Journal of Cleaner Production 71 (2014) 128e138

Contents lists available at ScienceDirect

Journal of Cleaner Production

journal homepage: www.elsevier.com/locate/jclepro

Heat integrated resource conservation networks without mixing prior

to heat exchanger networks
Yin Ling Tan a, *, Denny K.S. Ng b, Dominic C.Y. Foo b, Mahmoud M. El-Halwagi c, d,
Yudi Samyudia a
a
Chemical Engineering Department, Curtin University, Sarawak Campus, CDT 250, 98009 Miri, Sarawak, Malaysia
b
Department of Chemical and Environmental Engineering/Centre Excellence for Green Technologies, The University of Nottingham, Malaysia Campus, Broga
Road, 43500 Semenyih, Selangor, Malaysia
c
Chemical Engineering Department, Texas A&M University, College Station, TX 77843, United States
d
King Abdul-Aziz University, Jeddah, Saudi Arabia

a r t i c l e i n f o a b s t r a c t

Article history: This paper presents a generic approach for the synthesis of heat integrated resource conservation net-
Received 7 August 2013 works (HIRCNs) of the fixed flow rate problem, where process sources linked directly to process sinks
Received in revised form without any prior mixing. The mixed integer non-linear program (MINLP) formulation complemented by
7 January 2014
floating pinch concept was developed to determine the optimum fresh material resources as well as hot
Accepted 7 January 2014
and cold energy utilities. The proposed approach is applicable for both concentration- and property-
Available online 18 January 2014
based direct reuse/recycle system with variable operating parameters (i.e. flow rates, temperatures
and properties). Three literature case studies are solved to illustrate the proposed approach.
Keywords:
Process integration

2014 Elsevier Ltd. All rights reserved.
Targeting
Property integration
Heat integration
Resource conservation
Optimisation

1. Introduction Process Integration techniques for resource conservation and

Huge amount of energy and fresh resources (i.e. water, chem-
icals, solvents) are consumed by process industries to achieve the
desired product throughput and quality. The current drive toward
sustainability and business competitiveness has driven the process
industries to effectively use these resources. Thus, resource con-
servation activities have become the centre of attention as
compared to conventional end-of-pipe waste treatment system.
With enormous developments in the past three decades, Pro-
cess Integration techniques have been widely accepted as effective
tools for resource conservation and waste reduction for the pro-
cess industry. El-Halwagi (2006) defines Process Integration as a
holistic approach to process design, retrofitting and operation
which emphasises the unity of the process. Available tools of
* Corresponding author. Tel.: þ60 85 443833; fax: þ60 85 443837.
E-mail addresses: tanyinling@gmail.com (Y.L. Tan), Denny.Ng@nottingham.edu.
my (D.K.S. Ng), Dominic.Foo@nottingham.edu.my (D.C.Y. Foo), el-halwagi@tamu.
edu (M.M. El-Halwagi), yudi.samyudia@curtin.edu.my (Y. Samyudia).
waste reduction can be generally categorised as Heat, Mass and
Property Integration. Several important reviews can be found in
literature, e.g. Furman and Sahinidis (2002) for Heat Exchange
Network (HEN), Dunn and El-Halwagi (2003) for Mass Exchange
Network, and Foo (2009) for Material Resource Conservation Net-
works (RCNs).
In the past decades, active developments were seen for RCNs,
which includes Water, Hydrogen and Property Integration. In all
cases, the main objective is to reduce both fresh resource con-
sumption and waste generation. Review on RCNs synthesis can be
found in textbooks, including both in-plant and inter plant material
recovery systems (El-Halwagi, 2006; Foo, 2012).
It should be noted that most of the RCNs works do not consider
temperature effect in the process streams. There are many cases
where both mass and heat recovery are equally important. For
instance, when dry air is used to remove volatile organic com-
pounds from wastewater stream in a stripper, the air stream needs
to possess specific temperature and properties before entering the
stripper. Therefore, simultaneous consideration of mass, property
and heat recovery should be addressed.

0959-6526/$ e see front matter
2014 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.jclepro.2014.01.014
 Y.L. Tan et al. / Journal of Cleaner Production 71 (2014) 128e138
In the past decades, extensive work was reported for simulta-
neous energy and water minimisation. These techniques can also
be classified as insight-based and mathematical optimisation ap-
proaches. The former are developed based on sequential approach;
while the latter considered both sequential and simultaneous ap-
proaches. As reported in the literature, various insight-based
techniques were developed but they are limited to single prop-
erty RCN problems. These techniques include two dimensional grid
diagram and separate systems with (Savulescu et al., 2005b) and
without water reuse (Savulescu et al., 2005a), water energy balance
diagram (Leewongtanawit and Kim, 2009), source-demand energy
composite curves (Savulescu et al., 2002), heat surplus diagram
(Manan et al., 2009), superimposed mass and energy curves (Wan
Alwi et al., 2011), stream merging principles (Feng et al., 2008),
graphical thermodynamic rules (Sorin and Savulescu, 2004), ther-
modynamic principles for threshold problem (Polley et al., 2010),
energy recovery algorithm (Sahu and Bandyopadhyay, 2010),
modified problem table algorithm (Bandyopadhyay and Saha, 2010)
and temperature versus concentration diagram (Martínez-Patiño
et al., 2011).
On the other hand, mathematical optimisation techniques have
been established to overcome limitation associated with insight-
based techniques. Sequential linear programming models have
been developed to first determine the minimum fresh water con-
sumption followed by minimum energy requirement. Detailed
Heat Integrated Water Network (HIWN) is then identified via
MINLP models. Bagajewicz et al. (2002) initiated the sequential
approach for the synthesis of Heat Integrated Water Networks
(HIWNs). The fresh water and energy targets are firstly achieved
using an LP formulation based on the necessary conditions of
optimality. In the second stage, an MINLP heat transhipment model
is generated. These models incorporate non-isothermal mixing as
well as forbidden and compulsory flow connections and heat
transfer matches (Bagajewicz et al., 2002).
Feng et al. (2009) analysed the interconnections between the
design of a water allocation network and the design of a heat
exchanger network. The authors discovered that reducing the
number of temperature local fluctuations along the sub-streams in
water networks improves the energy performance of the system. As
a result, mathematical model with this consideration was proposed
to synthesise a HIWN.
However, the above-mentioned works are primarily applicable
to fixed load problems. To overcome this limitation, George et al.
(2011) established a sequential approach for the fixed flowrate
problems which is applicable for both single and multiple con-
taminants problems with the incorporation of isothermal and non-
isothermal mixing of streams. In this approach, a linear program-
ming model is formulated to identify the fresh water target. As for
HEN model, a linear transshipment model is formulated for
isothermal mixing problem while a nonlinear programming model
with a discountinuous derivative (DNLP) is formulated for non-
isothermal mixing cases.
On the other hand, Liao et al. (2011) presented an approach for
HIWNs that allow splitting of hot and cold streams. Based on Liao
et al. (2011), an MILP model that treats the direct and indirect
heat transfer separately is developed to identify the promising
matches between hot and cold streams. Next, an MINLP model with
consideration of the splitting and non-isothermal mixing features
inside the HEN is initiated to achieve the desired HIWN.
Recently, Sahu and Bandyopadhyay (2012) extended the
concept of modified problem table algorithm (Sahu and
Bandyopadhyay, 2010) to formulate HIWNs as linear program-
ming model. The authors formulated three LP models for targeting
the fresh water and energy consumptions for isothermal and non-
isothermal mixing situations. The proposed formulation avoids the
129
sub-optimality issues of MINLP and DNLP formulations for the case
of non-isothermal mixing as found in the previous work (George
et al., 2011). However, iteration of different pinch points is
needed to identify the minimum energy requirement (Sahu and
Bandyopadhyay, 2012).
On the other hand, the total cost of HIWNs can be minimised via
simultaneous techniques. Leewongtanawit and Kim (2008) initi-
ated mathematical models for synthesis of HIWNs with multiple
contaminants. The authors formulate the overall problem as an
MINLP optimisation problem. Decomposition approach is intro-
duced to decompose the overall MINLP problem into MILP and NLP
sub-problems which are solved in sequence using an iterative
process. The method has also considered an economic trade-off
between water network and HEN, non-isothermal and generation
of separate systems. Furthermore, Bogataj and Bagajewicz (2008)
developed another MINLP model for HIWNs synthesis. The estab-
lished MINLP model includes the NLP formulation of water network
superstructure and the MINLP formulation of heat exchange
network superstructure for non-isothermal stream mixing.
Note that the models presented by Leewongtanawit and Kim
(2008) and Bogataj and Bagajewicz (2008) utilised heuristics to
reduce the number of hot and cold streams in the HEN, which had
reduced the size of the model. However, the limitation of these
approaches is that, the potential promising HIWNs may be
excluded based on the heuristics. As a result, Dong et al. (2008)
modify the state-space superstructure to formulate an MINLP
model which covers a broader network structures. However, the
model is very large in size and it is computational extensive when
the problem scale increases. Thus, Dong et al. (2008) established an
integrated optimisation strategy to improve the solution quality
and efficiency. The potential global optimum may be identified by
applying an interaction method proposed.
Ataei and Yoo (2010) proposed a sequential approach for mul-
tiple contaminant systems with the consideration of flowrate
changes and heat loss in the HIWNs. Firstly, an NLP model is
established to identify the feasible non-isothermal mixing points
that provide the overall network with minimum fresh water and
energy consumptions. Next, HEN is simplified through a new
generation of separate system in HEN (Ataei and Yoo, 2010).
Ahmetovi c and Kravanja (2013) presented a HIWN superstruc-
ture with direct heat exchange by the mixing of streams and indi-
rect heat exchange in heat exchangers. The model is formulated as
non-convex MINLP and the objective is to minimise the total annual
costs. Furthermore, the authors developed a set of new constraints
to identify the interconnecting hot and cold streams between water
network and HEN. Later, this model was extended to include
process-to-process streams and other streams within the overall
network for heat integration (Ahmetovic and Kravanja, 2014). Two
strategies were proposed for heat integration of process-to-process
streams. This extended model was also a non-convex MINLP while
the objective is to minimise the total annual cost of the network.
Recently, some work on simultaneous Property and Energy
Integration has been observed. The first work on simultaneous
Property and Energy Integration was presented by Kheireddine
et al. (2011). The authors took into consideration of the thermal
constraints in mass and property-based RCN. A nonlinear pro-
gramming (NLP) model was presented to minimise the total cost of
RCN, while satisfying a set of process and environmental con-
straints. In addition, the model also accounts for heat of mixing and
the interdependency of properties. Nevertheless, the proposed
model does not allow temperature adjustment through heaters and
coolers.
Thus, Rojas-Torres et al. (2013) established a systematic
approach for the synthesis of property-based RCN with property
interceptors where heaters and coolers are modelled as thermal
130 Y.L. Tan et al. / Journal of Cleaner Production 71 (2014) 128e138
interceptors. Furthermore, the developed model incorporates the
dependence of properties on temperature, thermal effects as well
as a new set of property operators which is used to track key
environmental properties such as odor, toxicity. However, the
proposed model only involves heaters and coolers without heat
integration. In many problems, energy recovery need to be maxi-
mised before external heating and cooling utilities are used.
To overcome this limitation, Tan et al. (2013) presented a meth-
odology for Heat Integrated Resource Conservation Network
(HIRCN) synthesis, covering both concentration- and property-
based direct reuse/recycle systems. The proposed method is able to
handle HIRCN problems with variable operating range of process
parameters, such as stream flowrates, temperatures and properties.
For uncertain temperature, the authors adopted the floating pinch
concept (Duran and Grossmann, 1986) to identify hot and cold util-
ities. However, the main drawback of the previous work is that, the
proposed superstructure is too complex and required discretisation
approach (Pham et al., 2009) to solve the MINLP problem, which may
lead to global or near global optimal results. Furthermore, when the
discretisation approach is used for complex problems (e.g. with
multiple property operators), the MILP model becomes computa-
tional intensive. Hence, it is desired to develop a simple and yet
generic model for the synthesis of a cost effective concentration- and
property-based HIRCN. This is the subject of this paper.
In this paper, a simple superstructure is presented for HIRCN
of the fixed flow rate problem. The problem is formulated as
MINLP model, which satisfies a set of process and environmental
constraints. It makes use of the floating pinch concept (Tan et al.
(2014) in solving cases with varying process parameters. Three
literature case studies are used to illustrate the proposed
method.
2. Problem statement
The problem definition of a HIRCN is given as follows:
Given NSOURCES number of process sources that can be
considered for reuse/recycle in the process sinks, or be discharged
as waste. Given also NSINKS number of process sinks which are
units that can accept process sources. Each process source i, has
fixed flow rate (Wi), property operator p (ji,p) and temperature (Ti).
Each sink j, has an acceptable range of flowrate (Gj), property (jj,p)
and temperature (Tj), given as in Eqs. (1)e(3).
Gmin
j  Gj  Gmax
j j˛NSINKS
jmin
j;p  jj;p  jj;p j˛NSINKS p˛NPROP
max
Tjmin  Tj  Tjmax j˛NSINKS
where Gmin
j ; Gmax
j ; jmin
j;p ; jj;p ; Tj
max min ; T max are the respective lower
j
and upper bounds of the admissible flow rate, property operator p
and temperature for sink j. NFRESH number of external fresh re-
sources may be purchased to supplement the requirement of the
sinks. Each fresh resource r has property operator p (jr,p) and
temperature (Tr); and its flow rate is to be determined as part of the
solution model. Fresh resources and process sources may undergo
external heating and cooling in order to meet the desired tem-
perature range of the sinks and waste.
A general linearised mixing rule for property integration is given
as follows (Shelley and El-Halwagi, 2000):
X
jðpÞ ¼ xm jm;p
m where Gj is the total flow rate of sink j.
where jm,p and jðpÞ are the operators for property p of stream m
and mixture property p respectively; while xm is the fractional
distribution of stream m of the total mixture flowrate. Eq. (4) is
needed to identify all possible mixing patterns for each property.
Each source is split and supplied to all sinks or discharge as
waste. These streams are intended for the HEN and their flowrates
is to be determined simultaneously within the HIRCN. Note that
these streams can take the form as a set of NHOT hot streams or a set
of NCOLD cold streams. In this work, hot stream is defined as a
stream with supply temperature (Ths ) higher than target tempera-
ture (Tht ); while cold stream has supply temperature (Tcs ) lower than
target temperature (Tct ). Therefore, based on the target temperature
of sink j, source i can be considered as either hot or cold stream.
However, if source i temperature falls within the target tempera-
ture range of sink j, source i can be split into hot and cold streams.
For instance, for a sink with acceptable temperature range of 300e
400 K, a source with temperature of 350 K may be split into hot
(Ths ¼ 350 K and Tht ¼ 300 Ke349 K) and cold streams (Tcs ¼ 350 K
and Tct ¼ 351 Ke400 K). In other words, the temperatures before
HEN and after HEN are set or bounded based on the source i and
sink j temperatures respectively. This simplification enables one to
categorise the HEN streams as hot or/and cold streams and extract
the respective temperatures directly from the given source and sink
temperature limiting data. More importantly, it has significantly
reduced the number of temperature variables and also the search
space of this formulation.
Besides, external hot (Qh) and cold (Qc) utilities are available to
further heat/cold the process streams to meet the process sinks
requirement, after maximising energy recovery between hot and
cold streams. It is assumed that no mass transfer occur in the heat
exchangers. Therefore, the compositions and flowrates of process
streams remain unchanged after the HEN. Fig. 1 shows the source-
HEN-sink superstructure of the problem. The objective of this work
is to synthesise a HIRCN of minimum cost, which may take the form
of minimum operating or total annualised costs. Literature case
studies are solved to illustrate the proposed model.
3. Model formulation
The following sub-sections present the models for both con-
centration- and property-based RCNs, as well as the HEN section of
the HIRCN (see source-HEN-sink superstructure in Fig. 1).
(1) 3.1. Concentration- and property-based RCN
The mass and energy balances for various sources and sinks can
(2)
be defined as follows:
Splitting of fresh source r:
(3) X
Fr ¼ fr;j r˛NFRESH (5)
j˛NSINKS
where fr,j is the flow rate of fresh source r to sink j.
Splitting of process source i:
X
Wi ¼ wi;j þ Bi;waste i˛NSOURCES (6)
j˛NSINKS
where wi,j and Bi,waste are the flow rates of source i recovered to sink
j and discharged as waste.
Mass balances at the mixing point before sink j:
X X
Gj ¼ wi;j þ fr;j j˛NSINKS (7)
(4) i˛NSOURCES r˛NFRESH
 Y.L. Tan et al. / Journal of Cleaner Production 71 (2014) 128e138 131

Process Sinks
Fresh 1 M j=1

...
Fr fr,j

.
r =NFRESH
Gj

.
wi,j

.
Process Sources i
i=1 M j=NSINKS

Wi .
.
.

i=NSOURCES M Waste
wi,waste
Heat Exchanger
Network (HEN)

Fig. 1. Sourceesink representation for a HIRCN.

Mass balance of waste: Energy balances at the mixing point before the sink j:
X
Bwaste ¼ Bi;waste (8)
i˛NSOURCES   P 
H;out

Gj CPj Tj  To ¼ mH H
r;j CPr;j Tr;j
r˛NFRESH
 To
where Bwaste is the flow rate of waste. P 
C;out

Based on Eq. (4), property operator balance for property p at the þ mC CPr;j
r˛NFRESH r;j
C
Tr;j  To
mixing point before sink j: P H H

H;out

h i þ m CP T  To
X i˛NSOURCES i;j i;j i;j
jj;p Gj ¼ ji;p wi;j P C C

C;out

i˛NSOURCES
þ m CPi;j Ti;j
i˛NSOURCES i;j
 To j˛NSINKS
X h i
þ jr;p fr;j j˛NSINKS p˛NPROP (9) (14)
r˛NFRESH

H , CP H , CP C , and CP C are heat capacities of sink j, hot

where CPj, CPr;j
Property operator balance at the mixing point before waste: i;j r;j i;j
H;out
h i and cold streams from fresh resource r and source i to sink j; Tr;j ,
X H;out C;out C;out
jwaste;p Bwaste ¼ ji;p Bi;waste p˛NPROP (10) Ti;j , Tr;j and Ti;j are the target temperatures of hot and cold
i˛NSOURCES streams from fresh resource r and source i to sink j, while To is the
reference temperature.
where jwaste,p is the property operator p for waste. Energy balance at the mixing point before waste:

  X   X  
H;out C;out
Bwaste CP waste T waste  To ¼ mH H
i;waste CPi;waste Ti;waste  To þ mCi;waste CPi;waste
C
Ti;waste  To (15)
i˛NSOURCES i˛NSOURCES

Note that Eqs. (9) and (10) are used to determine the mean where CPwaste, CPi;waste
H C
and CPi;waste are the heat capacities of waste,
H;out
property value of each sink j and waste, and should be carried out for hot and cold streams from source i to waste; Twaste, Ti;waste and
C;out
all concerned properties for each process sink. As discussed previ- Ti;waste is the temperature of waste, target temperatures of hot and
ously, the same stream sent from fresh resource r and source i to sink cold streams from source i to waste.
j and waste can take the form as hot and cold streams, if its tem-
perature falls in between the operating temperature range of the
3.2. Heat exchanger networks
sink and waste. Therefore, Eqs. (11)e(13) are included in the model.

The HEN floating pinch method established by Tan et al. (2014) is

fr;j ¼ mH C
r;j þ mr;j r˛NFRESH j˛NSINKS (11)
applied in this work. This method is used because of the variable
flow rates and temperatures in the problem formulation, which
wi;j ¼ mH C cannot utilise the established heat integration models for fixed
i;j þ mi;j i˛NSOURCES j˛NSINKS (12)
streams conditions. The background concept of this method is
based on the shifted hot and cold composite curves. According to
Bi;waste ¼ mH C
i;waste þ mi;waste i˛NSOURCES (13) Duran and Grossmann (1986), the potential pinch candidates are
those corner points on the composite curves which corresponds to
where mH H C C
r;j , mi;j , mr;j and mi;j are flowrate of hot and cold streams the inlet temperatures of any hot and cold streams. Therefore,
from fresh resource r and source i to sink j; mH C
i;waste and mi;waste are overall energy balance equation and energy balance equation
the flowrate of hot and cold streams from source i to waste. above or below pinch are developed for each of the postulated
132 Y.L. Tan et al. / Journal of Cleaner Production 71 (2014) 128e138

pinch point candidates, which will then identify the true pinch NCOLD sets are j(r þ i) þ i respectively and each of the entries
point and also the minimum hot and cold utilities. Detail descrip- represents a vector column.
tion of HEN floating pinch method can be referred to Tan et al. The potential pinch candidate, Tp are taken as the inlet tem-
(2014). The following section presents the HEN model of the HIRCN. peratures of the hot and cold streams,
The supply and target temperatures of the hot and cold streams
from fresh resource r and source i to sink j and waste are first 8
shifted based on minimum temperature of driving force, DTmin < h˛NHOT
Ths
(subtract DTmin/2 for hot streams; add DTmin/2 for cold streams) as Tp ¼ s c˛NCOLD (20)
: Tc
follow,
p˛NPINCH

2 in  DTmin
3 The total energy balance is to be used together with energy
Tr;j 2
6 7 balance above or below the pinch point candidate to identify the
6 « 7 true pinch point and to ensure thermodynamic feasibility. In this
6 7
6 7
6 in  DT2min 7 work, energy balance above the pinch point candidate is used.
6 TðNFRESHþiÞ;j 7 i˛NSOURCES
6 7 To determine the energy balance above the pinch point candi-
6 « 7 j˛NSINKS
Ths ¼ 6 7 dates, stream locations parameterisation is needed. In this work,
6 7 r˛NFRESH
6 7 binary variables are used to parameterise the stream locations,
6 in
TðNFRESHþNSOURCESþiÞ;waste  DT2min 7 h˛NHOT
6 7 given by the constraints in Eqs. (21)e(24).
6 7
6 « 7
4 5

in
TðNFRESHþNSOURCESþNSOURCESÞ;waste  DT2min 1 if Tht > Tp h˛NHOT
bth;p ¼ (21)
(16) 0 if Tht  Tp p˛NPINCH

2 out  DTmin
3 
Tr;j 2 1 if Ths > Tp h˛NHOT
6
6
7
7
bsh;p ¼ 0 if Ths  Tp p˛NPINCH
(22)
6 « 7
6 7
6 out  DT2min 7
6 TðNFRESHþiÞ;j 7 i˛NSOURCES
6 7 
t 6 « 7 j˛NSINKS 1 if Tct > Tp c˛NCOLD
Th ¼ 6 7
6 7 r˛NFRESH gtc;p ¼ (23)
6 7 0 if Tct  Tp p˛NPINCH
6 out
TðNFRESHþNSOURCESþiÞ;waste  DT2min 7 h˛NHOT
6 7
6 7
6 « 7
4 5 
 DT2min 1 if Tcs > Tp c˛NCOLD
out
TðNFRESHþNSOURCESþNSOURCESÞ;waste gsc;p ¼ (24)
0 if Tcs  Tp p˛NPINCH
(17)
where bh;p ,bh;p ,gtc;p and gsc;p are the binary integer. The energy
t s

2 in þ DTmin
3 balance above the pinch point candidate is expressed as:
Tr;j 2
6 7
6 « 7 n
6 7 X    o
6 7 Qh  mc CPc gsc;p Tp Tcs  gtc;p Tp Tct
6 in þ DT2min 7
6 TðNFRESHþiÞ;j 7 i˛NSOURCES c˛NCOLD
6 7 X n
s 6 « 7 j˛NSINKS t  
Tc ¼ 6 7  mh CPh bh;p Tp Tht (25)
6 7 r˛NFRESH
6 7
þ DT2min
h˛NHOT
6 in
TðNFRESHþNSOURCESþiÞ;waste 7 c˛NCOLD o
6 7 s 
6 7  bh;p Tp Ths p˛NPINCH
6 « 7
4 5
in
TðNFRESHþNSOURCESþNSOURCESÞ;waste þ DT2min where mh, mc, CPc, CPh, Tht and Tct are flow rate, heat capacities and
(18) target temperatures for hot and cold streams in HEN. Readers may
refer to Tan et al. (2014) for the details of the usefulness of Eq. (25).
2 3 The following equation shows the total energy balance.
out þ DTmin
Tr;j 2
6 7 X   X  
6 « 7
6 7 mh CPh Ths Tht  mc CPc Tst Tss þQh Qc ¼ 0
6 7
6 out þ DT2min 7 h˛NHOT c˛NCOLD
6 TðNFRESHþiÞ;j 7 i˛NSOURCES
6 7 (26)
6 « 7 j˛NSINKS
Tct ¼ 6 7
6 7 r˛NFRESH
6 7 Note that the total energy balance (Eq. (26)) together with en-
6 out
TðNFRESHþNSOURCESþiÞ;waste þ DT2min 7 c˛NCOLD
6 7 ergy balance above the pinch point candidate (Eq. (25)) with the
6 7
6 « 7 respective constraints (Eqs. (21)e(24)) are used to identify the true
4 5
pinch point and the minimum hot and cold utilities of a given
out
TðNFRESHþNSOURCESþNSOURCESÞ;waste þ DT2min problem.
(19) In order to ensure the connection between property-based RCN
model and HEN model, Eqs. (27e33) are used to relate the flow
where Ths and Tht are shifted supply and target temperatures of the rates and temperatures in these models.
hot stream, while Tcs and Tct are shifted supply and target temper- Flow rate of hot and cold streams from fresh resource r and
atures of the cold streams in HEN. Note that the size of NHOT and source i to sink j are classified as follow,
 Y.L. Tan et al. / Journal of Cleaner Production 71 (2014) 128e138 133

2 3 are given as $80/kW.y and 20$/kW.y (Kim et al., 2009). In addition,

mHr;j
6 « 7 total annual operating hour (k) is taken as 8000 h.
6 7 i˛NSOURCES
6 mH 7
6 ðNFRESHþiÞ;j 7 j˛NSINKS
mh ¼ 6
6 « 7
7
4.1. Case Study 1
6 7 r˛NFRESH
6 mH
ðNFRESHþNSOURCESþiÞ;waste 7
4 5 h˛NHOT
« Case Study 1 is an ammonia recovery case study taken from Wan
mH
ðNFRESHþNSOURCESþNSOURCESÞ;waste
Alwi et al. (2011). It involves a process plant that utilises ammonia
as the mass separating agent in a sour gas absorption column and as
(27)
a dust-cleaning agent. On the other hand, ammonia waste is pro-
2 3 duced in the calcium chloride production section of the plant,
mCr;j which is currently sent to waste treatment system. Therefore, to
6 « 7 reduce the waste generation, ammonia can be recovered into the
6 7 i˛NSOURCES
6 mCðNFRESHþiÞ;j 7
6 7 j˛NSINKS process units which require ammonia. The limiting data for this
mc ¼ 6
6 « 7
7 case study is shown in Table 1. In addition, fresh ammonia may be
6 7 r˛NFRESH
6 mCðNFRESHþNSOURCESþiÞ;waste 7 purchased externally and is available at 30  C, with unit cost of
4 5 c˛NCOLD
« $500/t and waste ammonia has to be discharged at 40  C (Wan Alwi
mCðNFRESHþNSOURCESþNSOURCESÞ;waste et al., 2011). In this case study, the major component in all stream is
(28) ammonia; thus, heat capacities of all streams (CPn, CPi, CPr and CPj)
are assumed to take a constant value of 2.19 kJ/kg K; and DTmin of
Supply temperature of hot and cold streams from fresh resource 35  C is used (Wan Alwi et al., 2011).
r and source i to sink j are the same as that of fresh resource r and The objective of this case study is to minimise annual operating
source i. cost (AOC), which consists of operating costs for fresh resources, as
H;in C;in
well as hot and cold utilities in the HIRCN. The optimisation
Tr;j ¼ Tr;j ¼ Tr r˛NFRESH j˛NSINKS (29) objective is given as follows:
( )
H;in C;in X
NFRESH
Ti;j ¼ Ti;j ¼ Ti i˛NSOURCES j˛NSINKS (30) min AOC ¼ k* Costr Fr þ Costh Qh þ Costc Qc (35)
Fr ;Qh ;Qc
r¼1
Target temperature of hot and cold streams from fresh resource
r and source i to sink j are based on the temperature of sink j. where Costr is the cost of fresh resources r.
Eq. (35) is solved subject to the constraints in Eqs. (1)e(9) and
H;out C;out
Tr;j ¼ Tr;j ¼ Tj r˛NFRESH j˛NSINKS (31) (11)e(34). The optimised HIRCN is shown in Fig. 2. As shown, the
HIRCN consume a fresh ammonia flow rate (Fr) of 654.9 kg/s, and
H;out C;out with utility targets of Qh ¼ 132,927 kW and Qc ¼ 79,228 kW. To
Ti;j ¼ Ti;j ¼ Tj i˛NSOURCES j˛NSINKS (32)
further verify the results of Qh and Qc, the HEN for this case study is
Supply and target temperature of hot and cold streams from synthesised using the pinch design method (Linnhoff et al., 1982),
source i to waste are the same as that of source i and waste. and is presented in Fig. 3. Note that in this case study, the unit cost
of fresh ammonia is much higher than that of hot and cold utilities.
H;in
Ti;waste C;in
¼ Ti;waste ¼ Ti i˛NSOURCES (33) As a result, fresh ammonia is minimised while allowing using more
utilities. It is worth mentioning that the results are identical with
the reported result in Sahu and Bandyopadhyay (2012), if a two
H;out C;out
Ti;waste ¼ Ti;waste ¼ T waste i˛NSOURCES (34) stage LP model is solved such as that in Case Study 1.

It is worth noting that the above model is simpler as compared 4.2. Case Study 2
to that presented by earlier work Tan et al. (2013). This model only
involves one splitting point (for sources) and one mixing point (for Case study 2 is a single property-based water network which
sinks). However, the model presented by Tan et al. (2013) consists adapted from Sahu and Bandyopadhyay (2012). It consists of two
of two mixing points (i.e. one for HEN and the other for the sinks) process sinks and two process sources, with limiting data given in
and two splitting points (one for sources and the other for HEN) Table 2. Toluene is assumed as the key contaminant in this case
which is similar to a non-convex pooling problem formulation. study. The available fresh resource is given as pure fresh water
Besides, in Tan et al. (2013), the supply and target temperatures for (0 ppm), with unit cost of $0.45/t and available at 20  C. Besides,
HEN are unknown variables as they depend on the flow rate of wastewater has to be discharged at temperature of 30  C (Sahu and
sources being mixed. Hence, this leads to non-linearity with Eq.
(25), which makes the model to be highly non-convex and Table 1
computational intensive. As a result a solution strategy is needed. Data for Case Study 1.
However, in this work, the proposed model avoids such situation, as Flowrate (kg/s) Temperature ( C) Concentration (ppm)
the supply and target temperatures are directly based on the
Sink
temperature of sources and sinks of known values.
SK1 350 30 0
SK2 677 187 40
4. Case studies SK3 126 55 75
SK4 202 98 100
To illustrate the proposed model, three literature case studies Source
are solved. All cases are solved using Extended LINGO v11.0 with SR1 530 21 30
Global Solver. Solution strategy is not required in this proposed SR2 68 43 150
model as the solver is able to provide the global optimal solution. In SR3 1130 130 300
SR4 36 35 500
all case studies, the costs of hot and cold utilities (Costh and Costc)
134 Y.L. Tan et al. / Journal of Cleaner Production 71 (2014) 128e138

Fig. 2. Optimal solution for Case Study 1.

Fig. 3. HEN for Case Study 1.

Bandyopadhyay, 2012). The minimum approach temperature for
HEN (DTmin) is assumed as 10  C for the case study.
In this case study, all the heat capacities of sink, source, fresh
resource as well as hot and cold streams (CPj, CPi, CPr and CPn) can
be determined via equation below:
X (gmol K2), 151.27  105 J/(gmol K3), and 1630  109 J/(gmol K4)
CP ¼ xk CPk k˛NCOMP (36)
where xk is the mole fraction of component k and CP for each
component can be determined using Eq. (37). Note that the heat
capacity values are temperature-dependence.
CPk ¼ ak þ bk T þ ck T 2 þ dk T 3 k˛NCOMP (37)
where ak, bk, ck and dk are parameters in temperature-dependent
expression for heat capacity of each component. a, b, c and d parame-
ters for toluene are given as 1.8083 J/(gmol K), 81.222  102 J/
while a, b, c and d parameters for water are 18.2964 J/(gmol K),
47.212  102 J/(gmol K2), 133.88  105 J/(gmol K3), and
1314.2  109 J/(gmol K4).
The objective of this case study is to minimise the AOC, which
consists of operating costs for fresh water, as well as hot and cold
 Y.L. Tan et al. / Journal of Cleaner Production 71 (2014) 128e138 135

Table 2 the sinks are to be fulfilled. Table 3 shows the limiting data for
Data for Case Study 2. process sinks and sources for this case study.
Flowrate (kg/s) Temperature ( C) Concentration (ppm) Equations that follow outline the mixing rules for toxicity (Tox),
Sink
THOD, density (r) and viscosity (m) (Nápoles-Rivera et al., 2010),
SK1 100 100 50 P
SK2 40 75 50 Tox ¼ x Toxi
i i
(38)
SK3 166.67 100 800

Source
P
SR1 100 100 100
THOD ¼ x THODi
i i
(39)
SR2 40 75 800
SR3 166.67 100 1100 1 P 1
¼ x (40)
r i ir
i

P
utilities in the HIRCN. Solving the optimisation objective in Eq. (35), log m ¼ x log
i i
mi (41)
subject to the constraints in Eqs. (1)e(9), (11)e(34), (36) and (37)
yield the minimum AOC of $1,302,222. The optimal HIRCN is In addition, mixing rules of pH (Hortua et al., 2013) for different
shown in Fig. 4. The fresh water flow rate (Fr) is determined as pH range are also given as follow.
79.76 kg/h; while the hot (Qh) and cold utilities (Qc) for the HIRCN For acid mixing (0  pH  7):
are determined as 3348.38 kW and 34.80 kW. To further verify the P
results of Qh and Qc, the HEN for this case study is synthesised using 10pH ¼ x 10
i i
pHi
(42)
the pinch design method (Linnhoff et al., 1982), and is presented in
Fig. 5. As shown, only one heater with a duty of 3348.38 kW is For base mixing (7  pH  14):
needed. This shows the same result as the energy targets obtained P pHi 14
by the proposed model. 10pH14 ¼ x 10
i i
(43)
On the other hand, the same case study was solved by Tan et al.
(2013) with constant heat capacities of 4.2 kJ/kg K and reported an For neutralisation between acid and base streams:
AOC of $1,383,822 (Fr ¼ 77.27 kg/h, Qh ¼ 4473 kW and P P
Qc ¼ 1227 kW). It is worth mentioning that if this case study is 10pH ¼ x
Acid acid
10pHacid  x
Base base
10pHbase (44)
solved using the proposed model in this work with the same con-
stant heat capacities (solving Eq. (35) subject to constraints in Eqs. Unit cost for fresh resource is given as $0.0009/lb (Costr)
(1)e(9) and (11)e(34), higher fresh water flow rate (Fr ¼ 79.7 kg/h), (Nápoles-Rivera et al., 2010). In addition, all the heat capacities in
but lower utility targets (Qh ¼ 3347.5 kW and Qc ¼ 0 kW). This leads this case study (CPn, CPi, CPr and CPj) can be determined via Eqs. (36)
to a lower AOC of $1,300,764. and (45).

CPk ¼ ak þ bk T k˛NCOMP (45)

4.3. Case Study 3
Case Study 3 is adopted from Nápoles-Rivera et al. (2010), where
multiple properties are considered, i.e. concentration, toxicity,
THOD, pH, density, viscosity and temperature. The lower and upper
bound constraints on concentration, toxicity, THOD, pH, density,
viscosity and temperature ensure that the operational conditions of
where ak and bk are parameters in linearised temperature-
dependent expression for heat capacity of each component. Since
this case study involves a binary system with phenol and water, a
and b parameters for phenol are taken as 0.4685 J/(g K) and
0.0044 J/(g K2) while a and b parameters for water are1.3724 J/(g K)
and 0.0083 J/(g K2).

Fig. 4. Optimal solution for Case Study 2.

136 Y.L. Tan et al. / Journal of Cleaner Production 71 (2014) 128e138

Fig. 5. HEN for Case Study 2.

Table 3
Data for Case Study 3.

Sink Flowrate (lb/h) Composition (ppm) Toxicity (%) THOD (mg O2/l) pH Density (lb/l) Viscosity (cP) Temperature ( F)

SK1 3000 0e0.013 02 075 5.9e8.0 1.8e2.8 0.871e1.202 157e203

SK2 1900 0e0.011 02 075 5.7e7.9 1.7e2.5 0.782e1.430 113e135
Waste e e e e e e e 93

Source Flowrate (lb/h) Composition (ppm) Toxicity (%) THOD (mg O2/l) pH Density (lb/l) Viscosity (cP) Temperature ( C)

SR1 2900 0.033 0.8 75 5.3 2.000 1.256 176

SR2 2450 0.022 0.5 88 5.1 2.208 1.220 149
Fresh e 0 0 0 7.0 2.204 1.002 77

Fig. 6. Optimal solution for Case Study 3.

 Y.L. Tan et al. / Journal of Cleaner Production 71 (2014) 128e138
Fig. 7. HEN for Case Study 3.
Solving the optimisation objective in Eq. (35), subject to the
constraints in Eqs. (1)e(3), (5)e(34), (36) and (38)e(45) results
with the minimum AOC of $253,917/y, with the optimal values of Fr,
Qh and Qc determined as 3523.07 lb/h, 0 Btu/h and 43,679.75 Btu/h.
The optimal HIRCN and the HEN design for this case study are
shown in Figs. 6 and 7. As observed from the HEN design in Fig. 7,
four coolers are needed for this HIRCN, with a total duty of
43,679.75 Btu/h. This matches the targeted value as obtained via
the proposed model.
5. Conclusion
A new generic model for the synthesis of HIRCN is presented in
this work. An MINLP formulation has been developed to identify
the minimum cost of a HIRCN, which simultaneously optimised the
fresh resources as well as the external hot and cold utilities.
Furthermore, the methodology is able to solve for problems with
varied process parameters (e.g. flow rates, temperatures and
properties). Three case studies of single and multiple properties are
used to demonstrate the proposed methodology.
References
Ahmetovi c, E., Kravanja, Z., 2013. Simultaneous synthesis of process water and heat
exchanger networks. Energy 57, 236e250.
Ahmetovi c, E., Kravanja, Z., 2014. Simultaneous optimization of heat-integrated
water networks involving process-to-process streams for heat integration.
Appl. Therm. Eng. 62, 302e317.
Ataei, A., Yoo, C.K., 2010. Simultaneous Energy and water optimization in multiple-
contaminant systems with flowrate changes consideration. Int. J. Environ. Res.
4, 11e26.
Bagajewicz, M., Rodera, H., Savelski, M., 2002. Energy efficient water utilization
systems in process plants. Comput. Chem. Eng. 26, 59e79.
Bandyopadhyay, S., Saha, G.C., 2010. Modified problem table algorithm for energy
targeting. Ind. Eng. Chem. Res. 49, 11557e11563.
Bogataj, M., Bagajewicz, M.J., 2008. Synthesis of non-isothermal heat inte-
grated water networks in chemical processes. Comput. Chem. Eng. 32,
3130e3142.
Dong, H.-G., Lin, C.-Y., Chang, C.-T., 2008. Simultaneous optimization approach for
integrated water-allocation and heat-exchange networks. Chem. Eng. Sci. 63,
3664e3678.
Dunn, R.F., El-Halwagi, M.M., 2003. Process integration technology review: back-
ground and applications in the chemical process industry. J. Chem. Technol.
Biotechnol. 78, 1011e1021.
137
Duran, M.A., Grossmann, I.E., 1986. Simultaneous optimization and heat integration
of chemical processes. AIChE J. 32, 123e138.
El-Halwagi, M.M., 2006. Process Integration. Academic Press, New York, USA.
Feng, X., Li, Y., Shen, R., 2009. A new approach to design energy efficient water
allocation networks. Appl. Therm. Eng. 29, 2302e2307.
Feng, X., Li, Y., Yu, X., 2008. Improving energy performance of water allocation
networks through appropriate stream merging. Chin. J. Chem. Eng. 16,
480e484.
Foo, D.C.Y., 2009. State-of-the-Art review of pinch analysis techniques for water
network synthesis. Ind. Eng. Chem. Res. 48, 5125e5159.
Foo, D.C.Y., 2012. Process Integration for Resource Conservation. CRC Press, Boca
Raton, Florida, USA.
Furman, K.C., Sahinidis, N.V., 2002. A critical review and annotated bibliography for
heat exchanger network synthesis in the 20th century. Ind. Eng. Chem. Res. 41,
2335e2370.
George, J., Sahu, G.C., Bandyopadhyay, S., 2011. Heat integration in process water
networks. Ind. Eng. Chem. Res. 50, 3695e3704.
Hortua, A.C., El-Halwagi, M.M., Ng, D.K.S., Foo, D.C.Y., 2013. Integrated approach for
simultaneous mass and property integration for resource conservation. ACS
Sustain. Chem. Eng. 1, 29e38.
Kheireddine, H., Dadmohammadi, Y., Deng, C., Feng, X.A., El-Halwagi, M., 2011.
Optimization of direct recycle networks with the simultaneous consideration of
property,mass, and thermal effects. Ind. Eng. Chem. Res. 50, 3754e3762.
Kim, J., Kim, J., Kim, J., Yoo, C., Moon, I., 2009. A simultaneous optimization approach
for the design of wastewater and heat exchange networks based on cost esti-
mation. J. Clean. Prod. 17, 162e171.
Leewongtanawit, B., Kim, J.-K., 2008. Synthesis and optimisation of heat-integrated
multiple-contaminant water systems. Chem. Eng. Process 47, 670e694.
Leewongtanawit, B., Kim, J.-K., 2009. Improving energy recovery for water mini-
misation. Energy 34, 880e893.
Liao, Z., Rong, G., Wang, J., Yang, Y., 2011. Systematic optimization of heat-integrated
water allocation networks. Ind. Eng. Chem. Res. 50, 6713e6727.
Linnhoff, B., Townsend, D.W., Boland, D., Hewitt, G.F., Thomas, B.E.A., Guy, A.R.,
Marshall, R.H., 1982. A User Guide on Process Integration for the Efficient Use of
Energy. IChemE, Rugby, UK.
Manan, Z.A., Tea, S.Y., Wan Alwi, S.R., 2009. A new technique for simultaneous water
and energy minimisation in process plant. Chem. Eng. Res. Des. 87, 1509e1519.
Martínez-Patiño, J., Picón-Núñez, M., Serra, L.M., Verda, V., 2011. Design of water
and energy networks using temperatureeconcentration diagrams. Energy 36,
3888e3896.
Nápoles-Rivera, F., Ponce-Ortega, J.M., El-Halwagi, M.M., Jimenez-Gutierrez, A.,
2010. Global optimization of mass and property integration networks with in-
plant property interceptors. Chem. Eng. Sci. 65, 4363e4377.
Pham, V., Laird, C., El-Halwagi, M., 2009. Convex Hull discretization approach to the
global optimization of pooling problems. Ind. Eng. Chem. Res. 48, 1973e1979.
Polley, G.T., Picon-Nunez, M., Lopez-Maciel, J.D., 2010. Design of water and heat
recovery networks for the simultaneous minimisation of water and energy
consumption. Appl. Therm. Eng. 30, 2290e2299.
Rojas-Torres, M.G., Ponce-Ortega, J.M., Serna-González, M., Nápoles-Rivera, F., El-
Halwagi, M.M., 2013. Synthesis of water networks involving temperature-based
property operators and thermal effects. Ind. Eng. Chem. Res. 52, 442e461.
138 Y.L. Tan et al. / Journal of Cleaner Production 71 (2014) 128e138
Sahu, G.C., Bandyopadhyay, S., 2010. Energy conservation in water allocation net-
works with negligible contaminant effects. Chem. Eng. Sci. 65, 4182e4193.
Sahu, G.C., Bandyopadhyay, S., 2012. Energy optimization in heat integrated water
allocation networks. Chem. Eng. Sci. 69, 352e364.
Savulescu, L., Kim, J.K., Smith, R., 2005a. Studies on simultaneous energy and water
minimisation e Part I: Systems with no water re-use. Chem. Eng. Sci. 60, 3279e3290.
Savulescu, L., Kim, J.K., Smith, R., 2005b. Studies on simultaneous energy and water
minimisation e Part II: Systems with maximum re-use of water. Chem. Eng. Sci.
60, 3291e3308.
Savulescu, L.E., Sorin, M., Smith, R., 2002. Direct and indirect heat transfer in water
network systems. Appl. Therm. Eng. 22, 981e988.
Shelley, M.D., El-Halwagi, M.M., 2000. Component-less design of recovery and
allocation systems: a functionality-based clustering approach. Comput. Chem.
Eng. 24, 2081e2091.
Sorin, M., Savulescu, L., 2004. On minimization of the number of heat exchangers in
water networks. Heat. Transfer Eng. 25, 30e38.
Tan, Y.L., Ng, D.K.S., El-Halwagi, M.M., Foo, D.C.Y., Samyudia, Y., 2013. Synthesis of
heat integrated resource conservation networks with variable operating pa-
rameters. Ind. Eng. Chem. Res. 52, 7196e7210.
Tan, Y.L., Ng, D.K.S., El-Halwagi, M.M., Foo, D.C.Y., Samyudia, Y., 2014. Floating pinch
method for utility targeting in heat exchanger network (HEN). Chem. Eng. Res.
Des. 92, 119e126.
Wan Alwi, S.R., Ismail, A., Manan, Z.A., Handani, Z.B., 2011. A new graphical
approach for simultaneous mass and energy minimisation. Appl. Therm. Eng.
31, 1021e1030.