GreenPro-I: a risk-based life cycle assessment and decision-making methodology for process plant design

Environmental Modelling & Software 17 (2002) 669–692 www.elsevier.com/locate/envsoft

GreenPro-I: a risk-based life cycle assessment and decisionmaking methodology for process plant design Faisal I. Khan ∗, Rehan Sadiq 1, Tahir Husain Faculty of Engineering and Applied Science, Memorial University of Newfoundland, St. John’s, Newfoundland, Canada, A1B 3X5 Received 24 September 2001; received in revised form 11 January 2002; accepted 26 March 2002

Abstract In recent years, significant attention and emphasis has been given to cleaner and greener technologies in processes and product manufacturing. This is recognized as a key element in pollution prevention (P2) and development of sustainable strategies. Life cycle assessment (LCA) is a systematic approach that enables implementation of cleaner and greener product and process concepts in industry. In recent times substantial progress has been made in the use of LCA for product evaluation and selection. However, its use in cleaner and greener process design and decision-making has not been explored to a great extent. Process design and decision-making are challenging activities that involve trade-off of conflicting objectives, namely costs, technical feasibility and environmental impacts. These conflicting objectives can be analysed at the early design and decision-making stage by considering the full life cycle of a process or a product. A cleaner and greener process is the one that is cost optimal, technically feasible, and environmentally benign. To obtain these results LCA requires various tools and techniques in a systematic methodology. This paper proposes a holistic and integrated methodology GreenPro-I for process/product design by combining the traditional LCA approach with multi-criteria decision-making methods. This methodology is simple and applicable at the early design stage and is more robust against uncertainty in the data. Application of the methodology has been demonstrated in the paper through a urea production case study.  2002 Elsevier Science Ltd. All rights reserved. Keywords: LCA; Pollution prevention; Cleaner production; Risk-based design; MCDM

1. Introduction Life cycle assessment (LCA) is a systematic approach to estimate environmental impacts associated with products, processes and services. It is described as a process to evaluate environmental burdens by identifying and quantifying energy and materials used and wastes released into the environment, and to identify and evaluate opportunities to affect environmental improvements. The assessment includes the entire life cycle of the product, process or activity encompassing extraction and processing of raw materials, manufacturing, transportation and distribution, use, recycle, and final disposal (Fava et al., 1991; USEPA, 1995; Tukker 2000). The Corresponding author. Tel.: +1-709-737-7652; fax: +1-709-7374042. E-mail address: [email protected] (F.I. Khan). 1 Present address: Institute for Research in Construction (IRC), National Research Council, Ottawa, Ontario, Canada K1A 0R6. ∗

LCA is a decision-making tool for designer, regulatory agencies, and business organizations. It is used to evaluate the environmental impacts of products and process and also identifies a section within a product or process’s life cycle where the greatest reduction in resource requirements and emissions can be achieved. Generally the LCA employs a product-centered approach, but recent efforts suggest a great potential of the LCA in process design. For a designer it has been serious question how to accommodate various constraints such as economic, technology and environment to the design and operation of cleaner and greener process. Earlier design approaches were economics centered and based on cost–benefit analysis. These approaches attempt to trade-off environmental and economic assets with an aim to maximize differences between socioeconomic benefits of an activity against the financial and environmental liabilities. Such practices ultimately seek to attach a monetary value to the environment and are therefore fraught with difficulty (Jacobs, 1991).

1364-8152/02/$ - see front matter  2002 Elsevier Science Ltd. All rights reserved. PII: S 1 3 6 4 - 8 1 5 2 ( 0 2 ) 0 0 0 2 8 - 2

670

F.I. Khan et al. / Environmental Modelling & Software 17 (2002) 669–692

The LCA may prove to overcome many of the problems faced in the conventional approaches and therefore establish a link between the environmental impacts, operation, and economics of a process. It offers an expanded environmental perspective, considering impacts from resource extraction to the end product use and disposal. The LCA relates these effects to the mass and energy flows into, out of, and within a process (Kniel et al., 1996; Choong and Sharratt, 2000). According to the ISO 14040 series standards, the LCA should assess the potential environmental issues and aspects associated with a product or service by compiling an inventory of relevant inputs and outputs; evaluating the potential environmental impacts associated with those inputs and outputs; and interpreting the results of the inventory and impact phases in relation to the objectives of the study. Considering these, we believe that the LCA has an ability to focus both process operational feasibility and environmental concern along with other attributes. The present work focuses on the development of process selection and design methodology considering assessment and minimization of risks/impacts of process systems by embedding the LCA principles within a formal process design and decision-making framework. It has implications to process synthesis as it includes environmental objectives together with technology and economics at the design stage to determine cost efficient solutions. Authors believe that employing the LCA with a process design and decision-making would yield an optimal design and a best management alternative. This paper presents a detailed description of a revised methodology for cleaner and greener process design. The applicability of methodology has been demonstrated through a case study.

2. Past work on cleaner process selection and design In the past, attempts have been made by chemical industries to reduce pollution by implementing cleaner technologies or processes. Many times it has been observed that the efforts made to optimize the waste treatment process reduces the quality and/or quantity of waste discharge at the end-of-the-pipe, but increases the total environmental burden and impacts. Therefore, it is very important to consider the environmental burden and adverse impact caused due to any change or modification in the process and allied facilities in a holistic way. This is possible only when there is a vivid comprehension of the relationships among design/operating parameters with environmental impacts for a complete life cycle of the process. A few attempts have been made in this direction that are summarized in the following paragraphs.

El-Halwagi and Manousiouthakis (1989a,b, 1990a,b) have introduced the concept of mass pinch as a tool to derive cost optimal mass exchange networks with minimum emissions waste. Using this concept, Wang and Smith (1994) developed a method to obtain design targets for minimum wastewater generation in process plants. In their latest publications El-Halwagi and coworkers (1998 and 1999) have proposed interval based targeting procedure for waste minimization considering energy and mass inputs. The waste minimization was achieved by recycling of the target stream or species back to the process or selecting a replacement of the same with less severe one. Flower et al. (1993) have introduced the idea of graphical mass balance — a visual means of mass balance manipulations for an initial screening of process not complying with environmental regulations. Stephanopoulous et al. (1994) have introduced the concept of zero avoidable pollution in their design toolkit for batch processes. Pistkopoulos et al. (1994) have proposed a methodology to estimate various environmental impacts. These environmental impacts were minimized for a particular process through a cost function (a function describing a relationship between impacts and process model). Kniel et al. (1996) presented the LCA case study of a nitric acid process considering environmental and economic constraints. Though the article emphasizes the use of LCA in a design stage but does not provide any details of the methodology and steps involved. In further efforts, Hernandez et al. (1998) proposed a mathematical model to minimize the environmental impacts that are subsequently used for deciding the optimal degree of pollution abatement. Azapagic and Clift (1999a,b,c,d) appreciably advocated for the LCA based design and decision-making. They demonstrated the application of the LCA in various processes and product selection and decision-making. Azapagic (1999) and Azapagic and Clift (1999b,c) have discussed the LCA application in the evaluation of process performance for various Boron products. Young and Cabezas (1999) have proposed a sustainable design methodology using WAR (waste reduction design). Their methodology is focused on minimizing the waste across the process boundary. Spath et al. (1999) and Mann and Spath (1997) have conducted a detailed LCA for various options of power plant. In the latest report, Spath and Mann (2000) have demonstrated successful application of the LCA in evaluating different stages of natural gas combined cycle power generation system. This report clearly emphasizes the strength of the LCA in identifying and screening environmental burdens at various life stage of the plant, which enables development of targeted remedial goals. In order to streamline the LCA approach, a workgroup (SETAC North America LCA streamline workgroup) has been formed. The workgroup has recently submitted

F.I. Khan et al. / Environmental Modelling & Software 17 (2002) 669–692

671

their final report that discusses in detail streamline efforts of various steps of LCA and environmental parameters — eco vectors (Todd and Curran, 1999). The impact of any corrective or remedial measure may be beneficial for the process viewpoint, but modification required to achieve this may cause more severe environmental impacts or burdens outside the boundary of process. Khan et al. (2001) have proposed a design methodology GreenPro for cleaner and greener process. The methodology comprises four steps. The first two steps deals with design problem formulation considering life cycle, and the last two steps involves a design problem solution with multi-objective optimization and multi-criteria decision-making (MCDM). Application of this methodology was demonstrated by applying it to a vinyl chloride monomer (VCM) process. The GreenPro methodology is good and effective in design, however it application is restricted at early design stage due to extensive computational load and large data requirements. In the latest effort, these authors have reworked the GreenPro and tried to overcome earlier limitations. The revised methodology, GreenPro-I, is an efficient, easy to use, applicable to early design stage and robust (comparatively less sensitive to the reliability of the data). Detailed description of the methodology and recommended tools/techniques are presented in subsequent sections.

3. GreenPro-I: the proposed methodology for a cleaner and greener process design The GreenPro-I methodology comprises two main steps 1. risk-based life cycle assessment (RBLCA); and 2. risk-based MCDM. These two steps further consist of many sub-steps as shown in Fig. 1. A step-by-step description of the methodology is presented. 3.1. RBLCA The major objective of RBLCA is to help in decisionmaking and selecting the best management alternative available, which can reduce the risk and other environmental burdens at the optimal cost. The application of risk management tools aid in selection of prudent, technically feasible and scientifically justifiable actions that will protect environment and human health in a costeffective way. The RBLCA is a process of weighting policy alternatives and selecting the most appropriate action, integrating the results of risk assessment with engineering data and with social, economic, and political concerns to make a decision. The RBLCA determines

Fig. 1. Basic framework for GreenPro-I.

the alternatives, which cause minimum environmental damages and evaluates the costs and benefits of proposed risk reduction programs. The RBLCA integrates political, legal and engineering approaches to manage risks and environmental burdens of a process or product. The RBLCA provides information about human health, ecological, safety and economical risks in quantitative or qualitative terms for each alternative. Subsequently, designers and/or decisionmakers use this information to select the best management alternative. Generally, the traditional risk assessment is carried out objectively, whereas the RBLCA involves preferences and attitudes, which has objective and subjective elements. The RBLCA for process design has two main objectives: 1. quantifying and evaluating the environmental performance of a process from cradle to gate and 2. helping to choose the best process design combinations.

672

F.I. Khan et al. / Environmental Modelling & Software 17 (2002) 669–692

Thus the RBLCA acts as a useful tool for identifying the best practicable environmental option. Further the RBLCA helps in identifying the options for improvement of the environmental performance of a system. The RBLCA considers the whole material and energy supply chains within system boundary. It includes material and energy flows into the system, outflows from the system and what used within the system. It classifies the environmental burden for all activities within the system boundary right from material extraction/excavation, refining, transportation, construction of the plant, commissioning of the plant, production and decommissioning to the ultimate disposal. 3.1.1. Defining the scope and boundary Prior to getting the RBLCA underway, its scope must be clearly articulated and system boundaries should be decided to ensure that no relevant parts of the system are omitted. As shown in Fig. 2, this requires backtracking from the conventional process system to the natural state of pure raw materials, which are available at no environmental penalty. Different technological routes for the production of the same set of raw materials should be included in this expanded boundary. The

advantage of defining an expanded global process system boundary is that input (to the conventional process) together with their routes can be accounted with the output emissions forming an aggregated waste vector. Note that, although this definition is consistent with the one used in conventional LCA for products, but here it does not include the routes and stages of the product after leaving the process (cradle-to-gate approach instead of cradle-to-grave approach) (Fig. 2).

3.1.2. Inventory analysis In order to accomplish the RBLCA goals, the energy and material inputs, wastes and emissions data must be collected and environmental load (EL) should be quantified. The inventory analysis sub-step collects data and quantifies EL by focusing on material and energy balances for each operation within the system and for the entire life cycle of the system. Inventory analysis has to be comprehensive in order to provide the necessary data to make the final decision. In instances where rough estimates have to be made, they must be conservative and clearly noted. Data quality will likely vary depending upon its source, therefore, an extreme care must be taken

Fig. 2. Definition of various life cycle boundaries.

F.I. Khan et al. / Environmental Modelling & Software 17 (2002) 669–692

in order to ensure the data used are in the best available form and the sources are contemporary. In inventory analysis, the ELs are estimated for entire life span of the process plant and its associated activities are based on scope and boundary of the RBLCA. The output streams from a process is classified either a product or a waste. A waste is an output stream from the process that is released to the environment such as vents, purges, litters, and effluent-discharge. The release rates and composition of a component in waste streams represent its EL. Product streams are non-waste output streams. The distinction between waste and product streams is the essence of inventory analysis. The guiding principle of inventory analysis methodology is that all input and waste streams in a system (or subsystem) must be allocated to the respective non-waste output streams. When one output stream becomes an input stream to the next process step, it carries its ELs into the next output streams. Inductively it follows that and final products are assigned the sum of the ELs of raw materials, energy, and wastes in this process chain. In single product system, the EL calculation involves simple balancing of input and waste streams. In multiple systems, however, we also need a method to allocate loads to each product. Fig. 3 illustrates a general framework of the methodology of inventory analysis. Detailed description of inventory analysis is given in Khan et al. (2001) and a brief account is presented below. The inventory analysis starts with the identification of environmental hazards expected at various units. These hazards are due to the chemical species involved in the process that upon release causes adverse effects to humans or to the environment. It also includes hazard due to severity of operating conditions like temperature and pressure. These chemical hazards are not limited to process chemistry, rather they include cleaning solvents, heating and cooling agents, and all other chemicals involved in any part of the process. For each environmentally hazardous chemical all possible release scenarios are identified. This step is very important as it determines the scope and the extent of the overall EL assessment. It would be impossible for a designer to guarantee a complete coverage of all possible release scenarios. In this regard, authors recommend the use of maximum credible scenario (MCS) generation approach, which is recently proposed by Khan (2001) for risk assessment. Use of this technique not only reduces the working load but also helps in focusing the study on the desired aspects. These developed release scenarios are assessed quantitatively for various ELs.

3.1.3. Impact and risk assessment Impact and risk assessment (IRA) examines the potential and actual environmental and human health effects from the use of resources (energy and materials) and

Fig. 3.

673

Methodology for inventory and risk/impact analyses.

environmental releases. An IRA is divided into three phases: classification, characterization, and evaluation. Classification is the process of assigning and aggregating results from the inventory into relatively homogeneous impact categories. This process involves identifying stressors and organizing them with respect to impact on the ecosystem. For example, carbon monoxide, chlorine and methane are all stressors with the potential to impact the environment under the category of global warming. Classification includes the creation of complex stressor/impact chains, because a single pollutant can have multiple impacts, and a primary impact can result in secondary (or greater) impacts as one impact results in another along the cascading impact chain. The general categories are ozone depletion potential (OD), acid rain potential (AP), photochemical oxidant impact (POI), global warming potential (GWPI), controlled toxic water mass (CTWM), solid mass disposal (SMD), safety risk (SR), human health risk (HHR), ecological risk (ER), and natural resource depletion (RD). Characterization assesses the magnitude of impacts for each of the stressor categories in order to translate

674

F.I. Khan et al. / Environmental Modelling & Software 17 (2002) 669–692

inventory analysis data into impact descriptors. For example, BOD and nutrient data for wastewater discharges may be translated into ER. Impact equivalency units are available for subsets of chemicals. New impact equivalency units are created for the chemicals for those no equivalence is available. Characterization methods of few categories are depicted below, for detail refer (Overcash, 1986; Rossiter, 1995; Khan et al., 2001). 앫 CTWM=controlled toxic water mass of water pollutant (kg pollutant/h)/ threshold allowable limit value (kg pollutant/kg water); 앫 solid wastes by a solid mass disposal (SMD)=kg solids/h; 앫 GWPI=mass of pollutant (kg / h)∗GWP (kg CO2 equivalence/ kg pollutant); where GWP is the global warming potential for each pollutant defined in terms CO2 is defined; 앫 similar to GWPI, POI is estimated in terms of equivalence of ethylene; POI=kg ethylene/h; 앫 ODP for each substance is expressed relative to the ozone depletion potential of CFC-11; 앫 acidification potential (AP) is based on the contributions of SO2, NOx, HCl, NH3, and HF to the potential acid deposition i.e. on their potential to form H+ ions; and 앫 the HRR is characterized as a concentration of contaminant at specified location exposed to human population/reference dose (RfD or no-observedeffect-concentration, NOEC). The SR is quantified as the individual risk factor at the study area due to any eventuality in operation and divided by a factor of 1.0×10⫺06. The ER is characterized as a concentration of contaminant at specified location as exposed to ecological community/reference dose (RfD).

Valuation means, assigning relative values or weights to different impacts, economic and performance measures, so that the total score for all decision criteria could be determined. The valuation method used in this methodology is known as the analytical hierarchy process (AHP). The AHP process involves a structured description of the hierarchical relationships among the problem elements, beginning with an overall goal statement and developing a decision tree. The branches of the tree include major and minor decision criteria. Assignment of weights are done as a group exercise, where an expert team is asked to reach on a consensus based on weight factors prior to being entered into the RBLCA model. The decision-makers’ team includes at least cost estimation personnel, a process specialist, an environmental engineer, and an ecologist/toxicologist. These authors believe that suggested weighting scheme would hold good for a range of chemical process.

3.1.4. Design problem formulations Development and designing of a process plant involves consideration of number of constraints (Fig. 4). These constraints are interactive and dictate the total design process. Many experts and researchers have argued that consideration of environmental performance and safety along with other design constraints would lead the cost optimal, cleaner and greener design. In a traditional design approach a chemical process design requires knowledge of mass and energy balances, thermodynamics, chemical reaction engineering, and economics. Cabezas et al. (1997 and 1999) have recommended a method to incorporate environmental impacts into process design model through mass balance equations describing potential environmental impact (EI) of the process. The EI balance simply states that environmental impact can enter, exit and generate/accumulate within the system. It is mathematically expressed as: ∂Isystem ep cp ep cp ep system ⫽ Icp in ⫹ Iin ⫺Iout⫺Iout⫺Iwe⫺Iwe ⫹ Igene ∂t

(1)

where Isystem is the amount of EI inside the system (chemical process and energy generation process); Icp in and Icp out are the input and output rates of EI to chemical ep process; Iep in , and Iout are the input and output rates of EI ep to the energy generation process; Icp we and Iwe are the outis puts of EI associated with waste energy; and Isystem gene the rate of generation of EI in the system. For steady state the expression reduces to: ep cp ep cp ep system 0 ⫽ Icp in ⫹ Iin ⫺Iout⫺Iout⫺Iwe⫺Iwe ⫹ Igene .

(2)

It is necessary to relate the conceptual EI to measurable quantities. A generalized linear theory proposed by Mallick et al. (1996) and Cabezas et al. (1999) has been adopted in this study. It relates EI to measurable quantities such as stream flow rates and compositions and chemical specific environmental impacts (jk). The expression for chemical process is described as following: Icp in ⫽ Icp out ⫽ Iep in ⫽ Iep out ⫽

冘 冘 冘 冘 冘 冘 冘 冘 冘 冘 冘 冘 Iin j ⫽

Iout ⫽ j

Iin j ⫽

Iout ⫽ j

Min j

Mout j

Min j

Mout j

Xjkjk ⫹ … Xjkjk ⫹ …

Xjkjk ⫹ … Xjkjk ⫹ …

(3) (4) (5) (6)

where Icp j is the rate of environmental impact into or out of the chemical process; Ij is the EI flow rate with stream j which may be an input or an output stream; Mj is the mass flow rate of stream j which may be an input or an output; Xjk is the mass fraction of component k in stream j; and jk is the EI for chemical k. Same notation with the superscript ‘ep’ used to denote energy production.

F.I. Khan et al. / Environmental Modelling & Software 17 (2002) 669–692

Fig. 4.

675

Constraints involved at various stages of process plant design and life cycle.

In order to analyze the relationship between environmental impact and process cost, a mathematical framework has been used. Where process is represented by a set of mathematical equations, which describe the properties of the inlet stream, waste stream, and equipment specifications, cost functions and the degree of pollutant removal. The total environmental impact of the system is then related through an environmental function, I ⫽ I(X). Any type of environmental impact (e.g. global warming, ozone depletion, etc.) can be incorporated into this function, usually as environmental indices. It allows all emissions from different compounds to be lumped together into a single number, which represents the specific impact. Thus, the environmental function is expressed as the sum of environmental impacts generated by both outputs and inputs from each species. 3.2. Risk-based MCDM Various multi-criteria or multi-attribute decision-making (MCDM or MADM) approaches can be employed in the RBLCA using fuzzy or crisp set theory concepts. The proposed MCDM approach of fuzzy composite programming (FCP) involving AHP is discussed in detail in this section.

Decision-making is an integral part of all design and management issues and it is part of our daily lives. The main concern in decision-making is that decision problems are multiple and diverse in nature and usually have conflicting criteria. In last 30 years, the research related to MCDM, has produced enormous literature in the fields of engineering, business and social sciences. Decision-making is broadly classified into MCDM and MODM (multiple-objective decision-making). The MCDM is associated with problems whose alternatives are predefined and decision-maker is to select or rank various alternatives. In contrary to that the MODM designs the most promising alternative with respect to limited resources. Hwang and Yoon (1981) have critically reviewed methods and applications of MCDM/MODM for a single decision-maker. For more than one decision-maker, the problem becomes complex and the best solution is the one, which will be accepted by all decision-makers. Hwang and Lin (1987) have also discussed group decision-making under multiple-criteria. The fuzzy set applications in decision-making are the extension of many methods used in crisp sets of multicriteria techniques. Chen and Hwang (1992) have summarized these MCDM methods and categorized them. Many research attempts have been made to incorporate the uncertainty of attributes into decision-making, which

676

F.I. Khan et al. / Environmental Modelling & Software 17 (2002) 669–692

involves probability theory and/or fuzzy set theory. Fuzzy set theory is an important tool for modeling uncertainty or imprecision, arising from human perception. The human judgment is involved in decision-making in different disciplines therefore a rational approach toward decision-making is to take into account human subjectivity. A traditional MCDM problem can be expressed in matrix form as follows: X1 X2 · · Xn A1 A2 · · Am

冢 冣 X11 X12 · · X1n

X21 X22 · · X2n ·

·

· · ·

·

·

· · ·

(7)

Xm1 Xm2 · · Xmn

where Ai ⫽ 1,2,…,m are possible course of actions or alternatives; Xi ⫽ 1,2,…,n are attributes, reference to which that alternative performances are measured; and Xij is the performance or rating of alternative Ai with respect to attribute or criteria Xi. In real life problems, it is common that Xij is not assessed precisely due to unquantifiable, incomplete, and non-obtainable information and partial ignorance. The unquantifiable information means subjective attributes, for example, good, poor, high, low, etc. The examples of incomplete information are about one million, less than 10 miles per hour, etc. Sometimes crisp data are obtainable but need lot of resources, but approximation can be achieved with less effort and time. Sometimes linguistic descriptions are employed because of unavailability of information or resources. The fuzziness is also attributed due to ignorance from factual conditions. These limitations in MCDM methods lead to fuzzybased approaches.

compared and rank orders of fuzzy numbers are determined which in some cases is not a trivial task. Therefore both phases are integral parts of fuzzy MCDM. The size of a MCDM problem depends on the size of alternatives and attributes. Generally fuzzy methods are suitable for 10 or less alternatives and attributes. The data could be in the form of: 1. all fuzzy; 2. all fuzzy singleton (single value and its corresponding membership function, m(x); 3. all crisp; and 4. mixture of fuzzy and crisp. The concepts of fuzzy methods are derived from classical methods of MCDM, e.g., simple additive weighting (SAW), AHP, multiple attribute utility function (MAUF), etc. The techniques for applying fuzzy MCDM methods include: the α-cut method; fuzzy arithmetic operations; possibility and necessity measures; the eigenvector method, etc. A brief description is presented here and details can be seen in Chen and Hwang (1992). 3.2.1.1. Weighting of basic attributes MCDM methods require information about relative importance of attributes or criteria. It is usually established by set of preference weights, which are normalized to a sum of 1. In case of ‘n’ criteria, a set of weights can be written as: WxuT ⫽ (w1,w2,…,wj,…,wn) and

冘 n

wj ⫽ 1.

(9)

j⫽1

The MCDM problem becomes X1

3.2.1. Fuzzy MCDM The applications of fuzzy set theory in various engineering disciplines especially in decision-making have gained a lot of popularity. To perform MCDM using fuzzy-based methods two main steps are followed: 앫 the aggregation of the performance scores with respect to all the attributes for each alternative; and 앫 the rank ordering of the alternative according to aggregated scores. These two steps are referred as final rating and ranking order, respectively. In crisp MCDM, the final rating are expressed as real numbers, therefore rating order can be easily explained. Therefore in traditional crisp MCDM, the first step is relevant only. In fuzzy MCDM problem, the performance score of each alternative are

(8)

A1 D⫽

A2 · · Am

X2

· · Xn

W

冢 冣冢 冣 X11 X12 · · X1n

W1

X21 X22 · · X2n

W2

·

·

· · ·

·

·

·

· · ·

·

Xm1 Xm2 · · Xmn

.

(10)

wn

Hwang and Yoon (1981) described four different techniques for the MCDM weighting schemes, which include: the eigenvector method; weighted least square method; the entropy method and LINMAP. The weighting coefficients represent the relative importance of each attribute. Saaty (1988) has proposed an AHP to estimate the relative weight of each attribute in a group based on a paired comparison. To compare criterion ‘i’ with ‘j’

F.I. Khan et al. / Environmental Modelling & Software 17 (2002) 669–692

the decision-maker can assign the value aij from Table 1. He proposed the following steps to assign the weights: 1. if aij ⫽ t, then aji ⫽ (1 / t), where t ⫽ 0 and i ⫽ j; 2. if i ⫽ j, then aij ⫽ aji ⫽ 1; and 3. construct matrix A ⫽ (aij|i ⫽ 1,…,n; j ⫽ 1,…,n). It is further shown that the eigenvector corresponding to the maximum eigenvalue of matrix A is a cardinal ratio scale for the criteria compared (Saaty, 1988). The eigenvalue problem can be solved by: A·W ⫽ ⌽max·W

(11)

where ⌽max is a scalar corresponding to the maximum eigenvalue and the unit vector ‘W’ corresponding to ⌽max gives the preference weights. Lu and Hu (1999) suggested a simple technique in calculating the weights for the water quality analysis. They used a decision matrix for three different factors: chlorophyll (C); phosphorous (P); and Secchi depth (S) for determining the eutrophication effects. The relative importance of these factors can be assigned using intensity of importance given in Table 1. P C S

冤冥 1

J⫽

1 2 2

P

WT ⫽ [P C S] ⫽ [0.297 0.539 0.164].

677

(14)

A double weighting scheme is proposed in FCP. The other types of weights used in fuzzy MCDM are balancing factor (p). The balancing factor (p) is assigned to groups of indicators to reflect the importance of the maximal deviations, which represents maximum difference, an indicator value and the best value for that indicator. The larger the value, the greater the concern with respect to the maximal deviation. When p ⫽ 1, all deviations are equally weighted, for p ⫽ 2, each deviation receives its importance in proportion to its magnitude (Torno et al., 1988). 3.2.1.2. Converting linguistic terms into fuzzy numbers In cases where we deal with linguistic terms, e.g., good, very poor, excellent etc., a numerical approximation is proposed to convert linguistic terms into corresponding fuzzy numbers. Chen and Hwang (1992) defined eight scales to convert linguistic terms into fuzzy numbers. Two important scales are reported in Fig. 5. The first scale has three levels, whereas the second scale contains five levels. The linguistic terms for first scale are ‘low’, ‘medium’ and ‘high’ in second scale two additional degrees ‘very low’ and ‘very high’ are intro-

(12)

2 1 3 C

S 1 1 1 2 3

A normalized matrix I is determined by normalizing the matrix J column wise. The final weighted vector can be derived from summing the elements of each row of matrix I and normalizing again for this vector.

冤

0.286 0.273 0.333

冥冤 冥 0.892

I ⫽ 0.571 0.546 0.500 ⇒ 1.617 0.143 0.182 0.167

(13)

0.491

After normalizing to 1 eq. (13) becomes Table 1 Linguistic measures of importance based on Saaty’s concept (Sadiq, 2001) Intensity of importance

Definition

1 3 5 7 9 2, 4, 6, 8

Equal importance Weak importance Strong importance Demonstrated importance Absolute importance Intimidate values

Fig. 5.

Conversion of linguistic terms into numerical terms.

678

F.I. Khan et al. / Environmental Modelling & Software 17 (2002) 669–692

duced. Same linguistic terms contain different meaning in different scales. The ‘high’ in first scale means [(0.6,0), (0.8,1.0), (1.0,1.0), (1.0,0)], i.e. most likely interval (when the membership function m(x) is 1) is in between 0.8 and 1 and largest likely interval is in between 0.6 and 1 (when the membership function, m(x) is 0). But in second scale the ‘high’ means [(0.6,0), (0.75,1.0), (0.9,0)]. This reflects the fact that the same linguistic terms may posses different meanings for different occasions. The principle of this system is to pick a scale that contains all verbal terms given by a decision-maker and use the fuzzy numbers in that scale to represent the meaning of the verbal terms. The determination of the number of conversion scales is intuitive. Miller (1965) pointed out that ‘ 7±2’ represents the greatest amount of information that an observer can give about the objects based on absolute judgment. 3.2.2. Grouping of attributes — FCP The FCP is an extension of compromise programming (Bogardi and Bardossy, 1983). Recently this approach is gaining popularity and many researchers have used this approach in environmental decision-making problems (e.g. Sadiq, 2001; James and Lee, 1992; Lee, 1992; Lee et al., 1991; Stansbury et al., 1989). It is difficult to select an appropriate RBLCA strategy when the values of input variables, such as ecological, safety and human health risks, cost, and qualitative terms like technical feasibility, are uncertain. Some MCDM optimization tools including linear multi-objective programming, goal programming, compromise programming, composite programming, multi-attribute utility theory and Electre programming are very popular for this type of problems. Stansbury (1991) has compared some of the decision-making models including MAUF, Electre and FCP and suggested that the choice of selecting appropriate MCDM method depends on the following factors: 1. type of attribute values; qualitative or quantitative; 2. types of alternatives; continuous or discrete options; 3. robustness of results with respect to changes in attribute values; 4. ease of computation; and 5. decision-maker an individual or a group. The compromise programming and composite programming, and multi-attribute utility theory can effectively deal with discrete and continuous options. However, linear multi-objective programming and goal programming can only handle continuous sets of options, and Electre programming can deal with only discrete sets of options (Lee et al., 1991). These MCDM tools are used to assist decision-makers in solving problems involving multiple attributes and conflicting objectives.

The uncertainties inherent in these problems are often treated separately from the decision process itself. The decision analysis is often carried out using point or crisp estimates, meaning no uncertainty within input variables. The FCP is used to assist decision-makers in solving problems of multiple attributes and conflicting objectives. Lee (1992) and Lee et al. (1991) have incorporated such uncertainties in MCDM processes for the management of dredged material disposal and for developing management strategies for nitrate control in drinking water. Stansbury et al. (1989) have developed a methodology for a risk–cost trade-off study of dredged materials for analyzing the ecological, human health and socioeconomic impacts of dredging programs using composite programming. Sadiq (2001) has also applied FCP for disposal management of drilling waste in the offshore environment to determine the best discharge scenario from risk, cost and technical feasibility viewpoints. The FCP is a step-by-step procedure of regrouping of a set of various basic indicators to form a single indicator (Bogardi and Bardossy, 1983). The FCP uses a composite structure of the basic indicators selected for RBLCA. In addition to cost and technical feasibility, risks to both human and non-human populations, environmental damages (global warming, ozone depletion etc.) can also be defined as basic indicators. The first step in FCP is the normalization of the basic indicators. This is necessary because all basic indicators have different units and are difficult to compare in their respective units. At the first level, human health risk (HHR), ecological risk (ER), and safety risk (SR) are determined, which are grouped into environmental risk reduction index (RI). Similarly, global warming potential and ozone depletion indices are determined to get the warming reduction index (WI). At level 2, the warming reduction, acidification reduction (AI), pollution reduction (PI) and risk reduction (RI) indices are grouped as the sink depletion–improvement index (SDI). The same procedure is repeated for other basic indicators, e.g. technical feasibility index of any alternative is developed from ease of operation (EO), efficiency (Eff), status of technology (ST) and control measures requirements (CM). The level 3 sink depletion and resource depletion (RDI) improvement indices are grouped to form the environmental–improvement index. The same procedure is repeated for cost saving index (CI). The grouping for different attributes is performed in steps, until we get the environmental, cost saving and technical feasibility indices to make a trade-off analysis among these conflicting objectives. This trade-off analysis for different alternatives can be made at all hierarchy levels. The system improvement index value at level 5 represents the contribution of environment, cost saving and final technical feasibility of operations. Fig. 6 shows a framework of composite programming procedure. A fuzzy set U is defined as set of objects or elements

F.I. Khan et al. / Environmental Modelling & Software 17 (2002) 669–692

Fig. 6. The framework for FCP.

x as U ⫽ {x}. Therefore a fuzzy set A in U is characterized by set of ordered pairs A ⫽ {(x),mA(x)},∀U苹x, where mA (x) is grade or degree of membership of x in A (Zadeh, 1965). The membership function mA (x) assumes its value [0, 1]. When m(x) ⫽ 0, it does not belong to A and if m(x) ⫽ 1, it indicates x is surely an element in A. The values of the basic indicators are designated by fuzzy numbers to characterize their uncertainties. By defining Zi (x) as a fuzzy number of the ith basic indicator with a trapezoidal membership function of m[Zi (x)], various management alternatives under uncertainty can be evaluated. The confidence level for an uncertain value can be determined using observed or measured variability. Since units of basic indicators are different, the actual value of each basic indicator should be transformed into an index, Si,h(x), using the best or the worst value of the indicator as shown in Fig. 7 (Lee et al., 1991). Using the normalized index values of basic indicators, the level 2 index values, Lj,h (x), of composite indicators can be defined by:

冘 nj

Lj,h(x) ⫽ {

wi,j[Si,h,j(x)]pj}(1/pj)

(15)

i⫽1

where nj is the number of elements in second-level group j; Si,h,j(x) is the basic value for ith basic indicator in the

679

second level group j of basic indicators with membership of h; wi,j is the weight reflecting the importance of each of basic indicator (⌺wi,j ⫽ 1); and pj is the balancing factor for group j. Furthermore, the index values Lk,h(x), of third level composite indicators can be calculated by the index values for second level composite indicators. This procedure is repeated till the final step, which compares three indices: environment; cost saving; and technical feasibility. Selection of weights (w) and balancing factors (p) depend on a single or group of decision-makers. The weights represent the relative importance of each indicator as viewed by a decision-maker, whereas the balancing factors account for maximum deviation of the indicators and limit the ability of one indicator to substitute for another. The balancing factor reflects the strength of presence for a particular objective and its relative importance. A weighting technique is used to group basic indicators into more general groups. The weights are allotted in each group based on their relative importance (Table 1). To determine the ranking for various management alternatives, assume L(x) as a fuzzy number, which is represented by a final composite indicator of alternative x. With the help of two index values, Lh ⫽ 1(x) and Lh ⫽ 0(x), the membership function, m[L(x)], of the fuzzy number can be approximately calculated. For m management alternatives there are m fuzzy numbers, [L(x),x ⫽ 1,2,…,m], to which any ranking method can be applied, e.g. Chen (1985). The Chen’s ranking method (1985) determines the ranking of m fuzzy numbers by using maximizing and minimizing sets. This method has been extensively used in ranking alternative studies (Lee, 1992; Stansbury, 1991; Sadiq, 2001). 3.2.3. Fuzzy ranking methods in decision-making In traditional MCDM methods, all Xij, and Wj (relative importance of attributes) values are taken as crisp values. A utility function U(x1,x2,…,xm) is defined by a decisionmaker. For alternatives Ai, the utility function aggregates its performance ratings Xij into a final rating, Ui. This final rating determines how well one alternative satisfies the decision-maker’s utility. A decision-maker prefers the alternatives with higher final ratings. If the final ratings are real numbers than it is straightforward to decide the best alternative. The alternative performance rating Xij can be crisp, fuzzy, and/or linguistic. When fuzzy data are incorporated into MCDM problem the final ratings are no longer crisp number rather they are fuzzy numbers. The fuzzy numbers are not straightforward to compare. In MCDM applications when the final ratings are fuzzy, different ranking methods can be used to compare these fuzzy utility values. Chen and Hwang (1992) have classified ranking methods into four major groups: preference

680

F.I. Khan et al. / Environmental Modelling & Software 17 (2002) 669–692

Fig. 7.

Transforming the actual value Zi,h(x) into the normalized index value Si,h(x).

relation; fuzzy mean and spread; fuzzy scoring; and linguistic expression. Fuzzy scoring techniques are the most popular. In that Chen’s ranking method (1985) is extensively used for risk–cost trade-off analysis. Another simple and popular method proposed by Yager (1980) which is based on estimating centroidal index. A brief description of Chen (1985) and Yager (1980) methods are given below.

The left utility value, UL(x), for the alternative Ai is defined as

3.2.3.1. Chen’s ranking method (1985) The first step in Chen’s ranking method (1985) is to estimate the final composite L(x) of alternative Ai. The final index value of the composite system is in the form of fuzzy numbers. With the help of index values, Lh ⫽ 1(x) and Lh ⫽ 0(x), the membership, m(x) can be estimated. If there are ‘m’ alternatives, there are m fuzzy numbers [L(x),x ⫽ 1,2,…,m]. The ranking of these ‘m’ fuzzy numbers is done by maximizing and minimizing set. The maximizing set M is a fuzzy subset with membership function mM(L) is given by:

The alternative, which has the highest ordering value, is selected as the best alternative. Chen (1985) has generalized the results [eq. (20)] for trapezoidal fuzzy numbers as shown in Fig. 8.

冤

(L⫺Lmin) , LminⱕLⱕLmax mM(L) ⫽ (Lmax⫺Lmin) 0, OTHERWISE

冥

(16)

UL(X) ⫽ max(min{mG(L),m[L(x)]})

(19)

The total utility or ordering value for the alternative Ai is UT(x) ⫽

UT(x) ⫽ ⫺

UR(x) ⫹ 1⫺UL(x) . 2

冋

1 (di⫺Lmin) ⫹1 2 (Lmax⫺Lmin)⫺(bi⫺di)

(20)

(21)

册

(Lmax⫺ci) . (Lmax⫺Lmin) ⫹ (ai⫺ci)

3.2.3.2. Yager’s (1980) centroidal index ranking method In this method the centroid of fuzzy number is determined. Each geometric center corresponds to x value on the horizontal axis. Yager (1980) has proposed a ranking index as follows

and Lmax ⫽ where Lmin ⫽ min[minLh ⫽ 0(x)] max[maxLh ⫽ 0(x)] for x ⫽ 1,…,m. Then the right utility value, UR(x), for the alternative Ai is defined as UR(X) ⫽ max(min{mM(L),m[L(x)]}).

(17)

The minimizing set G is a fuzzy subset with membership function mG(L) which is given by

冤

冥

(L⫺Lmax) , LminⱕLⱕLmax . mG(L) ⫽ (Lmin⫺Lmax) 0, OTHERWISE

(18) Fig. 8. Schematic of Chen’s ranking method for trapezoidal fuzzy numbers.

F.I. Khan et al. / Environmental Modelling & Software 17 (2002) 669–692

冕 1

xO ⫽

g(x)mi(x)dx

0

冕 1

(22)

mi(x)dx

0

681

urea in India is more than Rs. 8000 per ton (US$ 168.0/ton) while the cost of importing urea from US is about US$116 per ton. The higher cost of production is because of costlier inputs in the form of feedstock of petroleum origin (Roy, 2001). This case study is undertaken with an aim to conduct a detailed risk-based life cycle of each urea production process considering all design constraints (environment, cost and technical feasibility). Before applying the RBLCA to this problem, a brief description of the process is presented below.

where g(x) is treated as weight function measuring the importance of the value x; while the denominator serves as a normalizing factor whose value is equal to the area under the membership function mi. When g(x) ⫽ x (linear weight), eq. (22) gives the xO (geometric center). The value of xO is weighted mean value of fuzzy number L(x) and higher xO values are considered better. Fig. 9 shows a comparison of two fuzzy numbers and their centroids. The software Risk Calc (1995) can be used to estimate the centroid of a fuzzy number. The centroid is generally not the same as the mean except for scalars. Yager’s estimate for ranking fuzzy numbers is not considered very reliable method and later Murakami et al. (1983) have modified this method. Therefore in this research both Chen’s (1985) and Yager (1980) method are used to rank the alternatives.

Ammonia (NH3) is the basic raw material for the urea production. It is synthesized from hydrocarbon feedstock. Urea is produced from a chemical reaction involving NH3 and carbon dioxide (CO2). A simplified block diagram of urea production from different feedstock is presented in Fig. 10. Production of ammonia accounts for ⬎80% of the total energy consumed in urea production. Ammonia production involves feed treatment, steam and air reforming of hydrocarbons, gas purification, and ammonia synthesis (Fig. 10). The major reactions involve in the production of urea are:

4. GreenPro-I application to a urea plant

CH4 ⫹ H2O→CO ⫹ 3H2

4.1. Processes of urea production

CH4 ⫹ air→CO ⫹ 2H2 ⫹ N2 In India, urea is the most common fertilizer being the cheapest source of nitrogen and the most suitable for Indian soil. Presently India has 41 plants producing urea with a total capacity of 20.1 metric ton. Out of them 19 are based on natural gas (alternative 1), 14 on naptha (alternative 2), six on fuel oil (alternative 3) and two on coal, as feedstock (alternative 4). Percentage contributions of each feedstock to the total installed capacity are: natural gas 54.2%, naptha 26.1%, fuel oil 18.21%, and coal 1.54% (Roy, 2001). It is interesting to note that the cost of production of

CO ⫹ H2O→CO2 ⫹ H2

冧

(23)

Carbon dioxide is removed from the product hydrogen through caustic and amine absorption, with final polishing with methanation reaction. The purified hydrogen and nitrogen stream is then converted into ammonia: N2 ⫹ 3H2→2NH3.

(24)

Ammonia is further converted into urea CO2 ⫹ 2NH3→NH4CO2NH2→NH2CONH2

(25)

⫹ H2O.

Fig. 9. Schematic of Yager’s centroid index method for ranking fuzzy numbers.

Production of NH3 involves large emissions of CO2. In the production of NH3, hydrogen is chemically separated from the hydrocarbon feedstock and combined with nitrogen to produce NH3. A chemical solution (scrubber) is added which absorbs some of the CO2 produced in this reaction. After this, the remaining CO2 is converted into methane in a methanator, because methane is more inert as compare to CO2. The CO2 percentage on dry mole basis in synthesis gas varies from 28.6% (Lurgi process) to 6.2% (Koppers–Totzec process) and in Winkler process it is 20.5% CO2 .The sulfur and CO2 are removed from the synthesis gas before it is methanated. The scrubbing solution is then passed through a process of re-generation. This process releases the absorbed CO2 into the atmosphere. Emissions of CO2

682

F.I. Khan et al. / Environmental Modelling & Software 17 (2002) 669–692

Fig. 10. The urea manufacturing process.

depend upon the type of hydrocarbon used for the production of ammonia. SO2, NOX, CO and non-methane volatile organic compound (VOC) may also be emitted in addition to CO2. In the production of urea there is not much CO2 emissions as almost 99% of both CO2 and NH3 are converted to urea (Holmstrom and Comer, 1999). GreenPro-I methodology has been employed to evaluate various design option of the urea manufacturing. Detailed results are discussed for natural gas based process whereas summary of results is presented for other production options (naptha, stock, fuel oil, and coal as feed stocks) as well. 4.2. RBLCA The objective of this study is to evaluate available option of urea manufacturing considering all the possible constraints. As per section 3.1.1, boundaries and scope of the problem have been defined. Fig. 11 illustrates the boundary of the problem. This study deals with a process selection and a design problem, therefore cradle-to-gate approach is proposed. The subsequent impacts of product post-manufacturing are not considered here. In inventory and risk/impact analyses (sections 3.1.2 and 3.1.3), nine main environmental indicators for six

main processes are considered. Among these various environmental indicators RD, GWPI and environmental risks are of main concern. The RD is important due to the utilization of non-renewable resources such as natural gas and ores whereas the GWPI is due to large release of CO2 and CH4. The environmental risk is important because it accounts for adverse human and ecological effects in their respective communities. Fig. 12 illustrates the resource depletion potential for all the four options of urea production. It is evident from the plot that on average resource consumption is large in case of coal-based process, which is then followed by fuel oil and naptha. Though RD for natural gas option is also high but this value is comparatively lower than the other three alternatives. Detailed results of environmental impacts for various activities in urea production are plotted in Fig. 13. It is evident that natural gas production and transport, ore mining and processing and energy production are three major activities contributing to the RD. Ammonia production and energy production are the two major processes contributing to GWPI. Energy production is contributing about 30% while ammonia production about 25% of the total GWPI. Among various sources, CO2 is a major pollutant contributing about 78%, followed by methane 12%, and oxides of nitrogen and others about 4% of the total GWPI. After GWPI,

F.I. Khan et al. / Environmental Modelling & Software 17 (2002) 669–692

Fig. 11.

Simplified life cycle block diagram of urea production.

Fig. 12.

RD details for each process option.

683

684

F.I. Khan et al. / Environmental Modelling & Software 17 (2002) 669–692

Fig. 13.

Contribution of environmental impact by different activities of the urea production life cycle.

POI and OD are the next parameters of concerns. They are caused mainly due to light hydrocarbons (VOCs) release at various processes. For OD, natural gas production, transportation and ammonia production approximately 45, 25 and 15% contribute energy production, respectively. In case of POI natural gas production and transport, energy production, and ammonia production are contributing with the same order of magnitude (Fig. 13). In the case of risk factors — HHR, SR and ER — a similar situation is expected. Ammonia and urea production processes together, are major contributors to these risks. These processes contribute about 38% (average) of the total risk, followed by natural gas production, transport and energy production. Mining process is the next most significant contributor (20%). The results are shown in Fig. 13. It is clear from above discussion that ammonia and urea productions are the major contributors to the total environmental burden. These processes are further studied in-depth for subsequent analysis (Fig. 14). The reforming unit is the main contributor (about 35%) to the total GWPI which is then followed by Hydro-desulfurization and ammonia reactor. In case of OD and POI reforming unit is the main contributor. CO convertor unit and urea reactor are main contributors to SMD and CTWM. Ammonia reactor is the major source of concern for human health, safety and ecological risk, contributing 25, 30 and 30%, respectively. A meticulous study of safety risk reveals that ammonia reactor is the

most vulnerable (contributes 30% of the total risk). It is followed by hydro desulfurizing (15%), reforming unit (15%) and urea reactor (15%) (Fig. 15). In safety risk analysis, the hazards including fire, explosion and toxic release are of same importance, on average basis (Fig. 15). In short it can be concluded that, the ammonia reactor, reforming and hydro-desulfurization units are the most important units in ammonia and urea production processes. Improvement in the operational performance of these units would improve the environmental quality of the whole production process significantly. The human health risk is one of the most important concern therefore a detailed investigation has been conducted for HHR. The analysis shows that the fuel oil based process poses a maximum HHR followed by naptha, coal and natural gas based process. Among various exposure routes, air exposure route is of major concern, which contributes about 55% to the total HHR followed. It is followed by exposures through water (35%) and soil (10%). The chemicals of concern and their respective contributions through different exposure routes are presented in Fig. 16. It is evident that ammonia and VOC in air, urea, heavy metals and acids in water and soil are of major source of concern. According to the developed methodology of GreenPro-I, now the estimated environmental burdens (discussed above), technical and economic constraints will be transformed into a design problem (section 3.1.4). The design problem is solved and results are sub-

F.I. Khan et al. / Environmental Modelling & Software 17 (2002) 669–692

Fig. 14.

Contribution of environmental burden by different units of the urea and ammonia plant.

Fig. 15.

Contribution in SR by different units of ammonia and urea production.

685

686

F.I. Khan et al. / Environmental Modelling & Software 17 (2002) 669–692

Fig. 16.

HHR through three exposure routes caused by different chemicals.

sequently transformed into a MCDM problem. These results are discussed in the next sections. 4.3. Problem solution and risk-based MCDM A trade-off analysis for selecting the best design and management alternative from minimum cost, best technical feasibility and minimum environmental damage viewpoints, is a challenging task. These conflicting objectives are grouped together with their respective importance/weightage to determine the final system index, which should be robust for different weighting schemes. The best alternatives will be the one, which has the highest value of system improvement index (SI). The attributes considered in this study are a long list of basic indicators, which include technical feasibility and cost other than environmental parameters discussed before. Table 2 has summarized these basic indicators. The values and units of basic indicator are reported. Due to uncertainty in the processes, estimation and raw material quality, minimum, maximum likely (MLV) and maximum values of the basic indicators are reported. The resource depletion is expressed as kg of natural gas/kg of urea production, global warming potential is expressed by equivalents of kg of CO2 produced/kg of urea production and similarly, other basic indictors are defined based on unit urea production for each alternative. Some of the basic indicators are expressed in quali-

tative terms. The technical feasibility and cost saving indicators are assigned numerical scores based on a scale given in Fig. 5. This scale converts linguistic terms into numerical scores. The values obtained from this scale are unitless. Therefore, to convert all basic indicators into unitless terms, normalization is required as explained in an earlier section of the paper. Basic indicators are normalized based on their reported BEST and WORST values as presented in Table 3. This normalization is done considering criterion WORST ⬎ BEST and formulation given in Fig. 7. The values of basic indicators for other attributes are defined qualitatively based on a criterion of BEST ⬎ WORST. Here, the higher numbers mean better value. These values do not require normalization because they have already been defined in unitless terms, e.g. technical feasibility. Table 4 presents all basic indicators into their normalized form. The normalized values of the environmental basic indicators show that higher the value more will be the environmental protection. The advantage of converting all basic indicators into their normalized forms is that they are unitless and can be group together to form a general group depending on its relative importance. After normalization of basic indicators, the next step in FCP is to decide the weights of attributes based on their relative importance from a decision-maker point of

F.I. Khan et al. / Environmental Modelling & Software 17 (2002) 669–692

687

Table 2 The basic indicators for various management alternatives Basic indicators

Alternative 1

Resource depletion (RD)a Global warming potential (GWPI)a Ozone depletion (OD)a Acidification potential (AP)a Photochemical oxidation (POI)a Human health risk (HHR) Ecological risk (ER) Safety risk (SR) Solid mass disposal (SMD)a Controlled toxic water mass (CTWM)a Ease of operation (EO) Efficiency (Eff) Status of technology (ST) Control measures (CM) Working capital (WC) Operation and maintenance (OM) Capital investment (CI) a

Alternative 2

Alternative 4

Min.

MLV

Max.

Min.

MLV

Max.

Min.

MLV

Max.

Min.

MLV

Max.

0.630 7.50 0.078 0.250 0.005 4.000 4.000 20.0 0.340 0.375

0.830 10.00 0.104 0.333 0.007 5.500 7.500 100.0 0.450 0.500

1.040 12.50 0.130 0.416 0.009 7.000 11.00 150.0 0.560 0.625

0.640 9.00 0.078 0.300 0.002 4.000 4.000 15.0 0.340 0.375

0.850 12.00 0.104 0.400 0.003 5.500 7.500 80.00 0.450 0.500

1.063 15.00 0.130 0.500 0.004 7.000 11.000 125.00 0.560 0.625

0.660 6.00 0.078 0.300 0.002 5.000 4.000 15.00 0.488 0.375

0.880 8.00 0.104 0.400 0.003 6.50 7.50 70.00 0.650 0.500

1.100 10.00 0.130 0.500 0.004 8.000 11.00 110.0 0.813 0.625

0.713 11.25 0.150 0.488 0.004 5.000 6.000 10.0 0.938 0.563

0.950 15.00 0.200 0.650 0.005 7.500 10.50 60.0 1.250 0.750

1.188 18.75 0.250 0.813 0.006 10.00 15.00 100.0 1.563 0.938

0.600 0.600 0.600 0.300 0.100 0.300 0.300

0.750 0.750 0.750 0.500 0.250 0.500 0.500

0.900 0.900 0.900 0.700 0.400 0.700 0.700

0.300 0.600 0.600 0.300 0.300 0.300 0.100

0.500 0.750 0.750 0.500 0.500 0.500 0.250

0.700 0.900 0.900 0.700 0.700 0.700 0.400

0.100 0.300 0.300 0.100 0.300 0.100 0.300

0.250 0.500 0.500 0.250 0.500 0.250 0.500

0.400 0.700 0.700 0.400 0.700 0.400 0.700

0.100 0.000 0.000 0.000 0.300 0.000 0.100

0.250 0.100 0.100 0.100 0.500 0.100 0.250

0.400 0.200 0.200 0.200 0.700 0.200 0.400

Value/kg of urea production.

Table 3 The normalised values of the basic indicators for various management alternatives Basic indicators

BEST

WORST

Resource depletion (RD)a Global warming potential (GWPI)a Ozone depletion (OD)a Acidification potential (AP)a Photochemical oxidation (POI)a Human health risk (HHR) Ecological risk (ER) Safety risk (SR) Solid mass disposal (SMD)a Controlled toxic water mass (CTWM)a

0.630 6.000

1.188 18.750

0.078 0.250 0.002

0.250 0.813 0.009

4.000 4.000 10.00 0.340 0.375

10.000 15.000 150.00 1.563 0.938

a

Alternative 3

Value/kg of urea production.

view. Table 5 presents the relative importance matrix for grouping basic indicators into generalized groups. The human health, ecological and safety risk indices are grouped to develop a risk reduction index (RI). The higher value of risk reduction index shows lower risk. It is now consistent with all other environmental basic indicators whose higher values depict improvement in the system. The grouping is performed at four levels. At first levels all basic indicators are grouped. The weights are estimated based on their relative importance. The Lu and Hu (1999) method is used to estimate the w for each

attribute. At second level, WI, AI, RI and PI are grouped to determine sink depletion–improvement index (SDI). Similarly at level 3, resource depletion and sink depletion–improvement indices are grouped to determine the environmental–improvement index (EI). At this level all indices including the technical feasibility, cost saving and final environmental indices are grouped to determine the SI index. After assigning the weights to the basic indicators and generalized groups at various levels, the next step in FCP is to estimate the values of the indices. Triangular fuzzy numbers are used to define basic indicators (to account for uncertainty) therefore the estimated indices will also be in the form of a fuzzy number. The approach used in this paper emphasises on risk assessment therefore the estimated indices for risk reductions are presented in Fig. 17 for various alternatives. The risk reduction index is developed from human health, ecological damage and safety risks due to fire, spills and accidents. Fig. 18 shows the results of cost saving indices. The alternative 2 is the best management (cost saving) alternative in comparison to other options considering MLVs and base of the fuzzy number. The MLV of option is nearest to 1.0 representing high cost saving, and the base of the option 2 fuzzy number is the smallest representing smaller uncertainty in the estimates. Option 2 is followed by options 1, 3 and 4. The environmental basic indicators are grouped at four levels, and finally establish the EI index. Fig. 19 illustrates the results of environmental indices in which all four alternatives are compared. The Fig. 19 shows

688

F.I. Khan et al. / Environmental Modelling & Software 17 (2002) 669–692

Table 4 The normalised values of the basic indicators for various management alternatives Normalised values of basic indicators

Resource depletion (RD) Global warming potential (GWPI) Ozone depletion (OD) Acidification potential (AP) Photochemical oxidation (POI) Human health risk (HHR) Ecological risk (ER) Safety risk (SR) Solid mass disposal (SMD) Controlled toxic water mass (CTWM) Ease of operation (EO) Efficiency (Eff) Status of technology (ST) Control measures (CM) Working capital (WC) Operation and maintenance (OM) Capital investment (CI)

Alternative 1

Alternative 2

Alternative 3

Alternative 4

Min.

MLV

Max.

Min.

MLV

Max.

Min.

MLV

Max.

Min.

MLV

Max.

0.265 0.490 0.698 0.705 0.000 0.500 0.364 0.000 0.820 0.556

0.642 0.686 0.849 0.853 0.286 0.750 0.682 0.357 0.910 0.778

1.000 0.882 1.000 1.000 0.571 1.000 1.000 0.929 1.000 1.000

0.224 0.294 0.698 0.556 0.714 0.500 0.364 0.179 0.820 0.556

0.606 0.529 0.849 0.734 0.857 0.750 0.682 0.500 0.910 0.778

0.982 0.765 1.000 0.911 1.000 1.000 1.000 0.964 1.000 1.000

0.158 0.686 0.698 0.556 0.714 0.333 0.364 0.286 0.613 0.556

0.552 0.843 0.849 0.734 0.857 0.583 0.682 0.571 0.747 0.778

0.946 1.000 1.000 0.911 1.000 0.833 1.000 0.964 0.879 1.000

0.000 0.000 0.000 0.000 0.429 0.000 0.000 0.357 0.000 0.000

0.427 0.294 0.291 0.290 0.571 0.417 0.409 0.643 0.256 0.334

0.851 0.588 0.581 0.577 0.714 0.833 0.818 1.000 0.511 0.666

0.600 0.600 0.600 0.300 0.100 0.300 0.300

0.750 0.750 0.750 0.500 0.250 0.500 0.500

0.900 0.900 0.900 0.700 0.400 0.700 0.700

0.300 0.600 0.600 0.300 0.300 0.300 0.100

0.500 0.750 0.750 0.500 0.500 0.500 0.250

0.700 0.900 0.900 0.700 0.700 0.700 0.400

0.100 0.300 0.300 0.100 0.300 0.100 0.300

0.250 0.500 0.500 0.250 0.500 0.250 0.500

0.400 0.700 0.700 0.400 0.700 0.400 0.700

0.100 0.000 0.000 0.000 0.300 0.000 0.100

0.250 0.100 0.100 0.100 0.500 0.100 0.250

0.400 0.200 0.200 0.200 0.700 0.200 0.400

Table 5 Weighting schemes for FCP Basic Indicators

Level

Relative importance

w

p

Groups

GWPI OD AP POI HHR ER SR SMD CTWM EO Eff. ST CM WC OM CI WI AI RI PI SDI RDI EI TFI CI

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 3 3 4 4 4

1 1 1 2 1 4/3 1 1 1 1 2 1 1 1 1 4/3 3/2 1 3/2 1 1 1/4 1 2/5 3/5

0.5 0.5 0.3 0.7 0.3 0.4 0.3 0.5 0.5 0.2 0.4 0.2 0.2 0.3 0.3 0.4 0.3 0.2 0.3 0.2 0.8 0.2 0.5 0.2 0.3

1

Warming improvement index (WI)

1

Acidification improvement index (AI)

2

Risk reduction index (RI)

1

Pollution reduction index (PI)

2

Technical feasibility index (TFI)

1

Cost saving index (CI)

1

Sink depletion improvement index (SDI)

1

Environmental improvement index (EI)

2

System improvement index (SI)

that alternatives 2 and 3 give almost the same results. The MLV is same for both alternatives but base of the fuzzy number is bigger for alternative 2 in comparison to 3.

The final system improvement indices are shown in Fig. 20. The system improvement index shows that alternative 2 is the best management alternative, which is followed 1, 3 and 4. The alternative 1 is close to alterna-

F.I. Khan et al. / Environmental Modelling & Software 17 (2002) 669–692

Fig. 17.

Fig. 18.

Fig. 19.

Comparison of RI for various alternatives.

Comparison of CI for various alternatives.

Fig. 20.

Comparison of SI indices for various alternatives.

To make a trade-off among conflicting objectives environmental improvement and cost saving, a twodimensional fuzzy numbers are shown in Fig. 21. The x-axis represents the cost saving index (CI) whereas the y-axis represents the EI index. The base of the fuzzy numbers two-dimensional, where a line of rectangle parallel to x-axis represents the cost saving index range and line parallel to y-axis represents the environmental improvement index range. The most likely values are plotted as point values (Fig. 21). It can be noticed that alternative 2 rectangle is smaller in size and close to the ideal conditions (1.0,1.0). Therefore, considering cost saving and environmental improvement viewpoints, alternative 2 promises better results than other options. The next step in FCP is to apply formal ranking techniques for determining ranking orders for available management alternatives. Two methods are used here, namely, Chen (1985) and Yager (1980) as described in section 3.2.3. Table 6 summarizes the utility and centroidal index values, and ranking orders of alternatives. Both methods determine the same ranking order, which can also be predicted by visual observation. From utility

Comparison of EI indices for various alternatives.

tive 2. The alternative 4 is the least desirable design alternative (coal as feed stock based process). The uncertainty in estimates of alternatives 1, 2 and 3 are approximately same which can be observed from base of fuzzy numbers (when membership function is 0).

689

Fig. 21. Comparison of CI and EI indices.

690

F.I. Khan et al. / Environmental Modelling & Software 17 (2002) 669–692

Table 6 Ranking of alternatives for trial 1 Alternatives

Definition

Ut(x)1

Centroid

Rank

1

Natural gas based process Naptha based process Fuel oil based process Coal based process

0.668

0.611

2

0.711 0.599 0.227

0.648 0.528 0.281

1 3 4

2 3 4

the results have already been discussed. The trial 2 represents the case when cost saving is given more importance than environmental improvement. The trial 3 represents the case when environment and cost saving are given equal importance. The final utility values for all three trials are plotted in Fig. 22. The result shown in Fig. 22 proves that option 2 is the best management alternative, which is followed by alternatives 1, 3 and 4. 5. Summary and conclusions

values, it can be concluded that alternatives 1 and 2 are very close to each other. The last phase of FCP is to conduct a sensitivity analysis. The assignment of relative importance and balancing factors (p) depend on subjective human judgement. To avoid the human bias, different weighting schemes are used for this purpose and analysis is repeated. Table 7 shows the alternate weighting schemes to re-evaluate the system improvement indices. In trial 1 the environment was given first priority and

The RBLCA methodology GreenPro-I, is a systematic step-by-step approach to estimate environmental risks and impacts associated with products, processes and services. It is a process to evaluate environmental burdens by quantifying energy and materials used and wastes released into the environment, and to identify/evaluate opportunities to effect environmental improvements. The assessment may include the entire life cycle of the product, process or activity encompassing extraction and

Table 7 Weighting schemes for sensitivity analysis Trials

Groups

Relative importance

w

Ut(x)1 (pro-environment)

EI CI TFI EI CI TFI EI CI TFI

1 3/5 2/5 3/5 1 2/5 1 1 1/2

0.5 0.3 0.2 0.3 0.5 0.2 0.4 0.4 0.2

Ut(x)2 (pro-cost saving)

Ut(x)3 (equal weightage to cost saving and environment)

Fig. 22.

Sensitivity analysis for estimating the robust ranking order.

F.I. Khan et al. / Environmental Modelling & Software 17 (2002) 669–692

processing of raw materials, manufacturing, transportation and distribution, use, recycle, and final disposal. The GreenPro-I is proposed as a decision-making tool for designer, regulatory agencies, and business organizations. The GreenPro-I overcomes many of the problems faced in the conventional approaches and establishes a link between the environmental risks/impacts, cost, and technical feasibility of processes. It offers an extended environmental perspective, considering risks/impacts from resource extraction to the end product use and disposal. The GreenPro-I relates these effects to the mass and energy flows into, out of, and within the process. The present work has focused on the development of process selection and design methodology considering assessment and minimization of risks/impacts of process systems by embedding the conventional LCA principles within process design and decision-making framework. It has implications to process synthesis as it includes environmental objectives together with technical feasibility and economics at the design stage to determine cost efficient solutions. Authors believe that employing the GreenPro-I with a process design and decision-making would yield a best management alternative among options available and promise a cleaner and greener process design. The previous work GreenPro, was good and effective in design, but its application was restricted at early design stage due to extensive computational and large data requirements. In this paper, the authors have modified the earlier GreenPro methodology and overcame these limitations. The revised methodology, GreenProI, is an efficient, easy to use and robust (comparatively less sensitive to the reliability of the data) in decisionmaking. A detailed description of the methodology and tools/techniques are presented with the help of a urea production case study.

References Azapagic, A., 1999. Life cycle assessment and its application to process selection design and optimization. Chemical Engineering Journal 73, 1–21. Azapagic, A., Clift, R., 1999a. Linear programming as a tool in life cycle assessment. International Journal of LCA 4 (6), 305–316. Azapagic, A., Clift, R., 1999b. Allocation of environmental burdens in multiple-function systems. Journal of Cleaner Production 9 (2), 101–119. Azapagic, A., Clift, R., 1999c. The application of life cycle assessment to process optimization. Computers and Chemical Engineering 23, 1509–1526. Azapagic, A., Clift, R., 1999d. Life cycle assessment as tool for improving process performance: a case study on boron products. International Journal of LCA 4 (3), 133–142. Bogardi, I., Bardossy, A., 1983. Application of MCDM to geological exploration. In: Henson, P. (Ed.), Essays and surveys on multiple criterion decision-making. Springer-Verlag, New York. Cabezas, H., Bare, C.J., Mallick, S.K., 1999. Pollution prevention with

691

chemical process simulators: the generalized waste (WAR) algorithm-full version. Computers in Chemical Engineering 23, 625. Cabezas, H., Bare, C.J., Mallick, S.K., 1997. Pollution prevention with chemical process simulators: the generalized waste reduction (WAR) algorithm. Computers in Chemical Engineering 21, s305–s310. Chen, S.H., 1985. Ranking fuzzy numbers with maximizing set and minimizing set. Fuzzy Set Systems 17 (2), 113–129. Chen, S.J., Hwang, C.L., 1992. Fuzzy multiple attribute decision-making. Springer-Verlag, New York. Choong, P.M., Sharratt, P.N., 2000. A holistic approach to environmental issues in process development and design. IChemE Symposium Series 147, 41–54. El-Halwagi, M.M., Manousiouthakis, V., 1989a. Synthesis of massexchange networks. American Institute of Chemical Engineering Journal 35 (8), 1233–1244. El-Halwagi, M.M., Manousiouthakis, V., 1989b. Design and Analysis of Mass-exchange Networks with Multi Component Targets. American Institute of Chemical Engineering Annual Meeting, San Francisco. El-Halwagi, M.M., Manousiouthakis, V., 1990a. Automatic synthesis of mass-exchange networks with single component targets. Chemical Engineering Science 45 (9), 2813–2831. El-Halwagi, M.M., Manousiouthakis, V., 1990b. Simultaneous synthesis of mass-exchange and regeneration networks. American Institute of Chemical Engineering Journal 36 (8), 1209–1219. El-Halwagi, M.M., Spriggs, H.D., 1998. Employ mass integration to achieve truly integrated process design. Chemical Engineering Progress August, 22–44. El-Halwagi, M.M., Noureldin, M.B., 1999. Interval-based targeting for pollution prevention via mass integrating. Computers and Chemical Engineering 23, 1527–1543. Fava, J., Denison, R., Jones, B., Curran, M.A., Vigon, B., Selke, B., Barnum, J., 1991. A Technical Framework for Life Cycle Assessment. SETAC North America Office, 1010 North 12th Avenue, Pensacola, FL. Flower, J.R., Bikes, S.C., Johnson, S.W., 1993. The graphical mass balance in the early design of clean processes. Transactions IChemE Part B 71, 194–198. Hernandez, O.R., Pistikopoulos, E.N., Livingston, A.G., 1998. Waste treatment and optimal degree of pollution abatement. Environmental Progress 17 (4), 270–276. Holmstrom, K., Comer, P., 1999. Risk-based Verification of Green House Gas Emissions for Emission Reduction and Trading. Det Norske Veritas, Houston, TX. Hwang, C.L., Lin, M.J., 1987. Group Decision-making under Multiple Criteria. Springer-Verlag, New York. Hwang, C.L., Yoon, K., 1981. Multiple Attribute Decision-Making — Methods and Applications, A State of the Art Survey. SpringerVerlag, New York. Jacobs, M., 1991. The Green Economy — Environment, Sustainable Development and Politics of the Future. Pluto Press, London. James, H.P., Lee, Y.W., 1992. Selection of design/build proposal using fuzzy-logic system. Journal of Environmental Engineering, ASCE 118 (2), 303–317. Khan, F.I., 2001. Development of maximum credible accident scenarios for realistic and reliable risk assessment. Chemical Engineering Progress 97 (11), 56–64. Khan, F.I., Natrajan, B.R., Revathi, P., 2001. GreenPro: a new methodology for cleaner and greener process design. Journal of Loss Prevention in the Process Industries 14, 307–328. Kniel, G.E., Delmarco, K., Petrie, J.G., 1996. Life cycle assessment applied to process design: environmental and economic analysis and optimization of a nitric acid plant. Environmental Progress 15 (4), 221–228. Lee, Y. W., 1992. Risk assessment and risk management for nitrate contaminated groundwater supplies. Ph.D. Dissertation, Depart-

692

F.I. Khan et al. / Environmental Modelling & Software 17 (2002) 669–692

ment of Civil Engineering, University of Nebraska, Lincoln, Nebraska. Lee, Y.W., Bogardi, I., Stansbury, J., 1991. Fuzzy decision-making in dredged material management. Journal of Environmental Engineering, ASCE 117 (5), 614–630. Lu, R.S., Hu, J.Y., 1999. Analysis of reservoir water quality using fuzzy synthetic evaluation. Stochastic Environmental Research and Risk Assessment 13, 327–336. Mallick, S.K., Cabezas, H., Bare, J.C., Skidar, S.K., 1996. A pollution reduction methodology for chemical process simulators. Industrial Engineering and Chemical Research 35, 4128. Mann, M.K., Spath, P.L., 1997. Life cycle assessment of a biomass gasification combined-cycle system. Report No. NREL/TP-43023076, National Renewable Energy Laboratory, Golden, CO. Miller, G.A., 1965. The Magic Number Seven, Plus or Minus Seven. Psychological Review 63, 81–97. Murakami, S., Maeda, S., Imamura, S., 1983. Fuzzy decision analysis on the development of centralized regional energy control system. In: IFAC Symposium on Information, Knowledge Representation and Decision Analysis, Marseilles, France pp. 363–368. Overcash, M., 1986. Techniques for Industrial Pollution Prevention — A Compendium Waste Minimisation. Lewis Publishers, USA. Pistkopoulos, E.N., Stefenis, S.K., Livinhston, A.G., 1994. Pollution prevention vis process and product modifications. AIChE Symposium Series 303, 140–149. Risk Calc., 1995. Risk Assessment with Uncertain Numbers, v. 3.0. RAMAS, Ecological and Environmental Software, Setauket, NY 11733. Rossiter, A.P., 1995. Waste Minimisation through Process Design. McGraw Hill, New York. Roy, A.K., 2001. Why not coal-based fertilizer plants? EPW Commentary http://www.epw.org.in/36-21/commm4.htm (last verified on September 14, 2001). Saaty, T.L., 1988. Multicriteria Decision-Making: The Analytic Hierarchy Process. University of Pittsburgh, Pittsburgh, PA. Sadiq, R., 2001. Drilling waste discharges in the marine environment: a risk based decision methodology. Ph.D. Thesis, St. John’s, Memorial University of Newfoundland, Canada. Spath, P.L., Mann, M.K., 2000. Life cycle assessment of a natural gas

combined-cycle power generation system. Report No. NREL/TP570-27715, National Renewable Energy Laboratory, Golden, CO. Spath, P.L., Mann, M.K., Kerr, D.R., 1999. Life cycle assessment of coal-fired power production. Report No. NREL/TP-570-25119, National Renewable Energy Laboratory, Golden, CO. Stansbury, J.S., 1991. Risk–cost analysis under uncertainty for dredged material disposal. Ph.D. Thesis, Interdepartmental Area of Engineering, University of Nebraska, Lincoln, NE. Stansbury, J., Bogardi, Kelly W.E., 1989. Risk–cost, analysis for management of dredge material. In: Risk-Based Decision-Making in Water Resources, Proceedings of the 4th Conference Sponsored by the Engineering Foundation, American Society of Civil Engineers, Reston, Virginia. Stephanopolous, G., Linninger, A.A., Stephanopolous, E., Ali, S.A., Han, C., 1994. Generation and Assessment of Batch processes with Ecological considerations. Paper presentation at ESCAPE-5 Conference, Bled, Slovenia. Todd, J.A., Curran, M.A., 1999. Streamlined Life Cycle Assessment: A Final Report from the SETAC North America Streamlined LCA Workgroup. SETAC North America Office, 1010 North 12th Avenue, Pensacola, FL. Torno, H.C., Hartman, L., Bogardi, I., Verhoog, F.H., Higler, L.W.G., 1988. Training Manual for the Environmental Evaluation of Water Resources Management and Development. U.N. Educational, Scientific, and Cultural Organisation, Paris. Tukker, A., 2000. Life cycle assessment as a tool in environmental impact assessment. Environmental Impact Assessment Review 20, 435–456. USEPA 1995. Life Cycle Assessment: A Conceptual Framework, Key Issues, and Summary of Existing Methods, EPA/530/R-95/011. USEPA, Research Triangle park, NC. Wang, Y., Smith, R., 1994. Wastewater minimisation. Chemical Engineering Science 49 (7), 981–1006. Yager, R.R., 1980. On choosing between fuzzy subsets. Kybernetes 9, 151–154. Young, D.M., Cabezas, H., 1999. Designing sustainable processes with simulation: the waste reduction (WAR) algorithm. Computer and Chemical Engineering 23, 1477–1491. Zadeh, L.A., 1965. Fuzzy sets. Information and Control 8 (3), 338– 353.