Professional Documents
Culture Documents
Depcik 2013 RevChemEng PDF
Depcik 2013 RevChemEng PDF
0.4
Region 1 Region 3
1.2 Early history
ΔP (bar)
Region 2
0.3
Wark and Warner (1981) describe the use of fabric filters
for particulate control in industrial applications. The pres- 0.2
sure drop across these fabric filters is a function of both
the dust layer and the filter represented by Darcy’s law: 0.1
μut 0
Δp = , (1) 0 0.5 1.0 1.5 2.0 2.5 3.0
Κ Soot mass (g)
which is a phenomologically derived constitutive equa- Figure 4 Three regions of pressure drop during PM loading
tion that describes the flow of a fluid through a porous (Versaevel et al. 2000).
where c and x are empirical constants. Figure 5 illustrates Figure 5 Schematic of single-cell and associated variables (Depcik
how this article defines the geometric variables of interest. and Assanis 2008).
600 25
( )
Estimated particulate accumulation (g)
R = ACOx2 exp - Ea RuTf , (7)
Estimated particulate accumulation
480 20
A B
Oxidation occurs primarily at the front of the channel
800 Uniform soot distribution over the length of the channel
Soot oxidizes more rapidly at the rear of the channel
Exhaust volume flow
700
Temperature (°C)
600
400
0 1 2 3 4 5 Channel length
Time (min)
Figure 7 (A) Exhaust and filter temperature during regeneration and (B) influence of particle load variation on axial exhaust flow distribu-
tion (Pauli et al. 1983).
A Exhaust flow This model neglects the influence of the channel vari-
ations in temperature and species on the solution to
the equations in the PM and wall layers. Of particular
importance is the usage of constant mole fractions for
the chemical species inlet channel. This is because later
researchers find that the mass diffusion to the surface
w Deposit
from the channel is important. This is what will require
Substrate a merging of two 1D models into one complete 1+1D
x simulation.
where k = ATf exp(-Ea/RuTf ). The oxidation expression in with Wa as the molecular weight of the gaseous mixture
Eq. (12) includes a specific area term (Sp), which is a func- and the density of the PM before the derivative converts
tion of the geometric properties of the PM layer. the right-hand side into the proper units of thickness. This
The wall species equation for oxygen is equation assumes that all the deposit consumed in the
reaction shrinks the thickness of the deposit layer while
(
∂ ρw uw X w ,O2 ) =0. (14)
keeping the particle bulk density and particulate specific
∂y area constant. The subsequent simulation results illus-
trate that although reactions can occur throughout the
Note that the s notation on the mole fractions (X) indi- entire deposit later, often most PM depletion occurs in a
cates the values for the mole fractions in the PM layer, thin layer near the surface.
whereas w indicates the values for the mole fractions in
the wall layer. The authors indicate a specific nomencla-
ture difference here (indicated as regions 1 and 2 in the 1.3.3 Energy balances
Bissett and Shadman article) to prepare the readers for
later revisions in a following section. Moreover, Bissett Bissett and Shadman’s initial calculations showed that
and Shadman use average density (ρw) and velocity (uw) the typical length scale over which the gas temperature
values for both layers that will be revisited by future in the wall (Tw) adjusts to differences within the solid tem-
authors. perature (Tf ) is several orders of magnitude smaller than
typical deposit layer thicknesses. Bissett presents this cal-
culation in an internal report for General Motors (Bissett
1.3.2 Particle mass balance 1981). Therefore, they assume that the gas and the solid
temperatures are equal, except at the inlet face of the
By expanding Eq. (12) according to the chain rule, substi- deposit layer (y = 0):
tuting the equation for continuity in the wall via Eq. (10)
Tw = Tf , (19)
and using the method of separation of variables while
integrating over oxygen mole fraction generates the fol- where
lowing result:
Tw|y = 0 = TI . (20)
X s ,O2 With respect to the PM energy balance, they assume that
y =0
X s ,O2 = . (15)
y =t d
(
exp kS ptd uw ) heat losses to the surroundings are negligible due to
proper insulation:
∂td
ρd
∂t
( ( )
)
= - WC s Wa ρ w uw X I ,O ⎡ 1-exp -kS ptd uw ⎤ ,
2 ⎣ ⎦ ( ) (18) This model provides a method for computing the
regeneration event of the filter and they use it to predict
the temperature profiles as a function of time with differ- where the subscripts I and II represent the inlet and outlet
ent PM loading amounts. channels, respectively. In this equation, the terms on
the right-hand side of the equation are a measure of the
mass leaking and entering through the porous walls. In
1.4 Bissett’s channel model his article, Bissett uses only one average density, velocity,
and pressure to describe the thermodynamic properties in
Bissett (1984) follows up the previous effort by creating a the wall and PM layers as discussed prior. The following
model for the transport of nearly all thermodynamic vari- sections will indicate how to compute these values for this
ables through a wall-flow DPF while trying to avoid the model.
complexities of multidimensional compressible fluid flow.
He accomplishes this by representing the flow at each
axial position in 1D and incorporating the proper physics 1.4.2 Momentum equations
for the flow to and from the wall system. In his article,
Bissett makes the following assumptions: For the momentum equations through the channel, Bissett
– The properties of the gas entering the front face of the assumes that the gas flows only in the z-direction as it
monolith are spatially uniform; however, they can be enters or exits the channel; hence, there is no transport
time dependent. of the y-component of momentum at the wall. However,
– The model assumes perfect insulation for the because the no-slip condition holds at the walls in the
circumferential walls of the monolith. z-direction, friction will occur. Bissett states that this vari-
– A single inlet and outlet channel represents the ation should be very close to that observed when the walls
behavior of all inlet and outlet channels. are solid because only a very small fraction of the flow at
– He neglects radiative heat transport in the channels each axial position passes through the wall:
because the long thin channel walls have similar view
∂pI ∂ ( ρI uI ) ∂p ∂ ( ρII uII )
2 2
factors and should have similar temperatures. μu μu
+ = - F 2 I and II + = - F 2II . (24)
– The thickness of the PM layer is so thin compared ∂z ∂z d ∂z ∂z d
with the channel thickness during the initial period
A correlation for fully developed laminar flow in
of operation; the change in the channel geometry
a square channel relates the constant (F) on the right-
due to the presence of the PM layer can be neglected:
hand side to the local differential pressure. This simpli-
td < < d .
fied assumption has been addressed and withdrawn in
– The emitted particles do not interfere with the flow
a recent effort by Bissett et al. (2012) by solving the 3D
due to their low mass fraction within the bulk gas.
boundary layer in the channel.
– Although not explicitly stated in his article, Bissett
To simulate the pressure drop across the porous
omits the time derivatives of the gaseous state
wall, Bissett uses Darcy’s (1856) law and includes Forch-
variables, effectively stating that there is not any
heimer’s (1901) modification to allow for the possibility of
storage in the gas according to a quasi-steady-state
inertial flow effects caused by high velocities:
approximation.
μ u
pI -pII = ⎛ w w + βρw uw2 ⎞ ⎡⎣t f + φ (td )⎤⎦ , (25)
This effort becomes the definitive beginning to the ⎝ Κ ⎠
channel flow DPF model with all extensions and adap-
tations basing themselves off this seminal article. This where φ(td) is a function of the thickness of the deposit
section describes the theory included in this article along loaded on the filter. Although not expressly indicated
with the assumptions made helping lead into the exten- in his article, this equation is an algebraic result of the
sions presented by later authors. momentum equation through the wall when employing
Darcy’s law. In his model, the ideal gas law is valid for the
channels and the wall gases:
1.4.1 Continuity equations
patm = ρRT. (26)
Bissett writes the mass equations for the channels as However, he uses atmospheric pressure instead of
local pressure, because the assumption is that all pressure
∂ ( ρI uI ) -4 ∂ ( ρII uII ) 4 differences are small in comparison with the atmospheric
= ρw uw and = ρw uw , (23)
∂z d ∂z d pressure. Of particular note, the value for the average wall
velocity comes from the solution of Eq. (25), whereas the which is used in predicting the influence of exothermic
average wall density will come from the ideal gas law once reactions in the filter energy balance. Hence, although not
making an assumption for temperature of the gas in the expressly a 1+1D model, one can see the need to model in
wall in the next section. both the channel direction (z) and through the wall (y).
In the energy equations, because there is mass flowing To calculate the temperature of the filter, Bissett uses the
through the walls, there will be transport of energy traditional heat diffusion equation writing it as a func-
between the channels and the porous wall in addition to tion of the thicknesses of the wall and PM. He assumes
heat transfer between the gases and the wall. In Bissett’s that these layers have the same temperature due to their
original article, he writes these equations as proximity to each other and the fact that PM is present
within the filter wall due to deep-bed filtration. The model
ρI uI c p
∂TI 4
( ) ∂T
= ⎡⎣hg -ρw uw c p ⎤⎦ Tf -TI and ρII uII c p II includes the effects of conduction in the wall Q cond , the ( )
∂z d ∂z
(27) (
heat transfer from the gases in the channel Q conv , and the )
4
(
= ⎡⎣hg + ρw uw c p ⎤⎦ Tf -TII .
d
) effects of PM oxidation on the energy of the wall Q reac : ( )
∂
Bissett states that the formulation of his model ∂t ⎣
( ⎦)
⎡ ρf cf tf + ρd cd td Tf ⎤ = Q cond + Q conv + Q reac (30)
neglects the thin heat transfer layer in the wall (essentially
the boundary layer at the inlet channel interface and the It is important to note that Bissett did include a wall
wall) requiring all heat transfer to take place in the chan- ( )
energy flux in the derivation Q wall of this equation;
nels. As a result, the wall temperature is the equal tem- however, the two resulting terms cancelled each other out,
perature of both phases in the wall (gas as Tw and filter as and consequently, the influence of wall energy flux is not
Tf ) and independent of the radial direction (Tw = Tf ); it can included in the final equation. The authors will explain
still vary in the axial direction. Hence, by continuity, the this flux omission and later inclusion in a later section.
wall temperature is also the temperature of the channel Bissett’s derivation of the conduction term in this
gas just outside the wall system. In addition, conduction energy equation allows for the impact of different PM
is determined to be negligible relative to advective trans- thicknesses along the length of the filter:
port in the axial direction. However, because this is a 1D
∂ ⎛ ∂T ⎞ ∂ 2Tf
model, he states that an effective heat transfer coefficient Q cond = λd ⎜ td f ⎟ + λ f tf . (31)
(hg) in this equation includes the contribution of conduc- ∂z ⎝ ∂z ⎠ ∂z 2
tion in the radial direction to effectively account for 2D
The heat transfer to the filter wall balances according
effects.
with the channel energy equations (Eq. 27),
Drawing from his earlier effort, Bissett restates the equa- and the effects of PM oxidation include the updated
tion for the oxygen level in the PM layer as a balance derived effort of Eq. (18) to Eq. (29):
between advection and reaction kinetics but writing now
via mass fractions: ( ) (
Q reac = - ΔH reac WO2 ρw uwYI ,O2 ⎡ 1-exp -kS ptd uw ⎤ ,
⎣ ⎦ ) (33)
(
∂ ρw uwYs ,O2 ) = -kS ρwYs ,O2 . (28)
where ΔHreac is the heat of reaction of the carbon oxidation
∂y
p reaction (Eq. 13).
Bissett’s model can simulate complete PM regenera-
This equation is needed to calculate the change in tion events and calculate the time change of filter temper-
particle mass balance derived prior: ature as shown in Figure 9. As mentioned previously, this
work is of prime importance to the DPF modeling commu-
∂t d
ρd
∂t
( ( )
)
= - WC s WO2 ρw uwYI ,O2 ⎡ 1-exp -kS ptd uw ⎤ ,
⎣ ⎦ ( ) (29) nity and its importance has been widely recognized in the
technical literature. In the next section, another article by
1000
wall layer (uw). However, Haralampous et al. only use one
average velocity for both layers (uw), similar to the original
900
212 s
efforts of Bissett, but instead add a function to the species
31.8 s
800 equations to account for the change in cross-sectional
area that the flow encounters through these layers. Depcik
15.9 s
700 and Assanis take a slightly different approach by mode-
26.5 s ling the velocity profile through both layers as a function
600
0 0.2 0.4 0.6 0.8 1.0
of distance: us, uw = f(y).
z/L (-) Peters et al. describe the continuity equations for
square channels as
Figure 9 Profiles of wall temperature as a function of dimension-
less axial distance at selected times (Bissett 1984). ∂ ( ρI uI dI2 ) ∂ ( ρII uII dII2 )
= -4dI ρI us and = 4dII ρII uw . (34)
∂z ∂z
Bissett describes the flow through the filter PM and wall When applying continuity to the mass flow entering
layers in more detail. This work will eventually merge into the PM layer and leaving the porous wall into the outlet
the channel wall model right after the turn of the century. channel, they recover the following equation for the
velocities:
ρ I dI
uw = us , (35)
2 Updated model equations ρII dII
Bissett and Shadman published two seminal articles in the with the effective hydraulic diameters now computed as
DPF modeling community of which most others base their
dI = d-2td and dII = d . (36)
efforts. In this section, the authors indicate those who
expanded the model to account for area conservation, NO2 In this model, the right-hand side of the continuity
passive regeneration, catalyzed filters, and multidimen- equation uses the inlet density leaving the inlet channel;
sional effects. Because the two models have undergone for the outlet channel, they use the outlet channel density
a merging into one, this article presents these sections entering from the wall. According to the classic model,
according to the governing equations of flow making little this does not correspond correctly to the system because
distinction between the versions. For example, the next the density at the interface of the wall can have a differ-
section presents the mass equations for the PM and wall ent value than the channels. A discussion of this concept
layers alongside those for the channel equations. In addi- occurs later in this article illustrating how to either
tion, each area illustrates a brief historical summary of compute the density as a function of a local thermody-
who modified the equations and why they revised them. namic equilibrium or use an average pressure across these
layers.
In the Peters et al. articles, the surface areas per unit
2.1 Continuity equations length in the inlet and outlet channels equations are dif-
ferent because of the smaller gaseous control volume in
The assumption of changes in channel geometry due to the inlet channel when PM is present:
PM layer loading was revisited by Peters et al. (Peters
SI = 4dI and SII = 4dII . (37)
2003, Peters et al. 2004), Haralampous et al. (Haralampous
and Koltsakis 2004a,b, Haralampous et al. 2004a,d,e), A physical property that must be met by all models
and recently by Depcik and Assanis (Depcik and Assanis is that there should not be any mass storage in the PM
2008, Depcik 2010). In particular, these works reflect a or wall layers during steady-state conditions: (∂ρs/∂t = 0)
consideration of Bissett’s original concept that the thick- and (∂ρw/∂t = 0). Because Eq. (36) relates the velocities,
ness of the PM layer does not influence the flow properties the mass flux source terms cancel out and the model
(1984 article, assumption #5). To do so, the Peters et al. does not violate any mass conservation laws. However,
∂ ( ρI uI d 2
I ) = -4d ρ u ∂ ( ρII uII d 2
II ) = 4d ρ u
and . (38) 0.05
∂z ∂z
w w w w
Km=2.00×10-12m2
When using different surface areas for the inlet and
0
outlet channels when PM is present, to balance the 0 50 100 150 200
mass fluxes, the density, velocity, or density times velo- Accumulated deposit loading (g)
∂ ( ρsus As ) a function of the filter and not the bulk gas (Depcik and
= As Smass . (41)
Assanis 2005).
∂y
With respect to the original assumptions of Bissett and
The change in nomenclature here stems from the Shadman in their 1985 article, Depcik and Assanis found
fact that the Haralampous and Koltsakis article does not that not including the mass source term in Eq. (40) or Eq.
differentiate the gas properties between the PM and the (41) results in a small error of approximately 6% in the mass
wall layer (besides the chemical species), whereas mass flow rate and 1% in the final deposit mass during an experi-
equations are written for the PM and the wall layers in the mental test involving PM oxidation. They also found that, in
Depcik and Assanis articles. the cases of heavy PM loading of the filter, the wall velocity
For square channels, the change in cross-sectional could change throughout the wall by up to 20%. However,
area per unit length through the PM layer can be deter- in the cases of light to moderate loading under which most
mined from Figure 5: DPF filters see operation, as illustrated in Figure 11, the velo-
city of the gas through the wall does not vary significantly,
⎡ y ⎞⎤
As = 4 ⎢d-2 ⎛ td - , where tan θ = 1 (42) indicating that the singular velocity model of Haralampous
⎣ ⎝ tan θ ⎠ ⎥⎦
et al. will not generate appreciably different results.
Focusing on the variable filtration area per unit length
and the instantaneous change the cross-sectional area
variable As, three cases can be identified. When td/d is
comes from the derivative of this value with respect to the
close to zero, no variability needs to be taken into account
distance through the PM layer; that is, dAs/dy=8. For differ-
(Bissett and Shadman 1985). In the range of td/d < < 1, a
ent channel geometries, it is relatively easy to replace As and
“quasi 1D flow” can be assumed in a thin cake with the
its derivative with the respective new values while main-
variable filtration area useful in modeling activities.
taining the same visual identity of the governing equations.
Finally, the case of td/d = O(1) corresponds to filter clogging
Although the cross-sectional area that the flow sees
conditions. Inside the cake, the assumption of quasi 1D
is constant in the wall layer, writing the mass equation in
flow will be less accurate due to the velocity components
an area-conserved format allows for continuity with the
perpendicular to wall velocity. Most important during this
other equations:
situation is that the axial flow friction increases signifi-
d ( ρw uw Aw ) cantly, pressure and density may vary considerably inside
= 0, (43)
the cake, and the surface of the cake may deviate from the
dy
rectangular shape. These conditions are increasingly diffi-
where Aw = 4d. On inspection of the mass equations in the cult to model accurately and simulation results, although
channel Eq. (39), the mass flux terms do cancel each other useful, must be treated with caution.
out similar to the before mentioned references. Although
the properties in the PM and wall layers change, they
must remain consistent with the governing equations of 2.2 Momentum equations
mass in each respective layer. In all cases, the time deriva-
tives of the variables are neglected, which is a reasonable Similar to the continuity equations, the momentum equa-
assumption as long as the time scales in the problem are tion in the channel remained constant from Bissett’s
3.4 1
0.02 0.02
3.3 0.99
0.04 0.04
2.5 mm 0.98
3.2
0.06 0.06
0.97
3.1
0.08 0.08
0.208 mm 0.96
3 15.5 mil
0.95
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Axial distance (cm) Axial distance (cm)
Figure 11 Velocity and pressure in PM and wall layers under moderate loading (Depcik and Assanis 2008).
formulation until Peters et al. and Haralampous et al. where c is the slip flow correction factor and p0 is the refer-
revisited the change in channel volume due to PM loading. ence pressure for apparent permeability. The discussion
The Haralampous et al. model incorporates the change in of this equation and the impact of mean free path length
geometry only in the frictional component on the right- will come in a later section of this article, which presents a
hand side of the equation through the effective channel historical summary of the calculation of appropriate per-
diameter: meability for the momentum equation.
In these approaches, the derived algebraic equation
∂pI ∂ ( ρI uI ) ∂p ∂ ( ρII uII )
2 2
μu μu for the pressure drop comes from solving the momen-
+ = - F 2 I and II + = - F 2II . (44)
∂z ∂z dI ∂z ∂z dII tum equation. Following Wooding (1957) and Brinkman
(1947a,b), a theorized extension of the Navier-Stokes
In the Depcik and Assanis model, the equations momentum equation to porous media involves the use of
resemble those used in area-conserved, 1D flow (Liu et al. an effective viscosity to account for the diffusion of velo-
1996, Onorati et al. 1999, Depcik et al. 2005): city. However, for many practical purposes, there is no
need to include this viscous term (Nield and Bejan 2006).
∂ ( ρI uI2 dI2 ) ∂ ( pI dI2 ) d (dI2 ) In addition, Beck (1972) found that the inclusion of the
+ = - F μuI + pI (45)
∂z ∂z dz advection term, ∂(ρu2)/∂y, is erroneous when Darcy’s law
is employed. Depcik and Assanis include these concepts
∂ ( ρII uII2 dII2 ) ∂ ( pII dII2 ) d (dII2 ) in formulating their area-conserved PM and wall momen-
+ = - F μuII + pII . (46) tum equations:
∂z ∂z dz
d ( ps As ) dA μu A
Writing the above equations in this form requires = ps s - s s -βd ρs Asus2 (50)
including an additional term on the right-hand side to dy dy Κ d
balance the forces. Expanding the derivatives recovers the
same equations of Haralampous et al. when assuming the
d ( pw Aw ) dA μu A
change in PM thickness in the z-direction is negligible: = pw w - w w -βm ρw Aw uw2 . (51)
dy dy Κf
dtd/dz≈0. It is important to note that keeping the area
terms in the derivatives and solving the governing equa-
Again, it is important to note that the area change
tions as indicated is more accurate with respect to mass
term on the right-hand side allows a balance in the pres-
conservation (Roe 1986, Liu et al. 1996, Onorati et al. 1999).
sure forces when writing in this format. Expanding the
Konstandopoulos et al. (2000) revisited the momen-
derivative of the pressure force recovers the equations of
tum equation through the PM and wall layers by adding
Konstandopoulos et al. and Haralampous et al. but written
the effects of variable area within the PM layer. This was
as a differential as required by Depcik and Assanis’s
expressed the following year in a manner similar to Bissett’s
numerical scheme explained later. As a result, the choice
work through an algebraic formulation (Konstandopoulos
of methodology depends on the requirements of the
et al. 2001):
modeler within their computational framework.
μuw μu ⎛ d⎞
pI -pII = t f + βm ρw uw2 t f + w d ln⎜ ⎟ . (47)
Κf 2 Κd ⎝ dI ⎠
A 1400 B 0.30
Experimental
1300 Using cp(ΔT) Using cp(ΔT)
Using Enthalpy 0.25 Using Enthalpy
0.20
1100 12.7 cm
0.15
1000
900 0.10
2.54 cm 7.62 cm
800 0.05 PM Conversion
Enthalpy formulation=55.0%
Inlet
cp(ΔT) formulation=52.0%
700
0
1680 1720 1760 1800 1840 0 0.05 0.10 0.15
Time (s) Axial distance (m)
Figure 12 Comparing the use of cp(ΔT) for enthalpy with actual enthalpy in energy equations (Depcik 2010).
To conclude their model development, Depcik and coefficient found using a reference by Wakao et al. (1979).
Assanis model the gas temperature in the PM and wall When the filter is empty, they use a correlation developed
layers. Again, because the kinetic energy of the gas in the by Zukauskas (1987) for staggered cylinders used in heat
wall is much smaller than its enthalpy (hw uw2 ) , the fol- transfer studies for metal foams. This correlation should
lowing equations for the energy of the gas in the PM and give a heat transfer coefficient smaller than actual due to
wall layers can be written as (Calmidi and Mahajan 2000, the high porosity of the metal foams compared with a DPF
Ingham et al. 2004, Zhao et al. 2004, Lu et al. 2006) filter wall.
To check the assumption that the gas temperature in
∂ (ρsushs As ) ∂ ⎛ * ∂Ts ⎞ hc Asa the wall instantly assumes the filter temperature, Depcik
∂y
- ⎜ λ s As
∂y ⎝ ∂y ⎠⎟
=
εp
Tf -Ts ( ) (57)
and Assanis ran a sample simulation artificially raising
the inlet gas temperature while holding the filter wall
at a constant lower temperature. When comparing the
∂ ( ρw uw hw Aw ) ∂ ⎛ * ∂Tw ⎞ hc Aw a magnitudes of the three terms in Eqs. (57) and (58), they
∂y
- ⎜ λw Aw
∂y ⎝ ∂y ⎠⎟
=
εp
Tf -Tw . ( ) (58)
calculate the magnitude of convection to advection to be
approximately 105, whereas the magnitude of convection
In these equations, the close proximity of the filter to conduction is about 104. This illustrates that convection
affects the thermal conductivity of the fluid in the porous dominates the mode of heat transfer in the wall validat-
layers and, as a result, requires an effective thermal ing the common assumption that Ts = Tw = Tf. As a result,
conductivity calculation. The calculation of which is incorporating this finding in Eqs. (55) and (56) will result
described for metal foams by Boomsma and Poulikakos in Eq. (52) for the inlet channel and Eq. (27) for the outlet
(2001). Because no fundamental models of effective con- channel when making the assumption of constant pres-
ductivity exist for DPF purposes, Depcik and Assanis use sure specific heats.
the metal foam formulation in their article (see also Lu In a recent article from Bissett et al. (2012), the
et al. 2006 for more information on its application). authors solve the boundary layer equations for mass,
The term on the right-hand side accounts for the heat momentum, and energy in the square channel with the
transfer between the filter and the gas in each respective aim to derive appropriate “film” heat transfer coefficients.
layer. It has been stated that the calculation of the heat They solve their equations using a boundary condition
transfer coefficient (hc) is difficult to measure (Ingham of fixed temperature at the channel/wall interface (Tf via
et al. 2004); however, two correlations exist in the litera- the nomenclature in this article). The authors imply that
ture served as an approximate upper and lower bounds. this temperature value used in the boundary could be
In the situation where the filter wall is packed full of PM considered equal to the solid wall temperature due to the
and a cake layer exists in the inlet channel, Depcik and extremely efficient heat transfer between gas/wall in the
Assanis theorize that the heat transfer will be analogous solid phase. However, the gas temperature will become
to a packed bed reactor with a representative heat transfer equal to the wall temperature only as a result of heat
transfer inside the wall. At the channel/wall boundary, conversion of particulates to partial and complete prod-
the gas temperature will assume a different temperature. ucts of combustion:
Therefore, the derivation of transport coefficients in the
C(s)+NO2→NO+CO (60)
work of Bissett et al. (2012) is indeed valid for the case
of fixed gas temperature at the channel/wall boundary; CO+NO2→NO+CO2. (61)
however, this temperature is not a priori known in the
The use of NO2 to combust the particulate is adapted
standard formulation of the DPF model equations. The
a number of years later by Hawker et al. (1997) through
use of the solid wall temperature in the fixed temperature
the design of an upstream catalyst for the oxidation of NO
boundary condition is only acceptable when the “no-slip”
to NO2 to make use of its lower temperature DPF regene-
condition is valid, which is not the case in the DPF porous
ration characteristics. Because diesel exhaust consists
wall case.
mainly of NO (Heywood 1988), an oxidation catalyst (typi-
Note that the Nu number is subject to the selec-
cally platinum based) is needed to form NO2. They later
tion of the boundary condition even in the simpler case
use this catalyst for continuously regenerated DPF opera-
of flows in closed parallel ducts (no-slip condition). For
tion in which deposit is constantly being regenerated by
example, the Nu number in square ducts is 3.61 in the case
NO2 during its operation (Hawker et al. 1998).
of uniform heat flux boundary condition in contrast to
At the same time, Ahlström and Odenbrand (1989)
the typically used value of 2.98 for the case of fixed wall
find that PM does not proceed completely toward com-
temperature. This is to emphasize the importance of the
plete combustion in an oxidative environment. This is
proper inclusion of the boundary condition when trying
later collaborated by Aoki et al. (1993) when they find
to derive global heat transfer coefficients, suggesting that
that a significant portion of the exhaust flow measured
a deeper investigation of the boundary layer problem with
out of the DPF shows the formation of CO as illustrated by
the inclusion of the wall domain would be worth study in
Figure 13. As a result, they write the particulate combus-
the future (Sahraoui and Kaviany 1994, Ochoa-Tapia and
tion reaction as
Whitaker 1997, Depcik and Assanis 2008, d’Hueppe et al.
2011). In the meantime, the use of the standard Nu number C(s)+(1-α1/2)O2→(1-α1)CO2+α1CO (62)
for closed ducts is justified by the conceptual assumption
with an experimentally determined α1 between a third and
that the portion of the flow that is not sucked through
a quarter.
the porous wall exchanges convective heat with the wall
Koltsakis and Stamatelos (1996a,b) further explore
surface as if the wall is nonporous.
incomplete deposit oxidation and formulate the com-
monly used chemical balance:
5
⎛ ∂YI ,O2 ∂YI ,O2 ⎞
ε p ρI ⎜ + uI = -k ρwYw ,O2. (59)
⎝ ∂t ∂z ⎟⎠ 0
CO
This equation is reminiscent of the Euler species -5
equation that can replace Bissett’s constant O2 mole frac- O2
tion assumption. However, researchers never include it -10
Conventional
along with the classic model formulation. Reverse
At the same time, Cooper and Thoss (1989) find that -15
0 50 100 150
NO2 present in the exhaust is also effective at oxidizing Time (s)
the particulate at a significantly reduced temperature
level versus oxygen (~250°C vs. 550°C). In their work, Figure 13 Comparison between conventional and reverse regenera-
they theorize a two-equation reaction mechanism for tions with regard to exhaust emissions (Aoki et al. 1993).
Layer II W2 X
Layer I W1
Substrate Ws
oxidation), and R is an array that accounts for gaseous (Depcik and Assanis 2005). This is because they find that
catalytic reactions via the NO oxidation reaction (O2 = loss the mass-transfer coefficients can be on the same order of
and NO2 = gain). They obtain the local NO mole fraction magnitude as the wall-flow velocity. In comparison, recent
from nitrogen atom conservation. work by Depcik uses the species equations for the channel
It is not until the Peters et al. (Peters 2003, Peters et al. via a concentration format instead of mass fractions:
2004) and Haralampous et al. articles (Haralampous et al.
∂ ( ρ I uI YI dI2 )
2003, 2004a,b,c,d,e, Haralampous and Koltsakis 2004a,b) = -4dI ⎡⎣ ρ sus YI -km ( ρ s Ys -ρ I YI )⎤⎦ (73)
that the literature illustrates a merging of two models: ∂z
channel and wall. These authors decide to include the wall
species equations into the overall formulation instead of ∂ ( ρ II uII YII dII2 )
= 4dII ⎡⎣ ρw uw Yw + km ( ρw Yw -ρ II YII )⎤⎦ . (74)
solving them separately. They accomplish this to provide the ∂z
proper boundary conditions for the species equations in the
channel with the Haralampous et al. version shown here: This use of concentration equations ensures that
the equations properly conform to the mass equations
∂ ( uI YI ) uw YI km ( Ys -YI ) presented earlier. This can be seen by summing both
=- + (69)
∂z dfy2 dfy sides of the equations, taking into account that the
sum of mass fractions equals one and using the dilute
mixture simplification that is common in catalyst mod-
∂ ( uII YII ) uw Yw km ( Yw -YII )
= + , (70) eling (Koltsakis and Stamatelos 1997a,b). Specifically,
∂z dfy2 dfy
this assumption states that the summation of the dif-
fusing gas species concentration difference terms is too
where a geometrical parameter is defined as
low to affect the overall mass balance of the carrier gas:
fy =
b (y)
.
∑ km ( ρI YI -ρsYs ) ≈0 and ∑ km ( ρII YII -ρw Yw ) ≈0 . However,
(71) Figure 16 illustrates that Depcik found this term to be
d
important because of the open system architecture of the
The width available to the flow, b(y), varies in the par- channels. As a result, it should be included in the mass
ticulate layer and remains constant in the wall: equations via a summation term on the right-hand side
of Eq. (39). Moreover, future efforts warrant revising the
⎧⎪d + 2 y , 0 ≤ y ≤ td mass-transfer calculations (km) to include the influence
b (y )= ⎨ . (72)
⎪⎩ d , td ≤ y ≤ td + t f of porous walls similar to the heat transfer coefficients
presented earlier.
Note, for this model, that Ys is the array of species For the species equations through the PM and wall
evaluated at y = 0 and Yw is the array of species evaluated at layers, a transport-reactive coupling is used by both Peters
y = tm+td. They include the effects of diffusing species to and et al. and Haralampous et al. In the Haralampous et al.
from the channel to the wall similar to catalyst modeling version, the formulation for the PM and wall layer is
A 1400 B 0.30
Experimental Not using simplification
Not using dilute mixture simplification Using simplification
1300 Using simplification
0.25
Final soot thickness (mm)
AO2=0.002285 m K-1s-1
1200 EO2=125 kJ mol-1
Temperature (K)
0.20
αO2=0.72
1100 12.7 cm
0.15
1000
900 0.10
2.54 cm 7.62 cm
800 0.05 PM Conversion
Inlet Simplification=55.0%
Additional terms=57.5%
700
0
1680 1720 1760 1800 1840 0 0.05 0.10 0.15
Time (s) Axial distance (m)
A 1800 B 1800
With diffusion Without diffusion
1600 1600
Tinlet Tinlet
1400 Tmaximum 1400 Tmaximum
Temperature (°C)
Temperature (°C)
z/L=0.1 z/L=0.1
1200 1200 z/L=0.5
z/L=0.5
z/L=0.9 z/L=0.9
1000 1000
800 800
600 600
400 400
200 200
0 50 100 150 200 250 0 50 100 150 200 250
Time (s) Time (s)
Figure 17 Comparison of temperature predictions during uncontrolled regeneration with and without O2 diffusion.
Computed temperatures along the filter for 12 kg m-3 initial PM mass (Haralampous et al. 2004a).
A 3.50×10-4 B 3.50×10-4
Soot Wall Soot Wall
3.00×10-4 3.00×10-4
Concentration NO2 (-)
2.50×10-4 2.50×10-4
Concentration NO2 (-)
2.00×10-4 2.00×10-4
1.50×10-4 1.50×10-4
Near filter entrance Near filter exit
1.00×10-4 (x/L=0.25) 1.00×10-4 (x/L=0.95)
320°C 320°C
5.00×10-4 360°C 5.00×10-4 360°C
400°C 400°C
0 0
-0.10 0 0.10 0.20 0.30 -0.10 0 0.10 0.20 0.30
Wall depth (mm) Wall depth (mm)
Figure 18 Computed profiles of NO2 concentration in the PM layer and the filter wall for different time points during balance point testing
(Haralampous et al. 2004e).
SO = -α 1k1S p ρ sYs ,O for Depcik and SO = -α 1k1S p ρ wYs ,O for 0.18
2 2 2 2
Haralampous et al. The model then uses Eq. (63) to find the
∑Y = 1 ,
j=1
j
(81)
Changing the equation from mole fraction to mass
fraction derivatives and assuming that the molecular
with the net species diffusion flux equal to zero for the
weight change in the y-direction is negligible (assumed by
Depcik model:
Bissett and Shadman and later validated by Depcik) the
NM
diffusion velocity equals
∑V Y =0.
j =1
j j (82)
D ⎡ ∂Y WY ⎛ Wˆ ⎞ ∂p ⎤
V =- ⎢ + ⎜⎜ -1 ⎟⎟ ⎥ . (84)
One approach that simplifies the computational task Y ⎢⎣ ∂y pW ⎝ W ⎠ ∂y ⎥⎦
ˆ
is to calculate only the trace species involved in the reac-
tions to reduce the numerical burden of holding equal to Note that if the pressure derivative term was omitted,
the chemical species constraints of Eqs. (81) and (82). For incorporating this equation into the PM and wall species
example, Konstandopoulos and Kostoglou (2010) illus- equations of (77) and (78) would result in the same for-
trate this approach in their article involving the analytical mulation as Haralampous et al. but on a concentration
solutions of the chemical species equations. instead of mass fraction basis. Recently, Konstandopoulos
The full formulation of the diffusion velocity using and Kostoglou (2010) presented a work on the coupling
mixture-averaged diffusion coefficients is equal to (Coltrin of convection diffusion and reaction phenomena using
et al. 1990, Reaction Design 2003): the above equations. They derived analytical solutions for
simplified cases leading to approximations applicable to
D ⎡ ∂X ( X-Y ) ∂p ⎤ DT ∂T the classic channel model for faster calculation.
V =- ⎢ + ⎥- . (83)
X ⎣ ∂y p ∂y ⎦ ρ YT ∂y The diffusion values used here must take into account
the porosity of the wall because of its close proximity. As a
result, calculation of an effective diffusivity in the PM and
In the above equation, only Depcik discusses the impact
wall layers use the Knudsen and bulk diffusivities along
of the pressure and temperature derivative terms in the calcu-
with the tortuosity of the flow (Smith 1970, Reif 1975, Ho and
lation of the diffusion velocity. The energy equation section
Strieder 1980, Mason and Malinaukus 1983, Cussler 1995):
illustrates that the gas temperature is effectively equal to the
filter temperature; hence, the temperature impact on diffu-
1 1 ⎛ τ Kn τ g ⎞
= + .
D ε p ⎜⎝ D Kn D g ⎟⎠
sion is nonexistent. However, because the pressure differen- (85)
tial between the inlet channel and the outlet channel may
be significant, Depcik finds that it has a small impact on the
In this equation, the binary diffusion constants are
results. Figure 19 illustrates this effect on exit oxygen levels
during a regeneration event when using one diffusion con-
1-Yj
stant for all species and individual diffusion constants and Dg = , (86)
∑
NM
incorporating the pressure differential. l≠ j
X j Dlj
with the Knudsen and bulk tortuosities assumed to 1. The collection efficiency of the filter;
depend only on the porosity, 2. Thermal and catalytic oxidation of the PM through
oxygen, NO2, metal additives, and surface catalysis; and
τ Kn = - lnε p + ⎛⎜
9 1 13 9 ⎞ 2 5 1
− ε and τ b = 1- lnε p , 3. The impact on particulate properties, specifically the
⎝ 9 8 ⎟⎠ p
(87)
8 2 2 permeability for pressure drop calculations.
When particulates flow into a DPF, they first load within 0.1
Y=
the wall layer where the pores are small enough for typical
diesel aerosol to be captured by interception and Brown- 0.01 Y=9.838X1.82
ian diffusion (Konstandopoulos and Johnson 1989). 0.001
After the wall has filled with PM, it then loads on the
inlet channel surface in a growing cake layer. In Bissett’s 0.0001
0.001 0.01 0.1 1 10
original article, he only models the soot cake layer via Eq. NR
X=g(ε) Pe1/3
(18) as a function of deposit thickness on the surface and 1+NR
oxidation only occurs through thermal oxidation via O2.
To capture the complete PM loading and regeneration Figure 20 Correlation for the “collection” efficiency of ceramic
profile, it is important to simulate monoliths (Konstandopoulos and Johnson 1989).
where uw ε0p is an interstitial or pore velocity. with the pressure taken to be the average pressure in the
Their model computes the Peclet number based on a filter wall, often a standard average of pI and pII at the axial
characteristic dimension of a unit spherical collector: location.
They write the interception efficiency as a function of
dc0 =
(
3 1-ε p)
0
dp , (92)
the same variables:
2 ε 0p
g (ε p0 )
2 3
⎛d ⎞
ηr = 1.5 ⎜ d0 ⎟ 3- 2 ε0 . (97)
⎝ dc ⎠ ⎡
( )
p
where dc0 is the initial value for an unloaded wall.
1 + dd dc0 ⎤ 3 ε p
0
Figure 21 Schematic representation of filter wall discretization into slabs of “unit collectors” (Konstandopoulos et al. 2000).
ηc = 1-exp ⎢-
⎡ 3η 1-ε pt ⎤
ff
⎥.
( ) (100)
allows calibration of the model to an experimental time-
history loading profile. Later work by Mohammed et al.
⎢ 2ε pdc0 ⎥
⎣ ⎦ (2006) indicate that a deficiency of this model is using
the partition coefficient, which is a function of filter
The mass deposited at each slab modifies the local wall parameters, as a filtration parameter for the PM
unit collector size as follows: cake layer when the cake layer becomes a filter for the
13
particles.
⎡ 3 md ⎛ dc0 ⎞ 3 ⎤ Konstandopoulos et al. (2002) revisit this model and
dc = 2 ⎢ +⎜ ⎟ ⎥ , (101)
⎣⎢ 4 π ρd ⎝ 2 ⎠ ⎦⎥ investigate the impact of velocity for the porosity via Eq.
(102). In particular, they find that, for high Peclet numbers,
where md is the mass of deposit and ρd is the particu- the deposit porosity is independent of this number and an
late packing density in the cake deposit and filter wall, asymptotic constant is the result. At low Peclet numbers
respectively, at time t. This influences the local porosity in the diffusion-dominated regime, a power law describes
as follows: the porosity when Pe > 0.3:
3 εp = APe-B+C, (106)
⎛d ⎞
ε p = 1- ⎜ 0c ⎟ 1-ε 0p .
⎝d ⎠
( ) (102)
where A, B, and C are constants that depend on particle
c
morphology and deposition surface.
The geometric factor of Eq. (90) and Peclet number of A few years later, Konstandopoulos et al. (2005) revisit
Eq. (91) then use the updated porosity and unit collector this model to generalize the results down to the diffusion
size in computing the Brownian collection efficiency of limited deposition limit:
Eq. (89) and interception efficiency of Eq. (97). -n
⎛ Pe ∞ ⎞
They compute the second term in Eq. (99) using Eq. ε p = 1-( 1-ε ∞p ) ⎜ 1+ , (107)
⎝ Pe ⎟⎠
(100) by replacing the wall thickness with a differential
distance (i.e., replace tf with Δy) and the evolution of
with ε∞p as the Peclet number asymptote of the porosity
deposited mass via Eqs. (101) and (102):
and Pe∞ as the characteristic crossover Pe number defin-
d
∂m ing the scale beyond which the convective mechanism
d .
= -ηcm (103)
∂y will take over the diffusive mechanism of deposition. This
allows a correlation between PM porosity and the flow
At this point, the model is able to explain the collec- through the wall as illustrated in Figure 22.
tion of PM within the filter wall often called deep-bed fil- Fukushima et al. (2007) revisit the filtration efficiency
tration. The extension to the cake layer comes by defining of Eq. (100) to write it as a function of an effective length
a partition coefficient to determine the fraction of mass scale for deposition:
collected in the first “slab” that contributes to cake forma- -1
tion. Based on geometrical grounds, this value depends ⎛ 1 1 1 ⎞
deff = ⎜ + +
on a dimensionless blocked-area fraction expressed as ⎜ d d d ⎟⎟ , (108)
⎝ p c cdl ⎠
(d ) - (d )
2
0 2 where dcdl is a characteristic decay length measured quan-
c y =0 c
Φ= , (104) titatively from different filter material images:
(ω b ) - ( d ) 0 2
2
c
(
⎡ d -1.081 ) ⎤
2 3
1.0 the early 1990s regarding the use of fuel additives for
0.9 lowering the regeneration temperature of diesel particu-
late through catalysis (Harvey et al. 1994, Tan et al. 1996).
0.8
Koltsakis and Stamatelos (1996a,b) take these concepts
(1-ε)/(1-εinf) [-]
In the following year, Konstandopoulos et al. (2000) to account for the oxidation components. However, their
expand this work to incorporate NO2 passively regener- articles do not illustrate a merging of this mass fraction
ated DPFs by modifying their particulate evolution equa- equation and Eq. (116).
tion using Eq. (65) and rewrite the evolution equation as Haralampous et al. (2003) convert the thickness-
a function of mass instead (dmd/dt) of thickness in each based expression of Eq. (115) into a mass-based expres-
layer. In their article, they state the relationship between sion written as a function of PM mass per unit volume:
mass of PM and its thickness as
1⎛ md ,t ⎞
∂md
∂t
{
= -sF ∑ ⎡⎣WC s
k =O2 ,NO2
( )
(α kWk )⎤⎦ ρw uwYI ,k
td = ⎜ d - d 2 - ⎟, (117)
2 ⎜⎝
)}
N ic Lρd ⎟ (114)
⎠ (
⎡ 1-exp -α k kk S ptd uw ⎤ ,
⎣ ⎦
where md,t is the total PM mass in the DPF, Nic is the total
where sF is the specific area of the filter that depends on
number of inlet channels, and L is the effective channel
the number of cells per square inch converted into the
length.
proper units:
Kandylas and Koltsakis (2002) follow this work a
couple years later by expanding Eq. (110) to include NO2 SF = 2d·1550·cpsi. (118)
regeneration through Eqs. (66) and (67):
It is important to note that if the bulk density of the
PM is constant, the formulations via soot thickness and
ρd
∂t d
{
= - ∑ ⎡WC
∂t k=O2 ,NO2 ⎣ s ( )
( αkWk )⎤⎦ ρw uwYI ,k mass of soot cake are the same. As the next section illus-
. (115) trates, by writing in a mass-based manner, the model can
⎣ ( ⎦ )}
⎡ 1-exp - αk kk S ptd uw ⎤ . take into account how the properties inside the soot cake
can vary. Hence, future efforts can include modeling soot
Again, the difference between the Konstandopoulos property changes during oxidation.
et al. and the Koltsakis et al. models is a matter of nomencla- They modify this formulation on the following year
ture and presentation of the equations, but the physics are (Haralampous and Koltsakis 2004a,b, Haralampous et al.
similar and both follow Bissett and Shadman’s initial effort. 2004a,d,e) by simplifying the right-hand side:
In the following 2 years, Peters et al. (Peters
1 dmd
2003, Peters et al. 2004) simulate the PM mass per unit = - ∑ S pkk pk (119)
md dt k =O2 ,NO2
length on the surface by including a flow of PM to the
surface:
as a linear dependence on the partial pressure of the oxi-
∂md ∂ (uI md ) dation species.
+ cm = 4uw dI ρdc , (116) Recent work by Depcik and Assanis (Depcik and
∂t ∂z
Assanis 2008, Depcik 2010) use these earlier efforts to
where cm is a migration constant that accounts for drag simulate the propagation of PM along the filter with regen-
forces that result when PM translates along the channel eration occurring only via traditional thermal oxidation
(Figure 23) and ρdc is the PM density in the channel. PM from O2 in a cake layer. An equation for PM mass per unit
oxidation is included in their model by writing an equation length on the surface simulates the amount of PM in each
for PM mass fraction and using an Arrhenius expression numerical cell in one channel,
Migration forces
Inlet
Soot
layer
Wall
Outlet
Figure 23 Forces on PM particles in the Peters et al. models (Peters 2003; Peters et al. 2004).
dmd
=ς -χmd , 1⎛ m ⎞
(120) td = ⎜ d - d 2 - d ⎟. (125)
dt 2 ⎝⎜ ρd ⎟
⎠
which gives the total PM mass in the filter by multiplying
The result of these efforts is an investigation into the
this value by the total number of inlet channels. In this
PM loading profile during constant and random loading
equation, the PM mass flow rate per mesh interval, ς, is
events (as shown in Figure 24), which illustrate that chang-
determined as a function of changing surface area similar
ing the flow rate does not significantly impact the loading
to Peters et al.:
profile along the filter and subsequent regeneration.
ς = 4ρsusdIYd (121) This historical summary illustrates a number of differ-
ent models in the literature that simulate similar physical
using the assumption that a constant mass fraction of PM
concepts. One version of the model simulates a 1D particu-
occurs in the inlet channel:
late loading equation in the axial direction with the wall
dYd phenomenon lumped into source terms on the right-hand
= 0. (122)
dz side:
∂md ∂ (udmd )
The mass fraction of PM into the filter comes from the + ds + m
=m dr , (126)
mass flow rate of PM into the DPF as compared with the ∂t ∂z
total flow rate of gas:
where the time rate of change of particulate mass at each
PM
m axial location (point in z) is impacted by the flow of par-
Yd = . (123)
m in ticulate down the channel (possibly a function of channel
velocity and/or migration), the amount stored on the
They calculate a PM combustion time scale from the surface that may include a storage efficiency component
historical efforts via Eq. (64): (m ds ), and the thermal and catalytic oxidation reactions
occurring on the surface (m dr ) . Of interest, researchers
WC s k1S p ρ sYs ,O only model the PM flow to the surface a function of wall-
χ =- ( ) 2
(124)
WO
2
ρd flow rates (uw); however, as indicated in the species equa-
tion section, diffusion between the gas and the surface
using the localized gas density and mass fraction of might also be important and need to be included.
oxygen in the cake layer. The relationship between PM Alternatively, researchers simulate the time rate of
mass per unit length and its thickness are a function now change at each axial location in the cake and wall layers
of the local PM mass: also in 1D (normal direction):
A 1.02 0.515
B
Initial loading
1.00 Steady loading
Particulate layer thickness (mm)
Inlet channel
Random loading
0.98
Channel pressure (bar)
0.510
Initial loading (DP=0.1445)
0.96 Steady loading (DP=0.1444)
Random loading (DP=0.1445)
0.94 0.505
Soot loading event
0.92 20 min-goal=174.37 g
Steady=174.44 g (+0.04% difference)
Random=174.46 g (+0.05% difference)
0.90 0.500
Outlet channel
0.88
0.86 0.495
0 0.05 0.10 0.15 0.20 0.25 0.30 0 0.05 0.10 0.15 0.20 0.25 0.30
Axial distance (m) Axial distance (m)
Figure 24 Comparison of three different PM loading events with respect to (A) pressure and (B) particulate layer thickness (Depcik and
Assanis 2008).
( )
Κ f = f ε p dc2 , (128) In the following year, Konstandopoulos (2003)
extends the model to the Forchheimer coefficient
using the Kuwabara model for beds of spherical particles: by basing it as a function of the permeability of the
medium:
2 ⎣ 2- 5 ( 1-ε p ) -ε p - 5 ( 1-ε p ) ⎦
⎡ 9 13
1
2
⎤
f (ε p ) = . (129) βf =
c
, (134)
9 ( 1-ε p ) ε 1.5
p Κf
In the next decade, Sorenson et al. (1994) experimen- where c is a constant between 0.134 and 0.298 depending
tally measure the PM layer density, whereas Pattas et al. on the wall.
∂S p ∂S p ∂ρd
= S 0pκ ( ρd0 ) ( ρd )
κ 1-κ
+ y , (141)
Before compression | After compression ∂t ∂y ∂t
Figure 27 Intralayer density field before and after compression where the superscript0 indicates the initial value of that
(Koltsakis et al. 2006). parameter. Of importance, the problem of determining
the permeability of the soot-loaded filter wall is more 2.6.1 Axial filter energy equation
complicated than compared with the soot cake perme-
ability. The penetrating soot by depth filtration inside In the axial direction model, Bissett assumes that the filter
the wall micropores has a dramatic effect on wall per- wall and PM layer have the same temperature due to their
meability that is responsible for the steep pressure drop proximity to each other and the fact that PM is present
increase at the early stage of DPF loading before the soot within the filter wall due to deep-bed filtration. His model
cake layer starts to form. The literature illustrates other includes the effects of axial conduction in the wall, heat
formulations for the permeability of a loaded filter wall transfer from the gases in the channel, and the effects of
using the concepts in the previous two sections (Moham- PM oxidation on the energy of the wall. He does include
med et al. 2006): an enthalpy flow through the wall in the initial steps of his
formulation, but the terms involved cancel each other out
⎡ f (ε p ) ⎤ ⎛ 1-ε p0 ⎞
2
Κ d ⎛ dc y =0 ⎞ and do not appear in the final resultant equation.
=⎜ ⎟ ⎢ ⎥⎜ ⎟. (142)
⎢⎣ f (ε p ) ⎥⎦ ⎜⎝ 1-ε p ⎟⎠
Traditionally, researchers write this equation as a
Κ f ⎜⎝ dc0 ⎟⎠ 0
et al. 1993, Jørgensen and Sorenson 1997, Romero et al. and Koltsakis and Stamatelos 1997a,b, Konstandopou-
1997, Opris and Johnson 1998a,b, Miyairi et al. 2001). The los et al. 2001, 2005, Kostoglou et al. 2003). Based on the
first instance of 2D adaption of the classic Bissett model is original channel energy equations, the enthalpy flow
by Konstandopoulos et al. (2001), who simplify a multidi- through the wall cancelled itself out and, as a result, did
mensional DPF model by using a continuum description not appear in this equation. A few years later, this equa-
of the filter (Gropi and Tronconi 1996). This formulation tion was written in differential format including this term
involves implementing an effective thermal conductivity (Koltsakis and Stamatelos 1997a,b). Depcik and Assanis
tensor (Haralampous et al. 2003, Konstandopoulos et al. later modify this expression to account for the local
2003a, Kostoglou et al. 2003): density and velocity along with the impact of varying
surface area:
1 ∂ ⎡ ∂Tf ⎤
Q multi =
r ∂r ⎢⎣
(
r λeff Vf +Vd
∂r ⎥⎦
, ) (145)
Q wall = ⎛⎜ c ⎞⎟ ( 4 ρ sushI dI -4 ρ w uw hw dII ) .
N
(149)
⎝ 2 ⎠
where λeff is an effective radial conductivity that depends
on the honeycomb bulk thermal conductivity as well as The energy flow through the wall is often written as
its geometrical details (wall thickness, diameter, and unit a function of constant pressure-specific heat times a tem-
porosity). They modify this expression to account for the perature difference; that is, hI-hw = cp(TI-Tw). Because the
contribution of the PM layer: chemical species modeled has different heats of forma-
tion and sensible components, a more rigorous interpre-
λe f f =
1- ε p
⎢( f )
⎡ d + t 2 -d 2
λ +
d 2 - ( d + 2 td )
2
λ
⎤
⎥
tation requires the use of enthalpies. Depcik found that
2 d , (146)
(
1 + ε p - ε p ⎢ ( 1-ε ) d + t ) ( p ) f ⎥⎥⎦ ( ) making the constant pressure assumption underpredicts
2 f
⎢⎣ p f
2 1- ε d + t
the PM conversion during an oxygen regeneration case by
approximately 3%. In the case of models that do not cal-
where the instantaneous value of the macroscopic honey-
culate the flow profiles through the wall, the wall average
comb structure porosity is
density and velocity will replace the respective terms in
d 2 + dI2 the above equation.
εp = . (147)
( ) Because active heating is a possible method of regen-
2
2 d +t f
erating the filter, Romero et al. (1995) illustrate how to
include this ability in the model formulation:
The multidimensional version of the filter energy
equation effectively eliminates Bissett’s original assump-
tions of an adiabatic device with spatially uniform inlet
(
Q supply = q Vf +Vd ,) (150)
where F is the view factor of the channels with the above and the reaction enthalpy that results (Koltsakis and Sta-
equation modified to remain consistent with the volu- matelos 1996a,b). However, its inclusion does not appear
metric model description. Model results illustrate that in the dimensional version of this equation, but its
the effect of radiation becomes important above approxi- importance is noted with respect to the axial filter energy
mately 600°C–700°C; hence, it can improve model accu- equation.
racy during uncontrolled regeneration events. In the previously mentioned literature, research-
The contribution due to reaction of the particulate ers write the expression for the heat of reaction added
develops from the historical summary described in Sec- to the filter energy equation a number of different ways.
tions 2.4 and 2.5.2. Skipping ahead to recent efforts, This all depends on how they express the reaction rate
Depcik and Assanis express the contributions due to PM equations; for example, Haralampous et al. deviate from
combustion as the traditional Bissett version, whereas Konstandopol-
ous et al. follow Bissett’s method with a few differ-
N ⎛ m ΔH ⎞
Q reac = - ⎛⎜ c ⎞⎟ ⎜ d reac ⎟ SC s ,
ences in their molar PM combustion reactions. Compil-
(152)
⎝ 2 ⎠⎝ ρd ⎠ ( )
ing and analyzing these versions is beyond the scope of
this article; however, their contributions are noted. Of
with oxidation only occurring on half the number of interest, Depcik and Assanis use average values for the
total channels because of the lack of PM in the outlet density and oxygen mass fraction within their calcula-
channel. tion of Eq. (153). However, Haralampous et al. correctly
This expression includes the heat of reaction that is a note that this expression should include integration in
function of the combustion reaction modeled. Section 2.4 the y-direction over the cake layer and deep-bed filtra-
indicates the derivation of the gaseous amount of oxygen tion to capture the heat released as a function of local
lost as SO = -α 1k1S p ρ sYs ,O . Because the mass added to the oxidation mass fractions. This was done originally by
2 2
gas equals the mass of PM reacted, the generalized oxygen Bissett as he solved for the local O2 profile analytically
combustion Eq. (63) can be used to calculate the respec- in his efforts.
tive PM loss. Incorporating both oxidation options, the The catalytic reaction term involves the impact of
source term for heat addition equals direct PM conversion through a catalyzed surface (Kon-
standopoulos and Kostoglou 1999), metal additives
WC s WC s
SC s = -
( )
( )
k1S p ρsYs ,O2 - ( )
k2 S p ρsYs ,NO2 . (153) interacting with the PM in the cake and wall layers
WO2 WNO2 (Koltsakis and Stamatelos 1996a,b), and any gaseous
phase reactions occurring in the wall (Konstandopoulos
and Kostoglou 2000, Konstandopoulos et al. 2000,
This value is equal to the total amount of PM that
Kandylas et al. 2002 ). Summarizing the first two into a
undergoes combustion requiring a modification for use
generalized format for incorporation is difficult because
in the mass equations through the PM and wall layers of
of the different versions in the literature. However, one
Eq. (40) or Eq. (41) based on the number of discretizations
can look toward catalyst modeling to express the reac-
in the respective layer:
tion rate through the wall layer (Depcik and Srinivasan
2011):
SC s
Smass = ( )
. (154) td + t f
N s -1 S N NM ⎛ ⎞
Q cat = R direct PM + R additives + ca c ∑ ⎜ ∫ R j hj dy⎟ . (155)
t f j =1 ⎝ y =td ⎠
Note that the above discretization is for the Depcik
and Assanis model where PM loading only happens in a
cake layer on the surface. Assuming that catalytic reactions only occur in the
Koltsakis and Stamatelos (1996a,b) modify the Bissett wall layer, its impact includes a summation through
and Shadman (1985) model by accounting for incom- this layer as a function of local rates that is averaged
plete PM oxidation. In the energy equation formulation, over the thickness of the wall (effectively an average
this only influences the reaction enthalpy in Eqs. (21) heat of reaction over the wall later). The term Sca is the
and (152) by including a reduction in exothermic energy catalytic surface area of the reactive material in the
from the partial production of CO. In the same year, they wall.
modify the perturbation analysis involving this equation The last term is only important when computing the
to include a metal additive catalytic reduction component gas temperature profile through the wall as in Section 2.3
via Eqs. (57) and (58). Depcik and Assanis omitted the cor- Bissett and Shadman assumed a bulk density of
responding heat transfer from the gas to the wall in their 550 kg m-3 and a bulk PM conductivity of 0.84 W mK-1. As
model: mentioned above, these early estimations of PM properties
have been since then revised by the increasing knowledge
( ) (
a ⎡hcVd Ts -Tf + hcVf Tw -Tf ⎤
Q inner = ⎣ ⎦, ) (156)
of the structure of the deposited PM. It would be reason-
1-ε p able to use a typical average soot cake density of 75 kg m-3,
which corresponds a porosity value of 0.97. For this poros-
ity value, the bulk thermal conductivity is 0.108 W mK-1.
which uses average gas temperature values in the PM Ts ( ) Using these values, the parameter Ξ is expected to be almost
( )
and wall Tw layers, respectively. Because the gas tem-
two orders of magnitude higher than the values assumed in
perature becomes the filter temperature nearly instanta-
the justification of the Bissett and Shadman model.
neously, this term is not important but is required in this
Haralampous and Koltsakis solve for the advection-
review for completeness.
conduction energy equations of Eqs. (21) and (22) to cal-
culate for the temperature field within the soot cake for
various regeneration modes (Figure 29). The calculated
2.6.2 Temperature gradients across the soot cake and temperature gradients result from the inability of the low-
the wall conductivity PM to diffuse the heat generated in the layers
close to gas entrance. Bearing in mind that the reaction
Haralampous and Koltsakis (2002) explicitly indicate the rate depends exponentially on temperature, the front PM
porosity of the layers in the calculation of the density and layers will react faster than the layers close to the wall.
thermal conductivity. The apparent density of the PM layer Consequently, the overall PM consumption rate is always
is then a factor of the solid phase properties assumed as higher compared with what would be expected from
solid carbon: uniform soot cake temperature model. They conclude that
the differences could be significant in the case of regen-
ρd = ( 1-ε p ) ρC ,
( s)
(157) erations under extreme soot loadings of the order of 10 g l-1
or more, as demonstrated in Figure 30. The same figure
with the conductivity a weighted factor between the air shows a critical effect of soot porosity.
and carbon properties:
λd =ε pλair + ( 1-ε p ) λC .
( s)
(158) 2.7 Numerical methods
Of importance, Bissett and Shadman explicitly indi- In Bissett’s 1984 article, he nondimensionalizes the gov-
cate these variables as bulk values, taking into account erning equations and uses the differential algebraic solver
the porosity of the medium; hence, the variables in Eq.
(21) are equivalent to these above equations. 1800
10 s 40 s 42 s 46 s 60 s
As noted by Bissett and Shadman, the following 30 s 41 s 44 s 49 s
dimensionless number is an indication for the tempera- 1600
Filter 100/17 46s
Soot loading 12.5 gr lt-1 44s
ture uniformity in the PM layer:
Temperature (K)
1400
42s
c ρ u t 49s
Ξ= p w w d , 1200
Aflt λd 41s
(159)
40s
1000 60s
where Aflt is the filtration area and Ξ is the Peclet number 30s
calculated for the particulate layer width. When Ξ has a 800 10s
α=97%, ε=0.3682 α=80%, ε=0.0131 formulas) option is employed for the ODE solver. The
α=94%, ε=0.1118 α=97%(0D), ε=0.3682
α=90%, ε=0.0460
method of transforming a DAE system to a system of ODEs
1800 by differentiating the algebraic equation is a common
Filter 100/17 procedure for numerical integration of DAE sets (Sham-
Soot loading 15 gr lt-1
1600 pine et al. 1999). Both model solutions produced nearly
identical results; however, the ODE solver methodology
Outlet temperature (K)
1200
3 Using the DPF model
1000
A thorough history of the DPF model based on Bissett
and Shadman’s early work illustrates that the model now
800
20 25 30 35 40 45 50 55 60 includes a greater ability to simulate active and passive
Time (s) regeneration events along with soot loading within the
filter wall and on the cake layer. This section indicates the
Figure 30 Effect of soot porosity on the regeneration behavior lessons learned from using the model and how this allows
(Haralampous and Koltsakis 2002).
researchers to advance the DPF field of practical usage.
LSODI to solve the fluid dynamics and ideal gas law equa-
tions simultaneously (Hindmarsh 2002). This solver is
3.1 NO2 impact on regeneration
still in use today as evidenced by the Depcik and Assanis
articles. Koltsakis and Stamatelos (1997a,b) use a differ-
Filter regeneration with oxygen occurs with noticeable
ent approach starting with their 1997 article by using time
rates at temperatures above 500°C–550°C, which are
marching via a fourth-order Runge-Kutta technique and
rarely met under typical operating conditions. On the
employing a computationally efficient iterative approach
other hand, NO2 is highly reactive with soot (Cooper and
to solve the flow field in the inlet and outlet channels. The
Thoss 1989). NO2 is able to oxidize soot at temperatures as
models of Konstandopoulos et al. solve for the equations
low as 250°C, which can be encountered in diesel exhaust
using the combination of a two-point boundary value
during normal driving cycles. However, NO2 is present in
problem for the mass, momentum, and energy equations
diesel raw exhaust at very low concentrations (5–15% of
along with incorporating the method of lines via a second-
total NOx or < 50 ppm), which are not sufficient to provide
order finite difference approximation and a variable step,
the required reaction rates.
prespecified accuracy explicit Runge-Kutta integrator for
The concentration of NO2 in the exhaust gas enter-
the particle mass balance (Konstandopoulos and Kosto-
ing the filter can be increased by placing upstream of the
glou 1999). They employ the implicit Nyström method to
filter an oxidation catalyst, which oxidizes NO to NO2.
solve for the flow equations and incorporate a Netwon-
This is the main idea underlying the continuously regen-
Raphson method for convergence.
erating trap concept, which is the trade name of a system
Recent work by Sprouse et al. (2011) includes two dif-
composed of a selective oxidation catalyst placed directly
ferent numerical solvers incorporated in MATLAB: one for
upstream of the particulate filter (Cooper et al. 1990).
the differential algebraic equation (DAE) system of equa-
At temperatures of 300°C–350°C, the oxidation catalyst
tions and one through the differentiation of the algebraic
oxidizes a proportion of the NO in the exhaust stream to
constraint equation (ideal gas law) to complete an ordi-
form NO2, increasing the NO2 fraction to approximately
nary differential equation (ODE) system:
50% of total NOx (Hawker et al. 1997, Allansson et al.
dρ 1 ⎛ dp dT ⎞ 2002).
= ⎜ T -p ⎟. (160)
dz RT 2 ⎝ dz dz ⎠ To model efficiently catalyzed and uncatalyzed DPF
systems, the knowledge of the rate laws for the uncata-
In the numerical code within MATLAB, the ode15i lyzed soot reactions is necessary. The published works
option (solves implicit equations using backward differen- deal mostly with reaction rate studies at laboratory scale
tiation formulas) is employed for the DAE solver, whereas using synthetic soot or synthetic gas. A comprehensive
the ode15s (variable-order numerical differentiation study dealing with the reaction kinetics of various soot
types under a wide range of laboratory conditions was asymptotically from 600°C to 300°C with a response time
recently published by Tighe et al. (2011). Kandylas et al. of the order of 20 s.
(2002) presented an in situ methodology to determine the Figure 17 presents the computed wall temperatures
kinetics of soot oxidation with O2 and NO2 in real engine at different axial positions in the filter channel, assum-
conditions. The rate laws for soot oxidation with O2 and ing an initial PM loading of 12 kg m-3 with and without
NO2 were derived from measurements of soot consump- inclusion of O2 diffusion effects. Notably, this loading cor-
tion during steady-state operation at various conditions responds to an extreme case, whereas most researchers
and filter configurations. The soot consumption was cal- study loadings that range up to 7 g l-1 for production cases.
culated based on a methodology to calculate the soot The oxygen diffusion phenomenon affects the evolution
mass as a function of the measured pressured drop, mass of filter temperatures, especially the maximum tempera-
flow rate, and temperature, as described in Haralampous ture with a difference indicated between the two simula-
et al. (2004b,c). NO2 was additionally produced by an oxi- tions of approximately 400°C. Such deviations are critical
dation catalyst placed in front of the filter. The results of for model applications concerning the ability of the filter
these measurements can be summarized in the form of an material to withstand severe regeneration conditions
Arrhenius plot, as shown in Figure 31. The computed profiles of O2 concentration along
the filter for the two modeling assumptions presented in
Figure 32 reveal the O2 diffusion impact. The results cor-
3.2 Oxygen diffusion impact on regeneration respond to the time of maximum reaction rate, and in
the left graphs, the inlet channel O2 concentration at the
“Uncontrolled” regenerations typically occur at condi- specific time drops suddenly from 15% to almost zero at
tions of low flow rate and high oxygen content with high an axial position approximately 100 mm from entrance.
initial filter temperatures. These situations happen when The concentration again increases slightly toward the rear
the engine switches to idle mode after a period of high- part of the filter because of diffusion from the outlet back
temperature operation. The simulation work presented to the inlet channel. The figure also illustrates that O2 dif-
in Haralampous and Koltsakis (2004a,b) discusses the fusion enhances the reaction rate in the middle part of the
importance of including the transport-reaction coupling channel. It is also interesting to note that a higher O2 con-
Eqs. (69), (70), (75) and (76) for the calculation of species centration of the outlet channel near the filter exit exists
concentration in the DPF channels. To simulate an with respect to the corresponding concentration in the
“uncontrolled” regeneration condition of a filter 2.5 l in inlet channel. This concentration difference drives O2 dif-
volume, the researchers assumed that the initial filter tem- fusion back from the outlet channel to the PM layer, thus
perature is uniform at 600°C along with a constant mass enhancing the overall reaction rate.
flow rate at 0.01 kg s-1 and oxygen content in the incom- Moreover, Figure 32 illustrates the effect of O2 dif-
ing gas equal to 15%. The inlet gas temperature decreases fusion on the evolution of PM layer thickness along the
filter channel during regeneration starting with a high PM
loading. The left graph shows that the diffusion-induced
-2.0
C+O2
increase of O2 availability at the middle part of the filter
-2.5 E=152 +/- 5 kJ mol-1 results in a preferential PM consumption compared to the
rear part of the filter. This is not the case in the simulation
-3.0 result without O2 diffusion. Examining the O2 intralayer pro-
log rate (s-1)
0.16
Concentration (-)
0.12
Soot thickness
0.12 Wall flow rate 15
Wall flow rate
0.08 10
0.04 5
0 0
0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160
Length (mm) Length (mm)
Figure 32 Profiles of O2 concentration in the inlet and outlet channels during uncontrolled regeneration with 12 kg m-3 initial PM mass at
time = 42 s.
Also shown is the wall-flow velocity and the PM layer thickness at the same time (Haralampous and Koltsakis 2004a,b).
12 mils (0.3048 mm) wall thickness in a propane burner. gas temperature through postinjection to initiate the regen-
The regeneration temperature was 600°C and the flow rate eration process. FBCs act to accelerate the reaction of
equal to 0.014 kg s-1. The regeneration events involved dif- oxygen with soot by reducing the regeneration tempera-
ferent PM loadings and the data include measuring 20 dif- ture to approximately 500°C instead of 600°C in the case of
ferent axial and radial positions for the filter temperature. noncatalytic systems. The situation is somewhat different
To simulate these experiments, it was necessary to employ in the case of catalytically coated filters. Most commercial
the extended 2D regeneration model (Haralampous et al. technologies involve coating or impregnating the filter wall
2003) involving Eq. (145) in the axial filter energy equation. with platinum-based catalysts that are effective in promot-
Figure 33 provides the results of two different simulations ing the oxidation reaction of NO, which is present in the raw
with and without diffusion effects. For comparison, the diesel exhaust, to NO2. As mentioned previously, the latter
figure presents the maximum computed filter temperature is a strong oxidizing agent and is able to react with depos-
versus the respective measurement as a function of initial ited soot at temperatures as low as 250°C. However, because
PM loading. When neglecting diffusion, the calculated NO2 forms on the catalytic wall, which is downstream of the
maximum temperature agrees with the measured values
for low to medium PM loadings up to 8 kg m-3. At higher PM 1400
loadings, which are namely of interest for filter integrity in Measured
Computed with O2 diffusion
uncontrolled regenerations, the discrepancy is critical. On 1300 Computed without O2 diffusion
Maximum temperature (°C)
1000
3.3 Catalytic regeneration
900
Due to the relatively low temperatures of the diesel exhaust
gas, catalytic assistance in the form of fuel-borne catalysts 800
4 6 8 10 12 14 16
(FBCs) (Salvat et al. 2000) or catalytic filter coatings (cata-
Soot loading (kg m-3)
lytic DPF or CDPF) (Suresh and Johnson 2001) is currently
the state-of-the-art for commercial systems. In these cases, Figure 33 Comparison of experimental and predicted maximum
diesel engine fuel injection and control systems continue to temperatures during regeneration as a function of PM loading
support the regeneration system by increasing the exhaust (Haralampous et al. 2003).
soot cake layer, the reaction of this species with soot would 400 600
Table 1 Evaluation of the importance of regeneration mechanisms for different catalyst-based systems depending on the regeneration
mode (Haralampous et al. 2004a).
< 450°C > 450°C Low msoot, T High msoot, T < 450°C > 450°C Low msoot, T High msoot, T
C+O2 o + + o o ++ ++ +
C+NO2 + o o o ++ + o o
O2 diff o o o ++ o o o +
NO2 diff + o o o ++ o o o
C+FBC ++ ++ ++ + - - - -
NO↔NO2 - - - - ++ + o o
o, not important; +, important; ++, very important. Controlled regeneration: steady inlet conditions, medium to high flow rate; uncontrolled
regeneration, low flow rate, initial DPF temperature>>BPT. All cases assume upstream DOC.
12 1.2×10-4
Diffusion-reaction model
8 8.0×10-5
Soot mass (g l-1)
Reaction model
6 6.0×10-5
4 4.0×10-5
0
0 0 20 40 60 80 100 120 140 160
0 200 400 600 800 1000
Length (mm)
Time (s)
slowly, approaching chemical equilibrium at wall exit as Figure 38 presents the regeneration rate as a func-
the entering NO undergoes an oxidation reaction that is tion of temperature for two different initial soot loadings.
equilibrium limited, for example, NO+1/2O2⇔NO2. The reaction rate increases with temperature as expected,
In the diffusion-reaction model, the NO2 produced in although not in an exponential manner. This is because
the wall is “back diffused” in the soot layer and a slow the overall soot consumption rate is controlled by the rate
increase in NO2 concentration is evident from the soot of soot oxidation with NO2 along with NO2 production in
layer region. Meanwhile, a reduction of the concentra- the wall, which is limited by thermodynamic equilibrium
tions in the wall is observed due to the same phenomenon at higher temperatures. Figure 39 presents the percentile
additionally resulting in smaller concentrations for the contribution of back-diffusion phenomena to overall reac-
outlet channel. An added NO2 flow is induced due to the tion rate. This shows that the reaction-only model predicts
concentration gradient between the inlet channel and the reaction rates that are 25–35% lower compared with those
soot layer resulting in the axial profile outcome of Figure predicted by the reaction-diffusion model with the discrep-
36. Moreover, similar behavior is observed in Figure 37B ancy larger at lower temperatures. The effect of flow rate
(near filter exit). Because the inlet NO2 concentration is and NOx concentration in terms of the mean reaction rate
very low from the engine (as also shown in Figure 36), is indicated in Figure 40. This figure finds that the impact
back-diffusion is responsible for supplying the soot layer of the diffusion phenomena is more pronounced at lower
with additional NO2. flow rates. This is to be expected because back-diffusion
A
Intra-layer NO2 mole fraction (-)
0.0008 B 0.0008
0.0006 0.0006
0.0005 0.0005
0.0004 0.0004
0.0003 0.0003
Near filter entrance Near filter exit
0.0002 (z/L=0.1) 0.0002 (z/L=0.9)
0 0
-0.10 0 0.10 0.20 0.30 -0.10 0 0.10 0.20 0.30
Wall depth (mm) Wall depth (mm)
Figure 37 Computed profiles of NO2 concentration in the PM layer and the filter wall with and without accounting for diffusion: (A) near
filter entrance and (B) near filter exit (Haralampous and Koltsakis 2004a,b).
in the soot layer is dependent on the Peclet number of the of the oxidation catalyst upstream to the DPF ensures
flow through the porous layer and wall. The results dem- the efficient reduction of CO and hydrocarbons during
onstrate that there is not a large dependency on initial normal operation as well as during forced regeneration
NO concentration. In Figure 41, the effects of tortuosity with postinjection (Banno et al. 2004). In the latter case,
on the percentile contribution of diffusion are shown for the exothermic reactions in the oxidation catalyst subse-
tortuosity values ranging from 1 to 10. The sensitivity of quently transfer energy to the DPF through the exhaust
the results on tortuosity is very important, especially for gas enthalpy and aid in regeneration of the DPF. In com-
values below 3. These results emphasize the importance parison, the purpose of a CDPF is to facilitate regeneration
of accurate determination of the species effective diffusiv- while minimizing the CO emissions during this regenera-
ity in the porous wall and the soot layer. tion event. Therefore, because catalyzed filters are effi-
cient in oxidizing CO and hydrocarbons, it is theoretically
possible to completely eliminate the oxidation catalyst
3.4 Use of DPF as an oxidation catalyst (Maly et al. 2004, Mizutani et al. 2004).
Modeling of the combined mass-transfer and reac-
The installation of most commercial DPFs occurs down- tion phenomena in a CDPF was done originally by
stream of a diesel oxidation catalyst (DOC). The presence
38
60
Diffusion-reaction model:5 g l-1 36
Back diffusion contribution (%)
34
40
32
30
30
20
5 g l-1
28
10 g l-1
10
26
0 275 300 325 350 375 400 425 450 475
275 300 325 350 375 400 425 450 475 Temperature (°C)
Temperature (°C)
Figure 39 Percentile contribution of back-diffusion phenomena to
Figure 38 Mean reaction rate as a function of temperature for two overall reaction rate as a function of temperature for two different
different initial PM loadings (Haralampous and Koltsakis 2004a,b). initial PM loadings (Haralampous and Koltsakis 2004a,b).
38
modeling approach claiming that the flow-through-type
reactor is more efficient due to increased reaction resi-
Back diffusion contribution (%)
36 500 ppm NO
dence time (Miller and Li 2000). A recent article (Dardio-
1000 ppm NO
tis et al. 2006) involves modeling a CDPF using the trans-
34 port-reaction coupling model in parallel with respective
flow-through DOC simulations. This work uses DOC and
CDPF models to analyze both systems concerning their
32
gaseous species oxidation performance and is recapped
here. Further details about the setup and conditions are
30 given in Tables 2 and 3.
Two main factors differentiate the behavior of the
28 CDPF compared with the DOC when it comes to the oxi-
0 20 40 60 80 100 120
dation reaction. The first factor involves the axial velocity
Mass flow rate (g s-1)
component; because half of the channels are plugged in
a CDPF, the inlet channel axial velocity is higher com-
Figure 40 Percentile contribution of back-diffusion phenomena to
overall reaction rate as a function of flow rate for two different NOx pared with the DOC. The second factor regarding the
concentrations (Haralampous and Koltsakis 2004a,b). CDPF involves the continuous loss of axial inlet bulk
flow through the wall to the outlet channel until the axial
velocity drops to zero at the rear plug. As the flow passes
Knoth et al. (2005) using steady-state computational through the wall, oxidation of CO or HC happens, and as
fluid dynamics. In addition, their work presents perfor- soon as the wall flow exits the wall, it mixes with the axial
mance comparisons with flow-through DOC substrates flow in the outlet channel. As an example, the axial pro-
in comparing oxidation effectiveness. The authors of this files of CO and HC concentration along the outlet channel
review article took a simplified approach to this model to are shown comparatively for CDPF and DOC in Figure 42.
reduce the computational load. This was accomplished Interestingly, the CO concentration of the gas exiting the
by eliminating the species transfer effects from the inlet CDPF is approximately the same as the respective concen-
channel to the wall. The subsequent results indicate tration for the DOC, implying the same conversion poten-
that flow-through-type structures have better conver- tial for both devises, with the same overall dimensions
sion efficiencies than wall-flow style supports. In a more and catalyst loading. This is because, in both systems, the
recent work of the same group (Votsmeier et al. 2007), phenomenon is kinetically controlled.
results show that the above simplification is valid only To make a thorough comparison between the flow-
at high flow velocities. They present a less complicated through-type DOC and the wall-flow CDPF, Figure 43
50 DOC CDPF
Back diffusion contribution (%)
20
Table 3 Exhaust gas concentrations used for the simulations
15 (Dardiotis et al. 2006).
0 2 4 6 8 10
Tortuosity (-)
O2 concentration 10%
CO concentration 100 ppm
Figure 41 Contribution of diffusion phenomena as a function of tor-
HC concentration 90 ppm
tuosity at one temperature (Haralampous and Koltsakis 2004a,b).
presents CO and HC conversion maps obtained by the two mass-transfer and chemical reactions during transient
models for a range of different practical flow rate/tem- operation. This is particularly important during a cold-
perature combinations. There are no differences between start event, which typically governs the emissions per-
the two systems with respect to CO efficiency because it formance of a vehicle. For two systems of equal thermal
is not limited by mass transfer for the tested conditions. mass, the CDPF warms up faster because of the efficient
Concerning the HC, the CDPF is slightly superior com- heat transfer mechanism of the flow through the wall.
pared with the DOC at high flow rates. The mass-transfer This is illustrated in Figure 44, which shows the calcu-
limitation is more important in the case of hydrocarbons lated temperature profiles for both systems at different
because they have the lower diffusivity among the species times during the warm-up phase. By defining the “active
of interest. At high flow rate and high-temperature condi- length” as the fraction of the reactor length above the
tions, the HC conversion in the DOC becomes mass-trans- light-off temperature, the study shows that the DOC is
fer limited as opposed to the CDPF. more efficient despite the shorter “active length”. This
To continue this comparison of catalytic options, it is explained by the higher specific flow rate in the inlet
is important to examine how heat transfer interacts with CDPF channel compared with the respective flow rate in
A 120 B 120
DOC DOC
CDPF inlet CDPF inlet
100 CDPF inlet CDPF outlet 100 CDPF outlet
CO concentration (ppm)
80 DOC 80
60 60
DOC
CDPF outlet
40 40
20 20 CDPF
outlet
0 0
0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160
Channel length (mm) Channel length (mm)
Figure 42 Computed CO concentration profiles in the channels of the clean catalyzed DPF and the oxidation catalyst: (A) 150°C and (B)
225°C (Dardiotis et al. 2006).
150 170
140 160
130 150
120 140
110 130
100 120
-- -- -- -- DOC -- -- -- -- DOC
90 110
----------- Clean CDPF ----------- Clean CDPF
80 100
0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
Exhaust flow rate (kg s-1) Exhaust flow rate (kg s-1)
Figure 43 CO and HC conversion efficiency maps as functions of mass flow rate and temperature.
Solid line, clean catalyzed DPF; dashed line: flow-through oxidation catalyst (Dardiotis et al. 2006).
A 250 B 250
t=5 s
t=5 s
DOC t=15 s CDPF t=15 s
t=25 s
t=25 s t=35 s
200 t=35 s 200 t=45 s
t=45 s t=55 s
t=55 s
Temperature (°C)
Temperature (°C)
Time
150 150
Time
100 100
50 50
0 0
0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160
Length (mm) Length (mm)
Figure 44 Wall temperature profiles at different times during the warm-up phase along (A) DOC and (B) CDPF (Dardiotis et al. 2006).
the DOC channels in the frontal area of the reactor. As Summarizing, the conversion efficiency of the wall-
a result, the residence time of the gas flow through the flow catalyzed filter is a complex function of thermal,
warmer inlet part of both reactors is longer in the case of mass transport, and chemical phenomena occurring
the DOC. The longer residence time compensates for the simultaneously. Under steady-state conditions, the CDPF
shorter active length, leading finally to improved conver- has a higher oxidation conversion efficiency compared
sion of the DOC. with a respective DOC with same dimensions and catalytic
The temporal evolution of CO conversion efficiency loading under mass-transfer limited conditions (high tem-
is presented in Figure 45. As explained above, the DOC perature, high flow rate). Under kinetically limited condi-
exhibits superior performance during the initial warm-up tions, both systems perform identically. However, for tran-
period. However, after sufficient warming (45 s from sient operation from cold-start, it appears that the DOC
start), the CDPF equals or even exceeds the DOC perfor- would be preferred. In all cases, the results are relatively
mance. This is due to the lower mass-transfer resistance similar illustrating that the main differentiation between
at fully warmed conditions, as explained through the pre- the two options will likely occur due to other parameters,
sented steady-state results of Figure 43. The advantage of including variations in cell structure, thermal mass, and
the CDPF is shown more clearly in the case of hydrocarbon quality and amount of wash-coat and precious metals.
conversion efficiency, included in the same graph.
terms in their reaction rates assuming a linear depend- approaching the soot surface is a function of the axial posi-
ency on the local platinum group metals content. They tion due to the advective mass-transfer occurring in the
model the zoned-coated DPF (zCDPF) as an oxidation channel gas. The reaction of soot with NO2 is sufficiently
catalyst to make direct comparisons under transient con- fast at this temperature to consume NO2 in the soot phase.
ditions simulating cold-start engine operation and a cool- On the other hand, the catalytic coating promotes NO2 for-
down phase, which are most relevant for automotive emis- mation and storage on the LNT component. Concurrently,
sions control. the concentration gradient drives NO2 back-diffusion to
Their study compares the transient performance char- the soot layer. It is interesting to note that analogous phe-
acteristics of a zCDPF with different zoning schemes. They nomena also occur at the end of the lean operation mode,
find that a zoning scheme with more catalyst in the frontal as shown in the bottom graph of the same figure. The dif-
part is more beneficial for CO conversion in a simulated ferences arise from the fact that the LNT component has
cold-start transient experiment. In the same test case, been partially saturated, thus influencing the rate of NOx
mass-transfer limitations in the front part of the zCDPF storage, which in turn affects the overall NO2 profiles in
influence hydrocarbon conversion, which results in a the catalyzed wall and soot layer.
lower conversion of the zoned systems under full-warmed Another concept under evaluation for coupled DeNOx
conditions. In a simulated cool-down transient test, the and filter capabilities is a wall-flow selective catalytic
system performance over the same zoning scheme was reductant (WSCR; also referred to as SDPF or SCRF) system
consistently inferior for both CO and HC. An analysis of (Koltsakis et al. 2008, Lee et al. 2008, Ballinger et al. 2009,
the reactor performance through examination of the tran- Cavataio et al. 2009a,b), that is, a filter with integrated
sient temperature and species profiles in the inlet and SCR functionality. To achieve equivalent NOx conversion
outlet channels illustrate that the conversion efficiency is with the SCR catalyst, the authors use a similar volume
a complex function of combined thermal, species trans- to a traditional SCR catalyst. In the simulation effort,
port, and reaction phenomena. the model assumes uniform deposition of the catalytic
Further tests involving different zoning schemes coating in the WSCR wall. In real-world applications, the
include transient simulations with input data from a leg- NOx conversion efficiency may be different if the manu-
islated driving cycle. The conclusions are in line with the facturer deposits the wash coat on the outlet channels or
simplified transient analysis, showing that a preferential on the inlet channels of the filter. The investigation of this
coating of the frontal part reduces the cold-start emis- effect of wash-coat deposition pattern is within the capa-
sions. However, the cool-down modes occurring later in bilities of the mathematical model, but the authors did not
the driving cycle may negate this advantage, which was include it in the study. Using the same reaction kinetics
especially the case for hydrocarbons. for an SCR catalyst model, Figure 48 presents the DeNOx
performance of three configurations: DOC-SCR-DPF, DPF-
SCR, and DOC-WSCR. The combined system presents
3.6 Incorporation of DeNOx catalytic similar warm-up behavior, with respect to NOx conversion
functionality
300
Solving for both the chemical species equations the chan- Lean Rich Lean
nels and wall in the model simultaneously allows for the 250
calculation of both DPF and lean NOx trap (LNT)-related
Concentration (ppm)
at 350°C inlet temperature. The model allows for the cal- 100
culation of the complex interactions of LNT functional-
ity in conjunction with passive soot regeneration via the 50
NO2 pathway. Figure 47 shows the calculated NO2 profiles
through the soot layer and catalyzed wall at two differ- 0
50 70 90 110 130 150
ent axial positions in the filter for two different instants Time (s)
corresponding to the start and the end of the lean (NOx
storage) phase of the experiment shown in Figure 46. As Figure 46 Measured versus simulated NOx emissions during a
shown in the top graph of Figure 47, the NO2 concentration lean/rich cycle at 350°C inlet temperature (Koltsakis et al. 2007).
0.00014 3.0
Soot Engine out
Inlet channel Layer Wall Outlet channel
0.00012 Tailpipe SCR-DPF
Point 1 (z=10 mm) Tailpipe DPF-SCR
2.5
Gas 2.0
0.00008 flow
0.00006 1.5
NO2 Diffusion
0.00004
1.0
0.00002
0.5
0
-0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
Wall depth (mm)
0
0 200 400 600 800 1000 1200
0.00014
Soot Time (s)
Inlet channel Wall Outlet channel
Layer
0.00012
Point 1 (z=10 mm)
Figure 48 Comparison of the cumulative NOx emissions over the
NO2 Concentration (-)
Δm = (soot mass filtered)–(soot mass oxidized) [mg km-1] Figure 49 Passive regeneration performance of the two main
(161) exhaust line formulations (Koltsakis et al. 2008).
A B E
945.0
935.4
925.7
916.1
906.4
896.8
887.1
877.5
C D 867.9
858.2
848.6
838.9
829.3
819.6
810.0
Figure 51 Calculated inlet gas temperature (in K) profiles at DPF entrance for the four inlet pipe geometries: (A) long cone, (B) short cone, (C)
asymmetric cone, and (D) bent pipe (Koltsakis et al. 2009b).
and exhibits an exotherm much later compared with the Figure 52 Schematic of thermocouples positioning in the filter and
other channels. In the specific test, this is mainly due to picture of the filter exit face showing the exact channel locations
the thermal boundary layer at the inlet of the DPF, which (Koltsakis et al. 2005).
A 1000 B 1000
Post-injection Idle Post-injection Post-injection Idle Post-injection
900 900
Symbols-measured 2 Symbols-measured
Lines-DPF model Lines-DPF model
800 800
Temperature (°C)
Temperature (°C)
1
2
700 700
Inlet Inlet
600 600
1
500 500
3 3
400 400
300 300
0 50 100 150 200 0 50 100 150 200
Time (s) Time (s)
C 1000
Post-injection Idle
900
2 Symbols-measured
Lines-DPF model
800
Temperature (°C)
1
700
Inlet
600
500
3
400
300
0 50 100 150 200
Time (s)
Figure 53 Measured (symbols) and computed (lines) filter temperatures along a filter channel: (A) 4 g l-1, (B) 6 g l-1, and (C) 8 g l-1 (Koltsakis
et al. 2005).
A 1000
B 1000
Post-injection Idle Post-injection Post-injection Idle Post-injection
900 900
8 Symbols-measured 25 Symbols-measured
Lines-DPF model Lines-DPF model
800 800
Temperature (°C)
Temperature (°C)
22
700 700
3
600 11
600 28
3
31
500 500
400 400
14
300 300
0 50 100 150 200 0 50 100 150 200
Time (s) Time (s)
Figure 54 Measured (symbols) and computed (lines) filter temperatures at filter exit for 6 g l-1 soot loading: (A) 45° plane (3-8-11-14) and (B)
0° plane (3-22-25-28-31) (Koltsakis et al. 2005).
creates lower DPF inlet temperatures near the filter periph- on the surface temperature of the metal pipe upstream the
ery. Based on these measurements, the temperature dif- DPF and therefore is time dependent.
ference between the centerline and the periphery at filter It is interesting to note that the fastest heat-up rate
inlet can be as high as 90°C during the first few seconds of and the maximum temperature is seen in the channel
the heat-up phase. This thermal boundary layer depends located at the center of the filter segment (#8), whereas
Figure 56 Flow maldistribution effect on soot loading calculated by coupling Star-CD with Axitrap.
Top, flow distribution; bottom, soot distribution at one DPF slice.
Time (s) Outlet temperature (°C) Soot loading (g l-1) Flow distribution (kg s-1 m-2)
300 350 400 450 500 550 600 650 700 750 800 2E-06 6E-06 1E-05 1.4E-05 1.8E-05 2.2E-05
1 2 3 4 5 6 7 8 9 10 11 12
t=48 s
t=100 s
t=140 s
t=180 s
t=500 s
Figure 57 Computed outlet temperature profiles, soot loading fields, and inlet flow profiles during a transient regeneration of a filter with
bent inlet pipe geometry (Koltsakis et al. 2009b).
At the same time, neighboring regions start to heat up and 4.5 Coupling with stress analysis codes
regenerate, which results in a slight homogenization of the
flow. The final soot loading field remains highly nonuni- Due to the failure problems of DPFs, stress analysis was
form and a considerable fraction of the soot fails to regen- very early employed in the study of filters. Similar prob-
erate in the periphery. The regeneration implications of lems concerning flow-through substrates offered a useful
several inlet cone geometries are summarized in Figure 58, basis in this respect (Heck et al. 2009). The first DPF stress
which shows the peak temperatures and final soot fields in analysis calculations involved 2D geometries and temper-
correlation with inlet temperature profile. ate fields estimated directly from measurements and later
Coupling with CFD software is very important for reached 3D geometries with equivalent or actual meshes
certain applications where 3D effects require investiga- and accurate temperature fields taken directly from vali-
tion in detail. However, in this combined example, the dated DPF models. The setup of a stress analysis model
calculation time for a single time step increases signifi- involves, on the one hand, mesh and material properties
cantly. In the case of DPF regeneration simulations, typi- and, on the other hand, structural loads, namely tempera-
cally more than 1000 time steps are necessary. The CFD- ture conditions and mounting pressure.
integrated approach requires many hours of CPU time or Following the developments of DPF models, the first
even days on a single processor. In such cases, it is more stress analysis calculations involved 2D meshes (Umehara
practical to apply the second coupling approach men- and Nakasuji 1993, Miyairi et al. 2001), which evolved to
tioned above. This requires only one steady-state solu- 3D in the course of the last decade (Pontikakis et al. 2002,
tion from the CFD software program. The DPF model is Barataud et al. 2003, Kuki et al. 2004, Hajireza et al. 2010).
then responsible to calibrate the flow resistance parame- Related to the mesh selection are the structural properties
ters describing the inlet cone effect. The result is that the of the material. A common feature in the articles referenced
3D simulation can be performed with very low CPU times is the usage of an equivalent lumped material with aniso-
(approx. 1 h for a 10-min regeneration on a single proces- tropic structural properties as explained by Kim (2004)
sor). The validity of this approach has been verified at instead of the actual honeycomb geometry. This allows the
several flow conditions with nonuniform soot loading, as usage of simplified meshes skipping the channel details.
shown in Figure 59, whereas the accuracy of the results It is also convenient when comparing stresses with failure
is very satisfactory compared with the CFD simulation limits, which are commonly given for equivalent materi-
(Figure 60). Figure 61 presents such a regeneration simu- als due to the measurement techniques applied in prac-
lation example for the case of an oval-shaped DPF sub- tice (Stobbe et al. 1993). Of note, the equivalent failure
jected to a nonsymmetric flow and temperature profile. limits are the product of both channel geometry and solid
Figure 58 Effect of diffuser geometry on inlet gas temperature, temperature profile at the peak of the regeneration, and final soot loading
(Koltsakis et al. 2009b).
80
15 2001) interpolated data from several thermocouple meas-
Dots: CFD
60 Lines: DPF model
12 urements. The main contribution of validated DPF models
10
9 (Pontikakis et al. 2002, Kuki et al. 2004) in the frame of
40
Radial distance (mm)
5 g s-1 50 g s-1 80 g s-1 120 g s-1 8 direct coupling to stress analysis was the replacement of
7
20 these rough approximations with accurately calculated 3D
6
300 CFD
DPF model
200
The regeneration events used to remove soot result in
150 the accumulation of ash within the DPF via two separate
mechanisms. Depending on the type of regeneration, the
100 ash will form a plug at the end of the channel or build up
50
in the pores of the filter walls. Once regeneration occurs,
ash is composed mostly of phosphorous, sulfur, calcium,
0 and zinc (Ishizawa et al. 2009). Ash is an oxidized sub-
5 50 80 120
Mass flow rate (g s-1) stance where the total amount is contingent on the metal
elements in the lubricant additives in the engine oil and
Figure 60 Computed pressure drop at different flow rates for a how much is consumed in the engine for nonadditive
radially nonuniform loaded DPF (Koltsakis et al. 2005). systems. With additive-based systems, the majority of the
Figure 61 Soot loading (top), temperature (middle), and flow (bottom) distribution calculation during an oval DPF regeneration subjected
to asymmetric inlet flow.
ash comes from the FBC with the remainder stemming of the engine (Gaiser and Mucha 2004). Ash produced
from the engine oil components (Konstandopoulos et al. during passive regeneration forms small clusters and
2000, Gaiser and Mucha 2004, Konstandopoulos and stagnates in its area of production along the filter walls.
Kladopoulou 2004, Ishizawa et al. 2009, Liati et al. 2012). This regeneration undergoes a cooler combustion tem-
Ash building up in either system increases the risk of perature than active regeneration, making the ash parti-
uncontrollable combustion in the filter during regenera- cles much smaller but more dense. These small particles
tion (Ishizawa et al. 2009). then adhere to the filter wall creating a buildup across
Two types of ash deposits form through the different the entire filtration area. This creates a significant pres-
regeneration modes of soot as well as the driving cycle sure drop because the generated ash is not fully perme-
able; hence, this wall of ash restricts flow through the DPF
while influencing the working life of the filter (Ishizawa
et al. 2009). Under active regeneration, typically a larger
>1 890 s amount of soot builds before regeneration occurs. Once
0.810871 regeneration commences, ash forms a large cluster that
0.621742 the exhaust gas carries throughout the inlet channel and
0.432613 accumulates to form a plug at the end of this channel.
0.243484 This reduces the effective length of the filtering wall and
910 s
0.0543548 creates an increase in the pressure drop, although these
-0.134774 larger clusters are more permeable that those generated
-0.323903 through passive regeneration (Gaiser and Mucha 2004).
-0.513032 The main difference between the two mechanisms is
930 s
<-0.702161 the ash particle size. When regeneration occurs at lower
temperatures, smaller particles are formed that can fit
inside the pores of the filter and begin to form a wall,
reducing the effective permeability of the DPF wall and
950 s
increasing the pressure drop. Under higher temperatures,
the ash particles are larger and travel down the channel
until the end where they become lodged (Gaiser and
Figure 62 Left column, normal Z stresses (normalized values); right Mucha 2004, Ishizawa et al. 2009). In one study, Ishizawa
column, normal Y stresses (normalized values) (Hajireza et al. 2010). et al. compared the two types of regeneration and took
pictures of the cross-section of the DPF to illustrate the Although a large variety of geometries including tri-
differences. Each type of regeneration clearly influences angular and hexagonal shapes (ECOpoint 2012) has been
filter operation in a different manner. However, equal studied, mostly square-like solutions have been selected
amounts of ash from active regeneration forming a plug for commercialization. The first from Ibiden with the
do not create the same amount of pressure drop as ash acronym OS (OctoSquare) (Ogyu et al. 2004) features a
forming on the filter walls from passive regeneration. Ash large octagonal inlet channel and a smaller square outlet
spread uniformly across the filter walls creates a greater channel. The second from Corning with the acronym ACT
increase in pressure drop than when it forms a plug at the (Young et al. 2004) features a large inlet square channel
end of the channel. With respect to subsequent soot intro- and a smaller outlet square channel. An asymmetric solu-
duction after ash formation, soot primarily forms on top of tion has also been proposed by NGK with the HAC concept
the ash layer. However, it is possible for the soot and ash (Caroca et al. 2009).
to create one common layer. Moreover, under high engine To model asymmetric geometries, the ratio (dc2/dc1) is
loads or substantial acceleration events, it is possible for necessary as input parameter to calculate the inlet and
soot to form under the ash layer. This combination of soot outlet channel widths dc1 and dc2:
loading and ash generation affects DPF operation sig-
nificantly (van Setten et al. 2001, Konstandopoulos et al. 2 ⎛ 1 ⎞
dc 1 = ⎜ -t f ⎟
2005, Johnson 2007). ( )
1 + dc 2 dc 1 ⎝ cpsm ⎠ (162)
Ash accumulation is a time-dependent phenomena
that builds as a function of successive loading and regen- where cpsm is the channels per square meter. Addition-
eration events. At each snapshot in time, both the wall- ally, to define the full geometry, the corner side width dc3
generated and the plug-generated ash can be modeled has to be calculated for the octagonal channel, as shown
through a thickness variable and included in DPF simula- in Figure 63:
tions as accomplished by Gaiser and Mucha (2004) and
Konstandopoulos and Kladopoulou (2004). Moreover, the
area-conserved efforts of Depcik allow modeling of this
dc 3 =
2 ⎢⎣ (
2 ⎡ dc 2
c1
)
1- d dc 1 - 2- 2 t f ⎤⎥ .
( ⎦
) (163)
dc3
A 140 B 300
No ash loading 100 g of ash in the
form of rear plug
120 250
Pressure drop (mbar)
80 150
60 100
40 50
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
Soot mass (g l-1) Soot mass (g l-1)
Figure 64 Effect of soot loading on pressure drop for symmetric and asymmetric channel geometries (300 cpsi, 12 mil, 0.7 channel width
ratio): (A) no soot loading and (B) 100 g ash.
of approximately 30% in back-pressure at the maximum filters and accelerates the soot cake formation. Because
loading. of the membrane, the soot trapped inside the wall can
be decreased significantly. This subsequently improves
the fuel economy of the engine due to a reduced pressure
drop and restriction in the exhaust. In addition, Mizutani
5.3 Inhomogeneous wall structure
et al. (2010) noticed that the correlation of pressure drop
and soot becomes linear with the addition of a membrane,
Since 2008, an inhomogeneous wall structure technol-
which brings advantages in on-board estimations of the
ogy has been presented by several DPF suppliers (Ogyu
current soot load.
et al. 2008, Furuta et al. 2009, Wolff et al. 2010, Iwasaki
These functionalities have been explained theoreti-
et al. 2011). This technology concerns the addition of a
cally using a detailed multilayered wall model (Koltsa-
membrane or layer with small pores usually on top of the
kis et al. 2012). The filtration efficiency of a DPF is most
standard wall layer. For comparative purposes, Figure 65
crucial when it is empty. After forming a sufficient soot
sketches the soot distribution in the wall of a standard
cake, the cake acts as filter providing approximately
DPF compared with a DPF with this membrane. The thin
perfect filtration. Hence, the target is to reach the cake fil-
membrane provides highly effective filtration with empty
tration condition as soon as possible. With the membrane
layer added, the duration can be notably decreased while
at the same time reducing the soot slip and the amount
Flow of soot inside the wall. This embedded soot in the wall is
largely responsible for the high-pressure drop across the
Flow filter. Figure 66 compares the number-based filtration per
Soot
particle size of a conventional filter (single-layer) and a
layered (double-layer) filter at the beginning of a loading
Layer
event. The benefit of the layer is obvious. In particular,
the number of small particles with diameter of 23–70 nm
Filter wall
is reduced considerably. Figures 67 and 68 show the fil-
tration efficiency and resulting soot slip for the beginning
of the simulated loading case, with both benefitting sub-
stantially from the addition of this layer. With a higher
filtration efficiency of the empty filter and faster cake
formation, the soot slip is appreciably reduced. Figure
Figure 65 Schematic soot accumulation at standard DPF (left) and 69 displays the soot accumulation in the filter over the
membrane DPF (right) (Koltsakis et al. 2012). first 6000 s. The soot inside the wall is considerable lower
1×107 0.35
Single-layer wall
Double-layer wall 0.30
Inlet
6
8×10
0.25
Particle number (cc-1)
2×106 0.05
0
0 0 1000 2000 3000 4000 5000 6000
10-8 10-7 10-6 Time (s)
Particle diameter (m)
Figure 68 Soot slip during steady-state loading of empty filters
(Koltsakis et al. 2012).
Figure 66 Particle number-based filtration per particle size of clean
filter (Koltsakis et al. 2012).
with the layer addition. As a result, although the relatively understanding the fundamental phenomena inside the
dense membrane layer increases the initial pressure drop catalyzed filter and their contribution to overall perfor-
of the filter, the reduction in soot stored within the wall mance. The complexity stemming from the plethora of
quickly demonstrates the overriding benefit of adding a phenomena, the multiple scales, and the variety of cata-
membrane layer (Figure 70). lyst configurations and geometries is of special scien-
tific interest. All these challenges have been faced with
increasingly detailed and accurate models supported by
6 Conclusion extensive experimental validations. As a result, today’s
advanced computations tools are considered by many not
Catalyzed DPF modeling has advanced steadily for over only an exhaust system design and optimization aid but
three decades, supporting the market introduction of also an integral part of the development process, offering
efficient and reliable diesel aftertreatment concepts. It insight for further improvements and breakthroughs in
is currently accepted that modeling is indispensable for DPF technology. Despite the maturity of the modeling and
0.35
70
0.30
60
0.25 50
Pressure drop (mbar)
Soot slip (g)
0.20 40
0.10 20
0 0
0 1000 2000 3000 4000 5000 6000 0.0 0.5 1.0 1.5 2.0
Time (s) Soot mass (g l-1)
Figure 67 Filtration efficiency during steady-state loading of empty Figure 69 Soot mass during steady-state loading of empty filters
filters (Koltsakis et al. 2012). (Koltsakis et al. 2012).
References
Ahlström AF, Odenbrand CI. Combustion characteristics of soot Awara AE, Opris CN, Johnson JH. A theoretical and experimental
deposits from diesel engines. Carbon J 1989; 27: 475–483. study of the regeneration process in a ceramic diesel
Allansson R, Blakeman PG, Cooper BJ, Hess H, Silcock PJ, particulate trap using a copper fuel additive. SAE Paper
Walker AP. Optimising the low temperature performance and 970188, 1997.
regeneration efficiency of the continuously regenerating diesel Ballinger T, Cox J, Konduru M, De D, Manning W, Andersen P.
particulate filter (CR-DPF) system. SAE Paper 2002-01-0428, Evaluation of SCR catalyst technology on diesel particulate
2002. filters. SAE Paper 2009-01-0910, 2009.
Aoki H, Asano A, Kurazono K, Kobashi K, Sami H. Numerical Banno Y, Tanak Y, Hihara T, Nagata M. Pre-filter diesel oxidation
simulation model for the regeneration process of a wall-flow catalyst development for DOC-CSF dystem. SAE Paper
monolith diesel particulate filter. SAE Paper 930364, 1993. 2004-01-1430, 2004.
Barataud C, Bardon S, Bouteiller B, Gleize V, Charlet A, Higelin P. Depcik C, Assanis D. One-dimensional automotive catalyst
Diesel filter optimization. SAE Paper 2003-01-0376, 2003. modeling. Prog Energy Combust Sci 2005; 31: 308–369.
Beck JL. Convection in a box of porous material saturated with fluid. Depcik C, Assanis D. Simulating area conservation and the gas-wall
Phys Fluids 1972; 15: 1377–1383. interface for one-dimensional based diesel particulate filter
Benedict RP, Carlucci NA, Swetz SD. Flow loss in abrupt models. ASME J Eng Gas Turbines Power 2008; 130: 1–18.
enlargements and contractions. J Eng Power 1966; 88: 73–81. Depcik C, Srinivasan A. One + one-dimensional modeling of
Bissett EJ. Combustion of particles on regenerable filters with large monolithic catalytic converters. Chem Eng Technol 2011; 34:
conduction. Warren, MI: General Motors Research, 1981. 1949–1965.
Bissett EJ. Mathematical model of the thermal regeneration of a Depcik C, van Leer B, Assanis D. The numerical simulation of
wall-flow monolith diesel particulate filter. Chem Eng Sci 1984; variable-property reacting-gas dynamics: new insights and
39: 1232–1244. validation. Numer Heat Transfer A Appl 2005; 47: 27–56.
Bissett EJ. Thermal regeneration of particle filters with large ECOpoint. ECOpoint Consultants. http://www.dieselnet.com/
conduction. Math Model 1985; 6: 1–18. ecopoint, 2012.
Bissett EJ, Shadman F. Thermal regeneration of diesel particulate Forchheimer P. Wasserbewegung durch Boden. Z Ver Dtsch Ing
monolithic filters. AlChE J 1985; 31: 753–758. 1901; 45: 1736–1741, 1781–1788.
Bissett EJ, Kostoglou M, Konstandopoulos AG. Frictional and heat Fukushima S, Ohno K, Vlachos N, Konstandopoulos AG. New
transfer characteristics of flow in square porous tubes of wall- approach for pore structure and filtration efficiency
flow monoliths. Chem Eng Sci 2012; 84: 255–265. characterization. SAE Paper 2007-01-1918, 2007.
Boomsma K, Poulikakos D. On the effective thermal conductivity of Furuta Y, Mizutani T, Miyairi Y, Yuki K, Kurachi H. Study on next
a three-dimensionally structured fluid-saturated metal foam. generation diesel particulate filter. SAE Paper 2009-01-0292,
Int J Heat Mass Transfer 2001; 44: 827–836. 2009.
Brinkman HC. A calculation of the viscous force exerted by a flowing Gaiser G, Mucha P. Prediction of pressure drop in diesel particulate
fluid on a dense swarm of particles. Appl Sci Res A 1947a; 1: filters considering ash deposit and partial regenerations. SAE
27–34. Paper 2004-01-0158, 2004.
Brinkman HC. On the permeability of media consisting of closely Gallant T. Single channel DPF experiments to investigate soot cake
packed porous particles. Appl Sci Res A 1947b; 1: 81–86. structures. 11th Diesel Engine Emissions Reduction (DEER)
Calmidi VV, Mahajan RL. Forced convection in high porosity metal Conference, Chicago, 2005.
foams. J Heat Transfer 2000; 122: 557–565. Gantawar AK, Opris CN, Johnson JH. A study of the regeneration
Caroca J, Villata D, Fino D, Russo N. Comparison of different diesel characteristics of a silicon carbide and cordierite diesel
particulate filters. Top Catal 2009; 52: 2076–2082. particulate filters using a copper fuel additive. SAE Paper
Cavataio G, Girard JW, Lambert C. Cu/zeolite SCR on high porosity 970188, 1997.
filters: laboratory and engine performance evaluations. SAE Garner CP, Dent JC. A thermal regeneration model for monolithic and
Paper 2009-01-0897, 2009a. fibrous diesel particulate traps. SAE Paper 880007, 1988.
Cavataio G, Warner JR, Girard JW, Ura J, Dobson D, Lambert CK. General Motors. General Motors response to EPA notice of proposed
Laboratory study of soot, propylene, and diesel fuel impact on rulemaking on particulate regulation for light-duty diesel
zeolite-based SCR filter catalysts. SAE Paper 2009-01-0903, vehicles submitted to Environmental Protection Agency, 1979.
2009b. Gropi G, Tronconi E. Continuous vs. discrete models of nonadiabatic
Coltrin ME, Kee RJ, Rupley FM. Surface Chemkin: a Fortran monolith catalysts. AIChE J 1996; 42: 2382–2387.
package for analyzing heterogeneous chemical kinetics at a Hajireza S, Johannesen LT, Wolff T, Koltsakis GC, Samaras ZC,
solid-surface-gas-phase interface. Albuquerque, NM: Sandia Haralampous OA. A modeling and experimental investigation
National Laboratories, 1990. on an innovative substrate for DPF applications. SAE Paper
Cooper BJ, Thoss JE. Role of NO in diesel particulate emission 2010-01-0891, 2010.
control. SAE Paper 890404, 1989. Haralampous OA, Koltsakis GC. Intra-layer temperature gradients
Cooper BJ, Jung HJ, Thoss JE. Treatment of diesel exhaust gases. during regeneration of diesel particulate filters. Chem Eng Sci
Taylor, MI, 1990. 2002; 57: 2345–2355.
Cussler EL. Diffusion: mass transfer in fluid systems. Cambridge, Haralampous OA, Koltsakis GC. Back-diffusion modeling of NO2 in
UK: Cambridge University Press, 1995. catalyzed particulate filters. Ind Eng Chem Res 2004a; 43: 875.
d’Hueppe A, Chandesris M, Jamet D, Goyeau B. Boundary conditions Haralampous OA, Koltsakis GC. Oxygen diffusion modeling in diesel
at a fluid-porous interface for a convective heat transfer particulate filter regeneration. AIChE J 2004b; 50: 2008–2019.
problem: analysis of the jump relations. Int J Heat Mass Haralampous OA, Koltsakis GC, Samaras ZC. Partial regenerations
Transfer 2011; 54: 3683–3693. in diesel particulate filters. SAE Paper 2003-01-1881, 2003.
Darcy HPG. Les Fontaines Publiques de la Ville de Dijon. Paris, 1856. Haralampous OA, Dardiotis CK, Koltsakis GC, Samaras ZC. Study
Dardiotis C, Haralampous O, Koltsakis GC. Catalytic oxidation of catalytic regeneration mechanisms in diesel particulate
performance of wall-flow vs flow-through monoliths for diesel filters using coupled reaction-diffusion modeling. SAE Paper
emissions control. Ind Eng Chem Res 2006; 45: 3520–3530. 2004-01-1941, 2004a.
Dardiotis CK, Haralampous OA, Koltsakis GC. Catalytic oxidation in Haralampous OA, Kandylas IP, Koltsakis GC, Samaras ZC. Diesel
wall-flow reactors with zoned coating. Chem Eng Sci 2008; 63: particulate filter pressure drop. Part I: modeling and
1142–1153. experimental validation. Int J Engine Res 2004b; 5: 149–162.
Depcik C. Simulating the concentration equations and the gas-wall Haralampous OA, Kandylas IP, Koltsakis GC, Samaras ZC.
interface for one-dimensional based diesel particulate filter Experimental evaluation of apparent soot oxidation rates in
models. ASME J Eng Gas Turbines Power 2010; 132: 1–12. diesel particulate filters. Int J Vehicle Des 2004c; 35: 365–382.
Haralampous OA, Koltsakis GC, Samaras ZC, Vogt C-D, Ohara E, Kim JS. Substrate failure prediction during converter canning
Watanabe Y, Mizutani T. Modeling and experimental study process. SAE Paper 2004-01-1134, 2004.
of uncontrolled regenerations in SiC filters with fuel borne Knoth JF, Drochner A, Vogel J, Gieshoff J, Koegel M, Pfeifer M,
catalyst. SAE Paper 2004-01-0697, 2004d. Votsmeier M. Transport and reaction in catalytic wall-flow
Haralampous OA, Koltsakis GC, Samaras ZC, Vogt C-D, Ohara E, filters. Catal Today 2005; 105: 598.
Watanabe Y, Mizutani T. Reaction and diffusion phenomena in Kolaczkowski ST. Measurement of effective diffusivity in catalytic
catalyzed diesel particulate filters. SAE Paper 2004-01-0696, coated monoliths. Catal Today 2003; 83: 85–95.
2004e. Koltsakis GC, Stamatelos AM. Modeling catalytic regeneration of
Harvey GD, Baumgard KJ, Johnson JH, Gratz LD, Bagley ST, Leddy DG. wall-flow particulate filters. Ind Eng Chem Res 1996a; 35: 2–13.
Effects of a ceramic particle trap and copper fuel additive on Koltsakis GC, Stamatelos AM. Modeling thermal regeneration
heavy-duty diesel emissions. SAE Paper 942068, 1994. of wall-flow diesel particulate traps. AIChE J 1996b; 42:
Hawker P, Myers N, Hüthwohl G, Vogel HT, Bates B, Magnusson 1662–1672.
L, Bronnenberg P. Experience with a new particulate trap Koltsakis GC, Stamatelos AM. Catalytic automotive exhaust
technology in Europe. SAE Paper 970182, 1997. aftertreatment. Prog Energy Combust Sci 1997a; 23: 1–39.
Hawker P, Hüthwohl G, Henn J, Koch W, Lüders H, Lüers B, Stommel Koltsakis GC, Stamatelos AM. Modes of catalytic regeneration in diesel
P. Effect of a continuously regenerating diesel particulate filter particulate filters. Ind Eng Chem Res 1997b; 36: 4155–4165.
on non-regulated emissions and particle size distribution. SAE Koltsakis GC, Haralampous OA, Margaritis NK, Samaras ZC, Vogt
Paper 980189, 1998. C-D, Ohara E, Watanabe Y, Mizutani T. 3-Dimensional modeling
Hayes RE, Kolaczkowski ST, Li PK, Awdry CS. Evaluating the effective of the regeneration in SiC particulate filters. SAE Trans J Fuels
diffusivity of methane in the washcoat of a honeycomb Lubricants (SAE Paper 2005-01-0953) 2005.
monolith. Appl Catal B Environ 2000; 25: 93–104. Koltsakis GC, Konstantinou A, Haralampous OA, Samaras Z.
Heck RH, Farrauto RJ, Gulati ST. Catalytic air pollution control: Measurement and intra-layer modeling of soot density and
commercial technology. New York, NY: John Wiley and Sons, permeability in wall-flow filters. SAE Trans J Fuels Lubricants
2009. (SAE Paper 2006-01-0261) 2006.
Heywood JB. Internal combustion engine fundamentals. New York: Koltsakis GC, Haralampous O, Samaras Z, Kraemer L, Heimlich F,
McGraw-Hill, 1988. Behnk K. Control strategies for peak temperature limitation in
Hindmarsh AC. Serial Fortran solvers for ODE initial value problems. DPF regeneration supported by validated modeling. SAE Paper
http://www.llnl.gov/CASC/odepack, 2002. 2007-01-1127, 2007.
Hinterberger C, Olesen M, Kaiser R. D simulation of soot loading Koltsakis GC, Haralampous O, Koutoufaris IZ. Potential and
and regeneration of diesel particulate filter systems. SAE Paper challenges of integrating de-NOx catalysts on DPF substrates.
2007-01-1143, 2007. SIA Paper R-2008-05-22, 2008.
Ho FG, Strieder W. Numerical evaluation of the porous medium Koltsakis GC, Dardiotis CK, Samaras Z, Maunula T, Kinnunen T,
effective diffusivity between the Knudsen and continuum Lundorf P. Optimization methodology for DPF substrate-
limits. J Chem Phys 1980; 73: 6296–6300. catalyst combinations. SAE Paper 2009-01-0291, 2009a.
Howitt JS, Montierth MR. Cellular ceramic diesel particulate filter. Koltsakis GC, Samaras ZC, Echtle H, Chatterjee D, Markou P,
SAE Paper 810114, 1981. Haralampous OA. Flow maldistribution effects on DPF
Ingham DB, Bejan A, Mamut E, Pop I. Emerging technologies and performance. SAE Paper 2009-01-1280, 2009b.
techniques in porous media. Dordrecht: Kluwer Academic Koltsakis GC, Bollerhoff T, Samaras Z, Markomanolakis I. Modeling
Publishers, 2004. the interactions of soot and SCR reactions in advanced DPF
Ishizawa T, Yamane H, Satoh H, Sekiguchi K, Arai M, Yoshimoto N, technologies with nonhomogeneous wall structure. SAE Paper
Inoue T. Investigation into ash loading and its relationship to 2012-01-1298, 2012.
DPF regeneration method. SAE Paper 2009-01-2882, 2009. Konstandopoulos AG. Flow resistance descriptors for diesel
Iwasaki S, Mizutani T, Miyairi Y, Yuuki K, Makino M. New design particulate filters: definitions, measurements and testing. SAE
concept for diesel particulate filter. SAE Int J Engines 2011; 4: Paper 2003-01-0846, 2003.
527–536. Konstandopoulos AG, Johnson JH. Wall-flow diesel particulate
Jacquot F, Logie V, Brilhac JF, Gilot P. Kinetics of the oxidation of filters – their pressure drop and collection efficiency. SAE
carbon black by NO2. Influence in the presence of water and Paper 890405, 1989.
oxygen. Carbon 2002; 40: 335–343. Konstandopoulos AG, Kladopoulou E. The optimum cell density of
Johnson TV. Diesel emission control in review. SAE Paper wall-flow monolithic filters: effects of filter permeability, soot
2007-01-0233, 2007. cake structure and ash loading. SAE Paper 2004-01-1133, 2004.
Johnson TV. Diesel emission control in review. SAE Paper Konstandopoulos AG, Kostoglou M. Periodically reversed
2009-01-0121, 2009. flow regeneration of diesel particulate traps. SAE Paper
Jørgensen MW, Sorenson SC. A 2-dimensional simulation model for 1999-01-0469, 1999.
a diesel particulate filter. SAE Paper 970471, 1997. Konstandopoulos AG, Kostoglou M. Reciprocating flow regeneration
Kandylas IP, Koltsakis GC. Simulation of continuously regenerating of soot filters. Combust Flame 2000; 121: 488–500.
diesel particulate filters in transient driving cycles. Proc Inst Konstandopoulos AG, Kostoglou M. Microstructural aspects of soot
Mech Eng D J Automobile Eng 2002; 216: 591–606. oxidation in diesel particulate filters. SAE Paper 2004-01-0693,
Kandylas IP, Haralampous OA, Koltsakis GC. Diesel soot oxidation 2004.
with NO2: engine experiments and simulations. Ind Eng Chem Konstandopoulos AG, Kostoglou M. Computationally fast
Res 2002; 41: 5372–5384. implementations of convection, diffusion, and chemical
reaction phenomena in diesel particulate filters. SAE Paper Masoudi M, Konstandopoulos AG, Nikitidis MS, Skaperdas E,
2010-01-0890, 2010. Zarvalis D, Kladopoulou E, Altiparmakis C. Validation of
Konstandopoulos AG, Kostoglou M, Skaperdas E, Papaioannou a model and development of a simulator for predicting
E, Zarvalis D, Kladopoulou E. Fundamental studies of diesel the pressure drop of diesel particulate filters. SAE Paper
particulate filters: transient loading, regeneration and aging. 2001-01-0911, 2001.
SAE Paper 2000-01-1016, 2000. Miller RK, Li CG. Effect of catalyst support structure on conversion
Konstandopoulos AG, Kostoglou M, Housiada P. Spatial efficiency. SAE Paper 2001-01-0183, 2000.
non-uniformities in diesel particulate trap regeneration. SAE Miyairi Y, Miwa S, Abe F, Xu Z, Nakasuji Y. Numerical study on force
Paper 2001-01-0908, 2001. regeneration of wall-flow diesel particulate filters. SAE Paper
Konstandopoulos AG, Skaperdas E, Masoudi M. Microstructural 2001-01-0912, 2001.
properties of soot deposits in diesel particulate traps. SAE Mizutani T, Watanabe Y, Yuuki K, Hashimoto S, Hamanaka T,
Paper 2002-01-1015, 2002. Kawashima J. Soot regeneration model for SiC-DPF system
Konstandopoulos AG, Kostoglou M, Housiada P, Vlachos N, design. SAE Paper 2004-01-0159, 2004.
Zarvalis D. Multichannel simulation of soot oxidation in diesel Mizutani T, Iwasaki S, Miyairi Y, Yuuki K, Makino M, Kurachi H.
particulate filters. SAE Paper 2003-01-0839, 2003a. Performance verification of next generation diesel particulate
Konstandopoulos AG, Vlachos N, Housiada P, Kostoglou M. filter. SAE Paper 2010-01-0531, 2010.
Simulation of triangular-cell-shaped, fibrous wall-flow filters. Mogaka ZN, Wong VW, Shahed SM. Performance and regeneration
SAE Paper 2003-01-0844, 2003b. characteristics of a cellular ceramic diesel particulate trap. SAE
Konstandopoulos AG, Kostoglou M, Vlachos N, Kladopoulou E. Paper 820272, 1982.
Progress in diesel particulate filter simulation. SAE Paper Mohammed H, Triana AP, Yang S-L, Johnson JH. An advanced
2005-01-0946, 2005. 1D 2-layer catalyzed diesel particulate filter model to
Konstandopoulos A, Kostoglou M, Vlachos N, Kladopoulou E. simulate: filtration by the wall and particulate cake,
Advances in the science and technology of diesel particulate oxidation in the wall and particulate cake by NO2 and O2, and
filter simulation. Adv Chem Eng 2008; 33 213–294. regeneration by heat addition. SAE Paper 2006-01-0467,
Kostoglou M, Housiada P, Konstandopoulos AG. Multi-channel 2006.
simulation of regeneration in honeycomb monolithic diesel Nield DA, Bejan A. Convection in porous media, 3rd ed. New York,
particulate filters. Chem Eng Sci 2003; 58: 3273–3283. NY: Springer-Verlag, 2006.
Kuki T, Miyairi Y, Kasai Y, Miyazaki M, Miwa S. Study on reliability Ochoa-Tapia JA, Whitaker S. Heat transfer at the boundary between
of wall-flow type particulate filter. SAE Paper 2004-01-0959, a porous medium and a homogeneous fluid. Int J Heat Mass
2004. Transfer 1997; 40: 2691–2707.
Lee W-T. Local preconditioning of the Euler equations. Ann Arbor, Ogyu K, Ohno K, Hong S, Komori T. Ash storage capacity
MI: Doctor of Philosophy, University of Michigan, 1992. enhancement of diesel particulate filter. SAE Paper
Lee KO, Yang SY. Parametric examination of filtration processes in 2004-01-0949, 2004.
diesel particulate filter membranes with channel structure Ogyu K, Oya T, Ohno K, Konstandopoulos AG. Improving of the
modification. SAE Paper 2010-01-0537, 2010. filtration and regeneration performance by the SiC-DPF
Lee JH, Paratore MJ, Brown DB. Evaluation of Cu-based SCR/DPF with the layer coating of PM oxidation catalyst. SAE Paper
technology for diesel exhaust emission control. SAE Paper 2008-01-0621, 2008.
2008-01-0072, 2008. Oh SH, Cavendish JC. Transients of monolithic catalytic converters:
Li CG, Koelman H, Ramanathan R, Baretzky U, Forbriger G, Meunier response to step changes in feedstream temperature as
T. Particulate filter design for high performance diesel engine related to controlling automobile emissions. Ind Eng Chem Res
application. SAE Paper 2008-01-1747, 2008. 1982; 2: 29–37.
Liati A, Eggenschwiler PD, Gubler EM, Schreiber D, Aguirre M. Oh SH, MacDonald JS, Vaneman GL, Hegedus LL. Mathematical
Investigation of diesel ash particulate matter: a scanning modeling of fibrous filters for diesel particulates – theory and
electron microscope and transmission electron microscope experiment. SAE Paper 810113, 1981.
study. Atmos Environ 2012; 49: 391–402. Onorati A, Ferrari G, D’Errico G. Fluid dynamic modeling of the
Liu J, Schorn N, Schernus C, Peng L. Comparison studies on the gas flow with chemical specie transport through the exhaust
method of characteristics and finite difference methods for manifold of a four cylinder SI engine. SAE Paper 1999-01-0557,
one-dimensional gas flow through IC engine manifold. SAE 1999.
Paper 960078, 1996. Opris CN, Johnson JH. A 2-D computational model describing
Lu W, Zhao CY, Tassou SA. Thermal analysis on metal-foam filled the flow and filtration characteristics of a ceramic diesel
heat exchangers. Part I: metal-foam filled pipes. Int J Heat particulate trap. SAE Paper 980545, 1998a.
Mass Transfer 2006; 49: 2751–2761. Opris CN, Johnson JH. A 2-D computational model describing
MacDonald JS, Vaneman GL. Experimental evaluation of fibrous the heat transfer, reaction kinetics and regeneration
filters for trapping diesel-exhaust particulates. SAE Paper characteristics of a ceramic diesel particulate trap. SAE Paper
810956, 1981. 980546, 1998b.
Maly M, Claussen M, Carlowitz O, Kroner P, Ranalli M, Schmidt Pattas KN, Samaras ZC. Computational simulation of the ceramic
S. Influence of nitrogen dioxide based regeneration on soot trap transient operation. SAE Paper 890403, 1989.
distribution. SAE Paper 2004-01-0823, 2004. Pattas KN, Stamatelos AM, Koltsakis GC, Kandylas IP, Mustel W.
Mason EA, Malinaukus AP. Gas transport in a porous media: the Computer aided engineering in the design of catalytically
dusty gas model. Amsterdam: Elsevier, 1983. assisted trap systems. SAE Paper 970472, 1997.
Pauli E, Lepperhoff G, Pischinger F. The description of the Stobbe P, Petersen HG, Høj JW, Sorenson SC. SiC as a substrate for
regeneration behavior of diesel particulate traps with the aid diesel particulate filters. SAE Paper 932495, 1993.
of a mathematical model. SAE Paper 830180, 1983. Stratakis GA. Experimental investigation of catalytic soot oxidation
Peters BJ. Numerical simulation of a diesel particulate filter during and pressure drop characteristics in wall-flow diesel particulate
loading and regeneration. 2003 Spring Technical Conference filters. Doctor of Philosophy, University of Thessaly, 2004.
of the ASME Internal Combustion Engine Division, Salzburg, Suresh A, Johnson JH. A Study of dilution effects on particle size
Austria, 2003. measurement from a heavy-duty diesel engine with EGR. SAE
Peters BJ, Wanker RJ, Münzer A, Wurzenberger JC. Integrated 1D to Paper 2001-01-0220, 2001.
3D simulation workflow of exhaust aftertreatment devices. SAE Tan JC, Opris CN, Baumgard KJ, Johnson JH. A study of the
Paper 2004-01-1132, 2004. regeneration process in diesel particulate traps using a copper
Pontikakis G, Stamatelos A, Bakasis K, Aravas N. 3-D catalytic fuel additive. SAE Paper 960136, 1996.
regeneration and stress modeling of diesel particulate filters Tessier LP, Sullivan HF, Bragg GM, Hermance CE. The development
by ABAQUS FEM software. SAE Paper 2002-01-1017, 2002. of a high efficiency diesel exhaust particulate filter. SAE Paper
Punke A, Grubert G, Li Y, Dettling J, Neubauer T. Catalyzed soot 800338, 1980.
filters in close coupled position for passengers vehicles. SAE Tighe CJ, Twigg MV, Hayhurst AN, Dennis JS. The kinetics of
Paper 2006-01-1091, 2006. oxidation of diesel soots by NO2. Combust Flame 2011; 159:
Rajadurai S, Tagomori MK, Berryhill J, Baig A, Snider BJ. Single seam 77–90.
stuffed converter design for thinwall substrates. SAE Paper Umehara K, Nakasuji Y. Lifetime prediction of wall-flow type diesel
1999-01-3628, 1999. particulate filters using fatigue characteristics. SAE Paper
Reaction Design. Transport core utility manual. San Diego, CA, 930128, 1993.
2003. van Setten BAAL, Makkee M, Moulijn JA. Science and technology of
Reif FP. Statistical and thermal physics. New York: McGraw-Hill, 1975. catalytic diesel particulate filters. Catal Rev 2001; 43: 489–564.
Roe PL. Characteristic-based schemes for the Euler equations. Ann Versaevel P, Colas H, Rigaudeau C, Noirot R, Koltsakis GC,
Rev Fluid Mech 1986; 18: 337–365. Stamatelos AM. Some empirical observations on diesel
Romero AF, Castrejón-Rodríguez J, Serrano-Romero R. Self particulate filter modeling and comparison between
regenerating catalyzed diesel aftertreatment system. SAE simulations and experiments. SAE Paper 2000-01-0477, 2000.
Paper 950367, 1995. Votsmeier M, Gieshoff J, Kogel M, Pfeifer M, Knoth JF, Drochner
Romero AF, Gutiérrez R, García-Moreno R. Combustion processes A, Vogel H. Wall-flow filters with wall-integrated oxidation
of particulate matter and soot in ceramic filter traps for diesel catalyst: a simulation study. Appl Catal B Environ 2007; 70:
engines and numerical modeling. SAE Paper 970475, 1997. 233–240.
Sahraoui M, Kaviany M. Slip and no-slip temperature boundary Wakao N, Kaguei S, Funazkri T. Effect of fluid dispersion coefficients
conditions at the interface of porous plain media: convection. on particle-to-fluid heat transfer coefficients in packed beds.
Int J Heat Mass Transfer 1994; 37: 1029–1044. Chem Eng Sci 1979; 34: 325–336.
Salvat O, Marez P, Belot G. Passenger car serial application of a Wark K, Warner CF. Air pollution: its origin and control, 2nd ed. New
particulate filter system on a common rail direct injection York: Harper & Row, Publishers, 1981.
diesel engine. SAE Paper 2000-01-0473, 2000. Wolff T, Friedrich H, Johannesen L, Hajireza S. A new approach to
Shadman F, Bissett EJ. Analysis of thermal regeneration of fibrous design high porosity silicon carbide substrates. SAE Paper
diesel-particulate filters. Ind Eng Chem Prod Res Dev 1983; 22: 2010-01-0539, 2010.
203–208. Wooding RA. Steady state free thermal convection of liquid in a
Shampine LF, Reichelt MW, Kierzenka JA. Solving index-I DAEs in saturated permeable medium. J Fluid Mech 1957; 2: 273–285.
MATLAB and Simulink. SIAM Rev 1999; 41: 538–552. Young D, Hickman DA, Bhatia G, Gunasekaran N. Ash storage
Smith JM. Chemical engineering kinetics. New York: McGraw-Hill, concept for diesel particulate filters. SAE Paper 2004-01-0948,
1970. 2004.
Sorenson SC, Høj JW, Stobbe P. Flow characteristics of SiC diesel Zhao CY, Kim T, Lu TJ, Hodson HP. Thermal transport in high porosity
particulate filter materials. SAE Paper 940236, 1994. cellular metal foams. J Thermophys Heat Transfer 2004; 18:
Springer KJ, Stahman RC. Removal of exhaust particulate from a 309–317.
Mercedes 300D diesel car. SAE Paper 770716, 1977. Zukauskas AA. Convective heat transfer in cross-flow. In: Kakac
Sprouse III C, Mangus M, Depcik C. Diesel particulate filter model with S, Shah RK, Aung W, editors. Handbook of single-phase
detailed permeability analysis. 2011. convective heat transfer. New York: Wiley, 1987.