DPF Regeneration

This item was submitted to Loughborough’s Institutional Repository (https://dspace.lboro.ac.uk/) by the author and is made available under the

following Creative Commons Licence conditions.

For the full text of this licence, please go to: http://creativecommons.org/licenses/by-nc-nd/2.5/

215

A diesel particulate filter regeneration model with amulti-step chemical reaction schemeM C Law, A Clarke, and C P Garner*Wolfson School of Mechanical and Manufacturing Engineering, Loughborough University, Loughborough, UK

The manuscript was received on 23 April 2004 and was accepted after revision for publication on 17 August 2004.

DOI: 10.1243/095440705X5902

Abstract: Diesel particulate filters (DPFs) are considered necessary in order to meet future globaldiesel engine emissions legislation. Various regeneration methods have been developed to clean DPFsby periodic oxidation of trapped particulate matter (soot). To achieve this goal, it is important tounderstand the fundamentals of the regeneration process. Previous soot oxidation regeneration modelsrelied on tunable chemical kinetic parameters to achieve agreement between model and experimentalresults. In the work reported in this paper, a multistep chemical reaction scheme is incorporated ina model to study the thermal regeneration process. The regeneration model does not require tunableparameters and its results compare well with experimental findings. The effects on regeneration ofvarious gas species are also studied, in addition to O

2and N

2, such as CO and H

2O that are present

in the exhaust gas. The model is also used to demonstrate the effects of quenching the regenerationprocess and its impact on partial filter regeneration.

Keywords: diesel, engine, particulate, filter, trap, regeneration, soot, kinetics, chemical, reaction,oxidation

1 INTRODUCTION this configuration, the exhaust gas flow is forced topass through the porous ceramic walls between thechannels. The pore sizes of the wall are such thatDiesel engine exhaust emission standards arethey are able to filter some of the particulate in thebecoming increasingly strict. Figure 1, for example,wall. As time proceeds, the soot particulates start toshows the heavy-duty diesel emission legislation foraccumulate on the wall surfaces. Layers of soot areparticulate matter (PM) and oxides of nitrogen (NO

x)

formed on the wall surfaces that encourage thein Japan, Europe, and the United States. This legis-trapping of subsequent particulate. This gives rise tolation is driven in part due to concerns of the effectsthe high filtration efficiency.of diesel PM on human health.

If soot is allowed to accumulate too much, thenRecent years have seen significant advances inboth engine power and fuel economy deterioratesdiesel emission control technology to meet the ever-due to excessive exhaust gas backpressures. Hence,stringent legislation. One of the ways to reduce signi-the filter has to be cleaned or ‘regenerated’. Theficantly PM emissions from diesel engines is the useregeneration process, in which the trapped parti-of a diesel particulate filter (DPF) in the exhaustculate matter oxidizes, is still poorly understood andsystem. There are a variety of DPFs available, suchits control a particularly difficult technical challenge.as the wall-flow monolith, wire mesh, and foams.During regeneration, heat is liberated by the soot dueOf these, the wall-flow monolith DPF exhibits theto the exothermic oxidation process, which can leadhighest filtration efficiency (>95 per cent by mass)to high thermal gradients and high temperatures andfor the least pressure drop. The wall-flow monoliththese can result in filter failure by fracture or melting.comprises a honeycomb of square channels, typicallyConversely, the regeneration process can sometimes1–2 mm in square cross-section, with each end ofextinguish prematurely leading to incomplete filtereach channel plugged alternately (see Fig. 2). Inregeneration.

The purpose of the work reported in this paper is* Corresponding author: Wolfson School of Mechanical and Manu- to model the regeneration process in greater detailfacturing Engineering, Loughborough University, Loughborough, than previously, especially with the adoption of a

more detailed multichemical reaction scheme.Leicestershire, LE11 3TU, UK. email: [email protected]

D08304 © IMechE 2005 Proc. IMechE. Vol. 219 Part D: J. Automobile Engineering

216 M C Law, A Clarke, and C P Garner

Fig. 2 Schematic of a monolithic wall-flow dieselparticulate filter (DPF)

to various problems, such as thermal and catalyticregeneration and the prediction of exhaust gas flowand pressure drop across DPF, as well as optimizingthe filter structural design.

Bissett and Shadman [1] and [2] reported the firstmathematically rigorous one-dimensional models,which remain the foundation of current DPF model-ling. Based on this model geometry, Garner and Dent[3] developed a generalized DPF model, which wasapplicable to not only wall-flow but also fibrousmesh DPF. Computational power has increased overthe past ten years and this has allowed researchersto extend some models to multidimensions. Forexample, Opris and Johnson [4] developed a two-dimensional regeneration model and found thatregeneration can proceed from upstream to thedownstream end of the filter channels. They foundthat after completely oxidizing the soot at thedownstream end, the regeneration wave returnedto upstream. This process could lead to partialregeneration of the inlet end of the DPF. By usinga three-dimensional CFD model, Konstandopolouset al. [5] were able to study the temperature distri-bution for the whole DPF. Generally, two- and three-dimensional models could give a more completepicture of the regeneration process that occurs in theDPF. However, due to the uneven exhaust gas flowdistribution in the DPF, discrepancies are alwaysobserved between these models and the experimental

Fig. 1 Heavy duty diesel emission levels in Japan, findings, as discussed by Stratakis and Stamatelos [6].Europe, and the US [32] Moreover, two- and three-dimensional models are

usually quite intensive in computational effort andtime. Thus, in order to obtain reasonable accuracyand predictability without being computationally

2 DPF REGENERATION MODELintensive, one-dimensional regeneration models areof significant value.

2.1 Overview of the existing DPF modelsPrevious reported models lack detailed submodels

of the multiple chemical reactions that take placeDPF modelling has been actively researched for morethan two decades. Previous models span zero-, one-, during the regeneration process. For the thermal (i.e.

non-catalytic) regeneration case, all previous modelstwo-, and three-dimensions and have been applied

D08304 © IMechE 2005Proc. IMechE. Vol. 219 Part D: J. Automobile Engineering

217DPF regeneration model with a multi-step chemical reaction scheme

reported [3], [4], [7], and [8] have assumed one of or regeneration process are then examined. Finally, astudy of the effect of quenching the regeneration pro-a combination of the following basic reactionscess on the exhaust-gas composition is presented.

C+O2�CO22.2 Governing equationsC+1

2O2�CO

The model [3] has been generalized so that it isC+CO2�2CO

applicable to any type of DPF by only changing theappropriate physical properties and dimensions ofC+O2�aCO2+(1−a)COthe interested filter. A full description of the overallmodel is given by Garner and Dent [3]. Here theIt is noted that extended chemical reaction schemessalient details are given for completeness of the newfor catalytic regenerations have been developed bymultichemical reaction scheme. Figure 3 shows theother researchers [9–11].schematic of the model geometry used in the workOne of the limitations of the previous models isreported in this paper.the lack of gas-phase reactions. For example, the

effect of CO on the PM/soot reaction rate has beenignored. In addition, most researchers have ‘tuned’ 2.2.1 Regeneration mechanism(i.e. arbitrarily varied) kinetic parameters in the

Soot oxidation. Exhaust gas is composed of variousglobal reaction expressions to ensure that the model

species that could take part in the soot regenerationresults agree with experimental findings.

process. The rate of soot oxidation by reactive gasThe understanding of reaction kinetics during

species is determined by two main factors, namelythe regeneration process has become increasingly

the rate of gas species transport from bulk phase tocrucial, particularly due to the onset of new regener-

soot surface and the chemical reaction rate betweenation methods, such as catalyst-assisted [12] and

the soot surface and reactive gas species in thenon-thermal plasmas [13] and [14], etc. Thus, the

boundary layer. The former mechanism is termedobjective of the work presented in this paper is to

as diffusion-controlled and the latter kinetically-represent an extended series of chemical reaction

controlled [15]. Thus, the rate of oxidant con-schemes that can occur during the regeneration

sumption at solid phase, R∞∞∞S,j

of species j duringprocess.

quasi-steady-state conditions (i.e. when the rate ofIn the following sections, a review of the over-

consumption oxidants (e.g. O2

and NO2) equals the

all regeneration model will be presented. This istransport rate from bulk gas phase to solid surface)

followed by a more detailed description of the hetero-can be expressed as

geneous and homogeneous chemical reactions thathave been incorporated in the model, which are

R∞∞∞S,j=A KibKi+bBrGmG,j (1)

C+12O2�CO (R1)

where Ki=A

iexp(−E

a/R

uT) is the Arrhenius type

CO+12O2uCO2 (R2) chemical kinetic expression of reaction i. E

a, A, R

u,

and T represent the activation energy, frequencyC+O2�CO2 (R3)factor, universal gas constant, and temperature of

CO+H2OuCO2+H2 (R4) solid or gas, depending on whether the reactionis gas- or solid-phase reaction. If it is gas-phase2NO2+C�2NO+CO2 (R5)reaction, T=T

gand T=T

sfor solid-phase reaction.

NO2+C�NO+CO (R6) b is the mass-transfer coefficient of the reaction, rG

is the exhaust bulk-gas density, and mG, j

is the massNO2+CO�NO+CO2 (R7)fraction of gas species j.

2NO+O2�2NO2 (R8)

Gas-phase reaction. The soot oxidation process pro-C+12O2+NO2(+NO2)�NO+CO2(+NO2) (R9)

duces various gas species. Some of these species areC+12O2(+NO2)�CO(+NO2) (R10) not inert and therefore they react in the gas phase

(or with the solid soot) and heat is liberated (orabsorbed) during the reaction. In the case of gas-Both one-step and multistep oxidation reaction sub-

models are studied and compared with experimental phase reactions, gas specie diffusion effects are con-sidered negligible (i.e. a well-stirred reactor) and thedata. The effects of various gas compositions on the



Fig. 3 Schematic for a single channel wall on which the model is based

reaction rate is solely kinetically controlled. In order can be expressed as followsto keep the model as general as possible, reversiblegas-phase reactions are considered, thus log10 Kformation,j=C1 ln T+

C2T+C3+C4T+C5T2

∑N

j=1n∞jiYju ∑

N

j=1n◊jiYj (2) (7)

where C1, C

2, C

3, C

4, and C

5are the polynomialwhere n∞ and n◊ are the stoichiometric coefficients of

constants.reactant and product. Yj

is the chemical formula ofspecies j.

2.2.2 Mass balanceThe rate of gas reaction i, R∞∞∞

G,iis represented as

Soot. The mass of soot is depleted as it is oxidizedand the oxidation rate is proportional to the sumR∞∞∞G,i=kfi a

N

j=1[Yj ]n∞ji−kri a

N

j=1[Yj ]n◊ji (3)

of the solid-gas reaction rates. Thus, it can berepresented bywhere k

fiand k

riare forward and reverse reaction

rates of reaction i. kri

can be evaluated from therelation with equilibrium constant based on molar

qbpqt=−∑

i

R∞∞∞S,jri

(8)concentration, K

c,i, [15] which is

where bp

is the bulk density of soot (particulate) andr is the stoichiometric ratio of the reaction i.kri=

kfiKci

(4)

Exhaust gas stream. The mass balance of exhaustThis can be related to equilibrium constants basedgas species during the regeneration process can beon partial pressures, K

p,i, by

written as

Kc,i=Kp,iA P

RuTBW n◊ij−W n∞ij (5)−QG∞∞GAqmG,jqz B−QrGAqmG,jqt B=∑ R∞∞∞G,j+∑ R∞∞∞S,j

where P is the exhaust-gas pressure.(9)The equilibrium constants based on partial press-

ures for any arbitrary equilibrium reaction i can be where Q is the medium porosity, G∞∞G

is the gas flowratecalculated from per unit area, m

G, jis the concentration of gas con-

stituent j in the bulk gas. It dictates that the net ratelog10 Kp,i=∑jn◊ij log10 Kp,formation,j of production (or destruction) of a gas species j is

the sum of both gas and solid-gas phase reactions.−∑jn∞ij log10 Kp,formation,j (6)

2.2.3 Energy balancewhere K

p,formation, jis the equilibrium constant based on

partial pressure of species formation. By using stan- Gas stream. The overall filter is assumed to be wellinsulated and heat transfer in the gas phase is verydard thermodynamic expressions [16], log

10K

formation, j



small and hence neglected. The energy balance in actual diesel soot reactions in DPF, the model, how-ever, uses the kinetic parameters that have beenthe gas stream in both time and space equals to the

heat convected to the solid phase and also the heat obtained from fundamental data. In other words,the ‘tuning’ of global reaction that has been widelyreleased (or absorbed) during reactions. This can be

expressed as follows practiced in previous models has been eliminated inthis present new model.

To facilitate the kinetic modelling, detailed pro-−QG◊GcPGAqTGqz B−QrGcP

GAqTGqt B cesses such as surface reaction and elementary gas-phase reactions have been ignored. Instead, multi-=h.a(TG−TS)+∑

iR∞∞∞i Hi (10)

step global reactions for both solid-gas- and gas-phase reactions are implemented. The full set of

where h is the convective heat transfer coefficientreactions considered in this model, together with the

and cP

G

is the specific heat capacity of exhaust gas.kinetic data, is given in Table 1

R∞∞∞i

is the reaction rate, which includes R∞∞∞G, j

and R∞∞∞S, j

.Reactions R1 and R3 respectively represent the

partial and complete soot oxidation that producesSoot. Radiative heat transfer effects from the solid CO and CO

2. The kinetic empirical data of R1 for

surfaces can be neglected since the channels are diesel soot has not been reported to date. Therefore,slender and each of the four surfaces of the inlet it is difficult to know which of these two reactionschannel are assumed to exchange equal amounts of are more favoured during the DPF regeneration pro-radiative heat with each other. The convective heat cess. However, in the case of carbon graphite particletransfer to the solid from the gas stream increases combustion in the fluidized bed, it was found thatthe internal energy of the solid and hence its tem- CO is the main product between 1000 K and 1400 Kperature, setting up a thermal gradient in the flow [18]. Hence it is assumed that R1 is the main PMdirection. If the temperature of the solid is high oxidation path.enough, the particulate will oxidize rapidly liberating Another important issue is the disagreement insignificant energy. This is represented as reaction order of PM with respect to oxygen. A 2/3-

reaction order indicates that the oxidation occurs(1−Q)rScP

SAqTSqt B=∑i R∞∞∞i Hi+ha[TG−TS ] on the particle surface and particle shrinks in sizeas oxidation proceeds. One the other hand, a unityreaction order means that the reaction rate is

+kSAq2TSqz2 B (11) proportional to the mass of the particle [19]. Thereaction order of R1 and R3 with respect to O

2is

where kS

is the thermal conductivity and a is the assumed to be unity.specific area of either the soot layer or the ceramic R2 represents the oxidation of CO to CO

2.

filter wall depending on the position of z. Literature has shown that direct CO oxidation suchas CO+O

2�CO

2+O is extremely slow [15]. The CO

oxidation is aided by the presence of OH radicals2.2.4 Equation of statevia reaction CO+OHuCO

2+H. In this study, the

Exhaust gas pressure is considered constant in this source of OH radicals is assumed to be generatedmodel since experiments [17] have shown that the from R4. From Table 1, it can be observed that thepressure change is only a few centimeters mercury. CO oxidation is a complex reaction because of itsThe equation of state of the system is thus defined dependence on the CO, H

2O, and O

2concentrations

as raised to some empirical constants. The reversiblereaction is added in the reaction by introducing arGTG=constant (12)reversible reaction rate.

R4 is the water–gas shift (WGS) reaction. Since2.2.5 Chemical reaction kinetic parameters

water concentration aids the CO oxidation, the WGSreaction is included in the reaction scheme to betterThe reaction kinetic parameters, i.e. activation energy,

E, and frequency factor, A, of diesel soot oxidation estimate the amount of water vapour participatingin the regeneration process. The kinetic data waswith various gas species in the literature are not avail-

able in the literature. Since the majority of diesel soot obtained by Graven and Long [20] at a pressure of1 atmosphere and at 900 °C.consists of carbon, it is assumed, therefore, that

diesel soot is similar to coal or synthetic carbon R5–R10 are NOx-related reactions. Except for R7

and R8, the reactions are direction oxidation of PMblack. Although this does not precisely represent the



Tab

le1

Mu

lti-

chem

ical

reac

tio

ns

use

din

the

mo

del

and

thei

rre

spec

tive

rate

equ

atio

ns

and

kin

etic

dat

a

Rea

ctio

nR

ate

exp

ress

ion

Pre

-exp

on

enti

alco

nst

ant,

AA

ctiv

atio

nen

ergy

,E

Ref

eren

ce

C+

1 2O

2�

CO

R∞∞∞

S,1=

k 1b

k 1+

b[O

2]

A1=

2.64×

107

s−1

E1=

79.7

91kJ

mo

l−1

[28]

CO+

1 2O

2<k

2f

k2

r

CO

2R∞∞∞

G,2=

k 2f[C

O][

H2O

]0.5

[O2]0

.3−

k 2r[C

O2]

A2=

3.8×

106

s−1

E2=

66.8

8kJ

mo

l−1

[29]

C+

O2�

CO

2R∞∞∞

S,3=

k 3b

k 3+

b[O

2]

A3=

3.68×

106

s−1

E3=

142

kJm

ol−

1[2

3]

CO+

H2O<

CO

2+

H2

R∞∞∞

G,4=

k 4f[H

2O

][C

O]0

.5

(1+

379.

473[

H2])

0.5−

k 4f[C

O2][

H2]0

.5

1+

3.6[

CO

]

A4

f=1.

58×

101

1m

3/2

mo

l−1

/2s−

1

A4

r=3×

109

m3

/2m

ol−

1/2

s−1

E4

f=28

1.31

4kJ

mo

l−1

E4

r=23

8.26

kJm

ol−

1[2

0]

2NO

2+

C�

2NO+

CO

2R∞∞∞

S,5=A

k 5b

k 5y1

.13

NO

2

(1+

8y0

.55

H2

O)[

C]+

bBy1.13 N

O2

(1+

8y0

.55

H2

O)[

C]

A5=

462.

3s−

1E

5=

45.5

kJm

ol−

1[2

1]

NO

2+

C�

NO+

CO

R∞∞∞

S,6=A

k 6b

k 6y1

.05

NO

2

(1+

9y0

.74

H2

O)[

C]+

bBy1.05 N

O2

(1+

9y0

.74

H2

O)[

C]

A6=

1002

.22−

1E

6=

59.4

3kJ

mo

l−1

[21]

NO

2+

CO�

NO+

CO

2R∞∞∞

G,7=

k 7f[N

O2][

CO

]−k 7

r[NO

][C

O2]

A7

f=1.

2×

107

m6

mo

l−1

s−1

E7=

132

kJm

ol−

1[3

0]

2NO+

O2�

2NO

2R∞∞∞

G,8=

k 8f[N

O]2

[O2]−

k 8r[N

O2]2

A8

f=1.

2×

10−

3m

6m

ol−

2s−

1E

8=−

4.40

7kJ

mo

l−1

[31]

C+

1 2O

2+

NO

2(+

NO

2)�

NO+

CO

2(+

NO

2)

R∞∞∞

S,9=A

k 9b

k 9y1

.07

NO

2

y0.2

9O

2

(1+

7.5y

0.9

H2

O)[

C]+

bBy1.07 N

O2

y0.2

9O

2

(1+

7.5y

0.9

H2

O)[

C]

A9=

40.3×

103

s−1

E9=

65.4

kJm

ol−

1[2

1]

C+

1 2O

2(+

NO

2)�

CO

(+N

O2)

R∞∞∞

S,1

0=A

k 10b

k 10y N

O2

y0.2

9O

2

(1+

7.5y

0.9

H2

O)[

C]+

bBy N

O2

y0.2

9O

2

(1+

7.5y

0.9

H2

O)[

C]

A1

0=

2.3×

105

s−1

E1

0=

82.6

1kJ

mo

l−1

[21]



with NO2. The NO

2–C kinetic data was obtained by

Jacqout et al. [21] measuring the reaction rate ofcarbon black (Cabot Vulcan 6) in isothermal con-dition (300–450 °C). The influence of kinetic data ofthe reaction in the presence of water and oxygen wasalso measured. The kinetic data used in this study isbased on their published data and with no further Fig. 4 Thermocouple locations in wall-flow DPF usednumerical modification. by Higuchi et al. [17]

3 MODEL VALIDATION The exhaust gas condition at the outlet channelwas compared with the experimental results. From

The model has been compared with two sets of Fig. 5, it can be observed that reasonable agree-carefully controlled experimental results reported by ment was achieved between the multistep chemicalHiguchi et al. [17]. The boundary conditions of both reaction model and the experimental data obtainedexperiments are given in Table 2. by Higuchi et al. [17]. At the current stage, it is

The exhaust gas was assumed to be the com- unclear what factors contribute to the temperaturebustion product of diesel fuel (C

12H

26). The equilib- lag between the experimental findings and the model.

rium product composition of the diesel fuel was The temperature difference can be minimized bygenerated using TPEQUIL [15]. The soot bulk density arbitrary tuning of some input data or the heat trans-was experimentally found to be 56 kg m−3 [3] and fer submodel. Hence, experimental work is currentlythe soot bulk porosity is assumed to be 0.5. The being carried out by the authors to address this andexhaust composition is given in Table 3. will be reported at a later date.

The model results were compared with the wall A number of factors can contribute to the differ-temperature measurement at the centre and the end ence between experimental and model results. First,of the filter (points 2 and 3 respectively), as shown the difference of temperature profiles between thesein Fig. 4. two cases is due to the heat-transfer-mass coefficient

used in the model. The coefficient used is based onthat of Olson et al. [22], who studied the regenerationof adiabatic fixed beds whose carbon particles wereTable 2 DPF properties and exhaust-gas boundary

conditions by Higuchi et al. [17] larger than the mean diesel soot particle size. Second,diesel PM is different from other carbon materials,

Filter type Ceramic wall-flowsuch as carbon black, in term of chemical species

Overall filter dimensions (mm) diameter, 118; composition and microstructure. Diesel PM haslength, 152 chemical components that differ from artificially

Cell density (cells.cm−2 31Wall thickness (mm) 0.3048Wall porosity 0.40–0.50Mean pore diameter (mm) 15–30Inlet gas final temperature (K) 863Exhaust gas flow rate at STP (m3.min−1) 0.6Oxygen concentration (%) 10Soot mass loading (g) 14.4

Table 3 Base-line exhaust-gas species con-centration DPF inlet boundaryconditions used in the multi-chemical scheme simulation

Exhaust gas specie Mole fraction (–)

Oxygen 0.09085Water vapour 0.07803Carbon dioxide 0.07202 Fig. 5 Comparison of multi-step regeneration modelCarbon monoxide 0.0 with the experimental results by Higuchi et al.Nitrogen 0.7591

[17]



manufactured carbon. For example, diesel PM con- 4 INCORPORATING VARIOUS EXHAUST GASSPECIES INTO THE MODELtains organic compounds and traces of precious

metals. Precious metals can have catalytic effectson the regeneration process. For example, the PM The multistep chemical scheme was used to

investigate DPF regeneration behaviour under actualreaction rate with oxygen was found to be about20 kJ mol−1 by Otto et al. [23] compared to that of exhaust-gas conditions. The exhaust-gas composition

and hence DPF input operating conditions wereother material such as Printex-U (a flame soot) withan activation energy of 168 kJ mol−1 . Also, during the obtained experimentally and are summarized in

Table 4. The weighted engine emissions test-cyclepreheating of diesel PM, the absorbed volatile hydro-carbon is oxidized and heat is liberated, which in average values were used in the simulations here.

The enhanced soot oxidation rate by NO2

will be dis-turn increases the temperature in the filter.Ishiguro et al. [24] found that the inner core of cussed in a future paper; here the effects of CO, CO

2,

H2O, and H

2on the regeneration rate are also dis-diesel PM was less thermodynamically stable and

hence could be burnt off preferentially than the outer cussed, since they are of interest to some advanceddiesel engine combustion concepts currently beingshell. Hence, the diffusion of gas can affect the rate

of regeneration process. developed that exhibit higher engine-out levels ofthese species. The unvalidated results here give usefulThe above factors result in the observations made by

Yezerets et al. [25]. In the experiment they conducted, a priori guidance on the relative impact these speciesmight have on the regeneration process.two values of activation energies were obtained for

soot oxidation rate. The lower value of activationenergy was found at the range of 21–39 kJ mol−1 for 4.1 CO and CO

2effects

200–300 °C, whilst 92–97 kJ mol−1 for 400–550 °C.Engine exhaust gas can have varying amounts of CO

The previously reported one-step chemical reactionand CO

2and hence it is valuable to study the effect

model [3] was run for comparison with the newof CO and CO

2on filter regeneration. Such effects

multistep chemical reaction model. In the one-stepare indicated in Figs 7 and 8.

model, the global kinetic data for R3 (Table 1) wereFigure 7 shows the effect of varying CO

2concen-

obtained from under the engine exhaust gas con-tration in the exhaust gas. By increasing CO

2from

ditions. As can be observed from Fig. 6, the one-step5.575 per cent (by volume) to 8.363 per cent, i.e. a

model over-predicts the peak filter temperature, as50 per cent increase, the maximum exhaust gas tem-

well as the time when the peak temperature occurs.perature at channel outlet decreased from 879 K to

Thus, the multistep chemical reaction scheme model862 K. Due to the efficient fuel oxidation in modern

is a more promising solution compared to the pre-diesel engines, the concentration of CO in exhaust

vious one-step model. As will be seen later in thisgas is usually small, of the order of a few hundred

paper, it also provides significantly more detailedppm. Figure 8 shows that there is no change in

understanding of the regeneration process andchannel gas temperature when the CO was increased

increases its generality.from 0 to 100 ppm. A more significant increase in

Table 4 Weighted emissions test-cycle average valuesof diesel engine exhaust composition for heavyduty 4-cylinder, 4-stroke, 4-valve per cylinderDI engine rated at 95 kW

Exhaust gas bulk condition

Exhaust gas inlet temperature (K) 660.5Final steady-state exhaust temperature (K) 773.0Inlet pressure (kPa) 108.5Exhaust mass flow (kg.h−1) 393.7

Exhaust gas species composition (by volume)

Oxygen (%) 12.38Carbon dioxide (%) 5.575Water vapour (%) 4.8Nitrogen EquilibriumCarbon monoxide (ppm) 295.3Fig. 6 Comparison of one-step regeneration modelNitrogen dioxide (ppm) 31.6reported by Garner and Dent [3] with experi- Nitrogen monoxide (ppm) 282.65

mental results by Higuchi et al. [17]



temperature (see Fig. 9) but had negligible effect onthe overall rate of soot oxidation time (see Fig. 10).The former result can be explained as follows: watermolecules are dissociated to produce hydroxyl ions,which are reactive and can promote the oxidationof CO to CO

2. Second, as the water concentration

decreases, (water–gas) equilibrium shifts to the left,which favours the production of water and CO

2,

which is also an endothermic reaction. Nevertheless,examining the reaction rate of the backward reactionof the water–gas shift, it can be noted that the rateis small (due to low concentration of hydrogen in theexhaust gas) thereby the contribution of this factoris small. Despite higher water vapour concentrationFig. 7 The effect of carbon dioxide on the filterresulting in higher heat liberation, the soot oxidationtemperaturerate is not significantly faster. Hence, it can be con-cluded that the presence of water in exhaust gasstream aids the regeneration process and helps toconvert poisonous CO to CO

2. This is favourable,

especially since CO is becoming a more significantemission for diesel engines due to the ever-stringent

Fig. 8 The effect of carbon monoxide on the filtertemperature

temperature is observed by the case of 1000 and5000 ppm CO; such CO values might occur in someregeneration system strategies that require the engine

Fig. 9 Increasing water vapour concentration increasesto periodically run at rich air/fuel ratios.the filter temperatureGenerally, it can be concluded that higher CO

concentration will increase filter temperature. On theother hand, increasing CO

2will decrease the filter tem-

perature. These trends are understandable thermo-dynamically since formation of CO

2is more favoured

than CO and thereby CO is oxidized and heat isliberated. The specific heat capacity of CO

2is higher

and thus filter temperature is reduced as CO2

isincreased.

4.2 H2

O effects

A number of previous investigators [25–27] havereported that the presence of water vapour mightenhance the regeneration rate. To investigate this,the present model was used.

Increasing the water vapour concentration in the Fig. 10 Effect of inlet water vapour concentration onsoot oxidation (regeneration) rateexhaust flow entering the DPF increased the filter



legislation. However, its concentration should bekept to a minimum to lower the filter temperatureand avoid filter damage.

4.3 H2

effects

Since the concentration of hydrogen in exhaust gasis usually negligible, hence it does not play animportant role in the normal regeneration processes,as illustrated in Fig. 11. This result is expected sincethe WGS is a high temperature reaction, as indicatedby its high activation energy. In addition to thetemperature effects, the concentration of H

2is also

important in order for the WGS to be important in Fig. 12 Effect of premature quenching of regenerationthe regeneration process. If, for example, the H

2con- process on filter temperature and total soot

centration is sufficiently high, then by Le Chatelier’s mass in filter. Inlet gas temperature increasedto 763 K from 660 K in 5 s, kept constant forprinciple, WGS will shift such that production ofabout 5 s to promote rapid oxidation, thenCO and H

2O is more favoured. This in turn will

reduced to 423 K in 5 sencourage the CO oxidation and more heat will bereleased and hence increases the filter temperature.

when the exhaust gas flow was too cold (i.e. when itreached 423 K), leaving approximately 60 per cent ofthe original 14 g of soot remaining in the filter.5 TEMPERATURE EFFECTS ON EXHAUST GAS

For the case of carbon monoxide and carbonSPECIESdioxide (see Fig. 13), it can be observed that theconcentration of CO and CO

2decreased when theIn this section, the effect of the reducing exhaust-gas

exhaust gas was quenched. In addition, it can betemperature during the rapid regeneration period isobserved that the concentration of CO fell below itsstudied to simulate and investigate the quenchinginitial value while that of CO

2is the same as its inputprocess that can lead to incomplete regeneration.

value. This is because CO2

is more favoured thermo-Figure 12 shows the DPF inlet gas-temperaturedynamically than CO. Validation of these predictionsboundary condition used to simulate potentialis currently being addressed experimentally.quenching during rapid regeneration. The inlet gas

temperature was first raised from 660 K to 763 K in5 s to promote rapid regeneration (oxidation), con-

6 CONCLUSIONStinued for about 5 s and then reduced to 423 in 5 s.The filter temperature continued to increase to 809 K

A multistep chemical reaction scheme has beenafter the exhaust flow was quenched. After that, itincorporated into a diesel particulate filter (DPF)decreased gradually. The soot regeneration stoppedregeneration model. The model is based on funda-mental kinetic data and therefore it has no further

Fig. 11 The effect of inlet H2

concentration on the gasFig. 13 CO and CO

2profiles during quenchingtemperature in the filter



8 Jørgensen, M. W. and Sorenson, S. C. Aarbitrarily ‘tuned’ global reaction parameters. The2-dimensional simulation model for a diesel parti-model has been compared with other experimentalculate filter. SAE Paper, 1997, no. 970471.findings and the agreement is reasonably good, with

9 Haralampous, O. A., Koltsakis, G. C., Samaras,acknowledged limitations. Z. C., Vogt, C.-D., Ohara, E., Watanabe, Y., and

The effects of various gas species on the DPF Mizutani, T. Reaction and diffusion phenomena inregeneration behaviour have been studied. For the catalyzed diesel particulate filters. SAE Paper 2004,case of the thermal regeneration process, CO

2is no. 2004-01-0696.

10 Haralampous, O. A., Koltsakis, G. C., Samaras,found to have a heat sink effect, whereas CO is foundZ. C., Vogt, C.-D., Ohara, E., Watanabe, Y., andto increase filter temperature. On the other hand, theMizutani, T. Modeling and experimental study ofmodel suggests that the water concentration shoulduncontrolled regenerations in SiC filters with fuelbe kept to a minimum so that it does not cause highborne catalyst. SAE Paper, 2004, no. 2004-01-0697.

temperatures in the filter. Exhaust gas, H2, is found 11 Pontikakis, G. N. Modeling, reaction schemes and

to have a negligible effect on regeneration. kinetic parameter estimation in automotive catalyticThe model is able to predict more detailed chemi- converters and diesel particulate filters, Doctoral

cal processes occurring during regeneration. It has Thesis, University of Thessaly, 2003.12 Cooper, B. J. and Thoss, J. E. Role of NO in dieselalso been used to illustrate the effect of prematurely

particulate emission control. SAE Paper, 1989,quenching the regeneration process by reducing theno. 890404.filter-inlet gas temperature, thereby leading to partial

13 Hoard, J. Plasma-catalysis for diesel exhaust treat-filter regeneration.ment: current state of the art. SAE Paper, 2001,no. 2001-01-0185.

14 Hammer, T. Non-thermal plasma application tothe abatement of noxious emissions in automotiveACKNOWLEDGEMENTSexhaust gases. Plasma Sources Sci. Technol., 2002,11, A196–A201.

The authors gratefully acknowledge Perkins Engines 15 Turns, S. R. An introduction to combustion, 2ndand the Royal Academy of Engineering for their edition, 2000 (McGraw-Hill International Editions).financial support of the third author. 16 Stull, D. R., Chao, J., Dergazarian, T. E., Plessis,

L. A. D., Hadden, S. T., and Justice, B. H. JANAFThermochemical Tables. 1965 (National Bureau ofStandards, USA).

17 Higuchi, N., Mochida, S., and Kojima, M. OptimizedREFERENCESregeneration conditions of ceramic honeycombdiesel particulate filters. SAE Paper, 1983, no. 830078.1 Bissett, E. J. Mathematical model of the thermal

18 Hayhurst, A. N. and Parmar, M. S. Does solidregeneration of a wall-flow monolith diesel parti-carbon burn in oxygen to give the gaseous inter-culate filter. Chem. Eng. Sci., 1984, 39(7/8), 1233–1244.mediate CO or produce CO

2directly? Some experi-2 Bissett, E. J. and Shadman, F. Thermal regeneration

ments in a hot bed of sand fluidized by air. Chem.of diesel particulate monolithic filters. AIChE J.,Eng. Sci., 1998, 53(3), 427–438.1985, 31(5), 753–758.

19 Yezerets, A., Currier, N. W., and Eadler, H. A.3 Garner, C. P. and Dent, J. C. A thermal regenerationExperimental determination of kinetics of dieselmodel for monolithic and fibrous diesel particulatesoot oxidation by O

2—modeling consequences. SAEtraps. SAE Trans., J. Passenger Cars, 1988, 97(4), 9–24.

Paper, 2003, no. 2003-01-0833.4 Opris, C. N. and Johnson, J. H. A 2-D computational20 Graven, W. M. and Long, F. J. Kinetics and mech-model describing the heat transfer, reaction kinetics

anisms of the two opposing reactions of the equilib-and regeneration characteristics of a ceramic dieselrium CO+H

2O=CO

2+H

2. J. Am. Chem. Soc., 1954,particulate trap. SAE Paper, 1998, no. 980546.

76, 2602–2607.5 Konstandopolous, A. G., Kostoglou, M., Housiada, P.,21 Jacquot, F., Logie, V., Brilhac, J. F., and Gilot, P.Vlachos, N., and Zarvalis, D. Multichannel simu-

Kinetics of the oxidation of carbon black by NO2

lation of soot oxidation in diesel particulate filters.influence of the presence of water and oxygen.SAE Paper, 2003, no. 2003-01-0839.Carbon, 2002, 40(3), 335–343.6 Stratakis, G. A. and Stamatelos, A. M. Flow distri-

22 Olson, K. E., Luss, D., and Amundson, N. R.bution effects in the loading and catalytic regenera-Regeneration of adiabatic fixed beds. Indl Engngtion of wall flow diesel particulate filters. Proc. InstnChem. Process Design Development, 1968, 7, 96–100.Mech. Engrs, Part D: J. Automotive Engineering, 2004,

23 Otto, K., Sieg, M. H., Zinbo, M., and Bartosiewicz, L.218, 203–216.The oxidation of soot deposits from diesel engines.7 Pontikakis, G., Stamatelos, A., Bakasis, K., andSAE Paper, 1980, no. 800336.Aravas, N. 3-D catalytic regeneration and stress

24 Ishiguro, T., Takatori, Y., and Akihama, K. Micro-modeling of diesel particulate filters by ABAQUSFEM software. SAE Paper, 2002, no. 2002-01-1017. structure of diesel soot particles probed by electron



microscopy: first observation of inner core and outer ks

bulk thermal conductivity of porousshell. Combustion and Flame, 1997, 108, 231–234. solid (k./s−1.m−1.K−1)

25 Yezerets, A., Currier, N. W., Eadler, H., Popuri, S., Kc

equilibrium constant based on molarand Suresh, A. Quantitative flow-reactor study of

concentration (various units)diesel soot oxidation process. SAE Paper, 2002,K

pequilibrium constant based on partialno. 2002-01-1684.pressures (various units)26 Ahlstrom, A. F. and Odenbrand, C. U. I. Com-

Kp,formation

equilibrium constant for speciesbustion characteristics of soot deposits from dieselengines. Carbon, 1989, 27(3), 475–583. formation (various units)

27 Neeft, J. P. A., Nijhuis, T. X., Smakman, E., m mass fraction (–)Makkee, M., and Moulijn, J. A. Kinetics of the oxi- N total number of reactions (–)dation of diesel soot. Fuel, 1997, 76(12), 1129–1136. PM particulate matter

28 Kilpinen, P., Kallio, S., Konttinen, J., and Barisic, V.ppm parts per millionChar-nitrogen oxidation under fluidised bed com-r stoichiometric ratio (–)bustion conditions: single particle studies. Fuel,R

uuniversal gas constant (kJ.kmol−1.K−1)2002, 81, 2349–2362.

t time (s)29 Hottel, H. C., Williams, G. C., Nerheim, N. M., andSchneider, G. R. Kinetic studies in stirred reactors: T absolute temperature (K)combustion of carbon monoxide and propene. WGS water–gas shiftTenth symposium (international) on Combustion, Y

jchemical formula of species j

The Combustion Institute, 1965, pp. 111–121. yj

mole fraction of species j30 Johnston, H. S., Bonner, W. A., and Wilson, D. J.

z coordinate (m)Carbon isotope effect during oxidation of carbonZ total thickness (particulate, ceramic,monoxide with nitrogen dioxide. J. Chem. Phys.,

or combined) (m)1957, 26(5), 1002–1006.31 http://www.iupac-kinetic.ch.com.ac.uk last accessed

bp

mass concentration of particulate per16 December 2004.32 Johnson, T. V. Diesel emission control in review— unit volume (bulk density) (kg.m−3)

the last 12 months. SAE Paper, 2003, no. 2003-01-0039. n∞ gas-phase reactant stoichiometriccoefficient (–)

n◊ gas-phase product stoichiometriccoefficient (–)

r density (kg.m−3)APPENDIXQ medium porosity (–)

NotationSubscripts

a specific area (m−1) f gas phase forward reactionA frequency factor (s−1) G gas phaseA

filtotal filtration projected area (m2) i ith reaction

b mass transfer rate (s−1) j general gas constituentc

pspecific heat capacity at constant p particulatepressure (kJ.kg−1K−1) pore pore in ceramic substrate

C1, …, C

5polynomial constants (various units) r gas phase backward reaction

d diameter (m) Reg regenerationDPF diesel particulate filter S solid phaseE

aactivation energy (kJ.kmol−1) w wall

h convective heat transfer coefficient(kJ.s−1.m−2.K1) Superscripts

H enthalpy of reaction (kJ.kg−1) ◊ per unit area (m−2)G gas flowrate (kg.s−1) ∞∞∞ per unit volume (m−3)

. derivative with respect to time (s−1)k chemical reaction rate (s−1)


DPF Regeneration

Documents

Transcript of DPF Regeneration