water retrofit

RETROFIT OF WATER NETWORK WITHREGENERATION USING WATER PINCHANALYSIS

Y. L. Tan1, Z. A. Manan2, and D. C. Y. Foo3

1School of Engineering & Science, Curtin University of Technology, Sarawak, Malaysia.2Chemical Engineering Department, Universiti Teknologi Malaysia, Johor, Malaysia.3School of Chemical and Environmental Engineering, University of Nottingham Malaysia, Selangor

Darul Ehsan, Malaysia.

Abstract: This paper presents the development of a new systematic technique for the retrot ofwater network with regeneration based on water pinch analysis. The procedure consists of twoparts: retrot targeting and design for a water network with regeneration unit(s). In the targetingstage, retrot targets (utility savings and capital investment) were determined for a range of processparameters (total owrate and/or outlet concentration of the regeneration unit) to obtain a savingsversus investment curve. Next, the existing water network was re-designed to meet the chosen tar-gets. A case study on paper making process was used to demonstrate the new methodology.

Keywords: water minimization; pinch analysis; retrot utility targeting; retrot; network design;water regeneration.

INTRODUCTION

Water is extensively utilized in chemical, pet-rochemical, pulp and paper and many otherindustries. Stricter environmental regulations,scarcity of quality industrial water and therising cost of wastewater treatment haveencouraged process plants to reduce waterusage. Concurrently, the development of sys-tematic techniques for water reduction, reuseand regeneration within a process plant hasseen extensive progress.The advent of water pinch analysis (WPA)

as a tool for the design of optimal water net-work has been one of the most signicantadvances in the area of water minimizationover the past decade. Through WPA, wateruse can be minimized through:

(1) Process changes. The inherent demandfor water can be reduced through thereplacement of process equipment, e.g.,wet cooling towers by dry air coolers.

(2) Reuse/recycle.Wastewater can be directlyreused in other water-using operations orrecycled within an individual operation, pro-vided that the level of contaminant does notinterfere with the process.

(3) Regeneration. Wastewater can be puri-ed by partial treatment to improve itsquality in order to be reused or recycledin a water network. Different types of puri-cation techniques such as ltration,

activated carbon, biological treatment,membranes, and so on may be appliedindependently or in combination.

The power of WPA is in its ability to locate theminimum utility targets (fresh water consump-tion and wastewater generation) prior todetailed network design. This provides abase line for any water network to be syn-thesized. WPA has been relatively well estab-lished for synthesis of grassroots waternetworks (Wang and Smith, 1994; Polleyand Polley, 2000; Hallale, 2002; Mananet al., 2004) as well as for retrot situations(Tan and Manan, 2003, 2004) particularlyusing the approach of reuse/recycle. How-ever, the potential utility reduction that canbe realized through direct reuse/recycle canbe rather restricted. By coupling reuse/recycle strategy with regeneration, thisallows further utility reductions in a water net-work. Nevertheless, most of the work incor-porating regeneration strategy has beenfocused on the development of grassrootsdesign. These approaches may not be appli-cable for retrot since various constraints onan existing site including technical as well aseconomics needed to be taken into consider-ation during retrot. To achieve larger watersavings for existing processes, there is aclear need to develop a systematic techniquefor cost-effective water network retrot whichincorporates regeneration strategy.

305 Vol 85 (B4) 305317

Correspondence to:Dr Z.A. Manan, ChemicalEngineering Department,Universiti TeknologiMalaysia, 81310 Skudai,Johor, Malaysia.E-mail: [email protected]

DOI: 10.1205/psep06040

09575820/07/$30.00 0.00

Process Safety andEnvironmental Protection

Trans IChemE,Part B, July 2007

# 2007 Institutionof Chemical Engineers

PREVIOUS WORK ON WATER REGENERATIONREUSE/REGENERATION RECYCLING

A number of methods related to the synthesis of grassrootswater network involving regeneration have been developed.These methods can be generally classied into two maingroups. The rst group involves techniques based on graphi-cal WPA while the second group involves various mathemat-ical-based optimization approaches.Wang and Smith (1994) proposed the rst pinch-based

method to maximise savings in a water network with reuse,recycling and regeneration strategies. The concept of limitingcomposite curves that was originally developed for utility tar-geting in water reuse/recycling network was extended toinclude targeting for network with regenerationreuse andregenerationrecycling schemes. The minimum utility targetsare located prior to detailed network design.However, Kuo and Smith (1998) later pointed out that this

approach may fail to obtain the true utility targets when thepinch points are relocated after regeneration. They proposeda new methodology where the minimum water targets arerened by migrating streams that have been classied intodifferent water groups which include streams that are fedby fresh water and those that require regenerated water.The number of regeneration and efuent treatment unitstargets were also included in their approach.Castro et al. (1999) later extended the regenerationreuse

approach to take into consideration multiple pinch problems.Minimum utility targets are achieved by using the watersource diagram during the design stage. However, networkachieved by this method mostly does not contain the mini-mum number of units due to the splitting of operations. Toovercome this problem, additional utility is needed.The major drawback in the abovementioned approaches is

the assumption of water-using processes as mass transferoperations. Water that is used as a reactant or produced asreaction by-product; as well as water as cooling and heatingmedia in cooling towers and boilers may not be appropriatelyrepresented as mass transfer operations. To overcome thislimitation, Hallale (2002) established an alternative graphicaltargeting method called the water surplus diagram that isapplicable to non-mass transfer-based operations. Thisapproach locates the minimum utility targets for a grassrootswater network with reuse/recycle scheme. It also providessome guidelines for the placement of regeneration units topurify water sources and further reduce in utility consumption.A tabular approach known as water cascade analysis

(WCA) technique was recently introduced by Manan et al.(2004) to substitute the tedious graphical approach of watersurplus diagram. The WCA technique, which is based on theprinciples of water surplus diagram (Hallale, 2002) allowsquick and accurate determination of water targets as well asassessment of options for regeneration and process changes.The rst mathematical optimization-based approach for

water regeneration was introduced by Takama et al. (1980).They address the problem of designing optimal waterrecovery network for a petroleum renery by generating asuperstructure of all possible re-use and regeneration oppor-tunities. Optimisation is then performed on the superstructureto remove uneconomic design features.Alva-Argaez et al. (1998) later proposed an integrated

approach for combining the insights from water pinch andmathematical programming for the synthesis of grassroots

water network. All possibilities for water reuse, regener-ationreuse and regenerationrecycling are considered inthe model. The generated network features the minimumtotal annual cost and takes into consideration process con-straints such as geographical, control and safety. Otherrecently developed works for water network synthesis basedon mathematical optimization include those of Benko et al.(2000), Bagajewicz and Savelski (2001) and Xu et al. (2003).

PREVIOUS WORK ON HEAT, MASS AND WATERNETWORK RETROFIT

Techniques developed for the retrot of existing water net-work are mainly based on the established concept of heatand mass exchange network retrot. These techniques willbe briey reviewed here to provide basic understanding forthe newly developed work presented next.The rst pinch-based retrot approach was introduced by

Tjoe and Linnhoff (1986) for heat exchange network (HEN)problems. The idea is to minimise energy consumption in aHEN by installing additional heat transfer area and bymaking effective use of the existing heat exchangers. In thetargeting stage, an area versus energy plot for the optimumgrassroots design can be constructed for a given range ofminimum approach temperature, DTmin (Figure 1). The totalarea efciency, aArea is next utilized to determine the appro-priate retrot prole on the plot (Tjoe and Linnhoff, 1986).The total area efciency aArea is dened as a ratio betweenthe energy consumption for an optimum grassroots design(Atarget) to that of an existing network (Aexisting), as follows:

aArea AtargetAexisting

Energy

(1)

From the area versus energy plot, utility reduction and newarea requirement are converted into an annual utility savingsversus investment plot (Figure 2). Once an acceptable invest-ment limit is specied, a DTmin associated with the retrottedutility target is determined. In the retrot design stage, cross-pinch exchangers are eliminated and existing heat exchan-gers are revamped to achieve better heat recovery.Although Tjoe and Linnhoff (1986) argued that a conserva-

tive estimation is preferable for the retrot prole with con-stant aArea value, later works (Silangwa, 1986; Shenoy,1995; Ahmad and Polley, 1990) show that the constantaArea value may be too conservative, and hence a better esti-mation is needed. Silangwa (1986) pointed out that, when the

Figure 1. Area versus energy plot.

Trans IChemE, Part B, Process Safety and Environmental Protection, 2007, 85(B4): 305317

306 TAN et al.

value of aArea is low (i.e., aArea , 0.9), an incremental valueof DaArea 1 should be used. The incremental value ofDaArea is dened as

DaArea DAtargetDAexisting

DEnergy

(2)

where DAtarget is the minimum targeted extra area needed in agrassroots network for a reduction of DEnergy; while DAexistingis the newly installed area in the retrotted network for anenergy reduction of DEnergy. The retrot prole plotted usingDaArea is also shown in Figure 1. Note from Figure 1, theDaArea retrot prole is moving closer to the optimum grass-roots prole, as compared to the aArea retrot prole to yielda less conservative retrot target. Both of these HEN retrottargeting proles will be evaluated for their adaptability in thework involving water network retrot described in this paper.Fraser and Hallale (2000) later extended the HEN retrot

work into mass exchange network (MEN) problem that wasinitiated by El-Halwagi and Manousiouthakis (1989). Theirretrot work was based on the constant aArea efciencyapproach of Tjoe and Linnhoff (1986). They demonstratedthat pinch-based retrot approach for HEN can be appliedto the MEN problem with appropriate modications. The areaversus energy plot by Tjoe and Linnhoff (1986) was adaptedto become the stage versus load plot in MEN retrot. In thisstage-load diagram, a retrot path is chosen to allow one todetermine the savings achieved for extra new stages to beadded in the retrotted MEN. This yields the same savingsversus investment plot that originally developed for HEN retro-t (Figure 2). Lastly, elimination of cross-pinch mass transferand the evaluation of driving force use are the key principlesin the retrot design stage (Fraser and Hallale, 2000).Later development has seen the pinch-based retrot

approach that was extended for the special case of MEN,i.e., water minimization problems. The rst work on this wasdedicated for the retrot of non-mass transfer-based water-using operations by Tan and Manan (2003). This approachconsists of three main steps. Firstly, water network is rep-resented by a new concentration block diagram (CBD).This diagram provides the information to diagnose the poten-tial of retrot in an existing water network according to a set ofheuristics. Lastly, the water network will go through an evol-ution procedure in order to generate the nal retrot scheme.More recently,TanandManan (2004)establishedanother ret-

rot method for water network with mass transfer-based

operations which was adapted from MEN retrot method withconstanta efciency (Fraser and Hallale, 2000). The retrot tar-gets (minimum utility and number of stages for the massexchangers) are achieved through WPA grassroots targetingand MEN capital cost techniques. These targets are thenused to plot the number of stages versus fresh water owratediagram where a retrot path is formed by comparing the exist-ing designwith the targets.Givenanacceptable paybackperiodor investment limit, a globalminimum composition difference (1)in accordancewith these economic criteria is determined. In theretrot design stage, elimination of cross-pinch mass exchan-gers is performed and retrot of the existing water networkthrough Wang and Smith (1994) design rules is utilised to ef-ciently reuse the existing number of mass transfer stages.Alternatively, two recent water network retrot works based

on mathematical-based optimization approach are indepen-dently developed by Jodicke et al. (2001) and Huang et al.(1999). Various retrot constraints such as process location,number and size of the holding tanks, pipe work and treat-ment capacity are taken into consideration by these authorsto synthesize an economically feasible network.Yet other works on mathematical programming for water

network retrot problems have also been developed (e.g.,Parthasarathy and Krishnagopalan, 2001; Jacob et al.,2002; Thevendiraraj et al., 2003; Koppol et al., 2003). How-ever the biggest pitfall in these works is that, they weremainly based on grassroots synthesis approaches, wherethe main focus was aimed on the minimization of utility con-sumption. However, it is quite impossible to achieve an opti-mal retrot without taking into consideration the variousprocess and equipment design constraints. This may causemajor modications with long payback periods in some cases.Agood retrot approach should exploit opportunities tomaximizeusageof existing facilitieswhile trying tominimizeutility cost. Thisoften makes a retrotted network looks quite different from theoptimum grassroots design (Tjoe and Linnhoff, 1986).This paper presents a new systematic technique based on

pinch analysis for the retrot of water network with regener-ation unit(s). This technique, which is applicable to mass trans-fer and non-mass transfer-based operations consists of twostages namely targeting and network design. In the targetingstage, the utility savings versus capital investment plot weregenerated for a range of the regeneration units owrate oroutlet concentration. An optimum regeneration owrate and/or outlet concentration based on a maximum payback periodor capital investment limit can be determined from this plot.During the network design stage, the existing water networkwas revamped and new regeneration unit was introduced.

PROBLEM STATEMENT AND BASIC ASSUMPTIONS

The problem statement of retrotting water network withregeneration can generally be stated as follows:

Given a set of mass transfer-based and non-mass trans-fer-based water-using processes, it is desired to retrotan existing water network through wastewater regener-ation, process stream restructuring and making effectiveuse of existing process units to accomplish the bestsavings in operating costs, subject to a minimum paybackperiod or a maximum capital expenditure.

The following assumptions were made in developing the ret-rot procedure:

Figure 2. Savings versus investment plot.


RETROFIT OF WATER NETWORK WITH REGENERATION USING WATER PINCH ANALYSIS 307

(1) The system operates as a single contaminant system.(2) The system operates isothermally.(3) Regeneration reuse/recycling are allowed in the system.(4) Single type of regeneration treatment.

PAPER MILL CASE STUDY

Water reuse/recycling in paper mills is considered to be auniversal practice for recovering valuable paper bres from apaper machines excess water (Wiseman and Ogden, 1996).Apart from operating cost savings and reduction in theenvironmental impact of a process, water reuse and recyclingenables the recovery of raw materials from a water network.A local paper mill was used as a case study to demonstrate

the newmethodology. The mill produces paper from old news-papers andmagazines. Its rawwater treatment plant treat riverwater with high content of suspended solids and dissolvedsolids at an operating cost of $0.043 m23. The mill watersystem is served by a complex water network with a freshwater consumption of 1989.06 ton h21 and wastewater gener-ation of 1680.3 ton h21. A simplied version of the existingwater network for the case study is illustrated in Figure 3.In this paper mill process, used paper is fed to the pulpers

where it is blended with dilution water and chemicals to formpulp slurry called stock. The paper sheet formation beginswhen the stock from pulpers is sent to the forming sectionof the paper machine. Figure 3 shows that a total of986.52 ton h21 of fresh water is fed to the paper machinevia streams 1 and 2 to remove debris while wastewater isremoved from the stock during paper sheet formation. Partof these water sources (streams 5 and 6) are then sent tothe white water tank along with recycled water (stream 9)from other processes in the De-inking pulper (others).To remove printing ink from the main stock, de-inking

pulper (DIP) is fed with 751.32 ton h21 of fresh water(stream 3) and 398.5 ton h21 of reused water from whitewater tank (stream 8). Of the DIP source (stream 12),54 ton h21 is then mixed with 14.7 ton h21 of freshwater

(stream 11) to dilute the stock being pumped to the deculatorin the approach ow system (AF).Fresh water (stream 10) is also used to dilute de-inking

chemicals in the chemical preparation section (CP) beforethe chemicals are sent to the DIP unit to assist ink removal.In addition, other process in DIP (others) also receives201.84 ton h21 of freshwater (stream4). As for thewastewatercollected from the paper machine and DIP (streams 7 and 13),it is sent to the efuent treatment plant before being dischargedinto the river. The efuent treatment plant operating cost is$0.295 m23 as specied by the plant personnel.In the case study, TSS was the most dominant water qual-

ity parameter and was selected as the main contaminantupon discussion with the plant authority. TSS was monitoredonline and ofine for the purpose of water reuse, recyclingand regeneration. Table 1 summarizes the water demandand source data for this case study. Water demand refersto the water requirement of a water-using operation whilewater source refers to efuent stream leaving a water-usingoperation. Note that water sources can be considered forreuse and recycle to the water demands.

TARGETING THE MINIMUM UTILITY FORGRASSROOTS AND RETROFITTED WATER

NETWORK

One of the latest and widely used water targeting tech-niques known as water surplus diagram (Hallale, 2002) is lim-ited in its ability to estimate the minimum utility targets asit implements a graphical approach that involves time-consuming trial-and-error steps. This limitation has inspiredthe development of a numerical technique known as WCAthat eliminates the trial-and-error approach (Manan et al.,2004). The main objective of the WCA is to establish theminimum utility targets, i.e., the overall fresh water require-ment and wastewater generation for a process after lookingat the possibility of using the available water sources withina process to meet its water demands.

Figure 3. Existing water network for paper mill case study.


308 TAN et al.

Manan et al. (2004) provides a detailed description of howWCA was used to establishe the baseline water targets for achemical process involving mass transfer and non-masstransfer-based processes. Table 2 shows the results obtainedfrom WCA, i.e. water cascade table (WCT) for the paper millgrassroots network before any regeneration placement.Note that the plant consumed 848.12 ton h21 of minimumfresh water and generated 539.36 ton h21 of wastewaterwhen no regeneration is involved. On the other hand, ifone were to retrot the water network using the conventionalretrot technique for non-mass transfer-based process (Tanand Manan, 2003), one will achieve a saving of1140.94 ton h21 fresh water. Further reduction of utility targetscan only be achieved with the introduction of water regener-ation units. This will be described in the following section.

RETROFIT TARGETING OF WATER NETWORK WITHREGENERATION

In this section, a new technique to incorporate regenerationunits into water network retrot is presented. Two types ofregeneration units were considered and the one giving thebest savings was selected. Dissolved air otation (DAF)tank and saveall disc lter (SDF) were two typical physicaltreatments suggested to purify water by recovering brefrom the excess water of paper machines. Note that theseregenerators were the types actually used in the sections ofthe paper mill under study and therefore appropriate for the

purpose of partial upgrading of white water to meet the qualityrequired for the relevant processes. Table 3 shows the econ-omic data for DAF tank and SDF that was extracted from var-ious literature sources (Arundel, 2000; Koppol et al., 2003;Perry and Green, 1997; Peter and Timmerhaus, 1980; Tcho-banoglaus and Burton, 1991; Wiseman and Ogden, 1996).DAF tank is an equipment that removes suspended solids

(TSS) from wastewater and other industrial process streams.DAF tank is commonly used for wastewater pre-treatment,product recovery and thickening of biological solids in foodprocessing, pulp and paper as well as petrochemical indus-tries. This separation process is operated by introducingne gas (usually air) bubbles into wastewater to attach andlift the particles to the water surface to be removed. Hence,wastewater leaves the unit at higher purity. A portion of theDAF tank efuent is recycled, pressurised and semi-saturated with air before re-entering the tank.On the other hand, SDF is widely used as thickening device

in the pulp and paper industry to remove solids from waste-water. It is operated by passing wastewater stream throughlter mediums supported by disks. The solid content of thewastewater will be trapped by the ltermediums and nally dis-posed off. This leaves the wastewater at higher quality. Bothregeneration units will be assessed in the retrot situation.Consider a generic model of a regeneration unit shown in

Figure 4. Water produced or discharged from a water-usingoperation is treated in the regeneration unit to a higherpurity. Water source leaving the regeneration unit is then

Table 2. WCT without regeneration.

Interval, nConcentration,

Cn (ppm)Purity,Pn

SFD,j(ton h21)

SFS,i(ton h21) SFD,j SFS,i (ton h21) FC, (ton h21)

Pure watersurplus (ton h21)

Cumulative purewater surplus(ton h21)

848.120 1 0

0.00002 848.12 0.01696220 0.99998 2190.08 2190.08 0.016962

0.00006 658.04 0.03948280 0.99992 2831.12 2831.12 0.056445

0.00002 2173.08 20.003462100 0.9999 2201.84 155.4 246.44 0.052983

0.00007 2219.52 20.015366170 0.99983 201.84 201.84 0.037617

0.00003 217.68 20.000530200 0.9998 21218.54 21218.54 0.037087

0.00003 21236.22 20.037087230 0.99977 1305.78 1305.78 0

0.00002 69.56 0.001391250 0.99975 469.8 469.8 0.001391

0.99975 539.36 539.22603539.22742

Table 1. Water demands and sources for case study.

Process descriptionWaterdemand

Flowrate, F(ton h21)

Concentration, C(ppm) Water source

Flowrate, F(ton h21)

Concentration, C(ppm)

Pressing showers 1 155.40 20 1 155.40 100Forming showers 2 831.12 80 2 1305.78 230Others 3 201.84 100 3 201.84 170DIP 4 1149.84 200 4 469.80 250CP 7 34.68 20AF 8 68.70 200



allocated for further reuse or recycle in the water network.Hence, regeneration is performed such that

Cin . Cout (3)

Feng and Chu (2004) states that the capital and operatingcosts of a regeneration unit are normally a function of regen-eration owrate (Freg) and the outlet concentration of theregeneration stream (Cout). By studying the effects of thesevariables on the added regeneration unit, several retrotcases can be explored in combination to target the optimumretrot design with the addition of regeneration unit. Threecases considered in this work include:

(1) varying Freg with Cout xed;(2) varying Cout with Freg xed;(3) varying both Freg and Cout

Case (1) involves a situation where a xed Cout is requiredfrom a regeneration unit. Case (2) applies for situationswhere a xed amount of regenerated water is needed in cer-tain processes. Case (3) applies when there are no prefer-able Cout and Freg values during retrot. To yield a specicretrot target, it is necessary to select a reasonable paybackperiod for investment for each of the cases mentioned.For the paper mill case study, the minimum achievable

outlet composition, Cout,min for both types of regenerationunits (DAF tank and SDF) was set at 30 ppm (Wisemanand Ogden, 1996). There is virtually no limitation for thevalue of Freg for each regeneration unit since this value willonly affect the number of newly installed regenerationunit(s) in the network. For economic analysis, the maximumpayback period was set at 2 years. Application of thesecases on the paper mill case study is described next.

Case 1: Varying Freg with Cout Fixed

For Case 1, the minimum outlet concentration for the regen-eration unit,Cout,min was xed at 30 ppmwhile the regenerationowrate, Freg was varied. The objective of this case was tosearch for the optimum regeneration owrate, Freg,optimum forthe newly added regeneration unit(s). When a new regener-ation unit was installed in an existing water network, Fregamount of water at lower quality was regenerated to a higherquality for reuse and recycle, thereby reducing the utility targets

of the network. Consequently, a bigger regeneration unit with ahigher Freg enabled further utility savings. However, it will beshown in the later section that the utility savings for a networkwill remain constant at the maximum regeneration owratevalue. We dene the maximum regeneration owrate to beFreg,max. Hence, the optimum regeneration owrate,Freg,optimumwill fall in the range of 0 Freg,optimum Freg,max.To obtain the value of Freg, max, a plot of fresh water ow-

rate (FFW) target versus Freg shown in Figure 5 was gener-ated for a grassroots design of the paper mill waternetwork. This yields a Freg, max of 620.27 ton h

21 at Cout,minof 30 ppm (see WCT in Table 4). As shown in Figure 5, thefresh water target starts to level off at the Freg,max of620.27 ton h21. Note that at this point of Freg, max, the regen-erated water cannot be further reused since the water net-work had reached its limitation in terms of reusing/recyclingof the regenerated.Ideally, it is desirable to retrot an existing network to

achieve the minimum utility targets identied in a grassrootsdesign. Nevertheless, this is usually not possible during retro-t due to various process constraints and the need to mini-mize changes on the existing process structure. Toconsider these constraint, we have adapted a retrot target-ing procedure parallel to those developed for HEN andMEN problems as a basis for the newly proposed water net-work retrot technique. However, instead of adding moreexchanger areas/stages (such as in the case of HEN andMEN retrot works), capital investments were allocated onnew regeneration units installation to further reduce the utilitytargets, apart from modications of existing network structure.Figure 6 shows a few possible water network retrot proles

for this case. As compared to the retrot proles for HEN andMEN problems shown in Figure 1, the retrot prole for waternetwork is indeed unique. For water network retrot, the

Figure 4. A regeneration unit. Figure 5. FFW versus Freg (Case 1).

Table 3. Economic data for regeneration unit.

Dissolved air otation Saveall disc lter

Cout,min 30 ppm 30 ppmHydraulic loading rate 20 m3 m22 day21 6 m3 m22 day21

Operating cost $0.150 ton21 $0.179 ton21

Costing equation C 2310.6 Area 260 292 C 63 300(Area 9.3)0.48Maximum area per unit 400 m2 140 m2

% Recycle owrate 50% Piping estimation 16% of capital investment 16% of capital investment


310 TAN et al.

proles originate from the utility consumption (fresh water ow-rate) of the existing network at 1989.06 ton h21 at the upperleft portion of the FFW versus Freg plot and moves towardsthe lower right portion of the plot. The best retrot curve is typi-cally the curve that approaches the utility targets of the grass-roots network. Note that fresh water reduction (represented byy-axis) also led to the reduction of wastewater owrate.We next dene a new retrot parameter called the

fresh water efciency,aF. For a xed regeneration owrate(Freg), aF is dened as the ratio between fresh water targetfor a grassroots design (FFW,target) to the fresh water consump-tion of an existing network (FFW,existing) as given in equation (4):

aF FFW, targetFFW, existing

Freg

(4)

The aF value indicates how close the fresh water

consumption in an existing network as compared to that in agrassroots design. A value of unity for aF means that the exist-ing water network has achieved the utility targets of a grass-roots design. This is however almost impossible for mostretrot cases.We have also dened the incremental value of fresh water

efciency, DaF for an increment of regeneration owrate,DFreg by taking the analogy from the DaArea (Silangwa,1986; Shenoy, 1995; Polley and Polley, 2000). DaF wasdened as the ratio between the decrease in fresh watertarget in grassroots design (DFFW,target) to the decrease infresh water consumption of an existing network (DFFW,existing)as given by equation (5):

DaF DFFW, targetDFFW, existing

DFreg

(5)

Table 4. WCT with 620.27 ton h21 of Freg with 30 ppm Cout.

Interval,n

Concentration,Cn (ppm)

Purity,Pn

SFD,j(ton h21)

SFS,i(ton h21)

SFD,j SFS,i(ton h21) FC, (ton h

21)Pure water

surplus (ton h21)

Cumulativepure water surplus

(ton h21)

308.760 1 0

0.00002 308.76 0.0061820 0.99998 2190.08 2190.08 0.00618

0.00001 118.68 0.0011930 0.99997 620.27 620.27 0.00736

0.00005 738.95 0.0369580 0.99992 2831.12 2831.12 0.04431

0.00002 292.17 20.0018100 0.9999 2201.84 155.4 246.44 0.04247

0.00007 2138.61 20.0097170 0.99983 201.84 201.84 0.03276

0.00003 63.23 0.0019200 0.9998 21218.5 21218.5 0.03466

0.00003 21155.3 20.0347230 0.99977 1155.31 1155.31 0

0.99977 0 00

Figure 6. Two kinds of retrot proles (Case 1): (a) curve paths; (b) straight paths.



aF and DaF values for this case study were calculated at0.4264 and 1.0000 respectively using equations (4) and (5).aF and DaF values yielded two possible retrot paths for thewater network as shown in Figure 5. The retrot path withthe constant aF value of 0.4264 to which it approached the uti-lity targets of an ideal grassroots design is the better choice.For the retrot of water networks, two types of retrot proles

may exist The rst is a curved path similar to the conventionalproles for heat and mass integration [Figure 6(a)] and thesecond is a linear retrot path [Figure 6(b)]. The curved pathoccurs for problems with multiple pinches while the linear pro-le is found in water network problems involving a single pinchpoint (Tan and Manan, 2004). Since the paper mill case studyinvolves a single pinch point, we will focus our discussion tothe linear retrot prole for the remainder of the text.Figure 7 focuses on the left portion of Figure 5, i.e., portion

where the FFW versus Freg plot levels off at Freg,max 620.27ton h21. For the ease of demonstration, only the retrot proleof aF 0.4264 was shown. The plot consists of three mainregions. We termed the area below the optimum grassrootsdesign as the infeasible region since it was impossible to achieveutility reduction lower than that for the optimum grassrootsdesign. The region in between the optimum grassroots designand the retrot prole was termed as the economical designregion where the desired retrot targets will possibly fall. Finally,it was uneconomical to achieve retrot targets in the regionabove the retrot prole since no savings could be achieved.We next determined the economic performance of this ret-

rot case using a savings versus investment plot. Thisincluded the fresh water and wastewater savings as well asthe capital investment for retrot work. Utility savingsachieved during retrot was dened as the total waterutility cost savings minus the increment in operating costfor Freg. On the other hand, the capital investment for networkretrot covered the costs of newly installed regenerationunits as well as piping modications. The size of a regener-ation unit was estimated based on the following equation(Tchobanoglaus and Burton, 1991):

Size of regeneration unit Total regeneration flowrateHydraulic loading rate

(6)

Due to the difference in capital and operating costs, differ-ent savings versus investment plots were needed to assessthe two proposed regeneration units, i.e., DAF tank and

SDF. Figures 8 and 9 illustrate the savings versus investmentplot for the DAF tank and SDF respectively. Figure 8 showsthree different segments of savings versus investment plotfor the DAF tank. Each of these segments represented thedesired number of units based on the size calculated at differ-ent Freg value. Payback period lines were also identied inthe diagram to enable designer to choose the optimum retrottargets. A targeted investment limit of 1.68 years (below 2years with maximum utility savings) was identied for a capi-tal investment $3.91M and savings of $2.30M. Three DAFtanks were required to achieve this target.The savings versus investment plot for SDF is shown in

Figure 9. A targeted investment limit of 1.92 years (below 2years with maximum utility savings) was identied for a capi-tal investment and savings of $4.83 M and $2.52 M respect-ively. However, the installation of 18 units of SDF wasrequired to achieve this target.It is also worth to point out that Figures 8 and 9 are unlike

the retrot situation in heat and mass integration where thetotal area or number of stages was the only parameter todecide the retrot option. There were more factors to con-sider in this work. For instance, installing 18 units of SDFmay not be a practical retrot option for the case study,since a large area was required for all of these regenerationunits. Hence, it was up to the designer to decide which regen-eration units were to be chosen during network retrot. In anycase, Freg,optimum for both retrot options was 620.27 ton h

21.

Figure 8. Savings versus investment plot for DAF (Case 1).

Figure 9. Savings versus investment plot for SDF (Case 1).Figure 7. FFW versus Fregplot with constant a (Case 1).


312 TAN et al.

Case 2: Varying Cout at a Fixed FregFor the second case of network retrot, Freg was xed at

Freg,max while the outlet concentration of the regenerationunit, Cout was varied. The objective of this case was to deter-mine the optimum outlet concentration of the regenerationunit, Cout,optimum, which fell between Cout,min and Cout,max.Table 3 indicates that the Cout,min for both regenerationunits in this case study was xed at 30 ppm. On the otherhand, Cout,max was based on the highest concentrationamong the available sources in the water network. This is areasonable basis since no water shall be regenerated dirtierthan the available water sources in an existing network. FromTable 1, the concentration of source 4 at 250 ppm was ident-ied as Cout,max.The next step in this case involved the determination of

Freg,max for the water network, as done for Case 1. WhileCout was varied, Freg,max was xed at 620.27 ton h

21.Recall that this value was the optimal regeneration owrate(Freg,optimum) found at a xed Cout,min for Case 1.Next, WCA technique was used to locate the various targets

for the grassroots water network when a regeneration unit withxed Freg,max and various Cout was added. Figure 10 showstwo different kinds of FFW,min versus Cout plot, i.e., a curvedpath similiar to the conventional heat and mass integration pro-le [Figure 10(a)] or a straight retrot path [Figure 10(b)].These plots revealed that for an optimum grassroots design,a regeneration unit with lower Cout would consume less freshwater as a result of cleaner regenerated water that could bereused or recycled A similar trend could be expected for anexisting water network. However, during retrot, some capitalinvestments would be needed for newly installed regenerationunit as well as for existing network modications. We hencedened another retrot parameter called fresh water ef-ciency, aC as given by equation (7):

aC FFW, targetFFW, existing

Cout

(7)

where FFW,target and FFW,existing are the fresh water targets forgrassroots network and the existing network respectively fora given regeneration Cout value. The aC value provided a com-parison of the minimum fresh water targets in a grassrootsdesign with the existing fresh water consumption. An aC ofunity means that the existing water network had achievedthe grassroots utility targets. This is impractical for most retrotcases.An incremental value of fresh water efciency, DaC on the

other hand was dened as the ratio between a decrease offresh water target in grassroots design, DFFW,target to that ofthe decrease in fresh water consumption in an existing net-work, DFFW,existing for a given decrease in outlet concentrationof regeneration unit, DCout [equation (8)]:

DaC DFFW, targetFFW, existing

DCout

(8)

Figure11 represents the two retrot proles forCase2,whichwas plotted using the fresh water efciency values calculatedusing equations (7) and (8). These proles corresponds tothe values of aC 0.4264 and DaC 1. Note from Figure 11that these retrot proles originated from the utility target ofthe existing network at the upper right portion of the FFW,minversus Cout plot. Fresh water usage was reduced with adecrease in outlet concentration of the regeneration unit. Afeasible retrot path is the one leading to the optimum grass-roots design, i.e., towards the lower left portion of the graph.As shown in Figure 11, a retrot path with a constant a valuewould be a better choice. This represents the situation inCase 1. Figure 11 also shows that this retrot diagram consistsof three main regions, namely an infeasible design region, aneconomical design region and an uneconomic design region.Next, the savings achieved and the capital investment (due

to new regeneration unit placement) at each point inFigure 11 was calculated based on the economics data inTable 1. The savings versus investment plot for DAF tankand SDF are shown in Figures 12 and 13 respectively.Figure 12 shows that an investment of $3.91 M is needed

Figure 10. Two kinds of retrot proles (Case 2): (a) curve paths; (b) straight paths.



for the installation of three new DAF tanks to achieve anannual savings of $2.30 M. This corresponds to the paybackperiod of 1.7 years. On the other hand, capital investment of$4.83 M is needed for the installation of 18 new SDF units.The targeted annual savings achieved for this alternativewas at $2.52 M. This led to a payback period of 1.92 years.Note also that due to the xed value of Freg in this case,both savings versus investment diagrams in Figures 12 and13 appeared to be a vertical straight line at a xed capitalinvestment. The capital investment was independent of Coutbecause the selected regeneration unit mainly dependedon the operating conditions. Finally, Cout,optimum correspond-ing to both regeneration units were determined at 30 ppm.

Case 3: Varying Cout and FregFor Case 3, both Freg and Cout were varied. The objective

of this case was to search for Freg,optimum and Cout,optimum forthe added regeneration units. The rst step in this case wasto identify the boundary for Cout,optimum. Since the same casestudy and regeneration units were used, Cout,min of 30 ppmand Cout,max of 250 ppm were chosen. Next, the value ofFreg,max was determined at various Cout.Figure 14 shows the FFW versus Cout plot for a grassroots

network, calculated using the WCA technique. Note that thefresh water consumption remained constant at308.76 ton h21 when the regeneration unit produced regener-ated water at concentration lower than the Cout of 95 ppm.However, beyond the Cout of 95 ppm, a larger amount offresh water was required in the network due to the lower

quality of regenerated water. Increased fresh water led tohigher wastewater generated from the network.Our objective was to target the utility reduction when

regeneration units were installed in an existing water net-work. It was therefore necessary to focus on the constantfresh water region of the grassroots FFW versus Cout plot(i.e., Cout 95 ppm) where the minimum utility targets wereachieved. As a result, the boundary of Cout,max was shiftedfrom 250 ppm to 95 ppm.Figure 15 focuses the FFW versus Cout plot with the newly

dened Cout boundary at 95 ppm (from Figure 14). This plotconsists of the optimum grassroots design and a newlyadded retrot prole for the existing network, with a calcu-lated at 0.4264 [following equation (8)]. Note also that awas selected instead of Da in this case. This was due tothe optimum grassroots plot being a horizontal straight line,and hence no prole of Da can be plotted. Similar to the pre-vious cases, regions of infeasible design, economical designand an uneconomic design exist in this case.The operating cost savings and the required capital invest-

ment were calculated next to assess the economics of thiscase. Figures 16 and 17 illustrate the savings versus invest-ment plot for DAF tank and SDF respectively. Note fromFigures 16 and 17 that the savings decreased as investmentincreased when the regeneration owrate (Freg) and regener-ation outlet concentration (Cout) were simultaneously varied.As Cout increased, the maximum regeneration owrate(Freg,max) also increased. This resulted in a lower fresh waterretrot prole. Increased Freg,max also led to higher capital

Figure 11. FFW versus Cout plot with constant a (Case 2).Figure 13. Savings versus investment plot for SDF (Case 2).

Figure 12. Savings versus investment plot for DAF (Case 2). Figure 14. FFW, min versus Cout (Case 3).


314 TAN et al.

investment. The savings also decreased due to increased(Freg) and reduced fresh water as well as wastewater savings.A targeted investment limit of 1.7 years (below 2 years withmaximumutility savings) was identied for DAF tank for a capi-tal investment of $3.91 M and savings of $2.30 M. Three DAFtanks were required to achieve this target. On the other hand,18 units of SDFwere needed to achieve the same targets. Thiscorresponded to a capital investment of $4.83 M, an annualsavings of $2.52 M and a payback period of 1.92 years. TheCout,optimum and Freg optimum for both regeneration units were30 ppm and 620.27 ton h21.

DISCUSSION

It was shown that three cases achieved the same retrot tar-gets (Cout,optimum and Freg,optimum) for the paper mill case study.This was mainly due to the selection of xedCout and Freg vari-ables in the rst two cases. If a different Cout and/or Freg anddifferent payback period were specied for Case 1 and Case2, different retrot targets would have emerged. Therefore, itcould be concluded that retrot targets depended on Coutand Freg as well as on the payback period specications.Although Cout,optimum and/or Freg,optimum attained for both

regeneration units were 30 ppm and 620.27 ton h21 respect-ively, selection of DAF tanks as the new regeneration unitswas the better retrot option for all three cases. This wasdue to the more reasonable targeted number of DAF tanksas compared to SDF. Nevertheless, one can still considerinstalling SDF in this case study if the maximum water sav-ings constraint was neglected. Installing ve units of SDF,for instance, gave a total Freg of 139 ton h

21, Cout of30 ppm and a payback period of 1.91 years.All the retrot targets achieved in this section were merely

based on the limiting data for the case study and was inde-pendent of any particular network design. In order for thesetargets to be meaningful and effective, a network design tech-nique leading to the retrot targets is needed. The procedurefor network design is described in the next section.

RETROFIT DESIGN OF WATER NETWORK WITHREGENERATION

In the grassroots design of a maximum water recovery net-work, the pinch point plays an important role to ensure theestablished utility targets are realized. Water network is nor-mally divided into regions above and below the pinchduring the design stage. Network design is then carried out

independently in these regions using various networkdesign procedures (e.g., Wang and Smith, 1994; Feng andSeider, 2001; Prakash and Shenoy, 2005). Similarly, gettingthe pinch location(s) for an existing water network is alsoessential before any retrot design is carried out.To determine the pinch points, we rely again on the WCA

technique. Utility targeting was performed for a grassrootsnetwork based on the water demands and sources data(Table 1) along with the Freg,optimum and Cout,optimum of theregeneration unit(s) identied earlier. Table 4 shows theresulting WCT for the grassroots network. The water pinchfor this grassroots network existed at the lowest concen-tration level (230 ppm) where there was zero cumulativepure water surplus (Table 4). Water pinch at this location indi-cated that all sources and demands appear at the regionabove the pinch, and hence, a zero discharge process (nowastewater generation) was achieved. However, in revamp-ing an existing water network, achieving the targets as inthe grassroots design is not always possible.For retrotting water network involving non-mass transfer

processes, such as for the paper mill case study, many net-work design techniques may be used. These include thesource sink mapping diagram (El-Halwagi, 1997; Dunn andWenzel, 2001), sink-source allocation (Prakash andShenoy, 2004) or concentration block diagram (Tan andManan, 2003). In this work, we utilized the concentrationblock diagram (CBD) to represent the existing water network.CBD provides a clear representation of the existing water net-work in terms of the water owrate as well as contaminantconcentration.

Figure 16. Savings versus investment plot for DAF (Case 3).

Figure 15. FFW versus Cout plot with new Cout boundary (Case 3).

Figure 17. Savings versus investment plot for SDF (Case 3).



CBD of the paper making case study is presented inFigure 18. The vertical dashed lines represent the concen-tration-interval boundaries which correspond to a distinct lim-iting inlet or outlet concentration. The water-using operationsare represented by rectangles corresponding to their inletand outlet concentrations. The arrows in the diagram indicatethe water streams of the existing water network with thestream owrate in ton h21 and contaminant concentration inppm (gures in parentheses).The next step in the retrot design stage was to identify the

streams for regeneration. Sources at the highest concen-tration were always preferred as this reduced the massload accumulated in wastewater produced. One noticedfrom Figure 1 that, at the highest concentration of 250 ppm,469.8 ton h21 of wastewater is available from the source 4,i.e., the water source of DIP. However, the targeted Freg,maxthat was identied in the earlier stage amounts to620.27 ton h21. This means that part of the wastewater dis-charged from Forming Showers (at second highest concen-tration) was chosen to satisfy the remaining Freg amountedfed to the DAF tanks. The dotted lines in Figure 18 showthe streams which are identied for regeneration.54 ton h21 of water produced from DIP was originally sentto DAF tanks instead of being reused in AF.The existing water network was next redesigned to meet

the established retrot targets. This involved sending theidentied streams to the regeneration unit (DAF tanks) andfeeding the puried source to the water-using operations toreduce fresh water intake. The preliminary retrot design ispresented in Figure 19. As shown, 54 ton h21 of watersource was sent to AF unit from the DAF tanks instead offeeding from the DIP unit directly in the existing water net-work (refer to Figure 3). The remaining regenerated waterfrom the DAF tanks were also sent to the Pressing Showers,Forming Showers and for Chemical Preparation.The nal step of the retrot design stage involves optimis-

ation of the preliminary retrotted network for further utilityreduction. To achieve this, opportunity for further reuseand/or recycle of wastewater sources in the preliminary ret-rot design was explored. One option in this effort is toreuse and recycle the efuent from Forming Showers backto its own water demands as well as to other water-using pro-cesses. The nal retrot design for the paper making casestudy is shown in Figure 20. It is shown that after retrot,

the fresh water consumption has been reduced by 80% to401.82 ton h21 while the wastewater generation has beenreduced by 95% to 92.52 ton h21.Economic calculations show that a total savings of $3.16 M

has been achieved with this nal network, with the installationof three new DAF tanks. Though the savings had slightlysurpassed the targeted value ($0.86M), however thecapital investment of $3.91M remained within target. Theresulting payback period of 1.24 years is slightly better ascompared to the targeted value of 1.7 years. Finally, notethat the retrot design presented above is one of the manypossible solutions that can achieve the retrot target.Often, different network design congurations can beachieved with the use of different network design techniques(El-Halwagi, 1997; Dunn and Wenzel, 2001; Prakash andShenoy, 2005).

CONCLUSIONS

A new systematic procedure for the retrot of water net-work with regeneration has been developed. This procedureenables further reduction of the utility targets to be achievedin an existing water network via water reuse, recycling and

Figure 18. Existing water network in CBD with identied streams forregeneration.

Figure 19. Preliminary retrot design.

Figure 20. Final retrot design.


316 TAN et al.

regeneration. The optimum design of the regeneration unitwas based on two process parameters, i.e., regenerationowrate and the outlet concentration. The methodology con-sists of two stages namely, retrot targeting and design. Inthe targeting stage, fresh water versus regeration owrateand outlet concentration diagrams were introduced, and afeasible retrot path was identied to establish various retrottargets. Given a xed payback period or capital expenditure,the retrot targets were determined from the savings versusinvestment diagram. During network design, the existingwater network was revamped according to pinch designrules to meet the established retrot targets. This method-ology has successfully achieved the retrot targets prior todesign, and further minimized fresh water consumption andwastewater generation in an existing water network.

REFERENCESAhmad, S. and Polley, G.T., 1990, Debottlenecking of heat exchangenetworks, Heat Recovery Systems & CHP, 10(4): 369385.

Alva-Argaez, A., Kokossis, A.C. and Smith, R., 1998, Wastewaterminimisation of industrial systems using an integrated approach,Computers and Chemical Engineering, 22: S741744.

Arundel, J., 2000, Sewage and Industrial Efuent Treatment, 2ndedition (Blackwell Science, UK).

Bagajewicz, M. and Savelski, M., 2001, On the use of linear modelsfor the design of water utilization systems in process plants with asingle contaminant, Transactions of the Institute of ChemicalEngineers, Part A, 79: 600610.

Benko, N., Rev, E., and Fonyo, Z., 2000, The use of nonlinear pro-gramming to optimal water allocation, Chemical Engineering Com-munication, 178: 67101.

Castro, P., Matos, H., Fernandes, M.C. and Nunes, C.P., 1999,Improvements for mass-exchange networks design, Chem EngSci, 54: 16491665.

Dunn, R.F. and Wenzel, H., 2001, Process integration designmethods for water conservation and wastewater reduction in indus-try. Part 1: Design for single contaminant, Cleaner Production Pro-cesses, 3: 307318.

El-Halwagi, M.M., 1997, Pollution Prevention Through Process Inte-gration: Systematic Design Tools (Academic Press, San Diego,USA).

EL-Halwagi, M.M. and Manousiouthakis, V., 1989, Synthesis of massexchange networks, AIChE J, 35(8): 12331244.

Feng, X. and Chu, K.H., 2004, Cost optimisation of industrial waste-water reuse systems, Trans IChemE, Part B, Process Safety andEnvironment Protection, 82(B3): 249255.

Feng, X. and Seider, W.D., 2001, New structure and design methodfor water networks, Ind Eng Chem Res, 40: 61406146.

Fraser, D.N. and Hallale, N., 2000, Retrot of mass exchange net-work using pinch technology, AIChE J, 46(10): 21122117.

Hallale, N., 2002, A new graphical targeting method for water mini-misation, Advances in Environmental Research, 6(3): 377390.

Huang, C.-H., Chang, C.-T., Ling, H.-C., and Chang, C.-C., 1999, Amathematical programming model for water usage and treatmentnetwork design, Ind Eng Chem Res, 38: 26662679.

Jacob, J., Kaipe, H., Couderc, F. and Paris, J., 2002, Water networkanalysis in pulp and paper processes by pinch and linear program-ming techniques, Chemical Engineering Communication, 189(2):184206.

Jodicke, G., Fisher, U. and Hungerbuhler, K., 2001, Wastewaterreuse: A new approach to screen for designs with minimal totalcosts, Computer and Chemical Engineering, 25: 203215.

Koppol, A.P.R., Bagajewicz, M.J., Dericks, B.J. and Savelski, M.J.,2003, On zero water discharge solutions in the process industry,Advances in Environmental Research, 8: 151171.

Kuo, W.C.J. and Smith, R., 1998, Design of water-using systemsinvolving regeneration, Transactions of the Institute of ChemicalEngineers, Part B, 76: 94114.

Manan, Z.A., Foo, D.C.Y. and Tan, Y.L., 2004, Targeting the minimumwater owrate using water cascade analysis technique, AIChE J,50(12): 31693183.

Parthasarathy, G. and Krishnagopalan, G., 2001, Systematic reallo-cation of aqueous resources using mass integration in a typicalpulp mill, Advances in Environmental Research, 5: 6179.

Perry, R.H. and Green, D.W., 1997, Perrys Chemical EngineersHandbook, 7th edition (McGraw Hill, New York, USA).

Peter, M.S. and Timmerhaus, K.D., 1980, Plant Design and Econ-omics for Chemical Engineers, 3rd edition (McGraw Hill,New York, USA).

Polley, G.T. and Polley, H.L., 2000, Design better water networks,Chemical Engineering Progress, 96: 4752.

Prakash, R. and Shenoy, U.V., 2005, Targeting and design of waternetwork for xed owrate and xed contaminant load operations,Chem Eng Sci, 60(1): 255268.

Shenoy, U.V., 1995, Heat Exchanger Network Synthesis: ProcessOptimization by Energy and Resource Analysis (Publishing Co,Houston Gulf, USA).

Silangwa, M., 1986, Evaluation of Various Surface Area EfciencyCriteria in Heat Exchanger Network Retrots, MSc dissertation,University of Manchester Institute of Science and Technology.

Takama, N., Kuriyama, T., Shiroko, K. and Umeda, T., 1980, Optimalwater allocation in a petroleum renery, Computers and ChemicalEngineering, 4: 251258.

Tan, Y.L. and Manan, Z.A., 2003, Retrot of water network in apaper making process based on pinch analysis technique,17th Symposium of Malaysia Chemical Engineers (SOMChE2003), 2930 December, Penang: Universiti Sains Malaysia,6166.

Tan, Y.L. and Manan, Z.A., 2004, Retrot of water network for masstransfer based processes, 18th Symposium of Malaysia ChemicalEngineers (SOMChE 2004), 1314 December, Perak: UniversitiTeknologi PETRONAS, 6: 3339.

Tchobanoglaus, G. and Burton, F.L., 1991, Wastewater Engineering:Treatment, Disposal and Reuse, 3rd edition (McGraw Hill,New York, USA).

Thevendiraraj, S., Klemes, J., Paz, D., Aso, G. and Cardenas, G.J.,2003, Water and wastewater minimisation study of a citrus plant,Resources, Conservation and Recycling, 37: 227250.

Tjoe, T.N. and Linnhoff, B., 1986, Using pinch technology for processretrot, Chemical Engineering, April 28: 4760.

Wang, Y.P. and Smith, R., 1994, Wastewater minimisation, ChemEng Sci, 49(7): 9811006.

Wiseman, N. and Ogden, G., 1996, Zero liquid efuent technologiesfor the paper industry, Paper Technology, 37: 3138.

Xu, D., Hu, Y., Hua, B. and Wang, X., 2003, Manimization of theowrate of fresh water and corresponding regenerated water inwater-using system with regeneration reuse, Chinese Journal ofChemical Engineering, 11(3): 257263.

ACKNOWLEDGEMENTThe nancial support of Ministry of Science, Technology and

Environment (MOSTE), Malaysia through Intensied Research Pri-ority Area (IRPA) research grant and National Science Fellowship(NSF) scholarship is gratefully acknowledged.

The manuscript was received 5 May 2005 and accepted forpublication after revision 17 October 2006.



water retrofit

Documents