Sie sind auf Seite 1von 7

CONFIDENTIAL Error! Reference source not found.

: Computational modelling
and methodology
1
Chapter 1
Computational modelling and methodology
In this chapter, computation methods, convergence criteria and validation of codes are
described. Computation fluid dynamics is described in section 1.1. In this work all the
simulations are carried out with air as fluid which is considered to follow the ideal gas
law. Hence, the description is restricted to compressible flow only. In the CFD
section, a short description of the basic flow equations are described in section 1.1.1,
turbulence modelling used in this work is given in section 1.1.2, numerical
descritization and algorithms used in this thesis are described in section 1.2.
1.1 CFD
Computational fluid dynamics (CFD) is the analysis of system involving fluid flow,
heat transfer and associated phenomena such as chemical reactions by means of
computer based simulations (Versteeg (1996)). Current capabilities and future
demands of the CFD as a tool for solving gas turbine engines problems are illustrated
in Chew and Nick (2007 and 2009). In the next section 1.1.1 basic governing
equations for the fluid and heat transfer processes are described. In this work CFD is
used as a tool to solve the fluid flow and heat transfer process in the internal air
system of the gas turbine engine. Fluid motion is governed by basic
1.1.1 Flow governing equations
For continuum fluid flow assumption, most of the flow can be explained by basic
conservation laws of mass, momentum and energy balance equations. In CFD
literatures (Hirsch (2007)), the mass conservation law referred as continuity
equation, the momentum conservation law is generalize expression of Newton law
referred as equation of motion for the fluid and energy conservation law is referred
as first principle of thermodynamics. For generalise viscous fluid all these
equations are known as Navier-Stokes equations given as follow,
CONFIDENTIAL Error! Reference source not found.: Computational modelling
and methodology
2
)
V 0
p
p

+\ =
t
)&
(1-1)
)
) ) )
V
V V
e
pI f
p
p t p

+\ - = \ +\ +
t
)&
)& )& ))&

(1-2)
)
) ) ) )
V V V V
e
E
E p k t f q
p
p t p p

+\ = \ + \ \ +\ + +
t
)& )& )& )& ))& )&

(1-3)
where, is the gradient operator, is the divergence operator, I is unit tensor, t is
time, t is temperature, p is pressure, k is thermal conductivity,

is velocity vector, V
is velocity magnitude,

t

is shear stress tensor,
e
f
))&
is body force vector per unit mass,
E (= e + V
2
/2) is total internal energy per unit mass, e is static internal energy per unit
mass and q is rate of volumetric addition per unit mass.
The shear stress tensor for the Newtonian fluid can be expressed in terms of the
velocity gradient given as follow,
)
V
j
i
ij ij
i j
v
v
x x
P
t t o

+

= = + + \



' ' |
)&
(1-4)
where, is Kronecker delta (when i=j, =1 and ij, =0), is dynamic viscosity and


is second viscosity coefficient (volumetric deformation). According to Stokes
hypothesis and

are related by following expression.

2 3 0
P
+ = (1-5)
By rearranging Equations (1-1), (1-2) and (1-3) could be rearrange write in following
conservative form of equation for general variable 1,
)
)
) V S
p
p
1
1
+\ 1 = \ I \1 +
t
)&
(1-6)
This is equation also called the transport equation for scalar variable 1, which could
be either 1, V
x
, V
y
, V
z
or e. The first term in left side of this equation is rate of
change, second term is convective or advection and right side first term is
diffusion and second term is source of the scalar 1. These five Navier-Stokes
(one mass, three momentum and one energy balance) equations are for seven
unknown basic variables , p, V
x,
V
y,
V
z
and t. Additional two equations comes from
CONFIDENTIAL Error! Reference source not found.: Computational modelling
and methodology
3
thermodynamics laws. For an ideal gas assumption, following two expressions are
valid, which are closed the system of equation for seven variables.
p Rt p = (1-7)
v
e c t = (1-8)
where, R is the gas constant and c
v
is heat capacity at constant volume. For air R =
289.897 J/(kg K).
The dynamic viscosity , and thermal conductivity k, in above equations are also
depends on the fluid state. The is strongly influence by temperature than pressure.
For the air, widely used correlation given by Sutherlands formula (moinier (is
considered in this work as follow,
3 2
5
1.461 10
110.3
t
t


= -
+
(1-9)
As mentioned all these equation are for compressible flow, but for the incompressible
flow case, the energy equation is not linkage to continuity equation and momentum
equation. Because of the density is constant, not depends on the pressure or
temperature. Hence for incompressible flow, the energy equation can be solved
separately based on the calculated pressure and velocity field for heat transfer
calculations.
1.1.2 Turbulence modelling
At low Reynolds number (Re = UL/, where U and L are characteristic velocity and
length scale of the mean flow) flow is laminar. This laminar flow regime could be
explain well with the Navier-Stokes equation developed in previous section. But at
higher Reynolds number flow becomes turbulent in nature. Most of the engineering
interest flows are in the turbulence flow regime. Hence it is required to understand the
turbulence flow and develop mathematical modelling to obtain quantities of interest
and practical relevance.
In turbulent flow, velocity field is three dimensional, time-dependent and random.
There are rotational flow structures, called as turbulence eddies observed in turbulent
flow with large range of length scales and timescales (Pope (2010)). This length scale
could be as large as the characteristic length of the flow. The smallest length scale
CONFIDENTIAL Error! Reference source not found.: Computational modelling
and methodology
4
(l
e
), velocity scale and time scale (t
e
) could be estimated by Kolmogrov hypothesis as
follow (Pope (2010)).
3/4
~ Re
e
l L

(1-10)
1/ 4
~ Re
e
u U

(1-11)
1/ 2
~ Re
e
e
e
l L
u U
t

=
(1-12)
The larger eddies are dominated by inertia effects and viscous effects are negligible
but in the smaller eddies viscous effects are dominating. The larger edges are highly
anisotropic (direction dependent) and strongly interact with the mean flow. In small
edges strong viscosity effects wipe out the anisotropic, hence the smaller edges are
isotropic in nature. The larger eddies are unstable and breakup in to smaller eddies
and transfer their energy to smaller eddies. The smaller eddies are break up again and
transfer their energy to even smaller eddies. This way eddy breakup creates a cascade
of energy transfer and it is referred as energy cascade (Pope (2010)). At the
smallest length scale eddies, the Reynolds number based on length scale and velocity
scale is equal to one. Hence the viscous effects are important and finally energy
dissipates into the flow. These energy losses are account for turbulent flow losses.
For resolving complete turbulence, one has to model smallest scale of turbulence as
described above. Hence for the numerical calculations in CFD spatial discretisation
(grid spacing) and time step should be equivalent to smallest eddys length and time
scales. This method is referred as the direct numerical simulation or DNS method. For
typical Re= 10
6
and L = 0.1 m, grid spacing and time steps equivalent to 3 m and 0.7
sec respectively. These figures are computationally too expensive even for simple
cases. Hence the DNS simulations are limited to simple cases academic research
within the moderate limit of Reynolds number. For example, for pipe flow.
As mentioned the smaller eddies are containing less energy than the larger eddies.
The smaller eddies are not significantly affected by flow geometries and shows
universal characteristics. Hence in large eddy simulations (LES), the effect of smaller
eddies effect is modelled and larger eddies turbulent motion is directly solved. This
LES method of modelling turbulence is still required high computation but with
increasing computation power it is becoming very popular in academic research and
some extended industrial applications. For specific to internal air system problems,
CONFIDENTIAL Error! Reference source not found.: Computational modelling
and methodology
5
Sun et al. (2006) demonstrate benefits of LES calculation for high pressure disc cavity
with axial through flow case and Omahoney (2010) et al. presented LES results for
rim sealing cavity problem.
For engineering calculation purpose, the interested quantities are time averaged mean
quantities of the flow and hence, the models based on the time average quantities are
much popular in CFD community. Typical time history of velocity component (axial
velocity) measurement in experiment is shown in Figure 1 for turbulent jet case taken
from Pope (2010).

Figure 1: Time history of axial velocity on the centreline of the turbulent jet (Pope
(2010)
For this axial velocity (V
x
) measurement, V
x
can be written as sum of steady mean
value U
x
and time varying fluctuation value
x
V
d

x x x
V U V
d
= +
(1-13)
Similarly, for general scalar flow variable 1 can be written as sum of steady mean
value O and time varying fluctuation value d 1
d 1 = O +1 (1-14)
where, steady mean flow property O is calculated as follow,
V
x
(m/s)
t

(sec)
U
x
x
V
d
CONFIDENTIAL Error! Reference source not found.: Computational modelling
and methodology
6
0
1
d
(
O = 1
(

t
t
t
(1-15)
where, time interval t should be infinite for ideal calculation but in practical
calculations t is taken as large enough to exceed the slowest variation of variable 1.
It is obvious that the time average of d 1

is zero but the root mean square (rms) value
will not be zero as follow.
0
1
0 d
(
d d 1 = 1 =
(

t
t
t
(1-16)
)
1 2
2
0
1
0
rms
d
(

d d 1 = 1 =

(
|

t
t
t
(1-17)
One other assumuption is made for deriving the time average Navier stokes equation.
It is assumed that the density is
Similarly for other variables and its derivatives can be written as,
) ) p p 1 O
=
t t

(1-18)
) )
)
U V V p p p dd \ 1 = \ O +\ 1
)& )& ))&
(1-19)
)
)
) \ I \1 = \ I \1 = \ I \O (1-20)

Chapter 2 adabsdasd
sdsdsdsdsd
CONFIDENTIAL Error! Reference source not found.: Computational modelling
and methodology
7
2.1 sdsdsdsd

Das könnte Ihnen auch gefallen