Article

pubs.acs.org/IECR

Incorporating Property-Based Water Networks and Surrounding

Watersheds in Site Selection of Industrial Facilities
Luis Fernando Lira-Barragán,† José María Ponce-Ortega,† Fabricio Nápoles-Rivera,†
Medardo Serna-González,† and Mahmoud M. El-Halwagi*,‡,§
†
Chemical Engineering Department, Universidad Michoacana de San Nicolás de Hidalgo, Morelia, Michoacán, 58060, México
‡
Chemical Engineering Department, Texas A&M University, College Station, Texas 77843, United States
§
Adjunct Faculty at the Chemical and Materials Engineering Department, King Abdulaziz University, Jeddah, Saudi Arabia

ABSTRACT: This paper presents a new mathematical programming approach for including water usage and discharge in the
selection of installation sites for industrial facilities. Water integration within the industrial facility as well as interaction of the
discharged wastewater with the environment and the surrounding watershed are considered simultaneously through a material
flow analysis model. The model tracks the properties that affect the process sinks and the environment (e.g., pH, toxicity, density,
color, chemical oxygen demand, etc.) as well as compositions of targeted components. The model considers all the inlets and
outlets that affect the watershed as well as the interactions between the pollutants and the environment. The objective function is
aimed at minimizing the total annualized cost, which includes the installation cost of the new facility, the transportation of raw
materials, products, and utilities, the land cost, the wastewater treatment costs (including the piping cost), and the fresh sources
cost. The proposed methodology is applied to two case studies in Egypt and Mexico.

1. INTRODUCTION using the framework of property integration proposed by El-

The chemical and process industries are among the largest
consumers of water resources and dischargers of wastewater.
Recently, several techniques have been proposed to econom-
ically reduce the clean water consumption in the process
industry considering the possibility to recycle, reuse, and
regenerate the wastewater produced in the industrial facilities
(see Figure 1). These approaches are based on different
objectives (e.g., minimizing the fresh water consumption,
minimizing the wastewater discharge, minimizing the regener-
ation costs, etc.) and using diverse techniques (e.g., graphical,
algorithmic, and mathematical programming-based ap-
proaches). The heuristic approaches allow the identification
of targets ahead of the design (see for example Wang and
Smith,1 Dhole et al.,2 El-Halwagi and Spriggs,3 Polley and
Polley,4 Hallale,5 Manan et al.,6 El-Halwagi et al.,7 Feng et al.,8
Foo9) or to solve this problem algebraically (Sorin and
Bedard,10 Gomes et al.,11 Foo et al.12). Additionally, different
mass integration techniques based on mathematical program-
ming models have been proposed to obtain the optimal
solution for the water integration problem inside industrial
facilities (see, for example, Takama et al.,13 El-Halwagi et al.,14
Savelski and Bagajewicz,15−17 Alva-Argaez et al.,18,19 Benko et
al.,20 Teles et al.,21 Gabriel and El-Halwagi,22 Kuo and Smith,23
Doyle and Smith,24 Galan and Grossmann,25 Hernandez-Suarez
et al.,26 Gunaratman et al.,27 Karuppiah and Grossmann,28
Putra and Amminudin,29 Ponce-Ortega et al.,30 Nápoles-Rivera
et al.,31 and Faria and Bagajewicz32−40). More recently, for
addressing problems with several pollutants that are based on
stream properties other than contaminant concentration, Ng et
al.,41 Ponce-Ortega et al.,42−44 Nápoles-Rivera et al.,45
Kheireddine et al.,46 and Deng and Feng47 have developed
techniques for water integration inside the industrial facilities
Halwagi and co-workers.48,49
It is worth noting that the aforementioned approaches have
only considered the process activities taking place inside the
industrial facilities, without taking into account the activities
occurring outside the plant for the wastewater streams
discharged to the environment. On the other hand, despite
the increasing number of works that consider environmental
aspects in the synthesis of mass exchange networks (see for
example Ku-Pineda and Tan,50 Tan et al.,51 Lim and Park,52−54
Ponce-Ortega et al.44), these formulations do not take into
account the interrelationships between the wastewater streams
discharged to the environment and the surrounding watershed.
Therefore, the physical, chemical, and biochemical interactions
between the industrial discharges and other discharges and uses
such as agricultural, industrial, residential, precipitation,
filtration, and others have not been considered.
Since the watersheds constitute an important ecological
element through which the discharged industrial, residential
and agricultural effluents are transported to their final disposal
while undergoing physical, chemical, and biological changes, the
watersheds and their surroundings are interrelated as shown in
Figure 2. The discharges received by a watershed change the
chemical composition and the properties of the receiving water
bodies. The discharged species undergo chemical, biochemical,
and physical transformations that are caused by the flora and
fauna of the watersheds (see Brunner et al.,55 Baccini and
Special Issue: L. T. Fan Festschrift
Received: February 12, 2012
Revised: April 23, 2012
Accepted: May 1, 2012
Published: May 1, 2012

© 2012 American Chemical Society 91 dx.doi.org/10.1021/ie3003792 | Ind. Eng. Chem. Res. 2013, 52, 91−107
Industrial & Engineering Chemistry Research
Figure 1. Recycle and reuse networks.
Figure 2. Impact of a new industrial plant in a watershed system.
Brunner,56 Lampert and Brunner,57 El-Baz et al.,58,59 and
Lovelady et al.60). Recently, Lira-Barragán et al.61,62 proposed
two mathematical programming approaches for the optimal
water integration inside the industrial facilities considering the
surroundings through an MFA model. Nonetheless, these
studies were based on the composition of the wastewater
streams without considering stream properties.
This paper presents a mathematical model that overcomes
the drawbacks of the previous plant−watershed networks by
incorporating the property tracking and constraints in addition
to the other environmental, technical, and economic consid-
92
Article
erations. The primary focus is to include the design of water
networks with multiple pollutants and properties (rather than
just contaminant concentrations) in the site selection of new
facilities while incorporating MFA in the surrounding water-
shed. In this problem, a set of sites are considered as candidates
for locating a new plant in a particular watershed. Therefore,
the optimum water network and, consequently, the optimum
flow and quality of the wastewater discharged in a receiving
water body depend on the location selected to install the plant,
which depends on the interactions between the new plant and
the watershed. Thus, the new plant and all sources and uses of
dx.doi.org/10.1021/ie3003792 | Ind. Eng. Chem. Res. 2013, 52, 91−107
Industrial & Engineering Chemistry Research
the watershed are seen as interacting systems rather than
comprising isolated components in an overall system. A
material flow analysis (MFA) model is proposed for the
integrated assessment of water use options in order to
incorporate the sustainability of the watershed in the synthesis
of property-based water networks. A superstructure is
formulated with all the different alternatives for the in-plant
water recovery and the plant location. The mathematical model
is a disjunctive programming formulation that is reformulated
as an MINLP problem considering the minimization of the
total annual cost for the recycle and reuse network and the
installation of the new plant to satisfy the process and
environmental regulations as well as the sustainability of the
watershed system in terms of properties. The model is able to
consider several properties and carry out the selection of several
technologies to treat the process sources. The paper is
organized as follows: section 2 presents the definition of the
problem addressed in this paper, section 3 presents the model
formulation, section 4 presents the results and discussions of
the application of the proposed model, and finally section 5
presents the conclusions of the paper.
2. PROBLEM STATEMENT
Given is a set of alternatives to place a new industrial plant (L =
{l|l = 1, 2, ..., Nl}). Each site has associated an installation cost,
(Clandl). Also is given a set of process sources (process
streams), (I = {i|i = 1, 2, ..., Ni}). Each process source has a set
of given properties (P = {p|p = 1, 2, ..., Np}), which may include
composition, density, pH, toxicity, viscosity, etc. These streams
can be segregated to be recycled and/or reused in the process
sinks, and the process determines the values for the flow rate
and the properties for each process source given by Wi and
PInSource
i,p , respectively. Besides, given is a set of process sinks
(process units), J = {j|j = 1, 2, ..., Nj}. Each process sink requires
a given flow rate (Gj), with specific limits for the inlet
properties (PInSink ). In addition, there is available a set of fresh
j,p
sources (K = {k|k = 1,2,...,Nk}) that can be sent to the process
sinks (Fk), each one with given properties (PFresh k,p ). As can be
seen in Figure 4, for each property to be treated there is a set of
interceptors in the proposed superstructure to treat each of the
properties individually. Therefore, there is a set of interceptors
to treat the first property (INT1), then for the second property
exists a second set of interceptors (INT2), and for the N
property to be treated there is the last set of interceptors
(INTN). Finally, there is a set of environmental regulations
(PWaste,max
p ) for the wastewater discharged to the environment
(Waste). This new stream will be discharged to a river or
watershed, which is divided in several reaches (R = {r|r = 1, 2,
..., Nr}) and each reach can receive several effluents (T = {t|t =
1, 2, ..., Nt}). Thus, a set of sustainability constraints is imposed
for the properties of the stream discharged to the final disposal.
The problem consists in determining simultaneously the
recycle and reuse network based on properties and the optimal
location for the new plant that discharges wastewater to the
surrounded watershed avoiding the accumulation of pollutants
and maintaining under control the water discharged at the final
disposal (considering the properties of the streams in the
watershed) and, at the same time, satisfying the environmental
constraints for the pollutants through the watershed system.
The objective function is the minimization of the total annual
cost (TAC) that includes the installation costs for the new
plant, the treatment costs for the process sources, the costs of
fresh sources and the piping costs.
93
Article
The total flow rate discharged from the new plant and its
properties are optimization variables, which depend strongly on
its optimal location to be determined by the optimization
process. This means that the required treatment also depends
on the location of the new plant. To simulate the behavior of
the river, the proposed model uses the MFA technique
considering all inlet and outlet streams. Then, the required
treatment also depends on the constraints at the inlet of
process sinks and the treatment is carried out segregating the
process sources to treat them in the interceptors.
One important point included in this paper is the property
balances. Thus, besides the composition balances for key
compounds, the model includes property balances for pH,
toxicity, COD, color, viscosity, etc., which are based on the
property operator functions contained in Table 1 (Shelley and
El-Halwagi48).
Table 1. Mixing Property Operators
property operator
composition ψz(z) = z
toxicity ψTox(Tox) = Tox
chemical oxygen demand ψCOD(COD) = COD
pH ψpH(pH) = 10 pH
density 1
ψρ(ρ) =
ρ
viscosity ψμ(μ) = log(μ)
Reid vapor pressure
ψRVP(RVP) = RVP1.44
electric resistivity
ψR(R ) =
1
R
paper reflectivity 5.92
ψR (R ∞) = R ∞
∞
color ψColor(Color) = Color 0.606
odor ψOdor(Odor) = Odor
On the other hand, the main stream of the river is fed by
several streams called tributaries (FTr,t). As it can be seen in
Figure 3, the watershed system exchanges water during its
trajectory (i.e., water for agricultural use, wastewater discharged
to the river with and without treatment, industrial and
residential effluents, etc.) and, finally, discharges a total stream
to the final disposal. In addition, natural phenomena like
precipitation, filtration, and vaporization are taken into account
in the model. Since all these processes can modify significantly
the characteristics of the river, for tracking adequately the mean
properties, the river is sectioned in parts where the properties
can be considered constant (these sections are represented in
Figure 3 with ovals). These parts or sections are called reaches,
which represent zones where no big intermediate effluents are
discharged and/or extracted. Also, the model accounts for the
main effluents discharged in each reach, which are called
tributaries. The tributaries can be channels or branches of the
river, which may contain discharges with/without treatment,
industrial discharges, etc. The flow rates (FTr,t) and properties
(PTrib
p,r,t ) of the tributaries affect the reaches where they are
discharged. It should be noted that the discharges to the
watershed contain compounds that interact chemically and
biochemically with the system through the flora and fauna
established in the watershed.
dx.doi.org/10.1021/ie3003792 | Ind. Eng. Chem. Res. 2013, 52, 91−107
Industrial & Engineering Chemistry Research
Figure 3. Alternatives to locate the new industrial plant in a watershed system.
To ensure the sustainability of the surrounding watershed
system, it is required to carry out an analysis based on
properties before the optimization to determine the capacity of
the final disposal to decompose chemically and biochemically
the pollutants. The analysis yields the sustainability constraints,
which ensure that the pollutants will not be accumulated in the
system and maintain the properties of the watershed under
adequate limits.
Figure 4 shows the proposed superstructure to treat the
properties of the process streams. In this superstructure,
process sources are segregated to treat the first property,
without mixing different process sources. At the exit of the first
treatment, process sources are segregated again and sent to the
following interceptors to treat the second property. The same
sequence is used to treat all required properties in the property
interception network, where the mixing of different process
sources is avoided in all the steps of the treatment. Each process
source that leaves the treatment units is segregated to be sent to
each process sink or to the wastewater stream. The available
fresh sources can be used in the process sinks when they are
required.
To adjust the streams properties, a set of interceptors can be
used. Different units can be used for these purposes, including
separation units to recover some hazardous materials,
neutralization units to modify the pH, and aeration units to
modify the chemical oxygen demand, among others. Therefore,
a given adjustment factor is used to characterize the
performance of each available interceptor, and these adjustment
1 N
factors (γint ,...,int
) can be calculated prior to the optimization
p
process. This approach leads to linear models for the
interceptors. The adjustment factor depends on several
variables such as the value of the property at the inlet
conditions of the interceptor, design and operating parameters
as well as the flow rate. The available treatment technologies
considered in the property interception network can be
simulated (i.e., using process simulators) to get their adjust-
ment factors. It should be noted that the property operators
values for the inlet condition of the process sources are also
known prior to the optimization process, ψp(PInSource i,p ). There-
fore, the property values at the exit conditions of the property
interception network can be calculated before optimization as
follows:
94
Article
1 N 1 N
ψp(PiOutPIN,int ,...,int
) = γpint ,...,int ψp(PiInSource),
,p ,p
i ∈ I , p ∈ P , int1 ∈ INT1, ..., intN ∈ INT N
3. MODEL FORMULATION
The model is a combination of the material flow analysis
technique, a disjunctive model to place in the optimal position
the new industrial plant, and the superstructure model for the
water integration inside the plant. The next sections explain
these models.
3.1. Material Flow Analysis for the Watershed Based
on Properties. Recently Lira-Barragán et al.61,62 proposed an
MFA model for watersheds based on the composition of the
streams. This model is extended in this paper to consider
property balances that are suited for wastewater streams
constituted by several compounds. In addition, the environ-
mental constraints are given in terms of limits for specific
properties that affect the environment. The balances required
to model the watersheds in terms of properties are based in
Figure 3 and these are stated as follows.
Overall Balance for Each Reach. The exit flow rate (Qr)
from each reach r is equal to the inlet flow rate (Qr−1) plus
precipitation (Pr), direct industrial discharges (Dr), residential
discharges (Hr), the sum of all effluents entering to the reach
(FTr,t) and the possible discharge from the new plant if this is
installed near the reach r of the l allowable (QPNEWr(l)), minus
the extractions due to natural phenomena such as filtration and
evaporation (Lr) as well as used water (Ur) in the r section of
the river. Therefore, the overall balance for each reach can be
written as follows:
Nt(r)
Q r = Q r − 1 + Pr + Dr + Hr + ∑ FTr ,t + QPNEWr(l)
t=1
− Lr − Ur , ∀r∈R (1)
where Nt(r) refers to the total number of tributaries that are
discharged to the reach r.
Property Balances for Each Reach. The product of the
property operator times the flow rate (ψp(PReach
p,r )Qr)) for each
reach r is equal to the property operator times the flow rate at
the inlet (ψp(PInReach
p,r−1 )Qr−1)), adding the property operator
Prec
contained in the precipitation (ψp(Pp,r )Pr)), industrial
dx.doi.org/10.1021/ie3003792 | Ind. Eng. Chem. Res. 2013, 52, 91−107
Industrial & Engineering Chemistry Research Article

Figure 4. Superstructure for the recycle and reuse network.

discharges (ψp(PInd Resid

p,r )Dr), residential discharges (ψp(Pp,r )Hr), the change in the property operator associated to the chemical
the sum of tributaries (∑t = 1 ψp(Pp,r,t )FTr,t) and the property
Nt(r) Trib
reactions that are carried out in that section of the river
operator from the new plant in the case that the new plant is (∫ V V=r 0rp,r dV). The reactive term considers the chemical and
PNEW
located in this reach (ψp(Pp,r(l) )QPNEWr(l)), subtracting the biochemical reactions that take place in the river because of the
Loses Uses
loses (ψp(Pp,r )Lr), uses (ψp(Pp,r )Ur) and taking into account interaction between the pollutants and the system. Therefore,
95 dx.doi.org/10.1021/ie3003792 | Ind. Eng. Chem. Res. 2013, 52, 91−107
Industrial & Engineering Chemistry Research Article

the property balance for each section of the river r is stated as affecting the properties of the streams. The effect of the
follows: chemical or biochemical reactions over the properties for
InReach reaches (eq 5) and tributaries (eq 6) can be represented as
ψp(PpReach Prec
, r )Q r = ψp(P p , r − 1 )Q r − 1 + ψp(P p , r )Pr follows:
Nt(r)
V =0
+ ψp(PpInd
, r )Dr + ψp(PpResid
, r )Hr + ∑ ψp(PpTrib
, r , t )FTr , t ∫V rp , r dVr = kp[ψp(PpReach σp
, r )] Vr , ∀ p ∈ P, r ∈ R
t=1 r
(5)
+ ψp(PpPNEW Loses Uses
, r(l) )QPNEWr(l) − ψp(P p , r )Lr − ψp(Pp , r )Ur
V =0 V =0
− ∫V rp , r dVr , ∀ p ∈ P, r ∈ R ∫V rp , r , t dVr , t = kp[ψp(PpTrib σp
, r , t )] Vr , t ,
r (2) r ,t

Overall Balance for Each Tributary. For the tributary t that ∀ p ∈ P, r ∈ R, t ∈ T (6)
discharges to the reach r, the overall balance is written as
follows: where kp is the kinetic constant of degradation for each
property p measured experimentally, σp is the reaction order for
FTr , t = SrUntreat
,t + SrTreat
,t + Ir , t + Pr , t + Dr , t + QPNEWr , t(l) each property, ψp(PReachp,r ) is the property operator at the reach r,
while ψp(PTrib
p,r,t ) is the property operator at the tributary t, and Vr
− Lr , t − Ur , t , ∀ r ∈ R, t ∈ T (3) and Vr,t are the volumes for the reach and tributary,
respectively.
The total flow rate discharged from the tributary t to the reach r
Agricultural Discharges and Uses. Experimental informa-
(FTr,t) is the sum of discharges without treatment (SUntreated
r,t ),
tion is used to compute the agricultural discharges (Dr,t) as well
discharges with treatment (STreated
r,t ), industrial discharges (Ir,t),
as the agricultural uses (Ur,t) through a pair of experimental
pluvial discharges (Pr,t), direct discharges (Dr,t), and the new
parameters that depend on the agricultural area (Ar,t).
discharge if the new facility is located on this tributary
(QPNEWr,t(p)), subtracting the losses (Lr,t) and use (Ur,t) of Dr , t = αr , t A r , t , ∀ r ∈ R, t ∈ T
water.
Property balances for each tributary. The next balance that
Ur , t = βr , t A r , t , ∀ r ∈ R, t ∈ T
includes the reaction term is used to calculate the property
operator discharged from the tributary t to the reach r.
αr,t is the water required per area and its units are m3/ha·s,
ψp(PpTrib Untreat Untreat
, r , t )FTr , t = ψp(Pp , r , t )Sr , t + ψp(PpTreat Treat
, r , t )Sr , t
while βr,t is the discharged flow rate per area and its units also
are m3/ha·s.
+ ψp(PpInd Prec Direct
, r , t )Ir , t + ψp(P p , r , t )Pr , t + ψp(P p , r , t )Dr , t 3.2. Location of the New Plant. This work considers a set
of alternatives to locate the new industrial facility (L), which are
+ ψp(PpPNEW Loses
, r , t (l))QPNEWr , t (l) − ψp(P p , r , t )Lr , t specified before the optimization process. Then, for each
V =0 alternative, it is necessary to consider the possible discharge of
− ψp(PpUses
, r , t )Ur , t − ∫V rp , r , t dVr , t , the new wastewater flow rate (QPNEWl) and its property
r ,t operators (ψp(PPNEWp,l )). If the new industrial plant is located on
∀ p ∈ P, r ∈ R, t ∈ T the alternative 1, the wastewater flow rate and the property
(4)
operators associated to this site must be greater than zero;
Reactive Terms. The chemical reactions that are carried out otherwise, if the new plant is not located on the position 1, the
in the rivers have strong effects over the properties of the wastewater flow rate and the property operators discharged
streams in the watershed. Several components can be degraded would be zero. This is included in the following disjunction:

⎡ Y1 ⎤ ⎡ Y2 ⎤ ⎡ Yl ⎤
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢ ψ (P pPNEW
,1 ) ≥ 0, QPNEW1 ≥ 0 ⎥ ⎢ ψp(P pPNEW
,1 ) = 0, QPNEW1 = 0 ⎥ ⎢ ψ (P pPNEW
,1 ) = 0, QPNEW1 = 0 ⎥
⎢ p
⎥ ⎢ ⎥ ⎢ p
⎥
⎢ ψ (P PNEW ) = 0, QPNEW = 0 ⎥ ∨ ⎢ ψ (P PNEW ) ≥ 0, QPNEW ≥ 0 ⎥ ∨ ··· ∨ ⎢ ψ (P PNEW ) = 0, QPNEW = 0 ⎥ , ∀p∈P
⎢ p p ,2 2 ⎥ ⎢ p p ,2 2 ⎥ ⎢ p p ,2 2 ⎥
⎢ ⋮ ⎥ ⎢ ⋮ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⋮ ⎥
⎢ ψ (P PNEW ) = 0, QPNEW = 0 ⎥ ⎢ ψ (P PNEW ) = 0, QPNEW = 0 ⎥ ⎢ ψ (P PNEW ) ≥ 0, QPNEW ≥ 0 ⎥
⎣ p p,l l ⎦ ⎣ p p,l l ⎦ ⎣ p p,l l ⎦

In this disjunction, the Boolean variable associated to the
location l of the new plant is Yl, while ψp(PPNEW
p,l ) is the property
operator for the property p from the wastewater stream of the
new industrial facility (QPNEWl). As it can be seen, these
variables depend on the final location of the new plant. In this
sense, the disjunctive model establishes that if the optimal
location for the new plant is on alternative 1, then the Boolean
variable Y1 is set as true; therefore, ψp(PPNEW
p,l ) and QPNEW1
must be greater than zero, while all the others Boolean variables
96
(Yl) must be false and the ψp(PPNEW
p,l ) and QPNEWl associated
to such locations must be set as zero.
A similar situation occurs for the case when the selected
location is 2, 3, and so on for the l possible locations. It is
important to remark that the value for the flow rate of the
wastewater and the property operators of the new plant are
strongly influenced by the location through the material flow
analysis model, satisfying the environmental regulations given
to maintain under control the properties and the desired
constraints in intermediate reaches, as well as the sustainability
dx.doi.org/10.1021/ie3003792 | Ind. Eng. Chem. Res. 2013, 52, 91−107
Industrial & Engineering Chemistry Research Article

constraints at the final disposal. Notice that these variables also where Fk is the total flow rate for the fresh source k and f k,j is
depend on the mass integration inside the plant where there are the segregated flow rate from the fresh source k sent to process
sinks constraints to be met at the inlet to process units. sink j. Notice in Figure 4 that the fresh sources only are sent to
Previous disjunction is reformulated using the convex hull the process sinks.
technique (see Raman and Grossmann63 and Ponce-Ortega et Splitting of Process Sources to the Property Interceptors.
al.64). The interception network is segregated avoiding the mixing of
The reformulation includes the transformation of the different process sources. If a process source does not require
Boolean variables (Yl) into a set of binary variables (yl). any treatment, it is sent to a fictitious unit (located at the end of
When the Boolean variables are true, the associated binary the set of property interceptors) with effectiveness and cost
variables must be set as 1; otherwise, when the Boolean equal to zero. Next equation represents the segregation of each
variables are false, the associated binary variables are zero. process source to treat the first property on the first set of
Therefore, the following relationship is used to select only one available interceptors.
location to install the plant: 1
Wi = ∑ wiint , ∀i∈I
∑ yl =1 int1 ∈ INT1 (13)
l∈L (7)
Then, to treat the second property, the flow rates at the exit of
The next upper limits for the variables QPNEWl and ψp(PPNEW
p,l ) the first line of interceptors are segregated, avoiding the mixing
complete the reformulation. of streams, and again these are fed to the second set of
interceptors.
QPNEWl ≤ QPNEW Upyl , ∀l∈L (8) 1 1 2
wiint = ∑ wiint ,int , ∀ i ∈ I , int1 ∈ INT1
ψp(PpPNEW
,l ) ≤ ψp(P PNEW )Up yl , ∀ p ∈ P, l ∈ L (9)
int2 ∈ INT2 (14)

Notice that this approach does not allow the mixing of different
where QPNEWUp and ψp(PPNEW p,l )Up are the upper limits for process streams. This approach eliminates the nonlinearities
QPNEWl and ψp(Pp,l ), respectively. Notice that ψp(PPNEW
PNEW
p,l )Up involved in the property balances, although the problem size
is an upper bound for all properties p. These constraints are increases exponentially.
required to ensure that QPNEWl and ψp(PPNEW p,l ) take values Following the same sequence for N properties to be
greater than zero only when the new plant is located in a intercepted, the next equations model the segregations.
particular position l.
Equations to Interconnect the Models. The following 1 2
,...,intN − 1 1 2
,...,intN
wiint ,int = ∑ wiint ,int ,
relationships are required to interconnect the material flow intN ∈ INT N
analysis model with the superstructure model for the mass
integration inside the plant: ∀ i ∈ I , int1 ∈ INT1, ..., intN − 1 ∈ INT N − 1 (15)

∑ QPNEWl = Waste Splitting of Sources at the Exit of the Interception

l∈L (10) Network. After the treatment network, the process sources
are segregated and directed to the process sinks and to the
wastewater stream discharged to the environment.
∑ ψp(PpPNEW
,l ) = ψp(Ppdisc), ∀l∈L
(11) 1 N 1 N 1 N
l∈L
wiint ,...,int = ∑ giint,j ,...,int + giint ,...,int
,Waste
,
where “Waste” is the flow rate discharged to the environment j∈J
by the new industrial plant and ψp(Pdisk p ) is the property ∀ i ∈ I , int1 ∈ INT1, ..., intN ∈ INT N (16)
operator for each property of this stream. In this regard, eq 7
assures that only one location is selected, then just one of the l Overall Balance at the Mixing Point Prior to Any Sink. The
variables for QPNEWl and ψp(PPNEWp,l ) can be greater than zero, total flow rate at the inlet of any sink (Gj) must be completed
and eqs 8 and 9 are satisfied. by the sum of the flow rates from process sources intercepted,
3.3. Superstructure Modeling for in-Plant Treatment. as well as the flow rates from the fresh sources.
The proposed superstructure for the in-plant treatment is 1 N
shown in Figure 4, which is based on the ones reported by Gj = ∑ ∑ ··· ∑ giint
,j
,...,int
+ ∑ fk ,j ,
Ponce-Ortega et al.43 and Gabriel and El-Halwagi.22 The size i ∈ I int1 ∈ INT1 intN ∈ INT N k∈K
for this formulation increases exponentially as a function of the ∀j∈J (17)
number of treated properties (N); however, the model
obtained for this superstructure has a linear behavior, which Property Balances at the Mixing Point Prior to Any Sink.
is the main advantage of this formulation because it avoids the These balances are required to calculate the properties at the
mixing of different streams and, thus, it is very useful for the inlet to any process sink.
optimization process. As can be seen in Figure 4, for each
property to be treated there is a set of treatments units; ψp(P jInsin
,p
k
)Gj = ∑ ∑ ···
therefore, it generates N treatment units for each stream. i ∈ I int1 ∈ INT1
Splitting Fresh Sources. Each fresh source is segregated and 1 N 1 N

sent to any process sinks.

∑ [ψp(PiOutPIN,int
,p
,...,int
)giint
,j
,...,int
]
intN ∈ INT N

Fk = ∑ fk ,j , ∀k∈K + ∑ [ψp(PkFresh
, p )fk , j ], ∀ j ∈ J, p ∈ P
j∈J (12) k∈K (18)

97 dx.doi.org/10.1021/ie3003792 | Ind. Eng. Chem. Res. 2013, 52, 91−107

Industrial & Engineering Chemistry Research Article

Overall Balance for the Waste Stream. The total flow rate different purposes, it becomes necessary to impose a set of
of the wastewater stream discharged to the environment is property constraints on corresponding reaches to guarantee the
given by the sum of the process sources after the treatment quality of the water extracted. The proposed model is able to
units that are sent to the waste. consider these constraints through the following relationship:
1 N
Waste = ∑ ∑ ··· ∑ giint ,...,int
ψp(P pDesiredmin ) ≤ ψp(PpReach Desiredmax
,Waste , m(r ) , m(r )) ≤ ψp(P p , m(r ) ),
i ∈ I int1 ∈ INT1 intN ∈ INT N (19)
Property Balances for the Waste Stream. Next equation is ∀ p ∈ P , m ( r ) ∈ M (R ) (23)
used to calculate the properties for the waste stream discharged
to the environment. where M(R) represents a subset of reaches of all the reaches r
that require a specific water quality, while ψp(PDesiredMAX
p,m(r) ) and
ψp(PpWaste)Waste = ∑ ∑ ··· ψp(PDesiredMIN
p,m(r) ) are the limits for the property operators with
i ∈ I int1 ∈ INT1 desired properties in specified reaches to maintain the water
∑ (ψp(PiOutPIN,int ,...,int 1
)giint
N
,...,int
),
1 N
∀p∈P quality in them.
,p
intN ∈ INT N
,Waste Sustainability Constraints. To ensure the sustainability of
(20) the watershed and its surroundings, and to avoid putting at risk
the final disposal (lake, sea, or ocean), the next constraints must
Sink Constraints. The process sinks impose a set of be met because the environmental regulations imposed on the
constraints to work properly that usually are given in terms plant discharge could not be enough:
of specific properties like the composition of specific
compounds, density, pH, viscosity, thermal conductivity, etc.
ψp(P pSustainablemin) ≤ ψp(PpReach Sustainablemax
,final) ≤ ψp(P p ),
In this work, these constraints are stated in terms of the
property operators as follows: ∀p∈P (24)
ψp(P jInsin k,min
) ≤ ψp(P jInsin k
) ≤ ψp(P jInsin k,max
),
,p ,p ,p
where ψp(PReach
p,final ) is the property operator in the final disposal
j ∈ J, P ∈ P (21) after installing the new industrial facility, while ψp(PSustainableMIN
p )
and ψp(PSustainableMAX
p ) are the set of sustainability constraints for
Environmental Constraints. The environmental regulations property p. Notice that the sustainability constraints are
impose a set of constraints on the industrial plant in terms of generated making a sustainability analysis for the final disposal
properties as follows: prior to the optimization process. Satisfying these constraints
ψp(PpWaste,min) ≤ ψp(PpWaste) ≤ ψp(PpWaste,max ), ∀p∈P ensures that the natural degradation of the pollutants in the
final disposal will maintain the properties under sustainable
(22) limits.
This set of constraints includes upper and lower limits for Objective Function. Finally, the objective function consists
specific properties in the wastewater discharged to the in minimizing the total annual cost (TAC), which comprises
environment like pH, COD, toxicity, color, etc. the annualized installation costs for the new plant, the fresh
Constraints for the Quality of Specific Reaches. Because sources costs, the treatment cost (including operational cost,
the water in some specific zones of the rivers can be used for variable and fixed capital cost), and the piping costs.

1 1 N 1 N
min TAC = ∑ Clandlyl + HY ∑ FreCkFk + HY ∑ [ ∑ int int
VarCop wi + ··· ∑ int int ,...,int
(VarCop wi )]
1 1 N N
l∈L k∈K i∈I int ∈ INT int ∈ INT
int 1 int 1 intN int1 ,...,intN
+ HY ∑ [ ∑ VarCcap wi + ··· ∑ (VarCcap wi )]
i∈I int1∈ INT1 intN ∈ INT N
1 1 N 1 N 1 1 N
+ kf ∑[ ∑ int int
FixCcap zi + ··· ∑ int int ,...,int
(FixCcap zi )] + HY [TrPipC ∑ ( ∑ wiint + ··· ∑ wiint ,...,int )
i∈I int1∈ INT1 intN ∈ INT N i∈I int ∈ INT intN ∈ INT N
1 N 1 N
+ WaPipC ∑ ∑ ··· ∑ giint ,...,int
,Waste
+ EqPipC ∑ ∑ ··· ∑ ∑ giint,j ,...,int + FrPipC ∑ ∑ fk ,j ]
1 1 N N 1 1 N N
i ∈ I int ∈ INT int ∈ INT i ∈ I int ∈ INT int ∈ INT j∈J k∈K j∈J (25)

where HY is the hours per year that operates the new industrial being discharged to the environment. A set of technologies
plant and kf is the factor used to annualize the inversion. The must be selected to treat the process sources yielding capital
annualized installation cost (Clandl) takes into account the land and operational costs for each one of the interceptors required.
cost and the transportation costs for raw materials, products N

and services. Notice that this annualized installation cost takes The operational cost (Varint op ) for the interceptors depend on
the treated flow rate, while the capital cost contains a variable
into account simultaneously capital and operational costs, and N
that the factor used to annualized the capital costs is used to cap ) that also depends on the treated flow rate and a
cost (Varint
N
provide the associated cost per year. Fresh sources cost (FreCk) fixed component (Fixintcap ). Finally, the piping costs for the four
is caused by the required amount of fresh sources to satisfy the sections showed in the superstructure are considered. The first
water requirements of the process sinks as well as the properties section involves the pipes required when the process sources
constraints imposed on them. Also, process sources need to be are sent to treatment (TrPipC); then, once the process sources
treated prior to be sent to the process sinks, as well as prior to are treated they can be sent to the wastewater stream
98 dx.doi.org/10.1021/ie3003792 | Ind. Eng. Chem. Res. 2013, 52, 91−107
Industrial & Engineering Chemistry Research Article

(WaPipC) or process sinks (EqPipC). The final section covers

the pipes to connect the fresh sources with process sinks
(FrPipC).
Remarks. (a) The model presented considers simulta-
neously the water integration inside the industrial facility and
the sustainability of the surrounding watershed. (b) The model
presented in this paper is based on the properties of the
streams, considering property balances, as well as process and
environmental constraints given in terms of limits for the
properties. This situation is very useful for streams constituted
by several pollutants, where the component balances are not
suited to characterize them. (c) The effect of the chemical and
biochemical reactions over the properties can be represented by
a simple kinetic model. The kinetic constants can be measured
experimentally. (d) The annualized installation costs (Clandl)
include different costs: transportation cost of raw materials,
transportation cost of products, services, and the land cost for
the alternative l. These costs depend directly on the location
1
,...,intN
selected for the new plant. (e) Adjustment factors (γint p )
are parameters known for each unit to treat each property,
which can be determined experimentally or estimated by
simulation. Therefore, the property balances in the super-
structure for the treatment systems are not required in the
optimization model. Also, to determine the outlet property
from the interception network only the following relationship is
OutPIN,int1,...,intN 1
,...,intN
used: ψp(Pi,p ) = γint
p ψp(PinSource
i,p ).

4. RESULTS AND DISCUSSION

The model presented was coded in the general algebraic
Figure 5. Bahr El-Baqar system.
modeling system (GAMS).65 Two examples are presented to
show the implementation of the proposed formulation, one
N The natural degradation of pollutants along the system helps
from Egypt and the other one from Mexico. Fixint cap are fixed as to improve the properties like toxicity, COD, color, etc.
zero for both examples. Specifically, in this system the kinetic constants in terms of
4.1. Example 1. Bahr El-Baqar System. The Bahr El- properties are: kComp = 9.041909 × 10−6/s, kTox = 5.208473 ×
Baqar system is in Egypt and its description was given by 10−7/s, kCOD = 4.08396 × 10−6/s, kOdor = 2.083389 × 10−6/s
Lovelady et al.60 The system receives several types of discharges and kColor = 3.645931 × 10−6/s, which correspond to a kinetics
(i.e., agricultural, industrial, municipal effluents) and the of first order. In addition, the new plant operates 8000 h/year
properties restricted in the environment are the toxicity, and kf = 0.1. Table 3 shows the characteristics for the fresh
COD, odor, color, and the composition of a dangerous
pollutant. The properties restricted in the process sinks are the Table 3. Fresh and Process Sources for Example 1
composition and the toxicity. Four candidate sites are identified
to install a new industrial plant, where each one has an flow rate, composition, toxicity, COD,
sources m3/s ppm % mgO2/L odor color
associated installation cost (see Table 2). Figure 5 shows the
Process
1 0.316 88.7 0.14 139 6 150
Table 2. Installation Cost for the Different Sites to Locate
2 0.435 77.2 0.22 93 7 262
the New Plant for Example 1
3 0.402 52.4 0.16 128 5.5 290
site annualized installation cost, $/year Fresh
1 14,000,000 1 0 0
2 12,500,000 2 10 0.005
3 11,000,000
4 10,000,000
sources and the process sources; notice that the flow rates for
Bahr El-Baqar system, and the four available locations as well as the fresh sources are optimization variables and the
the properties considered and the constraints imposed over the specifications of COD, odor, and color are not presented
system. The set of constraints is explained by considering the because fresh sources can only be sent to process sinks, which
reach 6. Before installing the new plant, it has a composition do not contain constraints in terms of these properties for this
and toxicity of 1.312 ppm and 0.0076, and the maximum example.
composition and toxicity allowed after the installation of the The flow rate required for each process sink as well as the
plant are 1.34 ppm and 0.0078, respectively. A similar maximum composition and toxicity allowed are presented in
explanation applies to the constraints imposed on the Table 4, while Table 5 shows the available interceptors to treat
remaining reaches of Figure 5. the properties considered.
99 dx.doi.org/10.1021/ie3003792 | Ind. Eng. Chem. Res. 2013, 52, 91−107
Industrial & Engineering Chemistry Research Article

Table 4. Requirements for the Process Sinks for Example 1 installation cost, $4.816 × 106/yr of treatment cost, $573,330/
yr of fresh source cost and $827,532/yr of piping cost. The
process sink flow rate, m3/s compositionmax toxicitymax
results obtained for the solution of the proposed model shows
1 0.365 55.5 0.013 that the new plant must be installed in the site 3 with the mass
2 0.290 36.2 0.011 integration based on properties showed in Figure 6. Notice that
3 0.335 30 0.010 some properties for the stream discharged to the environment
are limited by the environmental constraints, whereas others
Table 5. Characteristics of the Available Interceptors for are lower than the environmental constraint. Specifically, the
Example 1 composition, the toxicity, as well as the color require a larger
interceptor adjustment factor cost, $/m3 reduction to satisfy the constraints imposed on the river or the
Composition final disposal to maintain the quality of the water around the
Rec1 0.05 0.090 system (see Figure 7). Therefore, the results show that if the
Rec2 0.22 0.072 new plant only satisfies the environmental regulations imposed
Toxicity by the legislation on the wastewater, the sustainability
Tox1 0.08 0.105 requirements of the system are not fulfilled since the associated
Tox2 0.17 0.075 environmental impact of the wastewater discharged from the
COD new plant is not in line with the estimated degradation capacity
Aer1 0.06 0.055 of the watershed.
Aer2 0.23 0.032 On the other hand, this example was solved previously by
Odor Lira-Barragán et al.62 while only considering the compositions
Ads1 0.02 0.065 of chemical compounds and without taking into account the
Ads2 0.16 0.047 properties. Lira-Barragán et al.62 found that the optimal
Color location of the new plant is site 4 with a total annualized
Dec1 0.10 0.070 cost of $12.121 × 106/yr. When the constraints based on
Dec2 0.20 0.051 properties for the in-plant sinks and the wastewater discharged
to the environment are not included, some constraints may be
The costs for the fresh sources 1 and 2 are $0.190/m3 and violated. For example, the properties in reach 6 (toxicity of
$0.142/m3, respectively. On the other hand, the environmental 0.0088, COD, odor, and color of 42.808, 2.936, and 134.69,
constraints for the wastewater stream in terms of properties are: respectively) represent a risk for the water extracted for human
PSustainable
Comp = 50, PSustainable
Tox = 0.1, PSustainable
COD = 75, PSustainable
Odor = 2.7, activities, and this problem is avoided in the solution given by
Sustainable the formulation proposed in this paper.
and PColor = 100. Then, to satisfy these constraints the
model considers a set of interceptors with different efficiencies 4.2. Example 2. Balsas Watershed. This case considers
to remove the pollutants and to improve the properties. Finally, the Balsas watershed that is one of the largest systems in
the piping costs for each section in the superstructure are Mexico. This system transports several types of discharges until
TrPipC = 0.003 $/m3, WaPipC = 0.006 $/m3, EqPipC = 0.010 the Pacific Ocean, and CONAGUA 67,68 reported the
$/m3, and FrPipC = 0.008 $/m3. information for it shown in Figure 8. In this case, the
The problem for this case study was implemented in the properties considered were the composition, toxicity, density,
software GAMS (Brooke et al.65), and it was solved using the COD, odor, and color, and the properties restricted for the
solver DICOPT (Viswanathan and Grossmann66). The optimal stream discharged to the environment are composition, toxicity,
solution requires a total annual cost of $17.317 × 106/yr, which COD, odor, and color, while the process sinks limit only the
is constituted by the next cost components: $11.1 × 106/yr of composition and density. The system offers 20 possible

Figure 6. Mass integration based on properties for Example 1.

100 dx.doi.org/10.1021/ie3003792 | Ind. Eng. Chem. Res. 2013, 52, 91−107

Industrial & Engineering Chemistry Research Article

Table 6. Installation Costs for Example 2

site annualized installation cost, $/year
1 8,000,000
2 8,500,000
3 9,000,000
4 10,500,000
5 9,500,000
6 9,000,000
7 15,500,000
8 14,500,000
9 17,700,000
10 22,000,000
11 18,000,000
12 17,500,000
13 24,000,000
14 28,000,000
15 23,500,000
16 17,000,000
17 20,000,000
18 19,000,000
19 16,000,000
20 17,000,000

the new plant is working. The property odor discharged to

Figure 7. Impact of the new wastewater stream over Bahr El-Baqar
system.
locations to install the new industrial facility, and their
installation costs are presented in Table 6.
Figure 8 shows the properties before installing the new
facility as well as the set of constraints imposed on the system.
For example, the composition discharged on the Pacific Ocean
prior to the installation of the new plant is 1.466 ppm, and it is
desired that this concentration does not exceed 1.47 ppm when
Pacific Ocean is 0.654 and once the new plant is operating must
be lower or equal to 0.68. This explanation has to be followed
to understand all the constraints imposed on the river. It is
important to note that in the restricted reaches, toxicity is
strictly regulated.
To model the reactive effects, the kinetic constants in terms
of properties for first order kinetics are as follows: kComp =
3.2376 × 10−6/s, kTox = 4.5205 × 10−7/s, kCOD = 1.9273 ×
10−6/s, kOdor = 8.315 × 10−7/s and kColor = 1.1423 × 10−6/s.
Remember that the density is not restricted in the environment.
Furthermore, the new plant will operate 8000 h/year with a kf =
0.1. Table 7 shows the characteristics for the fresh and process
sources; notice that for fresh sources only the specifications of
composition and density are presented because they cannot be

Figure 8. Balsas watershed system (possible locations and the constraints on the river).

101 dx.doi.org/10.1021/ie3003792 | Ind. Eng. Chem. Res. 2013, 52, 91−107

Industrial & Engineering Chemistry Research Article

Table 7. Fresh and Process Sources for Example 2

sources flow rate, m3/s composition, ppm toxicity, % COD, mgO2/L odor color density, g/mL
Process
1 0.363 64.6 0.11 149 6.3 250 0.94
2 0.241 58 0.04 88 7.7 234 2.41
3 0.046 45.3 0 108 5.1 298 1.14
4 0.447 60 0.13 185 4.8 202 1.20
5 0.196 52 0.21 120 7.6 194 2.70
6 0.204 62 0.25 76 8.2 164 1.64
7 0.281 54.5 0 243 3.7 179 1.85
8 0.329 47.8 0.13 164 7.4 346 1.75
9 0.282 46 0.09 297 4.8 264 2.45
10 0.266 82.1 0.7 326 8.5 394 0.89
11 0.215 26.4 0.3 146 5.7 195 2.15
Fresh
1 0 2.3
2 1 2.45
3 2 1.8

sent to the wastewater stream discharged to the environment piping costs for each section in the superstructure are TrPipC =
where the other properties are restricted. 0.001 $/m3, WaPipC = 0.002 $/m3, EqPipC = 0.005 $/m3, and
Table 8 shows the requirements for the flow rate and FrPipC = 0.003 $/m3.
maximum composition, as well as the minimum and maximum Then, the proposed model was applied for this case study
yielding an optimal solution where the new facility must be
Table 8. Requirements for the Process Sinks for Example 2 placed in alternative 6 with a total annual cost of $14.9 × 106/
yr, composed by $9 × 106/yr of installation cost, $3.6 × 106/yr
process sink flow rate, m3/s compositionmax densitymin densitymax of treatment cost, $1.5203 × 106/yr of fresh source cost, and
1 0.348 16 2.2 2.4 $779,040/yr of piping cost. Figure 9 shows the water network
2 0.428 19.4 1.95 2.1 based on properties for this example; notice that the value of
3 0.406 23.3 2.35 2.55 toxicity in the wastewater stream is zero, satisfying the
4 0.312 28.5 1.7 1.9 constraint in reach 8, where it is rquired to be zero (see Figure
5 0.336 18 2.2 2.3 10). Also, it is worth noting that most of the properties in the
6 0.294 14 1.8 2 wastewater streams almost take the upper limits of the
7 0.196 22.5 2.45 2.55 environmental constraints. Nonetheless, the composition
requires a greater reduction to satisfy the constraints imposed
density for each process sink. The unitary cost for fresh source on reach 8 to meet the quality of the water extracted for human
1 is $0.128/m3, for fresh source 2 is $0.113/m3, and for fresh activities. On the other hand, if the proposed model by Lira-
source 3 is $0.097/m3. Table 9 shows the characteristics Barragán et al.62 is applied to this example (i.e., if the properties
(efficiency and variable cost) for the property interceptors. For are not taken into account), the optimal solution to install the
this case, the variables costs are included in just one term. new plant is alternative 1, obtaining a total annualized cost of
The environmental legislation imposes the next constraints $10.214 × 106/yr. However, in that solution, some of the
on the wastewater stream: PSustainableComp = 50, PSustainable
Tox = 0.1, properties have values greater than the maximum allowed,
PSustainable
COD = 75, P Sustainable
Odor = 2.7, and PSustainable
Color = 100. While the yielding not sustainable solutions.
Finally, Table 10 shows the computation time for the
Table 9. Available Interceptors for Example 2 problems presented in this work, using a computer with an i5
processor at 2.3 GHz with 4 GB of RAM.
interceptor adjustment factor variable cost, $/m3
Composition 5. CONCLUSIONS
Rec1 0.11 0.026 This paper has introduced a water integration approach to
Rec2 0.23 0.021 consider simultaneously the integration inside the plant and the
Toxicity sustainability of the surrounded watershed during the selection
Tox1 0.00 0.054 of a construction site of a new industrial facility. A material flow
Tox2 0.08 0.049 analysis is developed and coupled with an optimization
COD formulation to track, manipulate, and assign properties and
Aer1 0.18 0.022 compositions while considering water integration alternatives
Aer2 0.43 0.018 and the impact on the surrounding watershed. The model also
Odor considers the integration of the wastewater discharged from the
Ads1 0.14 0.013 industrial facility with the other discharges in the regions that
Ads2 0.39 0.011 impact the watershed (e.g., agricultural, domestic, other
Color industrial discharges, evaporation, filtration, etc.). Phenomena
Dec1 0.10 0.019 occurring within the watershed are accounted for from the
Dec2 0.20 0.015 natural degradation of the pollutants through the watershed
102 dx.doi.org/10.1021/ie3003792 | Ind. Eng. Chem. Res. 2013, 52, 91−107
Industrial & Engineering Chemistry Research Article

Figure 9. Mass integration for Example 2.

Figure 10. Impact of the new wastewater stream over Balsas system.

103 dx.doi.org/10.1021/ie3003792 | Ind. Eng. Chem. Res. 2013, 52, 91−107

Industrial & Engineering Chemistry Research Article

Table 10. Size and Computation Time for the Problems
Considered
concept example 1 example 2
number of constraints 1,415 9,527
number of continuous variables 5,030 30,876
number of binary variables 4 20
CPU time (s) 4.36 19.96
and their effect on the specific properties that are constrained in
the system. Economic and environmental issues are included in
the optimization formulation. The devised model has been
applied to two case studies, one from Egypt and the other one
from Mexico. The results show that the traditional water
integration schemes that do not consider the surrounding
watershed may not be sufficient to ensure the sustainability of
the watershed. To satisfy the sustainability requirements, the
interaction of the wastewater streams discharged to the
environment with other discharges and uses through the
watershed must be considered in conjunction with the physical,
chemical, and biological phenomena occurring in the water-
shed. The results also show that the property-based approach
proposed in this paper is suited for problems with several
pollutants that are difficult to quantify in terms of the
composition of the streams. Finally, when only the composition
is considered, several property requirements may not be
satisfied.
r = reach
Sink = sink
Source = source
t = tributary
Treat = treated
Trib = tributary
Untreat = untreated
Waste = waste discharged to the environment
Sets
I = set for the process sources, {i|i = 1,...,Ni}
INTN = set for property interceptor N, {int|int =1,...,NintN}
J = set for the sinks, {j|j = 1,...,Nj}
K = set for the fresh sources, {k|k = 1,...,Nk}
M(R) = subset for the specific reaches that require properties
constraints
L = set for the sites to locate the new plant, {l|l = 1,...,Nl}
P = set for the properties, {p|p = 1,...,Np}
R = set for the reaches, {r|r = 1,...,Nr}
T = set for the tributaries, {t|t = 1,...,Nt}
Parameters
Ar,t = area cover by effluent t in reach r, acre or ha
Clandp = annualized installation cost for the new plant in site
p, $/year
Dr,t = agricultural discharges from tributary t to reach r, m3/s
Dr = direct discharges to the reach r, m3/s

■
EqPipC = piping cost to send process sources to process
sinks, $/m3
AUTHOR INFORMATION N

cap = fixed part of the capital cost for interceptor int ,

N
Corresponding Author
FixCint
*E-mail: el-halwagi@tamu.edu. Tel.: 979 845-3484. Fax: 979 $/year
845-6446. FreCk = unit cost for fresh utility k, $/m3
FrPipC = piping cost to send fresh sources to process sinks,
Notes $/m3
The authors declare no competing financial interest.

■
Gj = total flow rate for process sink j, m3/s
HY = operation time per year, hr/year
ACKNOWLEDGMENTS
Hr = total discharge (i.e., industrial + sanitary) to the reach r,
Authors acknowledge the financial support from the Mexican m3/s
Council for Science and Technology (CONACYT).

■
Ir,t = industrial discharge from the tributary t to the reach r,
m3/s
DEDICATION kf = factor used to annualize the capital costs
To Professor L. T. Fan with deep gratitude for his inspiration kp = kinetic constant for the degradation for property p
and seminal contributions in process systems engineering

■
Lr,t = total losses (filtration and evaporation) from tributary t
of the reach r, m3/s
NOMENCLATURE Lr = total losses (filtration and evaporation) from the reach r,
Indexes m3/s
Direct = direct Ni = total number of process sources
Ind = industrial NintN = total number of interceptors
i = process source Nj = total number of sinks
int = interceptor Nk = total number of fresh sources
In = inlet Nl = total number of sites allowable to locate the new plant
j = sink Np = total number of properties
k = fresh source Nr = total number of reaches
l = site to locate the new plant Nt = total number of tributaries
m(r) = specific reaches that require property constraints Pr,t = precipitation discharged for the tributary t to the reach
max = maximum r, m3/s
min = minimum Pr = precipitation discharged to the reach r, m3/s
N = number of properties to be treated PDesired
p,m(r) = desired properties in some reaches
Out = out PWaste,max
p = environmental regulation for the property to be
PNEW = new industrial plant discharged by the new plant
Prec = precipitation PSustainable
p = sustainability constraints at the final disposal
p = property Suntreated
r,t = residual wastewater discharged without treatment
Resid = residential to the reach r for tributary t, m3/s
104 dx.doi.org/10.1021/ie3003792 | Ind. Eng. Chem. Res. 2013, 52, 91−107
Industrial & Engineering Chemistry Research Article

Streated
r,t = residual treated wastewater discharged to the reach r (5) Hallale, N. A new graphical targeting method for water
for tributary t, m3/s minimization. Adv. Environ. Res. 2002, 6 (3), 377−390.
TrPipC = treatment piping cost, $/m3 (6) Manan, Z. A.; Tan, Y. L.; Foo D. C. Y. Targeting the minimum
Ur,t = water used from tributary t discharged to reach r, m3/s water rate using water cascade analysis technique. AIChE J.. 50 (12),
3169-3183.
Ur = water used from reach r, m3/s
N
(7) El-Halwagi, M. M.; Gabriel, F.; Harell, D. Rigorous graphical
VarCint cap = variable part for the capital cost for interceptor targeting for resource conservation via material recycle/reuse net-
intN, $/m3 works. Ind. Eng. Chem. Res. 2003, 42 (19), 4319−4328.
N (8) Feng, X.; Bai, J.; Zheng, X. S. On the use of graphical method to
VarCint op = variable cost associated to operation of the determine the targets of single-contaminant regeneration recycling
interceptor intN, $/m3 water systems. Chem. Eng. Sci. 2007, 62 (8), 2127−2138.
Vr,t = volume for tributary t from reach r, m3 (9) Foo, C. Y. State-of-the-art review of pinch analysis techniques for
Vr = volume for reach r, m3 water network synthesis. Ind. Eng. Chem. Res. 2009, 48, 5125−5159.
Wi = total flow rate for the process sources i, m3/s (10) Sorin, M.; Bedard, S. The global pinch point in water reuse
WaPipC = waste piping cost, $/m3 networks. Process Saf. Environ. Protect. 1999, 77 (B5), 305−308.
αr,t = agricultural flow rate per area, m3/ha*s (11) Gomes, J. F. S.; Queiroz, E. M.; Pessoa, F. L. P. Design
βr,t = agricultural use of water from tributary t, m3/ha s procedure for water/wastewater minimization: Single contaminant. J.
1
,...,intN Cleaner Prod. 2007, 15 (5), 474−485.
γint
p = efficiency factor to improve the property p for the (12) Foo, D. C. Y.; Kazantzi, V.; El-Halwagi, M. M.; Manan, Z. A.
interceptor intN Surplus diagram and cascade analysis techniques for targeting
ρ = density property-based material reuse network. Chem. Eng. Sci. 2006, 61 (8),
σp = reaction order for property p 2626−2642.
ψp = property operator for the mixing rule for property p (13) Takama, N.; Kuriyama, T.; Shiroko, K.; Umeda, T. Optimal
water allocation in a petroleum refinery. Comput. Chem. Eng. 1980, 4
Variables
(4), 251−258.
Fk = total flow rate for fresh source k, m3/s (14) El-Halwagi, M. M.; Hamad, A. A.; Garrison, G. W. Synthesis of
f k,j = segregated flow rate from fresh source k to sink j, m3/s waste interception and allocation networks. AIChE J. 1996, 42 (11),
FTr,t = flow rate discharged by the tributary t to the reach r, 3087−3101.
m3/s (15) Savelski, M. J.; Bagajewicz, M. J. On the optimality conditions of
N water utilization systems in process plants with single contaminants.
gint1,...,int
i,j = segregated flow rate from interceptors intN to Chem. Eng. Sci. 2000, 55 (21), 5035−5048.
process sink j for process source i, m3/s (16) Savelski, M. J.; Bagajewicz, M. J. Algorithmic procedure to
1
,...,intN
gint
i,Waste = segregated flow rate from interceptors intN to design water utilization systems featuring a single contaminant in
waste for process source i, m3/s process plants. Chem. Eng. Sci. 2001, 56 (5), 1897−1911.
p = property for the final discharge of the new plant to the
Pdisk (17) Savelski, M.; Bagajewicz, M. On the necessary conditions of
river optimality of water utilization systems in process plants with multiple
PNEW contaminants. Chem. Eng. Sci. 2003, 58 (23−24), 5349−5362.
Pp,r(l) = property of the waste stream, to be discharged for
(18) Alva-Argaez, A.; Kokossis, A. C.; Smith, R. Wastewater
the new plant installed on site l minimization of industrial systems using an integrated approach.
p,final = property discharged to the final disposal
PReach Comput. Chem. Eng. 2003, 22 (Suppl), S741−S744.
Qr = flow rate exit from the reach r, m3/s (19) Alva-Argaez, A.; Vallianatos, A.; Kokossis, A. A multi-
Qr−1 = flow rate inlet to the reach r, m3/s contaminant transshipment model for mass exchange networks and
QPNEWl = flow rate discharged for the new plant for the wastewater minimization problems. Comput. Chem. Eng. 1999, 23
location p (10), 1439−1453.
rp,r = reaction carried out in the reach r in terms of property (20) Benko, N.; Rev, E.; Fonyo, Z. The use of nonlinear
p programming to optimal water allocation. Chem. Eng. Commun.
rp,r,t = reaction carried out in the tributary t that discharges to 2000, 178, 67−101.
the reach r in terms of property p (21) Teles, J.; Castro, P. M.; Novals, A. Q. LP-based solution
strategies for the optimal design of industrial water networks with
TAC = total annual cost, $/year
multiple contaminants. Chem. Eng. Sci. 2008, 63 (2), 376−394.
1
,...,intN
wint
i = segregated flow rate from process source i to (22) Gabriel, F. B.; El-Halwagi, M. M. Simultaneous synthesis of
interceptor intN, m3/s waste interception and material reuse networks: Problem reformula-
Waste = total flow rate for the waste stream discharged to tion for global optimization. Environmental Progress 2005, 24 (2),
the environment, m3/s 171−180.
Yl = Boolean variable for the location of the new plant (23) Kuo, W. C. J.; Smith, R. Effluent treatment system design. Chem.
yl = binary variable for the location of the new plant Eng. Sci. 1997, 52 (23), 4273−4290.

■
REFERENCES
(1) Wang, Y. P.; Smith, R. Wastewater minimization. Chem. Eng. Sci.
1994, 49 (7), 981−1006.
(2) Dhole, V. R.; Ramchandani, N.; Tainsh, R. A.; Wasilewski, M.
Make your process water pay for itself. Chem. Eng. 1996, 103 (1),
100−103.
(3) El-Halwagi, M. M.; Spriggs, H. D. Solve design puzzles with mass
integration. Chem. Eng. Progress 1998, 94 (8), 25−44.
(4) Polley, G. T.; Polley, H. L. Design better water networks. Chem.
Eng. Progress 2000, 96 (2), 47−52.
(24) Doyle, S. J.; Smith, R. Targeting water reuse with multiple
contaminants. Process Safety and Environmental Protection 1997, 75
(B3), 181−189.
(25) Galan, B.; Grossmann, I. E. Optimal design of distributed
wastewater treatment networks. Ind. Eng. Chem. Res. 1998, 37 (10),
4036−4048.
(26) Hernandez-Suarez, R.; Castellanos-Fernandez, J.; Zamora, J. M.
Superstructure decomposition and parametric optimization approach
for the synthesis of distributed wastewater treatment networks. Ind.
Eng. Chem. Res. 2004, 43 (9), 2175−2191.
(27) Gunaratnam, M.; Alva-Argaez, A.; Kokossis, A.; Kim, J. K.;
Smith, R. Automated design of total water systems. Ind. Eng. Chem.
Res. 2005, 44 (3), 588−599.

105 dx.doi.org/10.1021/ie3003792 | Ind. Eng. Chem. Res. 2013, 52, 91−107

Industrial & Engineering Chemistry Research
(28) Karuppiah, R.; Grossmann, I. E. Global optimization for the
synthesis of integrated water systems in chemical processes. Comput.
Chem. Eng. 2006, 30 (4), 650−673.
(29) Putra, Z. A.; Amminudin, K. A. Two-step optimization approach
for design a total water system. Ind. Eng. Chem. Res. 2008, 47 (16),
6045−6057.
(30) Ponce-Ortega, J. M.; Nápoles-Rivera, F.; El-Halwagi, M. M.;
Jiménez-Gutiérrez, A. An optimization approach for the synthesis of
recycle and reuse water integration networks. Clean Technol. Environ.
Policy 2008, 14 (1), 133−151.
(31) Nápoles-Rivera, F.; Ponce-Ortega, J. M.; El-Halwagi, M. M.;
Jiménez-Gutiérrez, A. Global optimization of recycle−reuse mass
integration networks for processes with multiples contaminants.
Environ. Progress Sustain. Energy 2011.
(32) Faria, D. C.; Bagajewicz, M. J. A new approach for global
optimization of a class of MINLP problems with applications to water
management and pooling problems. AIChE J.. 2011, In press. DOI:
10.1002/aic.12754
(33) Faria, D. C.; Bagajewicz, M. J. Global optimization based on
subspaces elimination: Application to generalized pooling and water
management problems. AIChE J.. 2011, In press. DOI: 10.1002/
aic.12738
(34) Faria, D. C.; Bagajewicz, M. J. Planning model for the design
and/or retrofit of industrial water systems. Ind. Eng. Chem. Res. 2011,
50 (7), 3788−3797.
(35) Faria, D. C.; Bagajewicz, M. J. Global optimization of water
management problems using linear relaxation and bound contraction
methods. Ind. Eng. Chem. Res. 2011, 50 (7), 3738−3753.
(36) Faria, D. C.; Bagajewicz, M. J. Novel bound contraction
procedure for global optimization of bilinear MINLP problems with
application to water management problems. Comput. Chem. Eng. 2011,
35 (3), 446−455.
(37) Faria, D. C.; Bagajewicz, M. J. Comparative analysis of different
assumptions for the design of single-contaminant water networks.
Chem. Eng. Commun. 2010, 197 (6), 859−880.
(38) Faria, D. C.; Bagajewicz, M. J. On the degeneracy of the water/
wastewater allocation problem in process plants. Ind. Eng. Chem. Res.
2010, 49 (9), 4340−4351.
(39) Faria, D. C.; Bagajewicz, M. J. On the appropriate modeling of
process plant water systems. AIChE J. 2010, 53 (3), 668−689.
(40) Faria, D. C.; Bagajewicz, M. J. Profit-based grassroots design and
retrofit of water networks in process plants. Comput. Chem. Eng. 2009,
33 (2), 436−453.
(41) Ng, D. K. S.; Foo, D. C. Y.; Rabie, A.; El-Halwagi, M. M.
Simultaneous synthesis of property-based water reuse/recycle and
interception networks for batch processes. AIChE J. 2008, 54 (10),
2634−2632.
(42) Ponce-Ortega, J. M.; Hortua, A. C.; El-Halwagi, M. M.; Jiménez-
Gutiérrez, A. A property-based optimization of direct recycle networks
and wastewater treatment processes. AIChE J. 2009, 55 (9), 2329−
2344.
(43) Ponce-Ortega, J. M.; El-Halwagi, M. M.; Jiménez-Gutiérrez, A.
Global optimization of property-based recycle and reuse networks
including environmental constraints. Comput. Chem. Eng. 2010, 34 (3),
318−330.
(44) Ponce-Ortega, J. M.; Mosqueda-Jiménez, F. W.; Serna-González,
M.; Jiménez-Gutiérrez, A.; El-Halwagi, A. A property-based approach
to the synthesis of material conservation networks with economic and
environmental objectives. AIChE J. 2011, 57 (9), 2369−2387.
(45) Nápoles-Rivera, F.; Ponce-Ortega, J. M.; El-Halwagi, M. M.;
Jiménez-Gutiérrez, A. Global optimization of mass and property
integration networks with in-plant property interceptors. Chem. Eng.
Sci. 2010, 65 (15), 4363−4377.
(46) Kheireddine, H.; Dadmhammadi, Y.; Deng, C.; Feng, X.; El-
Halwagi, M. M. Optimization of direct recycle networks with the
simultaneous consideration of property, mass, and thermal effects. Ind.
Eng. Chem. Res. 2011, 50 (7), 3754−3762.
106
Article
(47) Deng, C.; Feng, X. Targeting for conventional and property-
based water network with multiple resources. Ind. Eng. Chem. Res.
2011, 50 (7), 3722−3737.
(48) Shelley, M. D.; El-Halwagi, M. M. Componentless design of
recovery and allocation systems: A functionality-based clustering
approach. Comput. Chem. Eng. 2000, 24 (9−10), 2081−2091.
(49) El-Halwagi, M. M..; Glasgow, I. M.; Eden, M. R.; Qin, X.
Property integration: componentless design techniques and visual-
ization tools. AIChE J. 2004, 50 (8), 1854−1869.
(50) Ku-Pineda, V.; Tan, R. R. Environmental performance
optimization using process water integration and sustainable process
index. J. Cleaner Prod. 2006, 14 (18), 1586−1592.
(51) Tan, R. R.; Foo, D. C. Y.; Ng, D. K. S.; Chiang, C. L.; Hul, S.;
Ku-Pineda, V. An approximated mixed integer linear programming
(MILP) model for the design of water reuse/recycle networks with
minimum energy. Asia-Pac. J. Chem. Eng. 2007, 42, 566−574.
(52) Lim, S. R.; Park, J. M. Environmental and economic analysis of a
water networks system using LCA and LCC. AIChE J. 2007, 53 (12),
3253−3262.
(53) Lim, S. R.; Park, J. M. Cooperative water networks system to
reduce carbon footprints. Environ. Sci. Technol. 2008, 42 (16), 6230−
6236.
(54) Lim, S. R.; Park, J. M. Synthesis of an environmentally friendly
water network system. Ind. Eng. Chem. Res. 2008, 47 (6), 1988−1994.
(55) Brunner, P. H.; Rechberg, H. Practical Handbook of Material
Flow Analysis; CRC: Boca Raton, FL, 2004.
(56) Baccini, P.; Brunne, P. Metabolism of the Anthroposphere;
Springer-Verlag: Berlin, 1991.
(57) Lampert, C.; Brunner, P. H. Material accounting as a policy tool
for nutrient management in the Danube Basin. Water Sci. Technol.
1999, 40 (10), 43−49.
(58) El-Baz, A. A.; Ewida, K. T.; Shouman, M. A.; El-Halwagi, M. M.
Material flow analysis and integration of watersheds and drainage
systems: I. Simulation and application to ammonium management in
Bahr El-Baqar drainage system. Clean Technol. Environ. Policy 2005, 7
(1), 51−61.
(59) El-Baz, A. A.; Ewida, K. T.; Shouman, M. A.; El-Halwagi, M. M.
Material flow analysis and integration of watersheds and drain systems:
II. Integration and solution strategies with application to ammonium
management in Bahr El-Baqar drain system. Clean Technol. Environ.
Policy 2005, 7 (1), 78−86.
(60) Lovelady, E. M.; El-Baz, A. A.; El-Monayeri, D.; El-Halwagi, M.
M. Reverse problem formulation for integrated process discharges with
wastewater and drainage systems: Managing phosphorus in lake
Manzala. J. Ind. Ecol. 2009, 13 (6), 914−927.
(61) Lira-Barragán, L. F.; Ponce-Ortega, J. M.; Serna-González, M.;
El-Halwagi, M. M. An MINLP model for the optimal location of a new
industrial plant with simultaneous consideration of economic and
environmental criteria. Ind. Eng. Chem. Res. 2011, 50 (2), 953−964.
(62) Lira-Barragán, L. F.; Ponce-Ortega, J. M.; Serna-González, M.;
El-Halwagi, M. M. Synthesis of water networks considering the
sustainability of the surrounding watershed. Comput. Chem. Eng. 2011,
35 (12), 2837−2852.
(63) Raman, R.; Grossmann, I. E. Modeling and computational
techniques for logic based integer programming. Comput. Chem. Eng.
1994, 18 (7), 563−578.
(64) Ponce-Ortega, J. M.; Serna-González, M.; Jiménez-Gutiérrez, A.
A disjunctive programming model for simultaneous synthesis and
detailed design of cooling networks. Ind. Eng. Chem. Res. 2009, 48 (6),
2991−3003.
(65) Brooke, A.; Kendrick, D.; Meeruas, A.; Raman, R. GAMS-
Language Guide; GAMS Development Corporation: Washington, DC,
2011.
(66) Viswanathan, J.; Grossmann, I. E. A Combined penalty function
and outer approximation method for MINLP optimization. Comput.
Chem. Eng. 1990, 14 (7), 769−782.
(67) CONAGUA. Mexican National Water Commission. Water
Statistics 2010. http://www.conagua.gob.mx/OCB07/Contenido/
Documentos/EstadisticasBALSAS.pdf (accessed 2010).
dx.doi.org/10.1021/ie3003792 | Ind. Eng. Chem. Res. 2013, 52, 91−107
Industrial & Engineering Chemistry Research Article

(68) Water Statistics in Mexico, ed. 2010; CONAGUA. Mexican

National Water Commission: Mexico, 2010.

107 dx.doi.org/10.1021/ie3003792 | Ind. Eng. Chem. Res. 2013, 52, 91−107