters, we select 𝝆, 𝒈 and 𝑷𝒐 to be scaling parameters used to define
dimensionless variables or groups. These choices will not affect the content of our data,
only the form of their presentation. Now use the scaling parameters 𝝆, 𝒈 and 𝑷𝒐 to define
dimensionless variables as:
𝑃 𝜌 𝜌𝑔
𝑃∗ = 𝑉∗ = ×𝑉 𝑍∗ = ×𝑍
𝑃𝑜 𝑃𝑜 𝑃𝑜

• Substitute these variables into previous equation. The result is:

1 Nondimensionalize 1
𝑃 + 𝜌𝑉 2 + 𝜌𝑔𝑍 = 𝑃𝑜 𝑃∗ + 𝑉 ∗ 2 + 𝑍 ∗ = 1
2 2
• Therefore, dimensional analysis reduced the problem to only three dimensionless variables.
The reduction 𝟔 − 𝟑 = 𝟑 should equal the number of fundamental dimensions involved in
the problem {𝐌, 𝐋, 𝐓}.
7
 • The selection of scaling variables should consider (1) they must not form a dimensionless
group among themselves, (2) selecting the scaling parameters that can appear in all your
dimensionless groups.
• Considering example (2), the variables are 𝑭 and 𝑽. Scaling parameters are 𝝆, 𝝁 and 𝑳.
Now use the scaling parameters 𝝆, 𝝁 and 𝑳 to define dimensionless variables as:

𝜌𝐹 𝜌𝑉𝐿
𝐹∗ = 𝑉∗ =
𝜇2 𝜇
• Another option is to select the variables as 𝑭 and 𝝁. Scaling parameters are 𝝆,𝑽 and 𝑳.
Now use the scaling parameters 𝝆, 𝑽 and 𝑳 to define dimensionless variables as:

𝐹 𝜌𝑉𝐿
𝐹 ∗∗ = 𝑉∗ =
𝜌 𝑉 2 𝐿2 𝜇
• Therefore, dimensional analysis reduced the problem to only two dimensionless variables.
The reduction 𝟓 − 𝟐 = 𝟑 should equal the number of fundamental dimensions involved in
the problem {𝐌, 𝐋, 𝐓}. This is the fundamental idea of the pi theorem.

8
 The Pi theorem
❑ There are several methods of reducing a number of dimensional variables into a smaller
number of dimensionless groups. The first method was proposed in 1914 by Buckingham
and is now called the Buckingham 𝚷 −Theorem.
❑ The 𝚷 −Theorem involves two parts; in the first part, we are reducing the number of
dimensional variables, while in the second part, we are finding the pi groups.
➢ The first part of the 𝚷 − theorem:
If a physical process satisfies the dimensional homogeneity and involves 𝒏 dimensional
variables, it can be reduced to a relation between only 𝒌 dimensionless variables. The
reduction 𝒋 = 𝒏 − 𝒌 is always less than or equal to the number of fundamental dimensions
describing the variables.
➢ The second part of the 𝚷 − theorem:
Find the reduction 𝒋 = 𝒏 − 𝒌, then select the scaling parameters such that their number is
equal to 𝒋 as well as they do not form a pi group among themselves. Each pi group will be the
product of these 𝒋 scaling parameters and a dimensional variable. Then, each pi group is
assigned any convenient nonzero exponent.

9
 ❑ Repeat example (1): suppose that the drag 𝑭 acting on a particular body immersed in a fluid
flow depends on the body length 𝑳, flow velocity 𝑽, fluid density 𝝆, and fluid viscosity 𝝁; such
that:
𝐹 = 𝑔 𝐿, 𝑉, 𝜌, 𝜇
• Rewrite this as a dimensionless relationship.
Solution
1. Count the number of variables: 𝒏 = 𝟓
2. List dimensions of each variable:

𝐹 𝐿 𝑉 𝜌 𝜇
{M L T −2 } {L} {L T −1 } {M L−3 } {M L−1 T −1 }

• Therefore, the number of dimensionless variables 𝒌 = 𝟓 − 𝟑 is equal 𝟐. There will be exactly

two independent dimensionless groups.
• We can use 𝑳, 𝑽 and 𝝆 (or) 𝑳, 𝝆 and 𝝁 (or) 𝑳, 𝑽 and 𝝁 as repeating parameters, as they cannot
form a dimensionless group among themselves.
• Select the repeating 𝒋 variables 𝑳, 𝑽 and 𝝆.
3. Combine 𝑳, 𝑽 and 𝝆 with one additional variable to find the two dimensionless groups:

10
 • First add force 𝑭 with any exponent as you want to find 𝚷𝟏 as:
Π1 = 𝐿𝑎 𝑉 𝑏 𝜌𝑐 𝐹1 = 𝐿 𝑎
𝐿𝑇 −1 𝑏
𝑀𝐿−3 𝑐
𝑀𝐿𝑇 −2 = 𝑀0 𝐿0 𝑇 0
Equate exponents:
Mass: 𝑐 + 1 = 0 𝑐 = −1
Length: 𝑎 + 𝑏 − 3𝑐 + 1 = 0 𝑎 + 𝑏 = −4
Time: −𝑏 − 2 = 0 𝑏 = −2 𝑎 = −2

𝐹
∴ Π1 = 𝐿−2 𝑉 −2 𝜌−1 𝐹 = = 𝐶𝐷
𝜌𝑉 2 𝐿2
• Then add force 𝝁 with any exponent as you want to find 𝚷𝟐 as:
Π2 = 𝐿𝑎 𝑉 𝑏 𝜌𝑐 𝜇1 = 𝐿 𝑎
𝐿𝑇 −1 𝑏
𝑀𝐿−3 𝑐
𝑀𝐿−1 𝑇 −1 = 𝑀0 𝐿0 𝑇 0
Equate exponents:
Mass: 𝑐 + 1 = 0 𝑐 = −1
Length: 𝑎 + 𝑏 − 3𝐶 − 1 = 0 𝑎 + 𝑏 = −2
Time: −𝑏 − 1 = 0 𝑏 = −1 𝑎 = −1

𝜇 1
∴ Π2 = 𝐿−1 𝑉 −1 𝜌−1 𝜇 = =
𝜌𝑉𝐿 𝑅𝑒
• Therefore, if we select 𝝁 with any exponent of −𝟏, 𝜫𝟐 would equal to 𝑹𝒆.

11
 ❑ Example (3): At low velocities, the volume flow 𝑸 through a small-bore tube is a function only
of the tube radius 𝑹, the fluid viscosity 𝝁, and the pressure drop per unit tube length 𝒅𝒑/𝒅𝒙.
Using the pi theorem, find an appropriate dimensionless relationship:

𝑑𝑝
𝑄 = 𝑓 𝑅, 𝜇,
𝑑𝑥
• Rewrite this as a dimensionless relationship.
Solution
1. Count the number of variables: 𝒏 = 𝟒
2. List dimensions of each variable:

𝑄 𝑅 𝜇 𝑑𝑝Τ𝑑𝑥
{L3 T −1 } {L} {M L−1 T −1 } {M L−2 T −2 }

• Therefore, the number of dimensionless variables 𝒌 = 𝟒 − 𝟑 is equal 𝟏. There will be exactly

one independent dimensionless group.
• We can use 𝑹, 𝝁 and 𝒅𝒑/𝒅𝒙, as they cannot form a dimensionless group among themselves.
• Select the repeating 𝒋 variables 𝑹, 𝝁 and 𝒅𝒑/𝒅𝒙.

12
 3. Combine 𝑹, 𝝁 and 𝒅𝒑/𝒅𝒙 with 𝑸 with any exponent as you want to find 𝚷𝟏 as:

𝑑𝑝𝑐 1
𝑎 𝑏 𝑎
Π1 = 𝑅 𝜇 𝑄 = 𝐿 𝑀𝐿−1 𝑇 −1 𝑏
M L−2 T −2 𝑐
L3 T −1 = 𝑀0 𝐿0 𝑇 0
𝑑𝑥
Equate exponents:
Mass: 𝑏 + 𝑐 = 0
Length: 𝑎 − 𝑏 − 2𝑐 + 3 = 0
Time: −𝑏 − 2𝑐 − 1 = 0 𝑐 = −1 𝑏=1 𝑎 = −4

−1
𝑑𝑝 𝜇𝑄
∴ Π1 = 𝑅 −4 𝜇1 = 4 = 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡
𝑑𝑥 𝑅 𝑑𝑝Τ𝑑𝑥
• Since there is only one pi group, it must equal a dimensionless constant.

❑ Example (4):The pressure drop due to friction for flow in a long smooth pipe is a function
of average flow velocity, density, viscosity, and pipe length and diameter:
∆𝑝 = 𝑓 𝑉, 𝜌, 𝜇, 𝐿, 𝐷
• Use the pi theorem to rewrite this function in a dimensionless form.
H.W.

13
 Nondimensionalization of the basic equations
• One of the major drawbacks of pi theorem is its need to a large guesswork of
dimensionless groups. The success of this method will depend on choosing the most
effective dimensionless quantities that could govern the problem.
• An alternative approach is by nondimensionalizing (normalizing) the conservation
equations, so we eliminate the guesswork and establish the dimensionless variables in a
rigorous fashion.

𝜕𝑃 𝜕2𝑢 𝜕2𝑢 𝜕𝑢 𝜕𝑢
− +𝜇 + = 𝜌𝑢 + 𝜌𝑣
𝜕𝑥 𝜕𝑥 2 𝜕𝑦 2 𝜕𝑥 𝜕𝑦

𝜕𝑃 𝜕2𝑣 𝜕2𝑣 𝜕𝑣 𝜕𝑣
− +𝜇 + − 𝜌𝑔 = 𝜌𝑢 + 𝜌𝑣
𝜕𝑦 𝜕𝑥 2 𝜕𝑦 2 𝜕𝑥 𝜕𝑦

• The momentum equation contains three basic dimensions 𝑴, 𝑳, and 𝑻. All variables 𝝆, 𝒖, 𝒗,
𝒙, 𝒚, and 𝒕 can be nondimensionalized by using density 𝝆 and two reference constants that
might be characteristic velocity 𝑼 and reference or characteristic length 𝑳.
• Now redefine all variables in their dimensionless form, using pi groups:

𝑢 𝑣 𝑥 𝑦 𝑃 𝑈𝑡
𝑢∗ = , 𝑣∗ = , 𝑥∗ = , 𝑦∗ = , 𝑃∗ = , 𝑡∗ =
14 𝑈 𝑈 𝐿 𝐿 𝜌𝑈 2 𝐿
 • Substitute the variables into the momentum equation:
∗ ∗ ∗ ∗ 2 ∗
𝐿 𝑡∗
𝑢 = 𝑈𝑢 , 𝑣 = 𝑈𝑣 , 𝑥 = 𝐿𝑥 , 𝑦=𝐿𝑦 , 𝑃 = 𝜌𝑈 𝑃 , 𝑡=
𝑈
• Since 𝝆, 𝑼, and 𝑳 are all constants, the derivatives in the differential equations can be handled in
dimensionless form as:

𝜌𝑈 2 𝜕𝑃∗ 𝜇 𝑈 𝜕 2 𝑢∗ 𝜕 2 𝑢 ∗ 𝜌𝑈 2 ∗
𝜕𝑢∗ ∗
𝜕𝑢∗
− × ∗+ 2 × + = × 𝑢 +𝑣
𝐿 𝜕𝑥 𝐿 𝜕𝑥 ∗ 2 𝜕𝑦 ∗ 2 𝐿 𝜕𝑥 ∗ 𝜕𝑦 ∗

𝜌𝑈 2 𝜕𝑃∗ 𝜇 𝑈 𝜕 2 𝑢∗ 𝜕 2 𝑢∗ 𝜌𝑈 2 ∗
𝜕𝑣 ∗ ∗
𝜕𝑣 ∗
− × ∗+ 2 × + − 𝜌𝑔 = × 𝑢 +𝑣
𝐿 𝜕𝑦 𝐿 𝜕𝑥 ∗ 2 𝜕𝑦 ∗ 2 𝐿 𝜕𝑥 ∗ 𝜕𝑦 ∗

• Divide both sides of the equations by 𝝆𝑼𝟐 Τ𝑳, we get the momentum equation in dimensionless
form as:

𝜕𝑃∗ 𝜇 𝜕 2 𝑢∗ 𝜕 2 𝑢∗ ∗
𝜕𝑢∗ ∗
𝜕𝑢∗
− ∗+ × + = 𝑢 +𝑣
𝜕𝑥 𝜌𝑈𝐿 𝜕𝑥 ∗ 2 𝜕𝑦 ∗ 2 𝜕𝑥 ∗ 𝜕𝑦 ∗

𝜕𝑃∗ 𝜇 𝜕 2 𝑢∗ 𝜕 2 𝑢∗ 𝑔𝐿 ∗
𝜕𝑣 ∗ ∗
𝜕𝑣 ∗
− ∗+ × + − 2= 𝑢 +𝑣
𝜕𝑦 𝜌𝑈𝐿 𝜕𝑥 ∗ 2 𝜕𝑦 ∗ 2 𝑈 𝜕𝑥 ∗ 𝜕𝑦 ∗

• The equations show two dimensionless numbers, Reynolds number 𝑹𝒆 and Froude number 𝑭𝒓:
𝜌𝑈𝐿 𝑈
𝑅𝑒 = & 𝐹𝑟 =
𝜇 𝑔𝐿
15
 Similarity
❑ Once the dimensional analysis is performed, we can do similarity between a small model
tested and a large-scale prototype to be designed if the flow conditions for a model and
prototype are completely similar.
❑ The flow conditions are similar when all relevant dimensionless parameters are the same
for the model and the prototype. 𝜫𝟏 = 𝒇 𝜫𝟐 , 𝜫𝟑 , 𝜫 𝟒
❑ Instead of complete similarity of fluid flow conditions around a model and a prototype, we
are interested in satisfying four types of similarity between model and prototype; geometric,
kinematic, dynamic, and thermal.
❑ Geometric similarity: a model and prototype are geometrically similar if and only if all body
dimensions in all three coordinates have the same linear scale ratio.

16
 ❑ Kinematic similarity: a model and prototype are kinematically similar if they have the
same velocity scale ratio and same length scale ratio. As a result, the time scale ratio
will also be the same for both.

𝑉∞ 𝑚 𝑉1 𝑚 𝑉2 𝑚
= = =𝛽
𝑉∞ 𝑝 𝑉1 𝑝 𝑉2 𝑝

17
 ❑ Dynamic similarity: a model and prototype are dynamically similar if the model and the
prototype have the same length scale ratio, time scale ratio, and force scale (or mass scale)
ratio.
❑ For incompressible flow, the dynamic similarity exists between the model and prototype if
Reynolds numbers is the same for both. In case of free surface, the dynamic similarity is
satisfied when both Reynolds number, and Froude number are the same for both.

18
 References
Frank White. “Fluid Mechanics” 2011

19