8.

29 BRANCHING PIPES
For convenience, let us consider three pipes connected to three reservoirs as in Fig. 8.27 and connected
together or branching at the common junction point J. Actually, we can consider that any of the pipes is
connected to some other des. tination than a reservoir by simply replacing the reservoir with a piezometer
tube in which the water level is the same as the reservoir surface. We shall sup. pose that all the pipes are
sufficiently long that we can neglect minor losses and velocity heads, so hL = hf which we shall designate as
h.

We name the pipes and flows and corresponding friction losses as in the di. agram. The continuity and
energy equations require that the flow entering the

junction equals the flow leaving it, and that the pressure head at J (which we shall represent schematically
by the imaginary open piezometer tube shown, with water at elevation P) is common to all pipes.

As there are no pumps, the elevation of P must lie between the surfaces of reservoirs A and C. If P is level
with the surface of reservoir B then h2 and Q2 are both zero. If P is above the surface of reservoir B then
water must flow into B and Ql + Q2 = Q3, If P is below the surface of reservoir B then the flow must be out of
Band Ql + Q2 = Q3, SO for the situation shown in Fig. 8.27 we have the following governing conditions:
1. Ql + Q2 = Q3
2. Elevation of P is common to all three pipes.
The diagram suggests several different problems or cases, three of which we will discuss below using
different methods of solution.

Rigorous Solutions
When we know the pipe wall material, we can estimate its e value (Table 8.1), and we know that the friction
factor f varies with the e/D of the pipe and the Reynolds number of the flow. Because we are not considering
minor losses, we can use the equations and methods of Sees. 8.12-8.17. In particular, using only a basic
scientific calculator, we can solve pipelines for h (Type 1 problems) using the Haaland equation (8.52) with
Eq. (8.13); we can solve for V or Q (Type 2 problems) using Eq. (8.56); and more rarely, we can solve for D
(Type 3 problems) using Eqs. (8.57)-(8.62). We prefer these equations because they avoid trial and error,
which can become quite confusing when combined with other trial-and-error techniques needed to solve for
branching flow.
The following three cases illustrate some of the "manual" trial-and-error methods used to solve the
different types of problems that can occur. In each case all the required pipe data (lengths, diameters, and
materials for e values) are known.

Case 1. Find the flow to or from two reservoirs, and the surface elevation of one of these, given the other
flow and two elevations and all the pipe data.
We can solve this problem directly, without trial and error. Suppose that Q1 and the elevations of A and B are
given. Then we can directly determine the head loss hI (Type 1), using Fig. 8.11 or Eq. (8.52) to find the
proper value off Knowing hI fixes P, so now we can easily obtain h2 Knowing h2 enables us to directly
determine the flow in pipe 2 using the Type 2 equation (8.56). Condition 1 (continuity at junction J) then
determines Q3 which in turn enables us to directly find h3 (Type 1), in the same way as for line 1. Finally, P
and h3 define the required surface elevation of C.

Case 2. Find the flow to or from two reservoirs, and the surface elevation of the third, given the other flow
and two elevations and all the pipe data.
Let us suppose that Q2 and the surface elevations of A and C are given. Then we know hI + h3 =
Δh 13 say. We may use various solution approaches;26 we shall discuss a more convenient one. In this, we
assume the elevation of P, which yields values for hI and h3, and so QI and Q3 via Eq. (8.56). If these do not
satisfy the discharge relation at J then we must adjust P until they do. To help us converge on the correct
elevation of P, we can plot the results of each assumption on a graph like Fig. 8.28. For ∑ Q at J,
inflows to J are taken as positive and outflows as negative. Two or three points, with one fairly close to the
vertical axis, determine a (near-straight) curve that intersects that axis at the equilibrium level of P, where
∑ Q = 0, as required. Last, we can determine h2 from Q2 and Eqs. (8.52) and (8.13), and find the
required surface elevation of B.

Case 3. Find the flow in each pipe, given the surface elevation of all three reservoirs and all the pipe data.

This is the classic three-reservoir problem, and it differs from the foregoing cases in that it is not
immediately evident whether the flow is into or out of reservoir B. We can readily determine this direction
by first assuming no flow in pipe 2; that is, assume the piezometer level P at the elevation of the surface of
B. The head losses hI and h3 then determine the flows QI and Q3 via Eq. (8.56). If
Ql > Q3, then we must raise P to satisfy continuity at J, causing water to flow into reservoir B, and we shall
have Ql = Q2 + Q3; if Ql < Q3, then we must lower P to satisfy continuity at J, causing water to flow out of
reservoir B, and we shall have Ql = Q2 + Q3, From here on the solution proceeds by adjusting P as for Case
2.

Note: For any of the above three cases, we can avoid manual trial and error by setting up the governing
equations and solving them simultaneously using equation solving software like Mathcad or Excel (see
Appendixes C.2-C.3 and D.2-D.3). There will be the usual four governing equations for each line, a flow
continuity equation for the junction, and, depending on the case addressed, one or two equations relating the
various head losses. With so many unknowns to solve for, the success of the procedure becomes more
sensitive to the guessed values, and we may have to try different guesses. The great advantage of this ap-
proach is that it is so straightforward. It is illustrated in part (b) of the following sample problem.

SAMPLE PROBLEM 8.14

Given that, in Fig. 8.27, pipe 1 is 6000 ft of 15 in diameter, pipe 2 is 1500 ft of 10 in diameter, and pipe 3 is
4500 ft of 8 in diameter, all asphalt-dipped cast iron. The elevations of the water surfaces in reservoirs A and
Care 250 ft and 160 ft, respectively, and the discharge Q2 of 60°F water into reservoir B is 3.3 cfs. Find the
surface elevation of reservoir B: (a) using only a basic scientific calculator; (b) using equation solving
software.
Solution
This is a Case 2 problem.
Table A.l for water at 60°F: v = 12.17 X 10-6 ft2/sec.

Pipe: 1 2 3
L, ft 6000 1500 4500
D, ft 1.25 10/12 8/12
e ,ft 0.0004 0.0004 0.0004
L/D 4800 1800 6750
A= D / 4, ft
2 2
1.227 0.545 0.349
e/D 0.00032 0.00048 0.0006

(a) Find the elevation of P by trial and error.

Elevation of P lies between 160 and 250 ft. Calculate V from Eq. (8.56) assuming turbulent flow. Trials:

Elev. P h1 h3 V1 V3 R1 R2 Q1 Q3 ∑ Q Move P?
200 5 40 6.444 4.481 662,00 245,000 7.907 1.564 +3.04 Up
0 0
230 2 70 4.013 5.984 412,00 328,000 4.925 2.088 -0.463 Down
0 0

Interpolation (Pig. 8.28): (230 - Elev. P)/(230 - 200) = 0.463/(0.463 + 3.04); Elev. P = 226.03.

Elev. P h1 h3 V1 V3 R1 R2 Q1 Q3 ∑ Q Move P?
226 2 66 4.412 5.805 453,00 318,000 5,414 2.026 +0.088 Up
4 0

Interpolation (Pig. 8.28): (230 - Elev. P)/(230 - 226) = 0.463/(0.463 + 0.088); Elev. P = 226.64.

Close enough! Note: These adjustments are very suitable for making on a spreadsheet (Appendix C.2).

Q2 3. 3 D2 V 2
V 2= = =6 . 055 fps ; R2 = =416 ,500
A 2 0. 545 v
All three R values are turbulent, so the use of Eq. (8.56) and these results are valid.

Eq. (8.52): f2 = 0.01761; Eq. (8.13): h2 = 18.05 ft

 Elev. B = Elev. P - h2 = 226.64 - 18.05 = 208.59 ft ANS
(b) Using equation solving software such as Excel or Mathcad (Appendixes C.2-C.3), we note that there are
14 governing equations. Two of these are

hI + h3 = 250 - 160 = 90, QI = Q2 + Q3

The remaining 12 equations are the four governing equations (Sec. 8.14) for each of the three pipes.
The complete output for the Excel and Mathcad solutions is provided in Appendixes D.2 and D.3,
respectively.
The solver provides values for the 14 unknowns (three values each for f, V, R, h, and two values for Q)
from which it calculates

Elev. B = 208.53 ft ANS

Note: Observe how simple and clear the Mathcad solution is.

SAMPLE PROBLEM 8.15

With the sizes, lengths, and material of pipes given in Sample Prob. 8.14, suppose that the surface elevations
of reservoirs A, B, and C are 525 ft, 500 ft, and 430 ft, respectively. (a) Does water enter or leave reservoir
B? (b) Find the flow rates of 600F water in each pipe. Use only a basic scientific calculator.
Solution
This is a Case 3 problem. Find the elevation of P by trial and error.
Table A.1 for water at 600F: v = 12.17 X 10-6 ft2/sec.
The tabulated pipe data are the same as for Sample Prob. 8.14. (a) Triali. First, try P at elevation of reservoir
surface B = 500 ft:

Pipe: 1 2 3
h,ft 25 0 70
2 gDh / L, fps 0.579 0 0.817
V,fps(Eq. 8.56) 4.51 0 5.98
Q=AV,cfs 5.53 0 2.09
At J,∑Q = inflow - outflow = 5.53 - 2.09 = 3.44 cfs. This must be zero, so P must be raised (to
reduce Ql and increase Q3); then water will flow into reservoir B. ANS
(b) Trial 2. Raise P. 500 ft < Elev. P < 525 ft. Try P at elevation 510 ft:
h,ft 15 10 80
2 gDh / L, fps 0.449 0.598 0.874
V,fps(Eq. 8.56) 3.46 4.49 6.41
Check R 355,000 307,000 351,000
Q=AV,cfs 4.24 2.42 2.24

At J, ∑Q = 4.24 - 2.42 - 2.24 = -0.42 cfs. By interpolation (using Fig. 8.28),

510−Elev . P 0 . 42
= ;
510−500 0 . 42+3 . 44 Elev. P =5.8.91ft.
Trial 3. Try P at elevation 508.9 ft:
 h,ft 16.1 8.9 78.9
2 gDh / L, fps 0.465 0.564 0.868
V,fps(Eq. 8.56) 3.59 4.19 6.36
Check R 339,000 287,000 348,000
Q=AV,cfs 4.40 2.28 2.22

At J, ∑Q = 4.40 - 2.28 - 2.22 = -0.10 cfs. Close enough!

Note: These repetitive adjustments are very suitable for making on a spreadsheet (Appendix C.2).

Nonrigorous Solutions
If the value of the friction factor (constant f, CHW, or Manning's n) is given for each pipeline, we must use the
nonrigorous head-loss equations of Sec. 8.19. Then, to solve the three cases just discussed, we follow
exactly the same steps as for the rigorous solutions, but instead of using the equations of Sees. 8.12-8.17, we
simply use Eqs. (8.65)-(8.69). Of course, we must first determine the appropriate K and n value for each
pipeline, and select accordingly from Eqs. (8.66)(8.68). Notice, however, that we can easily solve Case 2
directly if n = 2 (DarcyWeisbach or Manning) by writing the known elevation difference, Δ h13 = h1 + h3 =
2 2 2
K Q1 + K Q3 = K (Q + Q )2 + K Q3 , which is a quadratic equation in Q the only unknown.
1 3 1 2 3 3 3,

SAMPLE PROBLEM 8.16

In Fig. 8.27 pipe 1 is 300 mm diameter and 900 m long, pipe 2 is 200 mm diameter and 250 m long, and
pipe 3 is 150 mm diameter and 700 m long. The Hazen-Williams coefficient for all pipes is 120. The surface
elevations of reservoirs A, B, and Care 160 m, 150 m, and 120 m, respectively. (a) Does water enter or leave
reservoir B? (b) Find the flow rate in each pipe. Use only a basic scientific calculator.
Solution
This is a Case 3 problem. Find the elevation of P by trial and error. Use the Hazen-Williams form (8.67) of
Eq. (8.65) in SI units:
10 . 675 L
K= ,n=1. 852
C 1.HW852 D 4. 87

Pipe: 1 2 3
L, m 900 250 700
D, m 0.3 0.2 0.15
L/D 3000 1250 4667
CHW 120 120 120
n 1.852 1.852 1.852
K 480 961 10924

(a) Trial I. First, try P at elevation of reservoir surface B = 150 m:

h, m 10 0 30
Q,m3/s (Eq.8.69) 0.1236 0 0.1414
At J, ∑Q = inflow - outflow = 0.1236 - 0.0414 = 0.0822 m 3/s. This must be zero, so we must raise P (to
reduce Ql and increase Q3), then water will flow into reservoir B. ANS
(b) Trial 2. Raise P 150 m < Elev. P < 160 m, so try P at elevation 155 m:

h, m 5 5 35
Q,m3/s (Eq.8.69) 0.0850 0.0584 0.0450

AtJ, ∑Q = 0.0850 - 0.0584 - 0.0450 = -0.0184 m3/s. This must be zero, so we must lower P. By interpolation
(Fig. 8.28),

155−Elev . P 0 . 0184
= ;
155−150 0 . 0184+0 . 0822 Elev. P = 154.09m

h, m 6 4 34
Q,m3/s (Eq.8.69) 0.0938 0.0518 0.0443 ANS
V,m/s (Eq. 4.7) 1.327 1.649 2.51

At J ∑Q = -0.0023 m3/s. This is close enough.

All Vs are < 3 m/s (Sec. 8.18), so these solutions using Eq. (8.67) are valid.
Note: These adjustments are very suitable for making on a spreadsheet (Appendix C.2).

8.30PIPES IN SERIES
The discussion in Sec. 8.27 addressed the case of a
single pipe of constant diameter. If a pipeline is made
up of lengths of different diameters, as in Fig. 8.29,
conditions must satisfy the continuity and energy
equations, namely:
Q=Q1 =Q2 =Q3 =.. . .. .. .
h L=h L1 +h L2 +h L3 +.. .. . .. (8.87)
If we are given the rate of discharge Q, the problem is straightforward. We may find the head loss directly
by adding the contributions from the various sections, as in Eq. (8.87); as discussed in Sec. 8.14, hL =
Δ Q. If we are given empirical coefficients or constant fvalues, we can do the same thing using Eq. (8.65)
and the appropriate values of K and n selected from (8.66)-(8.68). If, however, we are given the pipe
material or e, the result will be more accurate, because it uses the rigorous Darcy-Weisbach approach. Then
we use Eq. (8.13) to find the individual head loss contributions after finding e/D, V, R, and f for each pipe.
If we are given the total head loss hL and want to find the flow, the problem is a little more involved.
Using the nonrigorous equations, we again substitute Eq. (8.65) into Eq. (8.87), to get
n n n
h L=h f =K 1 Q 1 + K 2 Q2 + K 3 Q 3 +. . .
But since all the Qs are equal from Eq. (8.86), this becomes
n n
h L=(K 1 + K 2 + K 3 +. .. )Q =( ∑ K )Q (8.88)
So, knowing the pipe information and which nonrigorous equation we must use, we can solve for Q. Last,
we must check that all velocities are in the required ranges for the empirical equations to be valid (Sec.
8.18).

If we wish to use the more accurate, rigorous Darcy- Weisbach approach to find Q, we must note that in Eq.
(8.88) each K has now become a function of a different f. The preferred manual method of solution is
similar to that just discussed, and we call it the equivalent-velocity-head method. Substituting from Eq.
(8.13) into Eq. (8.87) and including minor losses if we wish (usually if L/D < 1000),
2 2
L1 V1 L2 V2
(
h L= f 1
D1
+∑ k 1
2g ) (
+ f2
D2
+∑ k 2
2g
+. .. .
)
2 2 2
Using continuity Eq. (4.17), we know D 1 V 1 =D2 V 2 =D 3 V 3 , etc., from which we can express all the
velocities in terms of one chosen velocity. So, by assuming reasonable values for each f (e.g., from Eq.
(8.54) or Fig. 8.11), for any pipeline, however complex, we can write the total head loss as

2
V
h L=K
2g (8.89)
where V is the chosen velocity. We can solve this equation for the chosen V, and so can obtain the V and R
and f values for each pipe. For better accuracy, we should replace the assumed values of f by the values just
obtained, and obtain an improved solution. When the f values converge V is correct, and we can calculate Q.

SAMPLE PROBLEM 8.17

Suppose in Fig. 8.29 the pipes 1, 2, and 3 are 300 m of 300 mm diameter, 150 m of 200 mm diameter, and
250 m of 250 mm diameter, respectively, of new cast iron and are conveying 15°C water. If Δ z = 10 m,
find the rate of flow from A to B using only a basic scientific calculator. Neglect minor losses.
Solution
Table 8.1 for cast-iron pipe: e = 0.25 mm = 0.00025 m.
Table A.1 for water at 15°C: v = 1.139 X 10-6 m2/s.

Pipe: 1 2 3
L, m 300 150 250
D, m 0.3 0.2 0.25
e/D 0.000833 0.00125 0.00100
fmin (Fig. 8.11) 0.019 0.021 0.020

Assuming these friction factor values,

2 2 2
300 V 1 150 V 2 250 V 3
Δz=h L=h f =10=0 .019 ( )
0. 3 2 g
+0 . 021
0.2 2 g ( )
+ 0. 020
0 . 25 2 g ( )
2 2 4 2 2
V 2 V 1 D1 V 1 0 .3 4 V

From continuity,
=
2 g 2 g D2
=
( ) ( )
2 g 0 .2
=5 . 06 1
2g

2 2
V3 V
=2 . 07 1
Similarly, 2g 2g

2
V 1000 750 1000
and thus 2g (
10= 1 0. 019
0
+0 . 021
1
5 . 06+0 . 020
1
2. 07 )
2
V1
=0 . 0713 m
from which 2g

So
V 1 =√2( 9.81m/s 2 )(0 .0713 m)=1 .183 m/ s

The corresponding values of R are 0.31 x 106, 0.47 X 106, and 0.37 x 106; the corresponding friction factors
are only slightly different from those we originally assumed, since the flow occurs at Reynolds numbers
very close to fully-rough pipe flow. So
Q= A1 V 1=1/ 4 π (0 .30 )2 1. 183=0. 0836 m3 /s ANS

Note: We would have obtained greater accuracy if we had adjusted the friction. factors to match the pipe-
friction chart more closely (further trials) or calculated them by Eq. (8.52), and if we had included minor
losses. In that case Q=0.0821 m3/s.

We can avoid manual iteration for f by solving simultaneous equations using equation solving software like
Mathcad or Excel (Appendixes C.2-C.3 and D.2-D.3). There are the usual four equations for each pipe (Sec.
8.14), plus Eq. (8.87); if necessary, we may easily account for minor losses by using head loss equations
with the form ofEq. (8.80). For the three pipes of Fig. 8.29, for example, there are therefore 13 simultaneous
equations, which we may solve in the usual manner for 13 unknowns (see Sample Prob. 8.14b and
Appendixes D.2-D.3). The unknowns are either the flow rate or the total head loss, and three values each of
hL , V , R, and f.

PIPES IN PARALLEL

In the case of flow through two or more parallel

pipes, as in Fig. 8.30, conditions must satisfy the
following continuity and energy equations:
Q=Q1 + Q2 +Q3 h L=h L1 =h L2 =h L3 (8.91)
because the pressures at A and B are common to all pipes. Problems may be posed in various ways.

If the head loss hL is given, the problem is straightforward. We can find the total discharge directly by
adding the contributions from the various pipes, as in Eq. (8.90). If we are given empirical roughness
coefficients or constant f values, we can do the same thing using Eq. (8.69) and the appropriate values of K
and n selected from (8.66)-(8.68). If, however, we are given the pipe material or e, the result will be more
accurate, because it uses the rigorous Darcy- Weisbach approach. Then we have an independent Type 2
problem for each pipe (see Secs. 8.14-8.17), which we can solve directly by Eq. (8.56) for example.

If we are given the total flow Q and want to find the head loss and individual flows, the problem is a little
more involved.

Using the empirical equations, and neglecting minor losses, we again substitute Eq. (8.69) into Eq.
(8.90), to get
1/ n 1 /n 1 /n
h h h
Q= f 1
K1 ( ) ( ) ( ) + f2
K2
+ f3
K3
+.. . .

But since all the hfs (= hLs) are equal from Eq. (8.91), this becomes
1/n 1/n 1/n 1 /n
1 1 1 1
1 /n
Q=(hf )
[( ) ( ) ( ) ]
K1
+
K2
+
K3
+. .. . =(h f ) 1/n
∑( K ) (8.92)

So, knowing the pipe information and the empirical equation we must use, we can solve for hf' We can then
find the individual flows using Eq. (8.69). Last, we must check that all velocities are in the required ranges
for the empirical equations to be valid (Sec. 8.18).
If we wish to use the rigorous Darcy- Weisbach approach to find hL and the individual Qs, we must
note that in Eq. (8.92) each K has now become a
function of a different f. The preferred manual method of solution is similar to the preceding. Writing Eq.
(8.13) for each line, including minor losses if we wish,

2
L V
h L= f( D
+∑ k
2g )
where ∑k is the sum of the minor-loss coefficients, which we can usually neglect if the pipe is longer than
1000 diameters. Solving for V and then Q, we obtain the following for pipe 1:

2 gh L
Q1 =A 1 V 1 =A 1
√ f 1 ( L1 / D1 )+ ∑ k 1
=C1 √h L
(8.93)

where Cl is constant for the given pipe, except for the change in f with Reynolds number. We can similarly
express the flows in the other pipes, using assumed reasonable values of f from Fig. 8.11 or Eq. (8.54).
Finally, Eq. (8.90) becomes

Q=C 1 √ h L+C 2 √ h L +C3 √ h L =(C 1 +C 2 +C3 ) √ h L (8.94)

This enables us to find a first estimate of hL and the distribution of flows and velocities in the pipes. Using
these, we can next make improvements to the values of f, if indicated, and if necessary repeat them, until we
finally obtain a correct determination of hL and the distribution of flows.

If we use the turbulent-flow equation (8.51) or (8.52) to obtain f, we must remember to confirm that the
Reynolds number is in the turbulent range. We can precheck the likelihood of laminar flow occurring in any
of the pipes by calculating an "average" flow velocity from the total flow divided by the total area of all the
pipes, and using this velocity to obtain an indicator R for each pipe.

Sample Problem 8.18 Three pipes A, B, and C are interconnected as in Fig. S8.18. The pipe
characteristics are as follows:

Pipe: D (in) L (ft) f

A 6 2000 0.020
B 4 1600 0.032
 C 8 4000 0.024

Find the rate at which water will flow in each pipe. Find also the pressure at point P. All pipe lengths are
much greater than 1000 diameters, so neglect minor losses.

Solution
2 2 2
2000 V A 4000 V C V
0+200+ 0−0 . 020 −0. 024 =0+50+ C
Eq. (5.28) 6/12 2 g 8/12 2 g 2g

2 2
V V
150=80 A +145 C
i.e., 2g 2g (1)

Continuity: 62 V A +4 2 V B =8 2 V C

i.e.,
36V A +16 V B =64 V C (2)
2 2
2000 V A 1600 V B
h fa=hfb =0 . 020 =0. 032
Also, 6/12 2 g 4 /12 2 g

i.e., 80 V 2A =153 . 6 V 2B , V B =0. 722 V A

Substituting into (2):

36 V A +16(0 .722V A )=64 V C

47.5V A =64 V C , V A=1.346V C
Substituting into (1):

2 2 2
(1 .346 V C ) V V
150=80 +145 C =289. 9 C
2g 2g 2g
2
V C =2(32 .2 )150/289 . 9=33 . 3
V C =5 .77 fps ,QC = A C V C =(0 .349 )5 . 77=2 . 01 cfs ANS
V A=1 .346 V C =7 .77 fps ,Q A =(0 .1963)7 .77=1 .526 cfs ANS
 V B=0. 722 V A =5 . 61 fps

Q B = A B V B=(0 .0873 )5 . 61=0 . 489 cfs ANS

As a check, note that QA + QB = QC is satisfied.
To find the pressure at P:
2
V Pp
0+200+ 0−80 A = +120
Eq. (5.28): 2g γ
Pp (7 .77 )2
=80−80 =5 . 01 ft
γ 2(32. 2)

2 2
Pp V V
120+ −144 c =50+ c
Check: γ 2g 2g
Pp (5 .77 )2
=145 −70=5. 01 ft
γ 2(32. 2)
So Pp/γ=5 . 01 ft and Pp=(62. 4 /144 )5 . 01=2.17 psi. ANS
Note: In this example we were given the values of f for each pipe as known. Actually f depends on R [Fig.
8.11 or Eq. (8.52)]. Usually we know or assume the absolute roughness e of each pipe, and achieve an
accurate solution by trial and error until the fs and Rs for each pipe have converged.

We can avoid manual iteration for f by solving simultaneous equations using equation solving software as in
Mathcad or Excel (Appendixes C.2-C.3 and D.2-D.3). There are the usual four equations for each pipe (Sec.
8.14), plus Eq. (8.90); if necessary, we may easily account for minor losses by using head loss equations
with the form of Eq. (8.80). For the three pipes of Fig. 8.30, for example, there are therefore 13
simultaneous equations, which we may solve in the usual manner for 13 unknowns (see Sample Prob. 8.14b
and Appendixes D.2-D.3). The unknowns are either the head loss or the total flow rate, and three values
each of Q, V, R, and f.

It is instructive to compare the solution methods for pipes in parallel with those for pipes in series. The role
of the head loss in one case becomes that of the discharge rate in the other, and vice versa. Students should
be already familiar with this situation from the elementary theory of dc circuits. The flow corresponds to the
electrical current, the head loss to the voltage drop, and the frictional resistance to the ohmic resistance. The
outstanding deficiency in this analogy occurs in the variation of potential drop with flow, which is with the
first power in the electrical case (V = IR) and with the second power in the hydraulic case (hL α V2 α
Q2) for fully developed turbulent flow.

PIPE NETWORKS
In municipal distribution systems, pipes are frequently interconnected so that the flow to a given outlet may
come by several different paths, as in Fig. 8.31. As a result, we often cannot tell by inspection which way
the flow travels, as in pipe BE. Nevertheless, the flow in any network, however complicated, must satisfy
the basic relations of continuity and energy as follows:
1. The flow into any junction must equal the flow out of it.
2. The flow in each pipe must satisfy the pipe-friction laws for flow in a single pipe.
3. The algebraic sum of the head losses around any closed loop must be zero.

Most pipe networks are too complicated to solve analytically by hand using rigorous (variable f) equations,
as was possible in the simpler cases of parallel pipes (Sec. 8.31). Nowadays they are readily solved by
specially developed computer programs (Appendix C.S). However, in many cases we cannot predict the
capacity requirements of water distribution systems with high precision, and flows in them vary
considerably throughout the day, so high accuracy in calculating their flows is not important. As a result, the
use of nonrigorous equations (Sees. 8.18-8.19) are very acceptable for this purpose. The method of
successive approximations, due to Cross,27 is such a method that was popular before the advent of
computers. We will review it here to help students understand the fluid mechanics of pipe networks and
evaluate computer-generated solutions. It consists of the following elements, in order:
Step 1: By careful inspection assume the most reasonable distribution of flows that satisfies condition 1.
Step 2: Write condition 2 for each pipe in the form

h L=KQ n (8.95)

where K and n are constants for each pipe as described in Sec. 8.19. If minor losses are important include
them as in Eq. (8.93), which yields K = 1/C2 and n = 2 for constant f We may include minor losses within
any pipe or loop, but must neglect them at the junction points.
Step 3: To investigate condition 3, compute the algebraic sum of the head losses around each elementary
loop, ∑ hL = ∑.KQn. Consider losses from clockwise flows as positive, counterclockwise negative.
Only by good luck will these add up to zero on the first trial.
Step 4: Adjust the flow in each loop by a correction Δ Q to balance the head in that loop and give ∑KQn =
0. The heart of this method lies in the following determination of Δ Q. For any pipe, we may
write
Q = Q0 + Δ Q
where Q is the correct discharge and Qo is the assumed discharge. Then, for each pipe,
 h L=KQ n =K (Q 0 + ΔQ )n= K (Q n−1
0 ΔQ +.. . .)

If Δ Q is small compared with Qo, we may neglect the terms of the binomial series after the second one,
so that
h L=KQ n0 + ΔQKnQn−1
0

For a loop, ∑hL = ∑KQn = 0, so because Δ Q is the same for all pipes in that loop,

∑ KQ n0+ ΔQ ∑ KnQ n−1

0 =0
As we must sum the corrections of head loss in all pipes arithmetically (treating all terms as positive), we
may solve this equation for Δ Q,

−∑ KQ0 |Q n−1
0 | −∑ h L
ΔQ= n−1
=
n ∑|KQ 0 n ∑ |hL /Q 0 | (8.96)
n-l
since, from Eq. (8.95), hL/Q = KQ . We emphasize again that we must sum the numerator of Eq. (8.96)
algebraically, with due account of each sign, while we must sum the denominator arithmetically. Note that
n−1
the Q0|Q 0 | in the numerator gives this quantity the same sign as the head loss. The negative sign in Eq.
(8.96) indicates that when there is an excess of head loss around a loop in the clockwise direction, we must
subtract the Δ Q from clockwise Qo values and add it to counterclockwise ones. The reverse is true if
there is a deficiency of head loss around a loop in the clockwise direction.
Step 5: After we have given each loop a first correction, the losses will still not balance, because of the
interaction of one loop upon another (pipes which are common to two loops receive two independent
corrections, one for each loop). So we repeat the procedure, arriving at a second correction, and so on, until
the corrections become negligible.

We may use either form of Eq. (8.96) to find Δ Q. As values of K appear in both the numerator and
denominator of the first form, we can use values proportional to the actual K to find the distribution. The
second form is more convenient for use with pipe-friction diagrams for water pipes.

An attractive feature of this approximation method is, that errors in computation have the same effect as
errors in judgment and the process eventually corrects them.

SAMPLE PROBLEM 8.19

If the flow into and out of a two-loop pipe system are
as shown in Fig. S8.19, determine the flow in each
pipe using only a basic scientific calculator. The K
values for each pipe were calculated from the pipe
and minor loss characteristics and from an assumed
value of f, and n = 2,
Solution As a first step, assume a flow in each pipe
such that continuity holds at all junctions. Take
clockwise flows as positive. Calculate Δ Q for each
loop, make corrections to the assumed Qs, and repeat
several times until the Δ Qs are quite small.

Further corrections can be made if greater accuracy is desired.

L, As noted earlier, varying demand rates usually make high solution accuracy unnecessary with pipe
networks. However, if high manual accuracy is required for some reason, we can first solve the problem in a
similar manner to the preceding example using the Darcy-Weisbach K in Eq. (8.65) and constant f values.
Then we can use the resulting flows to adjust the f and K values, and repeat the process (more than once if
necessary) to refine the answers. The value of such refinement is questionable, not only because of
uncertainties in the demand flows, but also because of uncertainties in the e values (pipe roughness) we
must use (see Sec. 8.15). Usually when we adjust f values they change by only a few percent, but we can see
in Fig. 8.11 that for smoother pipe it is possible for them to change by as much as a factor of five.

We can solve simple networks without approximation and manual iteration by solving simultaneous
equations using equation solving software like that in Mathcad and Excel (Appendixes C2-C3 and D.2-D.3).
For networks containing i pipes, 5i equations are required if using the Darcy- Weisbach equation with
variable f, and 2i equations are required if using the simplified Eq. (8.95) with constant friction factors.
These required equations include (a) the usual (condition 2) flow equations for each pipe (four or one per
pipe, depending on the equations used); (b) flow continuity equations (condition 1) at all but one of the j
nodes (as these imply continuity at the last node); (c) equations for the sum of the head losses around i - j +
1 loops (condition 3). The unknowns we want to find for each pipe are hL Q, V, R, and fusing the Darcy-
Weisbach equation, or only hL and Q using Eq. (8.95).

The pipe-network problem lends itself well to solution by use of a digital computer. Programming takes
time and care, but once set up, there is great flexibility and it can save many hours of repetitive labor. Many
software packages are now available to simulate water distribution networks; see Appendix C