ACTA MECHANICA SINICA, Vo1.19, No.3, June 2003

The Chinese Society of Theoretical and Applied Mechanics

Chinese Journal of Mechanics Press, Beijing, China

Allerton Press, INC., New York, U.S.A.

ISSN 0567-7718

SI MULATI ON OF NOx FORMATI ON I N TURBULENT SWI RLI NG

COMBUSTI ON USI NG A USM TURBULENCE- CHEMI STRY MODEL*

ZHOU Lixing ( ~2q~) t QIAO Li ( ~ Sg) ZHANG Jian ( ~ f~)

(Department of Engineering Mechanics, State Key Laboratory of Clean Coal Combustion,

Tsinghua University, Beijing 100084, China)

ABSTRACT: A unified second-order moment (USM) turbulence-chemistry model for simulating

NOx formation in turbulent combustion is proposed. All of correlations, including the correlation of

the reaction-rate coefficient fluctuation with the concentration fluctuation, are closed by the transport

equations in the same form. This model discards the approximation of series expansion of the exponen-

tial function or the approximation of using the product of several 1-D PDF's instead of a joint PDF. It

is much simpler than other refined models, such as the PDF transport equation model and the condi-

tional moment closure model. The proposed model is used to simulate methane-air swirling turbulent

combustion and NOx formation. The prediction results are in good agreement with the experimental

results.

KEY WORDS: swirling turbulent combustion, NOx formation, second-order moment model

1 I NTRODUCTI ON

The numerical simulation of NOx formation

in turbulent combustion is frequently used in the

optimization design of low-NOx burners and fur-

naces. Although the direct numerical simulation

(DNS)[ 1], large-eddy simulation (LES)[ 2], probabil-

ity density distribution function ( PDF) t ransport

equation model[ 3] and the conditional moment clo-

sure (CMC) model [4], developed in recent years, can

well simulate the interaction of turbulence with de-

tailed chemistry, these refined models need rat her

large computer memory and comput at i on time. They

can be used only in simulating very simple flows

in fundamental studies. For engineering combus-

tion modeling, the widely used models, including

those adopted in the commercial software, are the

Eddy-Break-Up (EBU)-Arrhenius (denoted as E-A in

the following text) model, the presumed-PDF fast-

chemistry model for turbulent combustion [5] and the

presumed-PDF finite-rate chemistry model for NOx

formation [6] . However, the PDF-fast -chemi st ry model

cannot account for the finite-rate kinetics and the E-

A model is also not handy to account for the finite-

rate kinetics. The PDF finite-chemistry model using

Received 27 October 2001, revised 24 May 2002

a product of several single-variable PDF's instead of

a joint PDF leads to an under-prediction over the av-

eraged reaction rate. Alternatively, the second-order

moment turbulence-chemistry models, based on the

idea of second-order moment turbulence models, are

more reasonable t han the E-A model and the pre-

sumed PDF models, while they are more economical

t han other refined models.

It is well known t hat the difficulty in developing

the second-order moment turbulence-chemistry mod-

els lies in the t reat ment of the exponential function of

t emperat ure in the time averaging procedure. Previ-

ously, two versions of the second-order moment mod-

els were developed. In the first version a series ex-

pansion of the exponential t erm with an approxima-

tion of E/RT << 1 is made[ 5]. This model is used

to simulate methane-air and hydrogen-air turbulent

diffusion combustion Iv's]. The predicted t emperat ure

and mai n species concentration are in good agreement

with experimental results, but the NOx concentration

is under-predicted. The reason is t hat for NOx forma-

tion the activation energy is large and E/RT is much

larger t han unity, so the approxi mat i on of E/RT << 1

leads to a significant under-prediction over the aver-

aged reaction rate. The second version is the second-

* The project supported by the Special Funds for Major State Basic Research of China (G1999-0222-07)

t E-mail: zhoulx@mail.tsinghua.edu.cn

Vol.19, No.3

Zhou Lixing et al.: Simulation of NOx Formation in Swirl Combustion

209

order-moment PDF model, in which the concentration

correlation is closed using the second-order moment

equations, while the t emperat ure-concent rat i on cor-

relation is closed using the presumed PDF[ 9,1~ Sim-

ulation of methane-air jet diffusion combustion shows

t hat this version of the second-order moment model,

not using the series expansion approximation, is much

bet t er than the first version of the second-order mo-

ment model and the E-A model. However, it still does

not entirely get rid of the approximation of using the

product of two single-variable PDF's instead a joint

PDF.

In this paper, a third version of second-order

moment model, i.e. a unified second-order moment

(USM) model is proposed. The feature of this new

model is t hat when solving the time-averaged reac-

tion rate, the correlation of the reaction-rate coeffi-

cient fluctuation with the concentration fluctuation is

closed using the t ransport equation in the same form,

as t hat of the t ransport equations for all other cor-

relations, assuming t hat the production t erm of this

correlation is proportional to t hat of the t emperat ure-

concentration correlation. The validity of this closure

can be verified only by experiments. Besides, the

effect of reaction on the dissipation of correlations

is taken into account. The proposed model is used

to simulate methane-air swirling turbulent combus-

tion and NOx formation. The prediction results are

validated using the experimental results taken from

Ref.[ll].

Eq.(1) becomes

Ws -- +

rfuk'ro'x + Lxk'Yf'] (2)

where k = f exp(-E/RT)p(T)dT, p(T) is the tem-

perat ure PDF. The correlations MY' and Y~Yorx are

closed using a unified form of t ransport equations.

The generalized form of t ransport equations of these

correlations when accounting for the effect of chemi-

cal reaction on the dissipation using two time scales

is

0

05 -Sgx5 ] +

0~0~ ( ab)

Cgl"T OXj OXj Cg2 ~ ~- < /9r (3)

where

k

TTZ --

C

Cg I = 2.8

--1

Cg2 = 2.0 a+ b = 1

For the correlation of the fluctuation of the reaction-

rate coefficient with the concentration k'Y', it is

very difficult strictly deriving the t ransport equa-

tion. Assuming t hat its t ransport equation takes the

same form as t hat for the t ransport of t emperat ure-

concentration correlation and its production and dis-

sipation are proportional to those of T'Y', we have

2 THE USM TURBULENCE-CHEMISTRY

MODEL

For an elementary reaction of any two species or

a global one-step reaction with two species, for exam-

pie, fuel and oxygen, the instantaneous reaction rate

is

Ws = Bp2YfuYox exp(- e/RT)

where Yfu, Yox, T express the instantaneous values of

fuel mass fraction, oxygen mass fraction and t emper-

ature, respectively, E denotes the activation energy

and R the universal gas constant. After taking the

Reynolds averaging the time-averaged reaction rate

takes the following form

Ws = Bp2kYfuYox

= Bp2( + #)(?fu + U3( ox + Y'x) (1)

where, k = exp(-e/RT).

When neglecting the third-order correlation

0 0 (,;0k Y,)

Cgl/AT ~

Cg2p + (4)

When taking the t op-hat PDF of t emperat ure, the

time-averaged reaction-rate coefficient is

9T = Tt2

For the chemical kinetics of methane-air com-

bustion a global reaction kinetics of Arrhenius type is

taken as

Wfu = 1.0 X 101~ exp(-1.84 x 104/T) (5)

210 ACTA MECHANICA SINICA 2003

The time-averaged reaction rate is

~uk'Yo'x + ?o~k'y~'] (6)

The transport equation of Yf~Yot~ is

_

ot ~xj

o (~ oyi'Yg, x ~ o~ Ofox

aXj ~ ~Xj ] ~- ggl#t OXj OXj

i-a)

a y, y,

Cg2P -~ fu ox (7)

TA

where cgl = 2.8, Cg2 = 2.0, a = 0.5. The transport

equation of k'Y' is Eq.(4).

For NO formation the Zeldovich mechanism of

thermal NO formation and Finemore mechanism of

prompt NO formation are used as

Wz =9 x 1012TO.3exp ( 38440) [N2][O2] (8)

Wp = {3 x 1010 exp(-2 900/T).

F3/2G[N2][CH4][021 a/2 [H20]1/2/

([M]- [N2])([M]- [N2]- [CO2])]}.

[1 + 3 000 exp(-15 185/T)] (9)

where B1 = 6MNo/4MNHa, B2 = MNO/J ~f NHa, M is

the species "molecular weight.

Therefore the laminar reaction rate of NO for-

mation is

WNO ~- WNO,fuel -]- Wz -F Wp (12)

For the turbulent reaction rate of NO formation,

the time-averaged reaction rate using the USM model,

for example, for fuel NO is

WNO,fue I = WNO,fue 1 X (1 + Z1)

E l Y,' k' Y/ k' Y~

NHa O2 1 NHa ~_ 1 02

Zl -- YNH3Yo2 + ~IYNH3 kl?O2

(13)

- + +

W;o,f.,ol x + = W;o,f.o 1 (1 12)

Y/ Y/ k I Y/ k I E I

NH3 O2 2 NHa 2 02

Z2=- - +- - +- -

YNHa YO2 k2YNHa k2Yo2

where

13 588

1 [exp ( T+

f~l ~ = T,) +exp ( L3588~

]

= ~ - T'Jj

1 [ ( 16105) ( 16105~]

k2= ~ exp '~+T' +exp T- T']

(14)

The correlations in Eqs.(13), (14) are determined also

by Eq.(3).

where T is

centration (mole/cm3s), F and G are given by

F=I - I.1 x 10-Z~ 3exp - [02]

G- - l +40exp( - ~- ) +2.6xl O13T -4.

exp -- [H20]

For fuel NO the following mechanism is used

4NH3 + 6NO ~ 5N2 + 6H20

4NH3 + 502 ~ 4NO + 6H20

The fuel NO reaction rate is determined by

WNO,fue I = 1.8 x IOspYNHaYNoB1 exp ( - - - -

+ IOspYNHayb2B2 exp (

W~o,f ue 1 = 4.0 - - -

\

the temperature (K), [M] is the mole con-

(10)

)

1T5 )

(11)

3 SIMULATION OF NOx FORMATION IN

METHANE- AIR SWIRLING TURBU-

LENT COMBUSTION

The geometrical configuration and sizes of the

swirl combustor to be predicted are shown in Fig.1

and Table 1. The fuel--methane is supplied from

the central tube and the swirling air is supplied from

the annular tube with a swirler. The air flow rate is

8.gm3/h and the methane flow rate is 0.8932m3/h.

The swirl number is s = 0.43. A small amount of

]4 L~

methane+ammonia

Fig.1 The swirl combustor

Vol.19, No.3

Zhou Lixing et al.: Simulation of NOx Formation in Swirl Combustion

211

Tabl e 1 The geomet r i cal si zes of t he combus t or

D1/mm D2/mm Da/mm Df/mm Dout/mm Lf/mm

8 i0 30 160 180 900

ammoni a (4.91%) is added to t he fuel t o si mul at e fuel

NO in case of gas combust i on. The exper i ment al re-

suits for t ur bul ent combust i on and NO f or mat i on in

this combust or are t aken from Ref.[11].

The modi fi ed k-e t ur bul ence model wi t h a cor-

rect i on account i ng t he effect of swirl and unified

second-order moment t ur bul ence- chemi st r y model are

used t o si mul at e t ur bul ent swi rl i ng flows, combus-

t i on and NO format i on. The comput at i on domai n

is 0.9 m x 0.08 m. 80 x 45 st ager r ed gri d nodes are

adopt ed, as shown in Fig.2. The di fferent i al equat i ons

are di scret i zed i nt o finite difference equat i ons and t he

FDE's are sol ved using t he SI MPLEC al gori t hm. For

boundar y condi t i ons, uni form di st r i but i on of different

vari abl es is t aken at t he inlet; symmet r i cal condi t i ons

are t aken at t he axis and ful l y-devel oped flow con-

di t i ons are t aken at t he exit. No-sl i p condi t i ons are

t aken at t he wails. For near-wal l gri d nodes t he wall

funct i on appr oxi mat i on is used. The comput er codes

consists of about 4 000 st at ement s. Runni ng a case in

a Pent i num- 3- 550PC t akes about 10 h.

0.08~!!!!!!!!!!!!!!!![!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!]!!!!]!!![!! !! ' ! ! ! ! ! ! ! ! ! !!

':'''''''''''''']E[`'''''''''''''''''''':'::`''':'':::::::: ',: ', ', :', : ', ', ', : ', ', ', ', ',',

0.0 7 IHIIIIIIIIIIII]II[IIIIIIIIIIIIIIIIIII[II[II[Illll i i i i i i i i I [ I I I I I I I I I I I [ I I I

m......i,i,m..mmmmu,Hllll,,lll i i i i i i i i i i i i ~ i i ~ i i i i i i r

HIIIIIIIIIIIIII/II}IIIIIIIIIIIIIIIIIIIII[IIIIIIIll I I I I I I I I I I I I I I I I I i I I I I I I I

0 06 mIIIIIIImI~ImUIIUUUUlUUIIUHIIIII~IIII I I I t I I nell In ~l I J I I I I I IF

. IIllllUlllllfi]LEklllllllllllllllllllllilLIIIII]lll I I I [ [ I I k I I I I [ I I I I I I I I I I

Ullllml[]l]B[llllllllllllllllllllllll[llll]]lll [i i k [ [ i i i i I] [I I I [ I I I I II

0.05 ........................................................... II ........... IIIIIIIIIlIIIlIlilllllllllllllllllllllllllll[lllllll I I I I I I I I I I I I I I I [ I I

IIIIIIIIII/ILl[lllllllllllllllllllllllllllllllllll[I J I I I I I IIII I I I I

"~- 0.04 ........................................................... i]

mllullmHIIIIIIIIIIIIII]lllllll]llllll II ] I I I I I I I I I I I I I I I I I I

m,~,nnnnnnnnnnnmnmnl.ullllllnunnn ii i i n n n n n n n n n n ~ n n ii

0.03 ........................................................... ~[F~IIIIIIIIIIIIIIIIII]IIII]HIIIIII]I]II II I L [ ~ I I I I I I I I I I I I I 1

I::I:,~:::::::::::II:::ET, I:P,:IIIIIII W~ llll::::llll : I I I I I 1

0.02 ~:::::::::::::::I:['??,II,[F,I::2J, V,I N Ill III:::::I I I I I I I I I I l

::::::::::::::::::::::::::::::::::::::::::::: ::::: : : : : : : : : : : : :

0.01 ':':':': ': ':

0

0 0.2 0.4 0.6 0.8

x/m

Fig.2 Computation domain and grid arrangement

4 PREDI CTI ON RESULTS AND DI SCUS-

SI ON

Figure 3 gives the streamlines. It can be seen

that under the effect of swirl there is a large-size cor-

ner recirculation zone and a small-size central recir-

culation zone. Figure 4 shows the temperature maps.

The high temperature region developed in the corner

recirculation zone and the central recirculation zone

plays the role of flame stabilization. Lower tempera-

ture exists at the inlet and near-wall region. Figure

5 gives the comparison of predicted temperature pro-

files at 7 cross sections with the experimental results

in Ref.[ll]. The agreement is good. The comparison

of the predicted NO concentration with the experi-

mental results is shown in Fig.6. In general, the pre-

diction results are near to those measured. Quantita-

tively, the model over-predicts the NO concentration

1.0

0.8

0.6

0.4

0.2

0.0

o~

0 5

x/R

Fig.3 Streamlines

I

10

0.08

0.06

"--- 0.04

0.02

/~ 1339 1274 1218 1173

0.00

0.0 0.2 0.4 0.6 0.8

x/m

Fig.4 Temperature maps (unit: K)

~4

O ~

r '

600 1 200

600 1 200

-%.

x=17.5

,

600 1 200

x=27.5

600 1 200

T/K

predictions

i , vl

600 1 200

experiments

4

)

t

x----52.6 x=702

, ~ ,',

600 1200 600 1200

Fig.5 Temperature profiles

212

i

x~5

i , i v ,

600 1 200

ACTA MECHANICA SINICA

10 - 7 "~

@

600 1 200

~=40

i v

600 1 200 600 1 200 600 1 200

[NO]/ppm

predictions experiments

Fig.6 NO concentration profiles

I

-5

I

I

x=70

a----52 .~ I

.l

600 1 200 600 1 200

2003

in t he first 4 cross sections. Thi s di scr epancy may be

caused by t he adopt ed over-si mpl i fi ed r eact i on ki net -

ics. It can be seen t hat due t o t he effect of swirl t he

NO concent r at i on in t he corner regi on is higher.

5 CONCLUSI ONS

(1) The NO f or mat i on in swi rl i ng t ur bul ent combus-

t i on can be r easonabl y si mul at ed using t he uni-

fied second-order moment t ur bul ence- chemi st r y

model

(2) More det ai l ed ki net i cs shoul d be t aken into ac-

count t o i mprove t he predi ct i ons.

REFERENCES

1 Tanahashi M, Yu Y, Miyauchi T. Effects of turbu-

lence intensity on the structure of hydrogen-air tur-

bulent premixed flame. In: Nagano Y ed. Proc 3rd

Inter Symp on Turbulence, Heat and Mass Transfer,

Nagoya, 2000-4-2~6, Nagoya: Aichi Shuppan, 2000.

823~830

2 Park N, Kobayashi T, Taniguchi N. Application of

flame wrinkling LES combustion models to a turbulent

premixed combustion around bluff body. In: Nagano

Y ed. Proc 3rd Inter Symp on Turbulence, Heat and

Mass Transfer, Nagoya, 2000-4-2~6, Nagoya: Aichi

Shuppan, 2000. 847~854

3 Pope SB. PDF methods for turbulent reactive flows.

Prog Energy Combust Sci, 1985, l h 119~192

4 Bilger RW. Conditional moment closure for turbulent

reacting flows. Physics of Fluids, 1993, A5:436~473

5 Zhou LX. Theory and Numerical Modeling of Turbu-

lent Gas-Particle Flows and Combustion. Beijing: Sci-

ence Press and Florida: CRC Press, 1993

6 Zhou LX. Advances in numerical modeling of t urbu-

lent reaction rat e of NOx formation. Advances in Me-

chanics, 2000, 30:77~82 (in Chinese)

7 Liao CM, Liu Z, Liu C. NOx prediction in 3-D turbu-

lent diffusion flames by using implicit multigrid meth-

ods. Combustion Science and Technology, 1996, 119:

219~260

8 Zhou X, Sun Z, Brenner G, Durst F. Combustion mod-

eling of turbulent jet diffusion hydrogen/ai r flame with

detailed chemistry. Int J Heat and Mass Transfer,

2000, 43:2075~2088

9 Chen XL, Zhou LX, Zhang J. Numerical simulation of

methane-air turbulent jet flame using a new second-

order moment model. Acta Mechanica Sinica, 2000,

16:41~47

10 Zhou LX, Chen XL, Zheng CG, Yin J. Second-order

moment turbulence-chemistry models for simulating

NOx formation in gas combustion. Fuel, 2000, 79:

1289~1301

11 Chen XL. A second-order moment PDF turbulent

combustion model and studies on the NO formation

in swirl combustion [Ph D Thesis]. Department of

Engineering Mechanics, Beijing: Tsinghua University,

2001

## Comments 0

Log in to post a comment