Computers and Chemical Engineering 35 (2011) 13431358

Contents lists available at ScienceDirect

Computers and Chemical Engineering

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

A combined heuristic and indicator-based methodology for design of sustainable chemical process plants
Iskandar Halim a , Ana Carvalho b,c , Rajagopalan Srinivasan a,d, , Henrique A. Matos b , Raqul Gani e
a

Institute of Chemical and Engineering Sciences, A*STAR (Agency for Science, Technology and Research), 1 Pesek Road, Jurong Island, Singapore 627833, Singapore CPQ, Instituto Superior Tcnico, UTL, Av Rovisco Pais, 1049-001 Lisboa, Portugal CEG-IST, Instituto Superior Tcnico, Av Rovisco Pais, 1049-001 Lisboa, Portugal d Laboratory for Intelligent Applications in Chemical Engineering, National University of Singapore, 10 Kent Ridge Crescent, Singapore 119260, Singapore e CAPEC, Department of Chemical and Biochemical Engineering, Technical University of Denmark, Soltofts Plads 1, Building 229, DK-2800, Lyngby, Denmark
b c

a r t i c l e

i n f o

a b s t r a c t
The current emphasis on sustainable production has prompted chemical plants to minimize raw material and energy usage without compromising on economics. While computer tools are available to assist in sustainability assessment, their applications are constrained to a specic domain of the design synthesis problem. This paper outlines a design synthesis strategy that integrates two computer methodologies ENVOPExpert and SustainPro for simultaneous generation, analysis, evaluation, and optimization of sustainable process alternatives. ENVOPExpert diagnoses waste sources, identies alternatives, and highlights trade-offs between environmental and economic objectives. This is complemented by SustainPro which evaluates the alternatives and screens them in-depth through indicators for prot and energy, water, and raw material usage. This results in accurate identication of the root causes, comprehensive generation of design alternatives, and effective reduction of the optimization search space. The framework is illustrated using an acetone process and a methanol and dimethyl ether production case study. 2011 Elsevier Ltd. All rights reserved.

Article history: Received 9 September 2010 Received in revised form 27 January 2011 Accepted 9 March 2011 Available online 16 March 2011 Keywords: Sustainable development Process simulation Heuristic analysis Indicator analysis Energy and water minimization Sustainability metrics

1. Introduction Chemical industry has been the epitome of modern process industry. Its importance to the world economy is apparent with U$2.4 trillion annual sales of more than 70,000 different products, ranging from basic chemicals, specialty chemicals, life sciences, and consumer products (KPMG, 2010). Despite its importance, the general public perception of the chemical industry has not been favourable. A recent survey conducted by the European Chemical Industry Council revealed that the public rating of this industry was 49% positive and 44% negative (ICIS, 2007). Such a negative public perception is most likely due to the growing concerns over the possible ill-effects from the use of chemicals and chemical based products and the environmental damages (such as global warming, ozone depletion, etc.) caused by the pollutant emissions. For example, of the total CO2 emission discharged to the atmosphere, about 25% can be attributed to the chemical industry (ICCA, 2009). Another main concern is associated with the large-scale processing

Corresponding author at: Laboratory for Intelligent Applications in Chemical Engineering, Department of Chemical Engineering, National University of Singapore, 10 Kent Ridge Crescent, Singapore 119260, Singapore. Tel.: +65 65168041; fax: +65 67791936. E-mail address: chergs@nus.edu.sg (R. Srinivasan). 0098-1354/$ see front matter 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.compchemeng.2011.03.015

of non-renewable resources that are typical of the chemical industry. Representing 4% of the world economy, the chemical industry is currently responsible for 7% of global energy use (Lines, 2005). In the USA, the percentage is even higher with the chemical industry accounting for nearly 25% of the energy used for industrial activity; further, 88% of this is derived from fossil fuels (National Research Council, 2005). Certainly, such level of consumption cannot be sustained in the long run. With the recent emphasis on process sustainability, much improvement is expected from the chemical industry to minimize its raw material and energy consumptions and pollutant generation without compromising the economic value of the enterprise. Responding to these challenges then requires a new insight into the characteristics of a sustainable system and a fundamental rethinking of how a chemical production plant is to be designed, built and operated (Bakshi & Fiksel, 2003). Design is an iterative activity which requires decision making at various stages and at different levels of detail. Overall, a design activity involves accepting as input an abstract description of the desires of an organization and delivering a rened description of a concrete product, process, or system that will satisfy those desires (Cano-Ruiz & McRae, 1998). In the context of chemical process design, for example, the abstract goal can be to convert excess methanol from one plant into a dimethyl ether (DME) product. In this case, the more rened description will then be a chemical

1344

I. Halim et al. / Computers and Chemical Engineering 35 (2011) 13431358

Abstract description

Inputs and results Problem specification

Steps Concept generation

Design approaches

Alternative generation

Design alternatives Analysis Performance Evaluation and Comparison Cost, environmental, safety, etc. Optimization

Refined description
Fig. 1. Design synthesis activity (Biegler et al., 1999).

process plant to accomplish just that. Overall, such an activity of converting an abstract description into a more rened description involves a sequence of steps (see Fig. 1). The rst of these steps is concept generation. Here, different concepts on which to base the design need to be identied. Typical questions to be addressed include: should the chemistry for the methanol conversion process be based on ones found in the literature?, should unconventional solutions be sought after?, and should a particular design strategy (for example waste minimization as compared with end-of-pipe treatment) be adopted in the design?. Next is generation of alternatives. At this stage, different possibilities to improve the base design are considered. This is done by assessing different parts of the process to derive possible modications to improve the process, for example, higher conversion of methanol to DME, suppression of by-products formation, and reduction of energy usage. The third is analysis of each alternative to establish its performance. Principally, this equates performing mass and energy balances of the process to determine the impacts from the alternative. The use of process simulators (such as CHEMCAD, Aspen Plus, HYSYS, PRO/II and gPROMS) will be benecial at this stage. The fourth step involves evaluating and comparing the process performance. In this step, indicator metrics that measure the economic worth, environmental impact, safety impact, and so on need to be calculated for comparing the alternatives and shortlisting the promising ones. Finally, optimization can be performed for adjustment and renement of the process variables to further improve the design. This optimization can be cast as a multi-objective decision problem, where trade-offs between economic, environmental, and safety concerns need to be resolved.

Design synthesis is thus a complex, laborious, time consuming, expensive and knowledge-intensive activity. The availability of computer-based tools that can perform systematic and thorough alternative generation, evaluation, and optimization altogether is very attractive since they can assist the designers in achieving the design goals at reduced time, money and efforts. Since the early 1960s, computer-based tools have been successfully deployed in the oil and chemical industries to expedite development and optimize the design and operation of integrated processes (Petrides, Koulouris, & Lagonikos, 2002). Their important roles to sustainability studies are now apparent with approximately 300 software applications (commercials and prototypes) available for deployment (Tsoka, Johns, Linke, & Kokossis, 2004). One sucessful example is process simulators which have been extensively used to compare the environmental impacts from various plant modications (Cabezas, Bare, & Mallick, 1999; Chen & Shonnard, 2004). Building on the results of process simulators, mathematical optimization have been applied to ne-tune the decision variables controlling the process such as pressure, temperature and ow rate to simultaneously maximize the prot and minimize the environmental impact (Dantus & High, 1996; Fu, Diwekar, Young, & Cabezas, 2000). Process integration tools such as pinch-based analysis have also been successfully deployed for targeting energy and water reduction and designing heat- and mass-exchange networks (ElHalwagi, 1997; Linnhoff, 1995). Another important tool is articial intelligence-based technique including P2TCP a heuristic design based expert system for minimizing waste generation and energy consumption (Pennington, 1999) and ENVOPExpert a combined knowledge-based and optimization system for sustainable design and operation (Halim & Srinivasan, 2011). An indicator-based computer tool, called SustainPro, has also been developed by Carvalho, Matos, & Gani (2008) for generation and evaluation of sustainable design alternatives. While all these tools are benecial, their applications are limited to solving specic domains of sustainability problem, such as minimization of process or utility wastes, optimization of process variables, and generation of qualitative design alternatives. To deliver a comprehensive sustainability solution to a process, collaboration between the different tools is needed. However, such effort has proven to be challenging as each seeks to tackle the problem from a different viewpoint (Hilaly & Sikdar, 1994). For example, the analysis and evaluation methods implemented in ENVOPExpert is limited to solving waste problem. Further, a large optimization search space could result since the proposed design alternatives and decision variables for optimization are considered of equal importance to the process. In the case of SustainPro, while the alternatives proposed and decision variables can be ranked in term of potential for improvement, they are not optimized. This paper outlines a design synthesis methodology that claims merits for integrating two computer systems ENVOPExpert and SustainPro for sustainable design and operation. The term integration here means the two systems are deployed on parallel or sequentially (depending on the step of the methodology) in a complementary fashion for process design synthesis. For this, we have also used a set of sustainability metrics dened by the Institution of Chemical Engineers (Azapagic, 2002) which measures the process sustainability in terms of economic; energy, water and material utilization; and environmental impacts. For the economic measures, prot and/or value added are used. For the environmental impact measures, instead of using the denition Azapagic, the WAR (waste reduction) algorithm proposed by Cabezas et al. (1999) is used. The rationale behind this research work is that by combining the collective strength of each tool, the benets that can be gained would exceed those which could have been achieved individually. ENVOPExpert has the functionality of qualitative analysis of waste problem through heuristic solution and quantitative analysis through optimization algorithm. On the other hand, Sus-

I. Halim et al. / Computers and Chemical Engineering 35 (2011) 13431358

1345

Fig. 2. Flowsheet of acetone production process.

tainPro has the capability of concise identication and evaluation of sustainability bottleneck through metric analysis and prioritization of variables. An integration of ENVOPExpert and SustainPro would lead to more accurate identication of the root causes of the sustainability problem, comprehensive generation of design alternatives, quantication of the cost-benet of the design alternatives, and more focused optimization of design variables through screening of variables with most impact on the process. The rest of this paper is organized as follows: the next section begins with a description of acetone process which will be used to illustrate the approach. In Section 3, a methodology that combines ENVOPExpert and SustainPro is presented together with illustration of the main features through the acetone process. This is followed in Section 4 by a methanol-DME case study where application results of the combined methods and tools are highlighted. 2. Acetone process One possible alternative for manufacturing acetone is through dehydrogenation of isopropyl alcohol (IPA) or 2-propanol. The primary advantage of this process over the other routes is that the acetone produced is high in purity which makes it suitable for use as solvent in the pharmaceutical industry. Fig. 2 shows the owsheet of an acetone process (Turton, Bailie, Whiting, & Shaeiwitz, 2003). A feed stream comprising of an azeotropic mixture of IPA and water (88% mass of IPA) is initially mixed with a recycle stream containing IPA, water and acetone. The feed mixture is rst vaporized using high pressure steam before being passed to a reactor where an endothermic reaction takes place as follows: IPA Acetone + H2 . The reaction kinetics are rst order with respect to the concentration of IPA and can be expressed as follows: rIPA = k0 exp Ea RT CIPA kmol m3 reactor s (1)

scrubbed with water to recover the acetone. The liquid from the phase separator is next combined with the liquid from the bottom of the scrubber and sent to the separations section. Two distillation columns are used to separate the acetone product (99.9% mass) and to remove the excess water, which is then recycled back to the front-end of the process as an azeotropic mixture. Table 1 shows the economic and environmental data for this process (Martin & Young, 2008; Turton et al., 2003). There are two waste streams in this processthe vapour stream leaving the scrubber and the waste water from the second distillation column. Our objective in this case study is to propose design alternatives to improve the sustainability of the process. In this case, we measure the sustainability of the process through a set of sustainability metrics incorporating the economic; energy, water, material utilization; and WAR impacts.

Table 1 Cost and environmental impact data for acetone process. Description Cost of raw material ($/kg) Feed Fresh-water Price of product ($/kg) Acetone-product Byproduct-gas Cost of waste stream ($/kg) Waste-vapor To-treatment-plant Cost of energy ($/kW h) E-Pump100 E-Heater100 E-Reactor E-Cooler101 E-Cooler102 E-Cond101 E-Reb101 E-Cond102 E-Reb102 Environmental impact (per kg) Hydrogen Acetone IPA Water Value 0.63 4 105 0.88 1.24 103 0.005 0.001 0.05 0.011 0.011 6 104 0.008 6 104 0.007 6 104 0.007 0 0.478 0.1401 0

where Ea = 72.38 MJ/kmol, k0 = 3.51 105 (m3 gas)/(m3 reactor s) and CIPA = (kmol)/(m3 gas). During the reaction, heat is supplied using a circulating molten salt. The reactor efuent, containing unreacted IPA, acetone, water and hydrogen byproduct is cooled in two heat exchangers prior to entering a phase separator. The vapour leaving the separator is

1346

I. Halim et al. / Computers and Chemical Engineering 35 (2011) 13431358

Step 1: Process information collection Information source (Process simulator) Flowsheet Process chemistry Mass and energy balances Economic, environmental and safety data

P-graph Heuristic rules Flow-path analysis Indicators calculation Synthesis rules

ENVOPExpert

SustainPro

Step 2: Alternative generation Heuristic synthesis method Flowsheet decomposition Mass and energy indicators

Process simulator

Step 3: Analysis Implementation to the base case design through structural or parametric modification

Sustainability metrics Safety indices

SustainPro

Step 4: Evaluation and comparison Performance evaluation of new alternative through metrics Comparison between alternatives

Process simulator

Optimization module of ENVOPExpert

Step 5: Optimization Identification of variables affecting the objective function(s) Variable optimization and identification of tradeoffs Selection of best alternative

Step 6: Implementation Implementing new alternative to the process

Fig. 3. Sustainability design synthesis steps.

3. Methodology for sustainable process design This section outlines a methodology that exploits the features of ENVOPExpert and SustainPro for generating and evaluating design alternatives for sustainable chemical processes. The methodology has been developed following the sequence of steps proposed by Biegler, Grossmann, & Westerberg (1999) to identify the structural as well as operational changes suitable to the process (see Fig. 1). As such, the methodology is applicable to different stages of the process design life-cycle from conceptual design to retrotting. Fig. 3 shows the steps employed in the methodology. As shown in the gure, the methodology is systematic but not fully automated in a sense that the designer decisions are taken into consideration at each stage of the analysis and the designer is guided through the different steps. First, structural modications to the owsheet design are explored following the suggestions given by ENVOPExpert and SustainPro. The implementation of these alternatives (structural changes of the owsheet) will depend on the judgement of the designer in the next stage. Subsequently, interactions between key process variables are identied and changes to these variables are simulated. This is followed by optimization

of the variables using stochastic optimization to nd the process optima (local or global). All these sequential steps are described in details below. 3.1. Step 1process information collection Fig. 4 shows the schematic ow of process information for sustainability analysis. The procedure starts with collecting information about the base process including the owsheet and the mass and energy balances. In HYSYS, such information can be easily extracted in the form of XML data (Hyprotech, 2003). This HYSYSXML data, which consists of the building blocks of the process, contains the list of materials, unit operations, stream and unit operation connections and positions, reaction kinetics, thermodynamic model, and user-specied values of the operating variables and process parameters. Beside them, other data needed by ENVOPExpert are material and stream classication, chemical environmental impact (WAR algorithm) and economic data. These need to be specied explicitly if not available in the library of ENVOPExpert. Once the required process data is available, it can then be used to construct the P-graph model of the process.

I. Halim et al. / Computers and Chemical Engineering 35 (2011) 13431358

1347

Information flow scheme

HYSYS simulator

Microsoft Excel Material classification Stream status WAR impact data Economic data

HYSYS-XML

ENVOPExpert Mass balance data Energy balance data Unit and connection information SustainPro

P-graph Heuristic design method ENVOPExpert alternatives

Flowsheet decomposition Indicators and metrics calculation ISA algorithm SustainPro alternatives

Alternatives comparison

Implementation to base case process

Alternative generation scheme

Fig. 4. Schematic of process information ow and alternative generation.

The required input to SustainPro is the mass and energy balances of the process. Such information is delivered by ENVOPExpert using a text le format that is cast into the following syntax: streamtemperature pressure heat-ow name vapour-fraction array-of-components array-of-mass-of-components from-unit from-subunit to-unit to-subunit for the material stream and stream-name connection-status connecting-unit heat-ow for the energy stream. As an illustration, consider the ashed-vapour stream connecting V-100 and T-100 in the acetone process. The information about this stream is cast as follows: Flashed-vapour 1.0 20.0 C 188.586 kPa 1.697 106 kJ/h Hydrogen Acetone 2-Propanol H2 O 123.871 kg/h 400.834 kg/h 4.783 kg/h 12.057 kg/h V-100 none T-100 none . In the same way, the acetone product stream from T-101 can be structured as Acetone-product 0.0 54.565 C 130.0 kPa -1.444 107 kJ/h Hydrogen Acetone 2-Propanol H2 O 0.015 kg/h 3415.09 kg/h 1.804 104 kg/h 3.403 kg/h T-101 condenser none none . For the energy stream of heater E-100, the syntax is specied as E-Heater100 input E-100 6.1 106 kJ/h . This information combined with the unit information and economic and environmental impact data are then used by SustainPro to determine the ow path decomposition of the process.

3.2. Step 2alternatives generation and comparison In this step, new design alternatives are identied using ENVOPExpert and SustainPro. Fig. 4 highlights the complimentary approaches of ENVOPExpert and SustainPro for generation of design alternatives. These are explained in detail next.

3.2.1. Knowledge-based alternatives generation using ENVOPExpert A knowledge-based system, called ENVOPExpert, has been developed to assist in sustainability assessment of chemical processes (Halim & Srinivasan, 2002, 2008, 2011). It has been successfully tested on a number of industrial-scale process including hydrocarbons separation, alcohol washing process, hydrodealkylation (HDA) of toluene to benzene, chemical intermediate manufacturing process, and a biodiesel production and found to generate results comparable to the analysis by human experts. ENVOPExpert has been implemented in an object-oriented framework using G2 expert system shell with the following task: Given a ow diagram and chemistry of a chemical process plant, identify alternatives which simultaneously minimize waste generated in the plant and improve the process economic (i.e. prot). This makes it applicable to the various stages in the life of the process from conceptual design to retrotting. ENVOPExpert performs its analysis using the methodologies that have been embedded into its knowledge base including material ow representation, waste source diagnosis, heuristic qualitative alternative generation, environmental and economic analysis and multi-objective optimization. This section summarizes each of the methodologies for generation of design alternatives. The reader is referred to the above papers for more details. The rst step in the quest for waste minimization in any process plant is to identify each material component present in each unit of the plant. This information can be easily obtained from the plant operational data or, if available, process simulation results in the forms of owsheet, reaction chemistry, and material and energy balances. In addition to these, other data needed includes

1348

I. Halim et al. / Computers and Chemical Engineering 35 (2011) 13431358

T-100

Acetone IPA Acetone IPA

MIX-100

Acetone

Acetone IPA

Acetone IPA

V-100

Acetone

Acetone

Acetone Acetone

P-100

E-100

PFR-100

E-101

E-102

IPA Acetone (recycle)

Acetone (product stream)

MIX-101 T-101

T-102

Acetone Acetone

(a)
Acetone (waste stream)

T-100

IPA IPA IPA

MIX-100 P-100

IPA

IPA
E-100

IPA
PFR-100

IPA
E-101

IPA
E-102

IPA

V-100

IPA IPA (recycle)

MIX-101 T-101

(b)
IPA (waste stream)

T-102

IPA IPA

T-100

Water Water Water

MIX-100 P-100

Water Water Water

Water
E-100

Water

V-100

Water
E-101

Water
E-102

Water

PFR-100

Water (recycle)
T-102

MIX-101 T-101

Water Water

(c)
Water (waste stream)
Fig. 5. P-graph diagnosis for to-treatment-plant stream.

the classication of material components (raw material, utility, waste or product), status of each material component in output streams (desirable or detrimental), environmental impact factors of the different materials and economic data. Once the information is available, it is then used to represent the material ow in the process. In ENVOPExpert, this is done through P-grapha special directed bipartite graph for representing process structure suitable for the synthesis problem (Friedler, Varga, & Fan, 1994). In the P-graph model, a circle represents a material stream and a bar represents a unit operation. Connecting the stream and the operating unit nodes are arcs, which dene the precedence of units. The P-graph representation of a process provides a convenient framework for diagnosing the origins of waste in the process and for deriving waste minimization alternatives. Starting from each waste stream, sources of waste, such as impurities in inlet stream, excessive feed of materials, low conversion of materials, waste byproducts and ineffective separation of useful materials, can be identied by tracing each material in the waste stream in the upstream direction. This procedure is illustrated in Fig. 5 which traces the origins of each material component in the to-treatmentplant stream of the acetone process. In this case, ve waste sources can be diagnosed through the P-graph analysis:

(1) Excessive acetone in the recycle stream (Fig. 5a). (2) Excessive IPA in the feed and recycle streams (Fig. 5a and b). (3) Excessive water in the feed, recycle and fresh-water streams (Fig. 5c). (4) Low conversion of IPA in PFR-100 leading to its presence in the to-treatment-plant stream (Fig. 5b). (5) Inefcient acetone separation in T-101 column causing its escape to the to-treatment-plant stream (Fig. 5a). The subsequent step after diagnosing the waste sources is to derive waste minimization alternatives to the sources. For this, a set of design heuristics, which have been compiled from Douglas (1992) and Smith (1995), is applied for generating the design alternatives. The heuristics are implemented in ENVOPExpert through a set of IF-THEN rules which propose the modication required in the process unit or material stream to minimize waste generation and recover the valuable material from the waste stream. These can be summarized as follows: (1) If an impurity exists in a feed stream, then prevent it from entering the process.

I. Halim et al. / Computers and Chemical Engineering 35 (2011) 13431358 Table 2 ENVOP Expert waste minimization solution for acetone process. Stream/unit name Feed Fresh-water Feed-recycle PFR-100 Waste-vapour Waste diagnosis Raw material in waste stream Washing agent in waste stream Product in recycle Raw material in waste stream Washing agent and product in waste stream Waste minimization solution

1349

To-treatment-plant

Raw material, washing agent, and product in waste stream

Prevent excessive feed of IPA and water in the stream to avoid them from becoming waste Use alternative material in the stream Prevent excessive feed of water in the stream Use alternative material in the stream Reduce the recycle of acetone in the process Increase the conversion of IPA Add separation unit to recover acetone, water, and hydrogen from becoming waste Optimize the operating condition of the distillation (T-101) to avoid useful material escaping as waste stream Direct-recycle or recovery-recycle of acetone, water, and hydrogen in the stream Add separation unit to recover acetone, water, and IPA from becoming waste Optimize the operating condition of the distillation (T-102) to avoid useful material escaping as waste stream Direct-recycle or recovery-recycle of acetone, water, and IPA in the stream

(2) If a useful material in the feed stream exits through a waste stream, then prevent excessive feed of that material. (3) If a unit-operation generates waste, then eliminate or reduce that byproduct formation. (4) If an inefcient separation unit exists, then improve the separation process or include another separation unit. (5) If a useful material exists in a waste stream, then recover and recycle that material. (6) If a product material recycles in the process, then reduce the presence of that material in the recycle. Thus, by using these heuristics, design alternatives to eliminate or minimize the presence of acetone, IPA and water in the two waste
Table 3 Mass and energy indicators of SustainPro. Indicators MVA (material value added) Formula
(c ) MVAop

streams can be proposed. Table 2 shows the complete list of such alternatives derived by ENVOPExpert for this case study. 3.2.2. Indicator-based alternatives generation using SustainPro A systematic methodology for sustainable chemical process design and a software tool called SustainPro that implements the methodology have been developed (Carvalho et al., 2008; Carvalho, Matos, & Gani, 2009a). The tool generates, screens and identies sustainable alternatives in a given chemical process by locating the operational, environmental, economical and safety related bottlenecks inherent in the process. It has been applied to a wide range of chemical processes such as reneries, gas processing, and pharma-

Denition
(c ) c mop (PPop

c PRop )

EWC (energy and waste cost)

ECk =

(c )

U u=1

PEu Qu

m
UK

(c ) (c ) A (Tm ,pm ) k u,k

mop : owrate of the compound c in open-path op c PPop : sale price c PRop : purchase price k: open- or the closed-path u: sub-operations index uk: index of all compound path ows in u. U: total number of sub-operations UK: total number of compound path ows in a sub-operation PEu : specic price of the utility needed in sub-operation u Qu : energy consumption Au : allocation factor calculated at the average temperature Tm and pressure pm and it is normally a physical property MVA: material value added EWC: energy and waste cost
r,rk,k :

(c )

uk=1

mm,uk Au,uk (Tm ,pm )

TVA (total value added)

TVA = MVA EWC

R RK
(c ) r,rk,k Er,rk,k FP fp=1

RQ (reaction quality)

RQk =
r =1 rk=1

(c )

extent of reaction rk

n(fp)

AF (accumulation factor)

AF =

mcp
I i =1 EN

(c ) OP (c ) op=1 i,op

n(fp): molar owrate of the nal desired product fp FP: total number of desired products r: index of the reactive unit operations R: total number of reactive unit operations RK: total number of reactions in the unit operation (c ) Er,rk,k : effect of a given compound in the reactions involved in the path ow
d

(c ) f + k=1 i,k

mcp : owrate of compound c in the closed-path, cp fi,k , di,op : compound ows leaving the cycle ow
(c ) (c )

(c )

EAF (energy accumulation factor)

EAF =

I i =1

EN

eblec

fh + k=1 i,k

OP

I: total number of the vertices encountered in the energy closed-path ow i: index of vertices ec: index of the cycle energy path ows ebl: energy base level PE: utility cost EOP: owrate of open-path SS: total number of supplies that energy contributes are signicant to the demand, d

dh op=1 i,op

DC (demand cost)

DCsu,d = PEsu EOPsu,d

SS

TDC (total demand cost)

TDCd =
su=1

DCsu,d

1350

I. Halim et al. / Computers and Chemical Engineering 35 (2011) 13431358 Table 4 Calculated indicator values of the most sensitive closed-paths for acetone case study. Closed-path CP8 CP9 Component IPA Water EWC (3 $/y) 12 8 AF 0.014 0.007

Table 5 Indicator values of the most sensitive open-paths. Open-path Fig. 6. Example of closed- and open-path. OP1 OP12 OP 17 Component Hydrogen IPA Water MVA (103 $/y) 1247 1474 156 EWC (103 $/y) 0 41 498 TVA (103 $/y) 1247 1515 655

ceuticals. This section highlights each step of the methodology for solving the acetone case study. The reader is referred to the above references for more details. The methodology starts with collecting the information about the process in the forms of mass and energy balances, material costs and process operation costs. Such data can be obtained from the simulation results or operational plant data. In this acetone case study, data was obtained from the HYSYS simulator in the form of HYSYS-XML data. This owsheet information was then translated by ENVOPExpert into the text le format described in Section 3.1 for input to SustainPro. In the subsequent step, the owsheet information is decomposed into a set of closed- and open-paths, as shown in Fig. 6. The closed-paths are the process recycles with respect to each material in the process. In other words, they are the ow paths which start and end at the same unit of the process. On the other hand, an open-path consists of an entrance and an exit of a specic compound in the process. The entrance of the compound into the system can be due to its entrance through a feed stream or by its production in a reactor unit. The exit of a compound can be due to a demand (exit) stream or by its reaction in a reactor unit. For this acetone case study, the owsheet decomposition method generates 10 closed-paths and 18 open-paths. Once the owrates of each closed- and open-path are known, the next step is to calculate the corresponding indicators for the paths. These are listed in Table 3 which shows ve mass indicators (MVA, EWC, TVA, RQ and AF) and three energy indicators (EAF, DC and TDC) for identifying the locations within the process where material and energy paths face barriers with respect to costs,

benets, or accumulation. Such locations indicate high potential for process improvements. At the same time, safety indices (Heikkil, 1999) and a set of sustainable metrics (Azapagic, 2002) are also calculated to evaluate the process performance. To illustrate the indicator procedure, consider Figs. 7 and 8 and their corresponding Tables 4 and 5 which highlight the most sensitive indicators for the closed- and open-paths of this case study. Table 4 indicates high values of EWC due to the recycling of IPA and water in the closedpaths, which means high energy consumption. Table 5 shows the most sensitive open-paths due to Hydrogen ow (OP1), IPA ow (OP12) and water ow (OP17)these are indicated by the high negative values of MVA and TVA. A high positive value of EWC indicator in OP17 also points to high energy consumption in this path. As the MVA values for the open-paths are much higher than their closed counterparts, the search for new alternatives takes priority for these open-paths. Subsequent to the indicator calculation step is the indicator sensitivity analysis (ISA). Here, the most promising indicators for improvements are selected for further evaluationthese are OP12 and OP17 but not OP1. The reason can be explained as follows. While examination of the indicator values for OP1 indicates the possibility for improving the sustainability of the process through reducing the hydrogen ow in this path, such hydrogen reduction would concomitantly reduce the process prot since hydrogen is a valuable byproduct of the main reaction. In the next step, the indicator sensitivity analysis (ISA) algorithm is applied to OP12

Fig. 7. Most sensitive indicators for closed-paths of acetone process.

I. Halim et al. / Computers and Chemical Engineering 35 (2011) 13431358

1351

Fig. 8. Most sensitive indicators for open-paths of acetone process.

and OP17 to assess their impacts to the overall process prot. For this, the following mathematical equation that describes the prot function is used: Fobj = Cf1 VF 1 + Cf2 VF 2 + + Cfn VFn (2)

be applied with the objective of reducing the IPA and water in the waste stream. 3.2.3. Alternatives comparison A comparison between ENVOPExperts and SustainPros results (see Tables 2 and 7) indicates that ENVOPExperts results are in-line with the conclusions of SustainPro. Both systems suggest reducing the ow of feed water and installing a new separation to recycle IPA and water from the waste stream. They are thus the candidates for implementation. A comparison between the results also highlight that, ENVOPExpert performs a more generic level of analysis through the use of qualitative heuristics. On the other hand, the quantitative approach of SustainPro performs a more specic analysis by focusing on the most critical points in the process that offer the highest potential for improvement. For instance, using the material accumulation indicator, a high build-up of certain materials in a recycle stream can be diagnosed by SustainPro and hence suggestions for reducing this material in the recycle line can be proposed. In ENVOPExpert, since the analysis is done qualitatively, it does not distinguish between high and low build-up of material in the recycle. In this case, ENVOPExpert would analyze the classication of the material in the recycle stream and infer a heuristic solution with respect to that material. In this case, if the material present is a product, the ENVOPExperts suggestion will be to reduce the recycle of that product in the process. 3.3. Step 3analysis of proposed alternatives Here, the proposed design alternatives reducing the freshwater owrate and adding a new separation unit to recoverTable 7 Summary of the proposed alternatives based on indicator analysis. Bottleneck diagnosis Possible solution Improve or insert new separation to recycle the raw material Reduce the feed stream Improve or insert new separation to recycle the raw material

where Cf1 to Cfn are the coefcients (such as selling prices or costs) and VF1 to VFn are the process variables inuencing Fobj (prot value) and can be obtained from the mass and the energy balance information. A sensitivity analysis is then performed for determining the indicators that lead to the largest positive or negative changes in the prot value. Table 6 lists the scores for different paths obtained using this algorithm. From the table, it can be observed that MVA indicator related to OP12 for IPA is the most sensitive as it is the one with the highest score. The second highest score is OP17 for water. Consequently, these two indicators are considered as the target indicators for improvements. Next, a design sensitivity analysis to the operational variables that inuence the target indicators is performed. The objective of this step is to select the operational parameter that allows the highest improvement in the target indicator. In this acetone case study, it has been found that the most signicant operational variables affecting the MVA of OP12 and OP17 are the IPA ow and water ow, respectively. Consequently, a set of new design alternatives must be proposed to reduce these ows. These are shown in Table 7 which lists feed reduction and adding a new separation process as the most important alternatives to improve the sustainability of the process. In this case, direct recycle of IPA and water mixture cannot be implemented without additional separation unit to separate the mixture. To determine the type of separation to be used, the synthesis methodologies based on Jaksland, Gani and Lien (1996), DAnterroches and Gani (2005), and Harper and Gani (2000) can
Table 6 Scores for different paths using ISA algorithm for acetone case study. Path OP 12 OP 17 OP 17 Indicator MVA MVA EWC Compounds IPA Water Water Scores 40 19 19

Waste of Raw material (IPA) Waste of washing agent (Water)

1352

I. Halim et al. / Computers and Chemical Engineering 35 (2011) 13431358

Table 8 Sustainability metrics for the base and modied acetone process. Base case Total net primary energy usage rate (GJ/y) % Total net primary energy sourced from renewable Total net primary energy usage per kg product (kJ/kg) Total net primary energy usage per unit value added (kJ/$) Total raw materials used per kg product (kg/kg) Total raw materials used per unit value added Fraction of raw materials recycled within company Hazardous raw material per kg product Net water consumed per unit mass of product (kg/kg) Net water consumed per unit value added Value added ($) WAR Safety 215,844 1 7959 6.20 2.25 0.0017 0.01 1.13 38.00 0.03 4,352,296 43.23 28 Modied 218,228 1 7980 4.89 1.07 0.00065 0.02 1.07 30.17 0.02 5,579,857 22.69 28 % Improvement 1.1 0 0.3 21 52 62 100 5 21 33 28 48 0

recycle IPA and water in the waste water stream are implemented to the base process. This is done by modifying the owsheet followed by simulation of the new owsheet to generate mass and energy balance data. To identify a suitable separation process for the recovery of IPA and water, the separation synthesis methodology of Jaksland et al. (1996) is used at this stage. This methodology evaluates the properties of the mixture to be separated on the basis of its thermodynamic relations. The objective is to derive a set of feasible separation processes based on the equilibrium calculations and rank them in the order of preferences. Using the methodology, the list of possible separation technologies for IPA recovery can be identied as follows: liquid membrane, gas adsorption and pervaporation. The use of pervaporation process for recovering alcohol compounds from water has been studied by many researchers. The process involves contacting a liquid stream containing two or more miscible components with one side of a non-porous polymeric membrane (such as polydimethyl siloxane (PDMS)) or a molecularly porous inorganic membrane (such as zeolite) while applying a vacuum or gas purge in the other side to draw the permeate liquid and evaporate it. A suitable operating condition for IPA separation from water has been reported to be less than 30 C for 5.5 wt% IPA (Vane, 2005). Such operating range is considered suitable for implementing to this process. Fig. 9 shows the new owsheet incorporating pervaporator unit. In the owsheet, the wastewater stream from column T-102 is cooled to 28 C before being passed to a pervaporator unit (modeled by component splitter with a dummy energy stream). The vapour permeate (13 wt% IPA, 14 wt% acetone and 72 wt% water) is recovered and recycled to the feed stream. The liquid retentate obtained at 99 wt% water is recycled and mixed with fresh water as the washing liquid for T-100. To prevent a material build-up, a purge stream is provided. 3.4. Step 4sustainability evaluation and comparison In this step, the sustainability of the modied process is evaluated using SustainPro for comparison with the base case. This is done by calculating a set of performance metrics for the base and the modied process. In the proposed methodology, sustainability metrics from different sources including the sustainability metrics of Azapagic (2002), WAR algorithm (Cabezas et al., 1999) and the safety indices (Heikkil, 1999) are used as measures of economic, environmental, and safety impacts. The reader is referred to the respective papers for detailed explanation of the metrics. The new alternative is considered to be more sustainable if and only if it improves at least one of the metrics while ensuring negligible negative changes to all other metrics (in this case, the limit is set to be 1%). Table 8 shows that the new alternative improves all the economic and environmental indicator metrics while maintaining the other index of their base values. The recycling of IPA

and water results in an increase of 28% in prot and a decrease of 48% in the environmental impact. The raw material utilization metrics per unit of product and per value added are decreased by 52% and 62%, respectively. This is due to the recycle of IPA to the feed stream. The water recycling also leads to substantial decrease in the net water consumption (21% per kg of nal product and 33% per value added). On the other hand, the total energy consumption and the total energy consumption per unit product of the new process increases by 1.1% and 0.3%, respectively, due to the installation of the pervaporator system. However, this minor increase is considered acceptable. 3.5. Step 5variables optimization The previous modication step considers only structural changes to the owsheet to improve the sustainability of the process. In this step, operational changes in terms of variables modication are performed for process renement. To determine the decision variables for optimization, degree of freedom analysis is rst performed. Design degrees of freedom are the parameters or variables that must be specied in order to completely dene the design of the process (Luyben, 1996). Such parameters can be easily indicated from the steady-state simulation model of the process (i.e. process simulator). Table 9 lists the degrees of freedom of the modied acetone process (Fig. 9). The objective function in this case study is prot maximization which is described using the following
Table 9 Degree of freedom analysis of modied acetone process. Stream or unit Feed Fresh-water Process-water Feed-recycle Cooled-product Quenched-product Cooled-water Pervaporation-feed Permeate Retentate Permeate-recycle Compressed-permeate E-Pump100 E-Heater100 E-Reactor T-100 T-101 T-102 TEE-100 Pervaporator Degrees of freedom Temperature, pressure, owrate Temperature, pressure, owrate Temperature, pressure, owrate Temperature, pressure, owrate Temperature Temperature Temperature Pressure Temperature, pressure Pressure Temperature, pressure, owrate Pressure Heat ow Heat ow Heat ow Number of trays, pressure Number of trays, pressure, acetone fraction, vent rate, distillate rate Number of trays, pressure, reux ratio, condenser energy (E-Cond102) The owrate ratio of to-treatment-plant stream to bottoms-water Component split-ratio, temperature, pressure

I. Halim et al. / Computers and Chemical Engineering 35 (2011) 13431358

1353

Fig. 9. Structural modication of acetone owsheet.

expression: Prot = Product sales Raw material costs Treatment costs Energy costs (3)

followed by reverse approach analysis to determine the minimum and maximum ranges of these variables so as to keep the metric within the 1% constraint. This procedure is repeated for other metrics of interest to yield the minimum bounds of variables that satisfy all sustainability constraints.

Product sales are the revenues from the acetone product and byproduct-gas stream. Raw material costs constitute the costs of the feed and fresh-water stream. Treatment costs are the costs of treating the to-treatment-plant and waste-vapour stream. Energy costs consist of the following 12 energy variables: E-Pump100, E-Heater100, E-Reactor, E-Cooler101, E-Cooler102, E-Cond101, E-Reb101, E-Cond102, E-Reb102, E-Pump101, E-Cooler103, and EComp100. Comparison between these prot variables (Eq. (3)) and the degrees of freedom of the process (Table 9) reveals the following as the decision variables for optimization: fresh-water owrate, E- Pump100, E-Heater100, E-Cond102 (i.e. the condenser energy of T-102), and split-ratio of TEE-100. Once the decision variables have been identied, the subsequent step is to select the subset of optimization variables and nd the respective bounds for these variables. This step is done using SustainPro which screens the decision variables by prioritizing the ones that signicantly affect the process. The rationale behind this is that not all decision variables have equal impact on the process. The procedure for reducing the optimization search space is based on the indicator analysis of ISA algorithm (Carvalho et al., 2008). The algorithm applies reverse approach to screen variables with most impact to sustainability metrics (such as prot, environmental impact, water usage, etc.) and other operation constraints. The reader is referred to the respective paper for detailed description of the algorithm. Once the variables with the most impacts are found, the next step is to set the minimum and maximum ranges of these variables for optimization of the sustainability metrics. The procedure involves the following steps (see Fig. 10). First, a set of constraints for sustainability metrics of the process is xed. In our case, all of the sustainability metrics are set to be within 1% from the initial values. This means that a move to new variable values can only be allowed if (and only if) it does not worsen any of the metrics by more than 1%. Next, a mathematical expression describing the sustainability metric as function of these variables is dened. This is

Start Specify the decision variables as candidates for optimization ISA Algorithm for screening of process variables with most impact to the process Define sustainability metric of interest and write a mathematical expression describing it Perform reverse analysis by setting 1% constraint on the metric to yield the minmax ranges of variables

Is there other metric? No Select the minimum bounds of variables as the optimization space

Yes

Stop
Fig. 10. SustainPro procedure for reducing the optimization space.

1354

I. Halim et al. / Computers and Chemical Engineering 35 (2011) 13431358

Table 10 Comparison between the sustainability metrics for the modied and the optimized acetone process. Modied Total net primary energy usage rate (GJ/y) % Total net primary energy sourced from renewable Total net primary energy usage per kg product (kJ/kg) Total net primary energy usage per unit value added (kJ/$) Total raw materials used per kg product (kg/kg) Total raw materials used per unit value added Fraction of raw materials recycled within company Hazardous raw material per kg product Net water consumed per unit mass of product (kg/kg) Net water consumed per unit value added Value added ($) WAR Safety 218,228 1 7,980 4.89 1.07 0.00065 0.02 1.07 30.17 0.02 5,579,857 22.69 28 Optimized 214,530 1 7,845 4.80 1.07 0.00065 0.08 1.07 29.96 0.02 5,581,977 14.31 28 % Improvement 1.7 0 1.7 2 0 0 300 0 1 0 0.04 37 0

To illustrate the above procedure more clearly, consider a sustainability metric involving process prot as described in Eq. (3). Simulation using the decision variables reveals the impacts of fresh-water, E-Pump100 and E-Heater100 to be minimal. In that case, these three variables are set to their minimum possible values, leaving only the split-ratio of TEE-100 (or to-treatment-plant owrate) and E-Cond102 as the decision variables. SustainPro analysis on these two decision variables is next performed and reveals the following optimization ranges which limit the deterioration of sustainability metric by 1%: 918 kg/h to-treatment-plant owrate 1516 kg/h and 0.02 GJ/h E-Cond102 0.66 GJ/h. Simulation within these ranges is then carried out to ensure that owsheet convergence can be attained during the optimization run. Next, a single objective problem involving prot maximization is performed to these variables using a simulated annealing algorithm. This is done by capitalizing the simulationoptimization functionality of ENVOPExpert which uses HYSYS as the simulation engine and G2-ActiveXLink (Gensym Corp.) and the HYSYSBrowser (Excel-VBA) components as the communication bridge. Table 10 compares the sustainability metrics for the modied and the optimized acetone process. The value added (prot) of the optimized process increases by a small margin (0.04%). However, substantial improvement is observed in the environmental impact (37% decrease). Compared to the modied case, the amount of utilities of the optimized process also reduces. In this case, 1.7% reduction in the total energy consumption and 1% reduction in the total water usage can be observed. Overall, it can be concluded that the optimized process is more sustainable than the modied case. Compared to the superstructure-based mathematical optimization, our methodology offers an advantage in that the designer is not removed from the decision making process (Smith, 1995). In this case, the many intangibles in process design such as safety, which are difcult to include in the mathematical formulation, can be considered satisfactorily using our design synthesis approach. In fact, we have incorporated inherent safety metric as one indicator of the process sustainability. Another advantage of our approach is due to the fact that the traditional optimization methods would have to be based on multi-objective formulation to incorporate the process sustainability issues as well as the safety problems. This leads to a very difcult and complex problem, which would require the analysis of a Pareto front/surface. We have avoided such complexity with our proposed methodology. Further, while different design alternatives and variables manipulation can be simultaneously optimized using the superstructure approach, the resulting search space of the problem would increase exponentially with the number of alternatives and variables considered. This could result in long computational time or even failed convergence. Such problems are handled effectively using our reduced space optimization approach in which only variables that greatly impact the process are optimized. Moreover, this optimization is

done within the ranges which would improve the sustainability metric. 4. Case study: methanol and DME production from syngas In this section, we illustrate the application of the framework to methanol and DME production. The step-by-step procedure and their corresponding results are summarized below, starting with a description of the process. The interested reader can obtain the detailed analysis for this case study from the corresponding author. 4.1. Process (problem) description Methanol is one of the most important basic chemicals that can be produced using synthesis gas (syngas)a mixture of CO, CO2 and H2 . The typical methanol process involves a reaction of syngas in the presence of metal catalyst followed by purication of the methanol. Dimethyl ether (DME) is one of the chemicals that can be synthesized using methanol as feedstock. DME is used predominantly as a propellant and has the potential to be utilized as a clean alternative fuel such as diesel due to its high cetane number. Although a direct conversion of syngas to DME is already a commercially viable process (Kim, Kim, Cho, & Yoon, 2010), in this paper we consider a scenario of converting the excess methanol from one plant into dimethyl ether (DME) products. Fig. 11 shows the owsheet of the process. The upstream section of this process is a methanol synthesis plant described as follows. Initially, fresh syngas stream is combined with a recycled syngas stream and heated prior to being passed to a xed-bed reactor. The reaction, which takes place in the presence of metal catalyst (CuZnOAl2 O3 ), forms methanol according to the following reaction: CO + 2H2 CH3 OH (Graaf & Beenackers, 1996). The single-pass conversion of CO in the reactor is estimated to be 94%. As the reaction is highly exothermic, adiabatic quenched bed reactor is used. From the reactor, the efuent is cooled to ambient to condense the methanol product. The efuent is next sent to a phase separator (V-100) where the vapour stream is recycled back to the front end of the process. To prevent material build-up, a purge line is used. As the purge stream has low hydrogen purity (64% volume), it is then sent to the are system. From the separator, the liquid stream is passed to second phase separator (V-101) to further remove the non-condensible gases. The liquid stream from this separator which contains crude methanol is next sent to distillation column (T-102) which operates under partial condensation. Methanol product is obtained at 99.9% mole purity as liquid distillate while the water stream exiting from the bottom of the column becomes waste. The downstream part of the process is a DME plant which uses excess methanol as feedstock (Turton et al., 2003). From column T102, the vapour distillate which consists of 79% (mole) of methanol is combined with a recycled methanol stream and cooled to 10 C

I. Halim et al. / Computers and Chemical Engineering 35 (2011) 13431358

1355

Fig. 11. Flowsheet of methanol and DME production process.

in order to remove the non-condensible gas inside a ash separator (V-102). The liquid from V-102 is next vaporized, compressed and cooled for a dehydration process which proceeds as follows: 2CH3 OH (CH3 )2 O + H2 O. The reaction takes place in a xed-bed reactor over an acid zeolite catalyst. The single-pass methanol conversion in this reactor is assumed to be 80%. From the reactor, the efuent is cooled prior to being sent to a set of separation units. The rst unit is an adsorber which removes excess CO2 from the process. The second separation unit is a distillation column which recovers DME at 99.9% (mole) purity at the overhead. The third unit is a distillation column which separates water from the unused methanol. The unused methanol is next recycled back to the front end of the process to be mixed with the methanol reactant stream, while the water is sent to waste water treatment. 4.2. Application of the combined framework The step-by-step procedure employed to improve the sustainability of this process is described as follows. 4.2.1. Step 1information of the process Information about the process was obtained from a HYSYS simulation model in the form of HYSYS-XML. Here, conversion reactors were used as representation models for the methanol and DME reactor units. The economic data, material and stream classications and chemical environmental impacts were also required, as listed in Table 11. 4.2.2. Step 2alternatives generation Using ENVOPExpert and SustainPro, the following critical points of the process were identied (in the order of importance): (1) The low purity of H2 in the waste gas stream deems it unsuitable for reuse. Purication of this H2 stream can be considered.

Table 11 Cost and environmental impact data for methanol-DME process. Description Cost of raw material ($/kg) Syngas Price of product ($/kg) Methanol-product DME-product Cost of waste stream ($/kg) Wastegas Non-condensible Sour-gas To-treatment Cost of energy ($/kW h) E-Comp100 E-Comp101 E-Comp102 E-Heat100 E-Cool101 E-Cool102 E-Heat103 E-Cool104 E-Cool105 E-Adsorber E-MethReactor E-DMEReactor E-Cond100 E-Cond101 E-Cond102 E-Reb100 E-Reb101 E-Reb102 Environmental impact (per kg) Methane Methanol H2 O CO CO2 H2 DME Value 0.07 0.21 0.95 0.005 0.005 0.005 0.001 0.05 0.05 0.05 0.011 0.008 0.020 0.011 0.008 0.008 0.008 0.008 0.011 0.008 0.008 0.008 0.011 0.011 0.011 0.0157 0.4946 0 0.03476 3.593E04 0 0.462

1356

I. Halim et al. / Computers and Chemical Engineering 35 (2011) 13431358

Fig. 12. Modied methanol and DME owsheet.

(2) The CO2 circulating in the process also demands high amount of energy. This calls for removal from the syngas stream. (3) The presence of water circulating in the process leads to high energy wastage. This also calls for purication of the syngas stream. Hence, the following structural modications were proposed based on the above recommendations: installation of a separation unit to improve the H2 purity in the waste gas stream and purication of the syngas feed stream by removing CO2 and water components. 4.2.3. Step 3analysis of proposed alternatives To reduce the water content in the syngas stream, a condensation-separation process was proposed. On the other hand, membrane separation was considered for the removal of CO2 and purication of H2 . This membrane unit was identied following
Table 12 Sustainability metrics for the base case, modied and optimized methanol-DME process. Base case Value Total net primary energy usage rate (GJ/y) % Total net primary energy sourced from renewables Total net primary energy usage per kg product (kJ/kg) Total net primary energy usage per unit value added (kJ/$) Total raw materials used per kg product (kg/kg) Total raw materials used per unit value added Fraction of raw materials recycled within company Fraction of raw materials recycled from consumers Hazardous raw material per kg product Net water consumed per unit mass of product (kg/kg) Net water consumed per unit value added Value added ($) Prot ($) WAR Safety 4,503,571 0.92 12,731 6.70 2.28 0.0012 0.86 0 0 40.47 0.021 84,075,867 51,023,982 940.4 27

the separation synthesis methodology of Jaksland et al. (1996). As a matter of fact, the use of membrane process for CO2 removal from syngas has been commercially available (Membrane Technology & Research, 2010a). Using such technology, CO2 recovery up to 80% at 95% purity (volume) can be obtained. A membrane process for H2 purication has also been widely used in the reneries. The process has been claimed to be able to recover up to 95% of H2 at 99% purity (volume basis) (Membrane Technology & Research, 2010b). Hence, these alternatives were implemented in the HYSYS model of the base process (see Fig. 12) and a new simulation was carried out. 4.2.4. Step 4sustainability analysis of alternatives The new sustainability metrics of the modied process were calculated and compared with the base case. The new alternatives (modied process) were found to improve all the sustainability indicator metrics from their base values (see Table 12). In this case,

Modied case Value 3,925,194 0.92 11,107 5.58 2.13 0.00107 0.40 0 0 29.13 0.015 87,884,257 59,248,420 745.7 27 % Improvement 13 0 13 17 7 11 53 0 0 28 29 5 16 21 0

Optimized case Value 3,944,081 0.92 11,152 5.55 2.13 0.00106 0.40 0 0 29.36 0.0146 88,788,866 60,021,308 696.4 27 % Improvement 12 0 12 17 7 12 53 0 0 27 30 6 18 26 0

I. Halim et al. / Computers and Chemical Engineering 35 (2011) 13431358

1357

the prot increased by 16% while the energy, material and water metrics per kg of product improved by 13%, 7% and 28%, respectively. Further, the energy, material and water metrics per unit value added improved by 17%, 11% and 29%, respectively, while the environmental impact decreased by 21%. All these results indicate a more sustainable methanol-DME production process. 4.2.5. Step 5variables optimization To further improve the process, operational changes in terms of variables optimization were performed. The decision variables were found to be cooler energy E-Cool100, cooler energy E-Cool101, condenser energy E-Cond101, condenser energy E-Cond102 and reboiler energy E-Reb101. Reverse approach analysis using SustainPro on these variables revealed E-Cool101 to be the most important variable and its bound is [3 106 , 276 106 ] kJ/h. In this case, the objective function was prot maximization. Next, simulation of the cooler energy E-Cool101 within the bound was performed to ensure the owsheet convergence within this range. The simulation results showed that an increase in the cooler energy would lead to increased prot caused by higher DME throughput. However, this would be traded off with deterioration in the sustainability metrics due to the concomitant increase in the energy and cooling water requirement. Thus, the next step of the analysis was to reduce the optimization space further. In this case, the minimum and maximum range for E-Cool101 was found to be 2.07 108 cooling energy 2.09 108 kJ/h. This range is obtained by considering the maximum allowable prot loss (1%) and the constraints of the other sustainability metrics. The simulation convergence was also conrmed within this range. Table 12 also shows the sustainability metrics for the modied and optimized process. The maximum prot was obtained when operating at the maximum cooling energy (i.e. 2.09 108 kJ/h). Further, the table also shows an increased prot, a decreased environmental impact (WAR) and improved energy, material and water metrics per unit value added compared to the base-case value. The remaining performance indicators remain constant or show improvement as well. Overall, it can be concluded that the optimized methanol-DME case is the most sustainable among the three process options. 5. Conclusions Increasing resource constraints, public pressure to address negative environmental and social impacts and regulatory burdens over the plant operations call for sustainable products and processesones that are eco-friendly, contribute to societal welfare and equity and provide enduring economic value (Schwarz, Beloff, Beaver, & Tanzil, 2001). While computer tools are available to assist in the sustainability assessment of a process, their applications are limited to a specic domain of the design synthesis problem. Integrating these tools into one framework for simultaneous generation, analysis, evaluation and optimization of design alternatives is essential for ease of use and widespread adoption but poses a great challenge as each tool seeks to solve the sustainability problem from a different perspective. This paper presents a design synthesis methodology that is based on collaboration of two computer tools ENVOPExpert and SustainPro to derive comprehensive design alternatives for chemical processes. The heuristic based analysis of ENVOPExpert quickly identies broad-level design alternatives through its qualitative inference rules. The indicatorbased methodology of SustainPro highlights the most important bottlenecks in the process through the mass and energy indicators, identies the sustainable alternatives and screens them at an in-depth level. The reverse-design approach combined with sensitivity analysis of SustainPro also reduces the optimization

space by highlighting the critical variables that are candidates for manipulation. The minimum and maximum ranges for those variables can also be obtained by considering maximum allowable deterioration in sustainability metrics of 1%. The optimal manipulation is identied by ENVOPExpert using simulated annealing algorithm which uses the process simulator to evaluate the process objective. Depending on the nature of the problem, such problem can be cast as a single- or multi-objective optimization involving environmental and economic impact. The methodology has been applied to two case studies: an acetone process and a methanol and DME production from syngas. Another case studies which have been successfully solved are a hydrodealkylation (HDA) process for producing benzene (the result is available upon request) and an ammonia production process (Carvalho, Halim, Srinivasan, Matos, & Gani, 2009b). In all cases, the methodology has been shown to be capable of identifying the root causes of the sustainability problem, generating alternative designs in terms of structural and process variable modication and effectively reducing the optimization search space by screening the variables with most impact on the process. The results are improvement in the prot; energy, water, and material utilization; and environmental impact (i.e. more sustainable). References
Azapagic, A. (2002). Sustainable development progress metrics. IChemE sustainable development working group. Rugby: IChemE. Bakshi, B. R., & Fiksel, J. (2003). The quest for sustainability: Challenges for process systems engineering. AIChE Journal, 49, 13501358. Biegler, L. T., Grossmann, I. E., & Westerberg, A. W. (1999). Systematic methods of chemical process design. Upper Saddle River: Prentice Hall. Cabezas, H., Bare, J. C., & Mallick, S. K. (1999). Pollution prevention with chemical process simulators: The generalized waste reduction (WAR) algorithmfull version. Computers and Chemical Engineering, 23, 623634. Cano-Ruiz, J. A., & McRae, G. J. (1998). Environmentally conscious chemical process design. Annual Review of Energy and the Environment, 23, 499536. Carvalho, A., Matos, H. A., & Gani, R. (2008). Design of sustainable chemical processes: Systematic retrot analysis generation and evaluation of alternatives. Process Safety and Environmental Protection, 86, 328346. Carvalho, A., Matos, H. A., & Gani, R. (2009). Design of batch operations: Systematic methodology for generation and analysis of sustainable alternatives. Computers and Chemical Engineering, 33, 20752090. Carvalho, A., Halim, I., Srinivasan, R., Matos, H. A., & Gani, R. (2009). Sustainability analysis of chemical processes plants using a hybrid heuristic and indicator model. Computer Aided Chemical Engineering, 27, 837842. Chen, H., & Shonnard, D. (2004). Systematic framework for environmentally conscious chemical process design: Early and detailed design stages. Industrial Engineering Chemistry and Research, 43, 535552. DAnterroches, L., & Gani, R. (2005). Group contribution based owsheet synthesis, design and modelling. Fluid Phase Equilibria, 228(229), 141146. Dantus, M. M., & High, K. A. (1996). Economic evaluation for the retrot of chemical processes through waste minimization and process integration. Industrial Engineering Chemistry and Research, 35, 45664578. Douglas, J. M. (1992). Process synthesis for waste minimization. Industrial Engineering and Chemistry Research, 31, 238243. El-Halwagi, M. M. (1997). Pollution prevention through process integration. San Diego: Academic Press. Friedler, F., Varga, J. B., & Fan, L. T. (1994). Algorithmic approach to the integration of total owsheet synthesis and waste minimization. In M. M. El-Halwagi, & D. A. Petrides (Eds.), Pollution prevention via process and product modications. New York: American Institute of Chemical Engineers. Fu, Y., Diwekar, U. M., Young, D., & Cabezas, H. (2000). Process design for the environment: A multi-objective framework under uncertainty. Clean Products and Processes, 2, 92107. Graaf, G. H., & Beenackers, A. A. C. M. (1996). Comparison of two-phase and threephase methanol synthesis processes. Chemical Engineering and Processing, 35, 413427. Halim, I., & Srinivasan, R. (2002). Systematic waste minimization in chemical processes: Part I methodology. Industrial and Engineering Chemistry Research, 41, 196207. Halim, I., & Srinivasan, R. (2008). Designing sustainable alternatives for batch operations using an intelligent simulation-optimization framework. Chemical Engineering Research and Design, 86, 809822. Halim, I., & Srinivasan, R. (2011). A knowledge-based simulationoptimization framework and system for sustainable process operations. Computers and Chemical Engineering, 35, 92105. Harper, P. M., & Gani, R. (2000). A multi-step and multi-level approach for computer aided molecular design. Computers and Chemical Engineering, 24, 677683.

1358

I. Halim et al. / Computers and Chemical Engineering 35 (2011) 13431358 Martin, T., & Young, D. (2008). Waste reduction algorithm graphical user interface. Cincinnati: US Environmental Protection Agency. Membrane Technology and Research (2010a). CO2 removal from syngas. http://www.mtrinc.com/pdf print/renery and syngas/MTR Brochure CO2 Removal Syngas.pdf. Membrane Technology and Research (2010b). Hydrogen purication in reneries. http://www.mtrinc.com/hydrogen purication in reneries.html. National Research Council. (2005). Sustainability in the chemical industry: Grand challenges and research needsA workshop report. Washington, DC: National Academies Press. Pennington, D. W. (1999). Chemical process design using heuristics in the context of pollution prevention. Clean Products and Processes, 1, 170179. Petrides, D. P., Koulouris, A., & Lagonikos, P. T. (2002). The role of process simulation in pharmaceutical process development and product commercialization. Pharmaceutical Engineering, 22(1), 18. Schwarz, J. M., Beloff, B. R., Beaver, E. R., & Tanzil, D. (2001). Practical minimum energy requirements for chemical product manufacturing: A management decision tool for achieving sustainable products. Environmental Quality Management, 11, 7589. Smith, R. (1995). Chemical process design. New York: McGraw-Hill. Tsoka, C., Johns, W. R., Linke, P., & Kokossis, A. (2004). Towards sustainability and green chemical engineering: Tools and technology requirements. Green Chemistry, 6, 401406. Turton, R., Bailie, R. C., Whiting, W. B., & Shaeiwitz, J. A. (2003). Analysis, synthesis, and design of chemical processes. Upper Saddle River: Prentice Hall. Vane, L. M. (2005). A review of pervaporation for product recovery from biomass fermentation processes. Journal of Chemical Technology and Biotechnology, 80, 603629.

Heikkil, A. M. (1999). Inherent safety in process plant designAn index-based approach. Ph.D. thesis. VTT Industrial Systems. Hilaly, A. K., & Sikdar, S. K. (1994). Pollution balance: A new methodology for minimizing waste production in manufacturing processes. Journal of Air and Waste Management Association, 44, 13031308. Hyprotech (2003). HYSYS. Plant Version 3.1. Calgary: Hyprotech. ICCA (2009). Executive summary of the ICCA report: Innovations for greenhouse gas reductions. http://www.icca-chem.org/ICCADocs/ICCA A4 LR.pdf. (2007). Chemical industry struggles with public image. ICIS http://www.icis.com/Articles/2007/09/17/9062166/chemical-industrystruggles-with-public-image.html. Jaksland, C., Gani, R., & Lien, K. (1996). An integrated approach to process/product design and synthesis based on propertiesprocess relationship. Computers and Chemical Engineering, 20, 151156. Kim, I. H., Kim, S., Cho, W., & Yoon, E. S. (2010). Simulation of commercial dimethyl ether production plant. In S. Pierucci, & G. Buzzi Ferraris (Eds.), Proceedings of the 20th European symposium of computer aided process engineering: Escape-20. Elsevier: Oxford. (2010). Inorganic chemicals market and opportunities. KPMG http://ibef.org/download%5CInorganic Chemicals 170708.pdf. Lines, M. (2005). Views on key issues facing the chemical industry. In transforming sustainability strategy into action. In B. Beloff, M. Lines, & D. Tanzil (Eds.), Transforming sustainability strategy into actionThe chemical industry. New Jersey: John Wiley & Sons. Linnhoff, B. (1995). Pinch analysis in pollution prevention. In A. P. Rossiter (Ed.), Waste minimization through process design. New York: McGraw-Hill. Luyben, W. L. (1996). Design and control degrees of freedom. Industrial and Engineering Chemistry Research, 35, 22042214.