UNIVERSITY OF SOUTHAMPTON

SCHOOL OF ENGINEERING SCIENCES

AERODYNAMICS & FLIGHT MECHANICS GROUP
Governing equations for a new compressible
Navier-Stokes solver in general cylindrical
coordinates
by
Richard D. Sandberg
Report No. AFM-07/07
November 2007
COPYRIGHT NOTICE -
All right reserved. No parts of this publication may be reproduced, stored in a retrieval
system, or transmitted, in any form or by any means, electronic, mechanical, photocopying,
recording, or otherwise, without the permission of the Head of the School of Engineering
Sciences, University of Southampton, Southampton SO17 1BJ, U.K.
c School of Engineering Sciences, University of Southampton
The aim of this report is to derive the governing equations for a new compressible Navier-
Stokes solver in general cylindrical coordinates, i.e. the streamwise and radial directions
are mapped to general coordinates. A literature review revealed that simulations of com-
plex cylindrical geometries have mostly been conducted using general curvilinear coordi-
nates (e.g. DeBonis & Scott, 2002). The method presented in this report is chosen over
three-dimensional general curvilinear coordinates as it requires a smaller number of terms
to be computed. Furthermore, only two-dimensional metric terms need to be stored,
reducing the required allocated memory and therefore resulting in a more ecient code.
1 Governing equations in cylindrical coordinates
The compressible Navier-Stokes equations consist of conservation of mass, momentum
and total energy. The ow is assumed to be an ideal gas with constant specic heat
coecients. All quantities are made dimensionless using the ow quantities at a reference
location in the ow; here the free-stream/inow location is used. The radius of the body
is chosen as the reference length. The non-dimensionalization results in the following
dimensionless parameters:
Re =

, M =
u

, Pr =

.
With z, r, denoting the streamwise, radial and azimuthal directions, respectively, and
u, v, w denoting the velocity components in z, r, directions, respectively, the non-dimensional
compressible NavierStokes equations in cylindrical coordinates are
U
t
+
A
z
+
B
r
+
1
r
C

+
1
r
D = 0 (1)
U =
_
_
_
_
_
_
_
_

u
v
w
E
_
_
_
_
_
_
_
_
(2)
A =
_
_
_
_
_
_
_
_
u
uu + p
zz
uv
rz
uw
z
uH + q
z
u
zz
v
rz
w
z
_
_
_
_
_
_
_
_
B =
_
_
_
_
_
_
_
_
v
uv
rz
vv + p
rr
vw
r
vH + q
r
u
rz
v
rr
w
r
_
_
_
_
_
_
_
_
C =
_
_
_
_
_
_
_
_
w
uw
z
vw
r
ww + p

wH + q

u
z
v
r
w

_
_
_
_
_
_
_
_
D =
_
_
_
_
_
_
_
_
v
uv
rz
vv ww
rr
+

2vw 2
r
vH + q
r
u
rz
v
rr
w
r
_
_
_
_
_
_
_
_
,
where the total energy is dened as E = T/[( 1)M
2
] +1/2u
i
u
i
with = 1.4, and the
total enthalpy is H = E + p/.
1
The molecular stress tensor components are

zz
=
2
3Re
_
2
u
z

v
r

1
r
_
w

+ v
__
(3)

rr
=
2
3Re
_

u
z
+ 2
v
r

1
r
_
w

+ v
__
(4)

=
2
3Re
_

u
z

v
r
+ 2
1
r
_
w

+ v
__
(5)

rz
=

Re
_
u
r
+
v
z
_
(6)

z
=

Re
_
w
z
+
1
r
u

_
(7)

r
=

Re
_
1
r
_
v

w
_
+
w
r
_
. (8)
The heat-ux vector components are
q
z
=

Pr( 1)M
2
Re
T
z
(9)
q
r
=

Pr( 1)M
2
Re
T
r
(10)
q

=

Pr( 1)M
2
Re
1
r
T

, (11)
where the Prandtl number is assumed to be constant at Pr = 0.72. The molecular
viscosity is computed using Sutherlands law (c.f. White, 1991), setting the ratio of
the Sutherland constant over freestream temperature to 0.36867. To close the system
of equations, the pressure is obtained from the non-dimensional equation of state p =
(T)/(M
2
).
2 Transformation to general cylindrical coordinates
In order to allow for complex geometries, the (r, z) plane is mapped to general coordinates
(, ). Hence, streamwise and radial derivatives need to be expressed in terms of the new
variables. By using the chain rule, the following expressions can be derived

z
=
1
J
_
r

_
= r

, (12)

r
=
1
J
_

+
z

_
= z

, (13)
where J is the determinant of the coordinate transformation. For conciseness, the metric
terms are abbreviated as, e.g., r

=
r

so that J = z

, and the asterisk denotes

a metric term already divided by J.
2
Following Anderson (1995), the r and z derivatives of equation (1) can be transformed as
follows (provided that the grid transformation does not vary in time)
U
t
=
1
J
_

(Ar

Bz

) +

(Ar

+ Bz

)
_

1
r
C

1
r
D (14)
Unfortunately, it is not straightforward to plug in the vectors A and B into (14), as
they contain terms that are pre-multiplied by 1/r. As r = r(, ), when taking and
derivatives of A and B the chain rule needs to be applied. To illustrate this procedure,
the derivation of the radial momentum equation is chosen as example. Here, we only
focus on the derivatives in the and directions,

_
(uv
rz
) r

(vv + p
rr
) z

_
(uv
rz
) r

(vv + p
rr
) z

_
. (15)
The only term containing 1/r is the stress tensor component
rr
. Therefore, the equation
is separated into

_
(uv
rz
) r

_
vv + p
2
3Re
_

u
z
+ 2
v
r
__
z

_
1
r
2
3Re
_
w

+ v
_
z

_
(uv
rz
) r

_
vv + p
2
3Re
_

u
z
+ 2
v
r
__
z

_
+

_
1
r
2
3Re
_
w

+ v
_
z

_
.
(16)
Applying the chain rule to the last term of each row results in

_
1
r
2
3Re
_
w

+ v
_
z

_
=
1
r

_
2
3Re
_
w

+ v
_
z

1
r
2
r

2
3Re
_
w

+ v
_

_
1
r
2
3Re
_
w

+ v
_
z

_
=
1
r

_
2
3Re
_
w

+ v
_
z

1
r
2
r

2
3Re
_
w

+ v
_
Substituting these two expressions into (16) yields

_
(uv
rz
) r

_
vv + p
2
3Re
_

u
z
+ 2
v
r
__
z

_
(uv
rz
) r

_
vv + p
2
3Re
_

u
z
+ 2
v
r
__
z

1
r

_
2
3Re
_
w

+ v
_
z

_
+
1
r

_
2
3Re
_
w

+ v
_
z

_
+
1
r
2
2
3Re
_
w

+ v
_
(r

)
. .
J
. (17)
The last term of (17) occurs because of the partial derivative with respect to r of
rr
. In
the case of streamwise derivatives

z
, we get (r

) = 0, hence 1/r
2
terms cancel.
The resulting governing equations are now summarized. For brevity, the inner derivatives
are not written in terms of , , however the streamwise and radial derivatives need to be
computed with equations (12) and (13).
3
2.1 Continuity equation

t
=
1
J
_

(ur

vz

) +

(ur

+ vz

)
_

1
r
(w)

(v)
r
(18)
2.2 Streamwise momentum equation
(u)
t
=
1
J
_

__
uu + p
2
3Re
_
2
u
z

v
r
__
r

(uv
rz
) z

_
(19)

__
uu + p
2
3Re
_
2
u
z

v
r
__
r

(uv
rz
) z

_
+
1
r

_
2
3Re
_
w

+ v
_
r

1
r

_
2
3Re
_
w

+ v
_
r

_ _

1
r

_
uw
z
_

1
r
_
uv
rz
_
2.3 Radial momentum equation
(v)
t
=
1
J
_

_
(uv
rz
) r

_
vv + p
2
3Re
_

u
z
+ 2
v
r
__
z

_
(20)

_
(uv
rz
) r

_
vv + p
2
3Re
_

u
z
+ 2
v
r
__
z

1
r

_
2
3Re
_
w

+ v
_
z

_
+
1
r

_
2
3Re
_
w

+ v
_
z

_ _
+
1
r
2
2
3Re
_
w

+ v
_

1
r

_
vw
r
_

1
r
_
(vv ww)
2
Re
_
v
r

1
r
_
w

+ v
___
4
2.4 Azimuthal momentum equation
(w)
t
=
1
J
_

__
uw

Re
w
z
_
r

_
vw

Re
w
r
_
z

_
(21)

__
uw

Re
w
z
_
r

_
vw

Re
w
r
_
z

1
r

_

Re
_
u

_
v

w
_
z

__
+
1
r

_

Re
_
u

_
v

w
_
z

__ _

1
r
2

Re
_
v

w
_

1
r

_
ww + p

1
r
_
2vw 2
r
_
2.5 Energy equation
(E)
t
=
1
J
_

__
uH + q
z
u
2
3Re
_
2
u
z

v
r
_
v
rz
w

Re
w
z
_
r

(22)

_
vH + q
r
u
rz
v
2
3Re
_
2
v
r

u
z
_
w

Re
w
r
_
z

__
uH + q
z
u
2
3Re
_
2
u
z

v
r
_
v
rz
w

Re
w
z
_
r

_
vH + q
r
u
rz
v
2
3Re
_
2
v
r

u
z
_
w

Re
w
r
_
z

_
+
1
r

_

Re
_
u
2
3
_
w

+ v
_
w
u

_
r

Re
_
2
3
v
_
w

+ v
_
w
_
v

w
__
z

1
r

_

Re
_
u
2
3
_
w

+ v
_
w
u

_
r

Re
_
2
3
v
_
w

+ v
_
w
_
v

w
__
z

_ _
+
1
r
2

Re
_
2
3
v
_
w

+ v
_
w
_
v

w
__

1
r

_
wH + q

u
z
v
r
w

1
r
_
vH + q
r
u
rz
v
rr
w
r
_
5
3 Vector form of general cylindrical coordinates
The equations in general cylindrical coordinates can be written as
U
t
=
1
J
{
_

_

Ar

Bz

_
+

Ar

+

Bz

_
_
+
1
r
_

_

A
r
r

B
r
z

_
+

A
r
r

+

B
r
z

_
__

1
r
C

1
r
D
1
r
2

B
rr
, (23)
where

A =
_
_
_
_
_
_
_
_
_
u
uu + p
2
3Re
_
2
u
z

v
r
_
uv
rz
uw

Re
w
z
uH + q
z
u
2
3Re
_
2
u
z

v
r
_
v
rz
w

Re
w
z
_
_
_
_
_
_
_
_
_
,

B =
_
_
_
_
_
_
_
_
_
v
uv
rz
vv + p
2
3Re
_

u
z
+ 2
v
r
_
vw

Re
w
r
vH + q
r
u
rz
v
2
3Re
_

u
z
+ 2
v
r
_
w

Re
w
r
_
_
_
_
_
_
_
_
_
,

A
r
=
_
_
_
_
_
_
_
_
_
0
2
3Re
_
w

+ v
_
0

Re
u

Re
_
u
2
3
_
w

+ v
_
w
u

_
_
_
_
_
_
_
_
_
_
,

B
r
=
_
_
_
_
_
_
_
_
_
0
0
2
3Re
_
w

+ v
_

Re
_
v

w
_

Re
_
2
3
v
_
w

+ v
_
w
_
v

w
__
_
_
_
_
_
_
_
_
_
,

B
rr
=
_
_
_
_
_
_
_
_
_
0
0

2
3Re
_
w

+ v
_

Re
_
v

w
_

Re
_
w
_
v

w
_

2
3
v
_
w

+ v
__
_
_
_
_
_
_
_
_
_
.
In order to make the new code most ecient, a good compromise between the number
of arithmetic operations and the number of three dimensional arrays stored needs to be
found. The emphasis here lies on reducing the number of stored arrays to a minimum
while minimizing the number of repeated operations. One choice is to simply store all
nine velocity derivatives. However, for this option, the number of operations is high as the
stress tensor components constantly need to be reassembled. The most ecient choice of
6
arrays to store appears to be the following set
1)
zz
=
2
3Re
_
2
u
z

v
r
_
2)
rz
=

Re
_
u
r
+
v
z
_
3)
z
=

Re
w
z
4)
rr
=
2
3Re
_
2
v
r

u
z
_
5)
r
=

Re
w
r
6)
1
=
2
3Re
_
w

+ v
_
7)
2
=

Re
u

8)
3
=

Re
_
v

w
_
.
Using these eight arrays in addition to the primitive variables and the heat-ux vector,

A,

B,

A
r
,

B
r
and

B
rr
can be assembled. In order to evaluate C and D, the following
relations can be used

z
=
z
+
1
r

2
,
r
=
r
+
1
r

3
,
rr
=
rr

1
r

1
,

=
zz

rr
+
2
r

1
. (24)
Finally, all vectors required for (23) can be written as

A =
_
_
_
_
_
_
_
_
u
uu + p
zz
uv
rz
uw
z
uH + q
z
u
zz
v
rz
w
z
_
_
_
_
_
_
_
_
,

B =
_
_
_
_
_
_
_
_
v
uv
rz
vv + p
rr
vw
r
vH + q
r
u
rz
v
rr
w
r
_
_
_
_
_
_
_
_
,

A
r
=
_
_
_
_
_
_
_
_
0

1
0

2
u
1
w
2
_
_
_
_
_
_
_
_
,

B
r
=
_
_
_
_
_
_
_
_
0
0

3
v
1
w
3
_
_
_
_
_
_
_
_
,

B
rr
=
_
_
_
_
_
_
_
_
0
0

3
w
3
v
1
_
_
_
_
_
_
_
_
,
C =
_
_
_
_
_
_
_
_
w
uw (
z
+
1
r

2
)
vw (
r
+
1
r

3
)
ww + p (
zz

rr
+
2
r

1
)
wH + q

u(
z
+
1
r

2
) v(
r
+
1
r

3
) w(
zz

rr
+
2
r

1
)
_
_
_
_
_
_
_
_
,
D =
_
_
_
_
_
_
_
_
v
uv
rz
vv ww (
rr

1
r

1
) + (
zz

rr
+
2
r

1
)
2vw 2(
r
+
1
r

3
)
vH + q
r
u
rz
v(
rr

1
r

1
) w(
r
+
1
r

3
)
_
_
_
_
_
_
_
_
.
References
Anderson, J. 1995 Computational Fluid Dynamics, The Basics with Applications.
McGraw-Hill.
DeBonis, J. & Scott, J. N. 2002 Large-eddy simulation of a turbulent compressible
round jet. AIAA J. 40 (7), 13461354.
White, F. M. 1991 Viscous Fluid Flow. McGraw Hill.
7