Welcome to Scribd, the world's digital library. Read, publish, and share books and documents. See more
Download
Standard view
Full view
of .
Look up keyword
Like this
1Activity
0 of .
Results for:
No results containing your search query
P. 1
Fast Solution for Large-Scale 2-D Convection-Diffusion

Fast Solution for Large-Scale 2-D Convection-Diffusion

Ratings: (0)|Views: 4|Likes:
Published by iMiklae
2-D convection-diffusion, reacting flows in a single channel of catalytic
monoliths are investigated. The fluid dynamics are modelled by a steady
state, boundary-layer equations, which is a large system of parabolic partial differential
equations (PDEs) with nonlinear boundary conditions arising from the coupling
between the gas-phase and surface processes. The chemical processes are
modelled using detailed chemistry. The PDEs are semi-discretized by a method of
lines leading to a large-scale, structured differential algebraic equations (DAEs).
The DAEs are solved using a tailored BDF code. We exploit the structure of
the Jacobian and freeze the diffusion coefficients during approximation of Jacobian
by the finite difference. By applying our approach, the computation times
have been reduced by a factor of 4 to 10 and more depending on the particular
problem.
2-D convection-diffusion, reacting flows in a single channel of catalytic
monoliths are investigated. The fluid dynamics are modelled by a steady
state, boundary-layer equations, which is a large system of parabolic partial differential
equations (PDEs) with nonlinear boundary conditions arising from the coupling
between the gas-phase and surface processes. The chemical processes are
modelled using detailed chemistry. The PDEs are semi-discretized by a method of
lines leading to a large-scale, structured differential algebraic equations (DAEs).
The DAEs are solved using a tailored BDF code. We exploit the structure of
the Jacobian and freeze the diffusion coefficients during approximation of Jacobian
by the finite difference. By applying our approach, the computation times
have been reduced by a factor of 4 to 10 and more depending on the particular
problem.

More info:

Categories:Types, Research, Science
Published by: iMiklae on Dec 01, 2012
Copyright:Attribution Non-commercial

Availability:

Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See more
See less

12/04/2012

pdf

text

original

 
Fast Solution for Large-Scale 2-D Convection-Diffusion,Reacting Flows
Hoang Duc Minh
1
, Hans Georg Bock 
2
, Steffen Tischer
1
, and Olaf Deutschmann
1
,
1
Institute for Chemical Technology and Polymer ChemistryUniversity of Karlsruhe (TH)Engesserstr. 20, D-76131 Karlsruhe, Germany
hoang.minh@kit.edu
2
Interdisciplinary Center for Scientific Computing (IWR)University of Heidelberg, Germany
Abstract.
2-D convection-diffusion, reacting flows in a single channel of cat-alytic monoliths are investigated. The fluid dynamics are modelled by a steadystate,boundary-layer equations, whichisalargesystemofparabolicpartialdiffer-ential equations (PDEs)with nonlinear boundary conditions arising from the cou-pling between the gas-phase and surface processes. The chemical processes aremodelled using detailedchemistry. ThePDEsaresemi-discretized byamethod olines leading to a large-scale, structured differential algebraic equations (DAEs).The DAEs are solved using a tailored BDF code. We exploit the structure of the Jacobian and freeze the diffusion coefficients during approximation of Jaco-bian by the finite difference. By applying our approach, the computation timeshave been reduced by a factor of 4 to 10 and more depending on the particularproblem.
1 Introduction
Theoretical analysis and simulation of complicated processes in catalytic monolithswidely used in industries, are currently active research topics. The aim of this work ismodelinganddevelopingrobustnumericalmethodsforsimulationofphysical-chemicalprocesses in a single channel, as a first step to study the complex ones in the wholecatalytic monoliths. Earlier investigations can be found in [1,2,3,4]. Boundary layertheory is applied for approximation of the flow field. The Navier-Stokes equations aresimplified to get a system of large-scale parabolic partial differential equations (PDEs).For simplicity of presentation without lost of generalization, we limit our discusion of reacting flows without surface processes—no catalyst on the surface.Thispaperis organizedas follows.Themathematicalmodelusingtheboundarylayerequations are stated in Section2. In Section3,numerical methods are discussed. Some computational experiments are given in Section4.The paper concludes in Section5 with a summary of the obtained results and discussion of further work.
Wewould liketothank Dr.LubaMaier, Institutefor Chemical Technology and PolymerChem-istry, University of Karlsruhe for supplying the reaction mechanisms and helpful discussions.
O. Gervasi et al. (Eds.): ICCSA 2008, Part I, LNCS 5072, pp. 1121–1130,2008.c
Springer-Verlag Berlin Heidelberg 2008
 
1122 H.D. Minh et al.
2 Mathematical Model
2.1 Modeling the Fluid Dynamical Process
To model flows in a channel, one can employ the boundary layer equations which aresimplified from the Navier-Stokes equations. Since our considered channel is an ax-isymmetrical cylinder, we assume a simplification that the investigated flow in it is alsoaxisymmetrical, which can be described by two spatial coordinates, namely the axialone
z
and and the radial one
r
. By applying von Mises transformation
ψ
=
 
r
0
ρur
dr
,
where
ψ
is stream variable and
u
is axial velocity, there arises the following equationsystem, which we use for modeling reactive flow in a channel of monoliths [5].Momentum:
ρuu∂z
+
pdz
=
ρu∂ ∂ψ
ρuμr
2
uψ
,
(1)
∂pψ
= 0
.
(2)Energy:
ρuc
p
∂T z
=
ρu∂ ψ
ρuλr
2
∂T ψ
g
k
=1
˙
ω
k
k
h
k
ρur
g
k
=1
kr
c
p
k
∂ψ.
(3)Species:
ρu
k
z
= ˙
ω
k
k
ρu∂ ψ
(
rJ 
kr
)
,
(
k
= 1
,...,N 
g
)
.
(4)State:
 p
=
ρRT .
(5)Where
z
and
r
are the cylindrical coordinates,
u
and
v
are the axial and radial com-ponents of the velocity vector,
p
is the pressure,
is the temperature,
k
is the massfraction of the
k
th species,
μ
is the viscosity,
ρ
is the mass density,
c
p
is the heat capac-ity of the mixture,
λ
is the thermal conductivity of the mixture,
c
p
k
is the specific heatcapacity of the
k
th species,
g
is the total number of gas phase species,
g
is the totalnumber of elementary reactions,
kr
is the radial component of mass flux vector of the
k
th species,
˙
ω
k
is the rate of production of the
k
th species by the gas phase reactions,
h
k
is the specific heat enthalpyof the
k
th species,
k
is the molecularweight of the
k
thspecies,
is the mixture mean molecular weight, and
R
is the universal gas constant.
 
Fast Solution for Large-Scale 2-D Convection-Diffusion, Reacting Flows 1123
The diffusion mass flux
kr
appearing in (3) and (4)is given by
kr
=
D
m
k
k
ρ
2
ur
k
∂ψ
D
T
k
ρur∂T ψ,
(6)where
D
m
k
and
D
T
k
are diffusion coefficients. The effective diffusion coefficient
D
m
k
of the
k
th species into a mixture is approximated by the following mixture-averageddiffusion coefficient formula (see e.g., [1],[6] and [7])
D
m
k
=1
k
g
i
=1
,j
=
k
k
/D
ik
,
(7)where
k
and
k
are the mass fraction and the mole fraction of the
k
th species, and
D
ik
is the binary diffusion coefficients. The binary diffusion coefficients
D
ik
is alsoestimated based on the logarithm of its value, which is evaluated by the fourth-orderpolynomial of the logarithm of the temperature
ln
D
ik
=
d
1
,ik
+
d
2
,ik
ln
+
d
3
,ik
(ln
)
2
+
d
4
,ik
(ln
)
3
+
d
5
,ik
(ln
)
4
ln
 p,
where
p
is the pressure and
d
l,ik
are the coefficients of the fitted polynomial of thebinary diffusion coefficients.As mass fraction,
k
must satisfy
0
k
1 (
k
= 1
,...,N 
g
)
,
g
k
=1
k
= 1
.
(8)The relation between
ψ
and
r
will be treated later in the differential form
0 =
r
2
ψ
2
ρu.
(9)
2.2 Modeling the Chemical Process
To model the chemical kinetics of moving gas mixture, we use the detailed chemistrydescribing elementary reactions on molecular level (see, e.g., [8,9] for more details).The rate of production
˙
ω
k
of 
k
th species appearing in equations(3) and (4) is given by
˙
ω
k
=
g
i
=1
v
ki
k
i
g
j
=1
[
j
]
v
ji
(
k
= 1
,...,N 
g
)
,
where
g
: the total number of elementary reactions,
g
: the total number of gas phase species,
v
ki
: the stoichiometric coefficient of the
k
th speciesin the
i
th reaction at the
left 
side,
v

ki
: the stoichiometric coefficient of the
k
th speciesin the
i
th reaction at the
right 
side,
v
ki
:
=
v

ki
v
ki
,
[
j
]
: the concentration of the
j
th species,
k
fi
: the forward rate coefficient of the
i
th reaction.

You're Reading a Free Preview

Download
scribd
/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->