Dimensionless form of equations

Motivation: sometimes equations are normalized in order to

facilitate the scale-up of obtained results to real flow conditions
avoid round-off due to manipulations with large/small numbers
assess the relative importance of terms in the model equations
Dimensionless variables and numbers
t =

t
,
t0

x =

x
,
L0

v =

v
,
v0

p =

p
,
v02

T =

T T0
T1 T0

Mach number

M=

|v|
c

inertia
gravity

Strouhal number

St =

L0
v0 t0

convection
diffusion

Prandtl number

Pr =

Reynolds number

Re =

v0 L0

inertia
viscosity

Froude number

Fr =

v0
L0 g

Peclet number

Pe =

v0 L0

Model simplifications
Objective: derive analytical solutions / reduce computational cost
Compressible Navier-Stokes equations
= const

=0

Incompressible Navier-Stokes equations

Stokes flow

boundary layer

Compressible Euler equations

inviscid Euler equations

potential flow

Derivation of a simplified model

1. determine the type of flow to be simulated
2. separate important and unimportant effects
3. leave irrelevant features out of consideration
4. omit redundant terms/equations from the model
5. prescribe suitable initial/boundary conditions

Viscous incompressible flows

Simplification:

= const,

= const

continuity equation

inertial term

(v)
t

stress tensor

= (v + vT ) = ( v + v) = v

+ (v) = 0

+ (v v) =

Let

z0

v =0
 v
t


+ v v = dv
dt

g = gk = (gz) = p0

p0 = g(z0 z)
p =

pp0

= /

g(z0 z)

hydrostatic pressure
reduced pressure
kinematic viscosity

Incompressible Navier-Stokes equations

v
+ v v =
p + v g(T T0 )
|
{z
}
t

momentum equations

Boussinesq

v =0

continuity equation

Natural convection problems

Internal energy equation

= (T ),

v = 0,

= const

e
+ v e = T + q + v : (v + vT ),
t

Temperature equation

(convection-diffusion-reaction)

T
+ v T =
T + q,
t
Linearization:
0 = (T0 ),

e = cv T

=


(T ) = (T0 ) +
 

1
0 T

T =T

,
cv

T =T

q =

q
+
|v + vT |2
cv 2cv

(T T0 )

Taylor series

thermal expansion coefficient

Boussinesq approximation for buoyancy-driven flows

[1 (T T )] in the term g
0
0
(T ) =
0
elsewhere

Viscous incompressible flows

Stokes problem (Re 0, creeping flows)
dv
v

0
dt
t

v
=
p + v
t
v =0

momentum equations
continuity equation

Boundary layer approximation (thin shear layer)

y

pipe flow
v

v
t

v = (u, v)

0 and u v

xu2 can be neglected

p
y

0 p = p(x)

Navier-Stokes equations
p
2u
=
+ 2
x
y
p
0 =
y
u v
+
= 0
x y

u
u
u
+v
x
y

Inviscid incompressible flows

Incompressible Euler equations

Euler equations

potential ow
y

=0

Irrotational / potential flow

boundary layer

v
+ v v =
p
t
v =0

= onst
' = onst

=v =0

(vanishing vorticity)

such that v = and v = = 0

Laplace equation

in 2D there also exists a stream function such that u =

y ,

Computation of the pressure

1
v v = v v + (v v) =
2
v
=0
t

|v|2
p =
2

|v|2
2

Bernoulli equation

v =
x

Compressible Euler equations

Simplifications:

= 0,

= 0,

g=0

Divergence form

Quasi-linear formulation

U
+F=0
t

U
+ A U = 0
t

Conservative variables and fluxes

U = (, v, E)

F = (F 1 , F 2 , F 3 )

F = A U,

F=
v v + pI
hv

A = (A1 , A2 , A3 )

Jacobian matrices

F d
A =
,
U
d

d = 1, 2, 3

h=E+

= cp /cv

Equation of state


p = ( 1) E |v|2 /2

Classification of partial differential equations

PDEs can be classified into hyperbolic, parabolic and elliptic ones
each class of PDEs models a different kind of physical processes
the number of initial/boundary conditions depends on the PDE type
different solution methods are required for PDEs of different type
Hyperbolic equations

Information propagates in certain directions at

finite speeds; the solution is a superposition of multiple simple waves

Parabolic equations

Information travels downstream / forward in time;

the solution can be constructed using a marching / time-stepping method

Elliptic equations Information propagates in all directions at infinite speed;
describe equilibrium phenomena (unsteady problems are never elliptic)

Classification of partial differential equations

u
u
a0 + a1 x
=0
+
.
.
.
+
a
D
x
1
D

First-order PDEs

X u
2u
bk
+
+ cu + d = 0
aij

x
x
x
i
j
k
i,j=1

elliptic
hyperbolic
parabolic

aij = aji ,
n+

k=1

A = {aij } RDD ,

coefficient matrix:

PDE type

D
X

Second order PDEs

symmetry:

are always hyperbolic

otherwise set
n0

D1

D1

aij = aij (x1 , . . . , xD )

aij = aji :=

aij +aji
2

n+

number of positive eigenvalues

number of negative eigenvalues

n0

number of zero eigenvalues

n+ n

Classification of second-order PDEs

2u
2u
2u
a22 2 + . . . = 0
a11 2 (a12 + a21 )
x1
x1 x2
x2

2D example

D = 2,

A=

a11

a12

a21

a22

det A = a11 a22 a212 = 1 2

2

elliptic type

det A > 0

xu2

2u
y 2

=0

Laplace equation

hyperbolic type

det A < 0

2u
t2

2u
x2

=0

wave equation

parabolic type

det A = 0

u
t

2u
x2

=0

diffusion equation

mixed type

det A = f (y)

y xu2

2u
y 2

=0

Tricomi equation

Classification of first-order PDE systems

U
U
+ . . . + AD x
=B
A1 x
1
D

Quasi-linear form
Plane wave solution

eis(x,t) ,
U =U

U Rm ,

= const,
U

m>1

s(x, t) = n x

where n = s is the normal to the characteristic surface s(x, t) = const

B=0

D
P

d=1

Hyperbolic systems

= 0,
nd A d U

det

D
P

d=1

nd Ad = 0 n(k)

There are D real-valued normals n(k) , k = 1, . . . , D

(k) of the associated systems are linearly independent

and the solutions U
(k)
Parabolic systems There are less than D real-valued solutions n(k) and U
Elliptic systems

No real-valued normals n(k) no wave-like solutions

Second-order PDE as a first-order system

2

Equivalent first-order system for

a u + 2b u + c v = 0
x
y
y

u + v = 0
y

Matrix form

a x2 + 2b xy
+ c y2 = 0

Quasi-linear PDE of 2nd order

U
A1 U
x + A2 y = 0,

u=

x ,

v=


a 0
0 1
| {z }

A1

u
2b c
+
v
1 0
| {z } | {z }
U

A2

einx = U
ei(nx x+ny y)
U =U

u
=0
v
| {z }
U

plane wave

= 0 admits nontrivial solutions if

[nx A1 + ny A2 ]U


anx + 2bny cny
det[nx A1 + ny A2 ] = det
= 0 an2x + 2bnx ny + cn2y = 0
ny
nx

The resulting problem

nx
ny

2

+ 2b

nx
ny

+c=0

nx
b b2 4ac
=
ny
2a

Second-order PDE as a first-order system

Characteristic lines
y

s
s
dx +
dy = nx dx + ny dy = 0
tangent
ds =
x
y

dy
nx
b b2 4ac
=
=
curve
dx
ny
2a

y(x)

The PDE type depends on D = b2 4ac

D>0

two real characteristics

D=0

just one root

D<0

no real characteristics

dy
dx

hyperbolic equation

b
2a

parabolic equation
elliptic equation

Transformation to an unsteady system

U
U
+ A
= 0,
x
y
det[A I] = 2

A = A1
1 A2 =


2b
a

c
+ =0
a

1
a

0
0 1



2b c
1 0

1,2 =

2b
a

c
a

b2 4ac
2a

Geometric interpretation for a second-order PDE

which may influence the solution at point P
Domain of dependence: x
which are influenced by the solution at point P
Zone of influence: x
Hyperbolic PDE

A
domain of
dependence

B
C

1111111111
0000000000
0000000000
1111111111
0000000000
1111111111
0000000000
1111111111
0000000000
1111111111
0000000000
1111111111
0000000000
1111111111
0000000000
1111111111
0000000000
1111111111
0000000000
1111111111
0000000000
1111111111

0000000000000
1111111111111
1111111111111
0000000000000
0000000000000
1111111111111
0000000000000
1111111111111
0000000000000
1111111111111
0000000000000
1111111111111
0000000000000
1111111111111
0000000000000
1111111111111
0000000000000
1111111111111
0000000000000
1111111111111
0000000000000
1111111111111
0000000000000
1111111111111
0000000000000
1111111111111
0000000000000
1111111111111
0000000000000
1111111111111
0000000000000
1111111111111
0000000000000
1111111111111
0000000000000
1111111111111
0000000000000
1111111111111
0000000000000
1111111111111
0000000000000
1111111111111
0000000000000
1111111111111
0000000000000
1111111111111
0000000000000
1111111111111
0000000000000
1111111111111
0000000000000
1111111111111
0000000000000
1111111111111
0000000000000
1111111111111
0000000000000
1111111111111
0000000000000
1111111111111
0000000000000
1111111111111
0000000000000
1111111111111
0000000000000
1111111111111

zone of influence

Parabolic PDE

Elliptic PDE
domain of dependence

A
domain of
dependence

11111111111111
00000000000000
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111

zone of influence

1111111111111111111111
0000000000000000000000
0000000000000000000000
1111111111111111111111
0000000000000000000000
1111111111111111111111
0000000000000000000000
1111111111111111111111
0000000000000000000000
1111111111111111111111
0000000000000000000000
1111111111111111111111
0000000000000000000000
1111111111111111111111
0000000000000000000000
1111111111111111111111
0000000000000000000000
1111111111111111111111
0000000000000000000000
1111111111111111111111
0000000000000000000000
1111111111111111111111
0000000000000000000000
1111111111111111111111
0000000000000000000000
1111111111111111111111
0000000000000000000000
1111111111111111111111
0000000000000000000000
1111111111111111111111
0000000000000000000000
1111111111111111111111
0000000000000000000000
1111111111111111111111
0000000000000000000000
1111111111111111111111
0000000000000000000000
1111111111111111111111
0000000000000000000000
1111111111111111111111
0000000000000000000000
1111111111111111111111
0000000000000000000000
1111111111111111111111
0000000000000000000000
1111111111111111111111
0000000000000000000000
1111111111111111111111
0000000000000000000000
1111111111111111111111
0000000000000000000000
1111111111111111111111
0000000000000000000000
1111111111111111111111

zone of influence

steady supersonic flows

steady boundary layer flows

steady subsonic/inviscid

unsteady inviscid flows

unsteady heat conduction

incompressible flows

Space discretization techniques

Objective: to approximate the PDE by a set of algebraic equations

stationary (elliptic) PDE

Lu = f
in

u=g
on 0
Dirichlet boundary condition
0

n u = g1
on 1
Neumann boundary condition

n u + u = g on
Robin boundary condition
2

Boundary value problem

BVP = PDE + boundary conditions

Getting started: 1D and 2D toy problems

1. u = f

Poisson equation

2. (uv) = (du)

convection-diffusion

Computational meshes
Degrees of freedom for the approximate solution are defined on a computational
mesh which represents a subdivision of the domain into cells/elements

structured

block-structured

unstructured

Structured (regular) meshes

families of gridlines do not cross and only intersect with other families once
topologically equivalent to Cartesian grid so that each gridpoint (or CV) is
uniquely defined by two indices in 2D or three indices in 3D, e.g., (i, j, k)
can be of type H (nonperiodic), O (periodic) or C (periodic with cusp)
limited to simple domains, local mesh refinement affects other regions

Computational meshes
Block-structured meshes
multilevel subdivision of the domain with structured grids within blocks
can be non-matching, special treatment is necessary at block interfaces
provide greater flexibility, local refinement can be performed blockwise
Unstructured meshes
suitable for arbitrary domains and amenable to adaptive mesh refinement
consist of triangles or quadrilaterals in 2D, tetrahedra or hexahedra in 3D
complex data structures, irregular sparsity pattern, difficult to implement

Discretization techniques
Finite differences / differential form
approximation of nodal derivatives
simple and effective, easy to derive
limited to (block-)structured meshes
Finite volumes / integral form
approximation of integrals
conservative by construction
suitable for arbitrary meshes
Finite elements / weak form
weighted residual formulation
remarkably flexible and general
suitable for arbitrary meshes