Optimization of recovery processes for multiple economic and environmental criteria

OPTIMIZATIO OF RECOVERY PROCESSES FOR MULTIPLE ECO OMIC A D E VIRO ME TAL CRITERIA LEE SU-QI ELAI E ATIO AL U IVERSITY OF SI GAPORE 2009 OPTIMIZATIO OF RECOVERY PROCESSES FOR MULTIPLE ECO OMIC A D E VIRO ME TAL CRITERIA LEE SU-QI ELAI E (B. Eng. (Hons.), US) A THESIS SUBMITTED FOR THE DEGREE OF MASTER OF E GI EERI G DEPARTME T OF CHEMICAL A D BIOMOLECULAR E GI EERI G ATIO AL U IVERSITY OF SI GAPORE 2009 Acknowledgements This Masters thesis would never have been possible without the encouragement given by the people around me. It gives me great pleasure to be able to express thanks for their unconditional support. Firstly, I would like to extend my gratitude to my supervisor, Prof. G. P. Rangaiah, from the Department of Chemical and Biomolecular Engineering in the National University of Singapore (NUS). I am awed with his drive for research, and his constant advice and guidance have brought me to where I am today. He has given me many opportunities to pursue several works and to prepare manuscripts, pushing me to my maximum potential. I am indebted to him more than he knows. My lab mates – Masuduzzaman, Suraj, Lakshmi, Haibo and Sharliza – have also made my journey in NUS meaningful. The technical knowledge exchanged between one another enhanced our research capabilities. Also, the conversations and jokes shared have made the office livelier and the journey of pursuing graduate studies endurable. Special mention goes to my friends – Stella, Ming Juan and Phyllis. With them, I could always relive the undergraduate moments which were the most eventful moments of my life. Last, but not least, I would like to thank my parents and my boyfriend, Jia Le. When I am feeling down, their patience and understanding would lift my spirits up miraculously. Their unwavering love and concern have enabled me to complete my thesis. Elaine Lee January 2009 Table of Contents Page v Summary omenclature vii List of Tables xi List of Figures xiii 1. Introduction 1 1.1. Optimization of Chemical Processes 1 1.2. Objectives in Process Optimization 3 1.3. Motivation and Scope of Work 4 1.4. Organization of the Thesis 8 10 2. Sustainability 2.1. Introduction 10 2.2. Sustainability Indicators 13 2.2.1. Economic Indicators 14 2.2.1.1. Cost Estimation 14 2.2.1.2. Profitability Criteria 16 2.2.2. Environmental Indicators 19 2.2.2.1. Environmental Impact 20 2.2.2.2. Environmental Efficiency 22 2.2.3. Societal Indicators 23 2.2.4. Energy Sustainability 24 2.3. Steps Taken Towards Sustainability 25 i 2.3.1. Integrated Networks 25 2.3.2. Green Chemistry 27 2.4. Summary 3. Environmental Impact 28 30 3.1. Introduction 30 3.2. Overview of Environmental Concerns 32 3.3. Environmental Metrics 34 3.3.1. Material Metrics 35 3.3.2. Energy Metrics 38 3.4. Environmental Impact 40 3.4.1. Toxicity Potentials 40 3.4.2. Global Warming Potential 42 3.4.3. Ozone Depletion Potential 45 3.4.4. Photochemical Oxidation Potential 46 3.4.5. Acidification Potential 48 3.4.6. Eutrophication Potential 49 3.5. Aggregated Indicator 50 3.6. Conclusions 55 4. Process Applications Studied for Economic and Environmental Criteria 57 4.1. Introduction 57 4.2. Economic and Environmental Criteria 58 4.2.1. Petroleum Refining and Petrochemicals 58 4.2.2. Biotechnology, Pharmaceutical and Chemicals 64 ii 4.2.3. Downstream Processing 65 4.2.4. Energy Systems and Heat Integration 75 4.3. Environmental Criteria 78 4.3.1. Petrochemicals 78 4.3.2. Biotechnology, Pharmaceutical and Chemicals 79 4.3.3. Downstream Processing 79 4.3.4. Energy Systems 84 4.4. Conclusions 84 5. Optimization of Recovery Processes 87 5.1. Introduction 87 5.2. VOC Recovery 87 5.2.1. Background 88 5.2.2. The Present Study 89 5.2.3. Operation Optimization 91 5.2.3.1. Case A: Bi-Objective Optimization 92 5.2.3.2. Case B: Optimization for Many Objectives 94 5.2.3.3. Case C: Optimization for Several Objectives 98 5.2.4. Design Optimization 99 5.2.4.1. Case D: Bi-Objective Optimization 101 5.2.4.2. Case E: Optimization for Many Objectives 105 5.2.4.3. Case F: Optimization for Several Objectives 108 5.3. Solvent Recovery 5.3.1. Background 109 110 iii 5.3.2. Design Optimization 111 5.3.2.1. Case G: Bi-objective Optimization 112 5.3.2.2. Case H: Optimization for Many Objectives 116 5.4. Conclusions 6. Ranking of Pareto Solutions 119 122 6.1. Introduction 122 6.2. Net Flow Method 124 6.2.1. Ranking of Solutions for VOC Recovery 128 6.2.2. Ranking of Solutions for Solvents Recovery 129 6.3. Conclusions 7. Conclusions and Recommendations 131 133 7.1. Conclusions 133 7.2. Recommendations for Further Study 135 References 136 Appendix A: Interface used for MOO 152 A.1. Excel, Visual Basic for Applications and HYSYS Interface Appendix B: et Flow Method (Macros) 152 159 B.1. Net Flow Method in Visual Basic for Applications 159 B.2. NFM Applied to Some Examples 162 iv Summary “Sustainable development is development that meets the needs of the present, without compromising the ability of future generations to meet their own needs” as given by World Commission on Environment and Development. There are three spheres of sustainability – economic development, environmental stewardship and societal equity. This is often touted as the “triple bottom line”. Of the three, only the first two are quantifiable based on process design and operating variables. While economic criteria such as profit before taxes, payback period and net present worth are well established, environmental objectives are still novel and there is no general consensus on the method for calculating the environmental index. Environmental indices can be measured via environmental metrics or environmental impact indices. For the former, it mainly comprises ratios that indicate the efficiency of the plant in terms of production or energy. For the latter, many contributing factors have been identified for environmental impacts: impact on humans, ecosystem – terrestrial and aquatic, global warming, ozone depletion, photochemical oxidation, acid rain and eutrophication. Several aggregation methods for the environmental indicator have also been discussed. This study focuses on process optimization for multiple economic and environmental criteria, or otherwise termed as sustainability criteria. The different process applications that have been studied for both economic and environmental criteria are reviewed. Applications that considered only environmental criteria are also included as it is of interest to identify the different methods that have been used to quantify the environmental performance of a process. Many of the previous studies that employed environmental indices for optimization, used aggregated environmental index as the v objective, and in some papers, the process analysis is coupled with an economic objective. Hence, feasibility and usefulness of process optimization for more than two economic and environmental objectives are studied in this work. Two recovery processes have been selected for the optimization using sustainability criteria. They are: a VOC (volatile organic component) recovery system and a solvent recovery system. These processes are optimized for both economic and environmental objectives using the elitist non-dominated sorting genetic algorithm. For the environmental objectives, the contributing factors to the environmental impacts are optimized individually or grouped into a few indices where appropriate. The Paretooptimal solutions are obtained to elucidate the trade-offs present, and the decision maker would be better equipped in choosing one of them for implementation. Thereafter, net flow method is then used to identify the preferred Pareto-optimal solution based on the preferences declared by the decision maker. The preferences provided by the decision maker should be more objective since s/he is aware of the quantitative trade-offs present in the objective functions. Insights gained from considering a number of environmental objectives for process optimization are highlighted. Conclusions and recommendations for further research are provided at the end of the thesis. vi omenclature AEP AHI AHP AP ATMP ATP BOD C[i,j] CAHi CCP CCR CF CFCs ck[i,j] COD COM CPI CST95 CTAM CTWM CUi = = = = = = = = = = = = = = = = = = = = = D DC DDB Dk[i,j] DPBP EDIP EII EP EPA EPS EPW ETP Fabs FC FCI Fi,j = = = = = = = = = = = = = = = = Fi,j,k = FSi G GDi = = = Annual Equivalent Profit ($/yr) Atmospheric Hazard Index Analytic Hierarchy Process Acidification Potential Atmospheric Potential Aquatic Toxicity Potential Biochemical Oxygen Demand Global Concordance Index Chemical Atmospheric Hazard for Chemical i Cumulative Cash Position ($) Cumulative Cash Ratio Cash Flow ($/yr) Chlorofluorocarbons Individual Concordance Index for Criterion k Chemical Oxygen Demand Cost of Manufacturing ($/yr) Chemical Process Industries Critical Surface Time 95 Critical Air Mass Critical Water Mass Cooling Utility for Condenser of Column i (i = 1, 2, 3) or Cooler i (i = feed, sol, prod) (°C) Depreciation ($/yr) Direct Costs ($) Double-declining Balance Method Discordance Index for Criterion k Discounted Payback Period (yr) Environmental Design of Industrial Products Environmental Impact Index Eutrophication Potential Ecotoxicity Potential to Air Environmental Priority Stages Ecotoxicity Potential to Water Ecotoxicity Absorbent Flow Rate (kmol/hr) Fixed Costs ($) Fixed Capital Investment ($) Flow Rate of Stream i of Column j (where i = dist or btm; j = 1,2,3) (kg/hr) Flow Rate of Component k in Stream i of Column j (where i = dist or btm; j = 1,2,3) (kg/hr) Feed Stage for Column i (i = dist, prod, 1, 2, 3) Global Impact Score Green Degree of Chemical Compound i vii GDS GE GWP HCPW HEN Hji HNCPW HTP HTPE HTPI HUi IE IFj,k IINH Ij IMCSD IMPACT IPC IPCC Jk L LC50 = = = = = = = = = = = = = = = = = = = = = = LCA LD50 = = LP M MACRS MEN Mi MOO MP NFM NPV NPW NSGA ODP PAT PBP PBT PCOP PEI Pk PVR Qk = = = = = = = = = = = = = = = = = = = = Green Degree for a Stream General Expenses Global Warming Potential Human Carcinogenic Toxicity Potential to Water Heat Exchanger Network Impact Hazard Value for each chemical component i Human Non-carcinogenic Toxicity Potential to Water Human Toxicity Potential Human Toxicity due to Dermal Exposure Human Toxicity Potential by Ingestion Heating Utility for Column i (i = 1, 2, 3) Industrial Ecology Importance Factor for Impact Category j and Area k Inhalation Toxicity Index Magnitude of deterioration on Impact Category j Inter-Ministerial Committee on Sustainable Development Impact Assessment of Chemical Toxics Process Composite Index Intergovernmental Protection on Climate Change Value of Objective Function k Land Cost ($) Lethal concentration that would cause death in 50% of Pimephales promelas (mg/kg) Life Cycle Analysis/Assessment Lethal-dose that produces death in 50% of rats by oral ingestion (mg-min/m3) Low Pressure Mass Flow Rate of a Stream (kg/hr) Modified Accelerated Cost Recovery System Mass Exchanger Network Mass Flow Rate of Component i (kg/hr) Multi-Objective Optimization Medium Pressure Net Flow Method Net Present Value ($) Net Present Worth ($) Non-dominated Sorting Genetic Algorithm Ozone Depletion Potential Profit after Taxes ($/yr) Payback Period (yr) Profit before Taxes ($/yr) Photochemical Oxidation Potential Potential Environmental Index Preference Threshold for Objective k Present Value Ratio Indifference Threshold for Objective k viii Qr R rd RHKi RLKi ROI RSM rt S SCENE = = = = = = = = = = SGA SL SMD SOO SOYD SPI Stagei Tabs Tabs,ex TAC TAPPS TCI tD Tdist Tf,abs TRACI = = = = = = = = = = = = = = = = TTP TVOC TWA-TLV VBA Vk VOC WAR WC WCHi,j wj Wk WMO Yji = = = = = = = = = = = = = Greek symbols: = σi σ[i,j] = Energy Consumption (kW) Revenue ($/yr) Discount Rate (yr-1) Heavy Key Reovery for Column i (i = 1, 2, 3) Light Key Recovery for Column i (i = 1, 2, 3) Return on Investment Rough Set Method Tax Rate (yr-1) Salvage Value ($) Simultaneous Comparison of Environmental and Nonenvironmental Process Criteria Scaled Gradient Analysis Straight Line Method Solid Mass Disposal Single Objective Optimization Sum of the Years Digits Method Sustainable Process Index Stages for Column i (i = abs, dist, prod, 1, 2, 3) Absorbent Temperature (°C) Temperature of Lean Absorbent exiting the Heat Exchanger (°C) Total Annual Cost ($/yr) Total Annualized Profit per Service Unit ($/yr-unit) Total Capital Investment ($) Project Lifetime (yr) Rich Absorbent Temperature to Distillation Column (°C) Waste Gas Stream Temperature to Absorber (°C) Tool for the Reduction and Assessment of Chemical and other environmental Impacts Terrestrial Toxicity Potential VOC Temperature (oC) Time weighted average of the threshold limit values Visual Basic for Applications Veto Threshold for Objective k Volatile Organic Compound Waste Reduction Working Capital ($) Weighted Category Hazard for Impact Category j for Chemical i Weighting Factors for Impact Category j Weights for Objective k World Meteorological Organization Normalized Impact Factor Final Ranking Score Outranking Matrix ix ϕi,j = ϕ j, E = ϕ max j = ϕ i, j = Environmental Impact Potential of Compound i for Impact Category j Normalized Impact Factor of Impact Category j for an Energy Source Maximum Value of category j among all Substances Reported in the Literature Normalized Impact Category x List of Tables Table 2.1: Different definitions for Return on Investment (ROI) Page 19 Table 3.1: Material Metrics 36 Table 3.2: GWPs and ODPs of some substances; refer to the respective references in the foot-note for the extended list 44 Table 3.3: PCOPs of some organic substances (Heijungs et al., 1992) 47 Table 3.4: Acidification potentials (Heijungs et al., 1992) 48 Table 3.5: Eutrophication potential of compounds (Heijungs et al., 1992) 50 Table 4.1: Economic and Environmental Criteria – Petroleum Refinery and Petrochemicals 60 Table 4.2: Economic and Environmental Criteria – Biotechnology, Pharmaceutical and Chemicals 66 Table 4.3: Economic and Environmental Criteria – Downstream Processing 71 Table 4.4: Economic and Environmental Criteria – Energy Systems and Heat 76 Integration Table 4.5: Environmental Criteria – Petrochemicals 80 Table 4.6: Environmental Criteria – Biotechnology, Pharmaceutical and Chemicals 83 Table 4.7: Environmental Criteria – Downstream Processing 85 Table 4.8: Environmental Criteria – Miscellaneous 86 Table 5.1: Objectives, Decision Variables and Constraints for VOC Process – Operation Optimization 92 Table 5.2: Objectives, Decision Variables and Constraints for VOC Process – Design Optimization 101 Table 5.3: Comparison of Two Selected Pareto-optimal Solutions 108 Table 5.4: Compositions of Components in Spent Wash Solution 110 xi Table 5.5: Objectives, Decision Variables and Constraints for Solvent Recovery Process – Design Optimization 112 Table 5.6: Comparison of Two Selected Pareto-optimal Solutions 118 Table 6.1: NFM Parameters for Ranking VOC Recovery Application 129 Table 6.2: NFM Parameters for Ranking Solvent Recovery Application 131 xii List of Figures Figure 2.1: Three spheres of sustainability Page 11 Figure 3.1: Life Cycle Assessment (Masters, 1998) 33 Figure 5.1: VOC Recovery Process Flowsheet 88 Figure 5.2: Excel-VBA-HYSYS Setup for MOO of Processes 91 Figure 5.3: Selected Results for Operation Optimization of VOC Recovery for PBT and PEI 95 Figure 5.4: Selected Results for Operation Optimization of VOC Recovery for Eight Objectives 97 Figure 5.5: Selected Results for Operation Optimization of VOC Recovery for Five Objectives 99 Figure 5.6: Selected Results for Design Optimization of VOC Recovery for NPW and PEI 102 Figure 5.7: Selected Results for Design Optimization of VOC Recovery for Ten Objectives 106 Figure 5.8: Optimal Objective Values for Design Optimization of VOC Recovery for Five Objectives 109 Figure 5.9: Sequences 1 and 2 from Chakraborty and Linninger (2002) 111 Figure 5.10: Selected Results for Design Optimization of Solvent Recovery for NPW and PEI 115 Figure 5.11: Selected Results for Design Optimization of Solvent Recovery for Ten Objectives 120 Figure 6.1: (a) Individual concordance index, and (b) discordance index calculations used in NFM algorithm to determine ranking scores for the Pareto domain solutions. 127 Figure 6.2: Ranking of Pareto-optimal Solutions by Net Flow Method for VOC Recovery Design Optimization for Several Objectives 130 xiii Figure 6.3: Ranking of Pareto-optimal Solutions via Net Flow Method for Solvent Recovery Design for Many Objectives 132 Figure A.1: Excel-VBA-HYSYS Setup for MOO of Processes 152 Figure A.2: Flowchart for NSGA-II implemented in VBA; (*) indicate steps which require VBA and HYSY interface 158 Figure B.1: Net Flow Method for Williams and Otto Process 163 Figure B.2: Net Flow Method for Alkylation Process 163 Figure B.3: Net Flow Method for KUR Benchmark Problem 165 Figure B.4: Net Flow Method for SCH Benchmark Problem 165 Figure B.5: Net Flow Method for ZDT2 Benchmark Problem 166 Figure B.6: Net Flow Method for ZDT4 Benchmark Problem 166 Figure B.7: Net Flow Method for ZDT6 Benchmark Problem 167 Figure B.8: Net Flow Method for CONSTR Benchmark Problem 167 Figure B.9: Net Flow Method for SRN Benchmark Problem 168 Figure B.10: Net Flow Method for TNK Benchmark Problem 168 xiv Chapter 1: Introduction Chapter 1 Introduction 1.1. Optimization of Chemical Processes Optimization refers to finding one or more feasible solutions which correspond to the maximum and/or minimum of one or more objectives. The need to find such optimal solutions in a problem comes mostly from the purpose of designing and operating a plant for minimum fixed capital cost and/or operating cost, for maximum reliability, and others. As a result, optimization is one of the major quantitative tools in industrial decision making. In general, it has become an integral part of our society, without which many activities would not be as efficient as they are now. A plethora of problems in the design, construction, operation, and analysis of chemical plants (as well as many other industrial processes) can be resolved by optimization. As society evolves with ever changing economic and environmental landscape, there is still room to further optimize the current industrial operations. There are generally two types of optimization problems, namely, the single objective optimization (SOO) and the multi-objective optimization (MOO). The first type (SOO) considers only one objective in the optimization procedure. In employing this method of optimization, the decision maker would need to choose the objective that is of greatest relevance to the problem at hand. Since practical applications require several objectives to be considered simultaneously, there is growing interest in the optimization of more than one objective – commonly known as the MOO. Bhaskar et al. (2000) presented the background of MOO, different methods and their applications in the 1 Chapter 1: Introduction chemical engineering are up to the year 2000. Their review shows that there were around 30 journal publications on MOO of various processes before 2000; on the other hand, about 80 applications of MOO in chemical engineering have been published since 2000, according to Masuduzzaman and Rangaiah (2008). These two reviews provide a comprehensive summary of chemical engineering applications, and interested readers may refer to them for more detailed information. There are many techniques that can be used to solve MOO problems. These techniques can be classified into five different classes: (1) no preference methods; (2) a posteriori methods using scalarization approach; (3) a posteriori methods using multiobjective approach; (4) a priori methods; and (5) interactive methods (Miettinen, 1999, and Rangaiah, 2008). A summary of the methods used and the applications studied in the field of chemical engineering from 2000 to mid-2007 is provided by Masuduzzaman and Rangaiah (2008). The popular methods used by academia in the field of chemical engineering are a posteriori methods using scalarization approach (i.e. weighting and εconstraint method) as well as a posteriori methods using multi-objective approach (e.g. genetic and evolutionary algorithms). As the objectives may be partially or totally conflicting, the solution of a MOO problem will not be unique and there will be many optimal solutions, which are known as Pareto-optimal solutions. Each one of them, when compared to another, is better in at least one objective value and is worse in at least one other objective value. Thus, Paretooptimal solutions elucidate the trade-offs present among the objectives, and equip the decision maker in choosing one of them for implementation based on other information, his/her experience and preferences. 2 Chapter 1: Introduction For any operation or design case of a chemical process, only a single optimal solution would be required for implementation. Hence, with the plethora of solutions obtained for MOO problems, a choice for one solution has to be made. In order to choose a single solution, ranking methods such as rough set or net flow methods (Thibault, 2008) could be used. These methods require information on the Pareto-optimal solutions and inputs from the decision maker before the solutions can be ranked. The availability of Pareto-optimal solutions also provides a quantitative foundation in reducing the biasness of the decision maker when his/her inputs are required for the ranking methods (Deb, 2001). 1.2. Objectives in Process Optimization As Lord Kelvin, a physicist, once said: “When you can measure what you are speaking about and express it in numbers, you know something about it”. The formulation of objective functions is one of the crucial steps in the application of optimization to a particular problem. Hence, one must be able to translate a verbal statement or concept of the desired objective into mathematical terms. The choice in the number and type of objectives is dependent on the purpose of the study. There are many objectives available and they are briefly discussed below. Very often, as chemical industries are profit-driven, the objective functions are economics-related. They can be material metrics or profitability measures. Material metrics are ratios that measure the efficiency of the chemical process – e.g. amount of product per unit of feed, amount of waste emitted per unit of product. On the other hand, profitability measures are economic objectives commonly used by the industries to 3 Chapter 1: Introduction measure the performance of the chemical processes. They include revenues, manufacturing costs, profits, net present worth, payback period, etc. Chemical processes can also be optimized for objectives that are not related economics. These objectives can be product quality, energy efficiency, environmental impacts, sustainability, process safety, operation time, robustness, etc. 1.3. Motivation and Scope of Work The World Commission on Environment and Development (1987) defined sustainability as “the development that meets the needs of the present, without compromising the ability of future generations to meet their own needs”. This is often quoted in almost every article advocating sustainability (e.g. Azapagic and Perdan, 2000; Sikdar, 2003a and 2003b; Heinzle et al., 2006; Darton, 2006). Increasing attention and emphasis are being placed on sustainability. For example, an Inter-Ministerial Committee on Sustainable Development (IMCSD) has been established in Singapore in February 2008 (available at http://app.mewr.gov.sg/web/Contents/ContentsSSS.aspx?ContId=1034). IMCSD seeks to create a national framework and strategy for Singapore’s sustainable development in view of the rising domestic and global challenges. The journey towards sustainability is not possible if it only depended on environmentalists and/or the government; it requires the awareness of every individual and their efforts to realize this laudable objective. There are three spheres of sustainability: economic development, environmental stewardship and societal equity (e.g., Azapagic and Perdan, 2000; Sikdar, 2003a and 2003b; Heinzle et al., 2006). This is often touted as the “triple bottom line”. Economic 4 Chapter 1: Introduction indicators measure the profitability of a chemical process. Usually, these are the first criteria for companies; if the process is not economically feasible, the project would be aborted. Environmental indicators are concerned with efficient use of raw materials and energy in the process as well as the environmental impacts caused by emissions. The latter include human toxicology, ecotoxicity, global warming, ozone depletion, acidification, eutrophication and photochemical oxidation (Young and Cabezas, 1999; Jolliet et al., 2003; Bare et al., 2006). Societal indicators measure different aspects of working conditions and regulations; these indicators are driven by the government and company’s policies, but not by engineers per se. Hence, a chemical engineer can and should consider both economic and environmental performance of the process in order to make it sustainable. As Steve Johnson, the administrator of US Environmental Protection Agency, once said: “We have the responsibility to sustain – if not enhance – our natural environment and our nation’s economy for future generations” (http://www.epa.gov/Sustainability/, accessed on 1st Aug 2008). Identifying methods to quantify economic and environmental performance of chemical processes is essential. Economic objectives (e.g. payback period and net present worth) have always been the most important criteria to the industries, and are well established. On the other hand, environmental indices are still in the development stage and there exist a number of methods to calculate them. There are almost 80 journal papers since the year 1995 discussing the calculation of one or more environmental indices and then using them to quantify the performance of chemical processes. There are several differences among the environmental indices employed. A number of investigators chose to consider selected individual environmental components 5 Chapter 1: Introduction (e.g. critical air mass, CTAM based on LC50 by Stefanis et al., 1996; critical water mass, CTWM by Steffens et al., 1999; inhalation toxicity index, IINH using EFRAT methodology by Chen et al., 2002; human toxicity potential by ingestion, HTPI, aquatic toxicity potential, ATP and acidification potential, AP by Kim and Smith, 2005; weighted sum of benzene and carbon dioxide production by Janjira et al., 2007). Other researchers chose to work with aggregated indicators which are more popular. Examples of aggregated indices are the Potential Environmental Index (PEI) using the Waste Reduction (WAR) algorithm (Young et al., 2000; Chen and Feng, 2005), Eco-Indicator 99 (Hugo et al., 2004; Dominguez-Ramos et al., 2007), Atmospheric Hazard Index, AHI (Gunasekera and Edwards, 2003) and Green Degree (Zhang et al., 2008). Besides using aggregated or single environmental indicator, more than one environmental indicator can be employed since there are many contributing components. For example, Azapagic and Clift (1999) optimized the formation of boron products by minimizing global warming potential (GWP) and photochemical oxidation potential (PCOP) simultaneously via bi-objective optimization. They also employed MOO for GWP, ozone depletion potential (ODP), production rate and costs using the ε-constraint method. Steffens et al. (1999) used annual costs as the economic objectives with two environmental indicators (CTWM and SPI) for penicillin production. Kim and Smith (2005) optimized the recovery of acetic acid from aqueous waste mixtures using four objectives – profit, HTPI, ATP and AP. Results in Azapagic and Clift, 1999) and other works reviewed in Chapter 4, clearly show that minimization of one environmental component (e.g. GWP) does not necessarily minimize another environmental component (e.g. PCOP). 6 Chapter 1: Introduction In the studies described above, MOO has been employed for two conflicting objectives – economic and environmental type. Economic objectives are well-established and the choice of a profitability measure may be easy; even then, one may like to consider more than one profitability measure. On the other hand, the choice of environmental performance indicator requires more care. Most of the reported works have employed aggregated indicators, providing a final environmental performance index. There are, however, many contributing factors for the environmental performance index. For example, the toxicity impact on humans (HTP), ecotoxicity (ETP), GWP, ODP, etc. Several studies have illustrated that minimization of an environmental performance index does not guarantee minimization of each contributing factor. In addition, there are two main issues about the use of aggregated indicators. First of all, the method of normalizing impact categories, and whether it brings the impact factors on the same platform, is debatable. Secondly, unless the decision maker has an in-depth understanding of the process and the impacts it imposes, the weights given by the decision maker are highly subjective. Since it is desirable to minimize the contributing components of the environmental performance index, they should be optimized as individual objectives together with the economic performance index. One of the several MOO algorithms can be employed for this purpose. In this study, feasibility and usefulness of considering several economic and environmental objectives are investigated. For this, two case studies are chosen: a VOC (volatile organic component) recovery system (Chen et al., 2003) and a solvent recovery system (Chakraborty and Linninger, 2002). Seven groups were identified for environmental impacts – HTP, ETP, GWP, ODP, PCOP, AP, and eutrophication (EP). 7 Chapter 1: Introduction The last 5 environmental impact components can be lumped together as impact on the atmosphere, where necessary. In addition, economic aspects should be considered in the optimization. Potential economic criteria are profit before taxes (PBT), payback period (PBP) and net present worth (NPW). Some or all these objectives will be optimized simultaneously using the elitist non-dominates sorting genetic algorithm, NSGA-II (Deb et al., 2002) implemented in Excel. Net flow method can be used as a tool in identifying the preferred Pareto-optimal solution (Thibault, 2008); this requires the decision maker’s preferences. While the choice of a solution may be subjective, the generation of Pareto-optimal solutions provides a quantitative foundation in reducing the biasness of the decision maker (Deb, 2001). Pareto-optimal solutions and the preferred Pareto-optimal solution for the two case studies will be presented and discussed. Insights gained from considering several objectives instead of just two will be highlighted. 1.4. Organization of Thesis Chapter 2 gives an overview on the concept of sustainability and explains the three contributing factors – economic, environmental and societal. Thereafter, an overview of the different environmental objectives available as well as different aggregation methods is presented in Chapter 3. Following on, Chapter 4 provides a comprehensive literature review on the applications studied for both economic and environmental objectives. In Chapter 5, two applications are chosen for multi-objective optimization for bi-, many and/or several objectives. The two applications are VOC and solvent recovery process. The results obtained are presented and the decision variables that have an influence on 8 Chapter 1: Introduction the objective functions are discussed. MOO would provide decision makers with a myriad of solution; however, for an operation or a design case, a single point would have to be identified to determine the conditions at which the application is to be designed for or operated at. Hence, in Chapter 6, net flow method would be employed to rank the solutions after the decision maker has provided the necessary parameters. The preferred Pareto-optimal solution would be the one chosen for design or operation, wherever applicable. Finally, Chapter 7 summarizes the conclusions of this study and recommendations for future studies. 9 Chapter 2: Sustainability Chapter 2 Sustainability 2.1. Introduction Almost every article advocating sustainability (e.g. Azapagic and Perdan, 2000; Sikdar, 2003a and 2003b; Heinzle et al., 2006; Darton, 2007) quotes: “Sustainable development is development that meets the needs of the present, without compromising the ability of future generations to meet their own needs” from the report of the World Commission on Environment and Development (1987). Sustainability is not a goal which one can reach, but it is a journey through time that one should take (Sikdar, 2003b; Cabezas, 2007). It is not possible to journey towards sustainability if it only depended on environmentalists or the government; it requires the awareness of every individual and their efforts to realize this objective. As Lord Kelvin, a physicist, once said: “When you can measure what you are speaking about and express it in numbers, you know something about it”. Hence, to have a better understanding of the sustainability concept, it is preferable to quantify it. Based on the definition of sustainability, numerous indicators have been formulated. These include ecological footprint, maximum sustainable yield (MSY), net national product (NNP), emergy, exergy, environmental sustainability index (ESI) and index of sustainable economic welfare (ISEW) (Bartelmus, 2001; Mayer et al., 2004; Cabezas, 2007). Such indicators are applicable to the sustainability of a nation, region or globe; however, they are not useful for businesses or specifically for chemical processes. Hence, another set/type of indicators has to be formulated. 10 Chapter 2: Sustainability There are three spheres of sustainability: economic development, environmental stewardship and societal equity (Azapagic and Perdan, 2000; Sikdar, 2003a and 2003b; Heinzle et al., 2006). This is often touted as the “triple bottom line”. Economic affluence is desirable especially for the poverty stricken; however, economic growth should proceed in a controlled manner that is able to balance the need for social development and equity. Moreover, quality of the environment should not be compromised. Our environment is important as it supplies our needs while absorbing the wastes generated. Figure 1 illustrates how these three spheres overlap to achieve sustainable development (indicated by the red/shaded region). Operating in the “sustainable” zone does not imply that the world has to maintain a particular way of life that is deemed optimal; rather, this state would change with demands as well as any limitations involved over time (Darton, 2007). Figure 2.1: Three spheres of sustainability A couple of studies have suggested a fourth sphere, on top of the three shown in Figure 2.1. The first suggestion of the fourth sphere is resource efficiency (Afgan et al., 11 Chapter 2: Sustainability 2000 and 2007; Darton, 2006), which is defined as the proportion of the resources whose benefits have been realized by the end users. For example, in the context of chemical engineering, it would measure the proportion of the raw materials that has been converted to the desired product (e.g. the conversion of ethyl benzene to styrene would have some undesired products formed simultaneously). Azapagic and Perdan (2000) had classified resource use (or efficiency) in the “environment” sphere of sustainability. As resource efficiency indirectly indicates the amount of waste that would be produced, it could be used as a measure of the environmental impact a process has. Hence, in accordance with Azapagic and Perdan (2000), resource efficiency would be treated as an environmental indicator. Interestingly, Spangenberg (2002) suggested that the fourth sphere should be institutional. This includes citizens’ turnout at election polls, access to basic necessities, etc. Evidently, the institutional sphere is only applicable when analyzing a country or region. Companies interested in issuing sustainability reports could refer to the Global Reporting Initiative (GRI) guidelines which provide a comprehensive list of economic, environmental and social data to be recorded (Weiss and Funnell-Milnar, 2007). This list can be obtained from www.globalreporting.org (accessed in January 2008). Almost 1000 organizations have indicated their usage of the GRI Reporting Framework. The companies from the chemical industry that have employed the GRI Reporting Framework are Du Pont, Dow Chemical, BP, ExxonMobil and Shell. These reports can be used to benchmark organizational performance with respect to laws, norms, codes, performance standards and voluntary initiatives, to demonstrate organizational commitment to sustainable development, and to compare organizational performance over time. On the 12 Chapter 2: Sustainability other hand, specifically catered to chemical industries, IChemE presented a set of 50 sustainability metrics involving all three spheres (refer to www.icheme.org/sustainability, accessed in January 2008). Other industries can also employ these metrics with some modifications. Both these sustainability reporting guidelines are very similar: from the amount of profits (economic) to the amount of greenhouse gases produced (environmental) to the rate of employee turnover (societal). Very often, chemical engineers are faced with the design or operation of chemical processes. Thus, this chapter seeks to present a review of the various sustainability indicators available in the literature that could be applied to chemical processes. These indicators can be classified into the three spheres as shown in Figure 2.1. 2.2. Sustainability Indicators As mentioned above, sustainability incorporates economic, environmental and societal effects. Thus, it would be more appropriate if indicators for each aspect are considered separately. However, it should be noted that the indicators do not solely affect only one aspect; rather, they affect more than one aspect due to interactions in the society. For example, material intensity would be classified as an environmental indicator since it refers to efficient use of resources and the use of resources has a negative impact on the environment. Conversely, processing the resource would add value to it and would have a positive impact on the economy, and the society would benefit from the product formed while future generations would be deprived of the resource (Martins et al., 2007). Hence, although primary effect of an indicator is on the sphere in which it is classified, it could have secondary effect on the remaining two spheres. 13 Chapter 2: Sustainability 2.2.1. Economic Indicators In chemical engineering projects, profitability measures are always used in assessing the feasibility of a project. Moreover, as companies are profit-driven, businesses would usually review the economic benefits of a project or process unit before investing in the project/process. There is an established set of economic measures/indicators which is often presented in textbooks and used in the industry (e.g., see Turton et al., 2003; Edgar et al., 2001). Before providing the profitability measures, some concepts of cost estimation have to be introduced first as they will be required for the measures. 2.2.1.1. Cost Estimation There are two sets of costs that the engineer has to consider when designing a new plant or implementing a new fixture in the existing plant. They are (1) capital cost or investment and (2) manufacturing cost. 2.2.1.1.1 Capital Investment Total capital investment (TCI) is the investment required to purchase new equipment, build and install them such that it is ready for production. It is the sum of two components: (1) fixed capital investment (FCI) and (2) working capital (WC). FCI is the cost associated with building the plant. A small portion of FCI is for the land and is not depreciable. FCI can be calculated using capital cost (investment) programs such as CAPCOST© (Turton et al., 2003), Aspen Icarus Process Evaluator (Aspen Technology Inc., 2007) and Cost Analysis and Project Economic Evaluation by SuperPro and EnviroPro Designer (http://www.intelligen.com/costanalysis.shtml; Seider et al., 2004). 14 Chapter 2: Sustainability WC is the amount of capital required to start up the plant and finance the first few months of operation before revenues from the process start. This usually includes raw material inventories, salaries, cash and accounts receivable which can be recovered at the end of the project lifetime. WC is usually estimated as 15-20% of FCI (Seider et al., 2004; Turton et al., 2003). Depreciation is the reduction in the value of an asset. The total capital amount that can be depreciated is FCI minus the land cost (L) and the salvage value (S). Salvage value is a small proportion of the initial FCI and often it is assumed to be zero. Land cost is also a small fraction of the total and thus its contribution to FCI is negligible. There are various methods to calculate the depreciation amount for each year; for example, straight line method (SL), double-declining balance method (DDB), sum of the years digits method (SOYD) and modified accelerated cost recovery system (MACRS). For illustration purposes, depreciation (D) is calculated using the SL method over a project lifetime of tD years when the salvage value is recovered. D= FCI - L - S FCI ≈ tD tD (2.1) 2.2.1.1.2 Manufacturing Cost Manufacturing cost is the day-to-day cost incurred in the operation of the chemical process unit. There are three components in the cost of manufacturing, COM: (1) direct costs, (2) fixed costs, and (3) general expenses. COM = DC + FC + GE (2.2) Direct costs (DC) are costs that vary with the production rate (e.g. cost of raw materials). Fixed costs (FC), on the other hand, are costs that do not vary with the production rate 15 Chapter 2: Sustainability (e.g. depreciation). General expenses (GE) are related to the management and administration activities, and not directly to the production process. 2.2.1.2 Profitability Criteria Several profitability criteria are available and the choice of criteria to be used for evaluation of a project is dependent on the engineer/company. In analyzing the profitability of a process that is already online, capital costs would have been predetermined and already incurred, and so the following would most likely be used in the economic analysis. Profit before taxes (PBT) is the difference between the revenue earned from the production process and the total annual cost incurred. Revenue (R) is the sales of the product while TAC is the sum of COM and D. PBT = R – TAC = R – COM – D (2.3) Profit after taxes (PAT) is the amount of the profit that the businesses retains after accounting for the income tax. Let rt denote the tax rate for the region. Hence PAT is calculated via equation 2.4. PAT = PBT (1 - rt) = (R – COM – D) (1 - rt) (2.4) As depreciation is an amount to account for the initial investment, it is not a physical outflow of cash for the business for that particular accounting year. Hence, the term cash flow (CF) is used for the physical cash exchanges of the company. Thus, CF is the addition of D to PAT. CF = PAT + D = (R – COM – D)(1 - rt) + D = (R – COM)(1 - rt) + rtD (2.5) 16 Chapter 2: Sustainability 2.2.1.2.1 Time Criterion Payback period (PBP) is the number of years for the annual income ($/yr) to recover the initial investment ($) without considering the time value of money. There are a number of definitions for PBP. One common definition of PBP has the CF as the yearly income and the FCI as the initial investment (e.g., Pintarič and Kravanja, 2006; Heinzle et al., 2006; Turton et al., 2003). Heinzle et al. (2006) had an additional definition for PBP, which is the ratio of TCI to PAT. PBP = FCI CF PBP = or TCI PAT (2.6) Discounted payback period (DPBP) is the same as payback period except that it accounts for the time value of money (Turton et al., 2003). Time value of money involves a factor called the discount rate (rd) which represents the minimum acceptable rate of return that the company is willing to accept for any new investment. The future yearly incomes are brought to the present value using rd. As one would notice from the expression of DPBP in equation 2.7, it requires iterative methods to obtain its value as DPBP appears on both sides of the equation. [ DPBPi +1 ]   FCI rd (1 + rd )DPBPi ln  + 1 CF   =  ln (1 + rd ) (2.7) 2.2.1.2.2 Cash Criterion Cumulative cash position (CCP) is the sum of all positive cash flows net of the sum of all negative cash flows. However, if the engineer is comparing projects with different scales 17 Chapter 2: Sustainability of production (i.e. bearing dissimilar fixed capital investments), it is preferable to employ the cumulative cash ratio (CCR) which is the ratio of all positive to negative cash flows. CCP = - TCI + CF tD + WC + S + L ≈ - FCI + CF tD (2.8) CF ⋅ t D + WC + S + L CF ⋅ t D + WC ≈ TCI TCI (2.9) CCR = Similar to CCP is the net present value (NPV), also known as net present worth (NPW), which takes into account the time value of money. Thus all the positive and negative cash flows are brought to its present value before taking the positive cash flows net of the negative cash flows. In addition, the present value ratio (PVR) is the ratio of the present value of positive to negative cash flows (i.e. taking into account the time value of money). (1 + rd )t − 1 WC + S + L (1 + rd )t − 1 WC NPW = −TCI + CF + ≈ −TCI + CF + t t t (1 + rd ) (1 + rd )t rd (1 + rd ) rd (1 + rd ) D D D D D (2.10)  (1 + rd )tD − 1 WC + S + L  + CF  t rd (1 + rd ) D (1 + rd )tD    ≈ PVR = TCI  (1 + rd )tD − 1 WC  + CF  tD rd (1 + rd ) (1 + rd )tD   TCI (2.11) 2.2.1.2.3 Interest Rate Criterion Return on investment (ROI) is the non-discounted rate at which annual earnings is made from the initial investment. Several definitions for this available in the literature are presented in Table 2.1 in conjunction with equation 2.12. As illustrated above, profit before and after taxes differ by a fixed tax rate which cannot be altered by the engineer; 18 D Chapter 2: Sustainability also, TCI is usually approximated as a percentage (115 – 120%) of FCI and is thus not variable with process conditions. ROI = Annual Earnings Initial Investment (2.12) Table 2.1: Different definitions for Return on Investment (ROI) Reference Annual Earnings Initial Investment Turton et al. (2003) PAT FCI Seider et al. (2004) & Heinzle et al. (2006) PAT TCI Pintarič and Kravanja (2006) PBT TCI The discounted cash flow rate of return (DCFROR) or more commonly known as the internal rate of return (IRR) is the discount rate at which NPW is zero. This means that the expression for NPW in equation 2.10 is set to zero and the resulting equation is solved for the value of rd. 2.2.2 Environmental Indicators Azapagic and Perdan (2000) had compiled a list of environmental indicators based on the life cycle approach. These indicators are classified into three categories: (1) environmental impacts, (2) environmental efficiency, and (3) voluntary actions. As the aim of this chapter is to present sustainability indicators that are applicable to a chemical process in the context of its design and operation, only the first two categories will be discussed in the subsequent sections. The last category on voluntary actions (e.g. management systems or assessment of suppliers) is relevant mainly to larger entities such as businesses or industries, and thus will not be discussed here. 19 Chapter 2: Sustainability 2.2.2.1. Environmental Impact Environmental impact indicators comprise both local toxicology effects and global atmospheric effects. The former includes toxic effects on human beings, terrestrial and aquatic organisms caused by the chemical compounds involved in the process. On the other hand, the latter consists of atmosphere-related issues that involves the degradation in the quality of air, water and soil that surrounds us. Several environmental assessment tools have employed toxicity and/or atmospheric indicators. Examples of these tools are the waste reduction algorithm, WAR (Mallick et al., 1996; Cabezas et al., 1999; Young et al., 1999 and 2000); the critical surface-time 95, CST95 (Song et al., 2002); the tool for the reduction and assessment of chemical and other environmental impacts, TRACI (Bare et al., 2003 and 2006); and IMPACT 2002+ (Joillet et al., 2003). Only a brief overview of indicators for toxicology and global atmospheric effects will be presented here; more details are presented in the next chapter. 2.2.2.1.1. Local Toxicology Effects First of all, toxicity is the ability to inflict harm on to a living organism as well as to indicate the adverse effects caused by a chemical. It can be measured as the amount of toxic equivalents that is being emitted by the process (Heijungs et al., 1992; Schwarz et al., 2002; IChemE 1 ). Toxicology effects can be measured for three categories of organisms: (1) human beings, (2) terrestrial life forms, and (3) aquatic organisms. Toxicity effects on both human beings and terrestrial life forms can be triggered through the ingestion of toxic chemicals. The common factor used to measure toxicity via 1 The Sustainability Metrics: Sustainable Development Progress Metrics, recommended for use in the Process Industries (available at http://www.icheme.org/sustainability/, accessed in January 2008). 20 Chapter 2: Sustainability ingestion is the lethal-dose that produces death in 50 percent of rats by oral ingestion (LD50) (Young et al., 1999; Joillet et al., 2003). In addition, human beings are the ones who are directly exposed to the chemicals due to working in the chemical plant or as direct or indirect users of the products. As such, human beings can come into contact with the toxic chemicals via inhalation and exposure. This category of toxicity can thus be measured using the time weighted average of the threshold limit values (TWA-TLV) (Young et al., 1999). Another measure of toxicity on humans generally, is based on the rate of usage and the MSDS’s R-phrases of the chemical compounds (Vincent et al., 2005; Martins et al., 2007). Finally, for aquatic organisms, the toxicity effects can be measured by the lethal concentration that would cause death in 50 percent of Pimephales promelas, a representative fish species, (LC50) (Young et al., 1999). 2.2.2.1.2. Global Atmospheric Effects Besides toxicity, the atmospheric effects of a chemical process are important. This is because it is not socially responsible to mar the environment which will have undesirable effects on the current and future generations. For example, the release of carbon dioxide seemed harmless initially, but emitting more than what the ecosystem could absorb had proven to be disastrous with impacts such as greenhouse effects. Global atmospheric effects are listed below (Heijungs et al., 1992; Young et al., 1999; Azapagic and Perdan, 2000; Song et al., 2002; Joillet et al., 2003; Bare et al, 2003 and 2006). Each atmospheric effect would be further elaborated in the next chapter. • Global warming potential • Ozone depletion potential 21 Chapter 2: Sustainability • Photochemical oxidation potential • Acidification potential • Eutrophication/Nutrification potential 2.2.2.2. Environmental Efficiency Environmental efficiency is the effective use of the resources in the manufacture of the product. There are two classes of efficiencies for a chemical process – material efficiency and energy efficiency (Schwarz et al., 2002; Marteel et al., 2003; Martins et al., 2007); these can also be termed as material intensity and energy intensity. These efficiencies or intensities are defined as the amount of raw materials/energy required per unit output of the process. As the term ‘efficiency’ suggests, it would be desirable if the amount of raw materials and energy required per unit output for the process is minimized. This not only reduces cost of raw materials but also conserves the natural resources. Engineers could choose to be more specific by minimizing the use of nonrenewable resources due to their limited supply (Martins et al., 2007). Nonrenewable resources are resources that cannot be replenished in short periods of time; examples of nonrenewable resources are minerals, metals, fuel oil, natural gas and coal. Other than material and energy efficiencies, Azapagic and Perdan (2000) included three more indicators in this category of environmental efficiency: (1) material recyclability, (2) product durability, and (3) service intensity. Material recyclability is the potential of the product material to be recovered for other applications. This would reduce the amount of fresh raw materials required and the amount of waste disposed. Product durability indicates that the product can be used over a relatively long period 22 Chapter 2: Sustainability without being depleted or consumed. With higher durability, it would lead to a reduction in the volume throughput for the process, resulting in the decrease in consumption of raw materials. Lastly, service intensity would be associated with the reuse, remanufacture or recycle of products. Instead of selling their products to customers, companies would lease out their products. As such, these products would be returned to the companies and not be disposed off after the customers have obtained utility satisfaction. This allows the companies to lease the returned products to other customers (reuse), to restore and refurbish the return products (remanufacture), or to recover the materials for other purposes (recycle). 2.2.3. Societal Indicators This set of indicators focuses on corporate social responsibility (CSR) by relating human well-being to the activities of business. Various societal indicators have been provided by Afgan et al. (2000 and 2007), Azapagic and Perdan (2000), Sikdar (2003), Krajnc and Glavič (2005) and Hienzle et al. (2006). Some of these societal indicators are number of jobs/employees, accident frequency, income distribution and amount of capital invested by the company in social and community projects. As noted, this set of indicators pertains to working environments and social rights which are largely dictated by the company and/or government policies, and not by the process design/operation. Hence, the social indicators are not relevant in the design and/or operation optimization of chemical processes, which, however, should minimize the potential for accidents. 23 Chapter 2: Sustainability 2.2.4. Energy Sustainability On earth, the primary source of energy is the sun. Without solar energy, there would be no life forms on earth. To apply the concept of sustainability, it would require that the embodied solar energy in the resources used do not exceed the amount of energy supplied by the sun. There are two ways in which solar energy can be measured – emergy and area. Emergy is the embodied solar energy in a service or product, and it is measured in sejoules (Brown and Herendeen, 1996; Singh and Lou, 2006). For a sustainable system, the emergy flow out of the system boundaries should be more than or equal to the energy flow into the system boundaries. Transformation factors are required to bring all components to the same unit of sejoules; some transformation factors are provided in Brown and McClanahan (1996). Singh and Lou (2006) had employed this concept to convert all materials to the same unit (sejoules) prior to computing the ratios for the analysis of the system. Solar energy has to be intercepted by the earth’s surface before it can become useful; also, the area of the earth’s surface is limited. Hence, Narodoslawsky and Krotscheck (2000) had introduced ‘area’ as the unit for solar energy. For a process, the sustainable process indicator (SPI) has to be computed which comprises the ‘area’ for raw materials, energy, process installation, staff and product dissipation. The objective is to minimize the total ‘area’ required. The conversion or transformation factors of all components used in a process into emergy or area are not readily available; hence, the use of emergy and area is rarely applied to chemical processes. 24 Chapter 2: Sustainability 2.3. Steps taken towards Sustainability As discussed in Section 2, sustainability consists of three aspects: economic, environmental and social factors. For operations and technology development, the focus would be on the economic and environmental aspects. Solutions to improve environmental impacts are (1) integrated networks, (2) green chemistry and (3) white biotechnology, just to name a few. The economic feasibility of these solutions would usually be the first criteria for their implementation. Usually, in order to minimize the environmental impact, additional investments are incurred. Thus, there would be a tradeoff and multi-objective optimization can be implemented to analyze the process system. The decision maker would have to decide how much economic benefit he is willing to forgo for the addition environmental protection. As for social aspect, since it is policy dependent, manipulation of the technology has no significant effect over it; management teams as well as governmental bodies would have to step in to make improvements in this sphere of concern. 2.3.1 Integrated etworks Most of the time, feed streams have to be heated up to higher temperatures for chemical reactions to take place at a desirable rate in the reactor. The product stream, which is often a hot stream, has to be cooled for further processing, storage or selling. Here, the process first requires heat input and later heat extraction. The amount of heat input can be reduced if some of the heat from the hot product stream is transferred to the cold feed. As a result, the amount of fuel burnt to provide energy is decreased, reducing the production of greenhouse gases and the depletion of nonrenewable resources. With heat integration, 25 Chapter 2: Sustainability it reduces the amount of cooling water required to lower the temperature of the hot product stream. Warm water discharged into the sea has detrimental effects on aquatic life. Multiple exchanges of energy in a process require a heat exchanger network (HEN). Besides heat integration, components can also be transferred from one stream to another; complex exchanges would lead to a mass exchange network (MEN). Mass exchange operations include absorption, adsorption, extraction, ion exchange, leaching and stripping. The operations are performed to selectively transfer certain undesirable species from numerous waste streams to mass separating agents (MSAs). As such, the waste streams are almost free from pollutants which make them safe for disposal, promoting the idea of pollution prevention (Masters, 1998). Some examples of MSAs are solvents, adsorbents, ion-exchange resins and stripping agents (El-Halwagi, 1997). Another concept derived from mass exchanges is industrial ecology (or industrial ecosystems). Industrial ecology (IE) involves in the change of focus of a chemical process from an open system, in which raw materials enters the system and exits as products and wastes, to a closed system where wastes become inputs for other chemical processes. The industrial symbiosis at Kalundborg, Denmark is an excellent example of an IE (Jacobsen, 2006). In 1961, as the amount of ground water was limited, the oil refinery, Statoil, embarked on a project to draw its water supplies from Lake Tissø’s surface water. Subsequently, the number of partners increased while about 20 more collaborative projects were initiated (refer to http://www.symbiosis.dk, accessed in January 2008). These partners have built HEN and MEN amongst themselves. The main products and services are heat and power production, motor fuels, ammonium 26 Chapter 2: Sustainability thiosulphate, plasterboard products, remediation of contaminated soil, pharmaceutical products, enzymes and waste water treatment service. 2.3.2 Green Chemistry Green chemistry (or otherwise known as sustainable chemistry) refers to the chemistry that would utilize raw materials efficiently, minimize the production of waste and also avoid the utilization and generation of hazardous substances (Anastas and Warner, 1998; Sheldon, 2007). As a result, green chemistry would lead to (1) waste reduction, (2) elimination of costly end-of-the-pipe treatments, (3) safer products, and (4) reduction in the use of energy and resources. These would improve the competitiveness of chemical manufacturers and their customers. Under green chemistry, the reduction of hazardous materials through careful selection of feedstocks and reagents as well as the use of alternative solvents is emphasized. Green chemistry should also include the use of catalysts to produce desired products with high selectivity (Marteel et al., 2003). A subset of green chemistry is white biotechnology, also known as bio-catalysis. White biotechnology is a technology where cells and enzymes are employed to manufacture products. These techniques have reduced environmental burden, use renewable resources, and promise better products at lower manufacturing costs in terms of energy, water and capital costs (Villadsen, 2007). In Singapore, two biodiesel plants are being built on Jurong Island; the first plant belongs to Peter Cremer while the second is a joint venture between Wilmar Holdings and Archer Daniels Midland Company (refer to http://www.channelnewsasia.com/stories/singaporebusinessnews/view/175426/1/.html, 27 Chapter 2: Sustainability accessed in January 2008). These projects would boost the petrochemical industry in Singapore. An example of white biotechnology is the production of ethanol from pentoses (Villadsen, 2007). Jack Pronk and coworkers made an amazing discovery – a fungus found in elephant manure could convert xylose to xylulose. Prior to the finding, the isomerization of xylulose to xylose proceeds via a complicated reduction/oxidation path of around xylitol. With the discovery of the fungus, the associated gene was extracted from the manure and inserted into Saccharomyces cerevisiae. This modified yeast strain showed improvements in the specific production rate of ethanol; however, some time is still required before this process could be implemented on an industrial scale. 2.4. Summary In this chapter, the concept of sustainable development and several related issues are described. There are three spheres of sustainability: economic development, environmental stewardship and social equity. Indicators which represent the performance of a chemical process in each sphere are presented. Each of these indicators does not solely determine the impact in its own sphere; it has indirect impact on another or all spheres simultaneously. Economic indicators measure the profitability of a chemical process. These usually are the first criteria for companies; if the process is not economically feasible, the project would be aborted or vice versa. Environmental indicators are concerned with efficient use of raw materials and energy in the process as well as the environmental impacts caused by emissions. The environmental impacts include human toxicology, 28 Chapter 2: Sustainability ecotoxicity, global warming, ozone depletion, acidification, eutrophication and photochemical oxidation. Societal indicators measure different aspects of working conditions and regulations. As these indicators are based on company and governmental policies, it is not appropriate to include them in the analysis and optimization of a chemical process. Finally, various solutions available to promote sustainability are outlined. The exchanges of energy and materials, within the process or between processes, have promoted the reduction of pollutants emitted and raw materials (as well as energy) required. Green chemistry is used to identify alternative reaction schemes that uses and produces less toxic and harmful materials; in addition, energy requirements of the process can be reduced simultaneously. As a subset of green chemistry, white biotechnology is introduced where living cells are used to manufacture products that were once produced by chemical synthesis. With this change, less harsh operation conditions are required and usually lesser toxic or flammable materials are handled. 29 Chapter 3: Environmental Impact Chapter 3 Environmental Impact 3.1. Introduction Environmental issues had slowly evolved and taken centre stage in chemical engineering domain. Society, environmentalists and governments realized the importance of environment protection when such issues had become so severe that they could not go unnoticed. Environmental issues include the extinction of wildlife species (plants and animals) and the infamous issue of global warming. Recently, in December 2007, discussions held in Bali to renew the Kyoto Protocol agreement had seen numerous countries making a concerted effort to tackle the global warming concern (refer to http://unfccc.int/2860.php, accessed in January 2008). Tracing the footprints of global development, it is observed that it was the industrial revolution that exacerbated environmental problems, with chemical processes playing an important role in this revolution. Hence, steps have to be taken to make such development environmentally friendly so that the world can trod the path towards sustainable development. To incorporate environmental protection into chemical processes, more investment would generally be required. First of all, waste treatment plants would have to be incorporated to ensure that the output streams meet the environmental regulations on waste concentrations/amounts. Secondly, investment is required in research and development (R&D), to optimize the operating conditions or in greening the reaction chemistry. Although more money is required for environmental stewardship, why are businesses incorporating environmental metrics into managerial decisions? Firstly, with 30 Chapter 3: Environmental Impact respect to the region’s environmental regulations, environmental metrics help to accomplish regulatory compliance. Secondly, these metrics assist the company in monitoring their progress, seeking to target improvements at more problematic areas. Thirdly, the company could use the metrics to optimize internal operations to meet customer demand more effectively; simultaneously, the company would have a competitive advantage over others in the same industry. Fourthly, environmental metrics enhance corporate reputation for investors and insurers (Marteel et al., 2003). Lastly, contrary to the idea that more money is invested for environmental protection, it has been reported that pollution prevention efforts could result in savings for the company. Harris et al. (2005) noted that the environmental stewardship in Minnesota Mining and Manufacturing (3M) had resulted in US$950 million savings from 1975 to 2005. In the attempt to make chemical processes more environmentally friendly, there is a need to measure the environmental impact of the process. One would then ask: is it possible to measure the environmental impact of a process? Researchers have pondered over this question and derived a myriad of indicators to quantify the environmental impact of a process. The main reason for many different indicators is that environmental impact is a huge umbrella consisting of many areas of concern, and thus each indicator depends on how inclusive the analyst is and its significance. In this chapter, we aim to provide a review of the various environmental indicators that are available and useful to chemical engineers for process design and operation. Eventually, it is still dependent on the engineer, manager and/or company to choose the appropriate indicator(s) that would suit the needs of their chemical process. 31 Chapter 3: Environmental Impact 3.2. Overview of Environmental Concerns In order to obtain an overview of the environmental concerns of a process, the engineer can draw up a life cycle assessment (LCA) of the product (Masters, 1998). LCA involves the use of energy and material balances at every stage of the life cycle of a product, elucidating the air and water pollution and solid waste problems (Figure 3.1). It also identifies the inputs required at the start of the life cycle, making the entire process complete. Azapagic and Clift (1999) noted that the optimization of a process alone may be misleading as the reduction of emissions and environmental impact at the specific process may transfer their burden upstream or downstream of the cycle. Thus, arguably, performing a LCA for each product is ideal; however, huge amount of information is required and as a result, the assessment process would be time consuming, requiring a lot of man-hours. Noticing these limitations, Azapagic and Clift (1999) employed the ‘cradle to gate’ approach where the process considers the transport and preparation of raw materials, the transformation of the raw materials to the desired product, without considering how the product is utilized and disposed off. Since engineers are often faced with the need to optimize a chemical process, for a quick estimation, he/she should optimize the process alone (focusing on the pink box in Figure 3.1) for environmental concerns while qualitatively foreseeing the probable shift of environmental burdens to elsewhere. Demands for energy and waste disposals of most chemical processes are often met by facilities/organizations outside the battery limits of the process. As such, it is highly probable that there would be transfer of environmental burden from the chemical 32 Chapter 3: Environmental Impact process to these external bodies. Hence, the environmental impacts inflicted by those offsite activities are under scrutiny. For example, in order to reduce process emissions of the process by a small amount, more electricity is required for its equipment, resulting in the generation of substantial pollution at a power station. It is important for engineers to avoid such situations and to ensure that the design represents a sensible option for the environment as a whole (Sharratt, 1999). Energy, Raw materials, Air, Water INPUTS Raw materials acquisition/transport PRODUCT LIFE CYCLE Materials processing Product Recycling Product manufacturing Remanufacturing Packaging and distribution Product use Product reuse Disposal OUTPUTS Solid wastes, Air emissions, Water effluents, Waste heat and energy recovery Figure 3.1: Life Cycle Assessment (Masters, 1998) 33 Chapter 3: Environmental Impact Environmental indicators are classified into two groups: (1) environmental metrics, and (2) environmental impacts. The former consists of ratios mainly, indicating the efficiency of the process and other factors depending on the numerator and denominator chosen. The latter, on the other hand, consists of the various environmental impacts that the process could inflict; they include global warming, photochemical oxidation, etc. Examples of environmental assessment tools are the Eco-indicator99 (available at http://www.pre.nl/eco-indicator99/default.htm); CST95 used by Song et al., 2002; WAR (Mallick et al., 1996; Cabezas et al., 1999; Young et al., 1999 and 2000); the environmental priority strategies, EPS 2000 (available at http://eps.esa.chalmers.se/); the tool for the reduction and assessment of chemical and other environmental impacts, TRACI (Bare et al., 2003 and 2006); impact assessment of chemical toxics, IMPACT 2002+ (Joillet et al., 2003); and CML 2 (available at http://www.leidenuniv.nl/cml/ssp/projects/lca2/lca2.html). 3.3. Environmental Metrics The first set of environmental indicators are in the form of ratios, or generally known as metrics. A metric often indicates the efficiency of the process. Generally speaking, the more efficient a chemical process, the lesser material input and/or energy per unit of product is required. Typically, a chemical process would involve both inputs and outputs. Inputs would include the raw materials and the fuel/energy for manufacturing. Outputs encompass the products and waste material streams. 34 Chapter 3: Environmental Impact 3.3.1. Material Metrics Material metrics are measures to identify the efficiency of the process. Efficiency is defined as the ability of the process to convert a large proportion of the inputs into desired products. These metrics can be either in the form of ratios (dimensionless or otherwise) or flow rates of materials. The need for efficiency is not only to conform to the environmental regulations set by statutory boards, but also to minimize the manufacturing cost (Sharratt, 1999). Thus, there are strong reasons for an engineer to utilize material metrics as a yardstick for both new projects and existing process operations. To start off, the metrics in the ratio form would be presented. Ratios, as one can expect, would involve both a numerator and a denominator. As mentioned, efficiency is the key to any chemical process-cum-business; hence, the relevant numerator and denominator should be chosen. For a chemical process, the mass of the desirable product formed is important. The ratios that incorporate mass of product as the numerator or denominator are summarized in Table 3.1. If the system studied is an energy system whereby the product is the energy generated, the engineer should replace mass of product to the amount of energy produced. Energy sources such as fossil fuels or materials such as minerals and ores are considered as nonrenewable resources. The replacement rate of these resources is negligible, thus their usage is often of a concern. Some authors have incorporated the idea of nonrenewable resources and have segregated the raw materials utilized into renewable or nonrenewable. As such, alternative metrics may be written, where either the numerator or denominator is concerned with the amount of nonrenewable resource(s) used. Not only 35 Chapter 3: Environmental Impact would the ratios measure the efficiency of the process, they would also seek to minimize the rate of extraction of the nonrenewable resources. Examples of these alternative metrics are: (1) emergy of the nonrenewable resources per emergy of renewable resources (Lou et al., 2004; Singh and Lou, 2006); (2) mass of nonrenewable resources to mass of desired product (Martins et al., 2007). Table 3.1: Material Metrics Material Metric Mass of product Mass of raw material Optimize Reference Maximize Marteel et al. (2003) Mass of raw material Mass of product formed Minimize Mass of non - products or waste Mass of product Minimize Molecular weight of product Molecular weight of all substances produced Carbon in organic raw material Carbon in product Mass of inorganics formed Carbon in product Mass of inputs/out puts Mass of product Another set of material metrics deal Afgan et al. (2000 and 2007); Dietz et al. (2006) Schwarz et al. (2002); Dietz et al. (2006); Sheldon (2007) Maximize Sheldon (2007) Minimize Lange (2002) Minimize Lange (2002) Minimize Hienzle et al. (2006) with the rate of release of pollutants/wastes/non-products into the various material sinks – soil, water and air. Typically, it would be assumed that solid waste would be transmitted to soil media, liquid emissions to water media and gaseous emissions to air media. However, there is a limitation to this assumption. Mass transfer would occur naturally between water and air media through a process called vapour-liquid equilibrium. In addition, water can seep 36 Chapter 3: Environmental Impact into soil and wastes would enter the soil media through diffusion. Although such a limitation is present, for quick estimation, the assumption is taken to be valid. For solid waste disposals, there is a need to handle their disposals properly, either through landfill or incineration. Hence, it is important to track the amount of waste produced by the process. Stefanis et al. (1995) used a metric called solid mass disposal (SMD) which measures the mass of solid waste produced per unit time. Next, liquid nonproducts emissions would be released into water bodies. Prior to their disposal, there is usually a need to treat wastewater to remove undesirable substances that would threaten the survival of marine organisms. The metric that could be used to measure such emissions is called the critical water mass (CTWM) which measures the mass of water pollutant per unit time (Stefanis et al., 1995). Finally, for air emissions, each country would have its own set of regulatory guidelines with regards to the permissible emission rates of various components. The engineers could either improve their process chemistry or operating conditions to meet the expected standards. Otherwise, post treatment such as amine scrubbers could be used to remove potentially threatening gases before they are released into the atmosphere. Stefanis et al. (1995) introduced the term critical air mass (CTAM) which measures the amount of air pollutants per unit time. Suppose the engineer strongly believes that the assumption of solid waste is transmitted to soil media, liquid emissions to water media and gaseous emissions to air media is not valid, s/he could incorporate the factors provided by Martins et al. (2007). These factors are the ratios of the solid wastes that could be intercepted by soil, water and air bodies respectively. Similar ratios are also provided for liquid discharges and air emissions. Using these ballpark figures, the engineer could calculate the total mass of 37 Chapter 3: Environmental Impact wastes that enter each media. Alternately, to simplify, the engineer could identify the most critical pollutant emissions of the process (which have larger environmental impact) and base his calculations on these pollutants (Cano-Ruiz and McRae, 1998). For example, carbon dioxide emissions via flue gas can be chosen for power generation plants. If necessary, emissions of more than one pollutant type can be analyzed simultaneously by employing multi-objective optimization. 3.3.2. Energy Metrics In a chemical process, in addition to the raw materials required for product formation, energy is also required in furnaces, steam for heating, refrigeration for cooling and electricity to drive pumps, compressors, etc. Energy metrics are formulated to evaluate the efficiency of a chemical process with regards to energy utilization. Energy, most of the time, is produced from the burning of energy sources such as coal and fossil fuels either within and/or outside the process. Consequently, carbon dioxide and other pollutants are released into the air, which have negative effects on the environment – global warming, acidification, etc. In addition, the utilization of nonrenewable resources depletes their availability to future generations. As a result, it would be desirable to minimize energy consumption of the process as much as possible. Hence, energy metric is another important criterion for analyzing the efficiency of the process. The energy metric suggested in various studies are often in the ratio form. There are basically two different ratios: (1) energy required per mass of product (Schwarz et al., 2002; Marteel et al., 2003); (2) nonrenewable energy per mass of product (Martins et al., 2007). The first metric can be computed easily by summing all the energy required for 38 Chapter 3: Environmental Impact the chemical process divided by the mass of product produced. The second metric requires the engineer to identify the energy sources, differentiating nonrenewable energy sources from renewable ones. Nonrenewable energy would comprise energy produced from the burning of fossil fuels or other nonrenewable sources; on the other hand, renewable energy would include hydroelectric power, solar energy, wind power, etc. In the absence of renewable energy, which is usually the scenario, the first metric is identical to the second. If one uses the energy metric that solely considers nonrenewable energy, it would seem to imply that the switch of nonrenewable sources of energy to renewable sources of energy (e.g. hydroelectric power) would have reduced environmental impact. However, this generalization does not always hold. Each energy source has its own set of undesired outcomes. The use of nonrenewable resources has direct environmental impacts such as global warming and acidification while simultaneously reducing its supply. On the other hand, alternatives such as hydroelectric power would result in flooding of coastal lines and also loss of biodiversity (e.g. Three Gorges Dam in China) (Sarkar and Karagöz, 1995). In addition, the build-up of water in dams traps wastes such as fertilizers, resulting in eutrophication and loss of marine diversity. Thus, the step that should be taken to reduce energy intensity is to have more efficient heat transfer equipment, reduce the energy requirements by process intensification and efficient operations (e.g., divided wall columns for distillation and reactive distillation columns) (Parkinson, 2005 and 2007), and/or by changing the chemistry of the process that has a lower energy requirement (Marteel et al., 2003). 39 Chapter 3: Environmental Impact 3.4. Environmental Impact In the earlier section, material and energy metrics, which consider the efficient use of raw materials and energy respectively, have been discussed. Besides efficiency of the process, the engineer should take into account the environmental impact imposed by the output streams (Young et al., 1999; Song et al., 2002). These impacts include: 1. Ecotoxicity Potential (ETP): (a) Human Toxicity Potential (HTP), (b) Aquatic Toxicity Potential (ATP), and (c) Terrestrial Toxicity Potential (TTP) 2. Global Warming Potential (GWP) 3. Ozone Depletion Potential (ODP) 4. Photochemical Oxidation Potential (PCOP) 5. Acidification Potential (AP) 6. Eutrophication Potential (EP) Each of the above 6 impact categories considers different aspects of the environment. The first impact category, ETP, is concerned with local toxicological effects of the output on all living organisms – humans, animals, plants and marine life. The next five categories are concerned with global atmospheric effects, which eventually may affect all living organisms. All these categories are explained individually and their respective indicators are explored below. 3.4.1. Toxicity Potentials From the list above, there are three areas of toxicity potentials – humans, aquatic and terrestrial organisms. Together, they are generally termed as ETP (Bare et al., 2006). The 40 Chapter 3: Environmental Impact three are closely related to the toxic emissions from the plant, threatening the survival of living things on earth. A general indicator for ecotoxicity is obtained by measuring the amount of 2,4-dicholorophenoxy acetic acid (2,4-D) equivalents emitted by the process (emissions to air and water medium). Toxicity of 161 compounds have been tabulated in terms of 2,4-D equivalents (Bare et al., 2003; Bare et al., 2006). Besides ETP, toxicity can be measured based on the medium in which the toxic materials are emitted. Emissions to the air are intercepted by organisms via inhalation or dermal exposure. The limits, which organisms can be exposed to, are measured by TWATLV. Hence, an indicator available in WAR algorithm is called the human toxicity potential by inhalation and dermal exposure (HTPE) and is measured by the inverse of the TWA-TLV values (Young et al., 1999 and 2000; Cano-Ruiz and McRae, 1998). Emissions to water and solid wastes may be consumed by organisms. As this affects humans and other terrestrial organisms, two indicators are used for each category: human toxicity potential by ingestion (HTPI) and TTP. Both indicators are related to LD50. For chemicals which do not have the LD50 classification, a value could be estimated using molecular methods (available in the WAR programme, http://www.epa.gov/ord/NRMRL/std/sab/war/sim_war.htm). HTPI downloadable and TTP at are calculated as the inverse of LD50; substances with higher HTPI and TTP values are more toxic than those with a smaller HTPI and TTP values. Another indicator similar to HTPE and HTPI combined is called the potential chemical risk (Vincent et al., 2005; Martins et al., 2007), which measures the toxicology risk humans face when handling and utilizing hazardous chemicals. It is calculated based on the MSDS’s R-phrases of the chemical, its material intensity and the usage frequency. 41 Chapter 3: Environmental Impact Other than terrestrial and human beings, aquatic organisms also need to be included in the analysis. The measure for these organisms is called ATP. In WAR, toxicity for aquatic species is coherent with LC50. ATP is calculated by taking the inverse of LC50; similar to HTPI and TTP, a high value of ATP indicates that the substance is more toxic and vice versa. Alternatively, Heijungs et al. (1992) provided a comprehensive list of substances along with their HTP, TTP and ATP values. The HTP values given are based on the medium into which the substance is emitted (i.e. air, water or soil). Thus, the HTP for a particular substance is calculated by summing the products of the emission of the substance with its respective HTP values for each of the three mediums. Thereafter, the overall HTP for the process is calculated by summing the calculated HTP values of each substance. The TTP for the process is simply calculated as the sum of the product of the TTP value of a particular chemical and its emission rate for each of all chemicals involved. For ATP, it is calculated using same formula as TTP except that ATP values are employed instead. 3.4.2. Global Warming Potential The Earth’s climate has been changing progressively throughout its history, ranging from ice ages to long periods of warmth. In the past, natural evolution such as the changes in the Earth’s orbit and natural disasters such as volcanic eruptions have contributed to the changes in the Earth’s climate. Since late 18th century, human activities associated with Industrial Revolution have also changed the composition of the atmosphere which inadvertently has an effect on the Earth’s climate. 42 Chapter 3: Environmental Impact Since Industrial Revolution, human activities such as deforestation and the burning of fossil fuels (e.g. coal and oil) have produced more heat-trapping greenhouse gases than the atmosphere could absorb, causing their concentrations in the atmosphere to increase significantly. These gases trap heat within the atmosphere rather than allowing the heat to escape into space, analogous to the function of glass panels of a greenhouse. Some examples of greenhouse gases are carbon dioxide (CO2), methane (CH4), halocarbons and nitrous oxide (N2O). Indeed, greenhouse gases are necessary to life as they keep the planet's surface warmer than it otherwise would be. However, as the concentrations of these gases continue to increase in the atmosphere, the Earth's temperature is climbing above past levels. In addition, global temperatures are rising due to indirect impact of the depletion of the ozone in the stratosphere, causing more radiation to be intercepted by the Earth’s surface. Ozone depletion, however, would be measured as a separate indicator described in the next section. GWP is determined by comparing the amount of infrared radiation a unit mass of a chemical can absorb in 100 years as compared to the amount of infrared radiation a unit mass of carbon dioxide can absorb over the same time span. This takes into account the decay of the chemical in the atmosphere over this time span as well. It is an indicator to measure the impact of the greenhouse emissions that the process releases. There is a general consensus to use CO2 as the common unit for GWP. Converting all greenhouse gases emissions to CO2 equivalents, comparisons can be made for different operating conditions, design schemes or even process chemistry. Table 3.2 provides the values of GWPs for several essential chemicals; see Daniel and Velders (2006) for the complete list. Although CO2 has a very low GWP potential when compared to the rest of the 43 Chapter 3: Environmental Impact components (about 2-3 orders of magnitude lower), its emission is of greatest concern because it is present in quantities many times greater than other pollutants (Masters, 1998; Elkamel et al., 2008); thus, it has the greatest impact on global warming. Table 3.2: GWPs and ODPs of some substances; refer to the respective references in the foot-note for the extended list Species Chemical Formula GWP (100 years)* 1 21*** 298 4750 10890 6130 10040 7370 7140 1890 1640 1400 146 1810 77 609 725 2310 122 595 5 ODP (Montreal Protocol)**    1 1 0.8 1 0.6 10 3 6 1.1 0.1 0.055 0.02 0.022 0.11 0.065 0.025 0.033 0.6 Carbon dioxide CO2 Methane CH4 Nitrous oxide N2 O CFC-11 CCl3F CFC-12 CCl2F2 CFC-113 CCl2FCClF2 CFC-114 CClF2CClF2 CFC-115 CClF2CF3 Halon-1301 CBrF3 Halon-1211 CBrClF2 Halon-2402 CBrF2CBrF2 Carbon tetrachloride CCl4 Methyl chloroform CH3CCl3 HCFC-22 CHClF2 HCFC-123 CHCl2CF3 HCFC-124 CHClFCF3 HCFC-141b CH3CCl2F HCFC-142b CH3CClF2 HCFC-225ca CHCl2CF2CF3 HCFC-225cb CHClFCF2CClF2 Methyl bromide CH3Br * Daniel and Velders (2006) ** From the Handbook for the Montreal Protocol on Substances that Deplete the Ozone Layer - 7th Edition (2006), Annexes A-E in http://ozone.unep.org/Publications/MP_Handbook/Section_1.1_The_Montreal_Protocol/ *** Wuebbles (1995) 44 Chapter 3: Environmental Impact 3.4.3. Ozone Depletion Potential Ozone (O3) is a molecule with three oxygen atoms. At the ground level, it is a component of smog which causes respiratory difficulties and eye irritations in human beings and animals. Ozone at ground level is not related to the environmental problem of ozone depletion, but it plays a major role in photochemical oxidation which is discussed in the next section. At the stratospheric level, ozone present is of great importance to all living things on the Earth. Its ability to filter potentially damaging ultraviolet light from reaching the Earth's surface protects us from direct exposure to the Sun. Stratospheric ozone is found to be depleting over the years and ozone holes have been discovered. This is also an indirect cause of global warming as larger amounts of radiation could reach the earth’s surface. The culprits for ozone depletion are chlorofluorocarbons (CFCs), chlorinated hydrocarbons and other ozone depleting substances. The mechanism for ozone depletion is given by (Masters, 1998). Overall: X + O3 XO + O2 XO + O X + O2 O + O3 2O2 The overall chemical reaction shows that oxygen radicals react with ozone to form oxygen as the eventual compound and this reaction is promoted by the presence of free radicals (X ), usually chlorine or bromine. Oxygen radicals are present in the atmosphere due to photo-dissociation of ozone by sunlight. ODP is the potential of a chemical to deplete the ozone layer in the stratosphere. It is determined by comparing the rate at which a unit mass of the chemical reacts with ozone to form molecular oxygen to the rate at which a unit mass of CFC-11 45 Chapter 3: Environmental Impact (trichlorofluoromethane, CFCl3) reacts with ozone to form molecular oxygen. Thus, ODP of CFC-11 is given a base factor of 1.0. The ODP of several chemicals are provided in Table 3.2; see the cited reference in this table for the complete list. The ODP of a process stream can be computed by multiplying the compound flowrates with their respective ODPs and summing them. ODP is also used for regulatory purposes to restrict the emissions of ozone depleting substances. The Montreal Protocol on Substances that Deplete the Ozone Layer and its subsequent Amendments, which place regulations on the production and use of halocarbons internationally, determine the phase-out of ozone depleting substances based on their ODPs. 3.4.4. Photochemical Oxidation Potential Photochemical oxidation is the process where photochemical smog is formed when volatile organic compounds (VOCs) react with oxides of nitrogen under the influence of sunlight according to the simplified reaction scheme: VOCs + NOx + Sunlight Photochemical smog The reaction scheme is, in fact, far more complex and many different photochemical oxidants can be formed. Some examples of the oxidants formed are ozone (O3), formaldehyde (HCHO), peroxybenzoyl nitrate (PBzN), peroxyacetyl nitrate (PAN) and acrolein (CH2CHCOH). As observed, these photochemical oxidants are secondary pollutants; thus, there is no direct method to calculate the emissions of photochemical oxidants from a chemical process. PCOP of a chemical compound is determined by comparing the rate at which a unit mass of a chemical reacts with hydroxyl radicals (OH ) to the rate at which a unit 46 Chapter 3: Environmental Impact mass of ethylene reacts with OH . The hydroxyl radical is the compound responsible for the initiation of VOCs oxidation. The PCOPs of some VOCs are given in Table 3.3. The extended list of compounds with their respective PCOPs is available on pages 83 to 86 in Heijungs et al. (1992). Since the reactivity of compounds is measured relative to the reactivity of ethylene, PCOP of ethylene is 1.0. Table 3.3: PCOPs of some organic substances (Heijungs et al., 1992) Species Chemical Formula PCOP (range in brackets)* Ethane C 2 H6 0.007 (0.000-0.030) Tetrachloroethylene C2Cl4 0.005 (0.000-0.020) Ethanol C2H5OH 0.268 (0.040-0.890) Acetone CH3COCH3 0.178 (0.100-0.270) Ethyl acetate C4H8O2 0.218 (0.110-0.560) Ethylene C 2 H4 1.000 (1.000-1.000) Acetylene C 2 H2 0.168 (0.100-0.420) Benzene C 6 H6 0.189 (0.110-0.450) Formaldehyde HCHO 0.421 (0.220-0.580) * PCOP values are based on three scenarios and nine days while the ranges are based on three scenarios and eleven days; the scenarios are: Germany-Ireland, France-Sweden, UK. Photochemical oxidants are notorious for causing many respiratory problems such as coughing, shortness of breath, headache, chest constriction, and irritation of eyes, nose and throat. The most abundant photochemical oxidant formed is ozone, which is known to cause damage to tree foliage and to reduce growth rate of certain sensitive tree species. There is also a drop in the yields of major agricultural crops such as staple food sources: corn and wheat. Ozone, however, is not responsible for the most common complaint of smog: eye irritation; it is caused by other photochemical oxidants, where HCHO, PBzN, PAN and CH2CHCOH are the main culprits. 47 Chapter 3: Environmental Impact 3.4.5 Acidification Potential Acidification is the change in pH values (hydrogen ion concentration, H+) of water and soil systems. Substances which contribute to the change in acidity are sulfur dioxide (SO2), oxides of nitrogen (NOx), hydrogen chloride (HCl), ammonia (NH3) and hydrogen fluoride (HF). Some processes that emit these substances include the burning of fossil fuels in furnaces and boilers, and the alkylation process where HF is used. AP is determined by comparing the rate of release of H+ ions in the atmosphere as promoted by a chemical to the rate of release of H+ ions in the atmosphere as promoted by SO2. In the WAR algorithm (Young and Cabezas, 1999), AP values for different components are obtained from Heijungs et al. (1992); these values are reproduced in Table 3.4. On the other hand, the acidification model in TRACI (Bare et al., 2003 and 2006) makes use of the results of an empirically calibrated atmospheric chemistry and transport model to estimate total North American terrestrial deposition of expected H+ equivalents as a function of the emissions location. Table 3.4: Acidification potentials (Heijungs et al., 1992) Species Sulfur dioxide Nitrogen monoxide Nitrogen dioxide Nitrogen oxides Ammonia Hydrochloric acid Hydrogen fluoride Chemical Formula SO2 NO NO2 NOx NH3 HCl HF AP 1.00 1.07 0.70 0.70 1.88 0.88 1.60 Acidification of water bodies would affect the survival and the reproduction of marine life and aquatic plants. In addition, acidification of soil systems (both topsoil and 48 Chapter 3: Environmental Impact subsoil, with a greater impact on the latter) would lead to lower vegetative yields, reduced grazing pasture and limitation in crop variety. It would also contribute to wider catchment problems such as weed infestations, salinity and erosion. 3.4.6. Eutrophication Potential Eutrophication, also known as nutrification, is the addition of mineral nutrients to the ecosystem which increases production (i.e. excessive plant growth and decay). This process can occur both on land and in water. The consequence of the production is a reduction in the amount of oxygen present in the atmosphere, impairment of the quality of water as well as reductions in fish and animal populations. The substances released to air that contribute to eutrophication are nitrogen monoxide, nitrogen dioxide, NOx, phosphorus and ammonia. In addition, other releases to water that contribute to eutrophication include phosphorus, phosphate, nitrogen, nitrate, chemical oxygen demand (COD), biochemical oxygen demand (BOD) and ammonia (Heijungs et al., 1992; Bare et al., 2003 and 2006). EP is the potential biomass (C106H263O110N16P) of the emitted substance relative to that of a reference substance (Heijungs et al., 1992). Bare et al. (2003) had chosen N2 as the reference component. On the other hand, the basis for EP is PO43- in other works (Heijung et al., 1992; Azapagic and Clift, 1999; Jolliet et al., 2003). Either basis can be used, although the latter is more common, as long as they are kept consistent throughout the analysis. Table 3.5 provides a list of compounds with their respective EP values. 49 Chapter 3: Environmental Impact Table 3.5: Eutrophication potential of compounds (Heijungs et al., 1992) Species Nitrogen monoxide Nitrogen dioxide Nitrogen oxides Ammonia Nitrogen Phosphate Phosphorus Chemical Oxygen Demand (as O2) Chemical Formula NO NO2 NOx NH3 N PO43P COD EP 0.20 0.13 0.13 0.33 0.42 1.00 3.06 0.022 3.5. Aggregated Indicator As seen above, there are many indicators for various environmental impacts. The concept of translating environmental objectives into a quantitative measure and to aggregate them into a single or several indicator(s) is still novel and not well established yet. Thus, several works had proposed different methods of calculating environmental indices and several of them are discussed below. The method of WAR algorithm was proposed by Young and Cabezas (1999) and Young et al. (2000), and was employed in several works, e.g., by Kim and Smith (2004 and 2005) in the recovery of acetic acid from aqueous waste mixtures and by Ramzan and Witt (2006) in optimizing a distillation column. The total potential environmental impact (PEI) is calculated by summing up the index or impact of the various impact categories multiplied by the weights assigned to each of them (see equation 3.1). The environmental impact categories (denoted by j) considered are: HTPI, HTPE, ODP, GWP, AP, PCOP, ATP and TTP. PEI = EnvCat ∑w I j j (3.1) j =1 where wj are the weighting factors (determined by Ecoindicator-99) and 50 Chapter 3: Environmental Impact I j = ∑ M i ϕ i , j + Qr ϕ j , E (3.2) i Here, ϕ i, j is the normalized impact factor of impact category j of component i, ϕ j, E is the normalized impact factor of impact category j for an energy source, Mi is the mass flow rate of component i, Qr is the energy consumption. Ij is the sum of the contributions of waste and energy streams for the impact category j, and indicates the magnitude of deterioration the process has on the particular impact category j. The WAR methodology is very similar to Eco-Indicator-99 (Goedkoop and Spriensma, 2001) and IMPACT2002+ (Jolliet et al., 2003), except that the impact categories considered are different. Another type of aggregated index is called atmospheric hazard index (AHI) by Gunasekera and Edwards (2003). A total of five environmental impact categories related to atmosphere are considered. They are toxicity, photochemical smog, acid deposition, global warming and stratospheric ozone depletion. For each of the impact category j, an impact hazard value (Hji) can be calculated for each chemical component i. Thereafter, the normalized impact factor (Yji) is computed; Yji has a range from 0 to 10 where a value of 0 indicates no impact of chemical component i while a value of 10 indicates catastrophic impact. The weighted category hazard (WCHi,j) gives the atmospheric environmental impact hazard associated with an impact category j for a chemical i, based on the amount of space affected on Earth, its indirect impacts and its reversibility. The formula for WCHi,j is as follows:  3  WCH i , j = Yji ∏ (1 + IFj ,k )  k =1  (3.3) where IFj,1 = importance of spatial environment (local, urban, regional, continental, global), IFj,2 = importance of indirect impacts (functional, ecological, species diversity) 51 Chapter 3: Environmental Impact and IFj,3 = importance of reversibility of the impact which is dependent on the affected level of the food chain (energy input to plants, primary producers, consumers). Chemical atmospheric hazard (CAHi) is total atmospheric hazard posed by the chemical i when it is released catastrophically into the environment and it is calculated using the equation: CAH i = ∑ WCH i , j (3.4) j Thereafter, the atmospheric hazard index (AHI) is computed by summing CAHi for all the relevant components: AHI = ∑ CAH i (3.5) i The method of AHI has been deemed as complicated, and a recent aggregation method called Green Degree has been proposed by Zhang et al. (2008). A total of nine environmental components have been considered. They are: (a) GWP obtained from the Intergovernmental Protection on Climate Change (IPCC); (b) ODP obtained from the World Meteorological Organization (WMO); (c) PCOP obtained from IMPACT 2002+ V2.01; (d) AP obtained from Environmental Design of Industrial Products (EDIP); (e) EP from IMPACT 2002+ V2.01; (f) ecotoxicity potential to water (EPW) obtained from TRACI 2.0; (g) ecotoxicity potential to air (EPA) obtained from TRACI 2.0; (h) human carcinogenic toxicity potential to water (HCPW) obtained from IMPACT 2002+ V2.01; (i) human non-carcinogenic toxicity potential to water (HNCPW) obtained from IMPACT 2002+ V2.01. These components are aggregated to obtain the environmental impact for (or Green Degree of) a chemical compound i (GDi) as shown below. 9 GDi = −∑ 100 w j j =1 ϕi, j ϕ max j (3.6) 52 Chapter 3: Environmental Impact 9 where wj is the weighting factor for impact category j with ∑w j = 1 ; ϕi,j is the j =1 environmental impact potential of compound i for impact category j and ϕ max is the j maximum value of category j among all of the substances reported in the literature. Green Degrees of chemical compounds are then aggregated into a single index for a stream (GDS) by: GD S = M × GD mix = M × ∑ GDi xi (3.7) i Where M is the mass flow rate of the stream (kg/h) and xi is the mass fraction of compound i in the mixture. The Green Degree for the entire plant by summing the Green Degrees for the relevant streams (e.g. waste streams or emitted streams): GD = ∑ GD S (3.8) S Comparing the WAR algorithm and Green Degree, the former adds the contribution of the relevant chemical compounds for a particular environmental component before summing the various components into an aggregated indicator. On the other hand, Green Degree adds the various environmental components for a particular chemical compound, obtaining the environmental impact for that chemical compound, before summing the environmental impact of various chemical compounds into an aggregated indicator. The former is preferred if it is used in the analysis of the performance of the plant especially if one is interested to find out the contribution to each impact category. The Green Degree is preferred for the analysis of the use of different chemicals/solvents. Hence, only the former is used in this study as there are no changes in the chemicals used. 53 Chapter 3: Environmental Impact Note that the environmental components considered may differ from one study to another. As an illustration, for human toxicity, Young et al. (2000) considered human toxicity potential by inhalation and by ingestion in the WAR algorithm while Zhang et al. (2008) considered human carcinogenic and non-carcinogenic potential to water in the Green Degree. Also, depending on the database(s) chosen, the environmental components considered and their values may differ. Another method of weighting is by using the analytic hierarchy process (AHP) method. First of all, the decision maker has to do a pair-wise comparison of all the different impact categories, providing a value judgment to determine the relative importance of one category over another. The values could range from more than zero (> 0) to 9. The value of 1 indicates that both categories are equally important. Values more than 1 would indicate that category A is more important than category B; a value of 2 indicates that it is slightly more important and the value of 9 indicating that category A is much more important than B with supporting evidence. Values less than 1 would indicate that category A is less important than category B. Assume that category A is slightly more important than category B and is given a score of 3; hence, when category B is compared in terms of category A, the value given is the reciprocal of the former score, which comes up to 0.333. Ramzan et al. (2007) applied the AHP methodology in conjunction with the WAR algorithm for the optimization of environmental impacts of a distillation column. As seen from the above, aggregating all impact categories into a single indicator requires value judgment from experts. This means that the aggregated indicator is subjective and it may vary between individuals. Further, it lacks transparency. Hence, 54 Chapter 3: Environmental Impact aggregated indicators should only be used for internal purposes, and not for public reporting (Goedkoop and Spriensma, 2001). Even if the decision maker decides to give equal weight to each environmental impact category, the aggregation of these categories into an overall environmental impact indicator is questionable. As illustrated in equation 2, the index for the component j in the impact category i is calculated by normalizing the scores with a base. However, there is no concrete evidence that, for example, an index of 1 for GWP has equivalent environmental impact as an index of 1 for ODP. Hence, it is recommended that each impact category should be analyzed individually rather than aggregating them into an overall environmental impact indicator. Another method of calculating PEI is based on four factors: (1) the physical state of chemical compound released, (2) the medium receiving the chemical compound, (3) the material intensity of the compound studied, and (4) the MSDS’s R-phrases of the compound. A PEI score for the particular compound could be obtained by identifying the appropriate category based on the data of the four factors. Summing the scores together for different compounds, an overall PEI score would be obtained for the chemical process (Vincent et al., 2005; Martins et al., 2007). 3.6. Conclusions In this chapter, the motivation for incorporating environmental considerations into the evaluation of any chemical process is presented. Primarily, the driving force would be the compliance to environmental regulation as well as corporate reputation amongst investors and insurers. A yardstick is thus required to measure the environmental performance of 55 Chapter 3: Environmental Impact each chemical process. There is, however, more than one yardstick available and each of them was discussed in this chapter. There are two classes of indicators: (1) environmental metrics, and (2) environmental impact categories. The former involves mostly ratios which measure the efficiency of the process in terms of the raw materials utilized as well as the energy requirement of the process. The latter class of indicators is a huge umbrella comprising different environmental aspects which would face detrimental impacts from chemical process releases. The seven environmental impacts are: (1) ecotoxicity, (2) global warming, (3) ozone depletion, (4) photochemical oxidation, (5) acidification and (6) eutrophication. Each of these impact categories has been discussed in detail, together with the method to calculate an index for each category. Given many different indicators which measure different aspect of environmental damage, aggregation is required to combine all aspects in a single indicator for easy reference. In order to aggregate into a single indicator, weights have to be introduced. These weights would thus be determined by the analyst, and hence the aggregated indicator would be rather subjective. This indicator should thus be used internally and not for public reporting. As a solution to subjectivity, MOO can be carried out where each metric and impact category would be optimized simultaneously. It will also provide more information on individual environmental metrics and impacts. 56 Chapter 4: Process Applications Studied Chapter 4 Process Applications Studied for Economic and Environmental Criteria 4.1. Introduction In the previous chapters, the concept of sustainability has been discussed. As mentioned, engineers have control over two spheres of sustainability – economic and environmental. Economic criteria are the fundamentals of any industry, and business players need to ensure that their businesses are profitable before investing. Henceforth, the economic criteria employed in countless journals will not be discussed and documented here. On the other hand, the concept of environmental criteria is relatively novel and not as well established as the economic criteria. As discussed in Chapter 3, there is no general consensus in the environmental criteria to be used when evaluating a process. To date, many indicators are available, and engineers have to choose one or more of them based on their own discretions. It is, thus, of great interest to identify and review the applications where sustainability indicators or solely environmental objectives are used (i.e. economic and environmental, or environmental only). Many applications reported in the chemical engineering field utilize the above-mentioned criteria. They include the petroleum refining and petrochemicals industry, biotechnology, pharmaceutical and chemicals industry, downstream processing, energy systems, heat exchanger networks etc. This chapter serves to provide a comprehensive summary of the relevant applications that have been analyzed so far for economic and environmental objectives (Section 4.1) or for environmental objectives (Section 4.2); use of optimization, if any, is also noted. 57 Chapter 4: Process Applications Studied 4.2. Economic and Environmental Criteria In this section, the focus is on applications which have applied both economic and environmental criteria in assessing the applications studied. There are many ways to measure the economic performance of any process. They include operating costs, capital costs, annualized costs, profits and NPW. These are the typical economic parameters the industries employ before changes are made to the processes (be it changes in operating parameters or the evaluation of a new project). Lately, there is a growing concern to reduce the environmental stress chemical processes create. This leads to the inclusion of environmental indices when measuring the environmental performances of the chemical processes. There are a total of 47 applications analyzed for both economic and environmental criteria. These are reviewed in the following sub-sections. 4.2.1. Petroleum Refining and Petrochemicals Petroleum refining and petrochemicals form the essential and indispensable foundations of chemical engineering. There are, in all, 17 applications studied in this category (Table 4.1). These cover the production of different components (e.g. vinyl chloride, acetaldehyde, 4-(2-methoxyethyl) phenol, stearyl-3-(4-hydroxy-3,5-di-tert-butylphenyl) propanoate, methyl chloride, allyl chloride, methyl ethyl ketone, propylene glycol, acrylic acid, benzene from toluene, styrene, maleic anhydride, aldehydes), scheduling of refinery processes, supply chain of vinyl chloride and ethylene glycol, distillation of atmospheric crude oil and catalytic reforming process. Economic indicators that were normally used are manufacturing costs, capital investments, profits, NPW/NPV and PBP. For the environmental aspect of sustainability, 58 Chapter 4: Process Applications Studied indicators considered include human, terrestrial and aquatic toxicity, and other atmospheric indicators like ozone depletion, global warming, photochemical oxidation and acidification; emission rates of pollutants and wastes were also deemed as environmental indicators. Single objective optimization was performed on two applications: vinyl chloride monomer plant and on an atmospheric crude oil distillation column. MOO was employed in nine applications: production of methyl chloride, allyl chloride, methyl ethyl ketone (MEK), benzene from toluene, propylene glycol, acrylic acid, and maleic anhydride, scheduling of refinery process and supply chain of vinyl chloride monomer and ethylene glycol. Different tools were used for MOO. Dantus and High (1999) used the compromise programming approach and the stochastic annealing algorithm hand-inhand. Another method is the summation of weighted objective functions (SWOF) utilized by Lim et al. (1999) and Song et al. (2002). On top of the weighted method, Lim et al. (1999) used other MOO tools: goal programming (GP) and parameter space investigation (PSI). A popular and more reliable option over SWOF is the epsilon-constraint method, used by Fu et al. (2001) and Hugo and Pistikopoulos (2005). Chen and Shonnard (2004) used analytic hierarchy process (AHP) to obtain the weights for the objective functions, and then genetic algorithm was used for optimization. Finally, normal boundary intersection (NBI) method was employed by Kheawhom and Hirao (2002). 59 Stearyl-3-(4hydroxy-3,5-di-tertbutylphenyl) propanoate production (Koller et al., 1998) Methyl chloride production (Dantus and High, 1999) 4 5 3 Acetaldehyde production (Stefanis et al., 1996) 4-(2-methoxyethyl) phenol (Koller et al., 1998) Application (Reference) Vinyl chloride monomer plant (Stefanis et al., 1995) 2 1 o. Economic: Annual equivalent profit including costs associated with the process and wastes; Environment: Environmental impact index which comprise the chemicals’ toxicities and their release potential. Economic: Cost index (CI) for input streams, waste treatment costs and equipment costs; Environment: Mass loss index (MLI) and environmental index (EI). The latter is analyzed for both inputs and outputs. For detailed calculation steps, refer to Hienzle et al. (1998) Economic: CI for input streams, waste treatment costs and equipment costs; Environment: MLI and EI. The latter is analyzed for both inputs and outputs. For detailed calculation steps, refer to Hienzle et al. (1998) Economic: Annual operating cost; Environment: Critical air mass (CTAM) based on LC50. Economic: Operating cost; Environment: Six indices including critical air and water emissions, solid disposals, photochemical oxidation, global warming and stratospheric ozone depletion. Metrics Used Manufactured via the thermal chlorination of methane. MOO was carried out by combining the compromise programming approach and the stochastic annealing algorithm. The production of stearyl-3-(4-hydroxy-3,5-ditert-butylphenyl) propanoate was separated into two sections; one is the production to the intermediate and the second section is from the intermediate to the product. 60 Three chemical synthesizing routes are considered: (1) from p-nitrotoluene, (2) from styrene, and (3) from o-chlorophenol. Single objective optimization was carried out for each index. Two different systems were considered: (1) conventional VCM process consisting only the plant, and (2) global VCM production system which considers the cradle-togate scenario. Between methanol and butanal, the former is more environmentally benign. Thereafter, three different blends were considered. Remarks/Comments Table 4.1: Economic and Environmental Criteria – Petroleum Refinery and Petrochemicals Chapter 4: Process Applications Studied Allyl chloride production (Young et al., 2000) Methyl ethyl ketone (MEK) production (Lim et al., 2001) Hydrodealkylation of toluene to benzene (Fu et al., 2001) Hydrodealkylation of toluene to benzene (Smith et al., 2004) 6b 7 8a 8b 6a Application (Reference) Allyl chloride production (Lim et al., 1999) o. Economic: Revenue of allylchloride product; Environment: Potential environmental impact index which comprises human, terrestrial and aquatic toxicity. Economic: Revenues, operating costs, net profit and capital costs; Environment: Using WAR algorithm consisting of human toxicity by inhalation and exposure as well as ingestion, terrestrial toxicity, aquatic toxicity, global warming, photochemical oxidation, ozone depletion and acidification. Economic: Operational costs minus product revenue, calculated as per kmol of SBA (sec-butyl alcohol or 2-butanol); Environment: Environmental impact per kmol of SBA, calculated based on Mallick et al. (1996) and Cabezas et al. (1999). Economic: Annual profit; Environment: Six impact categories from WAR were considered except acidification and ozone depletion potential (which are zero for all components in this process). Economic: Economic Potential (EP) is the annualized profits minus costs of chemicals and equipment, including hazardous waste treatment and fugitive emission costs; Environment: PEI calculated via WAR algorithm. Metrics Used Epsilon-constraint method was used for MOO, where annual profit was kept as the objective function. Two scenarios were considered: (a) diphenyl as a pollutant, and (b) diphenyl as a byproduct. Two configurations were considered: (1) diphenyl removed, and (2) diphenyl recycled. EP and PEI were computed for conversion values from 0 to 1. Fugitive emissions were assumed to be 0.1% of each stream. Normal boundary intersection method was employed for MOO. Three cases were considered. The first is the base case, the second case tackles the high energy consumption issue and the last case attempts to improve the yield of allyl chloride. MOO was performed using summation of weighted objective functions, goal programming and parameter space investigation. Remarks/Comments Table 4.1: Economic and Environmental Criteria – Petroleum Refinery and Petrochemicals 61 Chapter 4: Process Applications Studied Propylene glycol production process (Kheawhom and Hirao, 2002) Acrylic acid production (Hugo et al., 2004) Styrene production (Smith, 2004) Maleic Anhydride production from nbutane (Chen and Shannord, 2004) 11 12 13 Weights for each impact category are assumed as unity. The resulting single-objective optimization problem was solving using CPLEX in GAMS. Remarks/Comments 62 Propylene glycol is produced by hydration of propylene oxide. MOO was performed using normal boundary intersection method. Robustness of the process was measured by calculating failure probability and deviation ratio. Epsilon-constraint method was used for MOO. Two solvents, di-isopropyl ether and n-propyl acetate, were analyzed individually, and the former has a better performance. Economic: EP calculated as the annualized savings Feed is toluene. Hierarchical design is proposed to achieved; Environment: (PEI) calculated using the generate waste-recycle feeds. WAR algorithm (Young and Cabezas, 1999) Economic: Net present value, uniform annual Found that n-butane as raw material is superior to worth, conversion and yield; Environment: A benzene. Analytic hierarchy process was used to composite environment index and nine individual aggregate both economic and environment scores. impact categories indices using Environmental Optimization for single objective (economic, Fate and Risk Assessment Tool assessment environment or aggregated) was performed using framework. Valuation is done based on Ecogenetic algorithm. indicator 95. Economic: Summation of capital and operating costs and subtraction of product revenues; Environment: Sustainable Process Index calculated by the amount of wastewater discharged and steam consumed. Economic: Total annualized cost; Environment: Eco-indicator 99 score. Application Metrics Used (Reference) Refinery processes – Economic: Total Profit; Environment: Global Scheduling problem Impact Score calculated using Critical Surface(Song et al., 2002) Time 95 (CST95) methodology. 10 9 o. Table 4.1: Economic and Environmental Criteria – Petroleum Refinery and Petrochemicals Chapter 4: Process Applications Studied Catalytic reforming process (Janjira et al., 2007) 17 16 15 14 Application (Reference) Supply chain of vinyl chloride monomer and ethylene glycol (Hugo and Pistikopoulos, 2005) Atmospheric crude oil distillation column (Gadalla et al., 2005) Hydroformylation process, also known as aldehyde production (Fang et al., 2007) o. Economic: Profit; Environment: Environment Impact (EI) which is calculated based on the production rates of benzene and CO2. Economic: TCI and Total Production Cost; Environment: E-factor which is the mass of waste generated per mass of product. EFRAT tool was employed to analyze the environmental impacts via process composite environment index (IPC). Economic: Capital investment and PBP; Environment: CO2 emissions Economic: Net present value; Environment: EcoIndicator 99, carcinogenic plant emissions and network resource depletion Metrics Used 63 Since Eco-indicator 99 has inherent subjectivity, weighting factors are neglected and the more significant components (i.e. carcinogenic plant emissions and network resource depletion) are optimized separately. MOO using epsilonconstrain method was implemented. Crude oil processed is of Tia Juana Light (Venezuela) assay. Two hot utilities, flue gases from a fuel oil-fuelled fired heater and highpressure steam, are required. Single objective optimization was performed. Two processes were considered: (1) conventional method (based on Exxon hydroformylation), (2) catalytic olefin hydroformylation in CO2-expanded liquid (CXL) media. The former has aldehydes and alcohols as products, while the latter produces linear and branched aldehydes. For EI, impact inflicted by benzene was 3.5 times larger than CO2 as it has carcinogenic effects. Both economic and environmental metrics were calculated for different design specifications and plant capacity. Uncertainties were incorporated into the analysis. Remarks/Comments Table 4.1: Economic and Environmental Criteria – Petroleum Refinery and Petrochemicals Chapter 4: Process Applications Studied Chapter 4: Process Applications Studied 4.2.2. Biotechnology, Pharmaceuticals and Chemicals There are a total of 18 applications in this category (Table 4.2). Only one biotechnology application in the production of food was explored: the cottage cheese production chain (Stefanis et al., 1997a). Another four studied the production of chemicals; chemicals are compounds that are made of inorganic raw materials. They include formation of boron products (Azapagic and Clift, 1999), hydrogen cyanide (Hoffmann et al., 2001 and 2004), urea (Khan et al., 2002) and chlorine (Martins et al., 2007). The remaining applications are mainly related to the production of specific chemicals using biotechnology. These processes utilize enzymes or bacteria to produce the desired products – penicillin, sodium pyruvate, citric acid, pyruvic acid, L-lysine, etc. In most of the applications, the economic and environmental indicators were used to measure the performance of the processes. Optimization was implemented for six applications only; they are: cottage cheese production chain, penicillin production (Steffens et al., 1999; Heinzle et al., 2006) and multiproduct batch plants (Dietz et al., 2005; Dietz et al., 2006; Dietz et al., 2007a/b). MOO was carried out using epsilon-constraint method and multi-criteria decisionmaking. Most of the bioprocesses in Table 4.2 were obtained from the book by Heinzle et al. (2006) which focuses on the topic of sustainability and the assessment of bioprocesses using sustainability criteria. The criteria used are both economic and environmental. Economic criteria that were used are unit production costs (UPC), yield, total capital investment (TCI), annual operating costs, return on investment (ROI), PBP, NPV and internal rate of return (IRR). Environmental criteria comprised 6 impact groups which are in turn made up of 15 impact categories – raw material availability, land use, complexity 64 Chapter 4: Process Applications Studied of synthesis, thermal risks, acute toxicity, chronic toxicity, ecotoxicity, GWP, ODP, AP, PCOP, odor, EP and organic carbon pollution potential. 4.2.3. Downstream Processing While reaction processing is the heart of any chemical process, downstream processing is by no means playing a second fiddle to it. The purification of products or waste streams plays an important role in either meeting customers’ demands or governmental regulations respectively. Henceforth, 10 applications related to downstream processing were studied in the context of sustainability (Table 4.3). Economic criteria include profits, capital costs, annual operating costs, NPW and uniform annual worth (UAW). Environmental criteria would consist of toxicity (based on lethal-dosage or lethalconcentration), energy and material intensity. Many of the applications that employed optimization used multiple objectives. MOO tools employed are epsilon-constraint method, goal programming, AHP, normal-boundary intersection (NBI) method, nondominated sorting genetic algorithm (NSGA-II), parallel multi-objective steady-state genetic algorithm (pMSGA) and multi-criteria decision analysis (MCDA). 65 3 2b 2a 1 o. Metrics Used Penicillin production Economic: Unit production costs; Environment: (Heinzle et al., Material intensity and environmental index 2006) (consists of 15 impact categories in 6 impact groups). Boron products Economic: Production rate and life cycle operating (Azapagic and Clift, costs; Environment: Burdens which are given by 1999) emission rates and environment impacts (Heijungs et al., 1992) Economic: Total Annual Cost (TAC); Environment: Biological Oxygen Demand (BOD) in wastewater stream and the six indices - critical air mass, critical water mass, solid mass disposal, global warming impact, photochemical oxidation impact and stratospheric ozone depletion impact. The six indices were aggregated to a global environmental index (GEI) Penicillin production Economic: Annual costs; Environment: Critical (Steffens et al., water mass and sustainable process index. 1999) Application (Reference) Cottage cheese production chain (Stefanis et al., 1997a) Single and multi-objective optimization were performed. MOO was carried out using the epsilon-constraint method. Two boron minerals and five boron products were considered. 66 As the same quantity of biomass is produced, solid mass disposal is neglected. Multi-criteria synthesis procedure, carried out using Jacaranda system, was adopted to generate flowsheets that are environmentally and economically attractive. Penicillin in this study is specified as Penicillin V. Monte Carlo simulations were employed to examine the effects of parameter uncertainties. Single objective optimization was performed for TAC and BOD. A global batch plant was also provided, back-tracking the production plant to the raw materials required. Epsilon-constraint method was used to optimize TAC while the GEI was managed as a constraint. Different configurations were explored. Remarks/Comments Table 4.2: Economic and Environmental Criteria – Biotechnology, Pharmaceuticals and Chemicals Chapter 4: Process Applications Studied Urea manufacturing process (Khan et al., 2002) Sodium pyruvate formation (Biwer et al., 2005) Multi-product batch plants (Dietz et al., 2005, 2006, 2007a and 2007b) 6 7 Application (Reference) Hydrogen cyanide (HCN) production (Hoffmann et al., 2001 and 2004) 5 4 o. Economic: Working capital, operation and maintenance, and capital investment; Environment: Performance is dependent on ten impact categories. Thereafter, the pollution balance in WAR was incorporated. Economic: Energy consumption, unit contribution margin, sales return and return on investment; Environment: Material intensity, carbon oxygen demand, energy consumption Economic: Investment cost for equipment and storage vessels; Environment: Amount of biomass released and the volume of solvent used, both measured in terms of per unit product. Economic: Total Annualized Profit per Service Unit (TAPPS); Environment (Hoffmann et al., 2001): Material intensity per service (MIPS); Environment (Hoffmann et al., 2004): Ecoindicator 99 (EI99) which is a damage-oriented method for assessing adverse environmental effects on human health, ecosystems, and natural resources in Europe. Metrics Used MOO was carried out using multi-objective genetic algorithm (MOGA) with a Pareto-optimal ranking method. 67 For product purification, electrodialysis is superior to extraction. Hoffmann et al. (2001): 1250 alternatives, based on hierarchical approaches, were considered and their TAPPS and MIPS were evaluated. BMA performed better than Andrussow processes; the former produces HCN through an endothermic reaction while the latter employs an exothermic reaction. Hoffmann et al. (2004): Epsilon-constraint method was used for MOO. As there are three unit operations for removing unreacted ammonia and three different uses of hydrogen, nine alternatives were considered. Cradle-to-gate approach was employed. GreenPro-I comprises two steps: (a) risk-based life cycle assessment, and (b) risk-based multicriteria decision-making. Remarks/Comments Table 4.2: Economic and Environmental Criteria – Biotechnology, Pharmaceuticals and Chemicals Chapter 4: Process Applications Studied 12 11 10 9 8 o. Application (Reference) Citric acid from starch (Heinzle et al., 2006) Pyruvic acid from glucose using E. coli (Heinzle et al., 2006) L-lysine from glucose (Knoll and Buechs, 2006) Riboflavin production using E. ashbyii (Storhas and Metz, 2006) α-Cyclodextrin production (Heinzle et al., 2006) Economic: TCI, annual operating costs and UnitProduction Costs; Environment: Material intensity and environmental index (consists of 15 impact categories in 6 impact groups). Economic: TC, annual operating costs; Environmental: Material intensity and environmental index (consists of 15 impact categories in 6 impact groups). Economic: TCI, annual operating costs and ROI; Environment: Material intensity and environmental index (consists of 15 impact categories in 6 impact groups). Economic: Unit Production Cost and yield; Environment: Material intensity and environmental index (consists of 15 impact categories in 6 impact groups). Economic: TCI, payback period and annual operating costs; Environment: Material intensity and environmental index (consists of 15 impact categories in 6 impact groups). Metrics Used Both solvent and non-solvent processes were analyzed. As these processes have low potential environmental impact, and with the consideration for uncertainty, neither process supersede the other. Riboflavin is also known as Vitamin B2 or lactoflavin. Environment assessment was only performed on the outputs of the process. Process using electrodialysis for separation is environmentally and economically superior to a process with two extraction steps. Instead of molasses, starch was used to accommodate the downstream purification process. Remarks/Comments Table 4.2: Economic and Environmental Criteria – Biotechnology, Pharmaceuticals and Chemicals 68 Chapter 4: Process Applications Studied 16 15 14 13 o. Monoclonal Antibodies (Mabs) production (Heinzle et al., 2006) α-1-Antitrypsin (AAT) from Transgenic Plant Cell Suspension Cultures (Zapalac and McDonald, 2006) Application (Reference) Recombinant Human Serum Albumin (HSA) production (Kholiq and Heinzle, 2006) Recombinant human insulin production (Petrides, 2006) Economic: Unit-Production Cost, Internal Rate of Return (IRR) and NPV; Environment: Material intensity and environmental index (consists of 15 impact categories in 6 impact groups). Economic: TCI, annual operating costs, UnitProduction Cost, ROI and PBP; Environment: Material intensity and environmental index (consists of 15 impact categories in 6 impact groups). Economic: TCI, annual operating costs and UnitProduction Cost; Environment: Material intensity and environmental index (consists of 15 impact categories in 6 impact groups). Economic: TCI, Unit-Production Cost (UPC), ROI and PBP; Environment: Material intensity and environmental index (consists of 15 impact categories in 6 impact groups). Metrics Used The model studied is on the recovery and purification of rAAT from transgenic rice-cell suspension cultures using chromatography separation and diafiltration steps. Monte Carlo simulations were employed to perform parameter uncertainty analysis on the system. Human insulin is produced from recombinant E. coli. Expanded-bed adsorption (EBA) is preferred to packed-bed absorption (PBA) due to its better economic performance; however, EBA has a poorer ecological performance. Remarks/Comments Table 4.2: Economic and Environmental Criteria – Biotechnology, Pharmaceuticals and Chemicals 69 Chapter 4: Process Applications Studied Chlorine production (Martins et al., 2007) 18 17 Application (Reference) Plasmid DNA production (Freitas et al., 2006) o. Economic: TCI, Unit-Production Cost (UPC), Internal Rate of Return (IRR) and NPV; Environmental: Material intensity and environmental index (consists of 15 impact categories in 6 impact groups). Four sustainability metrics: (1) energy intensity, (2) materials intensity, (3) potential chemical risk, and (4) potential environmental impact. Last two metrics are calculated based on Vincent et al. (2005) which relies on the R-phrases of the chemicals given in their MSDS. Metrics Used Three process alternatives were assessed: (1) mercury cells, (2) diaphragm cells, and (3) membrane cells. Membrane cells had the lowest chemical risk and environmental impact while mercury cells had the lowest energy intensity. All of them had the same material intensity. 70 Three scenarios were considered: (1) base case, (2) base case with isopropyl alcohol recycle, and (3) base case with isopropyl alcohol free process. Remarks/Comments Table 4.2: Economic and Environmental Criteria – Biotechnology, Pharmaceuticals and Chemicals Chapter 4: Process Applications Studied 4a 3 2 1 o. Recovery of benzene and ethylene dichloride from mixture with acetone and toluene (Chakraborty and Linninger, 2002 and 2003) VOC recovery (Shonnard and Hiew, 2000) Application (Reference) Gas treatment system of an acrylic fiber plant (Buxton et al., 1999) Acetic acid separation from water (Kim and Diwekar, 2002) Economic: PBP; Environment: Process composite environment index (IPC) and nine individual impact categories indices using Environmental Fate and Risk Assessment Tool (EFRAT) assessment framework and the WAR algorithm. Economic: Variable operating expenditures; Environment: Global pollution index which includes the impact of output streams, fugitive emissions and raw materials, and global environmental impact vector. Economic: Acetic acid recovery and process flexibility; Environment: Environmental impact based on lethal-dosage or lethal-concentration, LD50 and LC50 respectively. Economic: Annual operating cost; Environment: Critical air mass (CTAM) based on LC50. Metrics Used 71 Removal of ethyl acetate and toluene from waste gas stream. Six alternatives were considered – adsorption or absorption, and different methods of adsorbent or absorbent regeneration. It was shown that adsorption had superior environmental performance in most indicators except smog formation. It was briefly mentioned that adsorption processes had better economic performance with low payback periods of 1 year. All blends considered were more environmentally friendly when compared to the base case (decanol); however, only 3 blends were economically superior. AspenPlus was employed to simulate the process. The environmentally benign solvent is chosen using the Hammersley stochastic annealing algorithm. A constraint multi-objective problem algorithm, similar to epsilon-constraint method, was used. Goal programming was used to obtain the Paretooptimal solutions of the MOO problem. Remarks/Comments Table 4.3: Economic and Environmental Criteria – Downstream Processing Chapter 4: Process Applications Studied VOC recovery (Chen et al., 2002b and 2003) Toluene recovery process (Kheawhom and Hirao, 2002) Toluene recovery process (Kheawhom and Hirao, 2004) 5a 5b Application (Reference) VOC recovery (Chen et al., 2002a) 4c 4b o. Economic: Sum of a portion of total fixed costs and the annual operation costs; Environment: Based on SPI, environmental performance index is measured by summing a portion of the total environmental performance associated with design variables and the yearly environmental performance related to control variables. Economic: Summation of capital and operating costs and subtraction of product revenues; Environment: Sustainability Process Index (SPI) Economic: Venture Profit (VP) which includes the sales revenue of ethyl acetate and toluene, capital cost of the absorber and distillation column, utility costs and taxes; Environment: Inhalation toxicity index (IINH) using EFRAT methodology. Economic: Net present value (NPV) and uniform annual worth (UAW); Environment: Process composite environment index (IPC) and nine individual impact categories indices using Environmental Fate and Risk Assessment Tool assessment framework and the WAR algorithm. Metrics Used 72 Process used is absorption. Absorbent is heavy oil (1-decanol). Sensitivity analyses of VP and IINH are performed. Uncertainties in physical and chemical properties of substances are incorporated into the analysis. Toluene and ethyl acetate are retrieved as products. Analytic hierarchy process was used to aggregate both economic and environment scores. Chen et al. (2002b) performed only sensitivity analysis whereas Chen et al. (2003) used genetic algorithm to find the global optimum of the single objective optimization. The study is based on a closed-loop process. Three recovery processes (membrane-based, condensation-based and adsorption-based) were considered. Robustness of the process was measured by calculating failure probability and deviation ratio (DR). Normal-boundary intersection method was used for MOO. Four membrane-based closed-loop configurations were considered: single-stage, two-stage enriching cascade, two-stage stripping cascade, and twostage enriching cascade with pre-membrane stage. Non-dominated sorting genetic algorithm (NSGAII) was employed for MOO. Hammersley sequence sampling was used for uncertainty analysis. Remarks/Comments Table 4.3: Economic and Environmental Criteria – Downstream Processing Chapter 4: Process Applications Studied 9 8 7 6 o. Distillation unit to separate acetone from water, acetone, methanol and acetic acid mixture (Ramzan and Witt, 2006) Activated sludge plant (Flores et al., 2007) Application (Reference) Acetic acid recovery from aqueous waste mixtures (Kim and Smith, 2004 and 2005) Propylene-propane splitter (Gadalla et al., 2005) Economic: Construction and operating costs; Environment: Impact on water. The period in which the total carbon oxygen demand, biological oxygen demand, total suspended solids, nitrogen and/or phosphorus exceeds the standards set by European directive. Economic: Total annualized cost; Environment: Potential environmental impact using the WAR algorithm. Economic: Net profit, capital costs and annual costs; Environment: CO2 emissions Economic: Total profit; Environment: Potential environmental impacts calculated via the WAR algorithm (Young and Cabezas, 1999). Metrics Used The purpose of the sludge plant is to achieve simultaneous carbon, nitrogen, and phosphorus removal. Design options are evaluated combining the hierarchical decision process with multicriteria decision analysis techniques. 73 MOO was carried out using a parallel multiobjective steady-state genetic algorithm. Kim and Smith (2005) reduced eight impact categories to three and performed a four-objective optimization with profit. The mixture of propylene and propane is equimolar. Other alternatives were considered (i.e. turbine using fuel oil or natural gas). The economic and environmental metrics were calculated for a range of reboiler duties. Single objective optimization was performed to reduce emissions and save energy; these results were provided in tables. Goal programming was used for MOO. Six alternatives with different number of stages and the availability of side stream, were studied. Weighting factors based on Eco-indicator 99 were employed. Remarks/Comments Table 4.3: Economic and Environmental Criteria – Downstream Processing Chapter 4: Process Applications Studied 10 o. Application (Reference) Separation of acetone and chloroform mixture (Martins et al., 2007) Four sustainability metrics: (1) energy intensity, (2) materials intensity, (3) potential chemical risk, and (4) potential environmental impact. Last two metrics are calculated based on Vincent et al. (2005) which relies on the R-phrases of the chemicals given in their MSDS. Metrics Used Two process alternatives based on different solvents (benzene or methyl-n-pentyl-ether) were evaluated. Methyl-n-pentyl-ether dominates benzene for all metrics. Remarks/Comments Table 4.3: Economic and Environmental Criteria – Downstream Processing 74 Chapter 4: Process Applications Studied Chapter 4: Process Applications Studied 4.2.4. Energy Systems and Heat Integration There two applications in this category (Table 4.4). The first involves the analysis of energy systems (Afgan et al., 2000 and 2007) whereby the performance of different energy systems (e.g. wind power plants, biomass power plants and combined cycle gas turbine power plant) were compared to one another. Heat exchanger network is the second application whereby the optimal minimum temperature difference (∆Tmin) was found. The optimum was determined by the weighted sum of both economic and environmental indices using AHP. 75 Energy systems (Afgan et al., 2007) 1b 1a Application (Reference) Energy systems (Afgan et al., 2000) o. Resource Indicator (RI): Fuel RI, Carbon steel RI, Copper RI, Aluminium RI ; Environment Indicator (EI): Carbon dioxide EI, Nitrogen oxide EI, Sulfur dioxide EI, Waste EI; Social Indicator (SI): New job indicator, Capital indicator, Diversity and vitality indicator; Economic Indicators (EcI): Efficiency EcI, Capital investment indicator, Community EcI Resource Indicator (RI): RI above, Stainless steel RI, Insulation RI; Environment Indicator (EI): EI above except Waste EI; Economic Indicator (EcI): EcI above except Community EcI; addition is Energy costs indicator; Social Indicator (SI): SI above except Capital Indicator. Metrics Used 76 Eight different energy systems were considered. They are: (1) Reconstruction of pulverized coal fired unit in condensing regime; (2) Reconstruction of coal fired unit in co-generation regime; (3) Fluidized bed combustion unit – New power station; (4) Combined cycle gas turbine power plant – New power station; (5) Reconstruction of big hydropower plant; (6) Power plants on solar energy (PV systems) – New power station; (7) Wind turbines power – New power units; (8) Biomass power plants – New power station. Four alternatives were considered: (1) solar PV unit, (2) wind power plant, (3) biomass power plant, and (4) thermal power plant. Remarks/Comments Table 4.4: Economic and Environmental Criteria – Energy Systems and Heat Integration Chapter 4: Process Applications Studied 2 o. Application (Reference) Heat exchanger network (Chen et al., 2002b; Wen and Shonnard, 2003) Economic: Annualized cost, inclusive of equipment and purchase costs and operating costs Environment: Process composite environment index (IPC) and seven individual impact categories indices. Life cycle assessment only done for cradle-to-gate. Pre-manufacturing impacts were calculated via the economic input/output life-cycle assessment method (EIOLCA® 2001). Impacts for process employ US EPA equation (Clearinghouse for inventories and emission factors, Air CHIEF). Metrics Used 77 Economic and environmental indices were combined into a single objective using AHP. The indices were plotted with different values of ∆Tmin. In Chen et al. (2002b), only one case study is explored. On the other hand, in Wen and Shonnard (2003), three cases were studied where supply and target temperatures of streams varied. Remarks/Comments Table 4.4: Economic and Environmental Criteria – Energy Systems and Heat Integration Chapter 4: Process Applications Studied Chapter 4: Process Applications Studied 4.3. Environmental Criteria The measure of chemical processes using environmental performance indicators is relatively recent. The number of applications employing only environmental studies is 19 applications thus far. Since environmental performance is one of the sustainability spheres, these applications are summarized below along with the type of environmental performance employed and the use of optimization tools, where applicable. Environmental performance could be measured by the amount of wastes produced, or via tools like WAR and EFRAT which provide the degree of impact on environmental categories such as global warming and human toxicity. 4.3.1. Petrochemicals Unlike the section of ‘Petroleum Refining and Petrochemicals’ for both economic and environmental criteria, no study related to only environmental criteria in the area of petroleum refining was found in the literature. Hence, this section has been reduced to applications in the category of petrochemicals. There are a total of 11 applications (Table 4.5), which include the production of methyl ethyl ketone (MEK), chloromethane, acetaldehyde, vinyl chloride, benzene from toluene, n-butylacetate, acrylic acid, ethanol, allyl chloride, acetic acid and methyl methacrylate (MMA). From Table 4.5, it can be observed that environmental metrics used evolved from measuring the amount of emissions to measuring the magnitude of impact in each environmental category (e.g. global warming). Other environmental metrics include resource conservation and energy consumption (Jia et al., 2004). Three of the listed applications were optimized for multiple objectives using ε-constraint method or AHP. 78 Chapter 4: Process Applications Studied 4.3.2. Biotechnology, Pharmaceutical and Chemicals Only a handful of applications are classified in this category; they include the production of ethanol from sugar beets, ammonia, 6-aminopenicillanic acid (6-APA) from penicillin G and biodiesel (Table 4.6). The environmental indicators used here consist of SPI, emission rates, material intensity, and a weighted sum of the scores of 15 impact categories. For biotechnology, in the case of the production of 6-APA (Biwer and Heinzle, 2004; Heinzle et al., 2006) and biodiesel (Niederl-Schmidinger and Narodoslawsky, 2008), both studies concluded that the products from bioprocess are more environmental friendly than those from chemical or petrochemical routes. 4.3.3. Downstream Processing As mentioned in Section 2.3, downstream processing is a very important part of either a chemical or biochemical process. Downstream processes mainly refer to treatment, recovery or purification units. A total of three applications were found and documented in Table 4.7. Different environmental tools were employed: EFRAT, SPI and WAR algorithm. Only Ramzan et al. (2007) performed MOO using AHP. 79 Vinyl chloride monomer plant (Khan et al., 2001) 4 3 Chloromethane production process (Stefanis et al., 1997b) Acetaldehyde production (Buxton et al., 1999) Application (Reference) Methyl ethyl ketone (MEK) production (Mallick et al., 1996; Cabezas et al., 1997 and 1999) 2 1 o. Global environmental impact (GEI) based on ten impact categories. Thereafter, the pollution balance in WAR algorithm was incorporated. 80 Methanol, with lowest CTAM, was chosen as the solvent. Ethanol dehydrogenation and ethylene oxidation are environmentally benign for gate-togate and cradle-to-gate analyses respectively. GreenPro comprised four steps. Steps 1 and 2 incorporate LCA in process design. Steps 3 and 4 uses multi-objective optimization (via ε-constraint method) and multi-criteria decision-making to find solutions to the design problem. Identifies both routine and nonroutine emissions and waste generations. Cabezas et al. (1997 and 1999) uses different impact categories in the computation of environmental impact. These categories are based on Heijungs et al. (1992). Critical air mass (CTAM) and global warming index (GWI) Critical air mass (CTAM) Process is simulated using Chemcad III. Secondary butyl alcohol is used as the feed. Modifications are made to the base case flow sheet by introducing a recycle stream. Remarks/Comments Mallick et al. (1996) uses three different methods to compute environmental impact: (1) no ranking system, (2) simple ranking system, and (3) modification of Total Hazard Value. Metrics Used Table 4.5: Environmental Criteria – Petrochemicals Chapter 4: Process Applications Studied Ethanol production (Jia et al., 2004) Allyl chloride production (Chen and Feng, 2005) 9 Application (Reference) Hydrodealkylation of toluene to benzene (Halim and Srinivasan, 2002) n-butylacetate production (Cardona et al., 2004) Acrylic acid production (Young and Cabezas, 1999) 8 7 6 5 o. Rate of generation environmental impacts from non-products and PEI which is calculated using the WAR algorithm (Young and Cabezas, 1999; Cabezas et al., 1999) Using the WAR algorithm consisting of human toxicity by inhalation and exposure as well as ingestion, terrestrial toxicity, aquatic toxicity, global warming, photochemical oxidation, ozone depletion and acidification. Process Environmental Performance Assessment was employed. PEI associated with releases, resource conservation and energy consumption. Impact categories in PEI are the same as those found in WAR with an addition of eutrophication potential. Potential Environmental Impact generated and in non-product streams; both are calculated in absolute terms and are normalized with the mass of product formed. All are calculated using the WAR algorithm. PEI is reported together with the individual impact categories. This is calculated via the WAR algorithm. Metrics Used Remarks/Comments 81 Multi-criteria decision making using the analytic hierarchy process is employed to aggregate as an integrated environmental index (IEI). Two processes were considered: (1) ethylene-derived feedstock process, and (2) straw cellulose-derived feedstock process. The latter is more environmentally benign. Graphs were plotted for the rate of increase of PEI per amount of propylene consumed at different concentrations of propylene for different reactor temperatures; reactor is CSTR). The impact factors are calculated for the base case and they are recomputed when the process flowsheet is modified gradually to implement waste minimization analysis. ENVOPExpert was employed for waste minimization analysis. Esterification of n-butanol with acetic acid gives n-butylacetate. The conventional process (i.e. reactor and separators) is compared with a reactive distillation column. Three cases were considered, differences lie in the reactor and separating column conditions. Multiobjective optimization was performed using the εconstraint method. Table 4.5: Environmental Criteria – Petrochemicals Chapter 4: Process Applications Studied 11 10 o. Methyl methacrylate (MMA) process (Zhang et al., 2008) Application (Reference) Acetic acid production (Hossain et al., 2007) Green Degree comprising of global warming potential, ozone depleting potential, photochemical ozone creation potential, acidification potential, eutrophication potential, ecotoxicity potential to air and water, human carcinogenic and noncarcinogenic potential to water. Three impacts are considered: (1) Human health, (2) Ecological health, and (3) Climatic change using the E-Green methodology. An overall indicator is computed for the cradle-to gate analysis. Metrics Used Remarks/Comments 82 The first three are applied to the gate-to-gate analysis while only climatic impact and the overall score is used for the cradle-to-gate approach. Cradle-to-gate approach is only concerned with the production of solvents from the raw form. Two different solvents (alternatives) were analyzed: (1) ethyl acetate and (2) isopropyl acetate. Four different process schemes for the production of MMA were evaluated and compared. In addition, green degree is computed for different sections of the MMA production (via isobutene) flowsheet. Table 4.5: Environmental Criteria – Petrochemicals Chapter 4: Process Applications Studied 4 3 2 1 o. Cleavage of penicillin G to 6-aninopenicillanic acid (6-APA) (Biwer and Heinzle, 2004; Heinzle et al., 2006) Biodiesel from used vegetable oil (UVO) and tallow methyl ester (TME) (Niederl-Schmidinger and Narodoslawsky, 2008) Application (Reference) Ethanol production from sugar beets (Krotscheck and Narodoslawsky, 1996) Ammonia production (Mallick et al., 1996; Cabezas et al., 1999) Process is simulated using Chemcad III. Modifications are made to the base case flow sheet by changing the purge ratio. The concept of SPI is based on the sustainable flow of solar exergy, which is calculated in terms of area. Remarks/Comments SPI measured in area equivalent per megajoule of biodiesel produced per annum. 83 SPI was calculated for biodiesel manufactured from both raw materials. Life cycle assessment was performed. It was concluded that biodiesel, in general, is superior to fossil fuels. Cabezas et al. (1999) uses different impact categories in the computation of environmental impact. These categories are based on Heijungs et al. (1992). Material intensity and environmental index (consists of Environmental performance of biocatalytic 15 impact categories in 6 impact groups). route is better than the chemical route. Mallick et al. (1996) uses three different methods to compute environmental impact: (1) no ranking system, (2) simple ranking system, and (3) modification of Total Hazard Value. Sustainable Process Index Metrics Used Table 4.6: Environmental Criteria – Biotechnology, Pharmaceutical and Chemicals Chapter 4: Process Applications Studied Chapter 4: Process Applications Studied 4.3.4. Energy Systems There is only one application in this category, and involves a steam and power generation plant (Eliceche et al., 2007). Single objective optimization was carried out for this application using CONOPT++ and OSL in GAMs. An overall environmental impact score was calculated as the weighted sum of different impact categories as given in Heijungs et al. (1992). 4.4. Conclusions This chapter has put forth a comprehensive summary of the process applications studied by researchers on the topic of sustainability. A total of 66 applications have been mentioned. These ranged from applications in the petroleum refining and petrochemical industries to biotechnology, pharmaceutical and chemicals industry. Of the 66, 47 applications considered both economic and environmental criteria while the remaining 19 applications looked solely at the environmental criteria. Of these, 32 applications employed single and multi-objective optimization tools such as CONOPT++ and OSL in GAMS and NSGA-II respectively. In our study, the application that is of interest is recovery systems. The main purpose of recovery systems is to reduce the impact the process has on the environment by recovering valuable or harmful compounds. In view of this, MOO for recovery systems will be discussed in the upcoming chapters. 84 3 2 1 o. Fluoride removal from wastewaters from AlF3 production (Dominguez-Ramos et al., 2007) Distillation unit to separate acetone from a mixture of water, acetone, methanol and acetic acid (Ramzan et al., 2007) Application (Reference) VOC recovery (Lapkin et al., 2004) Environmental Protection Index (EPI); EPI is a combination of the components in WAR algorithm, energy consumption, resource conservation and fugitive emissions. Individual impact categories are computed; three environmental methodologies were employed: Eco-indicator 99, CML 2 Baseline 2000 and IChemE Sustainability Metrics Energy and material intensity, production and process efficiency, ratio of renewable to total energy used, the amount of greenhouse gases in tonnes equivalent released, water consumption and sustainability process index. Metrics Used 85 The values of each component are evaluated for different combinations of a range reflux ratios and a range of steam flowrate. Weighting factors based on AHP were employed, performing multi-criteria decision making. Economic and environmental optimums are presented; however, discussion on the economic criteria was not evident. Removal of ethanol and ethyl acetate from gas stream. Two processes were considered: (1) VOC incineration and (2) monolith adsorption process. Since the above are gate-to-gate analyses, construction of the incinerator and the monoliths should be considered. The former’s impact is deemed negligible and not discussed while the latter is found to emit large amounts of greenhouse gases and is energy intensive. Two process schemes were analyzed: chemical precipitation and crystallization process. Crystallization process showed better environmental performace. Remarks/Comments Table 4.7: Environmental Criteria – Downstream Processing Chapter 4: Process Applications Studied 1 o. Application (Reference) Steam and power generation plant (Eliceche et al., 2007) Overall environmental impact (ψ) which is a weighted sum of the scores of each impact categories based on Heijungs et al. (1992) Metrics Used Remarks/Comments CONOPT ++ and OSL in GAMs were used to solve the nonlinear and mixed-integer programming respectively (single objective optimization). Life cycle impact assessment was carried out to include the generation of imported electricity. Table 4.8: Environmental Criteria – Energy Systems 86 Chapter 4: Process Applications Studied Chapter 5: Optimization of Recovery Processes Chapter 5 Optimization of Recovery Processes 5.1. Introduction As mentioned in Chapter 2, the concept of sustainability consists of three spheres – economic, environmental and societal. Chemical engineers can optimize the design and operation of their processes to make a difference in the economic and environmental impact posed by the processes. However, the engineers most probably do not control on societal aspects for they are dependent on the governmental and company regulations. Thus, in this chapter, two recovery processes have been optimized for both economic and environmental objectives in order to make them sustainable. Recovery processes are chosen as waste streams are often co-produced. These waste streams, together with the energy utilization, contribute to the environmental impact categories which are the objectives used in this study. The feasibility and usefulness of optimizing the recovery processes for a few, several and many objectives are also investigated. 5.2. VOC Recovery In this process, the objective is to recover and recycle two VOCs – namely toluene and ethyl acetate – from a gaseous waste stream. This gaseous waste stream originated from a cellophane production facility but could have come from any number of industrial processes (Shonnard and Hiew, 2000). The waste stream specifications are: nitrogen = 24,799.38 kg/h, toluene = 193.55 kg/h and ethyl acetate = 193.55 kg/h. From Chen et al. (2003), diethylene glycol monobutyl ether is the solvent used in recovering the VOCs 87 Chapter 5: Optimization of Recovery Processes from the gaseous waste stream at the absorber (see Figure 5.1). Thereafter, the rich solvent is preheated before the distillation column where the VOCs are recovered from the top of the column while the lean solvent is recovered at the bottom of the column. The lean solvent, together with some make-up solvent, is then recycled back to the absorber column. The product column is used to separate the recovered VOCs into relatively pure streams of ethyl acetate and toluene which are then sold. Figure 5.1: VOC Recovery Process Flowsheet 5.2.1 Background Shonnard and Hiew (2000) studied the different technologies (i.e. adsorption and absorption) for the removal VOCs from the gas stream. Thereafter, Chen et al. (2001) screened 847 different solvents and highlighted the solvents with superior economic and environmental performance. From their findings, diethylene glycol monobutyl ether, also known as B-Carbitol, was amongst the better ‘performing’ solvents and was thus chosen as the absorbent to be used for this process. Later Chen et al. (2002b) studied the process 88 Chapter 5: Optimization of Recovery Processes for both economic and environmental indices where there were only two decision variables: absorbent flow rate and temperature. The variables were reduced to these two with the use of scaled gradient analysis (SGA). In their analysis, pre-manufacturing life cycle analysis (LCA) was included. This accounted for the emissions produced in the fabrication of the process equipment. Thereafter, Chen et al. (2003) improved the search for the optimal values of decision variables by employing the software called Simultaneous Comparison of Environmental and Non-environmental Process Criteria (SCENE), linked with the process simulator, HYSYS. In SCENE, genetic algorithm was used as the optimizer. Analytic Hierarchy Process (AHP) was employed to aggregate multiple objectives into a single indicator which measures the fitness of the solution. 5.2.2. The Present Study In this study, pre-manufacturing emissions are not considered as the emissions are directly proportional to the dollar value of the equipment, which is already included in economic objectives such as PBT, NPW and PBP. Moreover, the emission data for the manufacturing of the equipment are either too generic (e.g. mass of air pollutants released without the breakdown of the components) or incomplete (e.g. giving the emissions of only a handful and more well-known pollutants such as carbon dioxide and sulphur dioxide). Use of AHP to aggregate the economic (NPW) and environmental (Process Composite Index, IPC) criteria by Chen et al. (2003) involves subjective inputs by the decision maker. In addition, aggregating the individual environmental components into IPC requires weighting, which is again subjective. Moreover, minimizing the weighted 89 Chapter 5: Optimization of Recovery Processes sum of the various factors does not necessarily minimize the contribution of each individual factor. Such problems can be avoided if the economic criteria and environmental components are not aggregated and optimized simultaneously. This application involves the emission of VOCs. One particular characteristic of VOCs is its significant contribution to the formation of smog via photochemical oxidation. Photochemical oxidation is the process where photochemical smog is formed when VOCs react with oxides of nitrogen under the influence of sunlight according to the simplified reaction scheme: VOCs + NOx + Sunlight Photochemical smog. In the discussion of the results below, the dominant influence of VOCs on PCOP would be elucidated. The elitist non-dominated sorting genetic algorithm (NSGA-II) was used as the MOO tool for this study. From the studies by Nandasana et al. (2003), Kasat et al. (2003) and Masuduzzaman and Rangaiah (2008), NSGA and related techniques are noted to be popular among researchers and hence NSGA-II is employed in this study. As the objectives (i.e. economic criteria and environmental components) may be partially or totally conflicting, Pareto-optimal solutions will be obtained. Three cases of two, several and many objectives are considered for both operation and design scenarios. These results are presented and discussed in the following sections. To enable MOO, a custom-made Excel-Visual Basic for Applications-HYSYS interface has been developed (Figure 5.2). This interface combines the process simulation capability of HYSYSTM with the mathematical computation and spreadsheet features of Microsoft Excel by linking the object libraries of these two applications using Visual Basic for Applications (VBA). Excel is first used to initiate the MOO program. NSGA-II, 90 Chapter 5: Optimization of Recovery Processes which has been implemented in VBA, is used to generate the values for the decision variables which are transported from VBA to HYSYSTM. After the convergence of the process flowsheet in HYSYSTM, the results are transported back to Excel via VBA. Within the Excel spreadsheet, the objective functions and constraints are calculated and the NSGA-II in VBA uses these values in the ranking of the solutions. This completes one generation of NSGA-II. The process is repeated for the number of generations specified by the user. EXCEL (User Interface, Objectives, Constraints) Object Library VBA (NSGA-II) Object Library HYSYSTM (Process Simulation) Figure 5.2: Excel-VBA-HYSYS Setup for MOO of Processes 5.2.3. Operation Optimization Although Chen et al. (2003) had performed SGA to reduce the number of decision variables, all possible decision variables are considered in this study for completeness sake. In the previous studies, it is not obvious if the system was simulated for design or for operation. If the system was simulated for design, size of equipment could vary to meet the process requirements. On the other hand, equipment size would limit the extent of cooling, distillation capacities, etc. in operation simulation and optimization. Hence in this study, we will have two sets of optimization, operation and design. In this section, the operation optimization is analyzed first while the design optimization is discussed in Section 5.2.4. For the operation case, the five decision variables are: 91 Chapter 5: Optimization of Recovery Processes absorbent flow rate (Fabs), absorbent temperature (Tabs), waste gas stream temperature to absorber (Tf,abs), rich absorbent temperature to distillation column (Tdist) and product temperature (TVOC). The bounds for the decision variables and constraints are given in Table 5.1, together with the objectives for three different cases considered. Table 5.1: Objectives, Decision Variables and Constraints for VOC Process – Operation Optimization Cases Objectives A Max J1 ≡ PBT Min J2 ≡ PEI B Max J1 ≡ PBT Min J2 ≡ HTP Min J3 ≡ TTP Min J4 ≡ ATP Min J5 ≡ GWP Min J6 ≡ PCOP Min J7 ≡ EP Min J8 ≡ AP C Max J1 ≡ PBT Min J2 ≡ HTP Min J3 ≡ ETP Min J4 ≡ PCOP Min J5 ≡ ATMP Decision Variables Constraints a) 110 ≤ Fabs ≤ 170 kmol/hr a) (∆Tj)i ≥ 5 °C, where j=1, 2 and i=feed, sol, VOC b) 30 ≤ Tabs ≤ 40 °C b) (∆Tj)i ≥ 25 °C, where j=1, c) 30 ≤ Tf,abs ≤ 40 °C 2 and i=hx, dist d) 200 ≤ Tdist ≤ 230 °C c) Ft ≥ 0.75 for all e) 30 ≤ TVOC ≤ 40 °C exchangers d) TCU,i ≤ 45 °C, i=feed, sol, VOC e) Vi ≥ weeping rate, where i=abs, dist, prod f) Vi ≤ flooding rate, where i=abs, dist, prod g) Heat and Material Balances solved in HYSYS 5.2.3.1. Case A: Bi-Objective Optimization Though MOO could handle many objective functions, considering too many objectives simultaneously may confuse the decision maker, who may eventually have too much information to make an informed choice of the optimal conditions for the plant. Since we are interested in both economic and environmental criteria, the number of objectives can be reduced to two – PBT and an aggregated environmental index, PEI (see Table 5.1). The choice of the economic objective is PBT because we are considering the operation 92 Chapter 5: Optimization of Recovery Processes mode whereby the investment cost is fixed and thus the other economic criteria, NPW and PBP, would improve or worsen with PBT. As for environmental categories, they are aggregated into a single performance index through normalization and using weights (which have been employed by Kim and Smith, 2004 and 2005; Ramzan and Witt, 2006; Zhang et al, 2008). The normalization factors are calculated using the arithmetic average of all chemical compounds in the impact category of concern. Equal weighting factors are employed as any deviation from unity would inherently involve some form of biasness. In addition, Young et al. (2000) explained that, if there is no specific site in mind for the process studied, it would be best to use equal weights. From Figure 5.3, it can be seen that PBT improves (increases) from 1.66×105 $/yr to 2.15×105 $/yr as PEI deteriorates (increases) from 1.05 to 1.87; in particular, PEI increases exponentially for PBT above 2.1×105 $/yr (see Figure 5.3a). Only one of the five decision variables (i.e. Fabs) had the largest influence on the objective functions. While TVOC varied slightly from 30°C to 30.3°C, the remaining decision variables took on constant values – Tabs varies slightly from 30.0 to 30.1°C Tf,abs from 31.3 to 31.6°C, TVOC from 30.0 to 30.4°C. Increasing trend of PEI with PBT is due to the decrease in Fabs from 149.6 to 140.6 kmol/hr (Figure 5.3b) and the decrease in Tdist from 215.1 to 212.5°C (Figure 5.3c). With a reduction in Fabs, the recovery of VOCs in the waste gas stream is reduced. This means that the amount of VOCs emitted has increased which would increase the environmental impact the system imposes. With the increase in VOCs emissions, the amount of VOCs recovered for sale as solvent is reduced, the revenues earned declines (Figure 5.3d). Simultaneously, the amount of utilities required in the heating and cooling of the absorbent through the system is reduced, which reduces the 93 Chapter 5: Optimization of Recovery Processes utilities costs incurred. In addition, the reduction in Tdist would indicate that the HP steam required at the distillation feed heater is lower. The reduction in utilities costs would directly reduce the cost of manufacturing, COM, as shown in Figure 5.3e. Although both revenues and cost of manufacturing declined concurrently, the reduction in revenues is offset by the decrease in the manufacturing cost; as a result, there is a net increase in PBT. There will be questions with regards to the aggregation of the different impact categories into a single impact factor. For example, does normalizing the impact factors with the arithmetic average bring them on the same platform for comparison? Also, the weighting factors are inherently subjective and should be avoided. Prior to the study of more than two objective functions, one can first observe the contribution of each impact category to the overall PEI (see Figures 5.3f to 5.3l). PCOP has the largest influence on the overall PEI and also conflicting with PBT. On the other hand, the indices for other impact categories improved (i.e., decreases) while PBT increases. 5.2.3.2. Case B: Optimization for Many Objectives Combining multiple objectives into a single objective function does not provide the decision-maker with information about trade-offs amongst the various objectives, or about alternative operating conditions (Thibault, 2008). Moreover, minimizing the weighted sum of the various environmental factors (i.e. PEI) does not necessarily minimize the contribution of each individual factor. Since the aggregation of impact categories into a single indicator is questionable, it is preferred to consider the environmental categories individually. Therefore, eight objective functions were 94 Chapter 5: Optimization of Recovery Processes (a) Fabs, kmole/h PEI 2 1.5 1 1.6 1.8 2 2.2 1.6 Revenue, $/yr , °C T dist 215 210 1.6 1.8 x 10 2 2.2 2.5 2 2.2 2.4 1.8 2 2.2 2.4 1.8 2 2.2 2.4 1.8 2 2.2 2.4 2 2.2 2.4 2.4 2.3 1.8 2 2.2 2.4 -5 3.5 ATP TTP x 10 2.2 1.6 3.8 3.6 1.8 -7 (f) 2.15 1.6 x 10 2.4 (d) 3 1.6 (e) 4 2.2 3.02 6 2.2 2 3.04 2.4 2.25 x 10 1.8 6 3.06 HTP COM, $/yr 120 (b) 3.08 (c) (g) 3.4 1.6 x 10 -5 3.4 3.3 (h) 1.8 2 2.2 3.2 1.6 2.4 0.08 0.32 0.078 0.31 0.076 AP GWP 140 2.4 220 2.3 160 0.074 0.3 0.29 0.072 (i) 0.07 1.6 (j) 1.8 2 2.2 2.4 1.6 1.5 0.105 0.1 EP PCOP (k) 1 0.095 (l) 0.5 1.6 1.8 2 PBT (105), $/yr 2.2 2.4 0.09 1.6 1.8 PBT (105), $/yr Figure 5.3: Selected Results for Operation Optimization of VOC Recovery for PBT and PEI 95 Chapter 5: Optimization of Recovery Processes considered for the operation optimization. They include PBT, and the different components of environmental impact – HTP, TTP, ATP, GWP, AP, PCOP and EP. Although the environmental components are not aggregated into a single index, they are normalized here for easy comparison with the results in Section 5.2.3.1. Readers should note that the normalization has no impact on the optimization results as each impact category is optimized individually. The results from MOO using NSGA-II are given in Figure 5.4. Figures 5.4a to 5.4g are the objective functions considered for this case study while the decision variables are given in Figures 5.4h to 5.4j. The remaining two decision variables, Tabs and TVOC, are not plotted as they both took on constant values of 30°C. It can be observed from Figure 5.4 that the Pareto-optimal solutions can be divided into two segments. The solutions denoted by circles indicate that as PBT increases from 1.24×105 to 2.35×105 $/yr, HTP (from 2.44×10-7 to 2.17×10-7), TTP (from 4.00×10-5 to 3.36×10-5), ATP (from 3.57×10-5 to 3.07×10-5), GWP (from 0.081 to 0.070), AP (from 0.32 to 0.27) and EP (from 0.104 to 0.089) also decrease. Although the above seven objectives improved simultaneously, the index for PCOP worsened (increased from 0.55 to 2.28). As a result, this set of solutions is part of Pareto-optimal solutions. The other set of solutions, denoted by triangles, shows that as PBT increases from 2.21×105 to 2.31×105 $/yr, PCOP improves as its index decreases from 3.15 to 2.29. However, the other impact categories had worsened. In comparison to Case A, a wider variety of Pareto-optimal solutions were obtained in Case B. 96 Chapter 5: Optimization of Recovery Processes 2.8 x 10 -7 4 TTP 2.4 3.6 3.4 2.2 (a) 2 1.2 (b) 1.4 x 10 1.6 1.8 2 2.2 3.2 1.2 2.4 1.6 1.8 2 2.2 2.4 1.4 1.6 1.8 2 2.2 2.4 1.4 1.6 1.8 2 2.2 2.4 1.4 1.6 1.8 2 2.2 2.4 1.4 1.6 1.8 2 2.2 2.4 0.09 3.5 3 2.5 1.2 1.4 -5 GWP ATP 4 -5 3.8 2.6 HTP x 10 0.08 0.07 (d) (c) 1.4 1.6 1.8 2 2.2 0.06 1.2 2.4 0.35 4 (f) PCOP AP 3 0.3 2 1 (e) 0.25 1.2 1.4 1.6 1.8 2 2.2 0 1.2 2.4 , kmole/h 0.11 EP 0.1 abs 0.09 F (g) 0.08 1.2 1.4 1.6 1.8 2 2.2 2.4 160 140 120 (h) 1.2 40 220 , °C (j) 215 T dist 35 T f,abs , °C (i) 30 1.2 1.4 1.6 1.8 5 2 PBT (10 ), $/yr 2.2 2.4 210 1.2 5 PBT (10 ), $/yr Figure 5.4: Selected Results for Operation Optimization of VOC Recovery for Eight Objectives For the first segment denoted by circles, from Figures 5.4h and 5.4j, PBT increases due to the decrease in both the Fabs from 152.3 to 136 kmol/h and Tdist from 215.8 to 211.2 °C. Reduction in the absorbent flow rate would decrease the amount of utilities required at the peripheral equipment. Also, the reduction of Tdist would reduce the 97 Chapter 5: Optimization of Recovery Processes amount of steam required in the heater. Further reduction in Fabs from 135.6 to 131.8 kmol/h and Tdist from 212.7 to 211 °C reduces the recovery efficiency of VOCs as denoted by triangles (Figures 5.4h and 5.4j). This causes the PCOP to worsen significantly as more VOCs are released into the air. In addition, with lower recovery, lesser amounts of the recovered VOCs can be sold, eroding the savings achieved with lower Fabs and Tdist. 5.2.3.3. Case C: Optimization for Several Objectives MOO with 8 objectives illustrated that some of the objectives followed similar trends. This indicates the potential to reduce the number of objectives. There is no incentive to reduce the number of objective functions if MOO for many objectives has already been carried out. Thus, assuming that the results in the previous section are not available, an attempt is made in grouping the appropriate categories into classes. First of all, human toxicity shall form a class on its own. Secondly, since aquatic and terrestrial beings form the ecosystem, it shall be classified into a group called “ecotoxicity”, denoted as ETP. This application studies the recovery of VOCs; if they are not recovered, they would be emitted into the atmosphere. The concern of VOC emissions is that it contributes to the smog formation. Thus, the potential for photochemical oxidation shall be an objective on its own. The remaining environmental categories would be grouped together and called “atmosphere”, denoted as ATMP. In summary, there are five objectives of interest – PBT, HTP, ETP, PCOP and ATMP (see Table 5.1). Since aggregation takes places for some of the environmental components, all components are normalized here. 98 Chapter 5: Optimization of Recovery Processes For the case of several objectives, PBT ranges from 1.48×105 to 2.37×105 $/yr, HTP from 2.08×10-7 to 2.42×10-7, ETP from 6.23×10-5 to 7.42×10-5, ATMP from 0.42 to 0.50 and PCOP from 0.54 to 3.25 (Figure 5.5). Generally, the trends for the several objectives optimization follow that for the many objectives case. Hence, from this case scenario, it could be observed that MOO, whether for several or many objectives, would not make a difference in the results analysis if the system is well understood and the objective(s) having greater importance than others are known in advance. 2.8 x 10 -7 7.5 (a) ETP HTP 2.4 2.2 1.4 1.6 1.8 2 2.2 7 6.5 6 1.2 2.4 0.5 1.4 1.6 1.8 2 2.2 2.4 1.4 1.6 1.8 2 2.2 2.4 4 PCOP (c) ATMP -5 (b) 2.6 2 1.2 x 10 0.45 0.4 1.2 1.4 1.6 1.8 2 5 PBT (10 ), $/yr 2.2 2.4 3 (d) 2 1 0 1.2 5 PBT (10 ), $/yr Figure 5.5: Selected Results for Operation Optimization of VOC Recovery for Five Objectives 5.2.4. Design Optimization In the above section, the optimization of the operation of the VOC recovery process was illustrated. The decision variables available for optimization are limited (i.e. 5 decision variables) – absorbent flow rate, absorbent temperature, feed temperature to the absorber, feed temperature to distillation and product temperature. In this section, we explore the optimization of the design of the VOC recovery process. In doing so, variables that were fixed in the operation case would be available as decision variables in the optimization of 99 Chapter 5: Optimization of Recovery Processes the design case. The decision variables considered on top of five variables used in the operation case are: cooling utility (refrigerant or cooling water, used for the feed, solvent and product cooler); feed stage for distillation and product column; number of stages for the distillation, product and absorber column; and temperature of the lean absorbent exiting the heat exchanger (Tabs,ex). Cooling utility is denoted by CUi, where i is feed, sol or prod depending on the cooler of concern. Feed stage is denoted as FSi, where i is can be dist or prod referring to the distillation or product column respectively. Lastly, the stages are denoted by Stagei where i is dist, prod or abs, depending whether it refers to the distillation, product or absorber column. Optimization for an operation and a design case has its fundamental difference. For the operation case, capital outlay is fixed as the equipment is deemed to be fixed in size, and only revenues and manufacturing costs will change which would result in the similar response of PBT, CF and NPW. On the other hand, for the design case, the size of the equipment can vary, the capital outlay for the design case will differ and so the profitability measures may not vary in tandem. Usually, when a design case is considered, the common profitability measure used is NPW, which calculates the present worth of receipts less the present worth of disbursements. NPW works well with all cash flow patterns, easy to compute and gives the correct ranking in most project evaluations. Table 5.2 gives a summary of the objectives considered for three cases and the corresponding decision variables, the constraints for the design case. 100 Chapter 5: Optimization of Recovery Processes Table 5.2: Objectives, Decision Variables and Constraints for VOC Process – Design Optimization Cases Objectives D Max J1 ≡ NPW Min J2 ≡ PEI E Max J1 ≡ NPW Max J2 ≡ PBT Min J3 ≡ PBP Min J4 ≡ HTP Min J5 ≡ TTP Min J6 ≡ ATP Min J7 ≡ GWP Min J8 ≡ PCOP Min J9 ≡ EP Min J10 ≡ AP F Max J1 ≡ NPW Min J2 ≡ HTP Min J3 ≡ ETP Min J4 ≡ PCOP Min J5 ≡ ATMP Decision Variables a) 110 ≤ Fabs ≤ 170 kmol/hr b) 30 ≤ Tabs ≤ 40 °C c) 30 ≤ Tf,abs ≤ 40 °C d) 200 ≤ Tdist ≤ 230 °C e) 30 ≤ TVOC ≤ 40 °C f) 200 ≤ Tabs,ex ≤ 230 °C g) CUi = 5 or 25 °C, where i=sol, feed, VOC FSi h) 0.3 ≤ ≤ 0.7, Stage i where i=dist, pdt i) 8 ≤ Stagedist ≤ 14 j) 16 ≤ Stagepdt ≤ 26 k) 30 ≤ Stageabs ≤ 60 Constraints a) (∆Tj)i ≥ 5 °C, where j=1, 2 and i=feed, sol, VOC b) (∆Tj)i ≥ 25 °C, where j=1, 2 and i=hx, dist c) Ft ≥ 0.75 for all exchangers d) Heat and Material Balances solved in HYSYS 5.2.4.1. Case D: Bi-Objective Optimization As in the operation case, the design case is first optimized for two objectives – namely, NPW and PEI (see Table 5.2). The Pareto-optimal solutions obtained can be segregated into four segments (see Figure 5.6): (1) NPW in the range of $-2.42×106 to $-1.89×106 denoted by triangles; (2) NPW in the range of $-1.76×106 to $-1.56×106 denoted by squares; (3) NPW in the range of $-1.53×106 to $-1.38×106 denoted by circles; and (4) NPW in the range of $-1.38×106 to $-1.07×106 denoted by diamonds. The three decision variables that are not shown in Figure 5.6 are Tabs,ex, CUfeed and CUpdt which took on constant values at 216.4°C, 25°C and 25°C respectively. 101 Chapter 5: Optimization of Recovery Processes PEI 160 120 140 abs 1.2 , kmol/h (a) 1.4 F 1.6 1 0.8 -2.5 -2 -1.5 -1 40 , °C f,abs , °C abs -1 -2 -1.5 -1 -2 -1.5 -1 -2 -1.5 -1 -2 -1.5 -1 -1.5 -1 35 T T -2 -1.5 30 -2.5 -1 230 40 (f) , °C (e) , °C -1.5 (d) 35 30 -2.5 VOC 220 210 200 -2.5 25 -2 -1.5 30 -2.5 7 -1 (h) 6.5 dist 20 15 FS CUsol, ° C (g) 10 -2 -1.5 14 (i) (j) Stage dist FSpdt 14 12 10 -2.5 -2 -1.5 10 60 (k) 22 20 18 16 -2.5 12 8 -2.5 -1 Stage abs Stage pdt 24 5 -2.5 -1 16 26 6 5.5 5 -2.5 18 35 T dist -2 40 (c) T (b) -2.5 -2 -1.5 6 NPW (10 ), $ -1 (l) 50 40 30 -2.5 -2 6 NPW (10 ), $ Figure 5.6: Selected Results for Design Optimization of VOC Recovery for NPW and PEI 102 Chapter 5: Optimization of Recovery Processes 1.75 x 10 -7 2.7 (m) TTP HTP 1.65 (n) 2.5 2.4 1.6 1.55 -2.5 -2 x 10 -1.5 -2 -1.5 -1 -2 -1.5 -1 -1.5 -1 -5 0.052 (o) 2.2 2.1 2 -2.5 0.21 2.3 -2.5 -1 GWP ATP -5 2.6 1.7 2.3 x 10 -2 -1.5 -1 0.048 0.046 -2.5 1.5 (q) AP PCOP 0.2 0.19 0.18 -2.5 -2 -1.5 -1 (p) 0.05 (r) 1 0.5 0 -2.5 -2 6 NPW (10 ), $ EP 0.07 (s) 0.065 0.06 -2.5 -2 -1.5 -1 6 NPW (10 ), $ Figure 5.6 (cont.): Selected Results for Design Optimization of VOC Recovery for NPW and PEI In the first segment, the increase in the NPW is due to the reduction in Stageabs from 45 to 33 trays (Figure 5.6l). The capital cost of the absorber contributes the most to capital investment; hence, a reduction in Stageabs has a larger effect on NPW than that for solvent and product columns (i.e. Stagedist and Stagepdt in Figures 5.6j and 5.6k). The reduction in the number of absorber trays, however, reduces the absorption capacity of the VOCs which results in the marginal increase in PEI (Figure 5.6a). 103 Chapter 5: Optimization of Recovery Processes In the second segment, the increase in NPW and PEI is due to the decrease in the Fabs from 161.8 to 155.5 kmol/h (Figure 5.6b). This reduces the capacity of the entire system, decreasing the utilities required for the peripheral equipment (e.g. pumps, heaters and coolers) and also the sizes required for the heaters, coolers and exchangers. Reduction in costs leads to the increase in cash flow (CF) and thus NPW. However, the reduction in Fabs compromises the absorbing capacity of the VOC-laden gaseous stream, resulting in increased emissions and higher PEI (Figure 5.6a). At the transition from squares to circles, though Stageabs increased to 33, PEI is about the same at 0.89 due to higher Tabs in Figure 5.6c. For this third segment, it is again largely due to the decrease in the Stageabs. The difference here is that the CUsol has changed from the refrigerant to cooling water, which is a cheaper alternative. This led to a slight increase in Tabs from 30 to 31°C. As a cheaper cooling utility is used, manufacturing costs are reduced leading to an increase in NPW. Finally, for the last segment denoted with diamonds, the largest contributor to the increase in NPW and PEI is the decrease in Fabs from 169.6 to 155 kmol/h (Figure 5.6b). The line of reasoning follows that given for the second segment denoted with squares; the only difference between the two segments is that the CUsol has changed from the refrigerant to cooling water (Figure 5.6g). The use of cooling water has reduced manufacturing costs as it is a cheaper alternative, resulting in higher NPW. On the other hand, it increases the Tabs (Figure 5.6c). This and increase in the Tf,abs (Figure 5.6d) led to the reduced recovery efficiency of VOCs, resulting in higher emissions and higher PEI. Other minor contributor to the increase in NPW is the increase in TVOC from 30 to 40 °C 104 Chapter 5: Optimization of Recovery Processes (Figure 5.6f) and the general increase in Tdist from 217 to 222 °C. Both lead to the reduction in the use of steam (heating medium), resulting in lower costs and higher NPW. From the plots of the individual components contributing to PEI (Figures 5.6m to 5.6s), it is evident that PEI is largely influenced by PCOP (Figure 5.6r) as both have similar trend. This is expected since the major contribution to PEI is due to VOCs emissions, which cause smog formation and photochemical oxidation. 5.2.4.2. Case E: Optimization for Many Objectives Optimizing for only two objectives in the above section has illustrated that the major variables affecting the objectives: NPW and PEI are Fabs, Stageabs and CUsol. However, the nature of aggregated indicators is that they only reflect the component that is most dominating; for example, minimizing PEI does not necessarily minimize the various components of PEI. Hence, in this case study, various economic performance indicators and the components of PEI will be optimized. The objectives to be optimized are – NPW, PBT, PBP, HTP, TTP, ATP, GWP, PCOP, EP and AP. From the results, it can be seen that the economic indicators generally move in tandem (Figures 5.7a and 5.7b). When NPW is maximized, PBP is minimized and PBT is maximized. Hence, these two economic objective functions are not conflicting in this case study. However, economic objective functions are not necessarily optimized simultaneously. Lee et al. (2008) provides a case study of a heat exchanger network illustrating conflicting economic objective functions, where the maximization of NPW leads to the deterioration of PBP (i.e. higher values of PBP). On the other hand, for the environmental indicators, five of the environmental components (namely, TTP, ATP, 105 Chapter 5: Optimization of Recovery Processes 20 10 x 10 5 PBT, $/yr PBP, yr (b) 15 10 (a) 5 -7 8 -6 x 10 -5 -4 -3 -2 0 -5 -7 -1 -7 3.5 (c) x 10 4 2 0 -7 1.5 -7 -6 x 10 -5 -4 -3 -2 -1 0.08 GWP ATP 2.5 2 1.5 1 -7 0.4 -6 -5 -4 -3 -2 (g) -6 -5 -4 -3 -2 -1 -6 -5 -4 -3 -2 -1 -6 -5 -4 -3 -2 -1 -6 -5 -4 -3 -2 -1 -6 -5 -4 -3 -2 -1 -6 -5 -4 -3 -2 -1 -6 -5 -4 -3 -2 -1 (d) (i) EP 0.08 0.06 0.04 -7 40 -6 -5 -4 -3 -2 (f) (h) 20 10 0 -7 -1 Fabs, kmol/h 0.1 -7 0.1 160 (j) 140 120 -1 -7 18 16 FSprod , °C -2 0.04 PCOP AP 0.2 35 T abs -3 0.06 0.02 -7 30 -1 0.3 30 -7 26 14 12 (k) (l) -6 -5 -4 -3 -2 -1 24 22 20 18 (m) 16 -7 -6 10 -7 60 Stage abs prod -4 -5 (e) Stage -5 2.5 2 3 -6 -5 3 TTP 6 HTP 5 50 40 (n) -5 -4 -3 NPW (106), $ -2 -1 30 -7 NPW (106), $ Figure 5.7: Selected Results for Design Optimization of VOC Recovery for Ten Objectives 106 Chapter 5: Optimization of Recovery Processes GWP, AP, ETP) move in sync – the objective values are about constant when NPW is less than US$ -4.45×106 and thereafter scattered when NPW is more than US$ -4.45×106 (see Figures 5.7d, 5.7e, 5.7f, 5.7g and 5.7i). HTP and PCOP show a different trend from the rest. The former is scattered when NPW is less than US$ -2.65×106 and about constant when NPW is more than US$ -2.65×106 (see Figure 5.7c). PCOP is scattered throughout the entire range of NPW with a decreasing trend of PCOP as NPW increases (see Figure 5.7h). Comparing the ranges of the objective functions with that for Case D, a wider variety of Pareto-optimal is presented to the decision maker in Case E. Two points ( and in Figure 5.7) are selected for comparison, they are chosen on the basis that they have very close NPW value but with a very different value for HTP (since it is the first environmental impact component on the list). Table 5.3 shows the values of the objective functions for each optimal solution. The solution indicated by filled square has marginally superior values for 7 out of 10 objectives (namely, PBT, PBP, NPW, ATP, GWP, AP and EP) while the solution indicated by filled triangle is superior for the remaining 3 objectives where 2 of them (HTP and PCOP) are significantly lower. An attempt is made to find the trends of the objective functions with their decision variables as follows. NPW increases as Stageabs decreases due to lower investment cost, and also as Tabs decreases due to improved recovery of VOCs for sale (Figures 5.7n and 5.7k respectively). The reduction in Tabs has a direct effect on PCOP as it improves the recovery of VOCs, resulting in lower emissions of VOCs and hence lower potential for smog formation. TTP, ATP, GWP, AP and EP have the same trend as Fabs, FSprod and Stageprod (Figures 5.7j, 5.7l and 5.7m respectively). Increase in Fabs increases the load on 107 Chapter 5: Optimization of Recovery Processes the entire VOC recovery system and the amount of steam required, resulting in an increase in TTP, ATP, GWP, AP and EP because they are more influenced by the energy consumption in generating steam rather than by the emission of VOCs. Table 5.3: Comparison of Two Selected Pareto-optimal Solutions PBT PBP NPW HTP TTP ATP GWP AP PCOP EP Units $/yr yr $ − − − − − − − -24640 10.28 -3.75E+06 2.35E-07* 1.5991E-05* 1.43E-05 0.033 0.129 14.07* 0.042 -23730* 10.27* -3.74E+06* 7.60E-07 1.5994E-05 1.41E-05* 0.032* 0.126* 21.36 0.041* Note: Better objective value is identified with * 5.2.4.3. Case F: Optimization for Several Objectives Just like in Section 5.2.3.3, the design case for VOC recovery is also optimized for several objectives – namely, HTP, ETP, ATMP and PCOP along with NPW. Since aggregation takes place for some of the environmental components, all components are normalized here. As NPW increases from US$ -6.28×106 to US$ -1.07×106, HTP decreased from 7.64×10-7 to 1.24×10-7, ETP increased from 3.00×10-5 to 5.02×10-5, ATMP increased from 0.20 to 0.34 and PCOP decreased from 21.41 to 0.53 (see Figure 5.8). It is observed that the trends follow that of MOO for many objectives and so the values of the decision variables are omitted in Figure 5.8 for brevity. Hence, the use of several objectives could well represent that for the many objectives optimization (or vice versa) if the decision maker has made an informed choice. In addition, the optimization 108 Chapter 5: Optimization of Recovery Processes for several objectives captures more of the plant performance rather than using only two objectives. 8 x 10 -7 6 (a) ETP HTP 6 4 2 0 -7 -5 -4 -3 -2 0.25 0.2 -7 -6 -5 -4 -3 6 NPW (10 ), $ -2 -1 (b) 5 4 30 (c) 0.3 -5 3 -7 -1 PCOP ATMP 0.35 -6 x 10 -6 -5 -4 -6 -5 -4 -3 -2 -1 -3 -2 -1 (d) 20 10 0 -7 6 NPW (10 ), $ Figure 5.8: Optimal Objective Values for Design Optimization of VOC Recovery for Five Objectives 5.3. Solvent Recovery In line with the above case study, another recovery process is chosen for analysis. The chosen solvent recovery process aims to separate a spent wash solution into its individual components – namely, acetone, benzene, ethylene dichloride and toluene. Their compositions in the spent wash solution and their normal boiling points are provided in Table 5.4. The boiling points for benzene and ethylene dichloride are very close and thus would be the hardest to separate. In the case study, benzene (B) and ethylene dichloride (C) should be recovered, while disposing of acetone (A) and toluene (D) as the former two components are present in larger amounts. Moreover, acetone and toluene, which are to be disposed off, are less hazardous than benzene and ethylene dichloride. 109 Chapter 5: Optimization of Recovery Processes Table 5.4: Compositions of Components in Spent Wash Solution Solvents (Abbreviations) Acetone (A) Benzene (B) Ethylene Dichloride (C) Toluene (D) Feed amount (kmol/hr) 3 10 15 6 Feed amount (kg/hr) 174 781 1484 552 Normal boiling point (K) 329.4 353.2 356.6 383.8 5.3.1. Background Chakraborty and Linninger (2002) had performed a detailed column sequencing analysis for the separation of the spent wash solution in Table 5.4 with economical and ecological considerations. The economic objective considered is the total cost operating costs, an equipment charge associated with the number of trays, and credits for recycled solvents. The environmental objective is the global pollution index using the chemical ranking methodology. Epsilon-constraint method was employed to solve the bi-objective optimization problem. It is desired to separate the spent wash solution having acetone, benzene, ethylene dichloride and toluene, into four streams, having high purity of each component in each stream. Five different column sequences were considered for this separation process. For each of them, an optimal Pareto front is obtained and presented. It was observed that out of the five suggested sequences, only two shown in Figure 5.9 provided better solutions than the remaining three. Sequence 1 (Figure 5.9) follows the heuristic of perform the easiest separation first and leave the most difficult to the last (Turton et al., 2003). On the other hand, Sequence 2 follows the heuristic that the largest product stream should be removed first so that the subsequent separation units are smaller (Turton et al., 2003). 110 Chapter 5: Optimization of Recovery Processes A A/B Sequence 2: Column 3 Column 2 Column 1 A/B/C A/B/C/D B C D Figure 5.9: Sequences 1 and 2 from Chakraborty and Linninger (2002) 5.3.2. Design Optimzation As mentioned in Section 5.3.1, both Sequences 1 and 2 of Figure 5.9 follow the heuristics given in the literature and the superiority of either sequence is not obvious. Sequence 1 has been chosen for this case study. In Chakraborty and Linninger (2002), number of stages and feed stages were not specified. Hence, only design optimization is studied for this case study. Shortcut columns in HYSYSTM were employed to give the first estimate of the number of stages and feed stage for each column before MOO for economic and environmental criteria. For each distillation column i, there are several decision variables available for optimization – namely the number of stages (Stagei), the feed stage (FSi), the recoveries of light and heavy keys (RLKi and RHKi), and the type of heating and cooling utility used in the reboiler and condenser respectively (HUi and CUi). It has been found that a variation of 25% or less in the total number of stages has minimal effect on the total 111 Chapter 5: Optimization of Recovery Processes annual cost incurred as long as the feed stream location has been optimized (Lek et al., 2004). Hence, the number of stages for each column is kept constant as provided by the shortcut columns in HYSYSTM at 31, 26 and 126 stages for Columns 1, 2 and 3 respectively; on the other hand, the feed stage will be optimized. The purpose of MOO of the solvent recovery system is to ensure that the process is at its most sustainable state. In other words, the system will be optimized using both economic and environmental criteria. Since it is at design stage, NPW is used as the main economic objective; on the other hand, one or more environmental indicators are employed (see Table 5.5). Table 5.5: Objectives, Decision Variables and Constraints for Solvent Recovery Process – Design Optimization Cases Objectives G Max J1 ≡ NPW Min J2 ≡ PEI H Max J1 ≡ PBT Max J2 ≡ NPW Min J3 ≡ PBP Min J4 ≡ HTP Min J5 ≡ TTP Min J6 ≡ ATP Min J7 ≡ GWP Min J8 ≡ PCOP Min J9 ≡ EP Min J10 ≡ AP 1. 2. 3. 4. 5. 6. 7. Decision Variables 19 ≤ FS1 ≤ 22 9 ≤ FS2 ≤ 11 52 ≤ FS3 ≤ 65 0.95 ≤ RLKi ≤ 0.999, where i = 1, 2, 3 0.95 ≤ RHKi ≤ 0.999, where i = 1, 2, 3 CUi = 5 or 25 °C HUi = 160,184 or 254 °C, where i = 1, 2, 3 Constraints 1. (∆Tj) i ≥ 5 °C, where j = 1, 2 and i = 1, 2, 3 for all condensers 2. (∆Tj) i ≥ 25 °C, where j = 1, 2 and i = 1, 2, 3 for all reboilers 3. Purityi ≥ 0.98, where i = EAcetate or Benzene 4. Heat and Material Balances solved in HYSYS 5.3.3.1. Case G: Bi-objective Optimization The solvent recovery system is first optimized for two objectives: NPW and PEI (see Table 5.5). As NPW increases from $ 1.86×106 to $ 2.80×106, PEI increases slightly from 11.75 to 12.03 (see Figure 5.10a). Optimal values of some decision variables are shown 112 Chapter 5: Optimization of Recovery Processes in Figures 5.10b to 5.10g. Optimal values of the other decision variables are as follows. The condenser in each column uses cooling water as the cooling utility; the reboilers in the first and third column, HU1 and HU3, use low pressure (LP) steam while the second column, HU2, uses medium pressure (MP) steam. The choice of refrigerant as the cooling utility and the high pressure steam as a heating medium are not favored for this process. FS3 is constant at stage number 65; RHK1 varies minimally from 0.987 to 0.988 and RLK3 also varies marginally from 0.978 to 0.979. It is observed from the optimization results that the decision variable that has the largest influence on PEI is RLK2, which corresponds to the acetone component flow rate emitted from the distillate stream of column 2. Acetone has the largest contribution to PCOP which contributes largely to PEI (Figure 5.10q). Corresponding to RLK1, the total product flow rate and thus revenue for the distillate stream of column 3 dipped at two points at NPW = US $2.20×106 and US$ 2.43×106 (Figure 5.10d); simultaneously, the demand for cooling water and LP steam also dropped simultaneously (Figures 5.10h and 5.10i), which resulted in the drop in manufacturing costs. Hence, PBT as well as NPW increased. The feed stage for each column (FS1, FS2 and FS3) does not have insignificant impact on the objective functions (Figures 5.10b and 5.10c). From Figure 5.10j, the increase in the MP steam utilization is directly associated with the increase in both RLK2 and RHK2. With increased recovery of acetone in the distillate stream for column 2, the amount of MP steam required is higher to provide better separation in column 2. For cooling water and the LP steam usage in Figures 5.10h and 5.10i, the data points are segregated into four segments where the first three segments generally decrease with increasing NPW and the last segment decreasing continuously 113 Chapter 5: Optimization of Recovery Processes with NPW. The decrease is due to the increase in RLK2 (Figure 5.10e) which led to the decrease in the utilities required in Column 3. However, within each of the first three segments, the amount cooling water and LP steam used increases. This is as a result of the higher RLK1 (Figure 5.10d) and thus column 3 would have to accommodate for the increase in component B and C entering it. Figures 5.10k to 5.10s provide the decision maker with an idea of how other economic criteria and environmental impact categories varied for the bi-objective optimization case. The economic objective functions as shown in Figures 5.10k and 5.10l varied in sync with NPW over the range of the Pareto-optimal solutions. The maximization of NPW corresponds to the maximization of PBT and the minimization of PBP. Hence, the use of only one of these in MOO would suffice in this application. For the environmental impact categories in Figures 5.10m to 5.10s, it can be observed that there are three generic groups of the categories that move in sync. The first group comprises HTP and TTP which is influenced by the emissions of the light key the bottom stream of column 1; the higher RLK1 is, the lower the values of HTP and TTP. The second group consists of ATP, GWP, AP and EP which follows the same trend as cooling water and LP steam as explained above. The last group consists of PCOP only which is contributed by the recovery and thus emissions of acetone in column 2 (i.e. RLK2 in Figure 5.10e). 114 Chapter 5: Optimization of Recovery Processes 1 FS PEI 21 11.9 11.8 11.7 20 (b) 1.8 2 2.2 2.4 2.6 19 1.8 2.8 (c) 0.99 RLK1 2 11 FS 22 (a) 12.1 12 10.5 2 2.2 2.4 2.6 2.8 2 2.2 2.4 2.6 2.8 2 2.2 2.4 2.6 2.8 2 2.2 2.4 2.6 2.8 2 2.2 2.4 2.6 2.8 2.4 2.6 2.8 (d) 0.98 0.97 0.96 10 1.8 2 2.2 2.4 2.6 0.95 1.8 2.8 (e) 0.99 RHK2 RLK 2 0.99 0.98 0.97 0.96 0.98 0.97 0.96 0.95 1.8 2 2.2 2.4 2.6 2.8 9 CW, $/yr RHK 3 0.99 0.98 0.97 0.96 2 2.4 2.6 2.8 MP Steam, $/yr x 10 2.2 6 1.85 1.8 (i) 1.75 1.8 PBT, $/yr 8.4 (h) 1.9 10 8.6 2 x 10 2.2 2.4 2.6 2.8 2.2 x 10 5 (j) 2.1 2 1.9 1.8 1.8 5 2.8 (k) 9 8 7 1.8 8.2 1.8 PBP, yr LP Steam, $/yr 1.95 x 10 4 8.8 (g) 0.95 1.8 (f) 0.95 1.8 2.6 2.4 2.2 (l) 2 2.2 2.4 6 NPW (10 ), $ 2.6 2.8 2 1.8 2 2.2 6 NPW (10 ), $ Figure 5.10: Selected Results for Design Optimization of Solvent Recovery for NPW and PEI 115 Chapter 5: Optimization of Recovery Processes x 10 -3 8 4 ATP HTP 6 2 (n) 2 2.2 2.4 2.6 7 1.8 2.8 0.0136 0.175 GWP TTP 0.0135 0.0134 0.0133 0.0132 1.8 2.2 2.4 2.6 2.2 2.4 2.6 2.8 2 2.2 2.4 2.6 2.8 2.4 2.6 2.8 (p) 0.17 0.165 (o) 2 2 0.16 1.8 2.8 0.7 (q) 11 10.9 0.68 AP PCOP 11.1 -5 7.5 (m) 0 1.8 x 10 0.66 10.8 10.7 (r) 1.8 2 2.2 2.4 2.6 2.8 0.64 1.8 2 2.2 6 NPW (10 ), $ EP 0.105 0.1 (s) 0.095 1.8 2 2.2 2.4 2.6 2.8 6 NPW (10 ), $ Figure 5.10 (cont.): Selected Results for Design Optimization of Solvent Recovery for NPW and PEI 5.3.3.2. Case H: Optimization for Many Objectives Optimizing for only two objectives in the previous section has illustrated that the major variables affecting the objectives: NPW and PEI are RLK1, RLK2 and the utilization rate of cooling water and LP steam. As mentioned in the VOC recovery process, aggregated indicators only reflect the environmental component that dominates the most. Hence, in this case study, various economic performance indicators and the components of PEI are optimized as a multi-objective problem. The reason for choosing many, rather than 116 Chapter 5: Optimization of Recovery Processes several objectives for this case study is because the behavior of the objective functions and the relative importance of each objective are not clear. This is unlike in the case of VOC recovery process whereby the main concern of VOC is the resultant smog formation. Here, for the solvent recovery, the most difficult case will be considered where multiple economic as well as environmental objectives are optimized simultaneously. The objectives considered are NPW, PBT, PBP, HTP, TTP, ATP, GWP, PCOP, AP and EP (see Table 5.5). Figure 5.11 shows the results for the optimization of the solvent recovery process for many objectives – Figures 5.11a to 5.11i are the objective functions and Figures 5.11j to 5.11v are all the decision variables except CU1 and CU3; all these are plotted against NPW. Note that the environmental impact categories are not normalized here. For more than 99% of the Pareto-optimal solutions, the choice for CU1 and CU3 is cooling water. Figures 5.11a and 5.11b indicate that the maximization of NPW would also lead to the maximization of PBT and minimization of PBP. Hence, it can be deduced that the economic objectives are not conflicting in this case study. On the other hand, for the environmental indicators, four of the environmental components (namely, ATP, GWP, AP and EP) share the same Pareto-optimal front (Figures 5.11e, 5.11f, 5.11h and 5.11i). HTP, TTP and PCOP show a different trend from the rest (Figures 5.11c, 5.11d and 5.11g). Two points in Figure 5.11 (shown as and ) have been selected for discussion to determine that the scattered plots are indeed Pareto-optimal solutions. They are chosen on the basis that they have approximately the same NPW value with a very different value for HTP (since it is the first environmental impact component on the list). Table 5.6 117 Chapter 5: Optimization of Recovery Processes shows the values of the objective functions for each solution. The solution indicated by the filled square has marginally superior values for 6 out of 10 objectives (namely, PBP, NPW, ATP, GWP, AP and EP). The environmental objectives having similar Pareto front (i.e. ATP, GWP, AP and EP) took on lower values while the remaining three environmental objectives (i.e. HTP, TTP and PCOP) took on higher (inferior) values. Table 5.6: Comparison of Two Selected Pareto-optimal Solutions PBT PBP NPW HTP TTP ATP GWP AP PCOP EP Units $/yr yr $ − − − − − − − 2.03E+05* 5.717 -2.79E+05 1.94E-04* 1.29E-02* 9.60E-05 0.22 0.86 10.80* 0.13 1.75E+05 5.716* -2.41E+05* 8.54E-03 1.37E-02 7.07E-05* 0.16* 0.62* 11.17 0.09* Note: Better objective value is identified with * Comparing the bi-objective optimization results with MOO results for many objectives, differences in the values undertaken by the decision variables are observed. FS3, CUi and HUi for i =1, 2 and 3 do not take on constant values but vary with the objective functions (see Figures 5.11l and 5.11s to 5.11v). From Figures 5.11g and 5.11o, the parallel trends are still observed between the RLK2 (which corresponds to acetone emissions) and PCOP. RLK3 and RHK3 decreased slightly at high values of NPW (Figures 5.11w and 5.11x). The drop in recoveries indicates that the separation of products is worse and hence the required reflux and vapor are reduced. As a result, the demand for cooling water and LP steam dropped (Figure 5.11y and 5.11z), which also resulted in the 118 Chapter 5: Optimization of Recovery Processes decrease of the following environmental criteria – ATP, GWP, AP and EP (Figures 5.11e, 5.11f, 5.11h and 5.11i). 5.4. Conclusions To measure the sustainability of a process, both economic and environmental indicators should be used simultaneously. In this chapter, two recovery processes have been optimized with at least one economic and one environmental objective. VOC and solvent recovery processes were used as case studies for sustainability optimization. Each of the recovery systems has been optimized for bi-objective and for many objectives. Only VOC recovery system was optimized for several objectives as well. From the results obtained, it is evident that the optimization using more than two objectives would provide the decision with wider spread of solutions, giving the decision maker an insight of the process and the trade-offs experienced for each objective. With better knowledge of the process, the decision maker would be less biased in choosing the preferred solution which is to be discussed in Chapter 6. 119 Chapter 5: Optimization of Recovery Processes x 10 5 5 0 -5 -3 0.01 -2 -1 0 1 2 -2 x 10 -1 0 1 2 3 -1 0 1 2 3 -2 -1 0 1 2 3 -2 -1 0 1 2 3 -2 -1 0 1 2 3 -2 -1 0 1 2 3 -2 -1 0 1 2 3 (d) 0.013 0.0125 -3 -5 (e) 0.24 (f) GWP ATP -2 0.0135 9 8 7 0.22 0.2 0.18 0.16 6 -3 11.5 50 0.014 0.005 10 100 0 -3 (c) 0 -3 (b) 150 3 TTP HTP 200 (a) PBP, yr PBT, $/yr 10 -2 -1 0 1 2 3 -3 1 (h) (g) AP PCOP 0.9 11 0.8 0.7 10.5 -3 0.16 -2 -1 0 1 2 3 -3 22 (j) (i) 1 21 0.12 FS EP 0.14 20 0.1 0.08 -3 11 -2 -1 0 1 2 19 -3 3 65 (k) (l) 3 60 10 FS FS 2 10.5 55 9.5 9 -3 -2 -1 0 1 6 NPW (10 ), $ 2 3 50 -3 6 NPW (10 ), $ Figure 5.11: Selected Results for Design Optimization of Solvent Recovery for Ten Objectives 120 Chapter 5: Optimization of Recovery Processes (m) 1 0.98 0.97 0.96 -2 -1 0 1 2 0.97 0.95 -3 3 (o) RHK 0.98 0.97 0.96 -2 -1 0 1 2 2 3 -2 -1 0 1 2 3 -2 -1 0 1 2 3 -2 -1 0 1 2 3 -2 -1 0 1 2 3 -2 -1 0 1 2 3 (r) 3 0.99 0.98 RHK 3 1 0.97 (q) RLK 0 0.98 0.95 -3 3 0.99 0.97 0.96 0.98 0.97 0.96 0.95 -3 -2 -1 0 1 2 0.95 -3 3 25 190 (s) (t) 20 HU , ° C 15 180 1 CU , ° C 2 -1 0.96 0.95 -3 10 5 -3 300 -2 -1 0 1 2 170 160 -3 3 300 (u) HU , ° C (v) 250 3 250 2 HU , ° C -2 (p) 0.99 2 0.99 2 0.98 0.96 0.95 -3 RLK (n) 0.99 RHK RLK 1 0.99 200 150 -3 x 10 -1 0 1 2 3 4 3 (w) 10 5 0 -3 150 -3 LP Steam, $/yr CW, $/yr 15 -2 200 -2 -1 0 1 6 NPW (10 ), $ 2 3 x 10 6 (x) 2 1 0 -3 6 NPW (10 ), $ Figure 5.11 (cont.): Selected Results for Design Optimization of Solvent Recovery for Ten Objectives 121 Chapter 6: Ranking of the Pareto-optimal Solutions Chapter 6 Ranking of the Pareto-optimal Solutions 6.1. Introduction The Pareto-optimal solutions obtained for the different cases have given us an insight into the two applications and how the various decision variables influence the objective functions. These solutions are plentiful and provide us many options for the design and operation of the process for the application. The myriad of solutions, however, does not provide the user a unique optimal solution for the application. Hence, in this section, the Pareto-optimal solutions are ranked and the preferred solution determined. As the decision maker has an insight into the behavior of the objectives, the inputs required from him/her for ranking the Pareto domain is expected to be more objective. Thibault (2008) described two different methods for the ranking of Pareto domains – (1) Net Flow Method, NFM and (2) Rough Set Method, RSM, and used them for chemical engineering applications in the production of gluconic acid (HalsallWhitney and Thibault, 2006) and in the manufacture of paper (Renaud et al., 2007). NFM uses the concept of outranking relation which is implemented by performing pair-wise comparisons of the Pareto-optimal solutions. The basis of ranking is in the use of concordance and discordance indices which are derived from user supplied parameters – weights and three thresholds (indifference, preference and veto). On the other hand, RSM is based on a set of preference and non-preference rules that are derived from ranking a small set of selected Pareto-optimal solutions by the decision-maker. These rules are then used to rank the complete set of Pareto-optimal solutions. 122 Chapter 6: Ranking of the Pareto-optimal Solutions For NFM, the user is required to provide the weights as well as three different threshold limits to enable ranking. The decision maker is thus required only once for his/her inputs before the NFM generates the results. On the other hand, RSM involves the selection of a handful of solutions which the decision maker has to rank the solutions. Whether the chosen solutions are representative of the pool of solutions available is debatable. Moreover, RSM requires more frequent feedback from the decision maker for he/she is required to state his/her preferences for several pairs of solutions until sufficient preference and non-preference rules are created for ranking. Hence, NFM is chosen for this study to rank the solution set for both applications – VOC recovery and solvent recovery processes. Net Flow Method (NFM) requires the decision maker to provide four ranking parameters for each objective. These are weight (Wk), indifference threshold (Qk), preference threshold (Pk) and veto threshold (Vk). Wk indicates the relative importance of objective k and it should be within the range 0 ≤ Wk ≤ 1 and that ∑Wk = 1. Qk is the range of each objective for which the decision maker is unable to state that one solution is preferred over another solution. Pk is the threshold whereby if the difference between two solutions for a given objective exceeds the threshold, the decision maker would then be able to state his/her preference for the solution with the better objective value. The last parameter, Vk, is used to penalize a solution relative to another solution when the difference between them for a particular objective is above Vk. The three threshold values for objective k should be such that: 0 ≤ Qk ≤ Pk ≤ Vk. The procedure for NFM including the calculation of concordance and discordance indices, leading to the ranking of the solutions can be found in Thibault et al. (2002) and Thibault (2008). 123 Chapter 6: Ranking of the Pareto-optimal Solutions 6.2. et Flow Method Net Flow Method (NFM) requires the decision maker to provide four ranking parameters for each objective. These parameters are weights (Wk), indifference threshold (Qk), preference threshold (Pk) and veto threshold (Vk). Wk indicates the relative importance of objective k and it should be within the range 0 ≤ Wk ≤ 1 and that ∑Wk = 1. Qk indicates the range of each objective for which the decision maker is unable to state that one solution is preferred over another solution. Pk is the threshold whereby if the difference between two solutions for a given objective exceeds the threshold, the decision maker would then be able to state his/her preference for the solution with the better objective value. The last parameter, Vk, is used to penalize a solution relative to another solution when the difference between both solutions for a particular objective is above tolerance. The three thresholds values for objective k should be given such that the following relationship holds: 0 ≤ Qk ≤ Pk ≤ Vk. After the values of the weights and the thresholds have been defined by the user, the NFM algorithm would be used to rank all the solutions. The details of NFM algorithm is as follows. 1. To begin with, for each combination of two solutions in the Pareto domain, the difference between two solutions i and j in their values of objective function k is calculated by using: ∆ k [i, j ] = Fk (i ) − Fk ( j ) i ∈ [1, M ]   j ∈ [1, M ] k ∈ [1, n]  where M is the total number of Pareto-optimal solutions and n indicates the total number of objectives. It should be noted that a minimizing criterion considers 124 Chapter 6: Ranking of the Pareto-optimal Solutions ∆ k [i, j ] , while a maximizing criterion considers its negative value, − ∆ k [i, j ] . When it is required for an objective to meet a specified target value, Fk(i) and Fk(j) would correspond to the absolute differences between the values of the criterion k and its target value; then, ∆ k [i, j ] is used directly since it is desired to minimize the distance of the criterion to its target value. 2. The values of ∆ k [i, j ] obtained in step 1 are now used in the calculation of the individual concordance index ck [i, j ] for each criterion. ck [i, j ] is calculated for all n objective functions and for each pair of solutions using the following relationships:  1   P − ∆ k [i, j ] ck [i, j ] =  k  Pk − Qk  0 if ∆ k [i, j ] ≤ Qk if Qk < ∆ k [i, j ] ≤ Pk if ∆ k [i, j ] > Pk ck [i, j ] measures the degree of how good solution i is when compared to solution j. When ∆ k [i, j ] between the solutions is less than the indifference threshold, ck [i, j ] is given a score of unity. On the other hand, when ∆ k [i, j ] is more than the preference threshold, ck [i, j ] is given a score of zero. Hence, for differences lying in between the indifference and preference thresholds, the score would vary linearly from 1 to 0, depending on the value of ∆ k [i, j ] . Figure 6.1(a) illustrates how the individual concordance index is determined using the values of the calculated differences, the indifference threshold, and the preference threshold. 125 Chapter 6: Ranking of the Pareto-optimal Solutions 3. The weighted sum of individual concordance indices for different objective functions k is calculated to give the global concordance index when solution i is compared to solution j. i ∈ [1, M ]  k =1  j ∈ [1, M ] 4. Another index, called the discordance index, Dk[i,j], is calculated for each criterion k n C[i, j ] = ∑Wk ck [i, j ] employing the preference and veto thresholds:  0   ∆ [i, j ] − Pk Dk [i, j ] =  k  Vk − Pk  1 if ∆ k [i, j ] ≤ Pk if Pk < ∆ k [i, j ] ≤ Vk if ∆ k [i, j ] > Vk The discordance index measures the degree of how significantly worse is solution i when compared to solution j. When ∆ k [i, j ] is less than the preference threshold, Dk[i,j] is given a score of 0. Between the preference and veto thresholds, Dk[i,j] varies linearly from 0 to 1. For a difference larger than the veto threshold, Dk[i,j] is given a score of 1. Figure 6.1(b) illustrates how the discordance index is determined using the preference and veto thresholds. 5. Employing both the global concordance and discordance indices, the relative performance of each pair of Pareto-optimal solutions is evaluated by calculating each element of the outranking matrix σ[i,j] using the following equation: [ ] n σ [i, j ] = C[i, j ] Π 1 − (Dk [i, j ])3   k =1 i ∈ [1, M ]   j ∈ [1, M ] Each value of σ[i,j] measures the quality of solution i relative to solution j for all n objective functions. When σ[i,j] has a value close to 0, it indicates that solution j outranks solution i. If the value is near 1, there are two possibilities – (1) solution i 126 Chapter 6: Ranking of the Pareto-optimal Solutions may outrank solution j, or (2) solution i is located in the vicinity of solution j. In the absence of discordant criteria, the outranking matrix would be equivalent to the global concordance matrix. However, it only takes one discordant criterion to make an element of the outranking matrix equal to zero. The definition of such a relation, called an outranking relation, involves the three thresholds mentioned above, and its function reflects the respective role played by each objective. 6. The last step involves the calculation of the final ranking score, σi. The final ranking score for each Pareto-optimal solution is computed by summing individual outranking elements associated with each solution as follows: M M j =1 j =1 σ i = ∑ σ [i, j ] − ∑ σ [ j , i ] The first term measures the extent to which solution i performs relative to all the other solutions in the Pareto domain, while the second term evaluates the performance of all the other solutions relative to solution i. Thereafter, the solutions are sorted from highest to lowest according to the ranking score, σi. The solution with the highest ranking is the one that best satisfies the set of preferences provided by the decision-maker. (a) (b) ck [i, j ] Dk [i, j ] 1 1 ∆ k [i, j ] ∆ k [i, j ] 0 Qk Pk 0 Qk Pk Vk Figure 6.1: (a) Individual concordance index, and (b) discordance index calculations used in NFM algorithm to determine ranking scores for the Pareto domain solutions. 127 Chapter 6: Ranking of the Pareto-optimal Solutions Finally, instead of relying on the unique solution of the Pareto domain having the best ranking score, it is preferable to use the results of NFM to divide the Pareto domain into zones containing high-ranked, mid-ranked, and low-ranked domain solutions in order to identify graphically where the optimal region is located. The decision variables associated with the preferred/selected optimal solution can then be implemented in the process. 6.2.1 Ranking of Solutions for VOC Recovery The Pareto-optimal solutions for VOC Recovery, which were obtained from the optimization for several objectives for the design case, are ranked by the NFM. Several objectives optimization and many objectives optimization gave very similar results; hence, to reduce the amount of information required from the decision maker, the results for several objectives optimization were used for this analysis. First, the solutions are presented to the decision maker in a graphical form (i.e., Figure 5.8). The decision maker would then need to decide on the weights to be given to each objective, together with the indifference, preference and veto thresholds (Table 6.1). S/he has decided to give highest priority to NPW since industries are always profit-driven. PCOP is given the second highest priority since it was noted that VOC play a large role in smog formation. The last three objectives were given equal weighting of 0.1 as their contribution to environmental impact is not very significant. The indifference, preference and veto thresholds were chosen based 10%, 20% and 80% of the optimal range of each objective. The preferred solution obtained after NFM ranking is the point shown as a black square in Figure 6.2. It provides the lowest HTP, ETP and ATMP attainable for the 128 Chapter 6: Ranking of the Pareto-optimal Solutions relatively high value of NPW; this resulted in slight compromises in NPW and PCOP (Figures 6.2a and 6.2d). The compromise in NPW and PCOP compared to the best value attainable are US $1.14×106 and 6.10 respectively. These values are close to the indifference thresholds and lesser then the preference threshold; thus, the compromise in NPW and PCOP are deemed acceptable. The rest of the points, which represent the remaining solutions, are highlighted in different shades of grey and white. The color coding identifies the solutions as the top 10%, next top 40% and last 50% solutions, according to the ranking given by NFM (see legend for details). Table 6.1: NFM Parameters for Ranking VOC Recovery Application Objective Goal Min Value Max Value Range Weights Indifference Threshold Preference Threshold Veto Threshold NPW HTP ETP ATMP PCOP Max Min Min Min Min -6.28E+06 -1.07E+06 5.21E+06 0.5 5.21E+05 1.04E+06 4.17E+06 1.24E-07 7.64E-07 6.40E-07 0.1 6.40E-08 1.28E-07 5.12E-07 3.01E-05 5.02E-05 2.01E-05 0.1 2.01E-06 4.02E-06 1.61E-05 0.20 0.34 0.14 0.1 0.01 0.03 0.11 0.53 21.41 20.88 0.2 2.09 4.18 16.71 6.2.2 Ranking of Solutions for Solvent Recovery Similarly, the Pareto-optimal solutions for Solvent Recovery, which were obtained from the optimization for many objectives, are ranked by NFM. To start with, the decision maker is presented with the Pareto-optimal solutions (i.e. Figure 5.11). The optimization of the solvent recovery for many objectives gave the following observations – the profitability measures (i.e. NPW, PBP, PBT) moved in tandem and four environmental impact categories (i.e. ATP, GWP, AP and EP) have similar Pareto fronts. Hence, one economic objective (i.e. NPW) and one of the four environmental impact categories (i.e. 129 Chapter 6: Ranking of the Pareto-optimal Solutions ATP) are chosen for ranking with the rest of the objectives. With fewer objectives to focus on, the decision maker would be able to provide the NFM parameters with more objectivity. 8 x 10 -7 6 (a) ETP HTP 6 4 2 0 -7 -5 -4 -3 -2 0.3 0.25 0.2 -7 5 4 30 (b) -6 -5 -4 -3 -2 -6 -5 -4 -3 -2 -1 -6 -5 -4 -3 -2 -1 (d) 20 10 0 -7 -1 -5 (c) 3 -7 -1 PCOP ATMP 0.35 -6 x 10 NPW (106), $ Preferred Solution NPW (106), $ Top 10% Next 40% Last 50% Figure 6.2: Ranking of Pareto-optimal Solutions by Net Flow Method for VOC Recovery Design Optimization for Several Objectives The decision maker decides on the weights to be given to each objective, together with the indifference, preference and veto thresholds (as given in Table 6.2). Since industries are naturally economically driven, NPW has been given the largest weight with the rest of the environmental objectives being given equal weight of 0.1. The values for the indifference, preference and veto thresholds were given based on 10%, 20% and 80% of optimal range of each objective. The results obtained by NFM ranking are given in Figure 6.3. The preferred solution obtained from NFM ranking is the point shown as a black square in Figure 6.3. It provides the lowest HTP and PCOP impact corresponding to a 130 Chapter 6: Ranking of the Pareto-optimal Solutions high value for NPW (Figures 6.3c and 6.3g). For the other two objectives TTP and ATP (Figures 6.3d and 6.3e), they are farther from the lowest value attainable at that value of NPW. For TTP, the difference is 0.06 which is less than the indifference threshold; for ATP, the difference is 31.7 which is close to the indifference threshold but much lower than the preference threshold. As a result, the slight tradeoff in TTP and ATP is found acceptable for the best solution given by NFM. The rest of the points, which represent the remaining solutions, are highlighted in different shades of grey and white. The legend provides the color coding which identifies the solutions as the top 10%, next top 40% and last 50% solutions, according to the ranking given by NFM. Table 6.2: NFM Parameters for Ranking Solvent Recovery Application Objective Goal Min Value Max Value Range Weights Indifference Threshold Preference Threshold Veto Threshold NPW Max -2.01E+06 2.82E+06 4.84E+06 0.6 HTP Min -8.11E+04 -1.60E+03 7.95E+04 0.1 TTP Min 9.89 10.62 0.73 0.1 ATP Min 313.67 447.54 133.87 0.1 PCOP Min 15.59 16.29 0.70 0.1 4.84E+05 9.68E+05 3.87E+06 7.95E+03 1.59E+04 6.36E+04 13.39 26.77 107.09 0.07 0.15 0.59 0.07 0.14 0.56 6.3. Conclusions In this chapter, NFM to rank Pareto-optimal solutions is described and used for ranking the Pareto-optimal solutions for both the VOCs recovery process and the solvent recovery process. The Pareto-optimal solutions from Chapter 5 have provided the decision maker an insight of the process dynamics and the resultant effect on the objective functions. Equipped with this information, the decision maker is able to give representative values for the different weights and threshold values required for ranking by NFM. Even with 131 Chapter 6: Ranking of the Pareto-optimal Solutions this said, it would not be surprising that people with different background, experience or locations may have varying opinions on the choice of values for the weights and thresholds. Nonetheless, with the inputs, NFM algorithm was successful in providing the preferred solution for the two applications studied. PBT, $/yr 5 x 10 5 200 -5 -10 -3 -2 -1 0 1 2 50 0 -3 4 x(a) 10 -1 0 1 2 3 -2 -1 0 1 2 3 -2 -1 0 1 2 3 -2 -1 0 1 2 3 11 -2 -1 0 1 2 10.5 10 9.5 -3 3 450 (c) 1100 (d) 1000 GWP 400 350 900 800 300 -3 -2 -1 0 1 2 700 -3 3 0.9 (f) 16.5 (e) (h) (g) 0.8 AP PCOP -2 (b) 5 0 -3 ATP 100 3 TTP HTP 10 150 PBP, yr 0 16 0.7 0.6 15.5 -3 -2 -1 0 1 2 3 0.5 -3 6 NPW (10 ), $ 0.12 (i) Preferred Solution 0.1 EP Top 10% Next 40% 0.08 0.06 -3 Last 50% -2 -1 0 1 2 3 6 NPW (10 ), $ Figure 6.3: Ranking of Pareto-optimal Solutions via Net Flow Method for Solvent Recovery Design for Many Objectives 132 Chapter 7: Conclusions and Recommendations Chapter 7 Conclusions and Recommendations 7.1. Conclusions As a process engineer, optimization for sustainability requires the consideration of two important areas – economic and environmental criteria. Economic objectives are well established and the choice of any one of them can be easily justified. For example, the optimization of an operation case would most probably need the evaluation of PBT, and the optimization of the design case would most probably need the evaluation of PBP and/or NPW. However, environmental objectives and the aggregation methods for them are novel and are not well established yet. A literature review of previous studies that employed environmental indices showed that most papers used an aggregate environmental index as the objective, and the process analysis is coupled with an economic objective in some papers only. Hence, feasibility and usefulness of process optimization for more than two economic and environmental objectives were studied in this research. The environmental objectives were optimized individually as much as possible along with one or more economic objectives, resulting in the optimization for several and many objectives. From the study, it was observed that the optimization for two objectives was unable to capture the entire Pareto-optimal set that could potentially be useful to the decision maker if aggregation was not used. This has been illustrated in both the case studies – VOC and solvent recovery processes. The optimization for several and many objectives has given the decision maker a larger range of Pareto-optimal solutions to 133 Chapter 7: Conclusions and Recommendations understand their trends and to choose from. Whether to optimize for several or many objectives is dependent on the particular situation. If the decision maker has already a preferred set objectives, those objectives can be optimized individually with the less important objectives aggregated into a separate objective function (e.g. among the environmental objectives in case of the VOC process, PCOP, human toxicity and ecotoxicity are deemed important, and the rest are aggregated into a separate indicator). Otherwise, if the decision maker has no preference, the environmental components can be optimized individually as separate objectives (e.g. solvent recovery process). With the myriad of Pareto-optimal solutions available, the decision maker has to choose one of them eventually for the design or operation. The use of NFM aided the decision making process whereby only the decision maker has to provide a few NFM parameters. Being equipped with the Pareto-optimal solutions facilitates the decision maker in providing the NFM parameters. From the results obtained for the two recovery processes, NFM is shown to be effective in locating the preferred Pareto-optimal solution. The present study shows the feasibility and usefulness of optimizing chemical processes for a number of economic and environmental objectives. For other case studies, if the researcher would want to optimize them to ensure that they are designed and/or operating sustainably, s/he can choose to optimize the case study for various economic and environmental objective functions. If the researcher is able to identify the critical economic or environmental objectives that are most relevant to the case study, it would reduce the number of objectives for the optimizer tool, which could reduce computational time. The use of NFM has helped us to identify the most preferred Pareto solution; 134 Chapter 7: Conclusions and Recommendations however, the different threshold values were given based on fixed percentages. Probably for future studies, the researcher can gather sentiments (whether locally or globally) to have more realistic threshold values for NFM evaluation. This would result in the preferred solution that would be more relevant in today’s context. 7.2. Recommendations for Further Study The following works are possible for future study. 1. To retrofit two columns of the solvent recovery process into a dividing-wall column (DWC) as it is known to have energy and economic savings. Thereafter, the economic and environmental evaluation and MOO can be carried out. The results from MOO may elucidate new findings. 2. The applications studied were limited to recovery processes. In future, the application chosen should be more complex, with more waste streams and the problem should be highly nonlinear. This would increase the chances for more economic and environmental objectives to be conflicting with one another. 3. The optimization of the application with different aggregation methods to determine if the aggregation method has any impact on the results. The best method could be identified, which may lead to consensus on the aggregation method in the academia field as well as in the industries. 4. For a design case study, life cycle analysis could be used to calculate the emissions during the manufacturing of grass-roots equipment. This would essentially capture all the impacts inflicted on the environment. The emissions, however, should be the actual emissions from the production of the fixed equipment and not cost dependent. 135 References References Afgan, N.H., Carvalho, M.G. and Hovanov, N.V. (2000). Energy system assessment with sustainability indicators, Energy Policy, 28, pp. 603-612. Afgan, N. H., Begić, F. and Kazagić, A. (2007). Multi-criteria sustainability assessment – A tool for evaluation of new energy system, Thermal Science, 11, No. 3, pp. 4353. Anastas, P. T. and Warner, J. C. (1998). Green Chemistry: Theory and Practice, Oxford University Press, New York. Azapagic, A., and Clift, R. (1999). Application of life cycle assessment to process optimization, Computers and Chemical Engineering, 23, pp. 1509-1526. Azapagic, A. and Perdan, S. (2000). Indicators of sustainable development for industry: A general framework, Process Safety and Environmental Protection: Transactions of the Institution of Chemical Engineers, 78, Part B, pp. 243-261. Bare, J. C., Norris, G., Pennington, D. W. and McKone, T. (2003). TRACI – The Tool for the Reduction and Assessment of Chemical and other environmental Impacts, Journal of Industrial Ecology, Vol. 6, No. 3-4, pp. 49-78. Bare, J., Gloria, T. and Norris, G. (2006). Development of the method and U.S. normalization database for life cycle impact assessment and sustainability metrics, Environmental Science & Technology, 40, pp. 5108-5115. Bartelmus, P. (2001). Unveiling Wealth: On Money, Quality of Life and Sustainability, Kluwer Academic Publishers, Netherlands. 136 References Bhaskar, V., Gupta, S. K. and Ray, A. K. (2000). Applications of multi-objective optimization in chemical engineering, Reviews in Chemical Engineering, 16, pp. 1-54. Biwer, A. and Heinzle, E. (2004). Environmental assessment in early process development, Journal of Chemical Technology and Biotechnology, Vol. 79, No. 6, pp. 597-609. Biwer, A. P., Zuber, P. T., Zelic, B., Gerharz, T., Bellmann, K. J. and Heinzle, E. (2005). Modeling and analysis of a new process for pyruvate production, Industrial and Engineering Chemistry Research, 44, pp. 3124-3133. Brown, M. T. and Herendeen, R. A. (1996). Embodied energy analysis and EMERGY analysis: a comparative view, Ecological Economics, 19, pp. 219-235. Brown, M. T. and McClanahan, T. R. (1996). EMergy analysis perspectives of Thailand and Mekong river dam proposals, Ecological Modelling, 91, pp. 105-130. Buxton, A., Livingston, A. G. and Pistikopoulos, E. N. (1999). Optimal design of solvent blends for environmental impact minimization, AIChE Journal, Vol. 45, No. 4, pp. 817-843. Cabezas, H., Bare, J. C. and Mallick, S. K. (1997). Pollution prevention with chemical process simulators: the generalized waste reduction (WAR) algorithm, Computers and Chemical Engineering, 21, pp. S305-S310. Cabezas, H., Bare, J. C. and Mallick, S. K. (1999). Pollution prevention with chemical process sinulators: the generalized waste reduction (WAR) algorithm – full version, Computers and Chemical Engineering, 23, pp. 623-634. 137 References Cabezas, H. (2007). Seminar on Environmental Management for Sustainability, presented at the Institute for Chemical and Engineering Sciences, Singapore, October 23, 2007. Cano-Ruiz, J. A. and McRae, G. J. (1998). Environmentally conscious chemical process design, Annual Review of Energy and the Environment, 23, pp. 499-536. Cardona, C. A., Marulanda, V. F. and Young, D. (2004). Analysis of the environmental impact of butylacetate process through the WAR algorithm, Chemical Engineering Science, 59, pp. 5839-5845. Chakraborty, A. and Linninger, A. A. (2002). Plant-wide waste management. 1. Synthesis and multi-objective design, Industrial and Engineering Chemistry Research, 41, pp. 4591-4604. Chen, H., Barna, B. A., Rogers, T. N. and Shonnard, D. R. (2001). A screening methodology for improved solvent selection using economic and environmental assessments, Clean Products and Processes, 3, pp. 290-302. Chen, H., Badenschier, S. M. and Shonnard, D. R. (2002a). Uncertainty analysis for toxicity assessment of chemical process designs, Industrial and Engineering Chemistry Research, 41, pp. 4440-4450. Chen, H., Wen, Y., Waters, M. D., Shonnard, D. R. (2002b). Design guidance for chemical processes using environmental and economic assessments, Industrial and Engineering Chemistry Research, 41, pp. 4503-4513. Chen, H., Rogers, T. N., Barna, B. A., and Shonnard, D. R. (2003). Automating hierarchical environmentally-conscious design using integrated software: VOC recovery case study, Environmental Process, 22, pp.147-160. 138 References Chen, H. and Shonnard, D. (2004). Systematic framework for environmentally conscious chemical process design: early and detailed design stages, Industrial and Engineering Chemistry Research, 43, pp. 535-552. Chen, Q. S. and Feng, X. (2005). A pollution reduction methodology in reactor design, Chemical Engineering and Processing, 44, pp. 13-21. Daniel, J. S., Velders, G. J. M., Douglass, A. R., Forster, P. M. D., Hauglustaine, D. A., Isaksen, I. S. A., Kuijpers, L. J. M., McCulloch, A., Wallington, T. J., Ashford, P., Montzka, S. A., Newman, P. A. and Waugh, D. W. (2007). Halocarbon scenarios, ozone depletion potentials, and global warming potentials, Scientific Assessment of Ozone Depletion: 2006, Chapter 8, World Meteorological Organization, Switzerland. Available at http://www.wmo.ch/pages/prog/arep/gaw/ozone_2006/ozone_asst_report.html. Assessed on 15 October 2007. Dantus, M. M. and High, K. A. (1999). Evaluation of waste minimization alternatives under uncertainty: a multiobjective optimization approach, Computers and Chemical Engineering, 23, pp. 1493-1508. Darton, R. (2006). Sustainable Development – A particular challenge for engineers, 11th APCChE, Kuala Lumpur, August 27-30, 2006. Darton, R. (2007). Measuring sustainability, The Chemical Engineer, 795, pp. 26-28. Deb, K. (2001). Multi-objective Optimization using Evolutionary Algorithms, John Wiley & Sons, New York. 139 References Deb, K., Pratap, A., Agarwal, S. and Meyarivan, T. (2002). A fast and elitist multiobjective genetic algorithm: NSGA-II, IEEE Transactions on Evolutionary Computation, Vol. 6, No. 2, pp. 182-197. Dietz, A., Azzaro-Pantel, C., Pibouleau, L. and Domenech, S. (2005). A framework for multiproduct batch plant design with environmental consideration: application to protein production, Industrial and Engineering Chemistry Research, 44, pp. 21912206. Dietz, A., Azzaro-Pantel, C., Pibouleau, L. and Domenech, S. (2006). Multiobjective optimization for multiproduct batch plant design under economic and environmental considerations, Computers and Chemical Engineering, 30, pp. 599-613. Dietz, A., Pantel, C. A., Pibouleau, L. G. and Domenech, S. (2007a). Ecodesign of batch processes: optimal design strategies for economic and ecological bioprocesses, International Journal of Chemical Reactor Engineering, Vol. 5, A34. Dietz, A., Azzaro-Pantel, C., Pibouleau, L. and Domenech, S. (2007b). Optimal design of batch plants under economic and ecological considerations: application to a biochemical batch plant, Mathematical and Computer Modelling, 46, pp. 109-123. Dominguez-Ramos, A., Aldaco, R. And Irabien, A. (2007). Life cycle assessment as a tool for cleaner production: application for aluminium trifluoride, International Journal of Chemical Reactor Engineering, Vol. 5, A33 Edgar, T. F., Himmelblau, D. M. and Lasdon, L. S. (2001). Optimization of Chemical Processes, McGraw-Hill, Singapore. 140 References El-Halwagi, M. M. (1997). Pollution prevention through process integration: systematic design tools, Academic Press, San Diego. Eliceche, A. M., Corvalán, S. M. and Martínez, P. (2007). Environmental life cycle impact as a tool for process optimization of a utility plant, Computers and Chemical Engineering, 31, pp. 648-656. Elkamel, A., Ba-Shammakh, M., Douglas, P. and Croiset, E. (2008). An optimization approach for integrating planning and CO2 emission reduction in the petroleum refining industry, Industrial and Engineering Chemistry Research, Vol. 7, No. 3, pp. 760-776. Fang, J., Jin, H., Ruddy, T., Pennybaker, K., Fahey, D. and Subramaniam, B. (2007). Economic and environmental impact analyses of catalytic olefin hydroformylation in CO2-expanded liquid (CXL) media, Industrial and Engineering Chemistry Research, 46, pp. 8687-8692. Flores, X., Rodriguez-Roda, I. and Poch, M. (2007). Systematic procedure to handle critical decisions during the conceptual design of activated sludge plants, Industrial and Engineering Chemistry Research, 46, pp. 5600-5613. Freitas, S. S., Santos, J. A. L. and Prazeres, M. F. (2006). Plasmid DNA. In Heinzle, E., Biwer, A. and Cooney, C. (Eds.), Development of Sustainable Bioprocesses: Modeling and Assessment (pp.155-167), John Wiley & Sons Ltd., England. Fu, Y., Diwekar, U., Young, D. and Cabezas, H. (2001). Designing processes for environmental problems. In Sikdar, S. K. and El-Halwagi, M. (Eds.), Process Design Tools for the Environment (pp. 295-317), Taylor & Francis, New York. 141 References Gadalla, M. A., Olujic, Z. Jansens, P. J., Jobson, M. and Smith, R. (2005). Reducing CO2 emissions and energy consumption of heat-integrated distillation systems, Environmental Science and Technology, 39, pp. 6860-6870. Goedkoop, M. and Spriensma, R. (2001). The Eco-indicator 99 A damage oriented method for Life Cycle Impact Assessment: Methodology Report. Available at www.pre.nl. Assessed on 3 January 2008. Halsall-Whitney, H. and Thibault, J. (2006). Multi-objective optimization for chemical processes and controller design: approximating and classifying the Pareto domain, Computers and Chemical Engineering, 30, pp. 1155-1168. Halim, I. and Srinivas, R. (2002). Integrated decision support system for waste minimization analysis in chemical processes, Environmental Science and Technology, 36, pp. 1640-1648. Harris, C. E., Pritchard, M. S. and Rabins, M. J. (2005). Engineering Ethics: Concepts & Cases, Thomson Wadsworth, USA. Heijungs, R., Guinée, J. B., Huppes, G., Lankreijer, R. M., Udo de Haes, H. A. and Wegener Sleeswijk, A. (1992). Environmental life cycle assessment of products: guide – October 1992, Centre of Environmental Science, Leiden. Heinzle, E., Biwer, A. and Cooney, C. (2006). Development of Sustainable Bioprocesses: Modeling and Assessment, John Wiley & Sons Ltd, England. Hoffmann, V. H., Hungerbühler, K. and McRae, G. J. (2001). Multiobjective screening and evaluation of chemical process technologies, Industrial and Engineering Chemistry Research, 40, pp. 4513-4524. 142 References Hoffmann, V. H., McRae, G. J. and Hungerbühler, K. (2004). Methodology for earlystage technology assessment and decision making under Uncertainty: Application to the Selection of Chemical Processes, Industrial and Engineering Chemistry Research, 43, pp. 4337-4349. Hossain, K. A., Khan, F. I. and Hawboldt, K. (2007). E-Green - A robust risk-based environmental assessment tool for process industries, Industrial and Engineering Chemistry Research, 46, pp. 8787-8795. Hugo, A., Ciumei, C., Buxton, A. and Pistikopoulos, E. N. (2004). Environmental impact minimization through material substitution: a multi-objective optimization approach, Green Chemistry, 6, pp. 407-417. Hugo, A. and Pistikopoulos, E. N. (2005). Environmentally conscious long-range planning and design of supply chain networks, Journal of Cleaner Production, 13, pp. 1471-1491. Jacobsen, N. B. (2006). Industrial symbiosis in Kalundborg, Denmark: a quantitative assessment of economic and environmental aspects, Journal of Industrial Ecology, 10, pp. 239-255. Janjira, S., Magaraphan, R. and Bagajewicz, M. J. (2007). Simultaneous treatment of environmental and financial risk in process design, International Journal of Environment and Pollution, Vol. 29, Nos. 1/2/3, pp. 30-46. Jia, X. P., Han, F. Y. and Tan, X. S. (2004). Integrated environmental performance assessment of chemical processes, Computers and Chemical Engineering, 29, pp. 243-247 143 References Jolliet, O., Margni, M., Charles, R., Humbert, S., Payet, J., Rebitzer, G. and Rosenbaum, R. (2003). IMPACT 2002+: A new life cycle impact assessment methodology, The International Journal of Life Cycle Assessment, Vol. 8, No. 6, pp. 324-330. Khan, F. I., Natrajan, B. R. and Revathi, P. (2001). GreenPro: a new methodology for cleaner and greener process design, Journal of Loss Prevention in the Process Industries, 14, pp. 307-328. Khan, F. I., Sadiq, R. and Husain, T. (2002). GreenPro-I: a risk-based life cycle assessment and decision-making methodology for process plant design, Environmental Modelling & Software, 17, pp. 669-692. Kheawhom, S. and Hirao, M. (2002). Decision support tools for process design and selection, Computers and Chemical Engineering, 26, pp. 747-755. Kheawhom, S. and Hirao, M. (2004). Decision support tools for environmentally benign process design under uncertainty, Computers and Chemical Engineering, 28, pp. 1715-1723. Kholiq, M. A. and Heinzle, E. (2006). Recombinant human serum albumin. In Heinzle, E., Biwer, A. and Cooney, C. (Eds.), Development of Sustainable Bioprocesses: Modeling and Assessment (pp.155-167), John Wiley & Sons Ltd., England. Kim, K. J. and Diwekar, U. M. (2002). Integrated solvent selection and recycling for continuous processes, Industrial and Engineering Chemistry Research, 41, pp. 4479-4488. Kim, K. J. and Smith, R. L. (2004). Parallel multiobjective evolutionary algorithms for waste solvent recycling, Industrial and Engineering Chemistry Research, 43, pp. 2669-2679. 144 References Kim, K. J. and Smith, R. L. (2005). Systematic procedure for designing processes with multiple environmental objectives, Environmental Science and Technology, 39, pp. 2394-2405. Knoll, A. and Buechs, J. (2006). L-Lysine – Coupling of bioreaction and process model. In Heinzle, E., Biwer, A. and Cooney, C. (Eds.), Development of Sustainable Bioprocesses: Modeling and Assessment (pp.155-167), John Wiley & Sons Ltd., England. Koller, G., Weirich, D., Brogli, F., Heinzle, E., Hoffmann, V. H., Verduyn, M. A. and Hungerbühler, K. (1998). Ecological and economic objective functions for screening in integrated development of fine chemical processes. 2. Stream allocation and case studies, Industrial and Engineering Chemistry Research, 37, pp. 3408-3413. Krajnc, D. and Glavič, P. (2005). How to compare companies on relevant dimensions of sustainability, Ecological Economics, 55, pp. 551-563. Krotscheck, C. and Narodoslawsky, M. (1996). The sustainable process index: A new dimension in ecological evaluation, Ecological Engineering, 6, pp. 241-258. Lange, J-P (2002). Sustainable development: efficiency and recycling in chemicals manufacturing, Green Chemistry, 4, pp. 546-550. Lapkin, A., Joyce, L. and Crittenden, B. (2004). Framework for evaluating the “greenness” of chemical processes: case studies for a novel VOC recovery technology, Environmental Science and Technology, 38, pp. 5815-5823. Lim, Y. I., Floquet, P. And Joulia, X. (1999). Multiobjective optimization in terms of economics and potential environment impact for process design and analysis in a 145 References chemical process simulator, Industrial and Engineering Chemistry Research, 38, pp. 4729-4741. Lim, Y. I., Floquet, P. And Joulia, X. (2001). Efficient implementation of the normal boundary intersection (NBI) method on multiobjective optimization problems, Industrial and Engineering Chemistry Research, 40, pp. 648-655. Lou, H. H., Kulkarni, M. A., Singh A. and Hopper, J. R. (2004). Sustainability assessment of industrial systems, Industrial and Engineering Chemistry Research, 43, pp. 4233-4242. Mallick, S. K., Cabezas, H., Bare, J. C. and Sikdar, S. K. (1996). A pollution reduction methodology for chemical process simulators, Industry and Engineering Chemistry Research, 35, pp. 4128-4138. Marteel, A. E., Davies, J. A., Olson, W. W. and Abraham, M. A. (2003). Green chemistry and engineering: drivers, metrics and reduction to practice, Annual Review Environmental Resource, 28, pp. 401-428. Martins, A. A., Mata, T. M., Costa, C. A. V. and Sikdar, S. K. (2007). Framework for sustainability metrics, Industrial and Engineering Chemistry Research, 46, pp. 2962-2973. Masters, G. M. (1998). Introduction to Environmental Engineering and Science, Prentice-Hall International Inc., New Jersey. Masuduzzaman and Rangaiah, G. P. (2008). Multi-objective optimization applications in chemical engineering. In Rangaiah, G. P. (Ed.), Multi-objective Optimization: Techniques and Applications in Chemical Engineering, World Scientific, Singapore, In Press. 146 References Mayer, A.L., Thurston, H.W. and Pawlowski, C.W. (2004). The multidisciplinary influence of common sustainable indices, Frontiers in ecology and the environment, 2(8), pp. 419-426. Miettinen, K. (1999). onlinear Multiobjective Optimization, Kluwer, Boston. Narodoslawsky, M. and Krotscheck, C. (2000). Integrated ecological optimization of processes with the sustainable process index, Waste Management, 20, pp. 599-603. Niederl-Schmidinger, A. and Narodoslawsky, M. (2008). Life cycle assessment as an engineer’s tool?, Journal of Cleaner Production, 16, pp. 245-252 Parkinson, G. (2005). Distillation: New wrinkles for an age-old technology, Chemical Engineering Progress, Vol., 101, No. 7, pp. 10-12. Parkinson, G. (2007). Dividing-wall column find greater appeal, Chemical Engineering Progress, Vol. 103, No. 5, pp. 8-11. Petrides, D. (2006). Recombinant human insulin. In Heinzle, E., Biwer, A. and Cooney, C. (Eds.), Development of Sustainable Bioprocesses: Modeling and Assessment (pp.155-167), John Wiley & Sons Ltd., England. Pintarič, Z. N. and Kravanja, Z. (2006). Selection of the economic objective function for the optimization of process flow sheets, Industrial and Engineering Chemistry Research, 45, pp. 4222-4232. Ramzan, N. and Witt, W. (2006). Multi-objective optimization in distillation unit: a case study, The Canadian Journal of Chemical Engineering, 84, pp. 604-613. Ramzan, N., Degenkolbe, S. and Witt, W. (2007). Evaluating and improving environmental performance of HC’s recovery system: A case study of distillation unit, Chemical Engineering Journal, 140, pp. 201-213. 147 References Rangaiah, G. P. (2009). Multi-objective Optimization: Techniques and Applications in Chemical Engineering, World Scientific, Singapore, In Press. Sarkar, A. U. and Karagöz, S. (1995). Sustainable development of hydroelectric power, Energy, Vol. 20, No. 10, pp. 977-981. Schwarz, J., Beloff, B. and Beaver, E. (2002). Use sustainability metrics to guide decision-making, Chemical Engineering Progress, 98(7), pp. 58-63. Seider, W. D., Seader, J. D. and Lewin, D. R. (2004). Product and Process Design Principles: Synthesis, Analysis and Evaluation, John Wiley and Sons Inc., United States of America. Sharratt, P. (1999). Environmental criteria in design, Computers and Chemical Engineering, 23, pp. 1469-1475. Sheldon, R. A. (2007). The E factor: fifteen years on, Green Chemistry, 9, pp. 1273-1283. Shonnard, D. R. and Hiew, D. S. (2000). Comparative environmental assessments of VOC recovery and recycle design alternatives for a gaseous waste stream, Environmental Science & Technology, 34, pp. 5222-5228. Sikdar, S. K. (2003a). Sustainable development and dustainability metrics, AIChE Journal, Vol. 49, No. 8, pp. 1928-1932. Sikdar, S. K. (2003b). Journey towards sustainable development: a role for chemical engineers, Environmental Progress, Vol. 22, No. 4, pp. 227-232. Singh, A. and Lou, H. H. (2006). Hierarchical Pareto optimization for the sustainable development of industrial ecosystems, Industry and Engineering Chemistry Research, 45, pp. 3265-3279. 148 References Smith, R. L. (2004). Hierarchical design and evaluation of processes to generate wasterecycled feeds, Industrial and Engineering Chemistry Research, 43, pp. 25082515. Smith, R. L., Mata, T. M., Young, D. M., Cabezas, H. and Costa, C. A. V. (2004). Designing environmentally friendly chemical processes with fugitive and open emissions, Journal of Cleaner Production, 12, pp.125-129. Song, J., Park, H., Lee, D-Y and Park, S. (2002). Scheduling of actual size refinery processes considering environmental impacts with multiobjective optimization, Industrial and Engineering Chemistry Research, 41, pp. 4794-4806. Stefanis, S. K., Livingston, A. G. and Pistikopoulos, E. N. (1995). Minimizing the environmental impact of process plants: a process systems methodology, Computers and Chemical Engineering, 19, Suppl., pp. S39-S44. Stefanis, S. K., Buxton, A., Livingston, A. G. and Pistikopoulos, E. N. (1996). A methodology for environmental impact minimization: solvent design and reaction path synthesis issues, Computers and Chemical Engineering, 20, pp. S1419S1424. Stefanis, S. K., Livingston, A. G. and Pistikopoulos, E. N. (1997a). Environmental impact considerations in the optimal design and scheduling of batch processes, Computers and Chemical Engineering, Vol. 21, No. 10, pp. 1073-1094. Stefanis, S. K. and Pistikopoulos, E. N. (1997b). Methodology for environmental risk assessment of industrial nonroutine releases, Industrial and Engineering Chemistry Research, 36, pp. 3694-3707. 149 References Steffens, M. A., Fraga, E. S. And Bogle, I. D. L. (1999). Multicriteria process synthesis for generating sustainable and economic bioprocesses, Computers and Chemical Engineering, 23, pp. 1455-1467. Storhas, W. and Metz, R. (2006). Riboflavin – Vitamin B2. In Heinzle, E., Biwer, A. and Cooney, C. (Eds.), Development of Sustainable Bioprocesses: Modeling and Assessment (pp.155-167), John Wiley & Sons Ltd., England. Thibault, J., Lanouette, R., Fonteix, C. and Kiss, L. N. (2002). Multicriteria optimization of a high-yield pulping process, The Canadian Journal of Chemical Engineering, Vol. 80, pp. 897-902. Thibault, J. (2008). Net flow and rough sets: two methods for ranking the Pareto domain. In Rangaiah, G. P. (Ed.), Multi-objective Optimization: Techniques and Applications in Chemical Engineering, World Scientific, Singapore, In Press. Trost, B. M. (1991). The atom economy: a search for synthetic efficiency, Science, 254(5037), pp. 1471–1477 Turton, R., Bailie, R. C., Whiting, W. B. and Shaeiwitz, J. A. (2003). Analysis, Synthesis and Design of Chemical Processes, Prentice Hall, New Jersey. Villadsen, J. (2007). Innovative technology to meet the demands of the white biotechnology revolution of chemical production, Chemical Engineering Science, 62, pp. 6957-6968. Vincent, R., Bonthoux, F., Mallet, G., Iparraguirre, J. F. and Rio, S. (2005). Méthodologie d’évaluation simplifiée du risque chimique: un outil d’aide à la décision, I RS Hygiène Sécurité du Travail, pp. 39-62 150 References Wen, Y. and Shonnard, D. R. Environmental and economic assessments of heat exchanger networks for optimum minimum approach temperature, Computers and Chemical Engineering, 27, pp. 1577-1590. World Commission on Environment and Development (1987). Our common future, Oxford University Press, Oxford. Wuebbles, D. J. (1995). Weighing functions for ozone depletion and greenhouse effects on climate, Annual Review of Energy and Environment, 20, pp. 45-70. Young, D. M. and Cabezas, H. (1999). Designing sustainable processes with simulation: the waste reduction (WAR) algorithm, Computers and Chemical Engineering, 23, pp. 1477-1491. Young, D., Scharp, R. and Cabezas, H. (2000). The waste reduction (WAR) algorithm: environmental impacts, energy consumption, and engineering economics. Waste Management, 20, pp. 605-615. Zapalac, E. and McDonald, K. (2006). a-1-Antitrypsin from transgenic plant cell suspension cultures. In Heinzle, E., Biwer, A. and Cooney, C. (Eds.), Development of Sustainable Bioprocesses: Modeling and Assessment (pp.155167), John Wiley & Sons Ltd., England. Zhang, X., Li, C., Fu, C. and Zhang, S. (2008). Environmental impact assessment of chemical process using the Green Degree method, Industrial and Engineering Chemistry Research, 47, pp. 1085–1094. 151 Appendix A Appendix A Interface used for MOO A.1. Excel, Visual Basic for Applications and HYSYS Interface To enable MOO, a custom-made Excel-Visual Basic for Applications-HYSYS interface has been developed (Figure A.1). This interface combines the process simulation capability of HYSYS with the mathematical computation and spreadsheet features of Microsoft Excel by linking the object libraries of these two applications using Visual Basic for Applications (VBA). EXCEL (User Interface, Objectives, Constraints) Object Library VBA (NSGA-II) Object Library HYSYSTM (Process Simulation) Figure A.1: Excel-VBA-HYSYS Setup for MOO of Processes The tool used for MOO is the elitist non-dominated sorting generic algorithm, NSGA-II (Figure A.2). It has been coded in VBA where Excel is used as the user interface for the user to launch the NSGA-II program. In the Excel file, the user is required to key in the NSGA-II parameters, the bounds for the decision variables, the calculation for the objective functions and constraints. As HYSYS is used for simulating the recovery processes in this project, VBA is required to link up the object library of HYSYS for the evaluation of the objective functions (see boxes with asterisk in Figure A.2). The decision variables generated by 152 Appendix A NSGA-II would be transferred to HYSYS where the simulation would be executed until convergence is reached. Thereafter, the results from HYSYS are transferred to Excel where the objective functions and constraints are evaluated. NSGA-II would use the results to rank the solutions. The VBA code written to link up the HYSYS object library for VOC recovery process is given below. Public Sub RunningHysys(NoDV As Integer) Sheets("From Hysys").Range("B3:E6").ClearContents Sheets("From Hysys").Range("B10:I23").ClearContents Sheets("From Hysys").Range("B27:D31").ClearContents Sheets("From Hysys").Range("A36:N95").ClearContents Sheets("From Hysys").Range("A98:N125").ClearContents Set hyApp = CreateObject("Hysys.Application") Currentpath = ThisWorkbook.Path Set hyCase = hyApp.SimulationCases.Open(Currentpath & "\VOC_REC_HEATINT_3.HSC") 'hyCase.Visible = True Dim StreamRange As Range, HeatExRange As Range, PumpRange As Range, ColRange As Range, DVRange As Range Dim Emission As ProcessStream, CoolN2 As ProcessStream, EAcetatePdt As ProcessStream, ToluenePdt As ProcessStream, SoltoAbs As ProcessStream, MakeupSol As ProcessStream Dim FeedCooler As HeatExchanger, SolCooler As HeatExchanger, PdtCooler As HeatExchanger, DFeedHeater As HeatExchanger, HeatInt As HeatExchanger Dim DistillCol As ColumnFlowsheet, PdtDisCol As ColumnFlowsheet, AbsCol As ColumnFlowsheet Dim DistillTS As TraySection, PdtDisTS As TraySection, AbsTS As TraySection Dim SolPump As PumpOp, DFeedPump As PumpOp 'PdtPump As PumpOp, Dim FlashD As Separator Dim EmFlow() As Double, CN2Flow() As Double, EAPdtFlow() As Double, TolPdtFlow() As Double, VapourFlow() As Double, LiquidFlow() As Double, MUSol() As Double, DVValue() As Double Dim Index1 As Integer, NoStage As Integer ReDim DVValue(NoDV) As Double Set StreamRange = Sheets("From Hysys").Range("B3") Set HeatExRange = Sheets("From Hysys").Range("B10") Set PumpRange = Sheets("From Hysys").Range("B27") Set ColRange = Sheets("From Hysys").Range("B36") Set DVRange = Sheets("VOC Recovery").Range("C3") Set Emission = hyCase.Flowsheet.MaterialStreams("Emission from Absorber") Set CoolN2 = hyCase.Flowsheet.MaterialStreams("Cool N2") Set EAcetatePdt = hyCase.Flowsheet.MaterialStreams("E-Acetate") Set ToluenePdt = hyCase.Flowsheet.MaterialStreams("Toluene") Set FeedCooler = hyCase.Flowsheet.Operations.Item("Feed Cooler") Set SolCooler = hyCase.Flowsheet.Operations.Item("Solvent Cooler") Set PdtCooler = hyCase.Flowsheet.Operations.Item("Product Cooler") Set DistillCol = hyCase.Flowsheet.Operations("ColumnOp").Item("T-101").ColumnFlowsheet Set PdtDisCol = hyCase.Flowsheet.Operations("ColumnOp").Item("T-104").ColumnFlowsheet Set DFeedHeater = hyCase.Flowsheet.Operations.Item("DisFeed Heater") Set SolPump = hyCase.Flowsheet.Operations("PumpOp").Item("Solvent Pump") 'Set PdtPump = hyCase.Flowsheet.Operations("PumpOp").Item("Product Pump") Set AbsCol = hyCase.Flowsheet.Operations("ColumnOp").Item("T-100").ColumnFlowsheet Set DistillTS = DistillCol.Operations.Item("Main TS") Set PdtDisTS = PdtDisCol.Operations.Item("Main TS") Set AbsTS = AbsCol.Operations.Item("TS-1") Set FlashD = hyCase.Flowsheet.Operations.Item("V-100") 153 Appendix A Set SoltoAbs = hyCase.Flowsheet.MaterialStreams("Solvent to Absorber") Set MakeupSol = hyCase.Flowsheet.MaterialStreams("Make-up Solvent") Set HeatInt = hyCase.Flowsheet.Operations.Item("Heat Integration") Set DFeedPump = hyCase.Flowsheet.Operations("PumpOp").Item("DFeed Pump") For Index1 = 1 To NoDV DVValue(Index1) = DVRange.Offset(Index1 - 1, 0).Value Next Index1 SoltoAbs.MolarFlow = DVValue(1) / 3600 SolCooler.ShellSideProduct.Temperature = DVValue(2) FeedCooler.ShellSideProduct.Temperature = DVValue(3) PdtCooler.ShellSideProduct.Temperature = DVValue(5) 'FeedCooler.TubeSideFeed.Temperature = (DVValue(6) - 32) * 5 / 9 'SolCooler.TubeSideFeed.Temperature = (DVValue(7) - 32) * 5 / 9 'PdtCooler.TubeSideFeed.Temperature = (DVValue(8) - 32) * 5 / 9 DFeedHeater.TubeSideProduct.Temperature = ClearContents DFeedHeater.TubeSideProduct.Temperature = DVValue(4) SolCooler.TubeLength = 4 SolCooler.TubeLength = 4.25 'Original Value EmFlow = Emission.ComponentMassFlow.GetValues("kg/h") CN2Flow = CoolN2.ComponentMassFlow.GetValues("kg/h") EAPdtFlow = EAcetatePdt.ComponentMassFlow.GetValues("kg/h") TolPdtFlow = ToluenePdt.ComponentMassFlow.GetValues("kg/h") MUSol = MakeupSol.ComponentMassFlow.GetValues("kg/h") ' Stream Data For Index1 = 1 To 4 StreamRange.Offset(Index1 - 1, 0).Value = EmFlow(Index1 - 1) StreamRange.Offset(Index1 - 1, 1).Value = CN2Flow(Index1 - 1) StreamRange.Offset(Index1 - 1, 2).Value = EAPdtFlow(Index1 - 1) StreamRange.Offset(Index1 - 1, 3).Value = TolPdtFlow(Index1 - 1) StreamRange.Offset(Index1 - 1, 4).Value = MUSol(Index1 - 1) Next Index1 ' For Coolers HeatExRange.Offset(0, 0).Value = FeedCooler.ShellSideFeed.Temperature.GetValue("C") HeatExRange.Offset(1, 0).Value = FeedCooler.ShellSideProduct.Temperature.GetValue("C") HeatExRange.Offset(2, 0).Value = FeedCooler.TubeSideFeed.Temperature.GetValue("C") HeatExRange.Offset(3, 0).Value = FeedCooler.TubeSideProduct.Temperature.GetValue("C") HeatExRange.Offset(4, 0).Value = FeedCooler.TubeSideFeed.MassFlow.GetValue("kg/h") HeatExRange.Offset(8, 0).Value = FeedCooler.UA.GetValue("kJ/C-h") HeatExRange.Offset(9, 0).Value = FeedCooler.HeatTransferArea.GetValue("m2") HeatExRange.Offset(10, 0).Value = FeedCooler.Duty.GetValue("kJ/h") HeatExRange.Offset(11, 0).Value = FeedCooler.ShellSideFeed.Pressure.GetValue("bar_g") HeatExRange.Offset(12, 0).Value = FeedCooler.TubeSideFeed.Pressure.GetValue("bar_g") HeatExRange.Offset(13, 0).Value = FeedCooler.FtFactor.GetValue HeatExRange.Offset(0, 1).Value = SolCooler.ShellSideFeed.Temperature.GetValue("C") HeatExRange.Offset(1, 1).Value = SolCooler.ShellSideProduct.Temperature.GetValue("C") HeatExRange.Offset(2, 1).Value = SolCooler.TubeSideFeed.Temperature.GetValue("C") HeatExRange.Offset(3, 1).Value = SolCooler.TubeSideProduct.Temperature.GetValue("C") HeatExRange.Offset(4, 1).Value = SolCooler.TubeSideFeed.MassFlow.GetValue("kg/h") HeatExRange.Offset(8, 1).Value = SolCooler.UA.GetValue("kJ/C-h") HeatExRange.Offset(9, 1).Value = SolCooler.HeatTransferArea.GetValue("m2") HeatExRange.Offset(10, 1).Value = SolCooler.Duty.GetValue("kJ/h") HeatExRange.Offset(11, 1).Value = SolCooler.ShellSideFeed.Pressure.GetValue("bar_g") HeatExRange.Offset(12, 1).Value = SolCooler.TubeSideFeed.Pressure.GetValue("bar_g") HeatExRange.Offset(13, 1).Value = SolCooler.FtFactor.GetValue HeatExRange.Offset(0, 2).Value = PdtCooler.ShellSideFeed.Temperature.GetValue("C") HeatExRange.Offset(1, 2).Value = PdtCooler.ShellSideProduct.Temperature.GetValue("C") HeatExRange.Offset(2, 2).Value = PdtCooler.TubeSideFeed.Temperature.GetValue("C") HeatExRange.Offset(3, 2).Value = PdtCooler.TubeSideProduct.Temperature.GetValue("C") HeatExRange.Offset(4, 2).Value = PdtCooler.TubeSideFeed.MassFlow.GetValue("kg/h") HeatExRange.Offset(8, 2).Value = PdtCooler.UA.GetValue("kJ/C-h") HeatExRange.Offset(9, 2).Value = PdtCooler.HeatTransferArea.GetValue("m2") 154 Appendix A HeatExRange.Offset(10, 2).Value = PdtCooler.Duty.GetValue("kJ/h") HeatExRange.Offset(11, 2).Value = PdtCooler.ShellSideFeed.Pressure.GetValue("bar_g") HeatExRange.Offset(12, 2).Value = PdtCooler.TubeSideFeed.Pressure.GetValue("bar_g") HeatExRange.Offset(13, 2).Value = PdtCooler.FtFactor.GetValue HeatExRange.Offset(0, 3).Value = DistillCol.MaterialStreams.Item("To Condenser").Temperature.GetValue("C") HeatExRange.Offset(1, 3).Value = DistillCol.MaterialStreams.Item("Reflux").Temperature.GetValue("C") HeatExRange.Offset(10, 3).Value = DistillCol.EnergyStreams.Item("CondEn1").HeatFlow.GetValue("kJ/h") HeatExRange.Offset(11, 3).Value = DistillCol.MaterialStreams.Item("To Condenser").Pressure.GetValue("bar_g") HeatExRange.Offset(0, 4).Value = PdtDisCol.MaterialStreams.Item("To Condenser").Temperature.GetValue("C") HeatExRange.Offset(1, 4).Value = PdtDisCol.MaterialStreams.Item("Reflux").Temperature.GetValue("C") HeatExRange.Offset(10, 4).Value = PdtDisCol.EnergyStreams.Item("CondEn2").HeatFlow.GetValue("kJ/h") HeatExRange.Offset(11, 4).Value = PdtDisCol.MaterialStreams.Item("To Condenser").Pressure.GetValue("bar_g") ' For Heaters HeatExRange.Offset(0, 5).Value = DFeedHeater.TubeSideFeed.Temperature.GetValue("C") HeatExRange.Offset(1, 5).Value = DFeedHeater.TubeSideProduct.Temperature.GetValue("C") HeatExRange.Offset(5, 5).Value = DFeedHeater.ShellSideFeed.Temperature.GetValue("C") HeatExRange.Offset(6, 5).Value = DFeedHeater.ShellSideProduct.Temperature.GetValue("C") HeatExRange.Offset(7, 5).Value = DFeedHeater.ShellSideFeed.MassFlow.GetValue("kg/h") HeatExRange.Offset(8, 5).Value = DFeedHeater.UA.GetValue("kJ/C-h") HeatExRange.Offset(9, 5).Value = DFeedHeater.HeatTransferArea.GetValue("m2") HeatExRange.Offset(10, 5).Value = DFeedHeater.Duty.GetValue("kJ/h") HeatExRange.Offset(11, 5).Value = DFeedHeater.ShellSideFeed.Pressure.GetValue("bar_g") HeatExRange.Offset(12, 5).Value = DFeedHeater.TubeSideFeed.Pressure.GetValue("bar_g") HeatExRange.Offset(13, 5).Value = DFeedHeater.FtFactor.GetValue HeatExRange.Offset(0, 6).Value = DistillCol.MaterialStreams.Item("To Reboiler").Temperature.GetValue("C") HeatExRange.Offset(1, 6).Value = DistillCol.MaterialStreams.Item("Boilup").Temperature.GetValue("C") HeatExRange.Offset(10, 6).Value = DistillCol.EnergyStreams.Item("RebEn1").HeatFlow.GetValue("kJ/h") HeatExRange.Offset(11, 6).Value = DistillCol.MaterialStreams.Item("To Reboiler").Pressure.GetValue("bar_g") HeatExRange.Offset(0, 7).Value = PdtDisCol.MaterialStreams.Item("To Reboiler").Temperature.GetValue("C") HeatExRange.Offset(1, 7).Value = PdtDisCol.MaterialStreams.Item("Boilup").Temperature.GetValue("C") HeatExRange.Offset(10, 7).Value = PdtDisCol.EnergyStreams.Item("RebEn2").HeatFlow.GetValue("kJ/h") HeatExRange.Offset(11, 7).Value = PdtDisCol.MaterialStreams.Item("To Reboiler").Pressure.GetValue("bar_g") HeatExRange.Offset(0, 8).Value = HeatInt.ShellSideFeed.Temperature.GetValue("C") HeatExRange.Offset(1, 8).Value = HeatInt.ShellSideProduct.Temperature.GetValue("C") HeatExRange.Offset(2, 8).Value = HeatInt.TubeSideFeed.Temperature.GetValue("C") HeatExRange.Offset(3, 8).Value = HeatInt.TubeSideProduct.Temperature.GetValue("C") HeatExRange.Offset(8, 8).Value = HeatInt.UA.GetValue("kJ/C-h") HeatExRange.Offset(9, 8).Value = HeatInt.HeatTransferArea.GetValue("m2") HeatExRange.Offset(10, 8).Value = HeatInt.Duty.GetValue("kJ/h") HeatExRange.Offset(11, 8).Value = HeatInt.ShellSideFeed.Pressure.GetValue("bar_g") HeatExRange.Offset(12, 8).Value = HeatInt.TubeSideFeed.Pressure.GetValue("bar_g") HeatExRange.Offset(13, 8).Value = HeatInt.FtFactor.GetValue ' For Pumps PumpRange.Offset(0, 0).Value = SolPump.FeedPressure.GetValue("kPa") PumpRange.Offset(1, 0).Value = SolPump.ProductPressure.GetValue("kPa") PumpRange.Offset(2, 0).Value = SolPump.FeedStream.ActualVolumeFlow.GetValue("m3/h") PumpRange.Offset(3, 0).Value = SolPump.FeedStream.MassDensity.GetValue("lb/ft3") PumpRange.Offset(4, 0).Value = SolPump.EnergyStream.HeatFlow.GetValue("kJ/h") 'PumpRange.Offset(0, 1).Value = PdtPump.FeedPressure.GetValue("kPa") 'PumpRange.Offset(1, 1).Value = PdtPump.ProductPressure.GetValue("kPa") 'PumpRange.Offset(2, 1).Value = PdtPump.FeedStream.ActualVolumeFlow.GetValue("m3/h") 'PumpRange.Offset(3, 1).Value = PdtPump.FeedStream.MassDensity.GetValue("lb/ft3") 'PumpRange.Offset(4, 1).Value = PdtPump.EnergyStream.HeatFlow.GetValue("kJ/h") PumpRange.Offset(0, 2).Value = DFeedPump.FeedPressure.GetValue("kPa") PumpRange.Offset(1, 2).Value = DFeedPump.ProductPressure.GetValue("kPa") PumpRange.Offset(2, 2).Value = DFeedPump.FeedStream.ActualVolumeFlow.GetValue("m3/h") PumpRange.Offset(3, 2).Value = DFeedPump.FeedStream.MassDensity.GetValue("lb/ft3") PumpRange.Offset(4, 2).Value = DFeedPump.EnergyStream.HeatFlow.GetValue("kJ/h") ' For Columns NoStage = AbsTS.NumberOfStages 155 Appendix A VapourFlow = AbsCol.NetMassVapourFlows.GetValues("kg/s") LiquidFlow = AbsCol.NetMassLiquidFlows.GetValues("kg/s") ColRange.Offset(0, 2).Value = AbsCol.MaterialStreams("Solvent to Absorber").MassDensity.GetValue("kg/m3") ColRange.Offset(NoStage - 1, 2).Value = AbsCol.MaterialStreams("To Distill Feed Heater").MassDensity.GetValue("kg/m3") ColRange.Offset(0, 3).Value = AbsCol.MaterialStreams("Emission from Absorber").MassDensity.GetValue("kg/m3") ColRange.Offset(NoStage - 1, 3).Value = AbsCol.MaterialStreams("Cool Feed").MassDensity.GetValue("kg/m3") ColRange.Offset(0, 4).Value = AbsCol.MaterialStreams("Solvent to Absorber").SurfaceTension.GetValue("dyn/cm") ColRange.Offset(NoStage - 1, 4).Value = AbsCol.MaterialStreams("To Distill Feed Heater").SurfaceTension.GetValue("dyn/cm") ColRange.Offset(0, 5).Value = AbsTS.TopStagePressure.GetValue("bar") ColRange.Offset(NoStage - 1, 5).Value = AbsTS.BottomStagePressure.GetValue("bar") For Index1 = 1 To NoStage ColRange.Offset(Index1 - 1, -1).Value = Index1 ColRange.Offset(Index1 - 1, 0).Value = VapourFlow(Index1 - 1) ColRange.Offset(Index1 - 1, 1).Value = LiquidFlow(Index1 - 1) Next Index1 NoStage = DistillTS.NumberOfStages VapourFlow = DistillCol.NetMassVapourFlows.GetValues("kg/s") LiquidFlow = DistillCol.NetMassLiquidFlows.GetValues("kg/s") ColRange.Offset(1, 9).Value = DistillCol.MaterialStreams("Reflux").MassDensity.GetValue("kg/m3") ColRange.Offset(NoStage, 9).Value = DistillCol.MaterialStreams("To Reboiler").MassDensity.GetValue("kg/m3") ColRange.Offset(1, 10).Value = DistillCol.MaterialStreams("To Condenser").MassDensity.GetValue("kg/m3") ColRange.Offset(NoStage, 10).Value = DistillCol.MaterialStreams("Boilup").MassDensity.GetValue("kg/m3") ColRange.Offset(1, 11).Value = DistillCol.MaterialStreams("Reflux").SurfaceTension.GetValue("dyn/cm") ColRange.Offset(NoStage, 11).Value = DistillCol.MaterialStreams("To Reboiler").SurfaceTension.GetValue("dyn/cm") ColRange.Offset(1, 12).Value = DistillTS.TopStagePressure.GetValue("bar") ColRange.Offset(NoStage, 12).Value = DistillTS.BottomStagePressure.GetValue("bar") For Index1 = 0 To NoStage + 1 If Index1 = NoStage + 1 Then ColRange.Offset(59, 6).Value = Index1 ColRange.Offset(59, 7).Value = VapourFlow(Index1) ColRange.Offset(59, 8).Value = LiquidFlow(Index1) Else ColRange.Offset(Index1, 6).Value = Index1 ColRange.Offset(Index1, 7).Value = VapourFlow(Index1) ColRange.Offset(Index1, 8).Value = LiquidFlow(Index1) End If Next Index1 NoStage = PdtDisTS.NumberOfStages VapourFlow = PdtDisCol.NetMassVapourFlows.GetValues("kg/s") LiquidFlow = PdtDisCol.NetMassLiquidFlows.GetValues("kg/s") ColRange.Offset(1 + 62, 2).Value = PdtDisCol.MaterialStreams("Reflux").MassDensity.GetValue("kg/m3") ColRange.Offset(NoStage + 62, 2).Value = PdtDisCol.MaterialStreams("To Reboiler").MassDensity.GetValue("kg/m3") ColRange.Offset(1 + 62, 3).Value = PdtDisCol.MaterialStreams("To Condenser").MassDensity.GetValue("kg/m3") ColRange.Offset(NoStage + 62, 3).Value = PdtDisCol.MaterialStreams("Boilup").MassDensity.GetValue("kg/m3") ColRange.Offset(1 + 62, 4).Value = PdtDisCol.MaterialStreams("Reflux").SurfaceTension.GetValue("dyn/cm") ColRange.Offset(NoStage + 62, 4).Value = PdtDisCol.MaterialStreams("To Reboiler").SurfaceTension.GetValue("dyn/cm") ColRange.Offset(1 + 62, 5).Value = PdtDisTS.TopStagePressure.GetValue("bar") ColRange.Offset(NoStage + 62, 5).Value = PdtDisTS.BottomStagePressure.GetValue("bar") For Index1 = 0 To NoStage + 1 If Index1 = NoStage + 1 Then ColRange.Offset(89, -1).Value = Index1 ColRange.Offset(89, 0).Value = VapourFlow(Index1) ColRange.Offset(89, 1).Value = LiquidFlow(Index1) Else ColRange.Offset(Index1 + 62, -1).Value = Index1 ColRange.Offset(Index1 + 62, 0).Value = VapourFlow(Index1) ColRange.Offset(Index1 + 62, 1).Value = LiquidFlow(Index1) End If Next Index1 156 Appendix A ColRange.Offset(62, 7).Value = FlashD.VapourProduct.MassDensity.GetValue("kg/m3") ColRange.Offset(62, 8).Value = FlashD.LiquidProduct.MassDensity.GetValue("kg/m3") ColRange.Offset(62, 9).Value = FlashD.VapourMassFlow.GetValue("kg/s") ColRange.Offset(62, 10).Value = FlashD.VapourProduct.Pressure.GetValue("bar") ColRange.Offset(62, 11).Value = FlashD.LiquidMassFlow.GetValue("kg/s") End Sub 157 Appendix A Set gen =0 PP: Generate parent (initial) population, pop, randomly. Evaluate objective functions and constraints* Classify and calculate Irank and Idist of chromosomes in PP BP: Copy better chromosomes DP: Do crossover and mutation on BP Evaluate objective functions and constraints* Combine PP and DP (2 pop) Classify the population into fronts EP: Select the best pop from this set (Elitism) Is gen < gen,max? No Yes = gen + 1 Replace PP by EP gen Stop Figure A.2: Flowchart for NSGA-II implemented in VBA; (*) indicate steps which require VBA and HYSY interface 158 Appendix B Appendix B et Flow Method (Macros) B.1. et Flow Method in Visual Basic for Applications In Chapter 6, Pareto-optimal solutions were being ranked by Net Flow Method (NFM) to identify the most preferred solution. The choice is made using the NFM parameters provided by the decision maker and the NFM methodology itself. Required information from the decision maker are (1) the Pareto-optimal solutions, and (2) the NFM parameters, which are provided in the Excel spreadsheet from which VBA will extract from. For this project, NFM methodology is being coded in Visual Basic for Applications (VBA) as shown below. Sub NetFlow() Dim NoObj As Integer, NoDV As Integer, NoCons As Integer, PopSize As Integer Dim Index1 As Integer, Index2 As Integer, Index3 As Integer Dim Diff() As Double, ConI() As Double, Weight() As Double, Indiff() As Double, Pref() As Double, Veto() As Double Dim GloCon() As Double, DisI() As Double, OutRank() As Double, RankScore() As Double Dim Summa As Double, Summa1 As Double, Summa2 As Double Dim ParetoRange As Range, InfoRange As Range, CheckRes As Range, ParResRange As Range Set ParResRange = ActiveWorkbook.Sheets("Results from MOO").Range("B3") Set ParetoRange = ActiveWorkbook.Sheets("Pareto Data").Range("B3") Set InfoRange = ActiveWorkbook.Sheets("Information Required").Range("C12") Set CheckRes = ActiveWorkbook.Sheets("Results").Range("B5") NoObj = ActiveWorkbook.Sheets("Information Required").Range("C2").Value NoDV = ActiveWorkbook.Sheets("Information Required").Range("C3").Value NoCons = ActiveWorkbook.Sheets("Information Required").Range("C4").Value PopSize = ActiveWorkbook.Sheets("Information Required").Range("C5").Value ReDim Diff(NoObj, PopSize, PopSize) As Double, ConI(NoObj, PopSize, PopSize) As Double, Weight(NoObj) As Double, Indiff(NoObj) As Double, Pref(NoObj) As Double, Veto(NoObj) As Double ReDim GloCon(PopSize, PopSize) As Double, DisI(NoObj, PopSize, PopSize) As Double, OutRank(PopSize, PopSize) As Double, RankScore(PopSize) As Double ' Copy solutions from Results from MOO to Pareto Data For Index1 = 1 To NoObj + NoDV + NoCons + 3 ParetoRange.Offset(-2, Index1 - 1).Value = ParResRange.Offset(-2, Index1 - 1).Value ParetoRange.Offset(-1, Index1 - 1).Value = ParResRange.Offset(-1, Index1 - 1).Value For Index2 = 1 To PopSize If Index1 > NoDV Then If Index1 NoDV + NoObj Then ParetoRange.Offset(Index2 - 1, Index1 - 1).Value = ParResRange.Offset(Index2 - 1, Index1 - 1).Value End If Next Index2 Next Index1 For Index1 = 1 To NoObj Weight(Index1) = InfoRange.Offset(0, Index1 - 1).Value Indiff(Index1) = InfoRange.Offset(1, Index1 - 1).Value Pref(Index1) = InfoRange.Offset(2, Index1 - 1).Value Veto(Index1) = InfoRange.Offset(3, Index1 - 1).Value 'MsgBox Weight(Index1) & " " & Indiff(Index1) & " " & Pref(Index1) & " " & Veto(Index1) Next Index1 ' To obtain difference For Index1 = 1 To NoObj For Index2 = 1 To PopSize ' i is the indication For Index3 = 1 To PopSize ' j is the indication If InfoRange.Offset(-1, Index1 - 1).Value = "Min" Then Diff(Index1, Index2, Index3) = ParetoRange.Offset(Index2 ParetoRange.Offset(Index3 - 1, NoDV + Index1 - 1).Value Else Diff(Index1, Index2, Index3) = ParetoRange.Offset(Index3 ParetoRange.Offset(Index2 - 1, NoDV + Index1 - 1).Value End If Next Index3 Next Index2 Next Index1 - 1, NoDV + Index1 - 1).Value - - 1, NoDV + Index1 - 1).Value - ' To obtain individual concordance index For Index1 = 1 To NoObj For Index2 = 1 To PopSize ' i is the indication For Index3 = 1 To PopSize ' j is the indication If Diff(Index1, Index2, Index3) Pref(Index1) Then ConI(Index1, Index2, Index3) = 0 Else ConI(Index1, Index2, Index3) = (Pref(Index1) - Diff(Index1, Index2, Index3)) / (Pref(Index1) - Indiff(Index1)) End If End If 'MsgBox Index1 & " i = " & Index2 & " j = " & Index3 'MsgBox Diff(Index1, Index2, Index3) & " " & ConI(Index1, Index2, Index3) Next Index3 Next Index2 Next Index1 ' To obtain global concordance index For Index2 = 1 To PopSize ' i is the indication For Index3 = 1 To PopSize ' j is the indication Summa = 0 For Index1 = 1 To NoObj Summa = Summa + Weight(Index1) * ConI(Index1, Index2, Index3) Next Index1 160 Appendix B GloCon(Index2, Index3) = Summa Next Index3 Next Index2 ' To obtain discordance index For Index1 = 1 To NoObj For Index2 = 1 To PopSize ' i is the indication For Index3 = 1 To PopSize ' j is the indication If Diff(Index1, Index2, Index3) Veto(Index1) Then DisI(Index1, Index2, Index3) = 1 Else DisI(Index1, Index2, Index3) = (Diff(Index1, Index2, Index3) - Pref(Index1)) / (Veto(Index1) - Pref(Index1)) End If End If Next Index3 Next Index2 Next Index1 ' To obtain outranking matrix For Index2 = 1 To PopSize ' i is the indication For Index3 = 1 To PopSize ' j is the indication Summa = 1 For Index1 = 1 To NoObj Summa = Summa * (1 - (DisI(Index1, Index2, Index3) ^ 3)) Next Index1 OutRank(Index2, Index3) = GloCon(Index2, Index3) * Summa Next Index3 Next Index2 ' To obtain final ranking score For Index2 = 1 To PopSize ' i is the indication Summa1 = 0 Summa2 = 0 For Index3 = 1 To PopSize ' j is the indication Summa1 = Summa1 + OutRank(Index2, Index3) Summa2 = Summa2 + OutRank(Index3, Index2) Next Index3 RankScore(Index2) = Summa1 - Summa2 ParetoRange.Offset(Index2 - 1, NoObj + NoDV + NoCons + 4).Value = RankScore(Index2) Next Index2 ' To sort solutions based on ranking score ActiveWorkbook.Worksheets("Pareto Data").Sort.SortFields.Clear ActiveWorkbook.Worksheets("Pareto Data").Sort.SortFields.Add Key:=Range( _ ParetoRange.Offset(0, NoObj + NoDV + NoCons + 4), ParetoRange.Offset(PopSize - 1, NoObj + NoDV + NoCons + 4)), _ SortOn:=xlSortOnValues, Order:=xlDescending, DataOption:=xlSortNormal With ActiveWorkbook.Worksheets("Pareto Data").Sort .SetRange Range(ParetoRange, ParetoRange.Offset(PopSize - 1, NoObj + NoDV + NoCons + 4)) .Header = xlGuess .MatchCase = False .Orientation = xlTopToBottom .SortMethod = xlPinYin .Apply End With Application.ScreenUpdating = True End Sub 161 Appendix B B.2. FM Applied to Some Examples Prior to employing NFM to the recovery processes discussed in the project, it is first applied to applications (i.e. Williams and Otto or alkylation processes) or benchmark test problems. These are used for NFM trial runs since their Pareto-optimal solutions have been made available from the testing of the NSGA-II tool which was used as the optimization tool. For the Williams and Otto (WO) process, it has been optimized using two economic objectives – NPW and PBP. In the design stage of the WO process, though both of the economic objectives are indicative of the profitability of the project, NPW is known to be a more popular choice as it considers the time value of money. Hence, a higher weight of 0.8 was given to NPW and thus PBP has a weight of 0.2. The indifference, preference and veto thresholds are given as 10%, 40% and 80% of the range over which the Pareto-optimal solutions spanned for each objective. The ‘best’ solution generated by NFM is one with high value of NPW with a reasonable value for PBP (see Figure B.1). This illustrates the usefulness of NFM as a tool for ranking Pareto-optimal solutions. Next, for the alkylation process, it was optimized using one economic objective (i.e. Profit) and one product quality indicator (i.e. octane number, ON). The range over which the octane number spanned is small and it is also assumed that any Pareto-optimal solution has met the minimum specification set for alkylated products. As a result, a higher weight of 0.8 was given to Profit while ON has a weight of 0.2. The indifference, preference and veto thresholds are given as 10%, 40% and 80% of the range over which 162 Appendix B the Pareto-optimal solutions spanned for each objective. NFM has generated the ‘best’ solution with high value of Profit with a reasonable ON value (see Figure B.2). Define NFM parameters Run NFM Max/Min Relative Weight Indifference Threshold Preference Threshold Veto Threshold Clear NFM NPW Max 0.8 0.35 1.4 2.8 PBP Min 0.2 0.1 0.4 0.8 Payback Period (yr) Williams and Otto Process 2.2 2 1.8 1.6 1.4 1.2 1 3 4 5 6 7 8 Net Present Worth (106 $) Remaining Points Top 50% Top 10% Preferred Solution Figure B.1: Net Flow Method for Williams and Otto Process Define NFM parameters Run NFM Clear NFM Profit Max/Min Max Relative Weight 0.8 Indifference Threshold 86.862 Preference Threshold 347.448 Veto Threshold 694.896 ON Max 0.2 0.167 0.668 1.336 Alkylation Process Octane Number 95.5 95.1 94.7 94.3 93.9 93.5 600 800 1000 1200 1400 1600 Profit ($/day) Remaining Points Top 50% Top 10% Preferred Solution Figure B.2: Net Flow Method for Alkylation Process 163 Appendix B For all the benchmark problems (i.e. KUR, SCH, ZDT2, ZDT4, ZDT6, CONSTR, SRN, TNK), equal weightages were given to each objective. As for the threshold values, 10%, 40% and 80% of the range over which the Pareto-optimal solutions spanned. The chosen solution provided desirable results as shown in Figures B.3 to B.10. In some of these figures, the top 10% solutions are at both ends. Since equal weightage was given to both objectives, it would mean that both objectives are equally important. The Pareto-optimal solutions here generally form a convex curve. If we start from the corner point, and move towards the centre of the Pareto-optimal solutions, we would notice that in order to improve one objective marginally, the other objective has to be sacrificed to a greater extent. Hence, the corner solutions are the most preferred solutions above the centre solutions. Overall, it is noted that if the Pareto-optimal solutions form a convex curve (if both objectives are minimized), the corner Pareto-optimal solutions are preferred. On the other hand, if the Pareto-optimal solutions form a concave curve, then the centre Paretooptimal solutions are preferred. 164 Appendix B Define NFM parameters Run NFM Max/Min Relative Weight Indifference Threshold Preference Threshold Veto Threshold Clear NFM f1 Min 0.5 0.56 2.23 4.45 f2 Min 0.5 1.16 4.65 9.30 KUR Problem -19 -18 -17 -16 -15 f2 -20 1 -1 -3 -14 -5 -7 -9 -11 -13 f1 Remaining Points Top 50% Top 10% Preferred Solution Figure B.3: Net Flow Method for KUR Benchmark Problem Define NFM parameters Run NFM Max/Min Relative Weight Indifference Threshold Preference Threshold Veto Threshold Clear NFM f1 Min 0.5 0.40 1.60 3.20 f2 Min 0.5 0.40 1.60 3.20 SCH Problem f2 4 3.5 3 2.5 2 1.5 1 0.5 0 0 0.5 1 1.5 2 2.5 3 3.5 4 f1 Remaining Points Top 50% Top 10% Preferred Solution Figure B.4: Net Flow Method for SCH Benchmark Problem 165 Appendix B Define NFM parameters Run NFM Clear NFM Max/Min Relative Weight Indifference Threshold Preference Threshold Veto Threshold f1 Min 0.5 0.100 0.399 0.798 f2 Min 0.5 0.099 0.395 0.790 f2 ZDT2 Problem 1.2 1 0.8 0.6 0.4 0.2 0 0 0.2 0.4 0.6 0.8 1 1.2 f1 Remaining Points Top 50% Top 10% Preferred Solution Figure B.5: Net Flow Method for ZDT2 Benchmark Problem Define NFM parameters Run NFM Clear NFM Max/Min Relative Weight Indifference Threshold Preference Threshold Veto Threshold f1 Min 0.5 0.100 0.400 0.800 f2 Min 0.5 0.100 0.400 0.800 f2 ZDT4 Problem 1.2 1 0.8 0.6 0.4 0.2 0 0 0.2 0.4 0.6 0.8 1 1.2 f1 Remaining Points Top 50% Top 10% Preferred Solution Figure B.6: Net Flow Method for ZDT4 Benchmark Problem 166 Appendix B Define NFM parameters Run NFM Max/Min Relative Weight Indifference Threshold Preference Threshold Veto Threshold Clear NFM f1 Min 0.5 0.072 0.288 0.575 f2 Min 0.5 0.092 0.368 0.737 f2 ZDT6 Problem 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0.2 0.4 0.6 0.8 1 1.2 f1 Remaining Points Top 50% Top 10% Preferred Solution Figure B.7: Net Flow Method for ZDT6 Benchmark Problem Define NFM parameters Run NFM Max/Min Relative Weight Indifference Threshold Preference Threshold Veto Threshold Clear NFM f1 Min 0.5 0.061 0.244 0.489 f2 Min 0.5 0.799 3.198 6.396 CONSTR Problem 10 8 f2 6 4 2 0 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 f1 Remaining Points Top 50% Top 10% Preferred Solution Figure B.8: Net Flow Method for CONSTR Benchmark Problem 167 Appendix B f1 Max/Min Min Relative Weight 0.5 Indifference Threshold 21.376 Preference Threshold 85.503 Veto Threshold 171.006 Define NFM parameters Run NFM Clear NFM f2 Min 0.5 21.977 87.908 175.817 SRN Problem 50 100 150 200 250 f2 50 0 -50 0 -100 -150 -200 -250 f1 Remaining Points Top 50% Top 10% Preferred Solution Figure B.9: Net Flow Method for SRN Benchmark Problem Define NFM parameters Run NFM Clear NFM Max/Min Relative Weight Indifference Threshold Preference Threshold Veto Threshold f1 Min 0.5 0.099 0.396 0.792 f2 Min 0.5 0.099 0.394 0.788 f2 TNK Problem 1.2 1 0.8 0.6 0.4 0.2 0 0 0.2 0.4 0.6 0.8 1 1.2 f1 Remaining Points Top 50% Top 10% Preferred Solution Figure B.10: Net Flow Method for TNK Benchmark Problem 168 [...]... Design Optimization of VOC Recovery for NPW and PEI 102 Figure 5.7: Selected Results for Design Optimization of VOC Recovery for Ten Objectives 106 Figure 5.8: Optimal Objective Values for Design Optimization of VOC Recovery for Five Objectives 109 Figure 5.9: Sequences 1 and 2 from Chakraborty and Linninger (2002) 111 Figure 5.10: Selected Results for Design Optimization of Solvent Recovery for NPW and. .. VOC Recovery Process Flowsheet 88 Figure 5.2: Excel-VBA-HYSYS Setup for MOO of Processes 91 Figure 5.3: Selected Results for Operation Optimization of VOC Recovery for PBT and PEI 95 Figure 5.4: Selected Results for Operation Optimization of VOC Recovery for Eight Objectives 97 Figure 5.5: Selected Results for Operation Optimization of VOC Recovery for Five Objectives 99 Figure 5.6: Selected Results for. .. Pharmaceutical and Chemicals 66 Table 4.3: Economic and Environmental Criteria – Downstream Processing 71 Table 4.4: Economic and Environmental Criteria – Energy Systems and Heat 76 Integration Table 4.5: Environmental Criteria – Petrochemicals 80 Table 4.6: Environmental Criteria – Biotechnology, Pharmaceutical and Chemicals 83 Table 4.7: Environmental Criteria – Downstream Processing 85 Table 4.8: Environmental. .. management and administration activities, and not directly to the production process 2.2.1.2 Profitability Criteria Several profitability criteria are available and the choice of criteria to be used for evaluation of a project is dependent on the engineer/company In analyzing the profitability of a process that is already online, capital costs would have been predetermined and already incurred, and so... employed for this purpose In this study, feasibility and usefulness of considering several economic and environmental objectives are investigated For this, two case studies are chosen: a VOC (volatile organic component) recovery system (Chen et al., 2003) and a solvent recovery system (Chakraborty and Linninger, 2002) Seven groups were identified for environmental impacts – HTP, ETP, GWP, ODP, PCOP, AP, and. .. and the choice of a profitability measure may be easy; even then, one may like to consider more than one profitability measure On the other hand, the choice of environmental performance indicator requires more care Most of the reported works have employed aggregated indicators, providing a final environmental performance index There are, however, many contributing factors for the environmental performance... Results for Design Optimization of Solvent Recovery for Ten Objectives 120 Figure 6.1: (a) Individual concordance index, and (b) discordance index calculations used in NFM algorithm to determine ranking scores for the Pareto domain solutions 127 Figure 6.2: Ranking of Pareto-optimal Solutions by Net Flow Method for VOC Recovery Design Optimization for Several Objectives 130 xiii Figure 6.3: Ranking of Pareto-optimal... Introduction 1.1 Optimization of Chemical Processes Optimization refers to finding one or more feasible solutions which correspond to the maximum and/ or minimum of one or more objectives The need to find such optimal solutions in a problem comes mostly from the purpose of designing and operating a plant for minimum fixed capital cost and/ or operating cost, for maximum reliability, and others As a result, optimization. .. depended on environmentalists and/ or the government; it requires the awareness of every individual and their efforts to realize this laudable objective There are three spheres of sustainability: economic development, environmental stewardship and societal equity (e.g., Azapagic and Perdan, 2000; Sikdar, 2003a and 2003b; Heinzle et al., 2006) This is often touted as the “triple bottom line” Economic 4... objectives available and they are briefly discussed below Very often, as chemical industries are profit-driven, the objective functions are economics-related They can be material metrics or profitability measures Material metrics are ratios that measure the efficiency of the chemical process – e.g amount of product per unit of feed, amount of waste emitted per unit of product On the other hand, profitability ... Setup for MOO of Processes 91 Figure 5.3: Selected Results for Operation Optimization of VOC Recovery for PBT and PEI 95 Figure 5.4: Selected Results for Operation Optimization of VOC Recovery for. .. Results for Operation Optimization of VOC Recovery for Five Objectives 99 Figure 5.6: Selected Results for Design Optimization of VOC Recovery for NPW and PEI 102 Figure 5.7: Selected Results for. .. Studied for Economic and Environmental Criteria 57 4.1 Introduction 57 4.2 Economic and Environmental Criteria 58 4.2.1 Petroleum Refining and Petrochemicals 58 4.2.2 Biotechnology, Pharmaceutical and