Manifold Flow in Pressure-Distribution Systems

S. David Graber, P.E., F.ASCE1

Abstract: Manifolds are used extensively in pipeline engineering applications. They may be classified as combining-flow in which the
flow rate increases along the length of the manifold, and dividing-flow in which the flow rate decreases along the manifold length. The
present paper addresses dividing-flow manifolds with the emphasis on pressure-distribution systems. The seminal theoretical and experi-
Downloaded from ascelibrary.org by University of Nebraska-Lincoln on 06/03/13. Copyright ASCE. For personal use only; all rights reserved.

mental work of McNown on discrete tee takeoffs, applicable to pressure-distribution-system manifolds, is discussed and elaborated upon
for flow across the run of the tee and into the tee branch. Pertinent theoretical and experimental work for discrete orifice takeoffs is
summarized. The information thus presented facilitates discussion of certain concepts presently incorporated in pressure-distribution-
system design. Certain erroneous concepts are identified. A perspective is provided and improvements in related aspects of manifold
design are suggested. A separate paper, presented at the ASCE/EWRI World Environmental and Water Resources Congress 2010 Con-
ference, builds on the present paper by specifically addressing wastewater pressure-distribution systems.
DOI: 10.1061/共ASCE兲PS.1949-1204.0000059
CE Database subject headings: Fluid mechanics; Hydraulic design; Pipes; Pipe flow; Pressure distribution; Pressure pipes.
Author keywords: Fluid mechanics; Hydraulic design; Pipes; Pipe flow; Pressure distribution; Pressure pipes.

Introduction from his paper are included here in Figs. 1 and 2, the former for
the flow continuing along the main conduit 共the “run” of the tee兲
Manifolds are used extensively in pipeline engineering applica- and the latter for the flow going into the branch of the tee. The
tions. They may be classified as combining-flow in which the abscissa on all the figures gives the flow in the branch Ql divided
flow rate increases along the length of the manifold, and dividing- by the main flow upstream of the flow division, and the ratio
flow in which the flow rate decreases along the manifold length. Dl / D is the ratio of the takeoff diameter to the conduit diameter.
Combining-flow manifolds are discussed in Graber 共2004a, On Fig. 1, the term on the ordinate is defined as
2007a,b兲. The present paper addresses dividing-flow manifolds,
with the emphasis on pressure-distribution systems. Some addi- ⌬h 共pu − pd兲/␥
tional examples of spatially decreasing flow and dividing-flow = 共1兲
V /共2g兲
2
V2u/共2g兲
manifolds include the “dispersion conduits” discussed in Camp
and Graber 共1968, 1969兲, outfall diffusers 共Camp and Graber in which pu and pd = pressure in the conduit upstream and down-
1970; French 1972b兲, subirrigation pipes 共Camp and Graber stream of the flow division, respectively; ␥ = weight per unit vol-
1972; French 1972a兲, and sodium hypochlorite generation sys- ume of fluid; Vu = velocity in the conduit upstream of the flow
tems 共Graber 2004a兲. division; and g = gravitational acceleration.
The seminal theoretical and experimental work of McNown on If one draws a control volume to encompass the takeoff and
discrete tee takeoffs, applicable to pressure-distribution-system flow upstream and downstream of the takeoff, an energy balance
manifolds, is discussed and elaborated upon for flow across the 共First Law of Thermodynamics兲 共Shames 1962兲 applied to the
run of the tee and into the tee branch. Pertinent theoretical and control volume 共neglecting energy loss兲 gives
experimental work for discrete orifice takeoffs, applicable to
pressure-distribution systems, is summarized. A separate paper,
presented at the ASCE/EWRI World Environmental and Water
Qu 冉 V2u pu
+
2g ␥
冊 冉
= Qd
V2d pd
+
2g ␥
冊 冉
+ Ql
V2l pl
+
2g ␥
冊 共2兲

Resources Congress 2010 Conference, builds on the present paper in which Q, V, and p = flow, velocity, and pressure, respectively,
by specifically addressing wastewater pressure-distribution sys- and the subscripts “u,” “d,” and “l” denote, respectively, condi-
tems 共Graber 2010兲. tions upstream of the flow division in the main conduit 关Q with-
out a subscript here and in McNown 共1954兲 is identical to Qu兴,
Discrete Tee Takeoffs downstream of the flow division in the main conduit, and in the
The theoretical and experimental studies of McNown 共1954兲 branch 关the subscript “l” is used here for consistency with
comprise a seminal work on manifold flow involving tees. Charts McNown 共1954兲, who used the term “lateral” for branch, the term
“branch” being used here, to avoid confusion with the use of the
1
term lateral in wastewater pressure-distribution systems, to mean
Consulting Engineer, 118 Larson Rd., Stoughton, MA 02072. E-mail: a particular type of manifold main conduit兴.
[email protected]
Fig. 1 has two theoretical curves by McNown 共1954兲 which
Note. This manuscript was submitted on January 31, 2010; approved
on April 16, 2010; published online on July 15, 2010. Discussion period will be explained and expanded upon. One approach is to apply
open until January 1, 2011; separate discussions must be submitted for Bernoulli’s equation and continuity 共conservation of mass兲 to the
individual papers. This paper is part of the Journal of Pipeline Systems portion of the flow continuing in the main conduit between sec-
Engineering and Practice, Vol. 1, No. 3, August 1, 2010. ©ASCE, ISSN tions upstream and downstream of the takeoff 共permissible be-
1949-1190/2010/1共3兲/120/7/$25.00. cause Bernoulli is applied along a streamline or streamtube兲.

120 / JOURNAL OF PIPELINE SYSTEMS ENGINEERING AND PRACTICE © ASCE / AUGUST 2010

J. Pipeline Syst. Eng. Pract. 2010.1:120-126.

Downloaded from ascelibrary.org by University of Nebraska-Lincoln on 06/03/13. Copyright ASCE. For personal use only; all rights reserved.

Fig. 1. Change in piezometric head in the conduit for dividing flow 共Reprinted from McNown 1954, with permission兲

Using the same notation as before and taking h f = head loss in the perimental data on Fig. 1 can be explained qualitatively by mo-
main conduit and ␣ = Coriolis energy coefficient 共accounting for mentarily disregarding the friction term and considering the effect
velocity-profile effects兲 of velocity profiles as reflected in the energy coefficient terms in
Eq. 共4a兲. The velocity profiles upstream and downstream of the
pu V2u pd V2d
+ ␣u = + ␣d + h f 共3a兲 branch will be approximately symmetrical. Considering first
␥ 2g ␥ 2g Dl / D = 1, one can readily envision for small Ql / Q that the lower
velocity portion of the upstream flow is drawn off to the branch so
Qu Qd 共Ql − Qu兲 that ␣u ⬎ ␣d or ␣u / ␣d ⬎ 1 which raises the theoretical curve of
Vu = , Vd = = 共3b兲
Ar Ar Ar −共pu − pd兲 / ␥ toward the data points. As Ql / Q becomes larger, by
this explanation the curve is raised less and less, in accordance
in which Ar = cross-sectional area of the main conduit or run.
with Fig. 1. At approximately Ql / Q = 1 / 2, the higher-velocity por-
Using Eq. 共3b兲 共continuity兲 to eliminate Vd from Eq. 共3a兲 共Ber-
tion of the upstream flow is drawn off to the branch, so ␣u / ␣d
noulli兲 yields

冋 冉 冊 冉 冊册
becomes less than unity which lowers the theoretical curve of
− 冉 ␥
冊
pu − pd
= ␣d
V2u Ql
2g Qu
2−
Ql
Qu
+
␣u
␣d
−1 − hf 共4a兲
−共pu − pd兲 / ␥, again in accordance with Fig. 1. The effect of Dl / D
can be similarly explained. For a given Ql / Q as Dl / D becomes
smaller, more of the branch flow will be drawn from the higher-
For ␣u = ␣d = 1 and h f = 0, the above equation reduces to velocity region of the main flow. That causes ␣u / ␣d to decrease as

− 冉 ␥
冊
pu − pd
= 冉 冊
V2u Ql
2g Qu
2−
Ql
Qu
共4b兲
Dl / D decreases, which, according to Eq. 共4a兲, causes −共pu
− pd兲 / ␥ to decrease. This again is in accordance with Fig. 1.
McNown 共1954兲 derived the lower theoretical curve of Fig. 1
which matches McNown’s “h f = 0” curve on Fig. 1. from the common relationship for a sudden expansion. A varia-
The friction term h f in Eq. 共4a兲 lowers −共pu − pd兲 / ␥ on Fig. 1, tion of that is derived here which provides additional insight. The
and the data points above the h f = 0 curve might appear to corre- branch exerts a force on the flow which is diverted into the
spond to an inexplicable increase in energy 共McNown 1954; Li branch. A judicious choice of control volume at the flow division
1973兲. However, the differences between the h f = 0 curve and ex- allows circumvention of the force associated with the change in

JOURNAL OF PIPELINE SYSTEMS ENGINEERING AND PRACTICE © ASCE / AUGUST 2010 / 121

J. Pipeline Syst. Eng. Pract. 2010.1:120-126.

 continuing flow 共Qd兲. Pressure variations and shear stresses along
the control surface in the vicinity of the branch are assumed neg-
ligible.
For this control volume, which avoids the forces that turn the
flow into the branch, there is an efflux of momentum ␤l␳QlVu in
which ␳ = fluid mass density 共=␥ / g兲 and ␤ = Boussinesq momen-
tum coefficient 共correcting for velocity-profile effects, in this case
for the flow that goes into the branch兲. The momentum equation
then becomes
共pu − pd兲Ar = − ␤u␳VuQu + ␤d␳VdQd + ␤l␳VuQl 共5兲
in which momentum coefficients are applied to the respective
Downloaded from ascelibrary.org by University of Nebraska-Lincoln on 06/03/13. Copyright ASCE. For personal use only; all rights reserved.

terms. Employing continuity 共Qu = Ql + Qd and V = Q / A兲 with the

above equation gives the following:

− 冉 pu − pd
␥
冊 冋冉
=2
V2u
2g
␤u
␤u ␤d
−
␤d ␤u
+ ␤u 冊
Ql ␤d ␤l
2 −
Qu ␤u ␤u
− 冉
Ql
Qu
冊 冉 冊册 2

共6a兲
Taking ␤u = ␤d, the above equation reduces to

− 冉 pu − pd
␥
=2冊 冋 冉 冊 冉 冊册
V2u
2g
␤u
Ql
Qu
2−
␤l
␤u
−
Ql
Qu
2
共6b兲

For ␤u = ␤d = ␤l = 1 Eq. 共6a兲 reduces to

− 冉 ␥
冊 冋 冉 冊册
pu − pd
=2
V2u Ql
2g Qu
1−
Ql
Qu
共6c兲

Eq. 共6c兲 is the same as McNown’s equation plotted as the

lower theoretical curve on Fig. 1. It is seen to represent the limit
to the data curves for small Ql / Q and small values of the branch
to pipe diameter ratio 共McNown’s Dl / D兲, and to give the quali-
tative shape of the pressure rise over a larger range than the h f
= 0 curve. Similar reasoning as before can be applied to Eq. 共6a兲
to demonstrate velocity-profile effects; e.g., for small Ql / Q for
which ␤u ⬵ ␤d and ␤l / ␤u ⬍ 1, Eq. 共6b兲 shows that −共pu − pd兲 / ␥
increases in the direction of actual observations on Fig. 1.
Considering now C.V. 1 on Fig. 3, the momentum equation
becomes
共pu − pd兲Ar − Tx = − ␤u␳VuQu + ␤d␳VdQd + ␤o␳VoQl cos ␾o
Fig. 2. Head loss between the conduit and the branch for dividing 共7兲
flow 共Reprinted from McNown 1954, with permission兲 in which Tx = turning force; subscript “o” denotes the efflux; ␾o
= angle of efflux relative to main conduit; and other terms are as
given above or indicated on Fig. 3. For the tee takeoff, cos ␾o
direction of the flow 共Camp and Graber 1968, 1969, Graber = 0. Then using continuity, rearranging as before, and taking ␤u
2004b兲. The control volume of choice is shown as C.V. 2 in Fig. = ␤d, gives

冉 冊 冋 冉 冊册
3, which is bounded by sections “u” and “d” and the conduit
walls except in the vicinity of the branch or orifice, at which pu − pd V2u Ql Ql 2
Tx
− =2 2␤u − − 共8兲
region the boundary is defined by the imaginary surface separat- ␥ 2g Qu Qu ␳Qu/Vu
ing the branch flow 共Ql兲 from the inflow at section u 共Qu兲 and the
With ␤u = 1 and Tx = 0, the above equation becomes identical to
Eq. 共4b兲 and corresponds to the upper theoretical curve on Fig. 1
共more will be said about the significance of this below兲. Note that
u d Tx in Eq. 共8兲 lowers the value of −共pu − pd兲 / ␥ from the upper
C.V. 2 theoretical curve on Fig. 1 in the direction of actual observations
for larger Ql / Q. It may also be noted that a momentum equation
can be derived for C.V. 3 on Fig. 3 to give an equation for
C.V. 3 Tx which can then be used to eliminate Tx from Eq. 共7兲 and yield
blrQlVu C.V. 1 Eq. 共5兲.
Of the two theoretical approaches discussed above, Bernoulli
l can be seen to provide a better approximation to the data over a
much wider range. The lesser accuracy of the theory correspond-
Fig. 3. Control volumes for analysis of dividing flow ing to the lower curve in Fig. 1 is attributable primarily to the

122 / JOURNAL OF PIPELINE SYSTEMS ENGINEERING AND PRACTICE © ASCE / AUGUST 2010

J. Pipeline Syst. Eng. Pract. 2010.1:120-126.

 neglect of the force required to turn the flow into the branch and higher branch entrance losses than for long branches 共contracted
secondarily to velocity-profile effects 共which affect the ␤ values兲. velocity head is not recovered兲. French 共1980兲 reported observing
The proximity of the upper curve 共h f = 0兲 to the data demonstrates stability problems in which the branch entrance loss will fluctuate
the degree to which energy is conserved in the main conduit in short branches.
across the takeoff. The increase in pressure in the main conduit
across the branch 共per Fig. 1兲 is primarily due to velocity-head
recovery. That is, because Qd ⬍ Qu and −共pu − pd兲 ⬎ 0 so pd ⬎ pu, Discrete Orifice Takeoffs
thus, from Eq. 共3a兲 共neglecting h f 兲, V2u / 共2g兲 ⬎ V2d / 共2g兲, so veloc-
ity head upstream of the branch is converted 共“recovered”兲 to The Fig. 1 data can be expected to reasonably represent the flow
pressure head downstream of the branch. along the main conduit with orifice takeoffs. In particular, the
Referring now to Fig. 2, the principle phenomenon affecting experimental data for small Dl / D, representative of orifice take-
head loss between the main conduit and branch, denoted by h⬘f , is offs, more closely matches the h f = 0 curve over a range of Ql / Q
Downloaded from ascelibrary.org by University of Nebraska-Lincoln on 06/03/13. Copyright ASCE. For personal use only; all rights reserved.

the contraction and subsequent expansion of the flow entering the values, and over that range the energy can accurately be assumed
branch due to separation at the sharp corner. That contraction and to be conserved in the main conduit across the orifice takeoff.
expansion is amenable to some theoretical analysis. The head loss Analogous to the tee branch relationship, is the equation giv-
due to a sudden expansion is given by ing the flow rate through the orifice as a function of main-conduit

冉 冊
variables. Based on a control-volume analysis similar to that dis-
Ac 2
V2c
hexp = 1 − 共9兲 cussed above for the tee, Camp and Graber 共1968, 1969, 1970兲
Al 2g used the following relationship:

q = C⬘da冑2gh
in which Al = cross-sectional area of the branch and the subscript
“c” denotes the velocity and cross-sectional area at the section of 共12兲
greatest contraction. By continuity, VlAl = VcAc which, substituted in which h = pressure head 共p / ␥兲 in the pipe at the orifice location
into the above equation, gives or differential static head across the orifice 共inside to outside the

冉 冊冉 冊
hexp = 1 −
Ac
Al
2
Al
Ac
2
V2l
2g
共10兲
pipe兲, and q, a, and C⬘d = orifice discharge, area, and discharge
coefficient, respectively. The notation C⬘d is used to distinguish the
discharge coefficient in Eq. 共12兲 from one defined differently as
in which Vl = velocity in the branch after expansion of the flow. discussed next.
Substituting the contraction coefficient Cc = Ac / Al into the above Vigander et al. 共1970兲 presented a relationship in essentially
equation yields the following form:

hexp = 共1 − Cc兲2
1 V2l
2
Cc 2g
=
1
Cc
冉 冊
−1
2
V2l
2g
共11兲 q = Cda冑2gE 共13a兲

McNown and Hsu 共1951兲 applied potential flow theory, using V2d
E= +h 共13b兲
a conformal mapping technique, to derive a relationship for Cc for 2g
a two-dimensional idealization of flow into the branch of a pipe
tee. They expressed Cc as a function of Ql / Q and Dl / D 共here Dl,
V2d
D, Ql, and Q are, respectively, branch diameter, main-conduit Cd = Cdo − r 共13c兲
diameter, branch flow rate, and conduit flow rate upstream of the 2gE
branch兲, and claimed to use that relationship, together with the in which Vd = velocity in the conduit just downstream of the ori-
above equation, to obtain the theoretical curves shown in Fig. 2. fice 共Brooks 1970兲; E = energy head 共pressure head plus velocity
The figure represents the head loss in question in multiples of the head兲 in the conduit at the orifice location; Cdo = orifice coefficient
branch and upstream conduit velocity heads as a function of Dl / D at 0 pipe velocity; r = coefficient; and other terms are as defined
and Ql / Q. The agreement between McNown’s experimental data above. Vigander et al. 共1970兲 determined experimental values of
and theory appears to be good except for Ql / Q ⬎ 0.5 with Dl / D Cdo and r for various types of pipes and orifices. Of particular
= 1; however, the equation published by McNown and Hsu interest here are the values for sharp-edged holes in a smooth-
共1951兲, which can be readily solved numerically by, e.g., graphi- walled 共as distinct from corrugated兲 pipe, for which they deter-
cal bisection, does not give their theoretical curves depicted on mined Cdo = 0.63 and r = 0.58, i.e., the discharge coefficient is
Fig. 2. Nevertheless, their notion that a two-dimensional idealiza- given by
tion of the dividing-flow problem could apply to a circular branch
does seem valid for the limiting case of small Dl / D for which the V2d
theoretical solution for the flow issuing from a large vessel by a Cd = 0.63 – 0.58 共14兲
2gE
two-dimensional aperture gives Cc = ␲ / 共␲ + 2兲 共Lamb 1945;
Streeter 1948兲; that limiting value, substituted into Eq. 共11兲, gives Based on theoretical demonstration and experimental data
h⬘f / 关V2l / 共2g兲兴 = 共2 / ␲兲2 ⬇ 0.405, which can be seen from Fig. 2共a兲 共Camp and Graber 1969, 1970, Graber 2004b兲, the discharge co-
to be accurately reached for Dl / D = 0.25 when Ql / Q ⬎ 0.5. efficient is more nearly constant when h is used 关Eq. 共12兲兴 than
McNown 共1954兲 reported that the branch measurements for when E is used 关Eq. 共13a兲兴 in the orifice equation. The theoretical
his data were made at a point five branch diameters along the demonstration relates to the derivation presented above corre-
branch. Hudson et al. 共1979兲 noted the difference in measured sponding to the lower curve of Fig. 1; although the neglect of the
data for short and long branches, and defined “long” branches as turning force, discussed above in connection with that derivation,
“substantially greater than three pipe diameters in length.” Hud- leads to lesser agreement with experimental data for tee branches,
son et al. 共1979兲 also presented data for short branches 共without it gives better agreement with experimental data for sharp-edged
quantifying the branch lengths兲, which indicated significantly orifices in which the turning force is small.

JOURNAL OF PIPELINE SYSTEMS ENGINEERING AND PRACTICE © ASCE / AUGUST 2010 / 123

J. Pipeline Syst. Eng. Pract. 2010.1:120-126.

 However, variation in Cdo may remain, and for more accurate velocity-head recovery. The significance of velocity-head recovery
work 共which can result in overall cost savings兲 Eqs. 共13兲 and 共14兲 depends on its magnitude in comparison to energy losses due to fric-
are useful. As previously demonstrated by Camp and Graber tion along the conduit. Velocity-head recovery can be quite impor-
共1970兲, the orifice discharge coefficients in Eqs. 共12兲 and 共13兲 are tant in the design of pressure-distribution conduits, and its neglect
related by in all cases is a major shortcoming of the Otis method.
By neglecting velocity-head recovery across each takeoff, Otis
Cd
共1982兲 in essence assumes that 共pu − pd兲 / ␥ = 0, which can be seen
冑
Cd⬘ = 共15兲
V2d with reference to Fig. 1 to be in poor agreement with the data. On
1−
2gE the other hand, negligible energy loss in the main conduit across
takeoffs was assumed in the lumped-parameter 共discrete takeoffs兲
approach of Rawn et al. 共1961兲 共who cited McNown 1954 as the
basis for that assumption兲 and Vigander et al. 共1970兲, and is implicit
Downloaded from ascelibrary.org by University of Nebraska-Lincoln on 06/03/13. Copyright ASCE. For personal use only; all rights reserved.

Relation to Present Concepts of Pressure-Distribution in the distributed-parameter 共continuous takeoff兲 approach of Camp
Design and Graber 共1968, 1969, 1970, 1972兲 and Graber 共1981, 2004a兲.
That is consistent with the upper curve 共h f = 0兲 on Fig. 1, based on
The above facilitates discussion of certain concepts presently incor- constant energy across the takeoff. That curve can be seen to pro-
porated in pressure-distribution-system design. Otis 共1982兲 states: vide a reasonable approximation to the data for Ql / Qu up to about
“The loss of energy at the junction causes a change in the liquid 0.5. The value of Ql / Qu increases in the downstream direction in the
pressure within the lateral 关here meaning a main conduit兴. The mag- conduit, reaching 0.5 共assuming uniform distribution兲 at the next to
nitude of change can be determined by equating the total energy the last takeoff. For the last takeoff, Ql / Qu = 1 共again, assuming uni-
upstream from the perforation to the energy in the liquid discharged form distribution兲, but the eddying that occurs downstream of that
through the perforation and the energy remaining in the lateral last takeoff in any short continuation of the conduit results in head
关main conduit兴 downstream of the perforation. The total energy in loss and brings the upper curve of Fig. 1 downward for Ql / Qu = 1.
each liquid stream is the sum of its velocity head, pressure, and The situation would be somewhat different if the last orifice is con-
elevation in relation to a fixed datum.” Based on this reasoning, Otis sidered to be a vent hole. However, it is not important to know about
共1982兲 presents an equation 关his Eq. 共3兲兴 which, in terms of the vari- conditions downstream of the last orifice, so any inaccuracy at that
able definitions used above, is as follows: location is irrelevant.
pd − pu 共V2u − V2l 兲 V2l A key question is how to consider the head loss in the branch
= − 共16兲 entrance. Unlike the relationships discussed above for orifice take-
␥ 2g 2g
offs, no good relationship is available for the branch head loss over
Contrary to the quoted portions of the previous paragraph, Eq. a wide range of conditions. The method and simple relationship
共2兲 shows that the velocity head and pressure terms have to be mul- proposed by Hudson et al. 共1979兲, to which Otis 共1982兲 refers and
tiplied by the flow rate of the respective stream. To see how Otis Berkowitz 共1985兲, White et al. 共1985兲, and Chaudhry and Reis
derived Eq. 共16兲 关his Eq. 共3兲兴, the flow terms are dropped from Eq. 共1992兲 use 共the first three in the context of wastewater pressure-
共2兲 共a temporary assumption兲 and the equation 共with pl = 0, i.e., 0 distribution systems, and the latter in the context of underdrains in
gauge pressure for an orifice discharging to the atmosphere or, e.g., rapid sandfilters兲, oversimplifies the estimation of branch entrance
an adequately ventilated wastewater leaching field兲 cast in terms of losses. Whereas McNown 共1954兲, as discussed above, found that
the pressure-head difference across the orifice in the conduit. How- the nondimensional branch entrance loss h⬘f / 关V2l / 共2g兲兴 is a function
ever, the dropping of the flow terms has no validity, and Eq. 共16兲 of Dl / D and Ql / Q, Hudson et al. 共1979兲 suggested that
关Otis’ Eq. 共3兲兴 is not valid. That does not matter for the development h⬘f / 关V2l / 共2g兲兴 may be expressed as a linear function of the single
of Otis 共1982兲 per se because he essentially abandons Eq. 共16兲 关his variable 共Vl / Vu兲2. They presented a plot of data from McNown
Eq. 共3兲兴 and neglects the velocity-head recovery anyway 共as do 共1954兲 and others which appeared to be adequately fit by that linear
Berkowitz 1985, Reese et al. 1985, and White et al. 1985兲. How- function with only modest scatter of the data about the line. How-
ever, it does matter if one is to properly include velocity-head ever, in the case of the data of McNown 共1954兲, Hudson et al.
recovery. showed only 7 of the 48 McNown data points shown on Fig. 2.
Otis interprets Eq. 共16兲 关his Eq. 共3兲兴 as suggesting that When all of McNown’s data points are shown, the scatter is much
“关w兴here…关the velocity兴 is high near the inlet to the lateral 关mani- greater. When only the 7 data points are considered, the average-
fold main conduit兴, the change in pressure is positive, but and maximum-absolute difference of the linear fit from McNown’s
the…关change in pressure兴 becomes negative towards the distal end data are about 6 and 12%, respectively. However, when all of
of the lateral 关manifold兴. Thus, the in-line pressure increases McNown’s data points are considered, the average- and maximum-
slightly before decreasing downstream within the lateral 关mani- absolute difference of the linear fit from McNown’s data are about
fold兴.” That is incorrect. As shown on Fig. 1, ⌬h / 关V2 / 共2g兲兴 is 12 and 37%, respectively. 关For the especially important case 共for
always positive, irrespective of the relative sizes of takeoff and pressure-distribution systems兲 of Dl / D = 1, the linear fit of Hudson
main conduit and irrespective of the flow ratio. Thus, from Eq. 共1兲, et al. 共1979兲 has an average- and maximum-absolute difference
the term 共pu − pd兲 is always negative, meaning that the upstream from McNown’s data of about 26 and 37%, respectively.兴 Similarly,
pressure is always less than the downstream pressure and the down- Hudson et al. showed only 14 of the 74 data points of Vennard and
stream pressure is always greater than the upstream pressure. The Dentoni 共1954兲, whose data exhibited good agreement with the
change in pressure is always positive, contrary to Otis’ assertion. McNown data shown on Fig. 2. When all the data are taken into
Eq. 共4b兲 also indicates that the pressure in the main conduit rises account, the linear function of Hudson et al. 共1979兲 does not provide
across the orifice 关because Ql ⬍ Qu so Ql / Qu ⬍ 1, then 共2 − Ql / Qu兲 a good fit to the data.
⬎ 0, −共pu − pd兲 ⬎ 0, and finally pd ⬎ pu兴; Eq. 共6c兲 indicates the same The method of Hudson et al. 共1979兲 also neglects both friction
by similar reasoning. As noted above, the pressure downstream of loss and velocity-head recovery in the manifold; the latter is also
the takeoff is greater than that upstream of the takeoff because of noted by French 共1980兲. McNown 共1980兲, in commenting on an

124 / JOURNAL OF PIPELINE SYSTEMS ENGINEERING AND PRACTICE © ASCE / AUGUST 2010

J. Pipeline Syst. Eng. Pract. 2010.1:120-126.

 unacceptable means of addressing this problem proposed by Hud- bution-system design. Certain erroneous concepts are identified.
son et al. 共1979兲, suggested that: “The manifold problem and all its Perspective is provided and improvements in related aspects of
complications disappear if the manifold acts like a quiescent reser- manifold design are suggested. A separate paper, presented at the
voir.” That requires the manifold to be large enough that the branch ASCE/EWRI World Environmental and Water Resources Con-
entrance loss is small compared to other pertinent head losses: by gress 2010 Conference, will build on the present paper by ad-
Fig. 2共b兲, small V makes h⬘f small. McNown 共1980兲 also noted that a dressing specifically wastewater pressure-distribution systems
large number of small branches can enable a near-constant branch 共Graber 2010兲.
entrance coefficient for many of the branches. That can be appreci-
ated with reference to Fig. 2共a兲: if Dl / D is small, then h⬘f / 关V2l / 共2g兲兴
becomes nearly constant, except for small Ql / Q which is most
Notation
likely to occur near the upstream end of the manifold.
One alternative for manifolds that feed branches which are all at
The following symbols are used in this paper:
Downloaded from ascelibrary.org by University of Nebraska-Lincoln on 06/03/13. Copyright ASCE. For personal use only; all rights reserved.

the same elevation is suggested by Fig. 2共b兲. If Dl / D = 1 then

A ⫽ flow area;
h⬘f / 关V2l / 共2g兲兴 ⬵ 1, so the branch entrance loss will approximately
a ⫽ orifice area;
vary only with the velocity head of the flow approaching the branch
Cc ⫽ contraction coefficient;
in the manifold and becomes relatively small 共compared to the
Cd ⫽ orifice discharge coefficient in Eq. 共13a兲;
branch entrance loss for smaller values of Dl / D兲. The analysis is
Cd⬘ ⫽ orifice discharge coefficient in Eq. 共12兲;
made significantly more tractable by having the branch entrance
Cdo ⫽ orifice coefficient at zero pipe velocity;
loss approximately independent of Ql / Q, and even more so by hav-
D ⫽ takeoff or conduit diameter 共latter if not subscripted兲;
ing the branch entrance loss be approximately equal to the velocity
E ⫽ energy head 共pressure head plus velocity head兲;
head in the manifold approaching the takeoff. By subtracting out the
g ⫽ gravitational acceleration;
velocity head, the available head becomes the pressure head, and by
h ⫽ pressure head 共p / ␥兲 or differential static head
using the pressure head in the analysis one automatically takes into
across orifice;
account the branch entrance loss. That is akin to the concept of
h f ⫽ pressure-head loss in the main conduit;
using the pressure head in the orifice equation, assumed by Camp
h⬘f ⫽ pressure-head loss between the main conduit and
and Graber as discussed above. However, this approach is conser-
the branch;
vatively accurate only when the manifold branch entrance loss is
p ⫽ pressure;
small compared to the total head loss from manifold through
Q ⫽ volumetric rate of flow in takeoff or main conduit
branch. 共It is, however, noteworthy that h⬘f / 关V2l / 共2g兲兴 ⬵ 1 fits
共latter if not subscripted兲;
McNown’s data within an average and maximum of about 21 and
q ⫽ orifice discharge;
37%, respectively, better than the fit by Hudson et al. as given
r ⫽ coefficient in Eq. 共13c兲;
above兲. The important point made here is that branch entrance
Tx ⫽ turning force;
losses can be significant even for Dl / D = 1.
V ⫽ average velocity;
A useful fit to McNown’s data for Dl / D = 1 is given by the fol-
␣ ⫽ Coriolis energy coefficient;
lowing fifth-order polynomial:
␤ ⫽ Boussinesq momentum coefficient;
h⬘f
V2/共2g兲
= 5.300
Ql
Q
冉 冊 5
− 9.234 冉 冊
Ql
Q
4
+ 3.605 冉 冊
Ql
Q
3 ␥ ⫽ weight per unit volume of fluid;
␳ ⫽ fluid mass density 共=␥ / g兲; and
␾o ⫽ angle of efflux relative to main conduit.
+ 2.072冉 冊 Ql
Q
2
− 1.605冉 冊Ql
Q
+1 共17兲 Subscripts
c ⫽ condition in the branch at the section of greatest
which has an average- and maximum-absolute difference from the
contraction;
data of about 0.40 and 5% 共essentially within the range of scatter of
d ⫽ main conduit downstream of flow division or just
the data points about a faired curve兲, respectively, the maximum
downstream of orifice;
difference decreasing to about 2% when only Ql / Q values less than
l ⫽ conditions in takeoff or branch;
0.5 and equal to 1.0 are considered 共the range of values of Ql / Q
o ⫽ efflux;
when distribution is uniform兲. A similar polynomial can be fit to the
r ⫽ conditions in the main conduit or run; and
curves for Dl / D equal to 0.25 and 0.5, and interpolation 共parabolic
u ⫽ main conduit upstream of flow division.
or second-order polynomial兲 can be made between the three poly-
nomials for a particular value of Ql / Q, all of which can be readily
implemented by computer once the polynomial fits to the curves for
the three Dl / D values are obtained. References

Berkowitz, S. J. 共1985兲. “Pressure manifold design for large subsurface

ground absorption sewage systems.” Proc., 4th National Symp. on
Summary and Conclusions
Individual and Small Community Sewage Treatment, American Soci-
ety of Agricultural Engineers, St. Joseph, Mich., 39–48.
The seminal theoretical and experimental work of McNown
Brooks, N. H. 共1970兲. “Ocean disposal.” Summer course notes on stream
共1954兲 on discrete tee takeoffs, applicable to pressure-dis-
and estuarine analysis, Manhattan College, New York.
tribution-system manifolds, is discussed and elaborated upon for Camp, T. R., and Graber, S. D. 共1968兲. “Dispersion conduits.” J. Sanit.
flow across the run of the tee and into the tee branch. Pertinent Engrg. Div., 94共1兲, 31–39.
theoretical and experimental work for discrete orifice takeoffs is Camp, T. R., and Graber, S. D. 共1969兲. “Closure to ‘Dispersion conduits’
summarized. The information thus presented facilitates discussion by T. R. Camp and S. D. Graber.” J. Sanit. Engrg. Div., 95共5兲, 943–
of certain concepts presently incorporated in pressure-distri- 947.

JOURNAL OF PIPELINE SYSTEMS ENGINEERING AND PRACTICE © ASCE / AUGUST 2010 / 125

J. Pipeline Syst. Eng. Pract. 2010.1:120-126.

 Camp, T. R., and Graber, S. D. 共1970兲. “Discussion of ‘Internal hydrau- Li, Y. T. 共1973兲. “Head losses in T-junctions and manifolds.” Water
lics of thermal discharge diffusers’ by S. Vigander, R. A. Elder, and N. power, I. P. C. Electrical–Electronic Press, London, 266–272.
H. Brooks.” J. Hydr. Div., 96共12兲, 2631–2635. McNown, J. S. 共1954兲. “Mechanics of manifold flow.” Trans. Am. Soc.
Camp, T. R., and Graber, S. D. 共1972兲. “Discussion of ‘Manifold flow in Civ. Eng., 119, 1103–1142.
subirrigation pipes’ by E. Zsák.” J. Hydr. Div., 98共4兲, 728–732. McNown, J. S., 共1980兲. “Discussion of ’Dividing-flow manifolds with
Chaudhry, F. H., and Reis, L. F. R. 共1992兲. “Calculating flow in manifold square-edged laterals,’ by H. E. Hudson, R. B. Uhler, and R. W.
and orifice system.” J. Environ. Eng., 118共4兲, 585–596. Bailey.” J. Envir. Engrg. Div., 106共4兲, 864–866.
French, J. A. 共1972a兲. “Discussion of ‘Manifold flow in subirrigation McNown, J. S., and Hsu, E.-Y. 共1951兲. “Application of conformal map-
pipes’ by E. Zsák.” J. Hydr. Div., 98共4兲, 732–736. ping to divided flow.” Proc., Midwestern Conf. on Fluid Dynamics, J.
French, J. A. 共1972b兲. “Internal hydraulics of multiport diffusers.” J. W. Edwards, Ann Arbor, Mich., 143–155.
Water Pollut. Control Fed., 44共5兲, 782–795. Otis, R. J. 共1982兲. “Pressure distribution design for septic tank systems.”
French, J. A. 共1980兲. “Discussion of ‘Dividing-flow manifolds with J. Envir. Engrg. Div., 108共1兲, 123–140.
square-edged laterals’ by H. E. Hudson, R. B. Uhler, and R. W. Rawn, A. M., Bowerman, F. R., and Brooks, N. H. 共1961兲. “Diffusers for
Downloaded from ascelibrary.org by University of Nebraska-Lincoln on 06/03/13. Copyright ASCE. For personal use only; all rights reserved.

Bailey.” J. Envir. Engrg. Div., 106共3兲, 675–677. disposal of sewage in seawater.” Trans. Am. Soc. Civ. Eng., 126共III兲,
Graber, S. D. 共1981兲. “Discussion of ‘Hydraulic design of flow distribu- 344–388.
tion channels’ by J.-L. Chao and R. R. Trussell.” J. Envir. Engrg. Div., Reese, L. E., Gold, A. J., and Loudon, T. L. 共1985兲. “Computer generated
107共1兲, 299–301. pressurized effluent distribution lateral systems.” Proc., 4th National
Graber, S. D. 共2004a兲. “Collection conduits including subsurface drains.”
Symp. on Individual and Small Community Sewage Treatment, Ameri-
J. Environ. Eng., 130共1兲, 67–80. can Society of Agricultural Engineers, St. Joseph, Mich., 49–58.
Graber, S. D. 共2004b兲. “Concepts of spatially varied flow.” Proc., HT- Shames, I. H. 共1962兲. Mechanics of fluids, McGraw-Hill, New York, 5.13.
FED04, 2004 ASME Heat Transfer/Fluids Engineering Summer Streeter, V. L. 共1948兲. Fluid dynamics, McGraw-Hill, New York, 174–
Conf., ASME, New York. 177.
Graber, S. D. 共2007a兲. “Full-flowing collection conduits with nonuniform Vennard, J. K., and Dentoni, D. I. 共1954兲. “Discussion of ‘Mechanics of
inflow.” J. Environ. Eng., 133共6兲, 575–580. manifold flow’ by J. S. McNown.” Trans. Am. Soc. Civ. Eng., 119,
Graber, S. D. 共2007b兲. “Hydraulic considerations in geosynthetic and 1136–1140.
aggregate subsurface drains.” J. Environ. Eng., 133共9兲, 869–878. Vigander, S., Elder, R.A., and Brooks, N.H. 共1970兲. “Internal hydraulics
Graber, S. D. 共2010兲. “Manifold flow in wastewater pressure-distribution of thermal discharge diffusers.” J. Hydraul. Div., Am. Soc. Civ. Eng.,
systems.” Proc., World Environmental and Water Resources Con- 96共2兲, 509–527.
gress., ASCE, Reston, Va. White, T. E., Bell, J. M., and Jones, D. D. 共1985兲. “Computer-aided
Hudson, H. E., Uhler, R. B., and Bailey, R. W. 共1979兲. “Dividing-flow design of pressure dosed on-site distribution systems.” Proc., 4th Na-
manifolds with square-edged laterals.” J. Envir. Engrg. Div., 105共4兲, tional Symp. on Individual and Small Community Sewage Treatment,
745–755. American Society of Agricultural Engineers, St. Joseph, Mich., 30–
Lamb, H. 共1945兲. Hydrodynamics, 6th Ed., Dover, New York, 98–99. 38.

126 / JOURNAL OF PIPELINE SYSTEMS ENGINEERING AND PRACTICE © ASCE / AUGUST 2010

J. Pipeline Syst. Eng. Pract. 2010.1:120-126.