Sddhand, Vol. 9, Part 4, December 1986, pp. 255-269. L~) Printed in India.

Noise reduction with perforated three-duct muffler components

K NARAYANA RAO* and M L MUNJAL t Department of Mechanical Engineering, Indian Institute of Science, Bangalore 560 012, India ::'Present address: Mechanical Engineering Department, University College of Engineering, Osmania University, Hydcrabad 500 0~7 India : Present address: Department of Mechanical Engineering, University of Calgary, 25{)0 University Drive NW, Calgary, Alberta, Canada T2N IN4 MS received 21 March 1986; revised 28 May 1986

Abstract. This paper describes the authors' dzstributed parameter

approach for derivation of closed-form expressions for the four-pole parameters of the perforated three-duct muffler components. In this method, three simultaneous second-order partial differential equations are first reduced to a set 9f six first-order ordinary differential equations. These equations are then uncoupled by means of a modal matrix. The resulting 6 6 matrix is reduced to the 2 x 2 transfer matrix using the relevant boundary conditions. This is combined with transfer matrices of other elements (upstream and downstream of this perforated element) to predict muffler performance like noise reduction, which is also measured. The correlation between experimental and theoretical values of noise reduction is shown to be satisfactory. Keywords. Mufflers; noise control; transfer matrix method.

I. Introduction
The majority of the commercial mufflers used on internal combustion engines contain one or more perforated tubes. These may interact with the surrounding annular cavity as in a Helmholtz resonator, or may conduct the gases as in plug mufflers and three-duct cross-flow or reverse-flow mufflers. Unfortunately, however, analysis of such perforated-element mufflers was not known until a few years ago, with the result that the design of commercial mufflers was based on trial and error. Recently, the authors developed and presented a generalized decoupling method for the analysis of perforated-element mufflers with and without mean flow, wherein the mean flow Mach numbers are allowed to have their actual (unequal) values (Rao & Munjal 1984). But the method was demonstrated for two-duct elements only. This method makes use of the distributed parameter approach in

256

K Narayana Rao and M L Munjal

contrast to Sullivan's segmentation approach (Sullivan 1979a, b), where a perforated clement is notionally divided into a number of segments. The present work extends the distributed parameter approach to three-duct muffler element configurations by removing the ambiguities in evaluation of the eigenvalues of the characteristic matrix. The three coupled partial differential equations for wave propagation have been solved using the authors' generalized decoupling approach with and without mean flow incorporating the relevant exact boundary conditions. Explicit expressions have been derived for four-pole parameters of the cross-flow expansion chamber as well as the reverse-flow expansion chamber, shown in figure 1. The transfer matrices so derived have been checked experimentally by comparing the predicted values of noise reduction with those observed experimentally.

2. The coupled governing equations

Implicit in the following equations, and the solution thereof, are certain assumptions that are typical of automotive muffler analysis, viz, (i) working in the frequency domain, all acoustic variables are made to have harmonic time-dependence everywhere; (ii) at the frequencies of interest, only plane waves would propagate; higher order modes could be ignored; (iii) mean flow velocities are small enough; the flow is incompressible; (iv) heat transfer and viscosity effects are neglected; all processes are isentropic; mic rophone M!

dl~

!'
it,)

!2

microphone
--.--.~M5

a i2 [I

i: 'I ~h['~
,

-I

microphone

MI
micro.phone
-q0----M3

d1.[

d3[

t3
Figure I. Three-duct muffler configurations: a. cross-flow expansion b. reverse-flow expansion chamber. chamber,

Noise reduction with perforated muffler components

257

(v) mean flow gradients along and across the duct are neglected. The same holds for mean pressure and mean density; (vi) velocity of cross flow is assumed to be uniform all over the perforate. This allows us to assume a uniform acoustic impedance for the perforate. For the three-duct model as shown in figure 2, the mass continuity equations may be written as (Sullivan & Crocker 1978)
C]W I C~Dl 4 P(1-- + Wo, +--poUl,

Opj
2 --

~z

c)z

d~

c)t

(1)

for duct 1 of diameter d~, P o - - +Wo, az Oz ape

014'2 ,9p2 4dl 4d3 poUl, 2 +

(d_~- d~ - d~)

(d~ - d~ - d~)

pou2, 3

Ot
for duct 2 of diameter d2, and

(2)

Ow3
P(1 - -

Oz

01)3 + W,,, - Oz

4
d3

Po u2.3

Op3
3t

(3)

for duct 3 of diameter d3. The notation is given in appendix D. The corresponding momentum equations are
Po

( ow, + wo,-c)%) z O---t

Opj az

"

j = 1''~ 3" -"

(4), (5), (6)

The radial momentum at interfaces of duct 1 and duct 3 result in the equations

ul,z(z, t) = [p~ (z, t ) - p 2 ( z , t)]/(po co ~1),

and

(7a) (7b)

u2.3(z, t) = [p2(z, t ) - p 3 ( z , t)]/(p,, co ~z),

respectively.

Assuming that the fluid is an ideal gas and the process involving wave propagation is isentropic, pj=pjc~, j= 1,2,3. (8)

For harmonic time dependence,

pj(z, t) = pi(z) e i't,

and

j = 1, 2, 3,

(9a)

ui, k(z, t) = Uj.k(Z) e i,o,, j = l , 2 ,

k=2,3.

(9b)

Substituting (7a), (7b), (8), (9a) and (9b) into (1) to (6) and eliminating Pl, P2, f13, //1,2, /,/2,3, Wl, w2 and w3, yields the following three coupled differential equations for ducts 1, 2 and 3, respectively (Munjal 1986),

258

K Narayana Rao and M L Munjal

[ d2 dz 2

iM,
1 - M21

(k ] +k2)
k

k2 ] + - p~(z) dz 1 - M2 (10)

"1-

I-M
I iM2 k~ - k 2 d

dz

a-M
k 2 -- k 2

1 - M--~ J + d dz 2

I pl(z)
2

iM2
1-M 2

(k~ + k 2)
k

d
dz

2 kb + k c 2- k ]

' l~M2~

p2(z)

[ iM2 (kc2.-k 2 ) d + t 1-M 2 \ k dz iM____2_ 3 k2 d l_M2( /~ k2) d z

k2c-k2 1 ]--~222 j p3(z) = 0,

k2-k2 ]

(11)

l_M2

p2(z)

+ where

dz 2

1-M 2

dzz + - I - M

p3(z)=O,

(12)

k--

O)
--, M1-

Wo,
, M2-

Wo
CO

:,

M3Co

Co k2 = k2 _ _

CO

and

i4k i4kdl , k2 = k2d, s r, (d 2 - d 2 - d2) Srl ' i4kd3 k2 = k2(d 2 - d~ - - d 32 )~"2 ' i4k k~ = k 2 - - d3~" 2

(13)

3. Decoupling analysis
The analysis runs exactly as for the two-duct perforated element mufflers (Rao & Munjal 1984). The three second-order differential equations (10), (11) and (12) are first reduced to six first-order linear equatiolas. Treating the differential operator as an algebraic variable, these six equations are decoupled through an eigenvalue analysis and the use of principal coordinates. Finally acoustic state variables at z = 0 are related to those at z -- l through a matrix, which is the required transfer matrix.

Noise reduction with perforated muffler c o m p o n e n t s

259

Denoting d / d z by prime ('), and defining

Pl = Y | , P 2 = Y e , P 3 = Y3, Pl = Y4,

P2 = Y5 and

P3 = Y6,

(14)

(10),.(11) and (12) may be written in the form

--1 0
D 0 0

0 0
0 D 0

0 0 -1
0 0 D

D 0 0
0:lD+a2 asD+0:6 0

0 D 0 0:3D + 0:.)
0:7D+0:~ 0:11D+0:)2

0 0 D 0
agD+al(i 0:13D+0:14
I

Yl
Y2 Y3 Y4

0 -1

0]
0

Y5 ,. Y%

1
0 0
k

0 0

(15a)

or

[~l{y} = {0}, where

0:1 - ~

(15b)

1 - Mi

, 0:-, - -

, 0:3-

----

1 - M~

1 - M~

k]-k
"0:4 - -

2
.~ , 0:5 - -

iM9
I - M~
-

[ k~--k2]
k 7 -) ' 0:6 --

( k ~ , - k 2)
I - M~

I - Mi

0:l"7-

1 - M?~

" 0:~ 1 M~

' 0:~

- (kc - k 2) 0:to
--

iM 3 , 0:11 -1 - M ~ "~

[ k~ - k2 ~
k

k,et - k 2
, 0:12-

1-M~ -iM3

k ,

f-M 2

k2 + k 2 ~ ,

a,4 -

k2

(16)

a13and

l_M2

1-M~ (17)

d D - --. dz

Making use of principal coordinates, these equations may be decoupled. The new variables, i.e., the principal co-ordinates F~, Fe, [`3, [`4, r'5 and I"6 are related to variables yj, Y2, Y3, Y4, Y5 and y,, through the modal matrix [~] as

260 where

K Narayana Rao and M L Munjal

~ l , j = 1.0

(say),
+ 0/l & + 0/2)/(0/3& + 0/4),

't'2,j

= - (~

'q-t3, j =
~4,j

(/3~ + 0/,/3j +0/~) (t~ +,~t3j +,~s)- (0/.,t~j +0/~)(0/5/3j +0/~)

(0/3~j +0/4) (0/9~j + O~10)

= 1"0//3i,

~II5,j = ~It2.j/ [~j ,

and

qr64 = ~3,j/[3j,
and subscript j takes the values 1, 2 . . . . polynomial ]al = 0,

(19) 6. /3's are roots of the sixth degree (20)

to be found numerically on a computer by means of one of the standard subroutines. The general solutions to first order equations (20) can be obtained as rj(z) = Cje ~'z, (21)

where subscript j takes values 1, 2 . . . . 6. Next, (4), (5) and (6) may be used to obtain expressions for wl(z), w2(z) and w3(z). Finally, we may write

p1(z)

p2(z)
p3(z)

pocowl(z)

=D]

C2 C~

(22)

poCoW2(Z)
poCow3(z)

G
C6

where elements Ai, j involving only M, k and l are given by expressions noted in appendix A. Now the pressure and velocity at z = 0 can be related to the pressure and velocity at z = 1; that is,

p,(0)
p2(O)

p,q)
p2(l)

p3(0) meow, (0)

poCow2 (0)

-- [TI

p~(t) p, :owl (1)

poco w2 (1)

poCoW3(O)

poCoW3(l)

(23) (24)

where the transfer matrix is given by [T] = [.4(0)1 [Aft)]-'.

Noise

reduction

with perforated

muffler

components

26l

4. Transfer matrix for cross-flow expansion chamber

To eliminate four of the six state variables in the foregoing general solution, (23), we make use of four boundary conditions. In the case of cross-flow expansion chamber with rigid end termination, the boundary conditions are (Thawani & Jayaraman 1983)
z = 0 : pocow2(O) = - i z = 0 " pocow3(O) = - i z = 1 : pocowl(1) z = l : pocow2(l) = -i = -i

tan(k l,) p2(0) tanik 1,,) p.~(O) tan(k ll,) p L ( l ) tan(k lb) p 2 ( l ) . (25)

Skipping the elimination details, the final expressions for the upstream state variables and the downstream variables can be written as (Rap 1984)

poCoWl (0)

T~.

T,I

pocow3(l)

The expressions for T are given in appendix B.

5. Transfer matrix for reverse flow expansion chamber

The following changes have to be noted in this element: (a) mean flow in duct three is negative with respect to the reference direction; (b) the boundary conditions are different; these are
z = 0 : pocow2(O) = - i z = l : poc~jwl(l) z = I : pocow2(1) = -i = -i

tan(k t.) p2(O) tan(k lt,) p l ( l ) tan(k lh) p2(l) (27)

z = l : pocow3(I) = - i tan(k 11,) p 3 ( l ) .

Skipping the algebraic details of the elimination process, the transfer matrix relation for this element may be written as (Rap 1984)

[ o~c~w~(0)
6. Experimental evidence

L,

Tt,
-

oocow3(O)

'3'' I

and the resulting expressions for T,,, Tb, Tc and T,t are given in appendix C (Rap 1984).

In order to test the validity of the mathematical model at various mean flow conditions, noise reduction measurements were made on typical three-duct perforated element muffler configurations of figure 1.

262

K Narayana Rao and M L Munjal

Noise reduction, NR, is the difference in sound pressure levels at two arbitrarily selected points in the exhaust pipe and the tail pipe, located upstream and downstream of the muffler proper, respectively. Referring to figure 1, NR
=

SeLl- SPL3

20 logm

IP~/P3[,

(29)

where p denotes the root mean square value of the acoustic pressure perturbation. For the purpose of calculation, since p~ is not known directly, one can write

P,/P3 = (Pl/Po) "(Po/P3),

(30)

where P0 is the acoustic pressure a t the radiation end. Applying the wave relationships for the pipe section of length le, one obtains (Munjal 1986)

(P3/Po) =

cos

kole+ ( ig3/zo)

where Zo is the radiation impedance, The other pressure ratio

kole, po/vo.
sin

(31) (32)

P~/Pomay be obtained
T:I

from the transfer matrix relation (33)

[ PTl l l v l= [

T12] [ P I T z vo 2 '

where v denotes acoustic mass velocity, poAu, and Tii (i, j = 1, 2) are the fourpole parameters of the overall transfer matrix relating the state variables at points 1 and 0. This matrix is a product of the transfer matrices of the upstream pipe, perforated element, and downstream pipe. From (32) and (33) we obtain

(Pl/Po) =

Tll @ T 1 2 1 Z o

(34)

Equations (29), (30), (31) and (34) may be combined to obtain NR = 20 log10
7. Results and discussion

r . + r,2/ Zo
cos

kole+ i Y3/Zo'sin kole I

(35)

General Fortran programs have been developed for computation (on D E C 10 computer) of noise reduction for three-duct element configurations shown in figure 1 over the frequency range covered by pure plane wave propagation, in steps of 20 Hz. The following perforated impedances were incorporated in the noise reduction prediction (Sullivan 1979),

~= (610-3+i4.810-5f)/o -, for ~r = (0-514 dM/lo-+i4.8 10-5 f)/~r,

M = 0, for M >I 0.05.

Figure 2 shows the typical curves of noise reduction versus frequency of three-duct cross-flow muffler represented in figure la for a stationary medium (M = 0 condition). Neglecting the temperature gradient and the tube wall thickness, there is good agreement between the predictions of the present investigation and those of experimental results. In figures 3 and 4 a similar exercise is repeated for inlet flow, Mach number M = 0.10 and M = 0.2. Though there are small discrepancies, yet the agreement between theory and experiment is generally good.

Noise reduction with perforated muffler components

dB 120

263

I mtosurcmeNt~s p l"edlcttons

100

80

60

40

A
m 0

20

-20

0.5

i 1.0

1.5 frequency

2.0

2.5

3.0 kHz

Figure 2.
M~ = 0

Noise reduction of three-duct cross-flow expansion chamber of figure la for

dB

120 j

,
meosurement$

/
100 L

predictions

80
tO

60

40
m o e-.

20

-2C 0

I 0.5

I lO

1 1-5 frequency

I 2.0

t 2.5

3.0 kHz

Figure 3.
M1 = 0-1.

Noise reduction of three-duct cross-flow expansion chamber of figure la for

264

K Narayana Rao and M L Munjal

d8 120

IOO

measurefneNtE

predictions

8o

6C

~q

" 20

-20

0-5

1.0

1.5

2-0

2.5

3.0kHz

frequency

Figure 4. Noise reduction of three-duct cross-flow expansion chamber of figure ia for M1 = 0-2.

Noise reduction at zero mean flow of the reverse-flow expansion chamber of figure lb is shown in figure 5. The agreement between theoretical predictions and experimental observations is good in terms of peak locations as well as magnitudes. Figures 6 and 7 present computed and experimental noise reduction curves for the same muffler with M = 0.1 and M = 0-2 respectively. Here too, the theoretical results match the experimental results well. In general, experimental values of noise reduction are in good agreement with the predicted results, thereby substantiating the validity of the transfer matrices derived and used in the predictions. These transfer matrices may be used along with those for other non-perforated elements like a tube, sudden contraction, sudden expansion, extended inlet, extended outlet, flow-reversal contraction and flow-reversal expansion (Munjai 1975; Panicker & Munjal 1981), to predict the overall performance (irt terms of insertion loss, transmission loss, level difference or noise reduction) of a given muffler. These performance curves for typical dimensions may be used for table-design of an efficient muffler, keeping in mind other non-acoustical considerations like overall size, weight, cost of fabrication, back pressure on the engine, durability etc. (Munjal 1986).

The first author would like to thank the authorities of the Osmania University for the study leave granted. The second author acknowledges with thanks the financial support received from the Stiftung Volkswagenwerk, Hannover, Germany, for a project in collaboration with Professor M Heckl and Professor M Hubert of the Technische Universitiit Berlin.

Noise reduction with perforated muffler components

dB

265

1213

I rneosuremen[s

10C

p redictaons

80

~o 6O

~- ~0 3
C

20

1
L

o o

--20

0.5

1.0

1.5 frequency

2.0

2.5

3-0 kHz

Figure 5. M] = 0.

Noise reduction of three-duct reverse-flow expansion chamber of figure lb for

d8 1"20 100

I -

[
meosuremen~s

p redlcQons

80
= o

6O

40
o c

20

-20

0.5

I-0

1-5 frequency

2.0

2.5

3.0 kHz

F i g u r e 6.

Noise reduction of three-duct reverse-flow expansion chamber of figure lb for

Mi = 0.1.

266

K Narayana Rao and M L Munjal

dB 120

l -

I meosur@rnc~nts predicU0ns

100

80
tO t)

60 /.0

tn o c

20

-20 0

I 0.5

I 1.0

1 1.5

I 2.0

t 2.5

3.0kHz

frequency
Figure 7. Noise reduction of three-duct reverse-flow expansion chamber of figure l b for M1 = 0-2.

Appendix A
The following relations hold for both the three-duct cross-flow expansion chamber and the three-duct reverse flow expansion chamber.

A 1 j = ~bn,je~jz, A2,i = ~5,jet3j~, A3, j = dd6,/eOjz, Z4,j = -e~jZ/(ik + Ml/3j),

A s j = -~2deOjZ/(ik+M2~y),
A6, j =

-~3,je t~jz/(ik + M3flj ).

Appendix B
The following relations hold for three-duct perforated element cross-flow expansion chamber:
T, = TTI,2+A3C3; 7",. = T T 3 , 2 + A 3 D 3 ;

Tb

= TTI,4+B3C3;

Td = TT3,4+B3D3;

A3 = (TT2,2X2- TT4,2)/~;

B3 = (TT2.4X2- TT4,4)/F2;

Noise reduction with perforated muffler components C3 = T T I , I + X I T T I , 3 ; D3 = TT3,1+X1TT3,3;

267

F2 = TT4.I "I-X1TT4.3 -- X2(TT2.1 + X I T T2.3). [ T T ] is an intermediate 4 x 4

matrix defined as
m

p2(O) poCowl (0)

pocow2 (0)

= [ T T]4x4

P3(/) poCow2(l)
O~cow2(l) _

with
TTl.1 = A1A2+ TI,2; TT1,3 = C1A2+ T1,5; TTe.1 = A1Be+ ire,2 ; TT2.3 = C1B2+ T2.5; TT3.1 = A~C2+ T4",; TT3.3 = C1C2 -t- T4.5; TT4.1 = AID2 + Ts.2; TT1,2 = BIA2 + T~.3; TTI.4 = DIA2+ Tl,6;

TT2.2 = B1B2+ T2,3;

TT2,4 = D1B2 + T2,6; TT3.2 = B1C2 + T4.3; TT3,4 = D1C2+ T4.6; TT4.2 = BID2+ T5.3;

T T4.3 = C1D2 + T5.5 ;

TT4. 4 = D1D2+ T5.6;

B1 = (T3,3X2 - T6,3)/F1;

A1 = (T3,2X2-T6,2)/F1; C1 = (T3.5X2- T6,5)/FI; A2 = TI.I+XIT1.4;

D1 = (T3.6X2- T6.6)/FI;

B2 = T2.1+XIT2.4;

C2 = T4.1+XIT4.4; D2 = Ts.l+X1T5.4;
F1 = T6,1 + X I T 6 , 4 - X 2 ( T 3 , 1 +X1T3,4);

x1 = i tan(klb);
and
X2 = - i tan

(kla).

Appendix C

The following relations hold only for three-duct perforated element reverse-flow
expansion c h a m b e r

7",, = Bl. IDl.l + BI.2D2.1+ BI.3D3.1,

Tt, = B~,IDI.2+ B~.2D2,2+ BI,3D3. 2, 7",. = B4,1Dt,I + B4,2D2,1 + Ba.3D3.1, Td = B4,1DI.2 + B4,2D2,~_+ B4,3D3.2,

268

K Narayana Rao and M L Munjal

Bil,i 2 = Til,i 2 + X 1 T/1,(i2+3),

i1=1,2 ..... i2 = 1 , 2 , 3 , )(1 = i tan(k/b).

Dl,1

6,

= Cl.lD2.1+Cl.3D3.1,

DI,z = C1,1D2,2 + C1,2D3,2,

1)2,1 = C3,2/F4; D3,1 = _C3,1/F4;
D2,2 = - C2,2/F4,

D3,2 = C2,1/F4,

F 4 = C2,1C3,2-C2,2C3,1, CI, 1 = ( B 5 , 2 - X 2 B 2 , i ) / F 3 ;
C1,2 = ( B s , 3 - X 2 B 2 , 3 ) / F 3 ,

C2,1 = B3,z+CI,IB3,1;
C3,1 -~ B6,2+Cl,lB6,1;

C2,2 = B3,3+Cl,2B3,~,
C3,2 = B6,3+CI,zB6,1,

F3 = B2,1X2-B5,1,
X2 = - i

tan (k la).

Appendix D. Notation
A1, Ao
Co

f
i k / M
Mi

go
Po P
t

temp
UI,2, U2,3

Wo
w

Y
z

po O
o)

~1, ~2

internal areas of inlet and exit pipes, respectively, velocity of wave propagation internal diameter of pipe frequency, iota, X/~-1, wave number, (tO~Co), length of pipe, Mach number, (Wo/co), inlet Mach number, exit Mach number, pressure of the undisturbed fluid, fluctuating pressure, time co-ordinate, temperature, radial fluctuating velocities at 1, 2 and 2, 3 interfaces of the control volume, respectively, velocity of the undisturbed fluid, fluctuating velocity, characteristic impedance, co~A, axial co-ordinate, density of undisturbed fluid, fluctuation in density, circular frequency, acoustical impedances at 1, 2 and 2, 3 interfaces, respectively.

Noise reduction with perforated muffler components

References

269

Munjal M L 1975 J. S o , n d Vib. 39: 1115-I19 Munjal M L 1986 Acousttcs o f duct,s attd mufflers (New York: John Wiley) (in print) Panickcr V B, Munjal M L 1981a J. India, ht~'t. SoL A63:1-19 Panicker V B, Munjal M L 1981b J. hulian hzst. Sci. A63:21-38 Rao K N 1984 Predictton and verlification ~f the aero-acoustic performance o f perfi)rated element ;m(l~flers. Ph.D. thesis, Indian Institute of Science, Bangalore

Rao K N, Munjal M L 1984 A gencralized decoupling method for analysing perforated element mufflers, Proceedings of the Nelson Acoustics Conference, Madison, USA Sullivan J W 1979a J. Acoust. Soc. Am. 66:772-778 Sullivan J W 1979b J. A~oust. Soc. Am. 66:779-788 Sullivan J W, Crocker M J 1978 J. Acoust, Soc. Am. 64:2(17-215 Thawani P T, Jayaraman K 1983 J. Acous7. Soc Am. 73:1387-1389