Hydromechanics

Ch.4 PPT 03: Turbulence

Basudev Biswal,
Civil Engineering Department, IIT Bombay 1
 Turbulence: still an unsolved mystery!
"When I meet God, I am going to
ask him two questions: Why
relativity and why turbulence? I
really believe he will have an
answer for the first.“ -- Werner
Heisenberg (apocryphal story)

Smoke rising from a candle:

http://www.cfdflow.com/wp-content/uploads/2014/10/F16-CFD-Real.jpeg

2
 Van Gogh’s starry night

https://www.amazon.in/Tamatina-Starry-Canvas-Paintings-Multicolour/dp/B071V9KVPN

The painting in fact has some of the key statistical characteristics of

turbulence.

3
 Reynold’s experiment

https://en.m.wikipedia.org/wiki/File:Reynolds_fluid_turbulence_experiment_1883.jpg

Remember how Navier and many after him were troubled with
observations that did not confirm theory. Well, that is because the
derivations we saw in the previous lectures are valid for laminar flow.

From Poissoulle’s experiment, we knew pressure drop to be linearly

proportional to velocity. However, Reynolds found for high velocity,
pressure drop is non-linearly related to velocity: pressure ∝ velocity𝑛
with 𝑛 varying between 1.7 to 2. 4
 Reynold’s number
Reynolds then injected dye in pipe flow and observed that for high
velocity the dye is undergoing a mixing process after a certain distance,
i.e. the flow is transitioning from laminar to turbulent type.

He conducted numerous experiments and observed that the many

factors determine laminar-turbulent transition and came up with an
dimensionless number that is now widely known as Reynolds number:
𝑹𝒆 = (𝒖𝒍𝝆)/𝝁.

For cylindrical pipe (his original experimental domain), 𝒍 = 𝑫 (pipe

diameter).

Reynolds observed that laminar-turbulent happens at 𝑹𝒆 = 𝟏𝟐, 𝟎𝟎𝟎. But

he later set a stringent criteria for the transition: 𝑹𝒆 = 𝟐, 𝟎𝟎𝟎.

Later experiments in special condition (no external disturbance) found

the upper limit of 𝑹𝒆 to be as high as 40,000. However, there is not
much meaning to such numbers from a partial point of view (we need
𝑹𝒆 for normal conditions). 5
 NS equations for turbulence

https://skill-lync.com/blogs/technical-blogs/design-what-is-turbulence-modelling

Turbulent flow can be characterized by expressing flow at any point and

at a particular time as:

𝑢 = 𝑢ത + 𝑢′,

where is the mean velocity at the point and 𝑢′ is the deviation of velocity
from 𝑢ത . Note that ∫ 𝑢 ⋅ 𝑑𝑡 = ∫ 𝑢ത ⋅ 𝑑𝑡 = 𝑢ത ⋅ ∫ 𝑑𝑡. That is ∫ 𝑢′ ⋅ 𝑑𝑡 = 0.

This concept can be utilized to refine the mass and momentum

conservation equations for turbulent flow.
6
 NS equations for turbulent flow
𝜕𝑢 𝜕𝑣 𝜕𝑤
Consider the continuity equation + + = 0, which can be written
𝜕𝑥 𝜕𝑦 𝜕𝑧
ഥ
𝜕𝑢 𝜕𝑣ത ഥ
𝜕𝑤 𝜕𝑢′ 𝜕𝑣 ′ 𝜕𝑤 ′
for turbulent flow as + + + + + = 0.
𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑥 𝜕𝑦 𝜕𝑧

𝜕𝑢ഥ 𝜕𝑣ത ഥ
𝜕𝑤 𝜕𝑢 𝜕𝑣 ഥ
𝜕𝑤 𝜕𝑢 𝜕𝑣ത ഥ
𝜕𝑤
Because + + = + + , + + = 0, which also means
𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑥 𝜕𝑦 𝜕𝑧
𝜕𝑢 ′ 𝜕𝑣 ′ 𝜕𝑤 ′
that + + = 0.
𝜕𝑥 𝜕𝑦 𝜕𝑧

𝑑𝑢 1 𝜕𝑝 𝜇 𝜕2 𝑢 𝜕2 𝑣
For the momentum equation in X direction, = 𝜉𝑥 − + ቀ + +
𝑑𝑡 𝜌 𝜕𝑥 𝜌 𝜕𝑥 2 𝜕𝑦 2
𝜕2 𝑤
ቁ can be written as
𝜕𝑧 2

𝑑𝑢 1 𝜕𝑝 1 𝜕 ഥ
𝜕𝑢 𝜕 ഥ
𝜕𝑢 𝜕 ഥ
𝜕𝑢
= 𝜉𝑥 − + ቆ 𝜇 − 𝜌𝑢′ 𝑢′ + 𝜇 − 𝜌𝑣 ′ 𝑢′ + ቀ𝜇 −
𝑑𝑡 𝜌 𝜕𝑥 𝜌 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑧 𝜕𝑧

𝜌𝑤 ′ 𝑢′ ቁቇ = 0 .

7
 The additional terms
𝑣 ′ 𝑑𝐴
Consider an element with thickness 𝑙 which is the distance 2
between the face 1 and 2. 𝑢′ 1

The velocity fluctuation between 1 and 2 is 𝑢′ and the

velocity fluctuation normal to the face is 𝑣 ′ .

Force on the face due to velocity fluctuation can be given as: 𝛿𝐹 =

𝜌𝑣 ′ 𝑑𝐴𝑢′.

Mass exchange is given by 𝜌𝑣 ′ 𝑑𝐴 and thus 𝜌𝑣 ′ 𝑑𝐴𝑢′ gives net force in X

direction (recall force = 𝜌𝑄𝑉; note that the 𝑣ҧ term will be taken care by
the original NS equation).

Note that we can similarly find expression for shear stress due to mass
exchange in X and Z directions but velocity fluctuation in X direction as
𝜌𝑢′ 𝑢′ and 𝜌𝑤 ′ 𝑢′ .

8
 A formal definition of Re
If we divide this force by 𝑑𝐴 in the previous derivation, it will give us a
stress term which is due to fluid turbulence (responsible for velocity
fluctuation): 𝜏𝑡 = 𝜌𝑣 ′ 𝑢′.

Thus, shear stress due to viscosity is 𝜏𝑣 = 𝜇 ⋅ 𝑢′ /𝑙 (the velocity gradient

here is 𝑢′ /𝑙).

Now the ratio of shear force due to turbulence to shear force due to
viscosity is 𝜏𝑡 /𝜏𝑣 = (𝜌𝑣 ′ 𝑢′)/(𝜇𝑢′/𝑙) = 𝝆𝒗′𝒍/𝝁, which is nothing but
Reynold’s number.

This derivation provides a generic definition of Reynold’s number. 𝑙 here

is the characteristic length, which will vary from system to system.

Thus, the critical Reynold’s number (i.e. when transition from laminar to
turbulence happens) of a system will depends on its characteristic
length 𝑙. In other words, critical Reynold number varies from one system
to other.
9
 Turbulent shear stress
The apparent shear stress or the total shear stress in turbulent flow is
expressed as 𝜏 = 𝜏𝑡 + 𝜏𝑣 , which can b expressed as

𝜏 = 𝜇 + 𝜂 ⋅ 𝑑𝑢/𝑑𝑦, where 𝜂 is the coefficient of eddy viscosity

caused due to turbulence.

Now revisit the expression for 𝜏𝑡 caused purely due to bulk momentum
exchange: 𝜏𝑡 = 𝜌𝑢′ 𝑣 ′ .

We talked earlier about molecular motion (molecules in both liuquid

and gas move randomly), which causes viscosity due to momentum
transfer between fluid layers.

In case of turbulent flow, unlike the case of laminar flow, momentum

exchange happen due to bulk motion of fluid parcels across layers.
Momentum exchange due to bulk movement

10
 Prandl’s mixing 𝑑𝑢
length
Velocity fluctuation 𝑢′ can be defined as u′ = ⋅ 𝑙.
𝑑𝑦

The characteristic length 𝑙 is called here as Prandl’s mixing length

(related to the concept of mean-free path length in gas, but not the
same thing).

Prandl then assumed that 𝑣 ′ ∝ 𝑢′ (the reasoning could be that more

turbulent motion could produce more velocity fluctuation in all
𝑑𝑢 2 𝑑𝑢 2
directions). Thus, 𝑣 ′ ∝ ⋅ 𝑙. That means 𝜏𝑡 = 𝜌𝑙 . We can also
𝑑𝑦 𝑑𝑦
see that 𝜂 = 𝜌𝑙 2 𝑑𝑢/𝑑𝑦.

Therefore, 𝜂 is not an intrinsic fluid property like dynamic viscosity

coefficient 𝜇; its value depends on variables mixing length and velocity
gradient.

For turbulent flow, there is a violent mixing, expect at the wall, i.e.,
𝑙 = 0 at the wall which also means 𝜂 = 0 at the wall. 𝜂 is also called as
coefficient of momentum transfer. Generally, 𝜂 ≫ 𝜇. 11
 Velocity distribution

https://ceng.tu.edu.iq/med/images/chapter_5_heat.pdf

Consider the case fluid flowing over a flat surface. 𝜂 ≪ 𝜇 near the
surface and far from the surface 𝜂 ≫ 𝜇, which makes turbulent flow so
complex.

Assuming that 𝜂 ≪ 𝜇 for 0 ≤ 𝑦 ≤ 𝛿, the shear stress at the surface can

be expressed as 𝜏0 = 𝜇 ⋅ 𝑢/𝑦 (note that for this region, shear stress is
constant since the velocity gradient is constant – velocity is varying
linearly with 𝑦).

Now Write this as 𝜏0 /𝜌 = 𝜇/𝜌 ⋅ 𝑢/𝑦. The term √(𝜏0 /𝜌) is often referred
to as shear velocity 𝒖∗ . Thus, 𝒖/𝒖∗ = 𝒖∗ 𝒚/𝝂 is valid for 0 ≤ 𝑦 ≤ 𝛿.
12
 Velocity distribution
𝑑𝑢 2
When 𝑦 > 𝛿, 𝜂 ≫ 𝜇, and thus 𝜏 = 𝜌𝑙 2 . Considering that shear
𝑑𝑦
2
2 𝑑𝑢
stress in the fluid flow domain is constant, 𝜏 = 𝜏0 = 𝑢∗ 𝜌 =
2
𝜌𝑙 .
𝑑𝑦
Further assuming that 𝑙 = 𝑘 ⋅ 𝑦 (the reasoning is that for a higher 𝑦
turbulence will be higher and, hence the mixing length will be higher),
where 𝑘 is a constant, we can find the expression for velocity by
𝑑𝑢
integrating the equation u∗ = 𝑘𝑦 or 𝑘 ⋅ 𝑑𝑢 = 𝑢∗ 𝑦 −1 𝑑𝑦:
𝑑𝑦

𝟏
𝒖/𝒖∗ = ⋅ 𝐥𝐧 𝒚 + 𝒄𝒐𝒏𝒔𝒕.
𝒌

It should be again noted that the equation is valid for 𝑦 > 𝛿.

13
 Power-law equation for velocity
If 𝑙 is having a slightly non-linear relationship with 𝑦 such that 𝑘 ⋅ 𝑑𝑢 =
𝑢∗ 𝑦 −1+1/𝑛 𝑑𝑦, the velocity profile will be given by the equation

𝟏
𝒖/𝒖∗ = 𝒚𝟏/𝒏
𝒌

assuming that the limits extends till 𝑦 = 0. If flow velocity is known for
a reference point, say at the center of the pipe with radius 𝑅 in case of
pipe flow (𝑈),

𝒖 𝒚 𝟏/𝟕
= .
𝑼 𝑹

Although the equation is pretty analogous to the logarithmic equation in

the previous slide, its handling is a lot more convenient in many
occasions (we will see later).

14