Integrated Decision Support Systemfor Waste Minimization Analysis inChemical ProcessesI S K A N D A R H A L I M † A N DR A J A G O P A L A N S R I N I V A S A N *

Laboratory for Intelligent Applications in ChemicalEngineering, Department of Chemical and EnvironmentalEngineering, National University of Singapore,10 Kent Ridge Crescent, Singapore 119260

The need to build and operate environmentally friendlyplants has challenged the chemical industry to considerwaste minimization or even elimination starting from theearly stages of process development. A thorough wasteminimization analysis requires specialized expertise and islaborious, time-consuming, expensive, and knowledge-intensive. This has caused a major technical barrier forimplementing waste minimization programs within the industry.Previously, we had reported a systematic methodologyand a knowledge-based system, called ENVOPExpert, foridentifying waste minimization opportunities in chemicalprocesses. In this paper, we propose an integratedqualitative-quantitative methodology to identify wasteminimization alternatives and assess their efficacy in termsof environmental impact and process economics. Aqualitative analysis is first conducted to identify the sourcesof wastes and to propose alternatives for eliminating orminimizing them. Environmental impact of each alternativeis then calculated by doing a quantitative pollutantbalance. The capital expenditure required for implementingthe alternative and the resulting plant operating costsare also calculated and used in the evaluation of the wasteminimization alternatives. Through this, practical and cost-effective options can be identified. This methodologyhas been implemented as an integrated decision supportsystem and tested using the hydrodealkylation process casestudy with satisfactory results.

IntroductionThe issue of clean production has challenged the chemicalindustries to initiate new approaches to tackle pollutionproblems. The traditional end-of-pipe treatment approachis no longer viewed as an adequate, stand-alone, pollutionproblem solver. Increasing public awareness of the impactof industrial pollution, more stringent discharge standards,and escalating waste treatment and disposal costs have placedenormous pressure on the chemical industries to shift theirparadigm of pollution prevention from the end-of-pipetreatment to waste minimization or even total eliminationat the point of generation.

Numerous techniques and methodologies for pollutionprevention have been published in the literature. In thebroadest sense, all of these available techniques can beclassified into qualitative and quantitative approaches. In

the qualitative approach, methods such as EnvironmentalOptimization (ENVOP) and Douglas’ hierarchical procedureare used to identify possible waste minimization alternatives.ENVOP technique is a waste minimization procedure thatfollows the approach of Hazard and Operability (HAZOP)analysis in process safety (1). During an ENVOP study, eachprocess line and unit operation is analyzed to identifypotential waste minimization alternatives that meet thedesired environmental objectives. These alternatives arederived by combining a set of qualitative guidewords (suchas more, less, etc.) with process variables (such as pressure,temperature, flow rate, etc). In Douglas’ procedure (2), thehierarchiacal decision structure for process design, developedby Douglas (3), is extended to incorporate potential strategiesto reduce waste generation right from the early stages ofdesign. The basic waste minimization solutions that can bederived through this procedure can be summed up as“changing the chemistry”, “changing the process”, “changingthe equipment”, “changing the solvent”, and “reuse andrecycle of the material”.

The quantitative approach to waste minimization involveschanging the process operating conditions or synthesizinga process structure through numerical optimization or byusing a process simulator. Hopper et al. (4) demonstratedthe opportunity for minimizing wastes through gradualchanges in the reactor and separator variables, using a processsimulator. Cabezas et al. (5) used a methodology called theWAste Reduction (WAR) algorithm for quantifying thepotential environmental impact of chemical processes. Intheir approach, the WAR algorithm is used in conjunctionwith a process simulator to evaluate the environmentalimpact of process modifications. Young and Cabezas (6)extended the WAR methodology to incorporate the envi-ronmental impact of energy generated and consumed withinthe process. WAR-based environmental analysis can becombined with economic evaluation of a process design asdemonstrated by Fu et al. (7). In their approach, a couplingbetween the economic and the environmental objectives ofthe process was presented as a multi-objective optimizationproblem and then solved using stochastic modeling andPareto set. Dantus and High (8) formulated a multi-objectiveoptimization problem that simultaneously targets maximiz-ing profit while minimizing environmental impact and solvedthis by combining compromise programming with simulatedannealing.

One common shortcoming of the above-mentionedquantitative approaches is due to the complexities involvedin modeling industrial-scale process with a large number ofinterconnections between the streams and the units. Theresulting optimization problem is usually quite large anddifficult to solve. Another shortcoming arises from the factthat these techniques require considerable skill and expertisein a number of areas. Recognizing each unit and variablewithin the process that contributes to waste generation inthe process is challenging and requires deep insight into theprocess and its operations. The numerical techniques usedto identify and evaluate the process modifications also requireconsiderable know-how of the specific techniques and theirnuances that is seldom available in chemical plants. In thispaper, we address this important problem of industrialsignificance by developing a methodology for identifyingprocess modifications that reduce waste generation and areeconomically prudent.

We have previously developed an intelligent system, calledENVOPExpert, for qualitative waste minimization analysis(9-13). ENVOPExpert is broadly based on the ENVOP

* Corresponding author e-mail: chergs@nus.edu.sg; telephone:+65 8748041; fax: +65 7791936.

† Present address: Environmental Technology Institute, Innova-tion Centre, Block 2, Unit 237, 18 Nanyang Dr., Singapore 637723.

Environ. Sci. Technol. 2002, 36, 1640-1648

1640 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 7, 2002 10.1021/es0155175 CCC: $22.00 2002 American Chemical SocietyPublished on Web 02/19/2002

technique and uses expert knowledge to automaticallyidentify the source of wastes in a chemical process andpropose process changes to eliminate or minimize them.ENVOPExpert has been tested on a number of industrial-scale processes including a hydrocarbons separation processand a chemical intermediate manufacturing process and wasfound to generate results comparable to the analysis byhuman experts (12, 13). In this paper, we propose a frameworkfor synthesizing waste minimization alternatives through theintegration of ENVOPExpert with quantitative environmentalimpact and process economics analysis. The outline of therest of this paper is as follows: Next, we provide overviewsof ENVOPExpert and the environmental impact calculationusing the WAR algorithm. In the next section, we proposethe integrated qualitative-quantitative methodology foridentifying waste minimization alternatives and evaluatingthem based on WAR and process economics. The integratedframework is illustrated in the last section through a casestudy involving the hydrodealkylation (HDA) process formanufacture of benzene from toluene.

Waste Minimization MethodologyIn 1988, the U.S. EPA published a document that establisheda systematic procedure for performing waste minimizationassessment in process plants (14). The procedure involvesthe following steps (see Figure 1):

(i) Planning and organization: to establish waste mini-mization goals, objectives, and tasks.

(ii) Assessment phase: where the evaluation team isorganized, plant and waste data collected, plant operationsreviewed, and options for minimizing wastes generated.

(iii) Feasibility analysis phase: when options are screenedon the basis of technical and economic feasibility.

(iv) Implementation phase: where the most promisingoptions are implemented and their performance evaluated.

A thorough waste minimization assessment is thuslaborious, time-consuming, expensive, knowledge-intensive,and requires specialized expertise of the team. This has causeda major technical barrier for implementing waste minimiza-tion within the industry. An intelligent system that canautomate waste minimization analysis would certainly bebeneficial since it can perform a systematic and thoroughevaluation of waste minimization options and reduce theteam’s time and effort. Such a system must be capable offirst identifying waste sources that arise in the process,assisting the nonexpert in terms of possible suggestions thateliminate or minimize the waste sources and then highlight-ing suggestions that are both environmentally friendly (interms of impact on environment) and cost-effective.

We have previously developed a waste minimizationmethodology that is amenable to automation (12). Thismethodology employs a two-step procedure: (i) wastedetection and diagnosis and (ii) waste minimization optionsgeneration. A detailed description of this methodology ispresented in the Supporting Information. This methodologyhas been implemented as a knowledge-based expert system,called ENVOPExpert, using Gensym’s expert system shell G2.ENVOPExpert is capable of providing technical and decisionsupport for identifying waste minimization alternatives (seeFigure 1). The qualitative analysis implemented in EN-VOPExpert offers a number of unique advantages. Since wasteminimization domain expertise is incorporated into EN-

FIGURE 1. Waste minimization assessment and the role of ENVOPExpert.

VOL. 36, NO. 7, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 1641

VOPExpert, it obviates the need for specialized expertise andmakes it easy to use. The knowledge library, which is currentlybuilt using literature solutions and design heuristics, can becontinuously updated with new knowledge. The systematicmethodology embedded in ENVOPExpert also makes itapplicable to different stages of the process life cycle fromconceptual design to retrofitting.

However, the inherent nature of the ENVOPExpert’squalitative analysis also leads to a number of shortcomings.It is possible for the qualitative analysis embedded inENVOPExpert to generate inconsistent results if counteractingprocess phenomena are present. For example, increasingtemperature inside a reactor to reduce the generation of aparticular waste byproduct may increase the formation ofanother waste byproduct in a downstream unit or decreasethe production of useful product. ENVOPExpert also doesnot account for the impact of process modifications on theprocess economics. These shortcomings can be addressedby incorporating additional quantitative knowledge aboutthe process into the analysis.

WAR Algorithm. The WAR algorithm is perhaps the mostpractical environmental impact calculation tool accom-plished to date (15). The WAR algorithm was first developedby Hilaly and Sikdar (16), who introduced the concept ofpollution balance based on the mass balance of pollutants.Cabezas et al. (5) later improved the original WAR algorithmand developed a generalized WAR algorithm based on thepotential environmental impact (PEI) balance of pollutants.From the PEI balance calculations, a relative indication ofthe environmental friendliness of the chemical process canbe obtained. The following are some of the key componentsof the WAR algorithm.

In the WAR algorithm, a potential environmental impact(I) of a chemical k in a nonproduct (NP) stream of j of aprocess is expressed as

where Mj is the mass flow rate of stream j, x kjNP is the mass

fraction of chemical k in the nonproduct stream j, and ψk isdefined as the overall potential environmental impact ofchemical k, which is developed using the following expres-sion:

where Rl is a relative weighting factor for impact categorytype l independent of chemical k, and ψk,l

s is the specificpotential environmental impact of chemical k for an envi-ronmental impact type l, which includes the followingcategories: global warming, ozone depletion, acid rain, smogformation, human toxicity, aquatic toxicity, and terrestrialtoxicity. Impact scores ψk of several chemicals used in theproduction of ammonia, methyl ethyl ketone, acrylic acid,and benzene have been quantified using this expression (5-7).

On the basis of steady-state balances, the environmentalimpact of any processes can be written as follows:

where Iin is the input impact rate of stream entering thesystem, Iout is the output impact rate of stream leaving thesystem, and Igen is the rate of impact generation by the system.For a balance involving only the nonproduct (NP) streams,the following analogous equation can be written:

Using the terminology in eq 1 and the impact balance of eq4, the potential environmental impact Igen generated by thenonproduct stream can thus be described as

To take into account the product stream of the process, anindex I gen

NP is introduced as follows:

where I genNP is a measure of the potential impact created by

all nonproduct streams in producing the products P. Theinterested reader is referred to Cabezas et al. (5) for a detaileddescription of the WAR algorithm.

The WAR algorithm provides a metric for the environ-mental friendliness of a process. It can also be used to evaluateprocess modifications for their environmental impact. Onedrawback of the WAR algorithm arises due to the difficulty,ambiguity, and subjectivity involved in combining thedifferent impacts generated by the process into a single value,ψk. Also, the WAR algorithm does not directly provide anyguidance on the actual origin of the waste in the process orthe modifications that would minimize the waste. WARshould therefore be used with methods for generating processalternatives.

Integrated Qualitative-Quantitative FrameworkIn this paper, an integrated framework that combines thewaste identification and alternative generation capabilitiesof ENVOPExpert with the quantitative alternative assessmentof the WAR algorithm is proposed. The framework comprisesof the following steps, as shown in Figure 2:

(i) Base-case process flowsheet simulation using a processsimulator.

(ii) Environmental impact calculation using WAR algo-rithm and process economic analysis.

(iii) Qualitative waste minimization analysis usingENVOPExpert to generate alternatives.

(iv) Modification to the base process based on theproposed alternatives.

(v) Comparison between the modified and the base-caseprocess in terms of environmental impact and economics.

Initially, the steady-state material and energy balances ofthe process are performed using a process simulator withthe main objective of calculating the flow rates of each wastestream in the process. We have used the HYSYS simulator(17) for this purpose, although other commercial simulatorscould also be used. On the basis of the simulation results,the PEI contributed by each waste stream is calculated usingthe WAR algorithm to obtain the overall environmentalimpact of this base-case process against which all processmodifications can be compared. The costs for this base-caseprocess are also calculated. In the next step, the process isqualitatively analyzed to derive alternatives that eliminateor minimize the waste generated within the process. This isdone using ENVOPExpert, which contains various heuristicrules, procedures, and P-graph, signed digraph, and func-tional models as described below (see Figure 3).

First, the sources of each material component that makeup the waste stream are identified. The P-graph representa-tion of a process provides a convenient framework fordiagnosing the origins of waste in the process and for derivingtop-level waste minimization alternatives. Starting from eachwaste stream and tracing upstream using the P-graph model,sources of waste, such as impurities in inlet stream, useful

INP ) Mjx kjNPψk (1)

ψk ) ∑l

Rlψk,ls (2)

Igen ) Iout - Iin (3)

I genNP ) I out

NP - I inNP (4)

I genNP ) ∑

j

M jout∑

k

x kjNPψk - ∑

j

M jin∑

k

x kjNPψk (5)

I genNP )

I genNP

∑p

Pp

(6)

1642 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 7, 2002

material transformed at low conversion rate, waste byproductproduced from reaction or phase change phenomena andineffective separation of useful material, are identified. Oncethe waste sources are detected, ways to eliminate them areproposed. For each waste origin, top-level waste minimiza-tion alternatives (such as remove impurities, optimize thereactor unit, improve the separation in the separator unit,

and recycle waste stream) that identify the broad modificationrequired in the process unit or feed material to minimizewaste are obtained through the use of domain knowledgeembedded in the form of heuristic rules. The broad sug-gestions from the P-graph analysis are distilled further usingdigraph and functional knowledge to derive detailed alter-natives. A digraph model represents the cause-and-effect

FIGURE 2. Integrated qualitative-quantitative waste minimization framework.

VOL. 36, NO. 7, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 1643

interactions among the different variables in each processunit and their influence on the underlying physiochemicalphenomena. In ENVOPExpert, the top-level alternativesgenerated by the P-graph heuristics are directly translatedto conclude a value for the phenomena nodes in the processunit digraphs. Detailed alternatives are identified by propa-gating this value from the phenomena node to the othernodes in the digraph. To link the digraph models of differentprocess units, a functional model of the process is used.

Detailed description of the methodology including theP-graph, digraphs, and functional models is provided in theSupporting Information. The final results from ENVOPExpertare a series of waste minimization alternatives that coverbasic suggestions such as materials substitution, streamrecycling, change of process chemistry, and optimizingcertain unit’s variables of the process. These alternatives haveto be assessed in detail using quantitative techniques toimprove the base-case process.

FIGURE 3. ENVOPExpert ’s algorithm.

1644 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 7, 2002

The following step involves implementing the proposedalternatives to the base-case process to evaluate their efficacy.In our approach, process modifications are implementedfollowing Douglas’ hierarchical decision procedure (3) andcover the input-output structure of the process, reactorsystem, and separator system. [Decisions related to heatexchanger network have not been considered explicitly inthe current work and is the subject of our future work.] Ateach hierarchical level, the environmental and the economicimpacts of the modification are calculated to measure thefeasibility of the proposed alternatives. The WAR algorithmis adopted for comparing the degree of environmentalfriendliness of each alternative applied to the base process.Process economics are measured using the overall profit afterimplementing the alternatives. This profit factor incorporatesthe required investment (capital and operating costs) andthe resulting production rates.

It might be necessary at this stage to search for theoptimum process condition that considers the tradeoffbetween production and waste generation, that is, profit andenvironmental impact. For this purpose, an optimizationtool, such as the HYSYS optimizer embedded inside HYSYSsimulator, can be utilized to search for the conditions thatfavor the waste minimization objective of the process. In thecontext of waste minimization, this objective may betranslated as minimize the concentration of waste byproductgenerated inside a reactor, optimize the separation inside aseparator, etc. These optimizations would be achieved bymanipulating the operating conditions in the unit subject toequipment and process constraints. For example, in a reactor,changes in reactant concentration, pressure, and temperature

may be necessary to minimize the concentration of a wastebyproduct or maximize the reaction selectivity. Similarly fora distillation unit, the tradeoff between increased reflux ratio,increased energy requirement, and improved separationefficiency could be investigated.

Case Study: Hydrodealkylation of Toluene to BenzeneTo illustrate the integrated framework, we describe itsapplication to a case study involving HDA of toluene tobenzene. This case study had been previously used by Fu etal. (7) to illustrate their multi-objective optimization frame-work for waste minimization. Figure 4 shows the flowsheetof this process as adapted from Douglas (3). Fresh tolueneand hydrogen are initially mixed with a recycle streamcontaining hydrogen, methane, benzene, and toluene. Thefeed mixture is heated in a furnace to about 641 °C beforebeing passed to an adiabatic reactor for the HDA reaction.In the reactor, toluene and hydrogen react to form thebenzene product. This main reaction is accomplished by thegeneration of methane byproduct and diphenyl waste. Thereactor effluent containing unreacted hydrogen, toluene, andreaction products is condensed using cooling water in acooler. This is followed by separation in a flash separator toremove the aromatics from non-condensable hydrogen andmethane. The vapor stream leaving the top of the separatorcontains a significant amount of methane and hydrogen. Afraction of this vapor is purged as saleable byproduct whilethe rest of the stream is recycled and mixed with the rawmaterials. The liquid from the flash separator is split intotwo streams. The first stream (with about 26% of the liquid)is mixed with the reactor effluent stream and recycled back

FIGURE 4. Base-case flowsheet of HDA process.

VOL. 36, NO. 7, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 1645

to the cooler. The second stream is passed through a seriesof distillation columns (stabilizer, benzene, and toluenecolumns) to separate the benzene product from the othercomponents. The benzene product is obtained at 99% (mol)purity. A feed-recycle stream from the toluene column, whichcontains high purity toluene, is sent to a storage tank.

There are two waste streams in this process: diphenylfrom the bottom of the toluene column and the small amountof vapor from the top of the stabilizer column and containingmainly methane and benzene. The process currently pro-duces benzene at the rate of 7762 kg/h, and the profit is $437 000 per annum as calculated using the unit cost basisshown in Table 1 (7).

Let us assume that the process plant built upon thisflowsheet has been in existence and we are interested inretrofitting the process to minimize wastes without adverselyaffecting process economics. Following the integrated meth-odology, we first use ENVOPExpert to derive the qualitative

waste minimization alternatives shown in Table 2. Hence,quantitative analysis based on those alternatives was inves-tigated using HYSYS simulator with the objective of mini-mizing the environment impact while keeping the overallprocess profit on focus. Table 3 shows the potential envi-ronmental index of the base process in terms of I out

NP , I genNP ,

I outNP , and I gen

NP . The negative values of I genNP and I gen

NP areobtained since some of the input materials are converted toproducts and products are not included in the impactcalculation (see eqs 5 and 6). The low I out

NP significantlymeans that the process has a considerably higher amountof raw materials transformed into products, i.e., low per-centage of waste material as compared with the amount ofproducts being produced. Because of space limitations, weillustrate only some of the important ENVOPExpert’s alter-natives using the quantitative analysis.

Input-Output Level. At this level, ENVOPExpert proposesthree alternatives: (i) direct recycle or recovery-recycle ofthe useful component from the waste streams, (ii) usealternative feed instead of hydrogen and toluene, and (iii)decrease the flow rate of hydrogen and toluene streams. Thefirst alternative, direct recycle of waste vapor stream back tothe furnace, has been evaluated, and the results are shownin Table 3. Compared with the base process, this option leadsto a small increment (2%) in I out

NP . This is mainly due to theslight increase of the waste quantity in the downstreamdiphenyl waste stream. However, the profit increases by 14%,thus making this alternative attractive to implement. Thesecond alternative, use of a different process chemistry,requires changing the entire process. This would entail majormodifications to the current process and is therefore rejectedas unsuitable. The third alternativesdecrease hydrogen andtoluene flow ratessmeans reduced benzene throughput, andsince this is in conflict with the production objective, it isalso rejected.

Reactor Level. At the reactor level, the alternatives canbe broadly expressed as “optimize the reactor’s variables toincrease the conversion of hydrogen and toluene and reducethe diphenyl production”. Examination of the reactor unitindicates that the diphenyl byproduct is formed from anexothermic reaction. In this case, one of the ENVOPExpert’ssuggestions is “decrease the furnace temperature”, which

TABLE 1. Cost Basis for the HDA Processa

unit cost

product and byproductbenzene productb $19.90/kmolmethane purge $3.37/kmolhydrogen purge $1.08/kmol

raw materialhydrogen raw material $2.50/kmoltoluene raw material $14.00/kmolutilitystream $8 000 000/kJcooling water $700 000/kJfuel $4 000 000/kJelectricity $0.05/kWh

equipment (fixed charges) annual basiscompressor $7155 + 815 × power (kW)stabilizer column $650 + 1000 × no. of traysbenezene column $16 300 + 1550 × no. of traystoluene column $3900 + 1120 × no. of traysfurnace $34 500 + 1.172 × heat duty (kJ/h)

× 10-6 × operating hour/yrreactor $74 300 + 1257 × vol (m3)

a Profit ) ∑product and byproduct - ∑raw material - ∑utility -∑equipment charges. b>95% mol.

TABLE 2. Qualitative Waste Minimization Alternatives of HDA Process

level ofhierarchy unit or stream waste minimization alternative

input-output hydrogen feed prevent excessive feed of hydrogen component in hydrogen streamstream use alternative material in hydrogen stream

toluene feed prevent excessive feed of toluene component in toluene streamstream use alternative material in toluene stream

waste vapor direct recycle or recovery-recycle of hydrogen, methane, and benzene component fromwaste vapor

diphenyl waste direct recycle or recovery-recycle of benzene and toluene component from diphenyl wastereaction reactor add recovery system reactor after reactor to recover toluene and hydrogen component

and recycle them back to reactoradd new reactor after current reactor to further transform toluene and hydrogen

component leaving reactoroptimize reactor operating variables to increase conversion of toluene and hydrogen

component inside reactorfurnace decrease temperature of furnace to optimum temperature to start reactionhydrogen stream decrease temperature of hydrogen stream to optimum temperature to start reactiontoluene stream decrease temperature of toluene stream to optimum temperature to start reaction

separation stabilizer optimize reflux ratio and other operating variables of stabilizerimprove control system of stabilizeruse further separation system after stabilizer to recover useful benzene

benzene column optimize reflux ratio and other operating variables of benzene columnimprove control system of benzene column

toluene column optimize reflux ratio and other operating variables of toluene columnimprove control system of toluene columnuse further separation system after stream-splitter to recover useful toluene and benzene

1646 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 7, 2002

would in turn reduce the reaction temperature. The simula-tion results show that reducing the furnace temperature from641 to 632 °C lowers the diphenyl formation by 17%. However,this was accompanied by decreased benzene production andefficiency in the downstream separation process. As can beseen from Table 3, the potential environmental impact ofthe process as given by I out

NP rises by 30%, and the overallprofit reduces by 5% as compared to the base process. Thisalternative is therefore not suitable for implementation.

Separator Level. The feasible suggestions at this level canbe summed up as “optimize the separator’s variables toreduce the presence of undesirable materials in each wastestream”. Here, three design changes to the separators’ refluxratio are considered.

Examination of the waste streams shows that the presenceof toluene in the diphenyl waste stream poses a significantcontribution to the overall environmental impact of theprocess. Thus, reducing the presence of toluene by increasingthe reflux ratio of the toluene column was simulated. Theresults showed that increasing the reflux ratio of the toluenecolumn from 1.1 to 1.5 reduced the I out

NP of the process from400 to 387. This was accompanied with increased toluenerecovery in the feed-recycle stream thus improving theprocess profit, in this case by 1%. The other contributor tothe environmental performance of the process is the benzenein the waste vapor stream. It is therefore desirable to reducethe presence of benzene in this stream by reducing the refluxratio of the stabilizer. In fact, decreasing the reflux ratio from1 to 0.2 reduced the I out

NP of the process by 2% and increasedthe profit by 7% through the energy saving of the column.Both of these alternatives are thus feasible for implementa-tion.

Increasing the reflux ratio of the benzene column willeventually increase the amount of benzene product recov-ered. However, the simulation results already showed a highbenzene recovery at the current reflux ratio. An increase inthe reflux ratio for this column was carried out from 1.0 to2.0. The results showed a very small increment in the benzeneproduction rate but required much higher energy consump-tion, thus making the process lose profitability. Interestingly,higher reflux ratio in the benzene column severely affectedthe separation efficiency of the downstream toluene column.This is shown in Table 3 as an increase in the environmentalimpact from 400 to 1018. Consequently, this alternative isalso rejected.

It is also logical to investigate the effect of simultaneouslydecreasing the stabilizer’s reflux ratio and increasing the refluxratio of the toluene column. Simulation results show thatthe I out

NP of the process was lowered by 6% and the profitincreased by 8%. It is apparent that this alternative is mostoptimal as compared with the previous alternatives since ityields the lowest environmental impact and the highest profit.

This shows that implementing waste minimization does notnecessarily reduce the profitability of the process. Instead,there is an economic incentive in implementing wasteminimization to a process. However, an interesting questionthat naturally arises is the following: Is the profit fromimplementing this alternative the maximum as compared toall the modifications possible for the process? To investigatethis, an increase in the furnace temperature was simulated.As expected, the results showed that increasing the furnacetemperature, in this case from 641 to 650 °C, resulted in anincrease in the environmental impact of the process (I out

NP )633). But this alternative showed a higher profit ($572 000)than the ones previously obtained. This highlights the tradeoffbetween environmental impact and process profitability forthe HDA case study. Optimal stream flow rates from thebenzene column that simultaneously minimize the envi-ronmental impact and maximize the process profit wasexamined using the HYSYS′ optimizer tool. The result showsa slight improvement both in the environmental impact (I out

NP

) 395) and the profit ($439 000) as compared to the baseprocess.

The above case study illustrates the following advantagesof the integrated framework: a qualitative waste minimiza-tion analysis using ENVOPExpert is a practical, fast, andcomprehensive approach to identify and generate feasiblealternatives for implementation in the process; a quantitativeapproach using process simulator and the WAR algorithm ishelpful in evaluating each of the alternatives on the basis ofcost and environmental impact analysis. Through theimplementation of this framework, the usually large, difficult,and expensive nonlinear optimization problem for solvingwaste minimization of the entire plant can be replaced bysmaller optimization problems that focus only on the relevantsection of the process. Another advantage is that designerscan also take into account other important nonquantifiablefactors such as safety, plant layout, etc. while deciding thebest process modification. The case study results also indicatethat the qualitative analysis embedded in ENVOPExpert canguide the nonexpert in identifying waste minimizationalternatives and establish the integrated methodology as apractical approach to material waste minimization in large-scale chemical processes. Encouraged by these positiveresults, we are currently working on improving the usabilityof ENVOPExpert. The current implementation requires theuser to switch between the G2-based ENVOPExpert and thesimulation in HYSYS. Recent developments in ComputerAided Process Engineering-Open Simulation Environment(CAPE-OPEN) have enabled collaboration between differentprocess engineering software through data and propertysharing capabilities (18). We intend to develop an integratedsolution with ENVOPExpert seamlessly communicating withany CAPE-OPEN compliant process simulator and using the

TABLE 3. Environmental and Economic Analysis of HDA Process

environmental impact (impact h-1)

waste minimization alternative I outNP I gen

NP I outNP I gen

NP profit ($)

base-case 400 -19 139 0.04 -1.85 437 000input-output level

direct recycle of waste vapor stream 408 -19 129 0.04 -1.85 496 000reactor level

decrease furnace temp from 641 to 632 °C 518 -19 020 0.05 -1.89 414 000increase furnace temp from 641 to 650 °C 633 -18 865 0.06 -1.75 572 000

separator levelincrease reflux ratio in toluene column from 1.1 to 1.5 387 -19 500 0.04 -1.85 442 000decrease reflux ratio in stabilizer column from 1.0 to 0.2 391 -19 147 0.04 -1.84 468 000increase reflux ratio in benzene column from 1.0 to 2.0 1018 -18 519 0.10 -1.86 -333 000decrease reflux ratio in stabilizer column from 1.0 to 0.2

and increase reflux ratio in toluene column from 1.1 to 1.5377 -19 160 0.04 -1.84 473 000

VOL. 36, NO. 7, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 1647

results for WAR and process economics calculations. Thiswill significantly enhance the ease of use of ENVOPExpert asthe “green design” decision support tool. In addition, we willalso augment the knowledge base in ENVOPExpert andextend our methodology to energy optimizations.

Supporting Information AvailableSequence of procedures (including figures) employed byENVOPExpert. This material is available free of charge viathe Internet at http://pubs.acs.org.

Literature Cited(1) Isalski, W. H. Environ. Prot. Bull. 1995, 34, 16-21.(2) Douglas, J. M. Ind. Eng. Chem. Res. 1992, 31, 238-243.(3) Douglas, J. M. Conceptual Design of Chemical Processes; McGraw-

Hill: New York, 1988.(4) Hopper, J. R.; Yaws, C. L.; Vichalak, M.; Ho, T. C. Waste Manage.

1994, 14 (3), 187-202.(5) Cabezas, H.; Bare, J. C.; Mallick, S. K. Comput. Chem. Eng. 1999,

23, 623-634.(6) Young, D. M.; Cabezas, H. Comput. Chem. Eng. 1999, 23, 1477-

1491.(7) Fu, Y.; Diwekar, U. M.; Young, D.; Cabezas, H. Clean Prod.

Processes 2000, 2, 92-107.(8) Dantus, M. M.; High, K. A. Ind. Eng. Chem. Res. 1996, 35 (12),

4566-4578.

(9) Halim, I.; Srinivasan, R. An intelligent system for identifyingwaste minimization opportunities in chemical processes. InEuropean Symposium on Computer Aided Process Engineering- 10; Pierruci, S., Ed.; Elsevier Science: Amsterdam, 2000; pp829-834.

(10) Halim, I.; Srinivasan, R. AIChE Annu. Meeting, Los Angeles, CA,November 12-17, 2000; Paper 233r.

(11) Halim, I.; Srinivasan, R. Regional Symposium on ChemicalEngineering (RSCE), Singapore, December 11-13, 2000; PaperPDD6.3.

(12) Halim, I.; Srinivasan, R. Ind. Eng. Chem. Res. 2002, 41, 196-207.(13) Halim, I.; Srinivasan, R. Ind. Eng. Chem. Res. 2002, 41, 208-219.(14) U.S. EPA. Waste Minimization Opportunity Assessment Manual;

U.S. Environmental Protection Agency, Hazardous WasteEngineering Research Laboratory: Cincinnati, OH, 1988.

(15) Yang, Y.; Shi, L. Comput. Chem. Eng. 2000, 24, 1409-1419.(16) Hilaly, A. K.; Sikdar, S. K. J. Air Waste Manage. Assoc. 1994, 44,

1303-1308.(17) Hyprotech. HYSYS.Plant, Ver. 2.1; Hyprotech: Calgary, Canada,

1999.(18) Braunschweig, B. L.; Pantelides, C. C.; Britt, H. I.; Sama, S. Chem.

Eng. Progr. 2000, September, 65-76.

Received for review May 10, 2001. Revised manuscript re-ceived December 19, 2001. Accepted December 20, 2001.

ES0155175

1648 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 7, 2002