Vous êtes sur la page 1sur 6

THE NAVIER-STOKES EQUATIONS

SIDDHARTH SRINIVASAN

This report is submitted towards the student’s fourth assignment in the course MAT431, and
could very well be called the culmination of this course, considering the importance of the
Navier-Stokes Equations. The video-link leads to my explanation of an OpenFoam tutorial
problem on YouTube. The instructor, Dr. Ajit Kumar is thanked for his valuable guidance in
clarifying key concepts as well as for introducing us to the primary reference [1] in this report.
No claim of originality is made regarding the content of this report or of the video, and the
student is obliged to the cited material, to Dr. Ajit Kumar, and to several sites whose
information the student has recalled from memory.

Abstract. This report attempts to explain the conservation laws for en-
ergy, mass and momentum, which together comprise the famous Navier-Stokes
Equations, used to model fluid flow. An attempt is made to convert all these
conservation equations from its traditional Lagrangian setting to its Eulerian
form. Moreover, a link to a video made by me on the stress analysis of a plate
with a hole is available later in this report.

Introduction
When you apply a large tangential stress, where stress is force per unit area, to
a solid object, the object resists the applied shear stress by deforming. However, in
the case of fluids (liquids and gases), a large stress is not resisted; in fact, it merely
sets them in motion. This can be easily visualized, when one realized that liquids
don’t have a rigid, well-defined shape, and therefore, are not bound to maintain
any spatial dimension under stress. While gaseous particles are light enough and
widely spaced enough to freely fill the entire domain they are present in, liquids
tend to form a free surface in the presence of a gravitational field.
For the purpose of this paper, we shall assume a fluid to be a continuim, not a
large number of particles that are very close together, since fluid flow is a macro-level
process, requiring, therefore, a macro-study of fluids, which makes the distances
between the fluid particles negligibly small. Newtonian fluids are those where the
shear stress is linearly proportional to the shear rate. The molecular viscosity, µ,
gives the slope of the linear function, a physically, represents the ability of the
fluid to resist deformation. In non-Newtonian fluids, however, the relation between
the shear stress and the shear strain is non-linear. In this report, we will study
the Navier-Stokes equations which describe fluid flow, and which are highly non-
linear P.D.Es in four independant variables, since fluid flows in general occur in
three-dimensional space in time.

Classical Mechanics Background


Introduction to the Two Models
In the Lagrangian model, the fluid is segmented into different sections or parcels
and each parcel is followed as it moves through space and time. The parcels are
1
2 SIDDHARTH SRINIVASAN

tagged using a time-independent position vector x0 , usually selected to be the


parcel’s centre of mass, and the flow is modelled by a function x(t, x0 ) of space and
time. The trajectory of a fluid parcel is obtained as the collection of its positions
at different times.
In the Eulerian model, the focus is on a specific region, during the passage of
time. Therefore, the flow is modeled as a function of position x and time t, and the
velocity is represented by v(x,t).
Now, since the derivative of the position of a fluid parcel x0 w.r.t time gives its
velocity, we have that:
δ
v(t, x(t, x0 )) = δt x(t, x0 ),
which relates the Lagrangian and Eulerian models.
Substantial and Local Derivatives
Let φ(t, x(t)) be some physical quantity, for instance density or velocity. The deriv-
ative of this with respect to time, in a fixed position in space, is called the Eulerian
derivative, written as δφ
δt . The derivative that follows a moving parcel is called the
Lagrangian substantial or the material derivative, since it doesn’t concern itself
with a fixed control volume, but with a material volume, and is denoted by Dφ Dt .
The substantial derivative of variable φ, can be obtained by direct application of
the chain rule for multi-variable differentiation, which accounts for changes of φ
w,r.t all the independent variables along the path, and is given below:
Dφ δφ dt δφ dx δφ dz δφ dy
= + + +
Dt δt dt δx dt δz dt δy dt
Dφ δφ δφ δφ δφ
= +u +v +w ,
Dt δt δx δy δz
δφ δφ
where u = δt , v= δt and w = δφ
δt .
Dφ δφ
= + ~v .∇φ,
Dt δt
where ~v = (u, v, w), and ∇φ is the gradient of φ.
Reynolds Transport Theorem
Since the conservation laws primarily apply to moving material volumes and not
to fixed control volumes, there is a need to convert these laws from the Lagrangian
model to the Eulerian approach. The reason for this is that it is much easier to
keep track of a fixed region in space, than to follow several parcels as required in the
Lagrangian model. Moreover, the Lagrangian approach does not allow us to control
the domain of interest, since the fluid particles will only travel as ascertained by
the flow. Let B be the physical property of the fluid that is of interest, say mass or
dB
momentum. Let b = dm be the amount of B per unit mass in any small packet of
the fluid. For an arbitrarily moving and deformable control volume, we know that
the instantaneous total change of B in the material volume MV, is equal to the sum
of the instantaneous total change in the control volume V and the net flow of B
into and out of the control volume V, through the control surface S. Let ρ be the
density of the fluid under consideration, n be the outward normal to the surface of
the control volume, v(t, x) be the velocity of the fluid, and vS (t, x) be the velocity
of the control volume surface. If vr (t, x) is the relative velocity with which the fluid
enters or leaves the control volume, we have that vr (t, x) = v(t, x) − vs (t, x). Now,
the Reynolds Theorem states that:
THE NAVIER-STOKES EQUATIONS 3

R 
dB d
 R
dt M V = dt V (t)
bρdV + S(t)
bρvr .ndS (1)
Leibniz Theorem: Let f : <2 → X be a function, such that f is continuous on S
and fx exists and is continuous on S, where S ⊂ <2 , and {z ∈ < : x0 ≤ z ≤ x1 }
x {y ∈ < : a(x) ≤ y ≤ b(x)}. Now, if a and b are continuous and have continuous
derivatives on {z ∈ < : x0 ≤ z ≤ x1 }, then:
d
R b(x) d d
R b(x) δ
dx a(x) f (x, t)dt = −f (x, a(x)) dx a(x) + f (x, b(x)) dx b(x) + a(x) δx f (x, t)dt
Now, putting a(x), b(x) to be constants, since the geometry of the region under
consideration in an Eulerian approach is time-independent, and also fixing the
control volume, resulting in vS = 0, we get:
R  R
d δ
dt V
bρdV = V δt (bρ)dV (2)
Hence, we obtain, from (1), we now get:
 
dB
R δ R
dt = V δt (bρ)dV + S bρv.ndS (3)
MV
Now, we transform the surface integral in the second integral to the volume integral
using Divergence Theorem, to get:
  R hδ i
dB
dt = V δt
(ρb) + ∇.(ρvb) dV (4)
MV
Alternatively, using the substantial term, we get:
  R hD i
dB
dt = V Dt
(ρb) + ρb∇.v dV (5)
MV

Conservation of Mass
The law of conservation of mass essentially states that in the absence of sources
or sinks (isolated), a system will conserve its mass at a local level. Let the material
volume MV of fluid be of mass m, density ρ, and velocityˇ, then, the conservation
of mass in the Lagrangian coordinate system can be written as:
 
dm
dt = 0 (6)
MV
dB
Now, if m = B, then, we have that dm = dm
dm = 1, and hence, we have, from
equation (5), that:
R h Dρ i
V Dt
+ ρ∇.v dV = 0 (7)
Now, since the integral is over any arbitrary control volume V, the integrand should
be 0 almost everywhere, and since we expect continuity, we expect the integrand to
be 0 everywhere, giving us the differential form of the law of conservation of mass,
or the continuity equation, which is as stated below:

Dt + ρ∇.v = 0 (8)
In order to get the flux form of continuity, we need to convert the equation to it
form using the Eulerian derivative. This is obtained by substituting (6) in (4) to
first get:
R  δρ 
V δt + ∇.[ρv] dV = 0 (9)
As before, since integral is 0 over an arbitrary control volume, the integrand is 0,
giving:
δρ
δt + ∇.[ρv] = 0 (10)
4 SIDDHARTH SRINIVASAN

If the absolute pressure and temperature do not change significantly, we can assume
the flow to be incompressible, that is, the density of the fluid ρ remains almost
constant with respect to changes in pressure. Therefore, we have that Dp Dt = 0, that
is, the density does not change with the flow, and therefore, substituting this in
(8), our continuity equation for incompressible fluids becomes:
R
∇.v = 0 (12) (or) S (v.n)dS = 0 (13)
The last equation, essentially says that in the case of incompressible flows, the net
flow across any control volume is zero. This doesn’t imply that density is constant,
but that each fluid element maintains its density as it moves.

Conservation of Linear Momentum


The law of conservation of linear momentum states that, assuming no action of
external force on the body under consideration, the total momentum of the body,
m~v , remains unchanged. Of course, the statement implies that the direction of all of
the momentum-components are retained. Newton’s Second Law of Motion, which
states that F~net = m~a, says that the momentum of the volume under consideration
can change only in the presence of a net force acting on it. Therefore, if MV is
the material volume, of mass m, density ρ, and velocity v, then, the second law in
Lagrangian coordinates gives:
  R 
d(mv)
dt = V
f dV , (14)
MV MV
where f is the net external force per unit volume acting on MV. Moreover, since
the R.H.S in (14) is the volume integral over material coordinates performed over
the instantaneous volume occupation by the fluid, we have that:
R  R
V
f dV = V f dV (15)
MV
Non-conservative Form of Momentum Conservation
We have that b = v, since d(mv)
dm = v. Now, using (5), we get:
 h i
D
R
V
f − Dt (ρv) + vρ∇.v
bigg)dV = 0 (16)
Like before, since the integral is 0, we have, by an argument analogous to the one
before, that the integrand is also 0, and hence, we have:
D
Dt [ρv] + [ρv∇.v] = f (17)
Expanding the substantial of momentum, and doing some rearrangement, we get:
 

ρ Dv
Dt + v Dt + ρ∇.v = f (18)
Now, the second term on the left hand side, by the continuity constraint is 0, and
we know the expansion of the material derivative, both of which now give us:
h i
ρ δvδt + (v.∇)v = f (19)
Conservative Form of Momentum Conservation
In this case, we apply the form of the Reynolds Transport Theorem that has been
given in (4), which gives that:
R hδ i
V δt
[ρv] + δ.ρvv − f dV = 0 (20)
Using a similar argument that the integrand will be 0, since the integral is 0 over
any arbitrary control volume V, we get the following result:
THE NAVIER-STOKES EQUATIONS 5

δ
δt [ρv] + ∇.ρvv = f (21)

Conservation of Energy
We define the total energy E of MV at a given time t to be the sum
 of theinternal
and kinetic energies of the material volume, thereby giving E = m û+ 21 v.v , where
û is the internal energy per unit mass. The 1st law of thermodynamics, applied to
a material volume states that the rate of change of the total energy of the material
volume is equal to the rate of heat addition or work done through its boundaries,
giving:
 
dE
dt = Q̇ − Ẇ (22)
MV
The net rate of heat transfer has two contributing components- one being the
transfer rate along the surface of MV and the other being the rate of creation or
destruction of heat within the system by sources or sinks- denoted by Q̇S and Q̇V
respectively. In addition, the work done by MV is a sum of the rate of work done
by surface forces ẆS and the rate of work done by body forces Ẇb . This gives us
a re-writing of (22) as follows:
 
dE
dt = Q̇V + Q̇S − Ẇb − ẆS (23)
MV
, which is essentially the form for the law of conservation of energy. Of course, one
may expand on the ẆS and the Q̇V terms further, but such computations would
only serve to expand the equation into several smaller computations, and hence,
which I exclude from this report.

Generalized Conservation Equation


Let us assume that φ is the physical quantity which is under consideration, i.e.,
the quantity that is being conserved. We have that change of φ in a time ∆t
within the control volume M V , call it α is equal to the sum of the surface flux of
φ over time ∆t across the control volume, call it β and the source/sink (creation
or destruction) of φ in time ∆t within the control volume, call it γ. Now, using
Reynolds Transport Theorem, we have that:
R  R h i
d δ
α = dt MV
(ρφ)dV = V δt (ρφ) + ∇.(ρvφ) dV , (24)
where ρ is the density of the fluid and V is the control volume, with surface area
φ
S. Now, we have that Jconvection = ρvφ, which represents the transport of φ by
the flow field. Moreover, as shown in an earlier assignment (Assignment 2, P.D.E,
φ φ
2017), using Fick’s First Law, we have that Jdif f usion = −I ∇φ, and therefore, we
have that:
R φ R φ R φ
β = − S Jdif f usion .ndS = − V ∇.Jdif f usion dV = V ∇.(I ∇φ)dV , (25)
where the negative sign is sign-convention, to denote that the flux is outward. We
also have that:
γ = V Qφ dV , (26)
R

where Qφ is the source or sink contribution to generation or removal of φ within


the control volume V. This results in the general conservation equation after some
apparent rearrangement of terms, as given below:
R hδ i
V δt
(ρφ) + ∇.(ρvφ) − ∇.(I φ ∇φ) − Qφ dV = 0 (27)
6 SIDDHARTH SRINIVASAN

As done before, since the integral is 0 on any arbitrary control volume, the integrand
must be zero, giving the following continuity equation:
δ φ φ
δt (ρφ) + ∇.(ρvφ) − ∇.(I ∇φ) − Q = 0 (28)
Thus, the objectives of this report have been realized in giving the Eulerian form
of the Navier-Stokes Equations in general.

Video Link
The video link to my tutorial explanation, titled ’Stress Analysis on a Plate with
a Hole’ is as given below:
https://www.youtube.com/watch?v=YvQU7njEEfMfeature=youtu.be

Additional Acknowledgements
I would like to thank OpenFoam, Cheese-Cam and Paraview, for their excellent
applications that made the video possible.

References
1. Moukallad, F., Mangani, L., Darwish, M..The Finite Volume Method in Computational Fluid
Dynamics, Ch. 3: Mathematical Description of Physical Phenomena, Pg. 43-67. Fluid Me-
chanics and Its Applications, Volume 113. Springer International Publishing Switzerland,
2016.
2. Tom M. Apostol, Calculus: Volume 1

Vous aimerez peut-être aussi