Thông tin tài liệu
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
Ngày đăng: 16/10/2015, 15:38
Xem thêm: Optimization of recovery processes for multiple economic and environmental criteria , Optimization of recovery processes for multiple economic and environmental criteria