Flow Distribution Manifolds

R. A. BAJURA
Flow distribution in the lateral branches of dividing, combining, reverse, and parallel
Associate Professor, flow manifold systems is studied both analytically and experimentally. Predictions
Department of Mechanical
Engineering and Mechanics, for the flow rates and pressures in the headers of any of the above four basic manifold
West Virginia University, configurations are obtained from the solution of two first order differential equations
Morgantown, West Va. involving tlie flow rate and the pressure difference across headers (pressure-flow equa-
Mem. ASME
tion set), or by the solution of a second order, nonlinear ordinary differential equation
involving the flow rate alone (flow distribution equation). Experimental results are
E. H. JONES, JR. presented for various manifold designs having different lateral/header area ratios,
1st Lieutenant, USAF, lateral flct.o resistances, and length /diameter ratios. Good agreement is obtained be-
Williams AFB, Ariz. tween the analytical and experimental results. Dimensionless parameters which af-
Mem.ASME
fect flow distribution are identified and discussed with respect to the generalized coeffi-
cients of the analytical model. The present method of analysis is proposed for geneTfil
application in evaluating the performance of flow distribution systems.

Introduction flow velocity head is related to the pressure differential by a flow

(or discharge) coefficient which accounts for frictional losses along
A manifold is defined here as a flow channel for which fluid the lateral flow path.
enters or leaves through porous side walls due to the action of a If the flow field is considered as one dimensional, the governing
differential pressure. Manifolds commonly used in flow dis- equations for the manifold are the continuity and momentum
tribution systems can be classified into four categorical types, equations for each header and the discharge equation for the
namely, simple dividing or combining flow manifolds and paral- lateral flows. Under suitable assumptions, the work-energy
lel or reverse flow manifold systems. These manifolds are il- equation can also be applied to the header flow stream. The
lustrated by Fig. 1. The parallel and reverse flow systems are prediction of the performance of a manifold depends on the proper
combinations of the basic dividing and combining flow manifolds selection of the momentum exchange and the discharge coeffi-
interconnected by lateral branches. In a dividing flow header, cients for the given system and the formulation of a valid physical
the main fluid stream is decelerated due to the loss of fluid model for the branching process. The prediction of the lateral
through the laterals. Therefore, pressure will rise in the direction flows for a manifold with many branches is accomplished move
of flow if the effects of friction are small as can be demonstrated readily by a continuous flow model as opposed to a discrete
by applying a frictionless Bernoulli equation to the header flow branch point model. The objectives of this paper are to present
stream. Frictional effects, however, would cause a decrease of
pressure in the flow direction. Therefore, the possibility exists
for obtaining a uniform pressure along the dividing flow header
by suitable adjustment of the flow parameters so that the pres-
sure regain due to flow branching balances the pressure losses
due to friction. The combining flow header is characterized by a DIVIDING FLOW COMBINING FLOW
falling pressure in the direction of flow. This characteristic
occurs due to the additive effects of both the frictional pressure
losses and the favorable pressure gradient required for accelera-
TTTTT TTTTl
tion of the main stream due to the inflow at the branch points.
The movement of fluid through the porous wall is governed by a REVERSE FLOW PARALLEL FLOW
discharge equation for the crossflow stream in which the cross- ^Piwpiinw ^ , a ^ »

" " " '' '' '' ' 1 ' ' ' 1
Contributed by the Fluida Engineering Division of T H E AMERICAN SOCIETY
OF MECHANICAL ENGINEERS and presented at the Gas Turbine and Fluids
Engineering Conference, New Orleans, La., March 21-25, 1976. Manuscript
received at ASME Headquarters, January 5, 1976. Paper No. 76-FE-7. Fig. 1 Four types of manifolds

654 / DECEMBER 1976 Transactions of the ASME

Downloaded From: https://fluidsengineering.asmedigitalcollection.asme.org on 07/02/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Copyright © 1976 by ASME
 valid physical and analytical model applicable to a wide energy theorem (First Law), or the conservation of momentum
variety of manifold system designs and to illustrate the applica- theorem. In applying either the Bernoulli or work-energy equa-
tion of this model to a sj'stem of uniform cross-sectional dimen- tions to the branching process, it was argued t h a t the mechanical
sions. The validity of the analytical model is demonstrated for energy before branching should be equal to the mechanical
pXperimental systems studied by the present writers and other energy after branching plus some losses due to friction. However,
investigators. McNown [1] has shown that the mechanical energy after branch-
ing for the dividing flow header can be greater than the approach-
ing energy. This result occurs due to the rearrangement of fluid
Literature
between the boundary layer and the main stream as the branch
Coefficient Data. Earlier discrete branch point analytical point is traversed, such t h a t fluid of low kinetic energy is dis-
models predicted the performance of manifolds by determining charged into the lateral with higher kinetic energy fluid remain-
the discharge at an individual branch point in terms of some ing in the header. T h e apparent violation of the First Law of
assumed local pressure and flow conditions. The discharges in Thermodynamics (work-energy equation) can be explained by
each lateral were iteratively adjusted until the overall discharge noting t h a t the energies of the three fluid streams are calculated
matched the given flow rate for the system. Therefore, the early on a per unit mass basis, whereas the energy conservation theorem
experimental work concentrated on determining flow coefficients is based on the overall energy flow rates in the control volume.
at single, isolated branch points. The major contribution to the If the specific mechanical energies of each fluid stream are mul-
data for pressure changes and flow loss coefficients at discrete tiplied by the relevant mass flow rate terms, then the overall
branch points was made by McNown [1, 2] 1 for circular pipes mechanical energy of the two fluid streams leaving the dividing
with right angled, sharp-edged junctions between the lateral and flow branch point is shown to be less than the approaching energy
the header. Other experimental data have been reported for flow rate. The loss of mechanical energy is accounted for in the
these geometries by Zeisser [3], Starosolszky [4], Ruus [5], and gain in internal energy of the fluid due to viscous dissipation.
Kubo and TJeda [6, 7], Flow coefficients for other lateral geom- Therefore the overall work-energy equation is satisfied.
etries, such as simple holes or short tubes, have been determined
The difficulty with applying a Bernoulli equation to the
by Oakey [8], Koh and Brooks [9], Acrivos, Babcock and Pigford
branching process lies in the ambiguity which exists in identify-
[10], Keller [11], Dow [12], and Dittrich [13]. Much of the data
ing a relevant streamline on which to conserve energy and esti-
for flow coefficients has been obtained for branch points which
mate frictional losses. Other authors have avoided this question
are infinitely spaced along the header and, therefore, the relevance
by applying a momentum equation along the header. Models
of this data to situations where branch points are closely spaced
have been proposed by Enger and Levy [14], Van Der Hegge
is questionable. McNown [1] has shown t h a t the pressure regain
Zijnen [15], Markland [16], and Acrivos, Babcock and Pigford
characteristics at a branch point are strongly dependent on the
[10] which relate the pressure changes to the momentum changes
spacing between laterals. The geometry of the branch point
in the main flow stream and frictional losses which are based on
itself was shown by Zeisser [3] to have no effect on the pressure
the local flow speed. In the above flow models, the effects of
regain characteristics but to have a profound effect on the flow
axial momentum transport by the lateral fluid stream are not
losses incurred by the branch stream turning into the lateral
considered. In addition, these models have been applied only
from the header. The results of Kubo and Ueda [6] illustrate that
to the case of simple dividing flow manifolds.
the flow coefficients may be considered as independent of Rey-
nolds number for a wide flow range. In an earlier paper, Soucek and Zelnick [17] proposed a model
for discharge ports in a lock system which included the effect of
axial momentum transport by the transverse flow stream. This
Analytical Models. The overall analysis of the performance of
model was developed b y Bajura [18] and applied to both dividing
a manifold system is based primarily on the analytical model
and combining flow manifolds. The flow model proposed by
chosen to represent the branching process. Considering first t h e
these latter two references is more physically acceptable than
flow stream in the header, the pressure rise in dividing flow (or
the previous models since the overall momentum balance (in-
pressure decrease in combining flow) has been analyzed tradi-
tegral equation) is satisfied for the control volume as a whole
tionally by the application of the Bernoulli theorem, the work and remains valid independent of the effects of friction or the re-
arrangement of streamlines due to the branching process,

'Numbera in brackets designate References at end of paper. Calculational Procedures. Calculational models to evaluate the

•Nomenclature-
I = lateral length
A = area, friction coefficient M = momentum coefficient A = differential between parameters
Ar = area ratio N = number of branch points along 6 = overall momentum coefficient for
B = momentum coefficient header header flow
turning loss coefficient for com- n = friction factor exponent ir = perimeter.
CTC — p = density
bining flow P = pressure
turning loss coefficient for divid- Q = volume flow rate $ = friction term
CTD =
ing flow T = wall shear stress Subscripts and Superscripts
header diameter V = velocity 0 = maximum velocity condition
D =
lateral diameter, differentiation x = distance along header 1 = dividing flow header
d =
Moody friction factor Z = dimensionless discharge coeffi- 2 = combining flow header
/ =
lateral resistance coefficient in cient 3 = lateral
H «= velocity heads (3 = momentum coefficient for header x = axial flow
local flow loss coefficient in ve- flow y = transverse flow
K = locity heads 7 = momentum coefficient for lateral (~) = overbar, average value
header length flow (') = differentiation
L =
Journal
Downloaded From: of Fluids Engineering
https://fluidsengineering.asmedigitalcollection.asme.org DECEMBER
on 07/02/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use 1976 / 65S
 performance of a manifold system can be formulated from several dividing the axial length, £,, by the total number of branch
viewpoints. I t is often desired to design a system with balanced points, iVi. The lateral area at the branch point, A3i, is assumed
flow in each lateral flow stream. Such designs can be accom- constant along the header, and the branch points are uniformly
plished by altering the size of the laterals, their flow resistance, distributed. The fluid is assumed to be incompressible and the
or the cross-sectional area of the duct. These systems have been header area, Ah is assumed constant. The flow conditions at
described by Howland [19] for round pipes, Perlmutter [20] and (xi -f A.Ti) are related to the flow conditions by xi by a first
Mardon, et al. [21] for tapered manifolds, Haerter [22] for air order Taylor series expansion. For the control volume of Fig. 2
conditioning systems, and Koh and Brooks [9] for ocean outfalls. the continuity equation is written as: • •
None of these models considers the loss of axial momentum from
the control volume due to the transverse flow. Other computa- 7 3 iA 3 i = — Ai
tional models approach the design problem from the standpoint dxi iVi (1)
of analyzing the performance of a given system. System models
of this type are described by Horlock [23] for slotted pipes, Olson where V3i and Vi are average velocities. Letting P represent the
[24], Huang and Yu [25], and Quaile and Levy [26] for porous pressure and T„ the wall shear stress, the momentum equation
ducts in laminar flow, and Bajura, LeRose and Williams [27] for in vector form is written for the control volume as:
manifold systems.
In view of the widely scattered values of flow coefficients and
- fpdA + f T m dA = f pV (VdA). (2
the different system geometries, it is clearly recognized t h a t each Surface Surface
manifold design must be evaluated based on its own charac- I t is desirable to express the momentum transport in terms of
teristics. However, a generalized method of analysis is required local average velocities. The following parameters are defined:
which can be applied to widely different manifold designs and /3i is an axial flow momentum correction factor, yi is a lateral
different computational viewpoints. T h e flow model described in flow momentum correction factor for axial momentum transport
the present document satisfies this need and is proposed as a through surface An, di is an overall momentum correction factor,
general model for manifold analysis. 7Ti is the perimeter of the header, 2\ is the wall shear stress, /,
is the Moody friction factor, and Pi is the pressure. These pa-
Analytical Model for Manifold Flow rameters are formally defined as:

In presenting the analytical model, the authors have chosen a

particular design relevant to superheater systems to illustrate
the development of the governing equations. In this section, the
A = (1/VM0 f
1/ Ai
VHAi)dAi (3)

system model is first defined for discrete branch points and is

then applied to parallel and reverse flow superheater systems. 7i = {1/ViVnAn) \ VAAnWyUnWAn (4)
<-/ An
^31
The system equations are later developed in a generalized, non-
dimensional form which can be applied to each of the four t?i = 2/3i - 7, (5)
manifolds identified by Fig. 1 by properly defining the relevant
parametric groups for the given design. Ti = /ipFiVS (6)

Basic Equations. Consider first the control volume illustrated The momentum equation for the dividing flow control volume
by Fig. 2 which describes the flow streams near a dividing flow in terms of the above parameters is:
branch point. The outflow of fluid a t surface An has velocity cl
components Vx and Vy since it is assumed t h a t the discharged I
p dxi
l} + (I ^
\fi + f,y, +e ivif^ (7)
fluid has not turned completely 90 degrees when crossing the p dxi \8Ai dxi I dxi
boundary o f ' t h e control volume. The length of the control Fig. 3 illustrates control volumes for the analysis of com-
volume in the streamwise direction is Axh which is calculated by bining flow headers. Fig. 3(a) shows the flow direction pertinent
to a reverse flow manifold system. Using an analysis similar to
the dividing flow header above, the momentum equation is:

dPi dVj.
VJ + 0272 = 0 (8)
dxi + l^ 8A2 dxi ) dxi
ix, = L^N,—J
where 02 is defined as (2/32 — 72) in an analogous fashion with
equation (5). The sign of the friction term is negative in equation
(8) since the flow direction is defined as being positive in the
negative x2 direction. Fig. 3(6) illustrates the control volume
I[-*•- V* I for the combining flow header of a parallel flow manifold system.
(A,) -j B^VJIX, + AX,)
The momentum equation for this configuration is:

Vx I dP1 (far* , dp\ \

+ n + 02F2 ^ = 0 (9)
dxi \8Ai dxt J <xx2
At this stage in the analysis, there are four unknowns, namely,
the pressures Pi and P 2 , and the velocities Vi and F 2 . The

© velocities are related by inter-manifold continuity equations

which ensure t h a t the flow from one header enters the other.
These inter-manifold continuity equations are:

Vt = 7I(AIMJ) (10)

for the reverse flow manifold system, and

Fig. 2 Dividing flow branch point control volume V, = (7io - 7i)(AiM2) (ID

Downloaded From: 656 / D E C E M B E R 1976

https://fluidsengineering.asmedigitalcollection.asme.org Transactions
on 07/02/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use of the ASME
 ' (Al REVERSE FLOW

Avv
V2(Xj) -« [_ V2(X2 + AX2I J [ "2

H
H
(B) PARALLEL FLOW

Fig, 4 Control volume for lateral discharge flow

7
^

- A X 2 = L2/N2 -
in the dividing flow header. I t is assumed t h a t the lateral fluid
enters the combining flow header as a stream with velocity F32
Fig. 3 Combining flow branch point control volume and not the velocity K2. Any mixing effects are accounted for
in the turning loss term CTC- The lateral flow resistance term,
H, is defined in terms of equation (12) as:

Pi - Pi APn TV H / Ai d?! U V
= H (13)
P 2 \ An dx! Ni )
for the parallel flow manifold system. The term VK is the
average velocity a t the inlet to the dividing flow header. T h e where AP12 is the differential pressure between headers. When
relationships defined by equations (10) and (11) are valid at the overall flow resistance, H, becomes large, the turning loss
correspondingly wealed distances Xi and x2 for each header. By terms assume minor significance in the discharge equation.
employing the inter-manifold continuity relationships, the ve- Equation (13) may be manipulated to obtain several relation-
locity Vi may be eliminated from the equations, leaving only the ships between the differential pressure and flow rate in the divid-
unknowns of Pi, P 2 , and F,. ing flow header. T h e first relationship is obtained by solving
A relationship between the pressure differential between equations (1) and (13) for the velocity gradient, dVi/dxi, as:
manifolds and the lateral flow rate is obtained in terms of a dis-
charge equation written in the following manner with respect to
the control volume illustrated by Fig. 4: £ - - £r (£)'"<ift"" (14)

Pi The minus sign is chosen since the velocity Vi must decrease with
- = ^ . Zs!
P p 2 \A3, ) distance Xi. This condition also requires that the pressure dif-
ferential always be positive or a reversal of flow will occur. (This
is not to say t h a t a reversal of flow cannot occur in actual sys-
CTD + Keq + (fl/d)eq + CTC (12)
W) tems.) T h e second relationship between the differential pres-
sure and the flow rate is obtained by differentiating equation
The term CTD is a turning loss for flow entering the lateral from (14) and solving for the differential pressure gradient. This re-
the dividing flow header; the terms K,q and (fl/d)eq represent sult is:
the equivalent loss coefficient (based on the velocity Vu) for
local upset flow losses and ordinary friction losses; and the term 1 d(AP:
(15)
CTC is a turning loss for flow into the combining header. The P dxi \A31N1 J dxx dxi2
area ratio term, (A31/A32) is an adjustment to allow for the pos-
sibility t h a t the area of the lateral changes from the inlet end to Equation (15) will be used later to eliminate the pressure terms
the outlet end. By writing the discharge relation as in equation from the governing equations.
(12), no credit is taken for any approaching velocity head (Fi 2 /2) Nondimensional Equations for Manifolds. The relevant equa-

Journal of Fluids Engineering DECEMBER 1976 / 657

Downloaded From: https://fluidsengineering.asmedigitalcollection.asme.org on 07/02/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
 tions for the analysis of flow in manifolds may be nondimen- manner similar to the reverse flow system above, the differential
sionalized by the following reference quantities: Fio, the entrance pressure is obtained by subtracting equation (9) from equation
velocity to the dividing flow header; L, and L-i, the lengths of (7); the inter-manifold continuity equation (11) is used to elimi-
each header; and Qo, the entering flow rate, F10A1. T h e nondi- nate the velocity F 2 ; and the remaining equations are nondimen-
mensional variables are defined in Table 1. sionalized with respect to the variables in Table 1. The pressure-
Table 1 Nondimensional variables
flow equation set which results is:

x = Xi/Li = X2/L2 Q = FiA,/Q„ dV_ NiAn

(2/#)i'?(AP)i« (19)
v = 7,/y,o AP = APn/pVit? dx At
rf(AP)
Reverse Flow Manifolds. I t is convenient to focus the develop- dx 8A1 + dx 8A2 dx
ment of the governing equations on each manifold system inde-
pendently. Consider the reverse flow manifold system which is dV /Ai V rfF
dl
(1 - F ? - B,V (1 - F).(20)
governed by equations (7, 8, 10, and 14). As a first step, subtract dx
equation (8) from equation (7) to eliminate the individual pres-
sures and obtain an equation in terms of the differential pres- These equations are subject to the boundary conditions F(0)
sure, AP12 = (Pi — P2), between the headers. The velocity = 1 and F ( l ) = 0. If the pressure is eliminated from equations
F> is eliminated by employing the continuity equation (10). (19) and (20), the resulting flow distribution equation is:
After nondimensionalizing the remaining equations as defined
by Table 1, a set of two dimensionless equations involving the
AR2
velocity F and the differential pressure A P is obtained in the
form:
(1 - Q)>
^ 8A2 dx J
dV_ _ N1A31
(2/ff)i'2(AP)> (16)
dx Al + d&Q' + 02(Ai/A 2 ) 2 Q'(l - Q) = 0- (21)

d(AP) 'flLiTTi fJ*n / A V dft _ dfc 1 This equation is solved subject to the boundary conditions
F2 Q(0) = 1 and Q(l) = 0.
dx 8A1 8A2 \At J dx dx J
Generalized Equations for Manifold Systems. The previous de-
Ai V velopments were presented to acquaint the reader with the basic
(17)
A, dx flow model and methods of analysis. The flow distribution in
any of the four types of manifolds shown by Fig. 1 can be ob-
Equations (16) and (17) constitute a coupled set of first order tained from the solution of a generalized set of equations given
equations called the Pressure-Flow Equations which must satisfy by the following forms:
the following boundary conditions. At the entrance, the dimen-
sionless velocity V must equal 1. At the dead end of the mani- (i) Pressure-flow equations
fold, the end wall imposes the physical requirement t h a t the
dV _
velocity F is 0. This condition requires all the flow to be dis- ZiAP)11* (22)
charged before the dead end of the manifold is reached. While dx
it is possible to prescribe a pressure boundary condition at the d[AP) .dV dV
inlet, the inlet pressure cannot be specified arbitrarily since the A1F2 + A 2 (l - Vf - HiV Ih(l - V)
dx dx dx
entrance pressure level controls the discharge from the header.
Therefore, the pressure a t the inlet is intimately tied in with the (23)
continuity equation and cannot be specified arbitrarily or the con-
dition of too much or too little discharge will result. T h e spec- (ii) Flow distribution equation
ification of the inlet pressure is equivalent to specifying the
Q'Q" + $iQ'2 + 2$ 2 Q + M\i 2' + M-iQ' = $2 (24)
derivative, dV/dx, at the inlet [see equation (16)] and would
amount to an over-specification of the problem since the bound- The definition of the coefficients for the pressure-flow equations
ary values F(0) and F ( l ) , and the derivative dV/dx(0) cannot is given in Table 2. The coefficients of the flow distribution
be specified arbitrarily for a problem which is only of second order. equation are given in Table 3. Both sets of generalized equations
If the differential pressure is eliminated from, equations (16) satisfy the boundary conditions:
and (17), a second order equation in the dimensionless flow rate
Q can be obtained as: 7(0) = Q(0) = 1 (25)

H 'dtk dp 2 7(1) = Q(i) = 0 (26)

Q'Q" Q2 +
As2 dx da 8A1
The above equations have been formulated under the assumption
that the lateral resistance, H, and the distribution of laterals along

+ 8A2 T\ <22 + 0i - ej AAi2 yV

The term A, is called the area ratio of the manifold and is de-
QQ' = 0. (18) the headers are constant, i.e., the porosity of the headers is
constant. If these parameters vary along the headers, additional
terms appear in the flow distribution equation. A discussion of
these considerations is given in reference [27].
fined as Ar = N1A31/A1. Physically, this term is the ratio of the
total lateral cross-sectional area to the cross-sectional area of the For many manifold systems, the flow regime is likely to be
header. fully turbulent. For these conditions, the friction factors may be
taken as a constant. For smooth surfaces and widely spaced
Equation (18) is a flow distribution equation and is subject
branch points, the friction factor may be a function of the
Io the boundary conditions: (2(0) = 1 and Q(l) = 0.
Reynolds number and will vary along the headers due to the
Parallel Flow Manifolds. The parallel flow manifold system is changing flow rates. Under these conditions, the friction factors
described by the dimensional equations (7, 9, 11, and 14). In a may be assumed to vary as the Reynolds number to a power,

658 / DECEMBER 1976

Downloaded From: https://fluidsengineering.asmedigitalcollection.asme.org on 07/02/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Transactions of the ASME
 TABU 2 COEFFICIENTS FOR PRESSURE-FLOW EQUATION SET headers by glueing commercial pvc saddles to the headers over
holes bored through the side walls. The junction was sharp-
Dividing Combining
Flow Flow Parallel Flow edged and was shaped to have a smooth internal diameter with
V 2 F no roughness to disturb the lateral flows. Individual sections of
T
the main headers were joined using commercial pvc couplers
p
» 10 "in
which presented a negligible discontinuity to the internal surface
II of the header. The spacing between branch points was uniform at
2.55 header diameters. Tests were conducted with either 20 or
10 branch points giving header length-to-diameter ratios of
51.0 and 25.5 respectively. The lateral tubes were of uniform
2D,
diameter and 1.56m (5 ft) in length. Orifices of various internal
diameters (0.96, 2.54 and 3.18 cm) were inserted at the midpoint
of the laterals to increase the &ow resistance.
Each header was instrumented by pressure taps located mid-
way between the branch points and on the opposite side of the
header. The static pressures in each header were referenced to
the static pressure at either the open or closed end of the header,
depending upon the configuration studied as shown by Fig. 5.
/r Differential pressures between headers were measured at the
outlet from the combining flow header. The overall flow rate at
the entrance to the dividing flow header was measured by a six
probe velocity rake using equal area averaging procedures. Flow
such as ( — 1/4). If the frictional exponent is given the symbol, rates were measured in two laterals using calibrated resistance
ii, then the friction factors can be replaced by the expressions: orifices to monitor the overall flow conditions in the laterals.
Reynolds numbers in the headers were of the order of 60,000-
Mx) = / , . y - = /„ (27) 80,000 and 8,000-10,000 in the laterals, depending on the mani-
folds studied. Data for each run were reduced by computer
Mx) = fwV f*>Q~ (28) analysis and polynomial curves were fitted to the raw data using
Where the (0) subscript indicates that the friction factor is a statistical analysis program for use in interpolating routines
evaluated at the maximum Reynolds number for the header. required to compare the experimental data with the analytical
The friction factor is now a function of the flow variable and model. The data presented here represents the average of three
Mutable alteration must be made in the dependent variable terms runs per test case. A more expanded description of the experi-
related to friction for the governing equations. mental facility and test conditions is given in Jones [28].

Experimental Apparatus Experimental Results

A schematic of the experimental apparatus is shown by Fig. Five configurations were studied experimentally for both
.">. The headers and laterals were fabricated from commercial parallel and reverse flow manifold systems. The parameters
pvc piping materials having diameters of 10.16 cm (4 in.) and varied were the lateral flow resistance, H, and the relative length,
3.81 cm (1.5 in.), respectively. The air flow through the system L/D, of the manifold. The number of branch points per manifold
was induced by connecting the combining flow header outlet to was either 10 or 20 depending upon the length of the headers
the inlet of a 3.73 kilowatt (5 horsepower) blower. Parallel and since the spacing between branch points was maintained con-
reverse flow systems were arranged by capping one of the ends stant at 2.55 header diameters. One set of experimental runs was
of the dividing flow header. Laterals were attached to the carried out for the case of infinite lateral resistance (i.e., constant

TABLE 3 COEFFICIENTS FOR FLOW DISTRIBUTION EQUATION

Dividing Combining
Parameter Flow Flow f everse Flow P a r a l l e l Flow

2
\2 ¥2
2
\ ¥, A
r
f
lh +
f L
2 2/ D lV A2
r
\t
1L1 f
2L2/")l\4
•i H 2D1 H 2Dj H 2D, 202 [DJ H 2D
l ~ 2D 2 W

(
2
0 0 0 A2 f2L2 /DA4
r
H 2o2 \r2)

M
l A2
"
r
6ft
1
-A2
_r_ .
H 9
2
A2
r
H ;, - »zfi) _ H r^fiT
M2 0 0 0 A
r
1
-ft)'
2
V
20

J 2

Journal of Fluids Engineering DECEMBER 1976 / 659

Downloaded From: https://fluidsengineering.asmedigitalcollection.asme.org on 07/02/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
 ;
HEADER REFERENCE
MANOMETER PRESSURE TAP
TUBES

REVERSE FLOW
10,16 CM D I A M ., P A R A L L E L FLOW _
I N L E T (x = 0) "
INLET (x = 0)

- 3 . 8 1 CM D I A M

HEADER
DIFFERENTIAL
PRESSURE RESISTANCE ORIFICES
TAPS

•DENOTES PRESSURE
TAP LOCATION

END
-TO BLOWER INLET- 10.16 CM D I A M
CAP

HEADER REFERENCE
PRESSURE TAP

Fig. 5 S c h e m a t i c of e x p e r i m e n t a l a p p a r a t u s

pressure headers) to obtain uniform discharge flows in the lat- Case B—Large A r e a Ratio a n d S m a l l L a t e r a l Resistance M a n i f o l d .
erals. The flow equations were then transposed by straight- The headers of the system described by Case A were separated
forward analytical methods (since the header velocity is now a and an orifice of diameter 3.18 cm was inserted in the laterals,
linear function of distance) to obtain values of the momentum giving a reduced flow resistance of 4.5 lateral velocity heads. The
coefficients and effective friction factors for each header. The
following coefficient data was obtained from the uniform flow
tests:

9, = 1.05 ± 0.05 (29) REVERSE FLOW

n - _ n P • q •
0s = 2.60 ± 0.05 (30)

This result is in agreement with the experimental values pre- O o u

O
dicted in references [18 and 27]. The friction factors were in o o „"_
agreement with the Moody charts for smooth tubes using the
Blasius correlation for which the friction factor varies with
Reynolds number to the (—1/4) power. The effects of the lateral
penetrations through the walls of the headers in increasing the <y A..
.' A'--.
the friction factor could not be detected from the data. A
o/ A A"-A---
A "A""A-A---A—A

Case A — L a r g e A r e a R a t i o a n d L a r g e L a t e r a l R e s i s t a n c e M a n i f o l d .
y
• DIVIDIN G FLOW MEADtH

The results of experiments for systems with an area ratio, Ar, 7 o COMBIN NG FLO '1 HEADER

A- NTIAL ETWEEN

of 2.810 and a lateral flow resistance, H, of 12.2 lateral velocity HEADER

heads (2.54 cm diameter orifice) are presented in Figs. 6(a) and ,1ENStONLESS DISTANCE X

6(b). There are 20 branch points along each header. The ordinate
for each figure is the dimensionless distance, Xi/Li, from the en-
trance of the dividing flow header. Three data sets are shown on
each figure. The pressures in each header are plotted as multiples
of a reference pressure, AP r , which is defined as the pressure dif-
ferential between the inlet to the dividing flow header and the .•-D-D-D-D-0
outlet from the combining flow header. For parallel flow,
APr = Pi(0) - JPj(l). For reverse flow, APr = P\(0) - P s (0). a—o—tr-
io - ~ a
The third data set illustrates the local pressure difference between
headers as normalized with respect to the maximum pressure dif- <L?._2_° o 0 o
-.o o
ference between headers, (Pi(x) - P2(z))mi.z. The third curve
can be interpreted as a measure of the relative discharge flow
since the flow rate in the laterals is proportional to the square
A
root of the differential pressure between headers.
Xo
Since the dividing flow header is dominated by the effects of N>
static pressure regain due to branching (i.e., small friction ef-
fects), the pressure characteristics of each header are better
matched to provide a more uniform discharge for the reverse \
flow system than for the parallel flow system. Note that the IMENSIOM.ESS DISTANCE X

only difference between these two manifolds is the orientation of

Fig. 6 Pressure profiles f o r m a n i f o l d s with a large a r e a r a t i o a n d
the outlet. large l a t e r a l r e s i s t a n c e

660 / DECEMBER 1976 Transactions of the ASME

Downloaded From: https://fluidsengineering.asmedigitalcollection.asme.org on 07/02/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
 a D I V I D I N G FLOW HEADER
D a
D Q D n n ° n D ' a O COMBINING FLOW HEADER

o A DIFFERENTIAL BETWEEN
HEADERS
-o—-—
O O'

(A) REVERSE FLOW

^A • '

• DIVIDING FLOW HEADER __ a _ •__•_—-a—c

- \ / 0
A
COMBINING FLOW HEADER
DIFFERENTIAL BETWEEN

ft "A"
A A"-A--r
: / ' % A
A A A A

; / "A.
o ~"A- - A_
-A. A .»-'
.-- 0 ''
1 1 i l 1 1

DIMENSIONLESS DISTANCE X./L.

D a_ a—D—a IB) P A R A L L E L FLOW

• D-
D4—a—
• n n D n—a~ • •—a—•—a—a

1 D
O
— DIVIDING LOW HEADER
COMBINING FLOW HEADER
\ / „ ..„"A A
A DIFFERENT AL BETWEEN
| HEADERS 1 o o o
i—..° ° o
A <A O
-*--« "~~Q-
"°-^
_A__A—A — A - - A - " DIMENSIONLESS OISTANCE X^L,

DIMENSIONLESS DISTANCE X , / L ( Fig. 8 Pressure profiles for manifolds with a small area ratio and
large lateral resistance
Fig. 7 Pressure profiles for manifolds with a large area ratio and
small lateral resistance

allowed to vary with Reynolds number and were calculated based

data for these experimental runs are presented in Figs. 7(a) and on the smooth tube correlation. The exponent, n, used for fric-
7(b). The reduction of the lateral flow resistance brings about tion factor variations was ( — 1/4). Second, allowances were
a poorer flow distribution for each system: however, the reverse made for variations in the momentum coefficient di to account
flow manifold still exhibits a better flow balance due to the for a readjustment of the velocity profile in the neighborhood of
matched pressure characteristics of each header. For the parallel the first few laterals in the dividing flow header. From the data
flow manifold, the differential pressure near x = 0 is almost zero of references [18 and 27], the value of 0i for laterals with infinite
and little discharge occurs in this region. spacing is taken as 1.30 and is reduced to 1.05 as the spacing
between laterals is decreased. The computations were carried
Case C—Small Area Ratio and Large Lateral Resistance Mani- out with a linear variation of 8i from 1.30 to 1.05 over the di-
fold. The length of the headers described above was reduced by mensionless distance, x, from 0.0 to 0.2. The value of B\ was as-
50 percent and tests were conducted for systems with 10 laterals. sumed constant for the remainder of the manifold. Third, a
Case C represents a system with an area ratio of 1.405 and a value of dfii/dx of ( — 0.3) was included in the computations for
lateral resistance of 12.2. This system is identical to Case A ex- the dimensionless distance, x, from 0.0 to 0.2. It was observed
cept for the shortened headers. The data are presented on Figs. that the pressure data for the dividing flow header suffered an
8(a) and 8(b). The flow distribution is nearly uniform due to abrupt increase near the entrance and then slowly increased
the large flow resistance of the laterals and the small area ratio. toward the dead end of the header. (Refer to Fig. 6(a), for ex-

Case D~Small Area Ratio and Small Lateral Resistance Manifold.

The manifold of Case C was altered to obtain a flow resistance of
TABLE 4 COEFFICIENTS OF THE PRESSURE-FLOW
4.5 velocity leads in the laterals by the insertion of the larger EQUATION SET FOR FIGURES 6 - 9

orifice. The data for these experiments are presented in Figs.

9(a) and 9(6). The flow distribution is poorer than Case C due to pure !!l
A,
h. h. 1

the smaller lateral resistance. 6A 0.81 0 -1.55 0 1.14

6B 0.40 0.40 1.05 2.60 1.14

Comparison With Analytical Model. The curves on Figs. 6

0 1.88
through 9 illustrate the pressure profiles predicted by the 7A 0.81 0 -1.55
1.05 2.60 1.88
7B 0.40 0.40
analytical model. Table 4 lists the values of the parametric coef-
ficients of the pressure-flow equations for each system. All curves 8A 0.40 0 -1.55 0 0.57

are normalized with respect to the reference pressure conditions 8B 0.20 0.20 1.05 2.60 0.57

described above. In programming the computer solutions, the 9A 0.40 0 -1.55 0 0.94
following assumptions were made. First, the friction factors were 9B 0.20 0.20 1.05 2.60 0.94

Journal of Fluids Engineering

Downloaded From: https://fluidsengineering.asmedigitalcollection.asme.org on 07/02/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
DECEMBER 1976 / 661
 o DIVIDING FLOW HEADER
O- .. ~ COMBINING FLOW HEADER
.!J.~ ~ -~ DIFFERENTIAL BE1WEEN
HEA[JERS

6,......... 0 0 0 0
", '__ 0 ~"-"--'
- _ 0 9.'-" .
E. ",
,/'
s..<-
",,,, 2----_
0., ", ", "'--"'-Z--'"
V
o )Y" o DIVIDING fLOW HEADER
Q-··-COMBINING FLOW HEADER
6-----DlFFERENTIAL BETWEEN
HEADERS

.2 .4 .6
DIMENSIONLESS DISTANCE X,/Ll

1 . 5 ,_ _ _ _ _ _ _ _ _ _ _ _ _ _----,

(B) PARALLEL FLOW

.8 1.0

~ -
_ _--:'-_ _ _ _-'-_ _ _ _.L.._ _ _......JI..

DIMENSIONLESS DISTANCE X 1/L 1

Fig. 10 Pressure protiles for the last 80 percent of a reverse flow

~ ~

manifold with large area ratio and large lateral resistance (reference
Flg.6(a»

decay of the entrance disturbance indicated that a value of -0.2

should be used for df3t/dx; however, the value of -0.3 above pro-
vided the best fit to the experimental data.
The above entrance conditions are relevant only to the presene
experimental configuration. Therefore, the adjustments madt
.2 .6 .8
DIMENSIONLESS DISTANCE X 1/L 1
1.0 in the momentmn coefficient {3, are not considered to be of gen-
eral applicability. The complications of the distorted entrance
Fig. 9 Pressure profiles for manifolds with a small area ratio and velocity profile can be removed from the experimental data,
small lateral resistance
however, by analyzing only the portion of the manifold down-
stream from the entrance where the velocity profile has adjusted
to a fully developed condition. Computer solutions for each of
the manifolds described in Figs. 6 through 9 were obtained for
a portion of the original manifolds corresponding to the distance
ample.) This increase in pressure at the entrance to the dividing interval from x = 0.2 to the end of the header (i.e., the last 80
flow header was e,ttributed to a nonsymmetrical velocity pro- percent of the headers). In these computations, the fully de-
file present in the neighborhood of the first few laterals. The veloped value of (J, = 1.05 was used and the momentum term
velocity profile distortion resulted from the sum total of the fol- d{3t/dx was taken to be zero. The agreement of the analytical
lowing effects. ~ir drawn through the manifold system by the model for fully developed flow with the shortened manifolds is
blower was recirculated in the laboratory to maintain a constant illustrated by Figs. 10 and 11. Fig. 10 is related to Fig. 6(1l) and
a.ir temperature. These recirculation currents from the blower Fig. 11 is related to Fig. 7(b). The agreement between the ana-
exhaust air would cause random vorticity to be present in the lytical solutions and the experimental data is good. Fig. 12 is
intake air stream of the dividing flow header, which consisted
of only a sharp-edged, short length of pipe with no other ductwork
or calming sections. (Limitations on the laboratory space avail-
able prohibited the construction of a smoother entrance section.)
Distortion in the entrance flow was verified by velocity rake pro- I.'
files; which were nonsymmetrical at the measuring cross-section.
This initial distortion was further accentuated by the tygon
tubing behind the velocity rake support which was used to
transmit the total pressure to a manometer tube bank. While
distortion in the entrance velocity profile is undesirable from an
academic viewpoint, nonuniform entrance conditions are typical-
ly found in many industrial manifold designs. Therefore, the ( ) _ DIVIDING FLOW HEADER
present experiments may be viewed as a realistic test of the 0-··- CQM8lNING FLOW HEADER
D--- OIFFERENTIAL BETWEEN
analytical model. HEADERS

Computer solutions were obtained for various values of the

friction factor, the momentum coefficients (It and (J2 and the
velocity profile adjustment factor df3t/dx. The comp~ter solu-
tions could not be adjusted to account for the jump in pressure
at the entrance to the dividing flow header without the inclusion
of the d{3t!dx term, even for wide ranges in the values of the other 100
.25 .50 .75
momentum coefficients and the friction factors. It was, therefore, DIMENSIONLESS DISTANCE X j /L 1

concluded that the adjustment of the entrance velocity profile

Fig. 11 Pressure profiles for the last 80 percent of a parallel flow
to a fully developed condition is a significant factor affecting the manifold with large area ratio and small lateral resistance (reference
flow distribution in a manifold. Analytical calculations for the Fig. 7(b»

662 / 0 E C E M B E R 1976 Transactions of the ASME

Downloaded From: https://fluidsengineering.asmedigitalcollection.asme.org on 07/02/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 taken from reference [18] and illustrates the application of the signed with lateral/header diameter ratios near 0.15 where Q2 is
fully developed flow concept described here to the last 16 laterals neai-ly constant. Therefore the momentum parameters #i and
(out of 20) for a simple dividing flow manifold. 02 cannot be considered as independent variables in a broad sense.
No adjustment was applied to the combining flow headers to Diameter Ratios. The diameter ratio between headers in a
account for variations in the momentum coefficient {i2 since the reverse or parallel flow manifold will affect both the friction and
velocity is nearly zero at stations in the combining flow header momentum coefficients of the flow distribution equation as shown
where changes in fi2 would be greatest. Hence, the correction by Table 3. The momentum coefficients are more important in
would contribute negligible pressure change (locally) in the com- determining flow distribution than the friction coefficients.
puter solutions. This procedure is supported by the agreement of Therefore, a better flow distribution can be obtained by reducing
the predictions of the analytical model with the data for a simple the magnitude of the momentum coefficients by increasing the
combining flow header taken from reference [18] as showr >n diameter of the combining flow header to offset the large value
Fig. 13- The value of 62 = 2.66 used in the calculation was taken of the momentum parameter 62. Systems in which the dividing
from the data of references [4 and 18]. flow header is larger than the combining flow header will exhibit
a poor distribution of flow in the laterals. The effect of varia-
Discussion tions in the header diameter ratio on the friction terms is not as
great in altering the flow distribution. Changes in diameter ratio
The data and analytical model presented above were developed between the header and the laterals will not affect the flow dis-
for the case of a system with uniform areas and lateral resistances. tribution directly except through minor changes in the values of
Some general conclusions relevant to the performance of mani- t h e lateral flow resistance coefficients, CTC and CTD, and the
fold systems can be illustrated with respect to these examples. momentum term d2.

Parameters Affecting Flow Distribution. The pressure-flow Momentum Parameter /?. The momentum parameter /3 is nec-
equations and the flow distribution equation were presented in essary to account for variations in the velocity profile at the
terms of dimensionless coefficients which were functions of many entrance to the manifold. The upstream history of the flow ap-
independent variables. These equations show that uniform flow proaching the manifold and the design of the entrance (i.e.,
distribution in the laterals is attained only when the headers act parallel or right angled impingement) determine the amount
as infinite reservoirs. The infinite reservoir condition is ap- of velocity profile distortion present in a given system. Therefore
proached when the dimensionless coefficients of the flow dis- the momentum parameter j3 cannot be generalized. As illustrated
tribution equation approach zero. T h e individual variables com- by the experiments reported in the present paper, one of the major
prising these coefficients affect the flow distribution in the follow- factors contributing to the uncertainty in predicting the perform-
ing manner. ance of a manifold is the entrance flow condition.
I t has often been observed t h a t a reversal of flow occurs in
Area i.atio, T h e square of the area ratio, Ar2, appears in the manifolds designed with out-of-plane bends leading to the
numerator of each term in the flow distribution equation. Hence, entrance of the dividing flow header. Under these conditions,
a large area ratio contributes to flow maldistribution in the mani- fluid can recirculate from the combining flow header into the
fold system since the value of the dimensionless coefficients is dividing flow header through the laterals nearest the entrance.
rapidly increased as the area ratio increases. A common design This effect has been demonstrated by Sherman [18] and LeRose
rule-of-thumb is to limit the area ratio to values less than one. (private communication) for similar entrance conditions. T h e
The area ratio is also viewed as the porosity of the header. momentum correction term dft/dx in the governing equations is
the only mechanism by which the flow reversal phenomenon can
Lateral Flow Resistance. The lateral flow resistance, H, ap- be explained analytically. For these poorly designed inlet condi-
pears in the denominator of each term in the flow distribution tions, the entrance pressure can be less than the pressure in the
equation. Therefore, the value of the dimensionless coefficients combining flow header of a parallel flow system. Therefore, fluid
decreases as the lateral flow resistance increases. An infinite will be drawn into the dividing flow header. A pressure recovery
flow resistance would cause even a small diameter header with will occur in the direction of flow due to the regain of pressure
large porosity to act as an infinite reservoir. Hence, large lateral from the distorted entrance profile. Otherwise, the pressure is
resistance is desirable for good flow distribution. However, large required to fall in the direction of flow as can be shown by a
literal resistances result in a high total pressure loss for the simple analysis of the governing equations. The experimental
manifold system which may be unacceptable if pumping costs parallel flow systems reported above (Figs. 6(6) and 7(6))
are an important design consideration. could have exhibited a flow reversal in the first few laterals
had the entrance to the dividing flow header been through a
Length/Diameter Ratio. The relative length of a header enters right angled bend rather than a straight inlet section. Flow
the governing equations only through the friction terms. For reversal is more likely to occur in parallel flow manifolds which
headers of relatively small length/diameter ratio, the effects of are subject to poor flow distribution since the differential pres-
friction may be neglected and the flow distribution equation can sure between headers is minimal at the entrance for these de-
then be solved analytically. For relatively long headers, the ef- signs.
fects of flow branching on the static pressure in the header can
be neglected. Some analytical solutions for the flow distribution Friction Factor. The selection of a particular pipe material or
are also possible for friction dominated manifolds. Analytical surface finish will affect the value of the dimensionless friction
solutions are discussed in references [18 and 27]. T h e effects of coefficients in the governing equations. The present experimental
header length/diameter ratio must be separated from the area results indicate that the friction factor can be calculated under
ratio effect for headers of constant porosity per unit length. the assumption that the branch points do not affect the friction
pressure loss characteristics of the header. This conclusion is
Momentum Parameters di and 62. The momentum parameters highly dependent on the spacing between laterals. Common de-
8i and 92 are relatively fixed. The fully developed flow value of di sign practice has been to evaluate the friction factor based on
is approximately 1.05 and is only weakly dependent on the ratio ordinary pipe friction calculations for the case of widely spaced
of diameters between the header and the laterals as shown by the branch points and to increase the value of the friction factor as
data of various investigators [29]. The value of d2 becomes the branch points become closely spaced. The present data in-
highly variable (see reference [27]) as the lateral/header diameter dicate t h a t a spacing of 6.8 lateral diameters may be taken as a
ratio increases to values of 0.5. However, most systems are de- case representative of widely spaced laterals.

Journal of Fluids Engineering DECEMBER 1976 / 663

Downloaded From: https://fluidsengineering.asmedigitalcollection.asme.org on 07/02/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
 Selection of Design Variables. The area ratio and the lateral for general computational use in determining the performance of
flow resistance are the variables which most significantly affect a given manifold.
the flow distribution in a manifold system. The effects of these Alternatively, the flow distribution equation has the ad-
two terms is clearly indicated in the experimental results re- vantage of being formulated in terms of only one dependent
ported by Figs. 6 through 9. In many system designs, the area parameter, namely the flow rate in the dividing flow header.
ratio aiid lateral flow resistance are fixed by other requirements. This equation may be solved analytically under some conditions
Therefore, the only other design parameters which may be altered [27]. However, if the design of the manifold is not constant along
to improve flow distribution are the relative length of the mani- the path length, then it is necessary to formulate functions such
fold, the friction factors, and the orientation between the inlet as dH/dx for lateral resistance, d(Ar)/dx for the area ratio, etc.,
and outlet headers. A reverse flow system will have a better flow when these parameters vary with distance. Stepwise changes in
distribution than a parallel flow system (other parameters the these parameters are difficult to accommodate analytically since
same) when the dividing flow header is dominated by pressure the derivatives become infinite. For systems in which manifold
recovery due to branching and friction effects are minimal. A flow coefficients are constants, the performance of the manifold
parallel flow system can give better flow distribution than a re- can be determined parametrically from the flow distribution
verse flow system if friction effects dominate the dividing flow equation (see reference [27]) for ease in design usage.
header (but this is not necessarily so). In general, the flow dis- In addition to predicting the performance of a given design,
tribution in a reverse flow system will be better than the parallel the governing equations can be solved alternatively to predict
flow system for most designs used in industrial applications. The the form of the lateral resistance variation along the headers
total pressure losses for the reverse flow system are typically necessary to give uniform flow distribution in the laterals. This
less than for the parallel flow system. design procedure can be accomplished by assuming that the di-
mensionless volume flow rate in a header is given by (l-x). Then
Application of Analytical Model. The analytical model has it is possible to solve for the local resistance at each value of (x)
been formulated in terms of both a pressure-flow equation set which satisfies the governing equations.
and a flow distribution equation. The advantages of each system Many manifold systems are designed with multiple outlet
of equations may be summarized as follows. T h e pressure-flow headers and a single inlet header to improve flow distribution.
equation set clearly shows the relationship between pressure and Other designs use inlet arrangements which are not symmetrical-
velocity changes in the headers and readily leads to a physical ly spaced with respect to the outlet header. One example of
interpretation of the manifold problem. The formulation of the such a design is illustrated by a system for which the inlet flow
equations in terms of first derivative expressions for pressure and enters the dividing flow header at its midpoint and leaves the
velocity allows for a variation of parameters like lateral resist- combining flow header at one of the ends. The calculation of the
ance, porosity, and the momentum coefficients along the header, flow distribution in manifolds of this type is facilitated by the
without the need to express these variations in terms of con- application of the proposed analytical model since the conserva-
tinuously differentiable functions. Therefore the pressure-flow tion of mass bookkeeping requirement is automatically satisfied
set of equations can accept system designs in which the number by the governing equations. Analytical models for nonsym-
of laterals at a branch point may vary discontinuously with metrical manifold designs are discussed in reference [29].
axial location. The pressure-flow equation set is recommended
The analytical model is formulated in terms of the momentum
equation as the governing conservation equation for the header
flow streams. An alternative approach is to use the Bernoulli
equation to compute the pressure changes in the headers as a
result of the branching process. The dashed line on Fig. 12 il-
lustrates the application of the Bernoulli equation (represented
by 0i = 1.00) to a simple dividing flow header. The momentum
model using the value of 0i = 1.06 clearly gives a better fit to
the data. The data of Fig. 13, which illustrates the application
of the momentum model to a simple combining flow manifold,
could not be adequately represented by a solution utilizing the
Bernoulli equation (02 = 1.00) to predict the pressure changes
in the header due to branching. These two examples clearly
demonstrate the superiority of momentum models over Bernoulli
models for branching process.

Summary
The analytical model for flow distribution in manifolds de-
scribed in the present paper was formulated from a first principles
approach to the problem which included the use of the continuity
and momentum equations for the header flows and a discharge
equation for the lateral flows. The approach has been generalized
MOMENTUM MODEL 0 , = 1.06
-• BERNOULLI MODEL 0 , = 1.00
in terms of dimensionless equations and flow coefficients. The
<> DATA FROM SHERMAN, TEST 5 model is applicable to a wide range of system designs. The ex-
perimental results are in substantial agreement with the analy-
tical model. The analytical model is recommended for general
application in the analysis of flow distribution in manifolds.

0 0.5 1.0
Acknowledgments
DIMENSIONLESS DISTANCE ALONG HEADER X/L
This research program was partially supported by the Babcock
Fig. 12 Comparison of analytical model with experimental data for & Wilcox Company. T h e authors appreciate the advice and dis-
a simple dividing flow manifold with large area ratio and small lateral
resistance (reference [18]) cussions offered by J. H. Kidwell, It. A. Lee, and M. Wiener .

664 / DECEMBER 1976 Transactions of the ASME

Downloaded From: https://fluidsengineering.asmedigitalcollection.asme.org on 07/02/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
 - - --_._- - - - _.. _ - - - tion anel Pressure Los~ with Circ:umfcrential Tnlct lind Outlet
?vllLnifolds," NASA Technic:nl 0iote TN D-66fl7, :'1'1111'. 1072.
o 1<1 Enger, M. L .. and Levy, ?vI. 1., "Presstll'es in Manifold
o Pipes," Journal of the A merican Water 1V0rks A.~sociation, May
1!J20, flP. 6;J!HiGl.
o 15 Van del' Hcgge Zijncn, B. G., "Flow Through Uniformly
Tltpped Pipes," Applied Science Research, Vol. Ail, 1!l52, pp.
144-162.
Hi Markland, 1<:., "Analysis of Flow from Pipe Manifolds,"
En(jinecrill(j, Vol. 187, ,Tan. :lO, 1959, pp. 150-15l.
17 Houcek, E., awl Zelnic:k, I<:. W., "Lock Manifold Expcri-
mcnts," Trans. ASCI<}, 1!J4!i, pp. 1357-1400.
IR Bnjura, lL A., "A :'I<{odel for Flow Distribution in Mani-
folds," JOl/rnal of Rn(jineering for Power, TitANS. AS ME, VoL !l3,
1!l71, Jl". 7-12.
lfl i1owlnnd, W. Eo, "Dcsign of Perforated Pipe for Uni-
formity of Diseharge," Proceedin(js oj the Third Midwest Con-
- - t,'O\\lrHIJM ',mDEL 0:1 766
ference on fihdcl Mechanics, l!Jii:l, pp. (lH7-70l.
- - - - fll.:flNOUlli MOOfL 02 - 100
20 l'earimut,t.er, lVI., "Inlet and Exit Headcr Sha.pcs for Uni-
200 -- o DATA I 110M SHEllMAN. HST I" form Flow Through a H(~sistance Pamllel to the :'I<{[lin StrCllm,"
J ollrnai of Nasic En(jinecring, TItANS. ASME, Vol. H3, Scpt. 1!lGI,
pp. aGI-:17o.
OF.AD
21 :'.<{[lrdon, ,T., Hauptm!tnn, E. G., Monahnn, R. E., ane!
[NO
___ ~.I ___ .1 _ _ _ -1
Brown, E. S., "The Extallt St.ate of the Manifold Problem,"
Tec:hnic:al Paper T34(\, Pulp and j'aper Magazine of Canada,
01\\1 ~jSIONII 5S UISfANC:1 Al.{)~JG IlfAOfH X/I Nov. l!l71, v. 72, n. II.
22 Hn.erter, A. A., "Flow Dist.ribution and Presstll'e Change
Fig.13 Comparison of analytical model with experimental data for a Along Hlot.t.ed or Brandlcd Ducts," ASHN AR Journal, Vol. 5,
simple combining flow manifold with large area ratio and large Jan. 10li;~, pp. 47-;)!J.
lateral resistance (reference [18])
2:l lIodock, ,/. II., "An Invcst.igation of t.hc Flow in :Vlllni-
folds with Opell !tnd Closcd Ends," Journal of the Royal Aero-
nautical Society, Vol. GO, Nov. l!l5G, pp. 74fl-7,j:t
24 Olscn, F. C. W., "Flow Through [I Pipe with a Porous
Wall," Journal of Applied Mechanics, TItANS. ASME, VoL 71,
Mar. 1fl4!l, pp. 5:l-54. (Discussion in TitANS. AS;\H~, Sept.
Hl49, Vol. 71, pp. 317-318.)
25 Huang, J. C. 1'., and Yu, H. 13., "Pressure Distribut,iolls
Purt of the I11I111erical (·"m]lll[.at.ioll~ wcrc carricd out. 1\[. [.hc \Vest. in PorOlI~ Ducts of Arbitmry Cross S(~ction," AS-:'I'[ I·; Paper
n-FI'>!l, Prcscnted at. thc AS:V[E ,Joint Applicd Mcc~hanic:s­
Virginia Univcrsit.y COlllpll(cr Ceniel' llndel' a fae-ulty resclIl'ch Fluids J.:ngilwering Confenmce, At.lanta, Ga., June 20 -22, 11)7:3.
gmnt.. 2(1 Qunilc, J. P., nnd Levy, I';. K., "Laminar Flow in a Porous
Tube With 8uction," AS:\-IE Paper n-WA/ITT-l, Presented at
the AS:'I-IJ<; Winter AnnlIal Meeting, Dctroit, Mic:h., Novem-
References bcr 11-1:), IOn.
McNown, J. S., "Meehanicti of :'.Ianifold Flow," Trans- 27 Bajum, n. A., LeRosc, V. F., nnd Williams, L. Eo, "Fluid
actions ASCE, Vol. 1 H), 1!J:)1, pp. IIO:l·111~. Distribut.ion in Comhining, Dividing and H.everse Flow M:lni-
2 l'I'I('Nown, ,J. S., and IIsll, KY., "Application of Confol'mal folds," AS:'IU; Paper 7:l-PWH.-I, Prcsented at the AS\I8-Il~EI';
IVlapping to Divided Flow," The First Midwcstern Conference on .J oi nt Power C;cnemtion Confcrence, N cw Orleans, La. Sep-
FllIid Dynami('.~, J\lay 12--1 :l, 1!J;)O, at tlw University of Illinois, tember 16-I!J, l!J73.
pp. H.:~-l!i!i, ](liiI. :21l JOlles, K II., Jr., "Pressurc [llld Flow Dist.ribution in a
:l Z!'isscr, M. II., "Sllmmary Heport of Singlc-Tubc Brnnch :'.Ianifold," i\IS thcsis, Dcp[lrtment of :VIec:hnnical Engineering
and Multi-Tube Bmnch \Vatcr Flow Tests Conduct.ed by the lInd :VlecIHLni!,s, West, Virginia University, :'I-Iorgnntown, West
lJniv('I'sity of Conl\cct icut.," Pratt al\d Whitncy Aircmft Division, Va., :\-lay 1974.
Cnited Aircraft Corporal.ion, Heport ;Vo. PIV AC-2,'Jl, USAEC 2\J LeHose, V. F., "An An:dytical St.udy of Flow Distribution
Contract AT (l1-1 )-22!J, May ID(i:~. in :\ianifolcb," \[S thcsis, Departmcnt of \leehllnical Engincering,
4 Sturowlszl<y, 0., "Pressure Conditions in Pipe Branches," Wcst Virginia Univer~ity, :'I'Iorgalitoll'lI, West Vn., :'IJay !!I 72.
Vizu(lYi f(ozlc1Ilenyek (Budapcst) No. I, pp. 115-121, l!Jii1l.
(Tml\slated by Language Hervice Bureau, Washington, D.C., on
tilc at thc Agricllitural Hcscardl Service, Ht. Anthony Falls
Hydraulic Laboratory, iVlillneapolis, Minn.)
5 Ruus, E., "Hcad Losscs III Wycs and MILIlifolds," Journal
of Ihe Hydmulics Division, ASCE, Vol. 3, Mnr. lfI70, pp. 593-608.
6 I\.ubo, T., and Ucdll, '1'., "On t.hc Chllraeteristics of Di-
vidcd I'low and Confluent Flow in Header~," nul/etin of JSillH
Vol. 12, No. 52, I!lG!J, pp. H02-HO!l. '
7 }(ubo, '1'., and Ueda, '1'., "On the Characteristics of Con-
fluent Flow of Gas-Liquid iVlixturcs in Headcrs," Bulletin of
JSME, Vol. Hi, No. 99, 197:3, pp. 1376-13R4.
H Oakey, J. A., "J[yclralllie Losses in Short Tllbes Dctcr-
mined by Experilllcnts,"'j;'n(jincerill(j News Record, .Junc 1, Ul3a.
pp.717-718.
\l Koh, I{. C. Y., lmd Brooks, N. n., "Fluid l\'[echn.l\i(~,; of
WIl,te-Wn.ler Disposal in the Ocenn," Annnal Heview of Flu.id
M ('(/,a1/ics, Vol. 7, 1!J7[i, pp. IH7-211, Annual Hevicws, [nc.'
Palo Alt.o, Calif.
10 Acrivos, A., Bnhcock, ll. D., and Pi~ford, J{. L., "Flow
Jli~t.l:ibllti()n in J\-[nnifolcb," Chemical Rn(linecrin(J Science, Vol.
10, No. 1/2, 1!);;!l, pp. II ~ 124.
11 Keller, ,/. D., "Thc l\Ianifold Problcm," Journal of Ap-
plu,d Jl! cc/lIl1!ics, THA:-;S. AH:V[ 1-:, Vol. 71, "'hr. 1!l4!l, pp. 77..·Hii.
(J)iSCllssioll ill THANS. AS:'I'IE, Vol. 71, Sept.. I (j·19, pp. :r,W-·;)2:2.)
12 I low, W. :'.'l., "Thc Uniform Dist.riblltion of a Fillid
Flowing Through a Pcrforated Pipe," .J onrl/al of Applied ill c-
chanic8, THANS. ASiVLE, Vol. 72, Dce. l!liiO pp.4:31-'1:l1l. (Dis-
eussion in THAl\S. ASJ\JE, l!l!il, Vol. 7:3, PI;. 221-224.)
13 Dittric:h, H. '1'., "I';xpcrimcnt:d Htuely of Flow Di,;trilm-

Journal of Fluids Engineering DEC E M B E R 1976 / 665

Downloaded From: https://fluidsengineering.asmedigitalcollection.asme.org on 07/02/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use