You are on page 1of 12

Computers

and electronics
in agriculture
ELSEVIERL Computers and Electronics in Agriculture 13 (1995) 75-86

Analyzing water distribution systems for swine growing


and finishing units
Rabi H. Mohtar a,*, Larry J. Segerlind b, Howard L. Person b
a Department of Agricultural and Biological Engineering, The Pennsylvania State Universiry,
University Park, PA 16802, USA
b Department of Agricultural Engineeting, Michigan State University, East Lansing, MI 48824, USA

Accepted 10 March 1995

Abstract

An analysis procedure for livestock water distribution systems with nipple waterers
is presented. The procedure uses the finite element analysis techniques to calculate the
pressuredistribulion in a systemwith severalbranchesand nipple waterers.The procedure
also optimize the systemfor flow uniformity and cost. The formulation allows the use of
the drop inlet as a dischargepoint to reduce the size of the solution matrix. The discharge
parameterswere developedat this inlet usingwaterer flow characteristics.The flow rate at
each nipple waterer can be then calculatedbasedon the drop inlet pressure.The solution
for a typical network is presented.The method can be usedasthe basisfor the designand
evaluation of water distribution systemsin swinegrowingand finishingunits.

Keywords: Pipenetwork design and analysis;Water distribution systems;Nipple waterer;


Livestock housingwater systems.

1. Introduction

A water distribution system is an integral part of livestock housing. Delivering


adequate quantities of water is a fundamental requirement for providing optimal
environmental conditions for animal housed in confinement facilities. Water stress
can include significant losses in the meat production as well as losing immunity
against certain diseases. Celis (1988) found that the rate of gain, feed intake and
feed conversion of growing pigs were significantly decreased if water flow was below
70 ml/min. Consequently, It is the responsibility of the designer to provide the
adequate water flow rate to all animals all the time at low cost. Significant savings

* Corresponding author. Fax: (814) 863-1031

0168-1699/95/$09.50 0 1995 Elsevier Science B.V. All rights reserved.


SSDI 0168-1699(95)00016-X
76 R.H. Mohtar et al. I Computers and Electronics in Agriculture 13 (1995) 75-86

in initial system and operating cost are achieved by proper system design. High
flow uniformity increases the production by reducing the water stress. Uniform
water distribution becomes an important issue. In order to maintain the minimum
adequate water supply in the entire livestock housing, some discharge outlets will
have higher than the minimum. In pipe network design, there is a trade-off between
flow uniformity and system capacity. For a water supply of constant flow rate
and pressure head, as the pipe sizes increase, the uniformity increases due to the
reduced friction losses in the pipes. On the other hand, larger pipe sizes increase the
initial system cost and therefore reduce profitability. Proper designs for livestock
water distribution systems will reduce the sum of initial fixed and maintenance cost
of the water distribution and ensure high uniformity of flow. This creates a need for
a procedure that allows engineers to adequately design livestock watering systems.
Water distribution systems are presently analyzed and designed using the finite
element method, Bralts and Segerlind (1985) Haghighi et al. (1989, 1992) and
Mohtar et al. (1991). These analysis and design procedures are applicable to
farmstead water systems but their application for this type of water systems has not
yet been presented.
The objective of the study is to develop an analysis procedure to assist in
designing water distribution systems that include nipple waterers commonly used in
swine growing and finishing units. The procedure should allow the designer to check
flow uniformity upon changing any system parameter and also evaluate already
existing system for their water flow uniformity and adequacy.
Fig. 1 represents a typical layout of a livestock water distribution system. It
comprises of a water source, main, submains, laterals, and drop inlets. A demand
node is defined as a node where water is being discharged. Each of these demand
nodes represents a drop inlet. The pipe segment connecting two nodes is defined as

Lateral
Submain

Fig. 1. Typical livestock water distribution layout.


R.H. Mohtar et al. /Computers and Electronics in Agriculture 13 (1995) 75-86 77

1
Lateral

Fig. 2. Typical drop unit. Bracketed numbers refer to element numbers while, non-bracketed numbers
refer to node numbers.

Water Source

Demand Node

Junction Node

Fig. 3. Example network layout. Bracketed numbers refer to element numbers while, non-bracketed
numbers refer to node numbers.

an element. Fig. 2 represents a drop inlet feeding four nipple waterers. Fig. 3 is an
example of a livestock farm distribution network. Solid nodes represent drop inlet
(demand nodes) and hollow nodes are junction nodes (non-demand nodes). The
sample network solutions will use the examples of Figs. 2 and 3 but the procedure is
general and can be implemented on any layout.

2. Theoretical development

2.1. Finite elementformulation forpressure head distribution

The element matrices were derived from the basic pipe flow energy equations
to a standard finite element form. Standard finite element procedure was then
employed to solve for the pressure distribution in the network. This section presents
78 R.H. Mohtar et al. I Computers and Electronics in Agrkulture 13 (1995) 75-86

a summary of this development. The pressure distribution in any hydraulic pipe


network is governed by the Bernoulli’s equation for mechanical energy conservation.
Applying this equation between any two nodes i and j in the network as upstream
and downstream nodes respectively yields
2

Zi+H~+$+h,=Z,+Hj+~+hl (1)

where i and j are any two nodes or junctions in the flow network, Zi and Zj are
elevations at nodes i and j (L), Hi and Hj are pressure heads at nodes i and j (L),
Vi and V, are velocities (L/T) at nodes i and j respectively, h, is the mechanical
energy input along path ij (L), hl is the frictional head losses along path ij (L), g is
the gravitational acceleration (LD?).
The general energy equation for computing head losses in a pipe network, hi was
developed by Mohtar et al. (1991) for all the components of the water distribution
network of Fig. 1 and is written as:
hl = k,Qm (2)
where h, is the head loss or gain (L) through the pipe network components: straight
pipe, pipe fitting, pump, or nozzle, kr is a component specific coefficient, m is a
component specific exponent, and Q is the volumetric flow rate (L3/I’). Values for k,
and m for different components were given by Mohtar et al. (1991).
Substituting (2) in (1) and using the flow continuity equation, Q = Ai Vi, where
Ai is the cross-sectional area (L2) of the pipe at node i and Vi is the flow velocity
(L/T) at node i, yield

(Zi - Z,;) + (Hi - Hj) = klQm - $(VF - Vy) -h,

or
Q = (Zi - Zj)C + (Hi - Hj)C (4)
where
1
C=
k,Q”-‘-%+&-hm
2gA; %A; Q
is a linearizing coefficient.
For any element [e} connecting nodes i and j, and by defining flow coming into a
node as negative while flow leaving a node as positive, the following nodal continuity
equations for mass conservation could be written as
R.H. Mohtar et al. I Computers and Electronics in Agriculture 13 (1995) 75-86 19

-Qin, + Qout,= 0
(5)
-Qin, + Qautj = 0
where Qin; and Qinj and are incoming flow to nodes i and j, Q,,,; and Qou, are
flows leaving nodes i and j.
The contributions of element {e} to (5) are Qouti and Qoutj which will be labeled
Q!“’ and Q!“l in later equations for consistency with the literature. Writing element
contribution to the above nodal equations and substituting for the flow as in (4) and
writing the results in a matrix form yield

(6)

where Q!e’ is the flow contribution of element {e} to node i (L3/T), Qy’ is the
flow contribution of element (e} to node j (L3/T), AZ is the change in elevation
(Zi - .Zi) (L),
[ 1
_‘, -,” is the element stiffness matrix, is the element force

vector, is the unknown nodal pressure head, C is the element coefficient from

the linearization of the energy (4).


Eq. (6) has the standard finite element form. It is derived for a single element
and can be written for all elements in the system. The set of element equations is
arranged to form a global system of equations using the direct stiffness procedure
(Segerlind, 1984). Direct stiffness algorithm yields a system of equations which have
the general matrix form

(7)
where {H} is the unknown nodal pressure head vector, [K] is the global stiffness
matrix, {F] is the global force vector.
The global system of equations in (7) is solved for the unknown heads at all
nodes. For more details in the procedure and formulation, refer to Haghighi et al.
(1989) and Mohtar et al. (1991).
The nodal pressure head values computed in (7) are at nodes representing drop
inlets feeding a series of the nipple waterer, Figs. 1 and 3. The pressure at each of
the waterers in Fig. 2 is computed using the following relation

h = Hi - h (drop) + El (8)

where h is the nodal pressure at the waterer (L) (nodes 7,8,9, and 10 in Fig. 2), Hi
is the total head at drop inlet (L) (node 1 in Fig. 2, demand nodes of Figs. 1 and 3)
hi (drop) is the sum of friction head losses of all elements from drop inlet node to the
waterers (L), El is drop length or the distance from drop inlet to the waterers, i.e.
elevation difference between drop inlet and waterer.
80 R.H. Mohtar et al. /Computers and Electronics in Agriculture 13 (199.5) 75-86

2.2. Nodal flow distribution

The outflow from any discharge nozzle is dictated by the head at the discharge
point. The nodal flow is related to the nodal head by the following relation
q = kh; (9)
where q is the flow at node i (L3D’), hi is the waterer head at node i (L), a
and k are exponent and coefficient of discharge respectively and are determined
experimentally.
If the drop unit is treated as a discharge nozzle, applying the orifice flow (9) to
any drop inlet node i in Fig. 1 yields
Q=KHi” (10)
where Q is the flow out of the drop inlet at node i, Hi is the head at drop
inlet at node i (L), K is the drop coefficient of discharge, a is the drop exponent
of discharge. Because the flow exponent changes only with the type of the flow
resistance at the discharge nozzle, the exponent a is the same in (9) and (10).
For a symmetrical drop with similar waterer characteristics, Q is related to a
single waterer discharge q by the following equation
Q = nq (11)
where IZ is the number of waterers per drop.
Assuming that the sum (-hl (drop)+ El) in (8) is negligible compared to the inlet
pressure Hi, then hi and Hi are assumed equivalent. Therefore, for the drop unit in
Fig. 2 and for the series of n waterers, the K term of (10) could be written as
K = nk (12)
The above assumption that hi and Hi are equivalent is only used to derive (12).
Eq. (8) is used throughout the analysis to compute the pressure head at the waterer
from the pressure head at the drop inlet.
The finite element formulation utilizes the orifice flow equation, (lo), which
solves for the variable outflow as a function of the pressure head. The coefficient
and the exponent of flow are considered constants for a given drop unit. The
formulation incorporates (10) into the analysis by treating the drop unit as a
discharge point and as a separate element. The linearized discharge coefficient,
c,,,,le, is added to the diagonal value of the stiffness matrix in (7), [Kii] where i
is the node number, at the demand node. The linearized discharge coefficient is a
positive non-zero linear function of the discharge coefficient, K, and exponent, a, of
(lo), Mohtar et al. (1991). The value of cnozzlereduce to zero for the nodes that do
not connect to any drop units (non-demand nodes).

3. Flow uniformity

The previous sections presented a procedure to determine the nodal pressure


and flow distribution in a pipe network. The two parameters are related by the
R.H. Mohtar et al. /Computers and Eiectronics in Agriculture 13 (1995) 75-86 81

Table 1
Physical parameters for the typical drop in Fig. 2

Element Node i Node j Description

(1) I 2 4” Galvanized steel pipe


(7-l 2 3 i” Rubber pipe
(3) 3 4 4” Galvanized steel pipe
(4) 4 6 1” Galvanized pipe
(5) 4 5 i” Galvanized pipe
(6) 5 I 4“ Rubber pipe
(7) 5 8 i” Rubber pipe
(8) 6 9 4” Rubber pipe
(9) 6 10 g” Rubber pipe
Nodes 7, 8, 9, and 10 have waterers.

Table 2
Physical parameters and equations for the example network of Fig. 3

Parameters Elements
Pipe length = 30.48 m (100 ft) (l)-(3), W(8), (lTj-(Wc (9, (lo), and (15)
Pipe size = SO.8 mm (2”) ’ (L)-(3), (6)-(g), (If)-(13% 4% (W’and (15)
Hazen-William’s coefficient, CR-w = ll.50 (1)-(3X (6&(g); WbU%f% P), altall
Flow exponent (a) = 0.5
Flow coefficient (k) = 1.33 em4 (for a single waterer)
Equations:
4 13 Q’.852
Hazen-William’s pipe friction: ht = ’ D4.9871 L
C-$2

where CH-w is a roughness coefficient, D is a pipe diameter (L)


Orifice flow: 4 = kh”.

orifice flow equations, (9) and (10). Uniform pressure distribution is needed to
ensure uniform flow at the nipple waterer. Excess pressure at the waterer leads
to a disruption of the uniformity and a high initial pipe network cost, Mohtar et
al. (1991). Uniform flow in a livestock housing and many other agricultural pipe
networks is desirable and lead to a reduction in the tied cost of the system. In order
to maintain the minimum pressure needed for adequate flow at all nodes of the
network, some nodes will receive excess pressure. Uniformity can be achieved by a
proper design and pipe size selection to consume this excess pressure. For a certain
pressure range installing pressure compensating waterers can add to the uniformity.
The major friction losses in a pipe network are generated in a straight pipe
section, although losses through ofher components in a livestock water distribution
network can be significant. For a certain network layout, reducing the pipe sizes
leads to :an increase in frictional losses (refer to the Hazen-Williams equation in
Table 2). Mohtar et al. (1991) presented a formulation to design and optimize the
82 R. H. Mohtar et al. I Computers and Electronics in Agn’culture 13 (1995) 75-86

network. The philosophy behind an optimized network is to deliver the required


amount of water at no excess pressure. The approach was to trim down on pipe sizes
and consume the excess pressure head in frictional losses. The same formulation
was used in the current study to design and optimize and re-size the network.

3.1. Solution procedure

A comprehensive program for the design and optimization of water distribution


networks, ANALYZER, was developed by Mohtar et al. (1991). The program uses
the finite element method as a numerical tool. ANALYZER was modified to include
peculiarities that are special to a swine housing water distribution system.
The demand nodes in the network shown in Fig. 3 correspond to inlets to the
drop units. All the drop units are lumped into a nodal discharge at the drop inlet.
Since all drops are usually identical and behave similarly, the output corresponding
to Fig. 3 will be used for uniformity evaluations.
The global linear system (7) solves for the pressure distribution in the network.
The flow in each element is calculated from the head values using (4) and (10).
The problem is non-linear since the coefficients are functions of the unknown flow,
therefore, the problem is solved iteratively and convergence is achieved within a
certain prescribed tolerance. After the flow parameters were solved at the drop
inlets, the pressure and flow at the discharge nozzle were traced and computed
using (8) and (9) respectively. Initial pipe sizes that will minimize the operating cost
of the system are determined. Excess pressure is calculated at each waterer and
the corresponding drop inlet is found. ANALYZER will try to down size the pipes
leading to that inlet until the pipe size that satisfies the desired flow uniformity
is found. Consequently, another set of iterations were needed re-size the network
links, and insure flow uniformity. The critical path concept was used in the search
for optimal pipe sizes. For a detailed description on the optimization procedure
refer to Kamand et al. (1987) and Mohtar et al. (1991).

3.2. Example network

The example network in Fig. 3, with corresponding physical parameters in


Table 2, was solved using the modified ANALYZER. Nodes 1, 2, 3, 6, 7, 8, 12, 13
and 14 are nodes feeding drops similar to the one shown in Fig. 2. Tables 3 and 4
show the results of the element flow rate and the nodal pressure for this example
network.
The distribution system layout and the waterer specifications, exponent and
coefficient of discharge and the number of waterers per drop, are user supplied
input. Fig. 4 shows an experimentally determined response of waterer flow rate as
affected by inlet pressure, (Celis, 1988). The curves generated follow the orifice
flow equation in (9). The parameters a and k in (9) were determined using an
exponential regression fit to the experimental data points of Fig. 4. Results of the
regression for a and k are shown in Table 2. These were used to evaluate the flow in
the network as described earlier.
R.H. Mohtar et al. i Computers and Electronics in Agriculture 13 (1995) 75-86 83

Table 3
Results of element flow rates for the example network of Fig. 3

Element Nodes Flow (I/s)


2 1 0.079
2 3 2 0.156
3 4 3 0.238
4 5 4 0.238
5 10 5 0.238
6 I 6 0.082
7 8 I 0.167
8 9 8 0.252
9 11 10 9 0.238 0.252
IO 16 11 0.493
II 13 12 0.099
I? 14 13 0.201
13 15 14 0.306
14 17 16 15 0.493 0.306
15 18 17 0.799

Table 4
Results of nodal pressure heads for the example network of Fig. 3

Node number Pressure head Cm)


1 8.13
2 8.13
3 8.96
4 10.37
5 10.39
6 9.03
I 9.45
8 10.17
9 11.75
10 11.81
11 11.8
12 13.26
13 13.85
14 14.89
15 17.14
16 17.21
I7 17.22
18 30.48
Elevation is zero for all nodes.

The critical nodes 1, 6, and 12 in Fig. 3 have a total head of 8.13, 9.03 and 13.26
m, respectively. These are defined as nodes with the highest pressure requirement
for a particular water flow path. Corresponding to that flow rates through critical
elements (1) (6) and (11) are 0.079, 0.082 and 0.099 l/s respectively. Assuming
all nodes have the same flow and pressure requirements of 0.079 l/s at 8.13 m of
pressure head, these values correspond to 11% and 60% variation in pressure and
84 R.H. Mohtar et al. /Computers and Electronics in Agriculture 13 (1995) 75-86

0.0800 --

0.0666
z

P 0.0530

?
2 0.0400

5
k 0.0266

PRESSURE, (kPa)
Fig. 4. Pressure flow rate relations for the water used in Fig. 2

3.8% and 25.3% variation in flow at nodes 6 and 1 respectively compared to node
1. The above situation is a major issue in uniformity evaluation and need to be
corrected. Optimizing the above results means consuming 0.9 m of pressure head
in the path from node 9 to node 6 and 5.13 m in the path from node 15 to node
12. This process of trimming is performed using ANALYZER using the critical path
concept and is described,by Mohtar et al. (1991). The pipe size availability and the
price per unit length are critical factors affecting the final pipe sizes in the network.
Using waterers with adjustable setting (Celis, 1988), pressure compensating
waterers, can also be used to improve flow uniformity. The problems with this
approach are, first, there is no reduction in fixed costs, second, most waterers have a
narrow pressure range where outflow is not affected by the inlet pressure.
The pressure distribution within the drop unit is determined from the drop
inlet pressure using (8). The frictional head losses hl for the waterer at node 10
of Fig. 2 is the sum of frictional losses in elements (1) (2) (3) (4) and (9).
Similarly, the frictional losses through paths leading to waterers at nodes 7, 8, and
9 are determined by summing the frictional losses through the corresponding water
path from the drop inlet to the waterer discharge point. For a symmetrical drop
layout and similar waterer characteristics, the following equations are established to
compute the head loss in the drop of Fig. 2

W) hi181
hi(drop) = hII11 + h1{2} + hl{3} + or + or (13)
hI51 hU’1
or
R.H. Mohtar et al. I Computers and Electronics in Agriculture 13 (1995) 75-86 85

Table 5
Results of element flow rates for the typical drop of Fig. 2

Element Nodes Flow rate (l/s)


1 1 2 0.12
2 2 3 0.12
3 3 4 0.12
4 4 6 0.06
5 4 5 0.06
6 5 7 0.03
I 5 8 0.03
8 6 9 0.03
9 6 10 0.03

Table 6
Results of nodal pressure heads for the typical drop of Fig. 2

Node number Elevation (m) Pressure head (m) Total head (m)
1 3.05 15.24 18.29
2 2.44 15.84 18.28
3 1.83 16.28 18.09
4 1.22 16.86 18.08
5 0.61 17.47 18.08
6 0.61 11.41 18.08
I 0.0 17.56 17.56
8 0.0 17.56 17.56
Y 0.0 17.56 17.56
10 0.0 17.56 17.56

where hi(i) is the frictional losses through element i and determined from (2).
Alternatively, the drop unit of Fig. 2 can be treated as a water distribution
network with a source pressure equal to the head at the drop inlet. Eq. (7) can be
solved for the pressure at the various nodes of the drop unit. The typical drop of
Fig. 2 with corresponding physical parameters in Table 1 was solved as a sample
drop example. Tables 5 and 6 show the element flow rate and nodal pressure for the
typical drop of Fig. 2. Similar output could be established as needed for the various
drops.

3.3. Features of the technique

The finite element method is a systematic numerical procedure for solving


complex engineering problems. The method has been implemented to solve the
hydraulic system for a water distribution system used for various types of livestock.
The presented technique was implemented on an IBM PC and is available in both
languages, Turbo Basic and FORTRAN. In addition to all the benefits of the finite
element method discussed by Haghighi et al. (1989, 1992), the technique is flexible
in handling all livestock housing systems of various configurations and complexity.
86 R.H. Mohtar et al. I Computers and Electronics in Agriculture 13 (1995) 75-86

It will allow the user to change any of the design parameters, particularly pipe
diameter and input source pressure. The goal is to adequately supply water to all
nipple waterers without any excess in pressure, since extra pressure is equivalent
to extra operational cost. The technique is well suited to access optimization
subroutines that will optimize system cost subjected to uniformity constraints.

4. Summary and conclusions

A procedure for livestock water distribution systems analysis is developed and


presented using the finite element method. The element matrices were derived from
the basic pipe flow energy equations. Standard finite element procedure was then
employed to solve for the pressure distribution in the network. The finite element
formulation utilizes the orifice flow equation which solves for the variable outflow
as a function of the pressure head. The coefficient and the exponent of flow are
fixed for a given discharge outlet. A linearized discharge coefficient is added to the
diagonal value of the stiffness matrix. The formulation is illustrated with practical
examples and used in the design of a distribution network and a drop unit. The
formulation allows the use of the drop inlet as a discharge point to reduce the size
of the solution matrix. The discharge parameters were developed at this inlet using
waterer flow characteristics. The flow rate at each nipple waterer is calculated based
on the drop inlet pressure. In addition to the design, the procedure uses the critical
path concept to optimizes the system cost and flow uniformity. The critical path
concept is an iterative procedure that search for the critical path, sub-critical path,
etc. At every iteration the excess pressure at each node are determined. Excess
pressure will be utilized by re-sizing the off-critical path pipes. The procedure goes
on until no more excess pressures exists. The method can be used as the basis for
the design of water distribution systems in swine growing and finishing units.

References

Bralts, V.F. and Segerlind, L.J. (1985) Finite element analysis of drip irrigation units. Trans. ASAE,
28(3): 809-814.
Cehs, J.E. (1988) Effect of water flow rates on performance of nursery pigs and influence of pressure
on flow rate from nipple waterer. M.S. Thesis presented to Michigan State University, East Lansing,
MI.
Haghighi, K., Bralts, J.F., Mohtar, R. and Segerlind, L.J. (1989) Modeling expansion/contraction, valve
and booster pump in hydraulic pipe network analysis. A finite element approach. Trans. ASAE,
32(6): 1945-1955.
Haghighi, K., Mohtar, R.H., Bralts, V.F. and Segerlind, L.J. (1992) A linear formulation model for pipe
network components. Comput. Electron. Agric., 7: 301-321.
Kamand, F.Z., Mohtar, R.H., and Baasiri, M. (1987) A model for the design and optimization of pipe
networks. ASAE Paper No. 87-2614, ASAE, St. Joseph, MI.
Mohtar, R.H., Bralts, V.F. and Shayya., W.H. (1991) A Finite element formulation for the analysis and
optimization of pipe networks. Trans. ASAE, 34(2): 394-401.
Segerlind, L.J. (1984) Applied Finite Element Analysis. John Wiley and Sons, New York.

You might also like