p r o c e s s s a f e t y a n d e n v i r o n m e n t p r o t e c t i o n 8 6 ( 2 0 0 8 ) 328346

available at www.sciencedirect.com

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

Design of sustainable chemical processes: Systematic retrot analysis generation and evaluation of alternatives
Ana Carvalho a,b , Raqul Gani a, , Henrique Matos b
a b

CAPEC, Department of Chemical Engineering, Technical University of Denmark, DK-2800 Lyngby, Denmark Instituto Superior T cnico, Av. Rovisco Pais, 1049-001 Lisboa, Portugal e

a r t i c l e
Article history:

i n f o

a b s t r a c t
The objective of this paper is to present a generic and systematic methodology for identifying the feasible retrot design alternatives of any chemical process. The methodology determines a set of mass and energy indicators from steady-state process data, establishes the operational and design targets, and through a sensitivity-based analysis, identies the design alternatives that can match a set of design targets. The signicance of this indicator-

Received 26 November 2007 Accepted 27 November 2007

Keywords: Process design Sustainability metrics Mass and energy indicators Safety index Indicator sensitivity algorithm

based method is that it is able to identify alternatives, where one or more performance criteria (factors) move in the same direction thereby eliminating the need to identify tradeoff-based solutions. These indicators are also able to reduce (where feasible) a set of safety indicators. An indicator sensitivity analysis algorithm has been added to the methodology to dene design targets and to generate sustainable process alternatives. A computeraided tool has been developed to facilitate the calculations needed for the application of the methodology. The application of the indicator-based methodology and the developed software are highlighted through a process owsheet for the production of vinyl chlorine monomer (VCM). 2008 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.

1.

Introduction

Increased industrial activities combined with new economic, environmental and societal constrains means that energy consumption, raw materials depletion and environmental impact are also receiving increased attention by modern society. Due to these factors, new as well as old industries need to achieve balance between the negative impacts from their activity and the positive benets (due to the products they provide) to the society. In order to achieve a sustainable development, much progress is needed for the identication, design and development of appropriate products and processes that will produce them. For the industry to remain competitive and to adapt to the new realities of globalization, process improvements with respect to waste, environment, safety and cost are necessary. Therefore, it is useful to develop systematic methods and tools, which enable the generation of new more sus-

tainable alternatives as well as ability to adapt to the future needs. In the past decades different methodologies have been proposed in order to determine the retrot potential of a chemical process with respect to improvement of the cost-efciency. Some methodologies were introduced using heuristic rules for the generation of the new design alternatives (Rapoport et al., 1994). Other methodologies based on mathematical concepts and optimization methods, such as mixed integer non-linear programming (MINLP) have also been proposed by Ciric and Floudas (1989) and Jackson and Grossmann (2002). Methodologies based on the resynthesize of the entire process by incorporating operating units with enhanced performance have been proposed by Liu et al. (2006). Finally, Lange (2002) has proposed a methodology that directly relates the process design alternatives to improvements in sustainability of the processes. This author claims that the optimal solution is a trade-off between the different performance criteria, and

Corresponding author at: CAPEC, Department of Chemical Engineering, Technical University of Denmark, DK-2800 Lyngby, Denmark. E-mail address: rag@kt.dtu.dk (R. Gani). 0957-5820/$ see front matter 2008 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved. doi:10.1016/j.psep.2007.11.003

p r o c e s s s a f e t y a n d e n v i r o n m e n t p r o t e c t i o n 8 6 ( 2 0 0 8 ) 328346

329

Nomenclature A AF CA Cf Cp DC ebl E EAF EC EN EWC f allocation factor accumulation factor cost allocation factor coefcients in the Fobj heat capacity demand cost energy base level parameter denoting the effect of a path compound on a reaction energy accumulation factor energy cost total number of leaving streams from a closepath excluding the demand streams energy and waste cost minimum owrate of compound c in the close-path cp within the streams that belong exclusively to that cycle (si) closed-path fraction relative to the other cycles number of desired products in the process objective function compound heat of combustion heat of vaporization of water total number of such vertices increment in the ISA scores scale inside battery limit area total inherent safety index chemical inherent safety index corrosivity index equipment index explosiveness index ammability index inventory index chemical interaction index process pressure index process inherent safety index heat of the main reaction index heat of the side reactions index process structure index process temperature index toxicity index number of closed-paths belonging to a given stream total number of paths selected in ISA algorithm owrate of the compound c in path ow op/cp molecular weights material value added total number of open-paths offside battery limit area pressure total number of products in the overall reaction unit price of the utility value outside the process boundaries price of a given compound purchase price prot term heat exchange in a given unit total number of reactive unit operations total number of raw materials involved in the overall reaction

RQ S SS Su SE SNL T TDC TVA TFobj U UEL UK V VA VI VT W WAR WC (c) WACj WAMj x y

(c) (c) WAVj

fr FP Fobj Hc Hvap I In ISBL ISI Ici Icor Ieq Iex I Ii Iint Ip Ipi Irm Irs Ist It Itox j K m M MVA OP OSBL p PD PE PP Pr PR PT Q R RM

reaction quality scores based on sensitivity of variables that inuence Fobj total number of supplies supply ows in the process graph stream belonging exclusively to one given closed-path, CP stream common to nested loops temperature total demand cost total value added terms of the objective function total number of sub-operations upper explosive limit total number of compound path ows in a suboperation volumetric owrate total number of variables inuencing Fobj variable that inuence the indicators variable that inuence the objective function terms scores based on the weights of variables that inuence Fobj waste reduction algorithm waste cost concentration allocation factor mass allocation factor volume specic allocation factor fraction of the owrate of path (k) by the total amount of a given variable fraction of a given term from the Fobj in the sum of all the terms related with the selected variables in ISA fraction of the coefcient Cf for the variable n, in the sum of all coefcients for all the variables selected in ISA

Greek letters stoichiometric coefcients extent of reaction density Subscripts or superscripts c compound index cp closed-path index d demand ows in the process graph ec index of the closed energy path ows fp desired product i index of vertices k path index n variable index r index of the reactive unit operations rk reaction index rm raw material index s stream index si stream exclusively belonging to one cycle op open-path index pd product index u sub-operations index uk index of all compound path ows in u va variable in study

330

p r o c e s s s a f e t y a n d e n v i r o n m e n t p r o t e c t i o n 8 6 ( 2 0 0 8 ) 328346

Fig. 1 Flow-diagram of the indicator-based methodology.

therefore, the improvement of sustainability in a process is a balancing of the three areas that most inuence it, such as, the environmental, the society and the economical aspects of the process. Recently, an indicator-based methodology was proposed by Uerdingen et al. (2003, 2005). This methodology is based on mass and energy indicators, which are able to identify and screen processes operating in the continuous mode. The indicators determine the critical points in the process and based on their values new (feasible) design alternatives are generated. These feasible alternatives are compared in terms of economic aspects that help to select the best alternative. Jensen et al. (2003) further extended this methodology where the previously dened indicators were retained but the choice of the best alternative was obtained using new parameters related to economic, safety and environmental factors. The sustainability metrics dened by Azapagic (2002), the safety indices developed by Heikkila (1999) and the waste reduction, algorithm, WAR, proposed by Cabezas et al. (1999) were incorporated in the new version as process performance criteria. The objective of this paper is to present a new generic and systematic methodology, which combines the previous versions, incorporates new features, as well as extend the application range to a while spectrum of chemical process. This is achieved through improvements in the owsheet decomposition technique identication, selection of target indicators as well as generation of new feasible alternatives. The new methodology is also able to avoid the typical trade-off between competing design decisions. That is, the environmental impact is reduced because less mass is released (in terms of emission) and because less energy is used, the waste and operational cost are reduced. Finally, the indicators also reduce (where feasible) the safety indicators indices by making the conditions of operation more exible. A sensitivity analysis algorithm, which is able to determine the design targets for a process has been developed together with a method to generate the design alternatives. The method to generate the new design alternatives combines three different algorithms for process-product synthesis Jaksland et al. (1996), Harper and Gani (2000) and DAnterroches and Gani (2005) into one integrated method. The systematic methodology has been implemented into an EXCEL-based software called Sustain-Pro.

2.

Methodology

The work-ow for the new extended methodology is organized in terms of six steps and is highlighted in Fig. 1. In the text below, each step is explained. Step 1: Collect steady-state data The objective of this step is to collect the mass and energy data corresponding to the steady-state operation of the process under investigation. This data can be provided as steady-state simulation results (for example, from commercial simulators like Pro II or Aspen) or operational data collected from the plant. Therefore, the methodology can be applied to any chemical process, for which steady-state operational data is available. Step 2: Flowsheet decomposition The objective of this step is to identify all the mass and energy ow-paths in the process by decomposing into openand closed-paths for each compound in the process. The closed-paths (cp) are the process recycles with respect to each compound in the process. In other words, they are the ow-paths which start and end in the same unit of the process. An open-path (op) consists of an entrance and an exit of a specic compound in the process. The entrance of the compound in the system can be due to its entrance through a feed stream or by its production in a reactor unit. The exit of the respective compound can be due to a demand (exit) stream or by its reaction in a reactor unit. Fig. 2 highlights the main steps of the owsheet decomposition algorithm.

Fig. 2 Main steps for the owsheet decomposition algorithm.

p r o c e s s s a f e t y a n d e n v i r o n m e n t p r o t e c t i o n 8 6 ( 2 0 0 8 ) 328346

331

Step 2.1: Flowsheet transformation into a process graph In this step, the owsheet is transformed into a process graph, where the units are represented as vertexes and while the edges joining two vertexes are the streams. The edges are represented by arrows in order to give a direction to the owrate sequence across the system. All the feed and demand (exit) streams are represented by bold arrows, including the compounds which enter and exit the system by reaction. More details about the transformation of the owsheet into a process graph are given in Uerdingen et al. (2003). Step 2.2: Determining the closed-paths Before identifying the closed-paths within the owsheet it is necessary to decompose the owsheet into different partitions. The Sargent and Westerberg algorithm, Sargent and Westerberg (1964), is implemented here. With all the partitions identied, it is possible to determine the closed-paths within each partition using the algorithm described in Biegler et al. (1997). Step 2.3: Determining the closed-paths owrate and determine the open-paths The owrates associated to closed-paths, needs to be identied before the identication of the open-paths and their associated owrates. This is because the mass accumulated in the cycles needs to be subtracted from the corresponding open-path streams. This allows dening a new graph only with the compounds amounts related with entrance and the exit of a given compound. The algorithm employed is highlighted in Fig. 3 and the explanation of each step of the algorithm is given as follows: Step A: The rst step of the new algorithm is to select the compound to analyze. This step needs to be repeated for all

compounds whose accumulation is different in each closedpath. Step B: Here a closed-path should be selected. All the closedpath should be analyzed for each compound. Step C: The stream, within the selected cycle, which corresponds to the minimum compound owrate, needs to be located. This value represents the amount of material, which is being recycled. Step D: Here it should be analyzed if the selected stream is belonging only to the closed-path in study or to more closed-paths. Depending on the result obtained by the previous analysis the way of calculating the closed-path owrate will vary. Step E: If this stream only belongs to one nested loop it is clear that this is the accumulated amount of the compound in that cycle and consequently this value is the compound owrate in the closed-path in study. Step F: If the minimum compound owrate appears in a stream that is common to more than one nested loop (stream called SNL ), then it is necessary to nd a way to quantify the percentage of the compound owrate that belongs to each closed-path.

(1) Here all the closed-paths that belong to the stream SNL (found in Step C) need to be determined. For those closedpaths, some procedures should be executed (described in the following steps). (2) For each closed-path determined in Step F-1 it is necessary to nd the streams which belong exclusively to them. This means streams, which are not belonging to other nested loops. (3) From the streams determined in Step F-2, the stream presenting the lowest owrate should be determined. This

Fig. 3 Schematic diagram of the algorithm for Steps 2.3 and 2.4.

332

p r o c e s s s a f e t y a n d e n v i r o n m e n t p r o t e c t i o n 8 6 ( 2 0 0 8 ) 328346

stream will be called SE . This procedure will be done for all the closed-paths passing in SNL , so there will be a SE stream for each closed-path determined in Step F-1. (4) Here the fraction of each closed-path, determined in Step F-1, relatively to the remaining closed-paths passing through SNL , is calculated using the following equation: fcp,S
(c)

Table 1 Mass balance for component A in owsheet shown in Fig. 4

Stream 1 2 3 4 5 6 7 8 9 10 Mass A (kg/h) 5 25 10 25 20 30 15 15 30 5

frcp =

(c)

j (c) f cp=0 cpj,SE (c)

(1)

where frcp is the fraction of the closed-path cp in the total of the owrates of the independent streams, SE , for (c) compound c. fcp,S represents the minimum owrate of E compound c in the close-path cp within the streams that belong exclusively to that cycle (SE ) and j is the number of closed-path which belong to the stream SNL . With this fraction it is possible to break the nested loops, determining the mass in SNL stream that corresponds to each closed-path. Without this approach negative owrates would appear when the accumulated mass would be discounted from the streams remaining to the open-paths. (5) The owrate for each closed-path passing in stream SNL should be determined by the multiplication of the owrate of stream SNL times the fraction of each closed-path calculated Step F-4. The expression used to calculate the owrate is the following one:
(c) mcp

only possible if the fraction of each closed-path in that stream is known. For this example the necessary data are the following: Stream belonging to more than one nested loop (SNL ): stream 3. Streams belonging exclusively to each cycle (SE )CP1: (4, 5); CP2: (7, 8). Stream with the minimum owrate, fcp,S , in the domain of E SE streamsCP1: S5 ; CP2: S7 . With the above data and applying Eq. (1) it is possible to calculate the fraction of each closed-path: fC1,S
(A) (c)

(c),cp fsNL (c)

(c) frcp

(2)

frC1 =

(A)

2 (A) f cp=0 cpj,si

20 = 0.57, 20 + 15 15 = 0.43 20 + 15

where mcp is the owrate for compound c in the closedpath, cp, which pass through
(c) (c),cp SNL . fsNL

is the owrate of

frC2 =

(A)

fC2,S

(A)

stream SNL and frcp is the fraction of the closed-path cp in the total of the owrates of the independent streams (see Eq. (1)). Step G: Once the owrate of the closed-path in study is already known (Step E or F-5) the accumulation in that closed-path is also known. Consequently, the stream determined in Step C should be removed from the owsheet in order to allow the determination of the open-paths. The closed-path owrate should be subtracted from all the streams belonging to the cycle, in order to leave only the ow that is entering and leaving the process and not the accumulated value. Step H: Here the entire open-paths related with that compound are going to be determined. Step I: In this step, it is checked if all the closed-paths have been analyzed. If not, Step B and the follows need to be executed again. Step J: Finally it is veried if all the compounds have been analyzed. If not, Step A and the follows need to be executed again. A short example Consider the owsheet represented in Fig. 4 and the respective mass balance for component A in Table 1. In this owsheet the owrate of stream 3 (f3 ) is the minimum owrate for closed-paths CP1: [378] and CP2: [2345]. Since f3 owrate represents both closed-paths, the total amount of f3 cannot be subtracted from both cycles, because stream 3 is a SNL stream. The owrate for each cycle needs to be determined, but this is

2 (A) f cp=0 cpj,si

With the fractions of each closed-path known, the closed-path owrates can now be calculated through Eq. (2): mC1 = fS mC2 = fS
(A) (A) (A),C1
3

frC1 = 10 0.57 = 5.7, frC2 = 10 0.43 = 4.3

(A)

(A)

(A),C2
3

The closed-path owrate of cycles C1 and C2 will be subtracted to the owrates of compound A in streams [245] and [78], respectively to each cycle. Removing stream 3, both cycles become open and the open-paths can be determined. Step 2.4: Determining the owrate of open-paths The open-paths owrates are calculated using the rules presented in Uerdingen et al. (2003), who also recommends the calculation of the open-path owrate as the multiplication of the owrate for the feed stream by the distribution factors across the path. The distribution factors are given by

Fig. 4 Process owsheet with nested loops.

p r o c e s s s a f e t y a n d e n v i r o n m e n t p r o t e c t i o n 8 6 ( 2 0 0 8 ) 328346

333

the coefcient of the exit stream that belongs to a given unit by the sum of the supply streams in that unit. Step 3: Calculation of indicators, sustainability metrics and safety index In this step the mass/energy indicators, the safety index and the sustainability metrics are calculated. Step 3.1: Calculate mass and energy indicators There are ve mass indicators and three energy indicators, Uerdingen et al. (2003). A brief description about these indicators is going to be given in the following paragraphs. 1. Material value added (MVA) This indicator gives the value added between the entrance and the exit of a given compound. In other words, this means the value generated between the start and the end point of the path. Consequently, this indicator is only applied to open-paths. To calculate this indicator it is necessary to know the purchase price or the costs related with the production of a given compound as well as its sale price. For the open-paths with compounds, which are not, involved in reactions the following equation is to be used for the indicator calculation: MVAop = mop (PPc PRc ) op op
(c) (c) (c)

consequently point to potential for improvements. The indicator value is given in terms of monetary units per year. 2. Energy and waste cost (EWC) This indicator is applied to both open- and closed-paths. It takes into account the energy costs, EC, and the costs related with the compound treatment, WC. The value of EWC represents the maximum theoretical amount of energy that can be saved in each path within the process. The expression to calculate this indicator is the following one: EWCk = ECk + WCk
(c) (c) (c)

(7)

The rst term of Eq. (7) is obtain from

U

ECk =
u=1

(c)

PEu Qu

mk Au,k (Tm , pm )
UK m A (T , p ) uk=1 m,uk u,uk m m

(c)

(c)

(8)

(3)

where mop is the owrate of the compound c in openpath op, PPc and PRc represent the sale price and the op op purchase price, respectively. When the open-path is related to a compound produced in the system the production price needs to be estimated from the raw material price. The equation to calculate the indicator is the following:
RM

MVAop = mop

(c)

(c)

PPop CAop
rm=1

(c)

(c)

(rm) (rm) op |M (c) (c) op M

PRop

(rm)

(4) where rm is the raw material index and RM is the total number of raw materials involved in the overall reaction equation, M(rm) are the molecular weights of raw materials, the (rm) are the stoichiometric coefcients from the reaction equation, the purchase price PR(rm) of raw materials, CA is the allocation factor. The allocation factor can be calculated using the following expression, if and only if one of the products generated in the overall reaction equation represents a value outside the process boundaries: CAop =
(c)

where k is the open- or the closed-path depending to the type of path that the indicator is being applied, u is the sub-operations index and uk is the index of all compound path ows in u. U represents the total number of sub-operations and UK the total number of compound path ows in a sub-operation. PEu represents the specic price of the utility needed in sub-operation u and Qu is the energy consumption in the same sub-operation. Au is an allocation factor calculated at the average temperature Tm and pressure pm and it is normally a physical property which depends of the unit operation. As a rule of thumb the properties used for each unit operations are Heat capacity, Cp : Heat exchangers (non-evaporators). Heat of vaporization, Hvap : Heat exchangers (evaporative and condensing units). Density, : Compressors and pumps. The second term of Eq. (7) involves the calculation of the waste cost, which is needed only for open-paths through which a compound leaves the process for waste treatment. It can be expressed as follows: WC = mass waste treatment cost (9)

The waste treatment costs often vary depending on the waste composition, total volume or mass. It is given by

WCs =mk

(c)

(c)

1 (c) ( s ) WAVs (volume allocation)

WAMs (Vs )
1 (c)

(mass allocation) (allocation by concentration) (10)

(c) WACs

PPop
PD PP(pd) pd=1

(c)

(5)

where pd denotes a product and PD is the total number of products in the overall reaction equation. Otherwise the allocation factor has to be determined using:
(c) Mop PD (pd) M pd=1 op

CAop =

(c)

(6)

Negative values of this indicator show that the component has lost its value in this open-path and

where s is the density and Vs is the volumetric owrate of the process stream s leaving the process boundaries, (c) WAVs is the volume specic allocation factor for com(c) pound c, WAMs is the mass specic allocation factor (c) and WACs is the concentration allocation factor for compound c. High values of this indicator show high consumption of energy and waste costs and consequently these paths should be considered in order to reduce the indicator value. This can be done in two ways, reducing the path owrate or changing the conditions in order to decrease the duties. The EWC value is given in terms of monetary units per year.

334

p r o c e s s s a f e t y a n d e n v i r o n m e n t p r o t e c t i o n 8 6 ( 2 0 0 8 ) 328346

3. Reaction quality (RQ) This indicator measures the inuence of a given path with respected to process productivity. This indicator is applied to both closed- and open-paths. Positive values of RQ show a benet of this path with respect to process productivity, while negative values show a decrease in the process productivity. The path where negative values appear is therefore a good target for improvement. The value of RQ is obtained by the following expression:
R RK (c) r,rk,k Er,rk,k FP n(fp) fp=1

RQk =
r=1 rk=1

(c)

EWC. Therefore, even if the TVA value does not present a very negative value, the values of EWC and MVA should be analyzed separately in order to conrm that they really do not show any problem. 6. Energy accumulation factor (EAF) This indicator determines the accumulative behaviour of energy in the closed-paths. Low values of this indicator highlight the potential for saving energy consumption in the system. The EAF can be calculated by using the following expression: EAF = eblec
I i=1 EN h f k=1 i,k

(11) +
OP dh op=1 i,op

(14)

where r,rk,k is the extent (in kmol/h) of reaction rk, n(fp) is the molar owrate of the nal desired product fp leaving the process boundaries, FP represents the total number of desired products in the process, r is the index of the reactive unit operations, R is the total number of reactive unit operations in the path k, rk is the index of the reactions and RK is the total number of reactions in the unit operation affected by the path k. (c) The parameter Er,rk,k is dependent on the effect of a given compound in the reactions involved in the path ow. The value can be determined using the following rule of thumb:

where I is the total number of the vertices encountered in the energy closed-path ow, and i is the index of these. ec is the index of the cycle energy path ows. ebl is the energy base level, i.e. the amount of energy recycled in the particular recycling. 7. Total demand cost (TDC) This indicator is applied only to open-paths and traces the energy ows across the process. For each demand in the process the sum of all DC, which pass through it,

Er,rk,k =

(c)

+1, If component c in k affects reaction rk such that the total productivity of the desired product n(fp) is favored
0, If component c in k has no effect on reaction rk and

total productivity of the desired product n(fp) is unchanged 1, If component c in k affects reaction rk such that the (fp)
total productivity of the desired product n is inhibited are calculated. DC can be calculated using the following equation: DCSu,d = PESu EOPSu,d (15)

4. Accumulation factor (AF) This indicator determines the accumulative behaviour of the compounds in the closed-paths. This corresponds to the amount that is recycled relative to the input to the process and not the inventory. High values of this indicator show high potentials for improvements. The expression representing this indicator is described by mcp
I i=1 (c) EN (c) f k=1 i,k (c)

where PE is the utility cost, in units of price/energy. The total cost for all the paths is expressed by
SS

AF =

OP (c) d op=1 i,op

(12)

TDCd =

DCSu,d
Su=1

(16)

where mcp is the owrate of compound c in the closedpath, cp, fi,k and di,op are the compound ows leaving the cycle ow and i is the index of vertices in the cycle from where the compound leaves, I is the total number of such vertices. 5. Total value added (TVA) This indicator describes the economic inuence of a compound in a given path. TVA joins two of the previous indicators EWC and MVA and the following expression determines its value: TVA = MVA EWC (13)
(c) (c)

Negative values of this indicator show high potential for improvements in terms of decrease in the variable costs. The values of this indicator should be analyzed carefully, due to the fact that MVA can have a high positive value and consequently hide the problems in

where SS is the total number of supplies that energy contributes are signicant to the demand, d. High values of this indicator identify the demands that consume the largest values of energy, so these are the process parts, which are more adapted to heat integration. Step 3.2: Calculate safety indices The safety of the process is another important parameter that should be taken into account. In this project the inherently safety index, developed by Heikkila (1999), has been implemented. In order to achieve the inherently safety index the value for some sub-indices need to be calculated. These sub-indices can be divided into two groups, one group, which takes into account the chemical inherent safety, and the other group that is dependent on the pro cess inherent safety. Heikkila (1999) dened a scale of scores for each sub-index. These scales are based on the values of some safety parameters, such as the explosiveness, the toxicity, the pressure of the process and so on.

p r o c e s s s a f e t y a n d e n v i r o n m e n t p r o t e c t i o n 8 6 ( 2 0 0 8 ) 328346

335

Table 2 List of safety indices and their scores

Score Total inherent safety index (ISI) Chemical inherent safety index, Ici Sub-indices for reactions hazards Heat of the main reaction, Irm Heat of the side reactions, Irs Chemical interactions, Iint Sub-indices for hazards substances Flammability, I Explosiveness, Iex Toxicity, Itox Corrosivity, Icor Maximum, Ici score Process inherent safety index, Ipi Sub-indices for process conditions Inventory, Ii Temperature, IT Pressure, IP Sub-indices for process conditions Equipment, Ieq IISBL IOSBL Process structure, Ist Maximum, Ipi score Maximum, ISi score

04 04 04

04 04 06 02 28

05 04 04

that either the impact of the process is less or the output of the process is more. For the environmental impact related metrics (see Fig. 5), instead of using the denition of Azapagic (2002); the definition proposed by Cabezas et al. (1999) have been used. Cabezas et al. proposed the WAR algorithm in order to calculate the environmental impacts from a chemical process. This algorithm has been implemented as part of the indicator-based methodology. To calculate these metrics, the owrates for each compound coming into the process and leaving the process are needed as known information. Summarizing, the indicators are applied to the entire set of open- and closed-paths. With their values the critical points of the process as well as the areas that should be improved in the process are determined. The sustainability metrics and the safety index are calculated using the steady-state data for the global process and they are used to measure the impact of the process in its surroundings. They will be used as performance criteria in the evaluation of the new suggested design alternatives.

04 03 05 25 53

The sum of all the sub-indices scores is the inherent safety index value; this parameter has the maximum value of 53. Note that the higher is the inherent safety index value the more unsafely is the process, so the aim in all the design alternatives is to try to reduce its value as much as possible. In Table 2 the entire set of sub-indices, as well as the respective scales, are specied. Step 3.3: Calculate sustainability metrics In this methodology the sustainability metrics dened by the Institution of Chemical Engineers by Azapagic (2002) have been used. The author has dened 49 metrics divided into three main areas: environmental, social and economical. Out of the 49 dened metrics, in this work we have used 23 of them. The sub-areas related to these metrics are highlighted in Fig. 5, for each sub-area, more than one metric is calculated. The use of the sustainability metrics follows the simple rule that the lower the value of the metric the more effective the process. A lower value of the metric indicates

Step 4: Indicator sensitivity analysis algorithm (ISA)target determination An algorithm to identify the design targets and to determine their desired target values has been developed for this step. With this algorithm it is possible to identify the indicators that have the potential to make signicant improvements in the process. These indicators are then designated as targets (design) for generation of new process alternatives. Since the variables which dene the indicators are the same that dene the process alternatives, it is also possible to simultaneously determine (without additional simulation or optimization) how much of the targeted process improvements can be actually achieved. The main steps of ISA algorithm are described in the diagram present in Fig. 6. Step 4-a: Analysis of indicators The objective here is to identify and select the indicators that show the highest potential for improving the process. That is, locate the critical points in the process owsheet with respect to potential design-operational deciencies. This determination is done comparing the indicators values and selecting the ones presenting the highest potential for improvements. Changing the process in located critical areas corresponds to sustainability metrics and safety indices guaranteed improvements in the related.

Fig. 5 Example of the sustainability metrics proposed by Azapagic (2002).

336

p r o c e s s s a f e t y a n d e n v i r o n m e n t p r o t e c t i o n 8 6 ( 2 0 0 8 ) 328346

Fig. 6 Main steps for the ISA algorithm.

As can be seen in Fig. 6, the indicators are related with a given path, P#, and they are depending on variables, VI (variables inuencing indicators). Step 4-b: Denition of the objective function The goal here is to derive an expression, which denes the criteria for process (performance) improvement objective. This equation should be a function of process-manipulated variables and process operational parameters such as the cost of materials and utilities. The general expression for an objective function (Fobj ) is given by Fobj = Cf1 VF1 + Cf2 VF2 + + Cfn VFn (17)

where, Cf are the coefcients, which multiply by the variables VF (variables inuencing Fobj ). Variables VF are taken from the mass and the energy balance (owrates or exchanged energies). Consequently it is possible to see that these variables are a function of some other variables (VF = f (operational variables)). This means that despite of the linearity shown in Eq. (17), the objective function is dependent on variables that were calculated through models that can be non-linear. Eq. (17) provides a exible denition of Fobj for different processes. This function can be divided in terms, TFobj , and each term is determined by Cf VF . For example, a process that produces large quantities of a desired product, Fobj could improve the prot margin while simultaneously improving waste reduction potential (and therefore, also the environmental impact) but keeping all other metrics and indices constants or within insignicant changes. In another case, such as a pharmaceutical product

manufacturing, where the process operational cost may be less important than, for example, the reliability of the product quality, the environmental impact and the time needed for production. Step 4-c: Incidence matrix analysis This analysis identies the variables VF that also inuence the selected set of indicators (Step 4-a) through an incidence matrix, where each row represents an indicator (model) equation and each column represents the variables VF . Step 4-d: Selection of target variables From the analysis of the incidence matrix of Step 4-c, identify the common set of variables that belong to the Fobj expression as well as the selected indicators. These variables will determine by how much the targeted value can be attained. The target variables, VT , are all the variables that satisfy the following condition: VI = VF These variables are the ones that are going to be modied in order to improve the target indicators (that is, match the target values); consequently these are the variables that are going to inuence our objective function. Step 4-e: Analysis of target variables The objective in the step is to identify the variables that will effect the largest changes in the objective function for the smallest changes in its values. This is done by analyzing the effect of each variable on the objective function and studying their sensitivity. Through this analysis, it is possible to identify the set of target variables that makes the largest impact on the objective function, and therefore, in the tar-

p r o c e s s s a f e t y a n d e n v i r o n m e n t p r o t e c t i o n 8 6 ( 2 0 0 8 ) 328346

337

get indicator. The following steps go through the problem in more detail. I. The aim here is to see the inuence of a change in target variable, VT , in the Fobj . These VT are not going to inuence the variables VF in their total amount, in this way it is necessary to dene VT as a function of P# for each indicator (VT = f (P#)). This means that just the fraction of the total amount of that variable that is related to each path can be changed to improve Fobj . In order to see the fraction of each VT that is going to be inuenced by each path, a owrate stream that represents that variable is going to it. This means that, if VT is already a owrate, this owrate will be the representative owrate of that variable. However if for example the variable VT is a heat duty exchange in a unit, the owrate of the leaving stream, coming out from the unit in study, is the representative owrate for variable VT . Once all the variables VT have a owrate associated to them, it is possible to see which inuence has the path, P#, in the variable VT . The expression to determine that fraction is the following one: xn,k = mk mn (18)

Table 3 Ranges and scores for the terms in the Fobj function
Weight Inferior limit Superior limit Scores

Weight in the prot expression Extremely high Min y + 4 InW High Min y + 3 InW Normal Min y + 2 InW Low Min y + InW Extremely low Min y

Max y Min y + 4 InW Min y + 3 InW Min y + 2 InW Min y + InW

5 4 3 2 1

changed, is determined using the following equation: y(n,k) = TFobj (n,k)

K k VA T va Fobj (va,k)

(20)

where xn,k is the fraction of the path (k) owrate by the owrate associated to a target variable, VT . mk is the path owrate and mn is the total owrate associated to the variable n which is being studied. When the variable in study is not a owrate than mk is assumed to be represented by the outlet owrate of the unit. When the unit has more than one outlet stream the outlet stream that is in the path need to be considered. II. In this step, for each target variable, VT , selected in Step 4-d, the terms of the objective function, TFobj , that are related to it will be determined. This means if the study is being done in variable VT1 the correspondent prot term it will be given by TVT1 = CfVT1 VT1 . The terms values represent the total amount of Fobj that is inuenced by a target variable; however here the objective is to determine which part of that term, T, is inuenced by the paths P# (see Step 4-a). In order to calculate the value of the term that is inuenced by a given variable in a given path, the term inuenced by that variable, should be multiplied by the fractions determined above. This is done using Eq. (18): T Fobj (n,k) = x(k) TFobj (n) (19)

where TFobj is the prot term and y is the fraction of a given term in the sum of all the terms related with the selected variables. va is the variable in study and VA is the number of variables selected in Step d. k is the path in study and K is the number of paths selected in Step a. With this factions, y, it is possible to dene which are the variables within a given path that can achieve a higher improvement in Fobj . III. Identify the most sensitive variables by determining the coefcients, which are multiplied by the variables VT , and then calculate their percentage in the sum of all coefcients. These coefcients are collected from Eq. (18), Cf. The equation used to determine the relative sensitivity between the Cf associated to each variable VT is the following one: zn = Cf(n) VA Cf va=0 (va) (21)

z is the fraction of the coefcient Cf for the variable n, in the sum of all coefcients related to all the variables selected in Step d. Trough this analysis is possible to conclude which are the variables that with smaller modications are going to achieve higher improvements in the objective function. Step 4-f: Scores scales determination Establish limiting values of the selected variables, based on a relation between percentage changes in the variables against improvements achieved, on a scale of 05. The increment is calculated trough: In(W) = In(S) = Max y Min y 5 Max z Min z 5 (22)

where T Fobj (n,k) is the amount of the Fobj term that is inuenced by a given path k and a given variable n. x is the fraction determined by Eq. (18). TFobj (n) is the total amount of the Fobj term that is inuenced by a given variable n. Summing all the terms inuenced by all the variables in the selected paths it is possible to achieve the total amount in the objective function that can be changedimproved. In order to make a comparison between the target variables and to see which of them have the relative higher effect in Fobj the fraction between the objective function terms, calculated using Eq. (19), by the total amount in the objective function that can be

(23)

where In is the increment used in the determination of the scores scale for the weights in the prot expression (W) or for the sensitivity of the variables (S). Min y and Min z are the minimum value presented for the fractions y and z, respectively. Max y and Max z are the maximum value presented for the fractions y and z, respectively. Tables 3 and 4 show how the scale is dened for both cases. Step 4-g: Target variables scores Analyze the percentages dened in Steps 4-e-II and 4-e-III and transform them to scores through the scales set in Step f.

338

p r o c e s s s a f e t y a n d e n v i r o n m e n t p r o t e c t i o n 8 6 ( 2 0 0 8 ) 328346

Table 4 Ranges and scores for the target variable sensitivity

Kind of sensitivity Sensitivity Extremely sensible Much sensible Sensible Small sensitivity Almost no sensible Inferior limit Superior limit Scores

Min z + 4 Ins Min z + 3 Ins Min z + 2 Ins Min z + Ins Min z

Max z Min z + 4 Ins Min z + 3 Ins Min z + 2 Ins Min z + Ins

5 4 3 2 1

Category 1: Operational variables associated with a separation. Category 2: Operational variables associated with owrate reduction in a closed-path. Category 3: Operational variables associated with a reaction. Category 4: Operational variables associated with owrate reduction in an open-path. Once the categories have been identied the corresponding synthesis algorithm is needed to generate sustainable alternatives. The following synthesis algorithms are going to be used in the given situations. Separation synthesis: Apply algorithm of Jaksland et al. (1996). Improvement in a separation unit: Apply algorithm of DAnterroches and Gani (2005). Improvement in a reactive unit: Apply algorithm of DAnterroches and Gani (2005). Selection of new solvents: Apply algorithm of Harper and Gani (2000). Table 5 summarizes the synthesis algorithms that can be apply to all the categories shown in the ow-diagram presented in Fig. 7. The nal step of this methodology is to validate and compare the new alternatives that match the design targets in terms of their improvements in the performance criteria. The proposed new alternatives are simulated using the new owsheet conguration or the new operational conditions. With this new data, the performance criteria are calculated again and a comparison between the new alternatives is done taking into account the following criteria. An alternative is considered more sustainable if and only if it improves the indi-

Step 4-h: Indicators scores Determine the score for all indicators by the sum of the points given by the target variables, which inuence them. Step 4-i: Determine the target indicators Set the highest scored indicator as the rst target for improvement. Step 5: Operational variables sensitive analysis With the target indicators and their variables identied in Step 4 the next task is to determine the process-operational variables that cause the biggest changes in the target indicators for smallest changes in their values. This analysis is done by checking the inuence of step increments in all the operational variables that inuence the selected target indicator and the consequent effect in the target. Note that the availability of an incidence matrix where the rows are the indicators, the sustainability metrics and the safety indices and the columns are the variables that inuence all of them makes this an easy task. Step 6: Generate and validate the new alternative The steps involved in the generation of sustainable design alternatives are highlighted in Fig. 7. The ow-diagram in Fig. 7 shows four categories where the operational variables can be considered.

Fig. 7 Work-ow for generation of sustainable alternatives (Step 6).

p r o c e s s s a f e t y a n d e n v i r o n m e n t p r o t e c t i o n 8 6 ( 2 0 0 8 ) 328346

339

Table 5 Synthesis algorithm that should be applied to each category in the ow-diagram represented in Fig. 8
Methodology DAnterroches and Gani (2005) Approach 1.2; 2.1.3; 2.1.4; 2.2.1; 2.3.1; 2.4.1; 2.4.2; 2.5.3; 3.1; 4.1.3; 4.1.4; 4.2.1; 4.3.2; 4.4.1; 4.5.2 1.1; 2.1.2; 2.1.4; 2.2.2; 2.3.1; 2.4.2; 2.5.3; 4.1.3; 4.2.1; 4.3.2; 4.5.2 2.5.2; 4.5.1

the user to identify the target indicators and guides the user trough the steps of the ISA algorithm and Steps 5 and 6 (these are under development for the next version).

2.2. Case studyVCM (vinyl chloride monomer) production

In order to illustrate the application of the indicator-based methodology and the use of Sustain-Pro tool the owsheet for the production of VCM is presented as a case study. This case study was chosen because it is not simple, has industrial signicance and provides interesting challenges for SustainPro. The VCM plant comprises of 31 unit operations which include 4 distillation columns and 3 reactor systems, 6 recycle streams, 52 streams and 35 compounds.

Jaksland et al. (1996)

ICAS (ProCamd)

cator targets without compromising the performance criteria in more than 1% comparing with their initial value. From the proposed alternatives the one with the better results will be the one selected.

2.1.

Sustain-Proas computer tool

2.2.1.

Process owsheet

Based on the work-ow, data-ow and calculations of the methodology described above, an EXCEL-based software (Sustain-Pro) has been developed. The inputs for Sustain-Pro are the mass and the energy balance data as well as the prices of the compounds present in the process. Sustain-Pro is able to read the mass and the energy balance from an EXCEL le, thereby making the transfer of data generated through process simulators or from real plants quite simple and easy. Sustain-Pro performs the owsheet decomposition, lists the open- and closed-paths, and calculates the indicators and the various metrics. It then allows

The VCM process owsheet (Casebook les, 1992) can be divided into ve sections as shown in Fig. 8. Section 1: Here it takes place the direct chlorination of ethylene. Part of 1,2-dichloroethane (EDC), which is the raw material used to produce VCM, is obtained by the reaction of ethylene with chlorine. Section 2: The remaining EDC, used in the process, is produced in this section by oxychlorination of ethylene with the recycled HCL and O2 . Section 3: EDC provided by the previous sections and recovered from the process presents around 20 compounds as

Fig. 8 Process owsheet for VCM process.

340

p r o c e s s s a f e t y a n d e n v i r o n m e n t p r o t e c t i o n 8 6 ( 2 0 0 8 ) 328346

Fig. 9 Flowrates of VCM in the nested loops of VCM process owsheet.

impurities that should be removed, so in this section it takes place the purication of EDC. Section 4: The thermal cracking of EDC to form VCM and HCL is present in this section. Section 5: Finally in this last section the VCMs purication takes place allowing the HCLs recover to the oxychlorination reactor. The unreacted EDC is also recycled to EDC purication system.

as shown below: fC
(VCM)
Dash line ,SBold

frC

(VCM)
Dash line

2 (A) f cp=0 cpj,si (VCM) Bold ,SBold 2 (A) f cp=0 cpj,si

0.88 = 0.51, 0.88 + 0.84

frC

(VCM)
Bold

fC

0.84 = 0.49 0.88 + 0.84

2.2.2.

Application of Sustain-Pro

Step 1: Collect steady-state data The required detailed process data for the VCM plant is given as a sample PROII (Version 8) les. The steady-state operational data were generated trough PROII. Step 2: Flowsheet decomposition For this case study it was obtained 8 different partitions with 252 closed-paths and 138 open-paths with owrate different from zero. In order to show the new features of the owsheet decomposition (see Step 2), Fig. 9 is presented. Here the decomposed owsheet, for VCM (compound selected) with the owrates in the respective streams is shown. There two of the closedpaths (nested loops) are highlighted by bold and dashed lines. Eq. (1) is used to calculate the fraction of the owrate of VCM in each of the two closed-paths,

Eq. (2) is used to calculate the owrates of VCM in each of the two closed-loops as follows: mC mC
(VCM)
Dash line

(VCM),CDash line = fBold xC (VCM)

Bold

(VCM)
Dash line

= 0.31 0.51 = 0.159,

(VCM)
Bold

(VCM),CBold = fBold xC

= 0.31 0.49 = 0.151

Here it is a good example to prove that without this new owsheet decomposition algorithm negative owrates would appear when the open-path determination would be done. In this way it is possible to quantify the real accumulation for each closed-path. Repeating this procedure, the corresponding values for all other compounds are obtained. Also using the calculation steps outlined in Fig. 3 and Section 2, the owrates for all open- and closedpaths are obtained. The selection of these paths is listed in Table 6.

Table 6 Identication of the most relevant pathsVCM case study

Path # OP19 OP28 CP3 CP39 CP111 CP219 Component EDC EDC EDC EDC EDC EDC Streams within the path PRX1 EDCAEDCDEDCEEDC1-EDC3EDC4EDC5BTM1HVYFBTM2 PRX3 OXYAOXYCOXYDEDC2EDC3EDC4EDC5BTM1HVYFBTM2 HVYFTOP2VAPP TOP2PYRFEFF1EFF2LF5QSP EFF2VF5Q2 EDC5BTM1HVYFTOP2PYRFEFF1EFF2VF5VF6VF7BTM3BTM4 Flowrate (kg/h) 163.32 233.92 4484.45 847.32 20090.08 27.96

p r o c e s s s a f e t y a n d e n v i r o n m e n t p r o t e c t i o n 8 6 ( 2 0 0 8 ) 328346

341

Table 7 Calculated indicator values of the most sensitive open-pathsVCM case study
Open-path OP19 OP28 Component EDC EDC MVA (103 $/year) 58,827 41,073 EWC (103 $/year) 272,167 189,108 TVA (103 $/year) 213,340 148,035

Table 8 Calculated indicator values of the most sensitive closed-pathsVCM case study
Closed-path CP3 CP39 CP111 CP219 Component EDC EDC EDC EDC EWC (103 $/year) 169.18 143.99 56.41 146.21

Step 3.1: Calculate mass and energy indicators The values of the indicators for all the open- and closedpaths were determined by Sustain-Pro and a selection of them are listed in Tables 7 and 8. Comparing the values obtained for all the indicators and their corresponding paths, those that have the most negative values of MVA, RQ and TVA and the highest values of AF and EWC were identied and listed in Table 7 (for open-paths) and Table 8 (for closedpaths). In the open-paths, it is possible to see that all sensitive open-paths are related to EDC. The negative values of TVA (total value added) demonstrate that a very high potential for improvements exist for these paths. These two open-paths were without doubt the ones with the lowest values of TVA and consequently the ones with higher chances of improvements. The reason why these open-paths have very negative TVA values is because of their very high values of EWC, which indicates that these paths are consuming a lot of energy comparing with the remaining paths and consequently they should be reduced. In the closed-path there is a big consumption of energy with the recovery of the raw material in all the places where it is being recovered. These values demonstrate that the utilization of EDC is not the optimal within the process. Step 3.2: Calculate safety indices The value of each sub-index was determined through the utilization of the process and compounds parameters and using the ranges available for each sub-index in Heikkila (1999). The scores for each index are presented in Table 9. The Chemical inherent safety index value is 22 while the process inherent safety index value is 17, giving a total safety index of 39. The data to calculate these indices was taken from MSDS (Anon, in press). Step 3.3: Calculate sustainability metrics In this work, only 23 sustainability metrics were calculated, because it is assumed that the corresponding parameters for the remaining 26 metrics would remain unchanged (for example, new investment, tax-benets, etc.) and also since data related to the social metrics was not available and/or too uncertain. The values for the 23 calculated metrics are given in Table 10. The environmental metrics were calculated using the WAR algorithm option in ICAS, ICAS (2006), for the data related to inlet streams (Cl2F, C2F1, WTR1, AIRF, C2F2) and outlet streams (WST1, WST2, VNT2, VNT3, WTR2, TOP1, BTM2 and TOP4). Stream VNT1 has not been considered because owrate is equal zero.

Step 4: Indicator sensitivity analysis algorithm (ISA)target determination Step 4-a: Analysis of indicators Comparing all the indicators calculated to this case study the ones that present the best potential for improvement are listed in Tables 7 and 8 for open- and closed-paths, respectively. It can be noted that the most sensitive openpaths correspond to the wasted mass of EDC, which is produced in the rst and in the third reactors, respectively. These open-paths show a negative value of TVA, indicating that a high potential for improvement exists. The high values of EWC mean high energy consumption initially a need for a reduction of the owrates for these paths or an improvement of the operational variables within the process. In the closed-paths, the only cycles that make a signicant impact are the recycle loops related with the recovery of EDC. In these loops, high values corresponding to EWC can be observed. These indicators point to a high potential for process improvement related with the recovery of the raw material. At this point, 6 paths are available as potential target indicators; they are OP19, OP28, CP3, CP39, CP111 and CP219. Step 4-b: Denition of the objective function To set values for the target indicators, it is necessary to rst dene Fobj . In this case study the main objective has been to increase the process prot without compromising the performance criteria.

Table 9 Inherent safety analysisVCM case study

Score Total inherent safety index (ISI) Chemical inherent safety index, Ici Sub-indices for reactions hazards Heat of the main reaction, Irm Heat of the side reactions, Irs Chemical interaction, Iint Sub-indices for hazardous substances Flammability, I Explosiveness, Iex Toxicity, Itox Corrosivity, Icor Ici Process inherent safety index, Ipi Sub-indices for process conditions Inventory, Ii Process temperature, IT Process pressure, IP Sub-indices for process system Equipment, Ieq IISBL IOSBL Process structure, Ist Ipi ISI

3 3 4

4 1 6 1 22

4 3 1

4 3 2 17 39

342

p r o c e s s s a f e t y a n d e n v i r o n m e n t p r o t e c t i o n 8 6 ( 2 0 0 8 ) 328346

Table 10 Calculated sustainability metricsVCM case study

Area Energy Sustainability metrics Total net primary energy usage rate Total net primary sourced from renewals Total net primary energy usage rate per kg product Total net primary energy usage per unit value added Total raw materials used per kg product 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 Net water consumed per unit value added Value added Value added per unit value of sales Gross margin Total PEI (potential environmental impact) HTPI (human toxicity potential by ingestion) HTPE (human toxicity potential by exposure both dermal and inhalation) ATP (aquatic toxicity potential) TTP (terrestrial toxicity potential) GWP (global warming potential) ODP (ozone depletion potential) PCOP (photochemical oxidation potential) AP (acidication potential) Value 2.42 105 GJ/y 100% 4.82 103 kJ/kg 8.38 103 kJ/$ 1.51 kg/kg 0.0033 kg/$ 0.384 kg/kg 0 kg/kg 1.24 kg/kg 47.27 kg/kg 0.01 kg/$ 3.17 107 $/year 0.77 $/year 2.89 107 $/year 11966 5116 1428 25 5116 0 0 178 104

Material

Water

Economic

Environmental

Table 11 Incidence matrix for use in the ISA algorithmVCM case study
Path OP19 OP28 C3 C39 C111 C219 Indicator EWC EWC EWC EWC EWC EWC mEDCD x x x x x x x x x x x x x x x mEFF2 mLF5 mQ2 mLF7 mEDC2 mTOP2 mPYRF mBTM1 x mBTM2 x x mBTM3 mBTM4

Due to a reason of space here only the variables present in indicators and the prot are represented. mx represents the owrate of stream x.

The prot expression selected for Fobj is the following: Prot = mVCM,TOP4 PrVCM,TOP4 + mEDC,BTM2 PEDC,BTM2 + mVNT1 PrFuel + mVNT2 PrFuel mET,C2F1 PrET,C2F1 mCl2 ,CL2F PrCl2 ,CL2F mET,C2F2 PrET,C2F2 mH2 O,WTR1 PrH2 O,WTR1 Qcooling Prcooling Qheating PrLP steam Qheating PrHP steam where mc is the owrate for compound c, Prc is the price for compound c, Q is the heat exchange in the units. Steps 4-c and 4-d: Incidence matrix and select the target variables The incidence matrix of target indicators expressions and their variables is given in Table 11. In Table 11 only appear the selected variables VT , in the columns of the table, instead of appearing, as normal, all VF variables. This fact is due to the high number of VF variables, which impossibilities the list of all of them. Step 4-e: Target variables analysis (i) For the selected target variables, VT (see column of Table 11), the value of the associated owrates are listed in Table 12. Using the path owrates (see Tables 7 and 8) and the respective variables owrates (see Table 12) the inuence of the paths in the

selected variables were calculated. The calculation of the fraction of variable mEDCD , which is inuenced by open-path 19, is given below as an example: xOP19 = FOP19 163.32 = 2.57% = FmEDCD 6355.5

The fractions of all the paths, which inuence all VT variables, were determined for the entire set of variables listed in the columns of Table 12.

Table 12 Calculated design variables owratesVCM case study

Variable mEDCD mEFF2 mLF5 mQ2 mLF7 mEDC2 mTOP2 mPYRF mBTM1 mBTM2 mBTM3 mBTM4 Flowrate (kg/h) 6355.5 69262.3 29094.0 22686.7 15047.4 5242.6 22892.8 18355.4 18144.9 699.6 13819.4 7534.7

p r o c e s s s a f e t y a n d e n v i r o n m e n t p r o t e c t i o n 8 6 ( 2 0 0 8 ) 328346

343

Table 13 Fraction of variable VA, which is inuenced by the path k (Step 4-e-I in ISA algorithm)VCM case study
Path OP19 OP28 C3 C39 C111 C219 Indicator EWC EWC EWC EWC EWC EWC mEDCD 2.6% 4.5% 1.2% 29.0% 2.9% 88.6% 0.2% 0.1% 0.2% 0.2% 0.2% 0.4% 19.6% 3.7% 4.6% mEFF2 mLF5 mQ2 mLF7 mEDC2 mTOP2 mPYRF mBTM1 0.9% mBTM2 23.3% 33.4% mBTM3 mBTM4

Table 14 Prot terms for each variable (Step 4-e-II in ISA algorithm)VCM case study
Variable mEDCD mEFF2 mLF5 mQ2 mLF7 mEDC2 mTOP2 mPYRF mBTM1 mBTM2 mBTM3 mBTM4 Total amount in prot ($/h) 2.639 0.414 0.414 7.342 6.259 5.280 10.229 157.508 25.551 58.727 20.080 20.066

Table 16 Coefcients to each variable (Step 4-e-III in ISA algorithm)VCM case study
Number v1 v2 v3 v4 v5 v6 v7 v8 v9 v10 v11 v12 Sum Variable mEDCD mEFF2 mLF5 mQ2 mLF7 mEDC2 mTOP2 mPYRF mBTM1 mBTM2 mBTM3 mBTM4 Coefcients 0.000 0.000 0.000 0.000 0.000 0.001 0.000 0.009 0.002 0.002 0.002 0.002 0.02

The list of the calculated values is present in Table 13. (ii) After the determination of all the fractions, it is necessary to determine for each variable present in Table 12 the prot terms that are inuenced by them and sum all of them. Table 14 gives the amounts inuenced by these variables. Using the calculated values from above, the maximum value by which the variables can be changed is determined. The calculations involved are highlighted for open-path 19 and variable mEDCD . Appling Eq. (19), the inuence of variable mEDCD , within open-path 19, in the prot term related with that variable is determined by the following expression: PT(mEDCD ,OP19) = x(OP19) PT(mEDCD ) = 0.0257 2.639 = 0.068 $/h

fraction is calculated as shown below: y(mEDCD ,OP19) = PT(mEDCE ,OP19)

K k V PT(v,k) v

0.068 = 0.134% 50.59

Table 15 shows the percentage of the prot that can be changed by each variable in each path with respect to the maximum amount that can be changed. (iii) The coefcients present in the prot expression that are multiplied by each target variable were determined and they are summarized in Table 16. These coefcients correspond to the sum of the utilities prices, which inuence each variable. The percentage of each coefcient in the total amount of the coefcients should be determined using Eq. (21). For open-path 19 and variable mEDCD the percentage of sensitivity coefcient is given by the following expression: Cf(mEDCD )
12 Cf v=0 (v)

zmEDCD = Applying the same expression to all the variables and paths and summing all the PT values the maximum amount in the prots that can be changed by these target variables is determined. Eq. (20) is next applied to determine the fraction represented by each variable within a given path in the maximum prot amount that can be changed. In the case of the open-path 19 the

4.2 104 = 2.27% 0.02

Applying the same expression for all the paths and variables the fraction of the coefcients in the sum of all the coefcients are presented in Table 18. Step 4-f: Scores scales determination The scales to give the scores to the target variables were determined calculating the increment following Eqs. (22)

Table 15 Percentage of the amount of each variable in the total prot amount determined by the sum of the values (Step 4-e-II in ISA algorithm)VCM case study
Path OP19 OP28 C3 C39 C111 C219 Indicator EWC EWC EWC EWC EWC EWC mEDCD 0.1% 0.5% 0.0% 0.2% 0.0% 12.9% 0.0% 0.0% 0.5% 0.1% 0.1% 0.1% 4.0% 0.7% 14.4% mEFF2 mLF5 mQ2 mLF7 mEDC2 mTOP2 mPYRF mBTM1 0.5% mBTM2 27.1% 38.8% mBTM3 mBTM4

344

p r o c e s s s a f e t y a n d e n v i r o n m e n t p r o t e c t i o n 8 6 ( 2 0 0 8 ) 328346

Table 17 Percentage of each coefcient in their sum (Step 4-e-III in ISA algorithm)VCM case study
Path OP19 OP28 C3 C39 C111 C219 Indicator EWC EWC EWC EWC EWC EWC mEDCD 2.3% 5.5% 0.1% 0.1% 0.1% 0.1% 2.3% 1.5% 47.0% 9.2% 9.4% 10.5% 1.5% 1.5% 47.0% mEFF2 mLF5 mQ2 mLF7 mEDC2 mTOP2 mPYRF mBTM1 9.2% mBTM2 10.4% 10.4% mBTM3 mBTM4

Table 18 Ranges and scores to the Step 4-e-II in ISA algorithmVCM case study
Weight Inferior limit Superior limit Scores

Weight in the prot expression Extremely high 31.05% High 23.29% Normal 15.53% Low 7.77% Extremely low 0.01%

38.81% 31.05% 23.29% 15.53% 7.77%

5 4 3 2 1

Table 19 Ranges and scores to the Step 4-e-III in ISA algorithmVCM case study
Type of sensitivity Sensitivity Extremely sensible Much sensible Sensible Small sensitivity Almost no sensible Inferior limit Superior limit Scores

37.64% 28.24% 18.85% 9.45% 0.06%

47.04% 37.64% 28.24% 18.85% 9.45%

5 4 3 2 1

and (23). The scales are in Tables 18 and 19, respectively. Steps 4-g, 4-h and 4-i: Target variables scores, indicators scores and Investment cost analysis In Step g the percentages presented in Tables 15 and 17 are transformed into scores using the scales available in Tables 18 and 19, respectively. The scores to each variable and the nal scores to each indicator are presented in Table 20. Step 4-j: Determine the target indicators According to Table 20, the most sensitive indicators were identied to be the EWC in the closed-path CP219, due to the highest score presented by this path. With this information, the target indicators were identied. Now the only thing remaining is to determine the most sensitive operational variable, which inuences the target indicators. Step 5: Operational variables sensitive analysis The operational variables that inuence the value of EWC in the closed-path CP219 are the following: o Reboiler heat duty: Units COL1, COL3 and COL4. o Condenser heat duty: Unit COL2.

o Heat exchange duty: Units FLS6, FLS7 and PYRO. o Flowrates: CP219 owrate. Variations of 5, 10 and 15% on the reference (design) values for all the identied operational variables. Table 21 gives a summary of the inuence in the target indicator from the changes in the operational variables. Analyzing the results it is possible to conclude that the reboiler heat duties for distillation column four is the most sensitive variable for the target indicator. However, since this unit operation is related with the recovery of the nal product and since the conditions in the distillation column are already optimal, nothing more can be done. The second most sensitive variable is the owrate of closed-path CP219. A variation in this variable will also change the owrate of EDC in other recycle streams and so this alternative could be investigated further by taking into account all the closed-paths together. Step 6: Generate and validate the new alternative Step 5 identied the operational variable to achieve an improvement in the target indicator as the owrate of a raw material (EDC) in a closed-path (CP219). From Fig. 7 it can be noted that the improvement category for this variable is the category 2.1. To improve the variables of category 2.1 there are four alternatives available: o Insert a purge (2.2.2.1). o Improve the already existent separation units (2.2.2.2). o Insert a new separation process (2.2.2.3). o Improve the conversion (2.2.2.4).

Insert a purge. This alternative is feasible if the raw material is recycled with small amounts of impurities that are becoming accumulated in the system. This alternative also involves practically no investment costs. However even though the indicators targets are improved due to the reduction in the EDC owrate reduction, the performance criteria became worse by more than 1%, thereby not satisfying the criterion used to dene a more sustainable alternative. The reason for this is because the loss of the raw material in the purge did not compensate for the reduction in the production of the prot. A simulation of process owsheet with this alternative conrmed the above analysis.

Table 20 Scores for the different paths (Steps 4-g, 4-h and 4-j in ISA algorithm)VCM case study
Path OP19 OP28 C3 C39 C111 C219 mEDCD 2 2 2 2 2 3 2 2 6 2 2 3 2 2 7 mEFF2 mLF5 mQ2 mLF7 mEDC2 mTOP2 mPYRF mBTM1 2 mBTM2 6 7 mBTM3 mBTM4 Total 10 9 2 13 5 17

p r o c e s s s a f e t y a n d e n v i r o n m e n t p r o t e c t i o n 8 6 ( 2 0 0 8 ) 328346

345

Table 21 Summary of the sensitivity analysis to measure the inuence of operational variables in EWC of closed-path CP219VCM case study
Cycle path Variables Decrease (%) 5 10 15 5 10 15 5 10 15 5 10 15 5 10 15 5 10 15 5 10 15 5 10 15 5 10 15 5 10 15 EWC improvements 0.03% 0.06% 0.09% 0.01% 0.01% 0.02% 0.03% 0.06% 0.09% 0.13% 0.25% 0.38% 0.00% 0.00% 0.00% 0.00% 0.01% 0.01% 0.01% 0.03% 0.04% 0.03% 0.06% 0.09% 4.76% 9.52% 14.28% 0.90% 1.86% 2.90%

QCol 1-R

QCol 2-C

QFLS4

QPyro

QFLS5 C219 QFLS6

the prot improved by 0.34%, while all the other metrics and indices remained constant. Based on a calculation required for an investment for a membrane-based pervaporation unit, it could be concluded that the investment cost of the membrane will be paid within 1 year. Improve the conversion. This alternative was not tested as there was not sufcient data available about the reaction kinetics. Also it is known that big changes in the reaction mean many changes in the down stream separation processes and therefore possible high investment costs. Summarizing, the insertion of a new membrane-based pervaporation unit appears to be a promising alternative to achieve improved sustainability of the process. In this (design) alternative none of the performance criteria became worse compared to their reference values, except for those that were improved. Therefore, Fobj was improved without any trade-off in the sense that something improved at the expense of others.

3.

Discussion and conclusions

QFLS7

QCol 3-R

QCol 4-R

FCP219

Improve the already existent separation units. The VCM process employs two distillation columns to purify the EDC before the production of VCM. The operational design of these two distillation columns was studied to identify the design corresponding to a reduction of the owrate in CP219. This alternative also was not sustainable as the energy and the water metrics became worse by more than 1% when compared to their reference (design) value. Again, simulation of process owsheet with this alternative, conrmed the above analysis. Insert a new separation process. Inserting a new operation technique (equipment) in closed-path CP219 becomes an interesting option if the investment cost for the new equipment is relatively low and if the owrate of CP219 can be reduced without reducing production rate. To determine the best unit to separate the impurities present in the EDC recycle loop, the methodology of Jaksland et al. (1996) was used. Following this methodology, four types of separation units were suggested to all pairs of compounds present in the recycle streamgas absorption, liquid absorption, liquid membrane and pervaporation. Since stream TOP2 that is going to be separated has very small owrates of the impurities and it is in the liquid phase, pervaporation appeared to be a promising alternative. The simulation to validate this new approach was done and the target indicators (EWC of EDC closed-paths in Table 8) were improved in 31%. At the same time all the performance criteria remained undisturbed while others actually improved. This means that this alternative respects the conditions (constrains) for being classied as a sustainable alternative. Further simulations reveal that the process is now more sustainable as the energy metrics decreased by 2%, the water metrics decreased by 2.8% and

The development of a systematic and generic indicator-based methodology for process analysis and improvement has been presented and highlighted through a non-trivial case study. One of the most important features of this methodology is the fact that sustainable (design) alternatives that avoid tradeoffs in design decisions can be identied if such opportunities exist. A useful feature of this methodology is that the indicatorbased analysis (Steps 13) needs to be made only once for any process with respect to a base-case or existing design and then when the process needs to be improved, the Steps 46 can be followed to identify the more sustainable alternatives. This methodology has also been applied to a number of case studies, for example, well known process such as the HAD process, natural gas purication plant, ammonia plant, etc., which have not been reported in this paper. With the development of a corresponding software (Sustain-Pro), the application of the methodology has been easier and less time consuming. As shown through the VCM case study (and also ammonia case studynot reported here), the size of the problem is not an issue as the objective is to locate the most sensitive indicators and their related variables, which are not that many. Current and future work is looking at generating operation alternatives for high-value batch-operation-based chemical products.

Acknowledgement
The authors gratefully acknowledge nancial support from Fundacao para a Ciencia e a Tecnologia (under grant no. SFRH/BD/24470/2005).

references

http://www.msds.com. A. Azapagic, Sustainable Development Progress Metrics, IChemE Sustainable Development Working Group, IChemE Rugby, UK (2002). Biegler, L.T., Grossmann, I.E. and Westerberg, A.W., 1997, Systematic methods of chemical process design, in Recycle

346

p r o c e s s s a f e t y a n d e n v i r o n m e n t p r o t e c t i o n 8 6 ( 2 0 0 8 ) 328346

Partitioning and Tearing, Amundson, N.R. (ed) (Prentice Hall PTR, New Jersey, USA), pp. 271282. Cabezas, H., Bare, J. and Mallick, S., 1999, Pollution prevention with chemical process simulators: the generalized waste reduction (WAR) algorithm. Comput Chem Eng, 23(4/5): 623634. Casebook les-ProII, 1992. Ciric, A.R. and Floudas, C.A., 1989, A retrot approach for heat exchanger networks. Comput Chem Eng, 13: 703715. DAnterroches, L. and Gani, R., 2005, The reverse approach for synthesis and design of chemical products & processes, In Proceedings of Seventh WCCE Harper, P.M. and Gani, R., 2000, A multi-step and multi-level approach for computer aided molecular design. Comput Chem Eng, 24(27): 677683. A.-M. Heikkila, Inherent safety in process plant designan index-based approach, Ph.D. Thesis, VTT Automation, Espoo, Finland (1999). ICAS documentation., (2006). CAPEC. (Department of Chemical Engineering, DTU, Lyngby, Denmark). Jaksland, C., Gani, R. and Lien, K., 1996, An integrated approach to process/product design and synthesis based on properties-process relationship. Comput Chem Eng, 20: 151156, colleague Jackson, J.R. and Grossmann, I.E., 2002, High-level optimization model for the retrot planning of process networks. Ind Eng Chem Res, 41: 37623770.

Jensen, N., Coll, N. and Gani, R., 2003, An integrated computer-aided system for generation and evaluation of sustainable process alternatives. Clean Technol Environ Policy, 5: 209225. Lange, J.P., 2002, Sustainable development: efciency and recycling in chemical manufacturing. Green Chem, 4(6): 546550. Liu, J., Fan, L.T., Seib, P., Friedler, F. and Bertok, B., 2006, Holistic approach to process retrotting: application to downstream process for biochemical production of organics. Ind Eng Chem Res, 45: 42004207. Rapoport, H., Lavie, R. and Kehat, E., 1994, Retrot design of new units into an existing plant: case study: adding new units to an aromatics plant. Comput Chem Eng, 18: 743753. Sargent, R.W.H. and Westerberg, A.W., 1964, Speed-up in chemical engineering design. Trans Inst Chem Eng, 2: 190197. Uerdingen, E., Gani, R., Fischer, U. and Hungerbuhler, K., 2003, A new screening methodology for the identication of economically benecial retrot options in chemical processes. AIChE J, 49: 24002418. Uerdingen, E., Fischer, U., Gani, R. and Hungerbuhler, K., 2005, A new retrot design methodology for identifying, developing, and evaluating retrot projects for cost-efciency improvements in continuous chemical processes. Ind Eng Chem Res, 44(6): 18421853.