Process Systems Engineering,

9. Domain Engineering
IOANNIS G. ECONOMOU, The Petroleum Institute, Department of Chemical Engineering,
Abu Dhabi, United Arab Emirates
EFSTRATIOS N. PISTIKOPOULOS, Imperial College London, Department of Chemical
Engineering, London, United Kingdom
JEN-PEI LIU, Imperial College London, Department of Chemical Engineering, London,
United Kingdom
YOSHIAKI KAWAJIRI, Georgia Institute of Technology, School of Chemical & Biomo-
lecular Engineering, Georgia I, Atlanta, United States
KRIST V. GERNAEY, Center for Process Engineering and Technology, Department of
Chemical and Biochemical Engineering, Technical University of Denmark, Lyngby,
Denmark
JOHN M. WOODLEY, Center for Process Engineering and Technology, Department of
Chemical and Biochemical Engineering, Technical University of Denmark, Lyngby,
Denmark
CONCEPCIÓN JIMÉNEZ-GONZÁLEZ, GlaxoSmithKline, Research Triangle ParkNorth
Carolina, United States
RENÉ BAÑARES-ALCÁNTARA, University of Oxford, Department of Engineering Science,
Oxford, United Kingdom

1. Molecular Modeling and Simulation for 2.3.2. Example 2–Hydrogen Infrastructure

Chemical Product and Process Planning . . . . . . . . . . . . . . . . . . . . . . 15
Design . . . . . . . . . . . . . . . . . . . . . . . . . 2 2.3.3. Example 3–Energy Systems in
1.1. Introduction. . . . . . . . . . . . . . . . . . . 2 Commercial Buildings . . . . . . . . . . . . 17
1.2. Elementary Statistical Mechanics . . 3 2.4. Conclusions and Future Directions . 18
1.3. Major Molecular Simulation Methods 3 3. Pharmaceutical Processes . . . . . . . . . 19
1.3.1. Molecular Dynamics (MD) . . . . . . . . 3 3.1. Introduction. . . . . . . . . . . . . . . . . . . 19
1.3.2. Metropolis Monte Carlo Simulation . . 4 3.2. Pharmaceutical Process Development
1.4. Applications . . . . . . . . . . . . . . . . . . . 4 and Operation . . . . . . . . . . . . . . . . . 20
1.4.1. Pharmaceuticals. . . . . . . . . . . . . . . . . 4 3.2.1. Crystallization . . . . . . . . . . . . . . . . . . 21
1.4.2. Polymer Membranes for Gas Separation 6 3.2.2. Chromatography . . . . . . . . . . . . . . . . 22
1.4.3. Ionic Liquids for Sustainable Chemical 3.3. Conclusion . . . . . . . . . . . . . . . . . . . . 25
Processes . . . . . . . . . . . . . . . . . . . . . 8 4. Biochemical Engineering . . . . . . . . . . 25
1.5. Conclusions . . . . . . . . . . . . . . . . . . . 9 4.1. Introduction. . . . . . . . . . . . . . . . . . . 25
2. Energy Systems Engineering . . . . . . . 10 4.2. Industrial Biotechnology Processes . 26
2.1. Introduction. . . . . . . . . . . . . . . . . . . 10 4.2.1. Fermentation Processes . . . . . . . . . . . 26
2.2. Methods/Tools/Algorithm . . . . . . . . 10 4.2.2. Microbial Catalysis . . . . . . . . . . . . . . 27
2.2.1. Superstructure-Based Modeling . . . . . 10 4.2.3. Enzyme Processes . . . . . . . . . . . . . . . 27
2.2.2. Mixed-Integer Programming (MIP) . . . 11 4.3. Modeling of Bioprocesses. . . . . . . . . 28
2.2.3. Multiobjective Optimization. . . . . . . . 11 4.3.1. Modeling of Bioprocesses–
2.2.4. Optimization under Uncertainty . . . . . 11 Mechanistic Models. . . . . . . . . . . . . . 28
2.2.5. Life-Cycle Assessment . . . . . . . . . . . 12 4.3.2. Modeling of Bioprocesses–Data-
2.3. Energy Systems Examples . . . . . . . . 12 Driven Models . . . . . . . . . . . . . . . . . 29
2.3.1. Example 1–Polygeneration Energy 4.4. The Role of Process Systems
Systems . . . . . . . . . . . . . . . . . . . . . . 12 Engineering . . . . . . . . . . . . . . . . . . . 30

2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

10.1002/14356007.o22_o13
2 Process Systems Engineering, 9. Domain Engineering

4.4.1. Evaluation of Process Options . . . . . . 30 5.7.4. Uncertainty . . . . . . . . . . . . . . . . . . . . 40

4.4.2. Evaluation of Platform Chemicals . . . 30 5.7.5. Emergent Behavior . . . . . . . . . . . . . . 40
4.4.3. Process Integration . . . . . . . . . . . . . . 31 5.7.6. Complexity of Causation . . . . . . . . . . 40
4.4.4. Biorefinery Design. . . . . . . . . . . . . . . 31 5.7.7. Objectivity in Policy Analysis . . . . . . 40
4.4.5. Biocatalyst Design. . . . . . . . . . . . . . . 31 5.8. Types of Models Used in the Analysis of
4.5. Assessing the Sustainability Policies. . . . . . . . . . . . . . . . . . . . . . . 41
of Bioprocesses. . . . . . . . . . . . . . . . . 32 5.8.1. Macroeconomic Models (Mainstream,
4.5.1. Life-Cycle Inventory and Assessment. 33 Descriptive, Aggregated, Mechanistic) 41
4.6. Future Outlook and Perspectives. . . 35 5.8.2. Optimization Models (Mainstream,
5. Policies and Policy Making . . . . . . . . 36 Normative, Aggregated, Mechanistic). 42
5.1. Introduction. . . . . . . . . . . . . . . . . . . 36 5.8.3. Control Models (Mainstream, Normative,
5.2. Policies and Policy Measures . . . . . . 36 Aggregated, Mechanistic) . . . . . . . . . 42
5.3. Policy Making and the Systems 5.8.4. Data-Based Models . . . . . . . . . . . . . . 42
Approach . . . . . . . . . . . . . . . . . . . . . 36 5.8.5. Game Theory (Descriptive) . . . . . . . . 42
5.4. Similarities between Policy Formulation 5.8.6. System Dynamics (Aggregated,
and Conceptual Process Design . . . . 37 Mechanistic) . . . . . . . . . . . . . . . . . . . 42
5.5. The Nature of Policy Formulation . . 38 5.8.7. Network Theory (Descriptive) . . . . . . 42
5.6. The Nature of Sociotechnical Systems 39 5.8.8. Agent-Based Approaches . . . . . . . . . . 43
5.7. Challenges for Modelers of 5.8.9. Some Conclusions on Models for the
Sociotechnical Systems. . . . . . . . . . . 39 Analysis of Policies . . . . . . . . . . . . . . 43
5.7.1. Multiple Stakeholders . . . . . . . . . . . . 39 5.9. Synthesis of Policies . . . . . . . . . . . . . 43
5.7.2. Incommensurable Values . . . . . . . . . . 39 5.10. Future Directions . . . . . . . . . . . . . . 44
5.7.3. Externalities . . . . . . . . . . . . . . . . . . . 40

1. Molecular Modeling and These are nontrivial changes that require con-
Simulation for Chemical Product certed effort at different levels: basic research to
and Process Design develop fundamental knowledge of physical phe-
nomena, applied research to develop physical
1.1. Introduction models and parameters, and development work
for the generation of new processes that meet the
Major chemical process industries (CPI) have requirements stated above. Accurate simulation
experienced a substantial transformation in re- and optimization methodologies are necessary at
cent years worldwide due to an increased com- all length and time scales, from the submolecular
petition at a global level and a significant pres- level all the way to the macroscopic level where
sure from national governments and interna- thermodynamic and computational fluid mechan-
tional organizations to develop new sustainable ics models together with advanced numerical
processes that consume significantly smaller methods are used in a concerted way. A consistent
quantities of energy and other natural resources hierarchical development of physical models is of
and operate under zero (or close to zero) waste outmost importance.
production. This chapter refers to the development of
In parallel, major multinational CPI shifted molecular modeling and simulation methods for
from low value commodity products to special- the design of new chemical products and the
ty products of high added value where the improvement of existing and design of new
underlined materials are of considerably higher processes. Molecular simulation was intro-
complexity in terms of: duced in the 1950s [1, 2] (! Molecular Model-
ing, Chapter 3) as an abstract physical applica-
. Chemical structure tion for the primitive computers of the time and
. Molecular and supramolecular architecture it evolved to a powerful engineering tool more
. Micro- and mesostructure than 50 years later. At the same time, the need
. Performance in the end-use environment for further development of simulation methods
 Process Systems Engineering, 9. Domain Engineering 3

and physically accurate models remains as it along the way in order that Equation (3) can
will be seen later in this chapter. lead to meaningful results [4].
Alternatively, one may calculate a macro-
scopic property P as a statistical average over
1.2. Elementary Statistical Mechanics all microstates of the system, that is:
Z
In statistical mechanics (! Molecular Dynam- dpN drN PðrN pN Þexp½
HðrN pN Þ/kB T
ics (MD) Simulation), the properties of a bulk hPi ¼ Z ð4Þ
dpN drN exp½
HðrN pN Þ/kB T
chemical system are calculated based on the
collective interactions between the molecules Even then, calculation of hPi using brute
that make up the system. Almost all of the force numerical integration requires extraordi-
systems of interest to process systems engineer- nary computing power. For example, for a 100
ing (PSE) follow Boltzmann statistics and so the molecule system using Simpson’s rule with just
partition function (Q) of a system of constant 5 equidistant points along each coordinate axis
number of molecules (N) in a specific volume one needs to evaluate the integrand of Equa-
(V) and temperature (T) is [3]: tion (4) at 10210 points [5].
Z
Q¼c dpN drN exp½
HðrN pN Þ/kB T ð1Þ A much more efficient approach is based on
the observation that some configurations of the
where rN and pN denote the coordinates and molecular system are much more important than
momenta of all N molecules, HðrN pN Þ is the others, so one should focus on sampling these
Hamiltonian of the system and c is a proportion- important configurations rather than random
ality constant. For a system of N identical (indis- configurations. This has been the basis of the
tinguishable) molecules: c ¼ 1/ðh3N N!Þ where h so-called Metropolis Monte Carlo simulation
is the Planck’s constant. The Hamiltonian pro- method discussed briefly below.
vides the total energy of the system as a function
of the coordinates and the momenta of the mo-
lecules and is given as the sum of the kinetic 1.3. Major Molecular Simulation
energy (K) and the potential energy (U), so that: Methods
X
HðrN pN Þ ¼ p2i /ð2mi ÞþUðrN Þ ð2Þ 1.3.1. Molecular Dynamics (MD)
i

The potential energy U depends strongly on In classical (Newtonian) mechanics, the follow-
the nature (complexity) of molecular interac- ing set of equations describes the evolution of
tions [3]. Intermolecular potentials range from the system over time [6]:
primitive potentials (such as hard sphere, square mi r€i ¼ fi
well, etc.) to potentials of moderate complexity i ¼ l; . . . ; N ð5Þ
fi ¼
rri UðrN Þ
(such as Lennard–Jones, Stockmayer, etc.) and
all the way to complex potentials that account where mi is the mass of molecule i and fi is the
for intra- and intermolecular interactions, many force exerted on it. MD consists of solving these
body effects (polarizable potentials), etc. N second order differential equations numeri-
From the partition function, one may calcu- cally using a number of different methods de-
late macroscopic thermodynamic properties veloped for this purpose. In this way, MD allows
using the so-called bridge function, which for monitoring of the evolution of the system with
the case of the constant NVT system (canonical time, and thus, time-dependent structure (poly-
statistical ensemble) is [3]: mer chain relaxation, etc.) and physical proper-
ties (such as viscosity, diffusion coefficient,
AðNVTÞ ¼
kB T ln QðNVTÞ ð3Þ etc.) can be calculated. MD is usually performed
in the microcanonical (NVE) statistical ensem-
where A is the Helmholtz free energy. Unfortu- ble; however, the method has been extended to
nately, the partition function Q can be calculated canonical (NVT), isobaric-isothermal (NPT)
analytically only for a very few simple systems and other statistical ensembles [6]. An impor-
and significant approximations are needed tant parameter concerning the robustness of the
4 Process Systems Engineering, 9. Domain Engineering

MD simulation is the time step used for the where DU ¼ U(new)
U(old) is the energy
numerical integration of the equations of mo- difference between the old and the new config-
tion. For systems characterized by a relatively uration. Thermodynamic properties are calcu-
stiff potential (e.g., the case of chain molecules), lated based on Equation (4). Additional moves
a typical time step is in the order of 0.1–1 fs. A in the NPT, mVT and other ensembles include
number of advanced simulation techniques al- volume fluctuation, random insertion and dele-
low the use of different time steps for different tion of particles and so on, and acceptance
types of forces. For example, a short time step is criteria are modified accordingly.
used for fast varying forces, such as bond A major breakthrough in molecular simulation
stretching and bond angle bending and a longer was the development of the Gibbs ensemble MC
time step is used for slowing varying forces, (GEMC) method which allows the simultaneous
such as nonbonded intra- and intermolecular simulation of several phases in equilibrium (e.g.,
interactions. Using state of the art computing vapor–liquid equilibrium) [8]. The method has
facilities, one may simulate a real system today been successfully applied to pure components,
(April 2010) for up to a few micro seconds. This binary and multicomponent mixtures, and differ-
is sufficient for the calculation of properties ent types of phase equilibria (vapor–liquid, liq-
such as chemical potential and self-diffusion uid–liquid, vapor–liquid–liquid, etc.) [9].
coefficient in systems that consist of small- and Development of efficient elementary moves
medium-size molecules. For the calculation for long chain molecules has also been a very
of dynamic properties of long chain molecules active area of research over the last two decades.
(e.g., polymers with a molecular mass higher A broad range of moves has been proposed for
than 10000), alternative methods are needed [7]. the efficient relaxation of chain tails, internal
segments, branch points, and even moves that
allow exchange of molecular segments between
1.3.2. Metropolis Monte Carlo two different chain molecules [10]. A combina-
Simulation tion of these moves allows today accurate sim-
ulation of polymer melts with a molecular mass
Metropolis Monte Carlo (MMC) simulation is a of the order of several thousand.
stochastic method that allows efficient sampling
of the multidimensional phase space of the
system. In other words, this method allows 1.4. Applications
‘‘jumps’’ in the phase space and so, no real time
monitoring of the system is possible. In MMC, 1.4.1. Pharmaceuticals
the different states of the system are visited with
a probability proportional to the Boltzmann Hydration energy plays a significant role in
factor of the energy of the system [5]. The system biological processes and is currently an impor-
goes from one configuration (state) to the next tant predictive index for molecule availability in
configuration (state) based on different types of the pharmaceutical industry. During the com-
moves that satisfy microscopic reversibility and plex process of driving a molecule from an
preserve the macroscopic properties of the sys- aqueous phase to a target protein active site,
tem that are set constant. In this way, MC the driving force is directly related to the differ-
simulations are performed in the NVT, grand ence between the hydration energy of the drug
canonical (mVT), NPT and many other statistical and the drug–protein association energy. More-
ensembles, depending on the system (pure fluid over, desolvation of both protein site and drug
or mixture) and conditions (one phase, two, or molecule occurs during this binding process,
more phases, etc.) examined. In a typical NVT and recently developed docking/scoring meth-
MMC simulation, particles are displaced ran- ods estimate this desolvation correction based
domly one at a time within the simulation box on free energy calculations. For some drug
and the new configuration is accepted or rejected molecules, solvation free energies may be esti-
according to the Boltzmann factor of the energy mated experimentally from concentration mea-
difference between the two states, that is: surements in two-phase systems. However, in
most cases this is not possible and so accurate
pNVT ¼ min½1; expð
DU/ðkB TÞÞ ð6Þ
theoretical or computational approaches are
 Process Systems Engineering, 9. Domain Engineering 5

needed. Molecular simulation using realistic

Dhyd G ¼ Dvacuum G
Dwater G
Ddummy G ¼ Dvacuum G
Dwater G ð7Þ
potential models is able to provide accurate
estimate of the property of interest and at the The term Dvacuum G contains only contribu-
same time a quantitative insight regarding tions from intramolecular nonbonded interac-
the molecular mechanisms associated with the tions (forces acting between atoms in the same
hydration. molecule separated by more than three bonds),
Recently, a simple thermodynamic cycle was which exist in the solute molecule but not in the
proposed to calculate the hydration Gibbs free dummy molecule. The thermodynamic integra-
energy, DhydG (P,T), of complex solute mole- tion approach was used by [11] to calculate
cules [11]: DhydG of barbituric acid and various substituted
Solute ðwaterÞ Dwater G
barbiturates at 298 K and 0.1 MPa. In Figure 1,
Dummy ðwaterÞ






! DhydG and the Lennard–Jones contribution to it,
Dhyd G" #Ddummy G
DLJG, are presented as a function of molecular
Solute ðvacuumÞ Dvacuum G Dummy ðvacuumÞ mass for various mono- and di-substituted






!
barbiturates.
Using the same methodology, DhydG of vari-
where, Dwater G is the Gibbs energy associated ous n-alkanes [12] and polar compounds [13]
with the mutation of the solute molecules into were calculated by [11]. An extensive evalua-
molecules of dummy atoms (atoms that do not tion of three widely used molecular models
interact with their environment) in water, (force fields) to describe the polar compounds,
Dvacuum G is the Gibbs energy associated with namely TraPPE, Gromos and OPLS-AA, was
the same process in vacuum, and finally performed. In all cases, the MSCP/E model was
Ddummy G can be seen as the hypothetical hydra- used for water. An overview of the predictions
tion Gibbs energy of dummy species. In prac- obtained from the different force fields for the
tice, these atoms have no intermolecular elec- polar compounds and a comparison with exper-
trostatic or van der Waals interactions, but imental data is shown in Figure 2. For the
their intramolecular bonded interactions are the relatively simple polar molecules, such as meth-
same as in the solute atoms. As a consequence, anol and propanol, all force field predictions are
Ddummy G is equal to zero and one can write in good agreement with experimental data. For
the following equation for the thermodynamic the case of more complex multifunctional mo-
cycle: lecules, including acetylsalicylic acid (ASA)
and ibuprofen (IBP) that are of interest to phar-

Figure 1. MD predictions of DhydG and of DLJG against molecular mass for various mono- and di-substituted barbiturates at
298 K
BA ¼ barbituric acid, MB ¼ methyl barbiturate, EB ¼ ethyl barbiturate, iPB ¼ isopropyl barbiturate, BAR ¼ barbital,
PRO ¼ probarbital, BUT ¼ butethal, PEN ¼ pentobarbital
6 Process Systems Engineering, 9. Domain Engineering

Figure 2. Experimental data and MD predictions of DhydG of various polar molecules using different force fields at 298 K
a) Methanol; b) Propanol; c) Ethylamine; d) Acetone; e) Acetic acid; f) IBP; g) ASA; h) Benzoic acid

maceutical industry, MD predictions from dif- and so ideal separation factors for various binary
ferent force fields deviate and the agreement gas and liquid mixtures have been measured.
with experiments is less satisfactory. The separation factor for n-C4H10/CH4 mixture
is used widely as a benchmark for hydrocarbon
1.4.2. Polymer Membranes for Gas mixture separation capability of a given mem-
Separation brane material. A new atomistic force field was
developed for PDMS that accounts for bond
Polymeric membranes (either glassy or elasto- stretching, bond angle bending, dihedral angle
meric) are used widely for separation of mix- torsion, and nonbonded intra- and intermolecu-
tures in chemical industry, medical applica- lar interactions [14]. For the nonbonded inter-
tions, etc. A major physical property in such actions, the Lennard–Jones potential for short-
a process is the permeability (P) of component i range van der Waals repulsive and attractive
in the polymer membrane defined as the product interactions together with a long-range Coulom-
of the solubility (S) and diffusivity (D), so that: bic potential were used. The model was shown
to predict accurately the thermodynamic prop-
P ¼ SD ð8Þ erties of polymer melts over a wide temperature
Separation of a binary mixture of compo- and pressure range [14]. It was further used for
nents i and j (where i is typically the most polymer–gas mixture simulations. In Figure 3,
permeable of the two components) by a given pure gas n-C4H10/CH4 solubility, diffusivity,
polymer membrane is characterized by the ideal and permeability selectivities in the range of
separation factor, which is the ratio of perme- 273–400 K calculated from MD simulations
abilities for components i and j according to the together with experimental data from [15] are
expression: shown.


 The solubility of hydrocarbons in the poly-
Pi Si Di
aid
ij ¼ ¼ ð9Þ mer matrix was based on the Widom’s test
Pj Sj Dj
particle insertion method that allows accurate
where the ratios aSij ¼ Si /Sj and aD ij ¼ Di /Dj calculation of the excess chemical potential of
represent the solubility selectivity and the dif- the solute in the solvent. MD simulation of the
fusivity selectivity, respectively. In rubbery polymer matrix for 5–10 ns followed by several
polymers aD ij is less than unity, while aij  1,
S
hundred thousands of solute molecule insertions
so ideal separation factor is governed by sel- in each polymer configuration (this is a relative-
ectivity of sorption. Polydimethylsiloxane ly fast post-processing calculation) provides an
(PDMS) is a widely used polymer membrane accurate estimate of the solubility.
 Process Systems Engineering, 9. Domain Engineering 7

Figure 3. n-C4H10–CH4 mixture behavior in PDMS as a function of temperature

A) Solubility (S) selectivity; B) Diffusivity (D) selectivity; C) Permeability (P) selectivity
Open symbols are experimental data [15] and closed symbols are MD predictions

For the calculation of diffusion coefficient, n-C4H10–CH4 mixture in PDMS showed an

significant longer MD simulations, on the order increase in CH4 solubility in the presence of
of 100 ns, are needed in order to ensure that the n-C4H10 in the polymer. On the other hand, only
hydrocarbon molecules diffusing through the a weak influence of CH4 on n-C4H10 solubility
polymer matrix reach the normal diffusing was reported. In Figure 4, experimental data and
(Fickian) regime [16]. In this case, the diffusion MD predictions are shown for the infinite dilu-
coefficient is calculated from the mean square tion solubility coefficient of CH4 in the PDMS–
displacement of the hydrocarbon molecules, n-C4H10 mixture at 300 and 450 K. Simulation
based on Einstein equation.
Figure 3 reveals that solubility selectivity
decreases significantly as temperature increases
while diffusivity selectivity increases but with a
smaller rate. Finally, the ideal separation factor
follows closely the trend exhibited by solubility
selectivity. In all cases, MD predictions are in
excellent agreement with experiments [15] over
the entire temperature range.
For the accurate design of a polymer mem-
brane for the separation of a real mixture, mix-
ture permeability data are needed. It is often
assumed that in rubbery polymers penetrants Figure 4. Mixed gas CH4 solubility in PDMS at 300 and
permeate independently of one another. How- 450 K as a function of n-C4H10 weight fraction in PDMS
a) Experimental data [15] at 300 K (open points); b) MD
ever, this behavior needs to be confirmed for a predictions at 300 K (closed points); c) MD predictions at
given system. Recent experimental data for the 450 K (closed points)
8 Process Systems Engineering, 9. Domain Engineering

specific application by the right choice of cation

and anion.
It is expected that ILs may revolutionize the
chemical process industry in the years to
come [18]. For example, they are increasingly
used as novel processing media in combination
with supercritical CO2. Due to the negligible
vapor pressure, it is possible to extract organic
products from ILs using supercritical CO2 with-
out any contamination by the IL. Despite the
Figure 5. Diffusion coefficient of pure and mixed n-alkanes wealth of experimental data available, more
in PDMS at ambient conditions. Solid circles correspond to data are needed for process design, and their
pure CH4 and n-C4H10 diffusion coefficient in PDMS. Open
symbols correspond to n-alkanes in mixture experimental determination is often difficult,
a) CH4 mixed with 2% n-C4H10 in PDMS; b) CH4 mixed time-consuming and expensive. Therefore, it is
with 10% n-C4H10 in PDMS; c) n-C4H10 mixed with 1% highly desirable to develop predictive methods
CH4 in PDMS for estimating the relevant thermodynamic,
phase equilibrium, and transport properties. At
results are consistent and in good agreement the molecular level, early molecular simulation
with experimental measurements [15]. studies focused on the development of accurate
Finally, the diffusion coefficients of a mix- force fields and validation towards the predic-
ture of CH4 and n-C4H10 in PDMS at ambient tion of structure and thermodynamic properties
conditions are shown in Figure 5 and are com- of ILs in melt. Recently, these models were used
pared to pure gas diffusion calculations. Clear- for the calculation of thermodynamic and trans-
ly, CH4 molecules move faster in the presence of port properties of IL melts and mixtures.
n-C4H10 molecules in PDMS matrix than in pure A powerful approach toward the develop-
polymer. The same behavior is observed for ment of accurate force fields is to start from the
n-C4H10 in the presence of CH4 molecules. The subatomic level with quantum mechanics cal-
presence of a second penetrant species swells culations. A recent example refers to ab initio
the polymer matrix resulting in an increase in density functional theory (DFT) calculations
the diffusion coefficient of the first penetrant. (! Process Intensification, 1. Fundamentals
The swelling behavior of PDMS in the presence and Molecular Level, Section 2.2.3.1) per-
of mixed gases and the consequent increase in formed on isolated IL molecules ([bmimþ]
diffusivity and permeability coefficients of the [Tf2N-], [hmimþ][Tf2N-], and [omimþ][Tf2N-])
corresponding gases has also been reported in order to evaluate the minimum energy struc-
experimentally by many investigators [15, 17]. ture and calculate charge density distribution of
the molecule [19]. In Figure 6, schematic repre-
sentation of DFT results are shown. DFT results
1.4.3. Ionic Liquids for Sustainable were used for the development of a realistic
Chemical Processes atomistic force field that was used subsequently
for MD simulations.
Ionic liquids (ILs) (! Ionic Liquids) have re- ILs have very long characteristic relaxation
ceived much attention for use as environmen- times and so long MD simulations are needed in
tally benign reaction and separation media. ILs order to obtain accurate thermodynamic and
are molten salts with melting points close to dynamic predictions. MD simulations of up to
room temperature. Their most remarkable prop- 50 ns on bulk ILs at various temperatures and
erty is that their vapor pressure is negligibly pressures were performed by [19]. Volumetric,
small, so that ILs are nonvolatile, nonflammable dynamic, and transport properties together with
and odorless. Other characteristics of ILs in- structure properties were calculated. Molecular
clude a wide liquid temperature range, a high conformations from MD simulations were in
thermal and electrochemical stability, a high good agreement with DFT results. The IL con-
ionic conductivity and good solvency proper- figurations generated from MD were used sub-
ties. In principle, ILs can be tailored for a sequently for the calculation of the excess
 Process Systems Engineering, 9. Domain Engineering 9

Figure 6. Relative electronic energies DE of the isolated ion pairs optimized at B3LYP/6-311þG* level Energies are given in
kJ/mol. Conformer III has been arbitrarily chosen as reference, thus relative energies are calculated by DEi ¼ Ei
E3. Energies
with zero point energy corrections are given in parentheses

Ionic liquid DE1 DE2 DE4 DE5 DE6

[C4mimþ][Tf2N
] 0.0; (0.0) 0.0; (0.0)
3.9; (
4.0) þ2.8; (þ2.9)
0.9; (
0.5)
[C6mimþ][Tf2N
] þ0.6; (þ0.1) þ0.2; (þ0.1)
3.8; (
4.4) þ2.7; (þ3.0)
0.6; (
0.5)
[C8mimþ][Tf2N
] þ0.1; (þ0.4) 0.0; (0.0)
4.0; (
4.1) þ3.2; (þ2.3)
0.3; (
0.9)

chemical potential, and thus solubility, of CO2 1.5. Conclusions

in the IL using the Widom’s test particle inser-
tion method. In all cases, excellent agreement Molecular simulation is a mature computational
with experimental data was obtained. Represen- tool that can be used reliably by material scien-
tative results concerning IL self-diffusion coef- tists and chemical engineers for industrial prod-
ficient and CO2 solubility in [bmimþ][Tf2N-] uct and process design. Highly robust and effi-
are shown in Figure 7 and Figure 8, respectively. cient computer codes have been developed by

Figure 7. Experimental data (markers) and MD simulation Figure 8. Henry’s law constant of CO2 in [bmimþ][Tf2N
]
results (lines) for the self-diffusion coefficient of various a) Experimental data (blue squares and line);b) Molecular
cations at 0.1 MPa simulation predictions from TraPPE (red squares); c) EPM2
a) C4 mim; b) C6 mim; c) C8 mim model (open diamonds) for CO2
10 Process Systems Engineering, 9. Domain Engineering

major academic and government laboratories tems of the future is introduced by [26].
worldwide and are freely available for research Energy systems engineering provides a meth-
and development purposes [20–23]. In addition, odological framework to address the complex
specialized computational chemistry and model- energy and environmental problems by an
ing software companies offer state-of-the-art integrated systematic approach which ac-
user-friendly interfaces to support the rather counts complexities of very different scales,
complex computer codes [24, 25]. In this respect, ranging from technology, plant, to energy
molecular simulations can be performed almost supply chain, and megasystem. Energy sys-
routinely by nonexperts. tems engineering employs systems-based
Despite the above, significant challenges still representations and methods, such as super-
exist: Accurate force-fields are of outmost im- structure-based modeling, mixed-integer pro-
portance in order that simulation results resem- gramming (MIP), multiobjective optimiza-
ble the real systems. Although significant ad- tion, optimization under uncertainty (see !
vances have been made in recent years, there is Process Systems Engineering, 3. Mathemati-
still need for developments of intra- and inter- cal Programming (Optmization) and ! Math-
molecular interaction models for highly com- ematics in Chemical Engineering, Chap. 10,
plex chemical compounds. and life-cycle assessment ! Waste, 2. Life-
As molecular simulation matures, the com- Cycle Assessment). These methodologies
plexity of the problems where it is applied have been applied in energy systems of very
increases. Very often, the detailed atomistic different nature and scale, including polyge-
representation of interactions is not necessary neration energy systems, urban energy sys-
anymore and a more coarse-grained representa- tems, hydrogen infrastructure, oil and gas
tion becomes more suitable. In such cases, a production, wind turbine, electric power in-
systematic hierarchical approach is needed in dustry, carbon dioxide capture and sequestra-
order to parameterize a model consistently. tion, and distillation columns [27, 28].
Finally, one should recognize the fact that com-
parison of simulation predictions against exper-
imental data at various time and length scales is
always necessary in order to validate the model 2.2. Methods/Tools/Algorithm
and the methodology used.
2.2.1. Superstructure-Based Modeling

2. Energy Systems Engineering Superstructure-based modeling is an approach

to simultaneously determine the optimal con-
2.1. Introduction figuration of a process and its optimal operating
conditions via mathematical programming [29].
Excessive energy consumption and conse- It was first proposed to address process synthesis
quent greenhouse gas (GHG) emissions have issues in heat-exchanger networks (HEN) [30],
become two major crucial global issues, and and widely used in process design thereafter,
this situation is most likely to continue in the and it is regarded as one of the most significant
next couple of years to come. Driven by this accomplishments in PSE [31].
urgent situation, technologies which can Superstructure-based modeling has been
facilitate a smooth transition from existing widely used in a broad range of fields. These
energy systems to more advanced ones are fields include heat-exchange networks [30],
receiving more and more serious attention. separation and distillation [32–34], reactor net-
However, although there already exist many works [35, 36], water usage and treatment net-
technical options, they usually differ greatly work [37], and energy systems [38, 39].
from one another in many aspects, and they Superstructure-based modeling usually in-
are often treated separately by their own tech- volves discrete decision making, e.g., inclusion
nical or political groups. of a certain type of reactor or not. Simultaneous
The concept of energy systems engineering modeling of discrete decisions and continuous
as an integrated approach for the energy sys- terms is usually implemented via MIP.
 Process Systems Engineering, 9. Domain Engineering 11

2.2.2. Mixed-Integer Programming (MIP) problem according to two or more (conflicting)

criteria subject to certain constraints (! Energy
An optimization model with both integer and Management in Chemical Industry, Section
continuous variables is denoted as a MIP prob- 3.1). Multiobjective optimization is suitable to
lem [40] see ! Process Systems Engineering, be applied to a problem where trade-offs exist
3. Mathematical Programming (Optmization) amongst its objective functions and optimal
and ! Mathematics in Chemical Engineering, decisions should be made in the presence of
Chap. 10. Integer variables in MIP problems these trade-offs. Multiobjective optimization is
usually refer to 0–1 variables, also known as widely used in various fields, including product
binary variables, only, due to the fact that any and process design, supply chain design, and
integer variable can be represented in terms of a energy systems engineering. A common multi-
set of binary variables. objective optimization problem involved with
MIP is widely used in PSE. Typical applica- energy system design is to maximize profitabil-
tions are, e.g., superstructure-based modeling, ity and minimize environmental impacts
facility location and allocation problems, sched- simultaneously.
uling problems. A canonical form of a MIP A generic mathematical from of a multiobjec-
problem is presented as follows: tive optimization problem is presented as follows:
8
min f ðx; yÞ >
> f1 ðx; yÞ
x;y >
>
s:t: hðx; yÞ ¼ 0 < f2 ðx; yÞ
min U
gðx; yÞ  0 x;y >
>
>
>
x  0; x 2 X  Rn :
fn ðx; yÞ
y 2 f0; 1gq x;y ð10Þ s:t: hðx; yÞ ¼ 0
gðx; yÞ  0
where x is a vector of n continuous variables,
x  0; x 2 X  Rn
and y is a vector of q 0–1 variables. y 2 f0; 1gq ð11Þ
Depending on specific forms of the objective
function f, equality constraints h, and inequality where x is a vector of n continuous variables,
constraints g, MIP problems can be classified into and y is a vector of q 0–1 variables.
twocategories:mixed-integerlinearprogramming The target of solving a multiobjective opti-
(MILP) problems, where the objective function mization problem is to obtain the utility function
and all constraints are linear, and mixed-integer U, where n scalar objective functions are to be
nonlinear programming (MINLP) problems, optimized simultaneously [44]. Usually, some
where either the objective function or some con- conflicts exist amongst the objective functions.
straints are nonlinear. MINLP problems can be However, if there are no conflicts, then a single
further classified as convex MINLP problems, solution can be obtained where every objective
where the objective function is a convex function function attains its optimum. In this case, opti-
and the feasible region is a convex region, and mizing the objective functions simultaneously
nonconvex MINLP problems, where either the or separately arrive at the same optimal solution.
objective function is a nonconvex function or the To avoid such simple cases, multiobjective
feasible region is a nonconvex region. optimization problems discussed hereafter al-
Themostcommonlyusedalgorithmforsolving ways involve conflicting objective functions.
MILP problems is branch and bound method [41]. Typical algorithms for solving multiobjec-
It has a huge number of varieties. Two commonly tive optimization problems are parametric
used algorithms for solving MINLP problems are programming [45] and the e-constraint method
generalized benders decomposition (GBD) [42] [46, 47].
and outer approximation (OA) [43], both of which
have a large amount of varieties.
2.2.4. Optimization under Uncertainty
2.2.3. Multiobjective Optimization
Uncertainty is inevitable and unpredictable in
Multiobjective optimization, or multicriteria process planning and design over a long-term
optimization, is to simultaneously optimize a horizon. Because of the very nature of these
12 Process Systems Engineering, 9. Domain Engineering

tasks, many parameters obtained at the planning methods start from a breakdown of a system
or design phase are subject to considerable under study into fundamental components
variability and cannot be predicted with a cer- and processes, then extract inventory data of
tain degree of accuracy. Optimization under these components and processes from a huge
uncertainty takes the impact of uncertain para- inventory database which contains inventory
meters into consideration at the planning and data of all primary products and processes,
design stage thus improves a plan or design in then multiply these inventory data with their
terms of both feasibility and operability (see capacity within the system under study and
! Process Systems Engineering, 3. Mathemat- sum them up to provide the LCA indicator.
ical Programming (Optmization)). . Economic input–output LCA. This method
estimates materials and energy requirements
2.2.5. Life-Cycle Assessment and environmental emissions in activities of
an economy. It uses information of industry
Life-cycle assessment (LCA), also known as transactions, i.e., purchases of materials by
life-cycle analysis, is to evaluate and quantify one industry from another industry, and infor-
the environmental impacts of a certain product mation of direct environmental emissions of
or production procedure caused by its existence. industries, to evaluate the entire environmen-
The definition and method of product LCA is tal impacts of a system or process under study.
described in detail in (! Waste, 2. Life-Cycle
Assessment). 2.3. Energy Systems Examples
Depending on the boundaries of a system
where LCA is applied, LCA can be classified The aforementioned energy systems engineer-
into the following four categories: ing methodologies can greatly facilitate the
planning or design of energy systems of differ-
. Cradle-to-gate. It accounts for the environ- ent types and scales, at different levels, from
mental impacts of a product produced at all different aspects, and according to different
stages before it is sent to the gate of a factory. criteria. Some of these methodologies have been
These stages usually consist of mining, pre- successfully applied in energy systems of very
processing, and transportation. different nature and scale, and have been sum-
. Cradle-to-grave. It accounts for the environ- marized as follows:
mental impacts of a product in its entire life
time, from manufacture up to disposal phase. . Polygeneration energy systems [38, 48–51]
. Cradle-to-cradle. It accounts for the environ- . Urban energy systems [52]
mental impacts of a product in a recycling . Hydrogen energy systems [53–55]
process, from the production of a product of a . Energy systems in commercial buildings [56,
certain type of material to the production of 57]
another product of the same material. . Electric power industry [58]
. Well-to-wheel. It is a specific type of LCA . Pulp and paper industry [59–63]
widely used in fuel and transportation LCA, . Oil and gas production [64]
accounting for the energy consumption and . Wind turbines [65]
emissions production from exploration to fi- . Carbon dioxide capture and sequestration [66,
nal consumption. According to the particular 67]
research interest, it can be further divided into . Separation and distillation [68]
well-to-tank and tank-to-wheel stages, or
well-to-station and station-to-wheel stages.
2.3.1. Example 1–Polygeneration Energy
Depending on the means an LCA impact Systems
factor is evaluated, LCA can be classified into
the following two categories: A polygeneration energy system is a multiinput
and multioutput energy system that coproduces
. Inventory-based LCA. Most conventional electricity and synthetic liquid fuels. Process
LCA methods belong to this category. These design of a polygeneration energy system in-
 Process Systems Engineering, 9. Domain Engineering 13

volves several typical energy systems engineer- dictable uncertainties. Design of a polyge-
ing issues, as follows: neration energy system under uncertainty
makes the task further complicated.
. A polygeneration energy system is a very
complex system which comprises many units A modeling and optimization framework
and pieces of equipment. For each of these for the optimal process design of polygenera-
units and pieces of equipment, there usually tion energy systems is proposed by [48–51],
exist many alternative technologies or types based on the energy systems engineering ap-
of equipment. Making the optimal selection proaches presented in the previous section.
from the many alternatives remains a First, a superstructure representation of a
challenge. polygeneration energy system is constructed,
. As public concern over fast increasing GHG as shown in Figure 9, where a polygeneration
emissions grows, environmental impact of an energy system is divided into many functional
energy system has become an important de- blocks. For each functional block, all alterna-
sign criterion. Designing a polygeneration tive technologies and types of equipment are
energy system according to multiple design included in the superstructure representation,
criteria (economic, environmental, etc.) poses thus all possible types of process design are
another challenge. captured.
. A polygeneration energy system usually has Based on the superstructure representation, a
an operating horizon of several decades, over MINLP design problem is developed in the
which there exist many inevitable and unpre- following form:

Figure 9. Superstructure representation of a polygeneration energy system (CCS ¼ Combined combustion system; GCS ¼
Gasification chamber and syngas scrubber; HRSG ¼ Heat recovery steam generator; ASU ¼ Air separation unit
14 Process Systems Engineering, 9. Domain Engineering

On obtaining these two objective functions, a

min f ðy; d; xÞ
y;d;x multi-objective MINLP problem is formed as
s:t: hdc ðy; dÞ ¼ 0 follows:
gdc ðy; dÞ  0 (
hoc ðy; d; xÞ ¼ 0 f1 ðy; d; xÞ ¼
NPV
min U
goc ðy; d; xÞ  0 y;d;x f2 ðy; d; xÞ ¼ GHG
d 2 Rm ; x 2 Rn ; y 2 f0; 1gq ð12Þ s:t: hdc ðy; dÞ ¼ 0

where gdc ðy; dÞ  0

hoc ðy; d; xÞ ¼ 0
goc ðy; d; xÞ  0
. Binary design variables are denoted as y,
d 2 Rm ; x 2 Rn ; y 2 f0; 1gq ð13Þ
which represent the selection (or not) of tech-
nologies or types of equipment for each func- where f1 is the objective function representing
tional block. the NPV, and f2 is the objective function repre-
. Continuous design variables are denoted as d, senting the GHG emissions.
which represent the capacities of the func- Equation (13) is solved using the e-constraint
tional blocks. method. Optimal results are presented on a
. Continuous operational variables are denoted pareto curve, as shown in Figure 10. For this
as x, which represent quantitative decisions to example, there exist 18 different combinations
be made at the operational stage, e.g., flow- of technologies, but only four of them appear on
rates, stream compositions and the like. the pareto curve, according to different econom-
. Equality design constraints are denoted as hdc, ic and environmental design criteria. Each point
which involve design variables only, e.g., on the curve represents a different process de-
evaluation of initial capital costs. sign. A decision-maker can thus pick up any
. Inequality design constraints are denoted as point from the curve according to their specific
gdc, which involve design variables only, e.g., interest or requirements.
logical relations between different functional In Equation (13), all time-variant parameters
blocks equality operational constraints hoc, are considered as piecewise constant functions
which involve design and operational vari- over the operation horizon, which is discretized
ables, e.g., mass and energy balances. into several time intervals. However, due to the
. Inequality operational constraints goc, which very nature of the long-term operation horizon,
involves design and operational variables, uncertainty is almost inevitable at the design
e.g., capacity constraints. stage, e.g., due to external factors, such as market
demands for products, prices of feedstocks and
In the MINLP design problem, there could be products. Here, all uncertain parameters can be
more than one objective function, i.e., design presented as random variables following given
criterion. Here, both the economic and environ- probability distribution functions p(x).
mental behavior of a polygeneration energy By incorporating the uncertainty into the
system is evaluated. Net present value (NPV) MINLP design problem, the following two-
is selected to be the economic design criterion, stage stochastic programming problem results:
which comprises the initial capital costs and the
min fd ðy; dÞþEu 2 Q½fs ðy; d; uÞ
discounted profit obtained over the entire oper- y;d

ating horizon. A cradle-to-gate GHG emissions s:t: hdc ðy; dÞ ¼ 0

indicator is selected to the environmental design gdc ðy; dÞ  0
criterion, mainly comprising three parts: d 2 Rm ; y 2 f0; 1gq
with :
fs ðy; d; uÞ ¼ min fs ðy; d; x; uÞ
. GHG emissions produced within the process x
s:t: hoc ðy; d; x; uÞ ¼ 0
during operation
goc ðy; d; x; uÞ  0
. GHG emissions produced throughout mining, x 2 Rn ; u 2 Q ð14Þ
extraction, and other processing stages of
feedstocks where the objective function is split into a
. GHG emissions produced during equipment deterministic term fd representing decisions at
production and plant construction the design stage, and the expectation of a sto-
 Process Systems Engineering, 9. Domain Engineering 15

Figure 10. Pareto curve for polygeneration energy systems design

a) H–CH–G–GTH; b) Q–CQ–L–GTH; c) RC–CRC–G–GTH; d) Q–CQ–G–GTH

chastic term fs which depends on the realization x are made. Equation (14) is solved using a
of uncertain parameters u at the operation stage. decomposition-based solution strategy.
Discrete variables y and continuous variables d
are ‘‘here-and-now’’ (design) variables which
should be decided at the first-stage problem 2.3.2. Example 2–Hydrogen Infrastruc-
before the realizations of uncertain parameters ture Planning
u occur, and x is a vector of ‘‘wait-and-see’’
(operational) variables which can be decided at Energy systems engineering methodologies
time interval t of the second-stage problem have been applied in hydrogen infrastructure
where all uncertain parameters have been ob- planning [39, 69]. The problem under study is
served. In the second-stage problem, the illustrated in Figure 11: given a specific region
recourse term based on a specific realization of where several potential production sites and
uncertain parameters is optimized and corre- markets (city as shown in the Figure 11) are
sponding corrective actions in terms of values of available, obtain the optimal infrastructure

Figure 11. Illustrative representation of a hydrogen infrastructure planning problem

16 Process Systems Engineering, 9. Domain Engineering

which connects the production sites to markets . Transitions from one type of supply chain
via a supply chain from primary feedstocks, structure to another over time
central production, distribution, forecourt re- . Planning according to both economic and
fueling, to the final product over a long-term environmental performance indicators
planning horizon.
This approach addresses the following issues A superstructure representation of the
involved in hydrogen infrastructure planning: modeling framework is shown in Figure 12. It
captures all possible types of primary feed-
. Planning over a long-term future horizon stocks, production sites, production technolo-
. Geological site allocation gies, distribution technologies, forecourt refill-
. Representing the state of existing infrastruc- ing technologies, and potential markets, and
ture, especially the natural gas distribution gives the optimal planning scheme over the
network, electricity grid, and existing hydro- entire future planning horizon.
gen production facilities Based on this modeling framework, a multi-
. All types of available primary feedstocks, objective optimization was conducted where
production, distribution, and forecourt refuel- NPV was selected as an economic objective and
ing technologies a LCA-based environmental impact factor as an
. Trade-offs between large-scale centralized environmental objective. A pareto frontier com-
production and small-scale distributed prising the full range of trade-offs between the
production economic and environmental objectives was

Figure 12. A superstructure representation of the modeling framework for hydrogen infrastructure planning
a) Gasoline equivalent WTW (well-to-wheel) emissions
SMR-LIQ ¼ Manufacturing of liquid hydrogen via steam methane reforming
SMR-GAS ¼ Manufacturing of gaseous hydrogen via steam methane reforming
NG COMP ¼ compression of natural gas (NG)
GAS-LIQ ¼ Manufacturing of liquid hydrogen via gasification
GAS-GAS ¼ Manufacturing of gaseous hydrogen via gasification
ELC-LIQ ¼ Manufacturing of liquid hydrogen via electrolysis of water
ELC-GAS ¼ Manufacturing of gaseous hydrogen via electrolysis of water
CNG ¼ Compressed natural gas
LIQ ¼ Liquid hydrogen
GAS ¼ Gaseous hydrogen
NG pipe ¼ Natural gas pipeline
H2 PIPE ¼ Hydrogen pipeline
SMR ¼ Onsite hydrogen production via steam methane reforming
ELC-N ¼ Onsite hydrogen production via electrolysis of water using nonrenewable electricity
ELC-R ¼ Onsite hydrogen production via electrolysis of water using renewable electricity
ELC-U ¼ Onsite hydrogen production via electrolysis of water using nuclear electricity
 Process Systems Engineering, 9. Domain Engineering 17

Figure 13. Pareto curve for hydrogen infrastructure planning (WTT ¼ well-to-tank)

obtained, shown in Figure 13. Any point on the . Selection of technologies. For each type of
pareto frontier represents an infrastructure design energy demands, several types of technolo-
with specific economic and environmental per- gies or types of equipment are usually avail-
formances, and decision-makers can pick up any able. Selecting the optimal combination of
point from this curve as the final design according them may become a challenging problem
to their own specific interest and preference. when facing with too many choices. This
issue could be further complicated when in-
volved with other design issues, e.g., integra-
2.3.3. Example 3–Energy Systems in tion between energy consumption and energy
Commercial Buildings production sectors.
. Integration. Integration amongst different
The applications of energy systems engineering energy consumption sectors within a system
methodologies in polygeneration energy sys- can reduce the entire energy demand of the
tems and hydrogen infrastructure planning fo- system. For example, heat produced in the
cus primarily on the energy production side. refrigeration sector of a supermarket could
However, energy systems engineering is not be collected to heat the aisle space, other-
confined within the scope of energy production. wise an extra amount of energy is required
It can also be applied to model and optimize the to meet the heating demand. The integration
energy consumption within a process or system. issue could become more complicated when
Next, its applications in design of the energy on-site production technologies are also
systems in commercial buildings [49] are pre- involved.
sented to illustrate its potential applications on . Building design. From an energy saving view-
the energy consumption side. point, building design should also be involved
The energy system in a commercial building at the design phase. For example, sizing and
usually comprises both an energy consumption positioning of windows could be considered
section and an energy supply section. Energy together with the lighting requirement of a
demands usually come from requirements for build to minimize it.
lighting, HVAC (heating, ventilating, and air . GHG emissions. From an LCA point of view,
conditioning), and refrigeration. The energy emissions from a commercial building come
supply is usually obtained from grid electricity, from two sources. One source is the emissions
district heat, and on-site energy generation, e.g., produced over the entire operation period, and
distributed power generation and boilers. Major the other one is the emissions produced in
issues to be addressed at the design stage are manufacturing and transporting equipment
summarized as follows: and construction materials. Emissions from
18 Process Systems Engineering, 9. Domain Engineering

Figure 14. Superstructure representation of the energy system in a commercial building

PE ¼ primary energy; GE ¼ grid electricity; DH ¼ district heat; OEG ¼ on-site energy generation; E ¼ electricity; h ¼ heat;
R ¼ refrigeration; L ¼ lighting; V ¼ ventilation; B ¼ bakery; SH ¼ space heating; P ¼ production; ES ¼ energy saving; D ¼ demand

both categories should be considered at the 2.4. Conclusions and Future

design phase to give an overall environmental Directions
impact indicator.
The introduced methodologies of energy sys-
To address these issues, a superstructure tems engineering cooperate with each other and
representation of the energy system in a com- provide a systematic solution strategy for the
mercial building is firstly constructed, as shown planning and design issues involved with any
in Figure 14. It comprises an energy supply energy system. These methodologies are illus-
section, an energy conversion section, and an trated via their applications in a simple example
energy savings section. The function of the of polygeneration energy systems design. It
energy supply section is to provide electricity shows that energy systems engineering is of
and heat for the entire energy system. The tremendous importance to guide the transition
energy conversion section converts electricity from our existing generation of energy systems
and heat obtained from the energy supply sec- to a more energy efficient and environmentally
tion to all energy demand tasks, such as refrig- benign one. It is certain that research in this field
eration, lighting, ventilation, bakery, and space will continue and prosper. Some recommenda-
heating. The energy savings section further tions for future research directions are summa-
involves available types of energy savings tech- rized as follows:
nologies, such as night blind and weir screen for
the refrigeration subsystem. . The generic modeling and optimization meth-
Based on the superstructure representation, a odologies presented in this section can serve
multiobjective MILP problem is formed and as a starting point, and more methodologies
solved to obtain the pareto curve, as shown in which are suitable for energy systems could
Figure 15. A decision-maker can pick up any be added into the scope of energy systems
point from the pareto curve according to their engineering. This would certainly extend its
specific design criteria or interest. Once a design applicable fields and enhance its capability.
point is selected (e.g., A, B, C, or D), the system . Modelingatamicrolevelcouldbeexplored.The
configuration behind it can be obtained directly methodologies introduced here enable model-
from the model results. ing at strategic planning and process design
 Process Systems Engineering, 9. Domain Engineering 19

Figure 15. Pareto frontier for the energy system design in a commercial building

levels, which can be regarded as modeling at methodologies and tools for the optimal de-
megalevel and macrolevel, respectively. sign of bioenergy value chains with the right
Modeling at a microlevel, e.g., at the molecular technologies at the right scale, and providing
level for biodiesel production, would give a policy suggestions to direct the development
much better insight to these systems. of bioenergy.
. The generic modeling and optimization meth-
odologies introduced in this section could be
used in the control field. Firstly, it can be used in 3. Pharmaceutical Processes
an integrated design and control scheme where
both operational and control requirements are 3.1. Introduction
taken into consideration at the design stage.
Secondly, the framework can be also used in the For pharmaceutical companies, drug develop-
context of model predictive control. ment requires a huge investment (! Pharma-
. Applications in energy value chain modeling ceuticals, General Survey, Chap. 4). A study
and optimization. For example, bioenergy is estimates that the cost to bring a single new drug
expected to play an important role in the to the market costs over $800 million [70]. The
ongoing transition from conventional energy clinical period of a new drug is complex, and
system to a more sustainable and environmen- divided into three phases. In phase I, normally a
tally benign one. There also have been many small number of healthy volunteers are tested to
controversies around bioenergy about its ca- find safe dosages. In phase II, the drug candidate
pability to ameliorate energy security and is given to a large number of patients. Phase III
climate change, concerning its life-cycle trials are more extensive, often carried out at
green-house gas (GHG) emissions and com- more than one clinical research centers. Figure 1
petition on land use with food crops. The shows the probability of entering the next phase
modeling and optimization methodologies from the previous phase. As can be seen, only
developed in this framework could be used one third of the drug candidates—which have
to guide the planning and design of a bioe- already passed the drug discovery stage and
nergy value chain in terms of analyzing and preclinical trials—reaches the final stage, phase
quantifying net profit of bioenergy, producing III. Furthermore, nearly 40% of the drugs that
20 Process Systems Engineering, 9. Domain Engineering

Figure 16. Drug development phases [70, 71]

a) Phase I; b) Phase II; c) Phase III; d) Registration launch

pass phase III do not get marketing approval. ty assurance. In this traditional approach, product
Therefore, wise decision-making at an earlier quality was achieved predominantly by restrict-
stage can reduce the development cost dramati- ing flexibility in the manufacturing process [80].
cally, and methods/tools that support such de- Manufacturing procedures are treated as being
cisions play a crucial role in drug development. frozen and changes in the process must be man-
The three clinical trial phases are closely aged through regulatory submissions. As a result,
related to the development of drug manufactur- many production processes are designed and
ing, and often proceed concurrently; the optimal operated inefficiently, and relatively little effort
production process design is explored during the has been devoted to innovate and improve them.
clinical trials, and scale-up is performed simulta- In some cases, the amount of product waste as a
neously. In phase I, a small-scale pilot facility–in result of mistakes in manufacturing was as high as
the order of one-hundredth of production scale–is 50% of the product manufactured [81]. This led to
usually sufficient, where as in Phase III a pilot significantly higher costs and even delays of new
plant of one-tenth production scale is often need- drug development.
ed to supply a sufficient amount of the new drug, In 2002, the US Food and Drug Administra-
which is administered for this large-scale clinical tion (FDA) launched the process analytical
trial [71]. Therefore, as the clinical tests proceed, technology (PAT) initiative to challenge the
the production volume goes up, and efficient hesitancy to innovate pharmaceutical manu-
manufacturing that can supply such an amount facturing [82]. This initiative, a paradigm
of the drug candidate becomes more critical. change of the FDA to inspect and approve
Process systems engineering (PSE) techni- pharmaceutical processes, promotes better un-
ques have been applied to these unique issues in derstanding of drug production processes. Here,
the pharmaceutical industry to aid high-level the term analytical should be interpreted broad-
decision-making. The complexity of these ly; it includes ‘‘chemical, physical microbiolog-
problems is recognized as one of the challenges ical, mathematical, and risk analysis conducted
by the PSE research community [72–74]. Some in an integrated manner’’ [83]. For example,
examples of recent research activities include PAT encourages fundamental process under-
portfolio management for drug research and standing for on-line or real time process control
development by multistage stochastic optimiza- to ensure product quality by reducing variability
tion [75], development of informatics infrastruc- in the process [82]. Application reports of PAT
ture for product development and manufactur- of specific processes include batch crystalliza-
ing [76], supply chain optimization [77], model- tion [84] (! Crystallization and Precipitation,
based design/analysis of PAT systems [78], Section 5.6), freeze drying [85], and fermenta-
and resource investment and scheduling for tion [86, 87]. A software for design of PAT
new drug product development [79]. systems has also been developed by [88].
Another change in pharmaceutical process
3.2. Pharmaceutical Process development and operation was brought by the
Development and Operation introduction of quality by design (QbD) [89]. In
QbD, the concept of Design Space has given a
Traditional pharmaceutical manufacturing most- significant impact to the pharmaceutical process
ly relied on extensive laboratory testing for quali- development. The design space is defined as
 Process Systems Engineering, 9. Domain Engineering 21

‘‘the multidimensional combination and inter- the CSD estimation techniques, the sieve analy-
action of input variables and process parameters sis is a primary offline technique, but this relies
that have been demonstrated to provide assur- on good sampling which cannot be always real-
ance of quality’’ [89], which is initially deter- ized. On the other hand, on-line measurement
mined during product development and re- techniques have been attracting attention in re-
ported to the regulatory agency. Change of cent years. In particular, focused beam reflec-
process parameter values within the design tance measurement (FBRM) has been employed
space usually does not require a regulatory in many crystallization studies. FBRM is a probe
post-approval process [89]. This enables flexi- which can be installed directly in a crystallizer
ble changes of manufacturing operations and eliminating the need for sampling (Fig. 17). This
wider applications of process automation, on-line measurement technique obtains the
which had been severely limited previously in chord length distribution (CLD). The challenge
this industry. here is to find the relationship between the CLD
With such unprecedented changes, a larger and CSD. This problem has been recognized by
number of advanced PSE techniques are begin- the PSE community, and techniques based on
ning to be applied to pharmaceutical processes. projections onto convex sets [93], Monte Carlo
Some specific PSE technologies are mentioned simulation [94], and principal component anal-
in the documents from the regulatory agen- ysis [95] have been proposed.
cies [83, 89], and studies in response to such Crystallization consists of two major me-
demands have been carried out including pro- chanisms, nucleation and growth (! Crystalli-
cess control [82, 86], process monitoring zation and Precipitation, Section 4.1 and !
[78, 85, 88, 90], and development of multivari- Crystallization and Precipitation, Section 4.2).
ate design tools [91]. Furthermore, computer- Nucleation is an event where the solute mole-
aided process design and simulation tools are cules in the solvent gather and form nuclei. The
beginning to be used in pharmaceutical process rate of nucleation can be modeled as a function
development [92]. of the concentration of the API in the mother
Another important trend in the pharmaceuti- liquor C, and the the equilibrium concentration
cal manufacturing is a revisit to continuous at temperature T, C ðTÞ:
processes. Traditionally, pharmaceutical pro- 0  2 1
duction relies on batch processing due to the B
16pg r
3 M
C
low production volume. However, understand- Rb ¼ kb exp@  2 A ð15Þ
3k3 T 3 ln CCðTÞ
ing the process dynamics and dealing with the
batch-to-batch fluctuation often become the
bottleneck for efficient production. Thus, the where g is the interfacial tension, and r the
reliability of continuous processes, in addition density, M the molecular mass, and k the
to the higher productivity, is attracting pharma- Boltzmann’s constant. On the other hand,
ceutical manufacturers. growth is the subsequent step where the size of
the nuclei increases. The growth rate GðL; tÞ is
usually expressed in terms of the degrees of
3.2.1. Crystallization supersaturation C
C ðTÞ [96, 97]:


Ea
Although crystallization is widely used in the G ¼ kg;0 exp
ðC
C  ðTÞÞg ð16Þ
RT
pharmaceutical industry, it remains one of the
most poorly understood processes. In particular, The mass balance can be described by the
controlling the size and shape of crystals is a big population balance equation (PBE). If breakage
challenge, which requires substantial experi- and agglomeration is ignored, and it is assumed
mental, modeling, and computational efforts that the crystal size can be represented by the
(! Crystallization and Precipitation). crystal length L, the PBE for a batch crystallizer
The crystal size in a crystallizer can be char- is given by [98]:
acterized by the crystal size distribution (CSD).
Estimating the CSD accurately is crucial for qnðL; tÞ qðGðL; tÞnðL; tÞÞ
þ ¼0 ð17Þ
qt qL
process development of crystallization. Among
22 Process Systems Engineering, 9. Domain Engineering

Figure 17. Batch crystallization of paracetamol [84]

A) Lab-scale crystallizer; B) Paracetamol crystals; C) Experimental setup of FBRM
a) Computer; b) Temperature probe; c) Agitation; d) Cooling fluid (water); e) Batch crystallizer; f) FBRM

where nðL; tÞ is the particle density. The bound- model-based control and optimization [104,
ary condition at L ¼ 0 is given by the nucleation 105] can be applied. These recent research
rate: activities on crystallization are influenced and
encouraged in particular by the PAT initiative.
Rb
nðt; 0Þ ¼ ; ð18Þ A generic model-based frameowrk for crystal-
Gðt; 0Þ
lization modeling, control, and monitoring has
It can be noted that this is a hyperbolic partial recently been developed by [106].
differential equation (PDE), which requires fine
discretization. Therefore, an efficient solution
technique must be employed to reduce the
computational cost to solve this equation, such 3.2.2. Chromatography
as a flux limiter and space-time conservation
element method [98, 99]. In addition to the Chromatography is often the only choice for
computational difficulty associated with the separation of thermally sensitive compounds
PBE, there are many model parameters such as proteins, or structurally similar com-
ðg; Ea ; kg;0 ; kb Þ which must be identified for each pounds; it requires only a slight difference in the
application. These parameters can be estimated affinity for adsorbent particles. Figure 3(a)
by advanced computational techniques such as shows a traditional batch chromatographic pro-
design of experiments (DOE) [100] and param- cess, where the purified products A, B, and C are
eter estimation techniques [101, 102]. In esti- fractionated at the outlet ðx ¼ LÞ of the column.
mating these model parameters, it has been The feed mixture is supplied at the top of the
reported that utilizing on-line measurement of column which is packed with solid adsorbent
the crystal size increases the reliability [102, particles. Then the mixture is eluted with des-
103]. After obtaining the crystallization model, orbent, which can be water or organic solvent
 Process Systems Engineering, 9. Domain Engineering 23

Figure 18. Chromatographic separation processes

A) Batch process; a) Chromatographic column
B) Simulated moving bed (SMB) process; a) Direction of valve switching

(s). Because of the difference in the affinity for process for active pharmaceutical ingredients
the adsorbent, the migration speeds of the com- (API) purification. An SMB system consists of
ponents are different and the components sepa- multiple columns connected to each other mak-
rate from each other as they move towards the ing a circulation loop (Fig. 18B). Between every
bottom. Due to its batchwise operation, the column, there are inlet ports for feed and des-
throughput, or the feed processing rate, is often orbent streams, as well as outlet ports for extract
small. Furthermore, it consumes a large amount and raffinate streams. The feed and desorbent
of desorbent which dilutes the product, and thus are supplied continuously and at the same time
the evaporation cost can be high. extract and raffinate are drawn continuously
To overcome these drawbacks, simulated through the ports. These four inlet/outlet ports
moving bed (SMB) chromatography (! Chro- are switched simultaneously at a regular interval
matographic Reactors, Chap. 2) has been devel- in the direction of the liquid flow. This system
oped by the Universal Oil Products (UOP) in the does not reach a steady state but a cyclic steady
1960’s, and applied to the separation of xylene state (CSS), where the concentration profiles
isomers. After their patent expired, its applica- change dynamically, but the profiles of both
tion areas expanded into sugar separations, in liquid and solid phase at the beginning of a
particular for the production of high-fructose cycle are identical to those at the end of the
corn syrup. In the past decade, applications in cycle.
the pharmaceutical industry have been gaining Established mathematical models for chro-
attention. In particular, separation of enantio- matography can be found in literature. Compre-
mers has been found to be one of the most hensive reviews are given by [107, 108]. One of
effective applications. Figure 19 shows an SMB the modeling approaches is the linear driving
24 Process Systems Engineering, 9. Domain Engineering

proaches have been developed based on PDE-

constrained numerical optimization of the rig-
orous dynamic chromatographic model.
Stochastic optimization approaches as well as
Newton-based approaches have been pro-
posed [109–111]. Alternatively, this problem
can be formulated as a multiobjective optimiza-
tion problem to evaluate more than one objec-
tives, such as throughput, purity, and desorbent
consumption, and obtain the pareto optimal
set [112, 113].
For SMB chromatography, many new im-
proved operations have been proposed to en-
hance the performance. Traditional SMB sys-
tems keep the liquid velocities constant during
a step, and then switch the four inlet/outlet
streams at the same time. In PowerFeed sys-
Figure 19. Simulated moving bed chromatographic pro- tems, however, the velocities become time-
cess for API purification (courtesy of AMPAC Fine variant. Optimization of PowerFeed can be
Chemicals, Rancho Cordova, CA, USA)
formulated as an optimal control problem [79,
114]. Moreover, VARICOL systems perform
asynchronous valve switching, where the four
force (LDF) model, where the mass transfer inlet/outlet ports are switched independently,
between the liquid and the adsorbent particles not simultaneously [115]. A comprehensive
is described by a linear relationship character- summary of modifications to SMB can be
ized by the mass-transfer coefficient. In the LDF found elsewhere [111, 116]. Since the number
model, the mass balance equations in the liquid of operating parameters is larger in such
and adsorbent phases are given by the following systems, finding the operating and design
two PDEs, respectively: parameters in such improved SMBs relies
on a systematic and efficient optimization
qCi ðx; tÞ qqi ðx; tÞ qCi ðx; tÞ
eb þð1
eb Þ þu ¼0 ð19Þ approach [117].
qt qt qx
Due to the complex dynamics of the chro-
matography, model-based feedback control
qqi ðx; tÞ
ð1
eb Þ ¼ Kapp i ðCi ðx; tÞ
Cneq ðx; tÞÞ ð20Þ of a batch process is very challenging. Fur-
qt
thermore, the poor observability of chro-
where eb is the void fraction, Ci ðx; tÞ is the matographic processes makes automatic con-
concentration in the liquid phase of component trol more difficult; although the temporal
i, qi ðx; tÞ is the concentration in the solid phase, concentration profiles at the outlet of the
u is the superficial liquid velocity, Cieq ðx; tÞ is the column can be observed by a detector, the
equilibrium concentration in the liquid phase, spatial concentration profiles inside the col-
and Kapp i is the mass-transfer coefficient, re- umns cannot be monitored directly. A few
spectively. In addition to the mass balance recent control strategies that tackle these
equations, the isotherm that describes the equi- challenges for batch processes can be found
librium between the liquid and adsorbent con- elsewhere [118–120]. For SMB chromatogra-
centrations must be specified. One of the most phy, there has been considerable advance in
widely employed isotherms is the Langmuir recent years in model-based control techni-
isotherm: ques including linear and nonlinear model
predictive controllers [121–123]. Further-
ai Cieq ðx; tÞ
qi ðx; tÞ ¼ : ð21Þ more, several experimental techniques for
1þbi Cieq ðx; tÞ
stable monitoring of the purity have been
To find the optimal design and operation of proposed which improve the controller
SMB and batch chromatography, several ap- performance [124].
 Process Systems Engineering, 9. Domain Engineering 25

3.3. Conclusion biofuels, which are the new sectors of industrial

(also called ‘‘white’’) biotechnology. Today
There are many exciting research challenges in there are significant new opportunities in white
the pharmaceutical industry where PSE techni- biotechnology for processes based on renew-
ques can contribute. This is being accelerated by able resources such as biomass and clean pro-
the recent changes in drug manufacturing initi- cesses with reduced solvent inventories, renew-
ated by the regulatory agencies. The introduc- able catalysts, and mild conditions for reaction
tions of PAT and QbD have brought a signifi- and separation [127]. In addition to the direct
cant impact to the PSE community, and current- process improvements, bioprocesses have also
ly many advanced PSE technologies such as frequently been justified on the basis that they
process modeling, control, optimization, and are processes with potentially lower environ-
decision-making support which can meet the mental impact than their chemical synthetic
unique demands from this industry are being counterparts. The main synthetic operations in
developed. bioprocesses include fermentation, microbial
catalysis, and enzyme catalysis (see Table 1).
Downstream options are dependent on the
4. Biochemical Engineering nature of the product (i.e., macromolecular or
low molecular mass compounds (‘‘small mole-
4.1. Introduction cules’’). Small molecules are frequently pro-
cessed in a similar way to other chemical pro-
Process systems engineering (PSE) offers many ducts, although dilute aqueous solutions bring
tools for the chemical engineer. Today, for specific problems which need to be addressed,
example, modeling, simulation, and process both from the viewpoint of process optimization
evaluation tools are routinely applied to design and the environmental footprint. For instance,
and optimization problems in the bulk chemi- the downstream processing of some small mol-
cals and fuels sector, where small process im- ecule bioprocesses could include large amounts
provements yield significant economic returns. of organic solvents for extraction from aqueous
In recent years there have been an increasing solutions. In these cases, the organic solvents
number of bioprocesses implemented and these require processing, recycle, control and ulti-
provide a different type of challenge for PSE. mately safe disposal. Macromolecules require
This chapter has a focus specifically on biopro- more specialist operations such as filtration or
cesses, and especially on the use of PSE tools for chromatography. However, in all cases the mo-
the design and development of bioprocesses. lecules are frequently sensitive to extremes of
Bioprocesses have found application in the pH and temperature, placing specific restric-
production of high-value products such as phar- tions and constraints on processing methods.
maceuticals (and their intermediates). The pro- Biocatalyst recovery (frequently for recycle)
cess engineering emphasis in these cases is on also necessitates filtration and centrifugation.
rapid process implementation, rather than opti- It is clear from the above that a range of
mized development [125, 126]. However, in questions need to be addressed when imple-
recent years bioprocesses have also been in- menting new processes, and specifically biopro-
creasingly applied to bigger volume products cesses. For example: When should a bioprocess,
such as fine chemicals, bulk chemicals, and rather than a chemical process be implemented?

Table 1. Process features of the three major types of bioprocess for chemicals production (I) represents options for immobilized enzyme

Feature Fermentation Microbial catalysis Enzymatic catalysis

Yield on substrate low medium high

Catalyst recovery and/or recycle potentially via a continuous process recycle via filtration via immobilization
Reactor options stirred tank stirred tank stirred tank
bubble column bubble column packed bed (I)
fluidized bed (I)
membrane
26 Process Systems Engineering, 9. Domain Engineering

If a bioprocess is to be implemented, can the namics, Control, Monitoring, and Identification,

existing infrastructure (feedstock, utilities, and Chap. 2 and will therefore not be discussed here.
plant) be used? How can a process plant be Process control issues are highlighted
adapted for different biomass sources available in ! Process Systems Engineering, 5. Process
in different geographical regions? What is the Dynamics, Control, Monitoring, and Identifica-
optimum biorefinery? What options exist for tion, Chap. 2 and ! Process Systems Engineer-
process integration? What are the environmen- ing, 5. Process Dynamics, Control, Monitoring,
tal, health, and safety issues of bioprocesses in and Identification, Chap. 3 for continuous and
comparison with chemical processes? What is batch/fed-batch processes, respectively. The
the environmental footprint of a bioprocess chapter only pays limited attention to data-driv-
compared with its chemical counterpart? How en modeling, where applications of data-driven
can bioprocesses be designed to maximize pro- models in the area of soft sensors is highlighted,
cess efficiency, minimize environmental im- since this is one of the areas where probably
pact, as well as maximize sustainability? Which data-driven models will become increasingly
variables can be measured on a bioprocess? important in the future. The main focus of this
How can control contribute to more efficient chapter, however, will be on mechanistic mod-
operation of the bioprocess? els, and on the current and future use of those
Many of these questions can currently be models within the design of sustainable biopro-
addressed qualitatively, but to have real value cesses. This is likely to be one of the most
it is necessary to assess the questions on a promising R&D areas at this moment, where
quantitative basis. In order to achieve this ef- PSE methods and tools will contribute tremen-
fectively therefore, computer-based tools are dously to embed design principles during the
required. In addition, models are needed. Over early stages of process development and design.
the last decades, PSE has already developed Future trends are highlighted as well, where
many of the appropriate tools, and those tools relevant.
often rely on models. Nevertheless, some fur-
ther developments are required. For example, in
the case of bioprocesses an extra option avail- 4.2. Industrial Biotechnology
able to the engineer is the improvement of the Processes
catalyst itself. This requires models which take
into account catalyst properties. In addition, one Before describing the PSE tools and some of
emerging consideration is measuring the rela- their applications in more detail, it is important
tive sustainability of processes, and one can see to highlight the most relevant industrial biopro-
life cycle inventory and assessment (LCI/A) cesses. Three major types of bioprocess can be
modeling tools and methods as a logical exten- identified dependent on the nature of the biocat-
sion of PSE. LCI/A methodologies allow for the alyst. These are outlined in the following sec-
estimation of environmental impact across the tions and the key process features are summa-
entire life cycle of a process or product. LCI/A rized in Table 1.
estimations rely heavily on the characterization
of the process and its unit operations using
modeling and simulation techniques, which are 4.2.1. Fermentation Processes
key competences within PSE.
This chapter will first introduce the three For a significant number of chemicals, the use of
different types of bioprocesses that are of in- fermentation has become a standard alternative
dustrial relevance. PSE methods and tools can to fossil-based feedstocks and technology. Nev-
be applied in the bioprocess design phase–using ertheless the possibility of growing microbial
process models and design PSE tools–as well as cells on a variety of sugars (derived from re-
to improve the process operation–using process newable biomass) has reinvigorated interest in
monitoring and control methods and tools. Pro- this area. The consequence is that fermentation
cess monitoring, with focus on applications at a large scale will become more common in the
on bioprocesses, is already summarized in ! future chemical industry. Many different types
Process Systems Engineering, 5. Process Dy- of fermentation process (using different strains
 Process Systems Engineering, 9. Domain Engineering 27

Figure 20. Alternative process scenarios for use of microbial cells for nongrowth associated biocatalysis
a) Route 1: Combined fermentation and microbial catalysis; b) Route 2: Fermentation separated from microbial catalysis; c)
Route 3: Fermentation separated from microbial catalysis with intermediate processing to change catalyst concentration

to produce different products) can take place in suitable media for effective product recovery
the same process plant which is a significant chosen. For processes requiring oxygen it can
advantage. The plant is relatively simple and the be highly important to select the optimal cell
challenges lie in adequate mixing (sometimes concentration in order to avoid oxygen-transfer
with materials having complex rheology), suit- limitations. Several tools are now available for
able oxygen input (for aerobic processes), and evaluating the oxygen supply issues in such
process control. Downstream, the separation reactions [129, 130]. The three potential routes
process depends on the product, but will nearly using microbial catalysis are shown in Figure 20.
always need to avoid high temperatures and
extremes of pH. The solvent in such a process
is water, meaning that the dilute product stream 4.2.3. Enzyme Processes
combined with the presence of many other
products presents a significant process engi- The presence of so many products at the end of
neering challenge. Both, large molecular mass fermentation or microbial catalysis is a conse-
and low molecular mass products can be made quence of the complexity of cells, where many
by fermentation. The processes either focus on enzymes catalyze reactions giving a spectrum of
low molecular mass products which can subse- products as well as decreasing the yield of the
quently be used as platform chemicals or bio- desired product on the reactant. An alternative,
fuels or high molecular mass compounds such for short pathways, is to isolate the enzymes and
as enzymes (for application in a range of indus- then immobilize them on a solid support or
tries, including detergents, textiles, and food behind a membrane or via aggregation, such
ingredients) or therapeutic proteins. that they are large enough and have the right
properties to be recycled (like a heterogeneous
catalyst). In this way a yield of product on
4.2.2. Microbial Catalysis catalyst of around 5–10 t/kg immobilized bio-
catalyst can be achieved, which typically is
In fermentation, by definition, the catalyst is sufficient to enable commercial implementation
growing during the process. This means that at an industrial scale. Such an approach has been
some of the reactant (or substrate) will inevitably widely used to assist in the synthesis of high-
be diverted from the product towards the cata- value compounds such as pharmaceuticals and
lyst, lowering the yield. An alternative (for a more limited number of well-known lower-
nongrowth associated products) is to grow the value products such as high fructose corn syrup
cells first and subsequently carry out the reaction (HFCS). Many of these processes have also
to increase the yield. This also enables the been modeled [131]. Enzymatic processes can
possibility of growth and reaction on different also be carried out using soluble enzymes (i.e.,
substrates (reactants) or under different condi- nonimmobilized), although these present chal-
tions (such as temperature or pH) in each stage. lenges in terms of separating and recycling the
Likewise the optimal cell concentration for con- catalysts (enzymes) when compared with the
version can be selected [128] after growth and immobilized enzyme processes.
28 Process Systems Engineering, 9. Domain Engineering

Traditionally, single enzymes were used for interested in investigating the process perfor-
catalysis. However, multiple enzyme mixtures mance under different operating conditions on
are nowadays becoming attractive for catalyz- the basis of simulations with a dynamic model.
ing the production of several compounds at an Sufficient process knowledge is a necessity
industrial scale. A classification of multien- in order to optimize the design or the operation
zyme-catalyzed processes was proposed recent- of a bioprocess. A mechanistic model captures
ly [132], including reaction and process con- that process knowledge in a structured
siderations for mathematical modeling of mul- way [134]. The model therefore has great value
tienzyme processes operated in a single reactor. in planning experiments, or in determining
which critical process variables necessitate tigh-
ter control [135–137].
4.3. Modeling of Bioprocesses Within the fermentation field, development
of mechanistic models has a long history. Early
Mechanistic and empirical (data-driven) model- work involving quantitative descriptions of bac-
ing approaches complement each other. Biopro- terial growth dates back to the 19th century, and
cesses are usually represented by a combination a broad spectrum of modeling techniques
of both where mechanistic models gradually are available today [138]. For more detailed
replace empirical models as more knowledge reviews on models and model types, a number
about a process or a unit operation becomes of key publications can be suggested on fermen-
available [133]. Empirical models represent tation models [139–142], on specific modeling
input–output relationships in a data set without approaches for enzyme production kinet-
requiring detailed knowledge of an underlying ics [143], on mechanistic model studies for
mechanism. Usually, an empirical model can biocatalytic processes [144], and on modeling
only accurately predict conditions represented cellulase kinetics [145]. Finally, the opinion
by the data set that was used to build the model. article of [146] highlights both the history and
Empirical models are useful in a process control the future of mathematical modeling in bio-
context, where software sensors often rely on chemical engineering.
empirical models for the prediction of variables Assuming a homogeneous reactor environ-
that are not measured directly due to on-line ment, a generally accepted classification of
measurement difficulty or excessive sensor cost mechanistic models of cell populations is pre-
as will be explained in more detail in the fol- sented in Figure 21 [139, 146]. Moving between
lowing sections. models in Figure 21 is determined by the as-
sumptions behind the mechanistic model. For
example, if the assumption of a homogeneous
4.3.1. Modeling of Bioprocesses– reactor environment does not hold, then a dis-
Mechanistic Models tributed model is needed (i.e., a model where not
only time, but also space (1-, 2- or 3D), forms an
Mechanistic process models (! Biotechnolo- independent variable). Unsegregated models
gy, 5. Monitoring and Modeling of Biopro- are common, and rely on an average cell de-
cesses, Section 5.3) for fermentation and bio- scription. Unstructured unsegregated models
catalytic processes are developed based on are the simplest models. They use a single
mass, heat, and momentum balances, supple- variable to describe the biomass [134, 142, 147].
mented with appropriate mathematical formu- Unsegregated structured models form an
lation of mechanisms (e.g., kinetic expressions important class. The distinguishing feature of
to reflect process dynamics). Specifically for the these models is that they describe the biomass
description of bioprocesses, the kinetic expres- as consisting of several variables (such as
sions themselves are often empirical, providing NADH, precursors, metabolites, ATP, bio-
a simplified and idealized view of a complex mass), and have been used for modeling com-
biological mechanism. Unlike empirical mod- plex processes. An example is a structured
els, mechanistic models usually offer better model of yeast intracellular metabolism [141].
extrapolation capabilities, which is critical in Morphologically structured models [141, 148]
a process modeling context, where one is often distinguish between different regions of the
 Process Systems Engineering, 9. Domain Engineering 29

Figure 21. Classification of mechanistic bioprocess models

hyphal elements of filamentous fungi, and were 3), especially for handling multivariate data
specifically developed to describe growth of this which are increasingly becoming available on-
important class of production organisms. line. In this contribution, the focus is on the use
Segregated models consider individual cells. of data-driven models in software sensors. The
They were developed in recognition of the fact estimation of the concentration of analytes of
that cells in a population–a pure culture–are interest using so-called ‘‘software sensors’’ is in
different, and are most often formulated as a many cases a fruitful alternative to direct (or
population balance model (PBM). An unstruc- analyte-specific) measurements using, for exam-
tured segregated model characterizes cells by ple, chromatographic and spectroscopic meth-
one distributed property (i.e., cell size or age of ods, as illustrated in a recent review on methods
individual cells [149]) without considering in- that allow the on-line measurement of the cell
tracellular composition. Obviously, structured mass concentration [152] ! Biotechnology, 5.
segregated models are more complex, since the Monitoring and Modeling of Bioprocesses, Sec-
distribution of one or more intracellular vari- tion 3.2: software sensors were considered as one
ables is also considered. Solving the resulting alternative approach that can compete with
multidimensional PBM is difficult, unless the methods such as dielectric spectroscopy, OD,
intracellular state can be captured with just a IR spectroscopy, and fluorescence for in situ
few variables [140]. One alternative to PBMs is measurement of the cell mass concentration.
cell ensemble modeling [140, 150], where the Software sensors can in fact be divided into
parameters of a single cell model are random- three classes:
ized to generate a cell population. Despite the
fact that segregated models are more complex, . Software sensors based on stoichiometry,
advances in data collection methods mean that it elementary balances and first-principles
becomes more and more relevant to develop models
segregated models in order to increase our . Software sensors based on data-driven model-
understanding of the complex interactions be- ing methods, i.e. black-box approaches
tween individual cells [151]. . Hybrid software sensors, which are partly based
on first-principles, but supplemented with
black-box approaches for parts of the system
4.3.2. Modeling of Bioprocesses–Data- that are not sufficiently well understood
Driven Models
Only the second class of software sensors
Data-driven models are extremely useful in a will be highlighted here, and these seek to
process monitoring and control context (see exploit correlations between the variables in
! Process Systems Engineering, Process Sys- the process, without seeking any mechanistic
tems Engineering, 5. Process Dynamics, explanation for the observed correlations. Meth-
Control, Monitoring, and Identification, Chap. ods such as artificial neural networks (ANN)
30 Process Systems Engineering, 9. Domain Engineering

and chemometric modeling techniques, for trics [158] and integrated with life-cycle analy-
example, partial least squares (PLS), belong sis. This will need to include evaluation of
here [153–156]. In a recent publication, a mov- feedstocks and products as well as processes,
ing window principal component analysis including energy and mass integration. This
(MW-PCA) method was successfully used to presents a fascinating set of alternative routes
identify phase changes in several different in- and technologies from a given feedstock and/or
dustrially relevant batch processes [157]. The to a given product(s). PSE has a particular role to
MW-PCA method was solely based on changes enable such evaluations on a quantitative basis,
in the statistical properties of on-line data, such not only from the process perspective, but also
as pH, dissolved oxygen concentration, agitator from the wider sustainability aspect. PSE also
speed and concentrations of CO2 and O2 in the brings the advantages of rapid computational
exhaust gas, and appears to be useful for any methods. Such simulations enable alternatives
case where even slight changes in process prop- to be quickly evaluated. The answer in a specific
erties must be identified. case to the problem formulated here will in
The main message is that black-box methods addition depend on regional factors. Feedstock
are certainly useful, on condition that the user is availability and cost is highly dependent on
aware of the limitations of the methods. Moreover, geographical location. A parallel set of evalua-
some of these methods have the advantage that tions concerning the need to retrofit existing
they can be applied rather easily in practice. Soft- plant, or build new plant, is also required.
ware for building chemometric models, to name
one example, is available from several software
vendors or can be downloaded as freeware. 4.4.2. Evaluation of Platform Chemicals

While the increasing cost of oil is driving partic-

4.4. The Role of Process Systems ular interest in the production of new biofuels
Engineering from biomass there is little doubt that today of
equal importance is the production of chemicals
4.4.1. Evaluation of Process Options from biomass. Indeed for the supply of fuels in the
future there are many potential sources aside from
For some higher value products a bioprocess biomass. In a world with limited (or very expen-
may in some cases be the only route to a given sive) oil it is less clear where the chemicals of the
product (to ensure correct folding of a therapeu- future will originate. There is currently an existing
tic protein, for example, or the synthesis of an infrastructure based on the use of the 7 established
optically pure pharmaceutical intermediate). platform chemicals (toluene; benzene; xylene;
However, the more usual situation is that there 1,3-butadiene; propene; ethylene; methane). In
are other competing routes to the same product. the short term one could consider if we can use the
Therefore, for now, biotechnology is just one of same infrastructure and just create the 7 chemicals
a number of options for the production of che- from alternative sources. However, in the longer
micals and biofuels. The economic drivers for term it will be necessary to devise new processes
implementation depend on existing infrastruc- based on a different set of platform chemicals.
ture, feedstock costs, feedstock availability as One group will be based around glucose (the
well as the efficiency of the relevant (bio)cata- hydrolytic product of starch and cellulose and
lyst and (bio)process technology. At the same therefore readily available from biomass). In a
time, there are environmental drivers and, in the biorefinery it will be necessary to develop a
wider sense, sustainability drivers for the selec- structure which can manage a range of feedstocks,
tion of different process alternatives. The sus- a range of technologies, and a range of products.
tainability drivers are by no means simple, as This presents a considerable challenge for design
they require the balance of different sets of goals and optimization as well as process integration.
and metrics that can present trade-offs in some An interesting example which illustrates the com-
cases. Objective functions to be optimized plexity and the challenge that lies ahead is the use
should not exclusively be based on economics of glucose or fructose to produce 5-hydroxy-
but increasingly also on sustainability me- methylfurfural (HMF) or 2,5-furandicarboxylic
 Process Systems Engineering, 9. Domain Engineering 31

acid (FDA) [159]. Greatest value is obtained by biomass utilization to provide a range of
going the whole way from glucose to FDA. sugars (for subsequent (bio)catalysis or fermen-
However even in this small reaction pathway tation) and oil-based material (from biomass)
there are many alternative technologies. Some ! Biorefineries–Industrial Processes and Pro-
can be integrated together, some give the required ducts. In each case the current research empha-
yield and selectivity, some are difficult to imple- sis on biorefineries is to ensure that all the
ment and others are untested at scale. This illus- fractions of a particular biomass in a given
trates very well the challenge that design engi- situation are fully exploited. Likewise the de-
neers face. velopment of downstream products is now be-
ing explored. For example, glycerol (as a by-
product of biodiesel production) can be used as a
4.4.3. Process Integration
platform chemical (e.g., via fermentation to
produce 1,3-propanediol). Another interesting
The solvent for most bioprocesses, with a few
example concerns the production of bioethanol.
exceptions, is water. Consequently the down-
This is widely developed as a biofuel although
stream process is frequently difficult and this is
there is considerable economic incentive for
exacerbated by the need to carry out separations
developing a range of other products (e.g.,
at moderate temperatures. Given the dilute na-
acetic acid) from ethanol, in other words using
ture of the streams it is frequently the case that
it as a platform chemical [160].
the majority of the costs and environmental,
health, and safety impacts are therefore in the
downstream process. For instance, in some fine
4.4.5. Biocatalyst Design
chemical and pharmaceutical applications of
biocatalysis large amounts of organic solvents
A particular feature of bioprocesses is the use of
may be used in the purification of a biocatalytic
biocatalysts ! Biocatalysis, 1. General, which
reaction. The dilution of the streams has histori-
may exist in several forms as indicated earlier
cally also driven the need for energy-intensive
and where options exist for modification. At the
separation. In the case of transport fuels removal
simplest level as a protein (isolated enzyme), the
of water becomes an essential requirement to
options for swapping amino acids via protein
reduce costs and avoid transporting significant
engineering exist. New enzymes which have
amounts of water. For example, in the case of
been modified may display new tolerance to
ethanol which forms an azeotrope, this can be a
reactor conditions such as temperature or pH
significant cost. In other cases the product may
and may also have improved selectivity or
be integrated within a biorefinery although at
reactivity (activity) on a given (nonnatural)
some point water will need to be removed.
substrate or reactant. Order-of-magnitude
Consequently the integration of water use and
improvements have been found in a number of
reuse via recycle is an essential part of the design
cases although understanding the most effective
of industrial bioprocess facilities. In addition,
method of making changes to the enzyme de-
bioprocesses need to be designed with process
pends on past precedent and, to some extent,
synthesis and process integration approaches,
structural knowledge [161]. In the case of mi-
thus avoiding a process that is efficient in one
crobial catalysts, individual enzymes can be
part and inefficient in another. Existing tools of
over-expressed (increasing reaction rate of a
mass and energy integration such as pinch tech-
given cell) and the regulatory control scheme
nology (! Pinch Technology) will have an
fixed to direct the carbon to give improved rates
important role. The issue of water use in a
and yields (via metabolic engineering). Some
biorefinery is in many ways analogous to the
start has also been made to the development of
issue of heat use in a conventional refinery.
pathways where enzymes coming from a variety
of sources are cloned into single host to make a
4.4.4. Biorefinery Design new pathway via a combination of genetic
engineering and de-novo pathway engineer-
Two major types of biorefinery have been iden- ing [162]. In all these areas it is clear that those
tified for the future, based on lignocellulose involved in PSE need to inform the biological
32 Process Systems Engineering, 9. Domain Engineering

engineers about what is required in a given case LCA methodologies for a more objective anal-
and set suitable targets. Philosophically it is ysis of their ‘greenness’ or sustainability. This
interesting to note that process implementation would allow to develop bioprocesses that are
may come via process improvements or alter- sustainable by design, in such a way that they:
natively via catalyst improvements. In many
cases both will be required. Understanding the . Optimize the use of material and energy
necessary balance between these areas, as well resources
as their integration with each other will be . Eliminate or minimize environment, health,
important for the future development of the and safety hazards by design
field. PSE is particularly powerful in its ability . Integrate life-cycle thinking in the design
to predict and can therefore be used to direct
decision-making and process development. Analyzing and comparing sustainability will
require a comprehensive assessment that bal-
ances the three different spheres of sustainabili-
ty (see Fig. 22). This can only be achieved
4.5. Assessing the Sustainability through a multivariate optimization that will
of Bioprocesses account for environmental performance, eco-
nomic viability, and social responsibility
As discussed before, bioprocesses have fre- (which include health and safety aspects).
quently been highlighted as greener chemistry Another important concept when assessing
or engineering, since they address many of the the efficiency and the sustainability of processes
green chemistry and green engineering princi- is the differences between new process perfor-
ples [163] by offering reactions that are poten- mance and retrofit performance. For instance, in
tially more atom economic, operate under mild comparing the sustainability or performance of
conditions, use mostly nonhazardous chemi- a well established process with a new biopro-
cals, and have less protection/deprotection cess, a situation that one often encounters is the
steps. However, this tends to be true mostly fact that initially, the new process may not have
when looking at the reaction part of the process, the same level of performance as the established
in other words, the biocatalysis. However, it technology, mainly because they are at different
cannot be generalized when analyzing the entire points in the development curve, and therefore
process that in some cases may include the use the new process is suboptimal. On the other
of a large amount of organic solvents for down- hand, the established process can be retrofitted
stream processing. This is one of the areas in to improve its performance. Retrofitting and
which systems engineering, in conjunction with new process development is not a new concept
a transparent application of life-cycle inventory from the systems engineering viewpoint. How-
and assessment (LCIA) methodologies can and ever, additional modeling work is needed to
must play a pivotal role. estimate the achievable performance limits of
Determining whether a process is sustainable a fully developed process and an established
or green is by no means a simple feat. It is more process that undergoes retrofit. This will allow
akin to a multivariable optimization that is very more meaningful comparisons without unnec-
familiar to systems engineers, and for which essarily penalizing the new process for its lack
several proposals and methods have been pre- of development, nor the established process for
sented [164–166]. For an objective assessment the lack of timely retrofitting.
of the sustainability of a process, there is the In the recent past there have been many
need to utilize the tools that system engineers attempts to measure the ‘greenness’ of synthetic
have developed during recent years and apply routes, and the approaches have generated a
them with a life-cycle approach. It is necessary series of ‘green metrics’. Most of the ap-
to move from the basic analysis of the biocatal- proaches have searched for a simple metric in
ysis alone and discrete unit operations (separa- an attempt to provide a low resolution view of
tions) and use a whole system engineering how green is a given process. ‘‘E-factor’’ was
approach instead. This implies utilizing multi- one of the first measures of greenness proposed
variate optimization techniques coupled with to highlight the amount of waste generated in
 Process Systems Engineering, 9. Domain Engineering 33

Figure 22. The three spheres of sustainability through a triple bottom line assessment

order to produce 1 kg of product [167, 168]. For instance, a technology comparison

This metric, while simple to understand, has framework [181] that accounts for environment,
several key drawbacks by focusing on waste health, safety, and life-cycle assessment im-
instead of efficiency, neglecting a view of the pacts, was used to compare the established
type of waste generated, not accounting for the chemical process and a two-enzyme biocata-
relative impacts, and the lack of life-cycling lytic process for the production of 7-ACA [158].
thinking. For instance, bioprocesses have in The conclusion of this assessment was that the
general large amounts of wastewater produced, bioprocess was greener when compared with the
which would indicate an extremely large E- chemical process. This was driven by the fact
factor, but at the same time the actual impact that the chemical process uses more hazardous
of the effluent might not be as large as an E- materials, requires about 25% more process
factor would suggest, since the waste is rela- energy than the enzymatic process and, has a
tively benign. Additional metrics have focused larger life-cycle environmental impact: it uses
on efficiency, especially on mass efficiency (or approximately 60% more energy, 16% more
its inverse, mass intensity) which addresses the mass (excluding water), has double the green-
first part of the disadvantages of the E-fac- house gas impact, amongst others.
tor [169, 170]. However, it has been widely However, although this type of assessment is
recognized that measuring the sustainability or useful and more common as time passes, there is
even the ‘greenness’ of a process is a multi- an ongoing need for modeling methodologies
objective optimization problem that must take that will seamlessly integrate sustainability fac-
into consideration the efficiency of the entire tors during bioprocess design and development.
process regarding the use of mass and energy, Systems thinking and systems engineering are
the environment, health and safety characteris- the skill and the discipline that will need to play
tics of materials, and the inclusion of life-cycle an important role to make this happen, and a
thinking amongst others [171]. holistic view of bioprocesses and their interre-
Some efforts have been made to comprehen- lations will be imperative.
sively address and compare the sustainability of
processes in general and bioprocesses in partic-
ular. These methodologies have been employed 4.5.1. Life-Cycle Inventory and
in some instances to assess the sustainability, Assessment
environmental, health and safety aspects of bio-
processes [172–181]. These methodologies at- One of the tools to analyze systems holistically
tempt to measure the sustainability of biopro- is LCIA (Fig. 23). LCIA is a methodology used
cesses and in some instances they compare to evaluate the environmental profile of an
bioprocesses with their chemical alternatives. activity or process from the extraction of raw
34 Process Systems Engineering, 9. Domain Engineering

Figure 23. Life-cycle assessment to evaluate the environmental profile of a process from the extraction of
a) Raw materials; b) Production; c) Transportation; d) Sales, distribution, and use; e) Final fate

materials to its end-of-life. The resource con- systematic and holistic method to evaluating
sumption and emissions are inventoried and waste impacts, and the nuances of renewabili-
assessed from the extraction of raw materials, ty [173, 174, 178, 179]. For instance, a compar-
production, transportation, sales, distribution, ison of a process using metal catalysts and one
use, and final fate. Depending on the goal and using biocatalysts for the enantioselective re-
scope of the assessment, the boundaries can be duction of ketoesters in pharmaceutical synthe-
set differently; for instance a ‘cradle-to-gate’ sis was performed using a streamlined LCIA
assessment might be adequate when comparing methodology. The analysis identified some
two processes to the same product; or a ‘gate-to- processes and reaction conditions that had the
grave’ boundary may suffice when comparing largest significance on the impact of the synthe-
two different end-of-life technologies. The sis. It was also concluded that whether the
results of these assessments can be reported as metal catalysts were better than bio-catalysts
direct inventory data (for example life cycle depended mainly on the work-up from the use
energy, life cycle mass, life cycle emissions), of organic solvents and energy-intensive
measures of individual potential impacts (such steps [172].
as global warming or acidification), or as an Developing life cycle inventories and asses-
aggregate score or index for high-level compar- sing the LCIA impacts of bioprocesses is not a
ison (for example Eco-Indicator 99). LCIA simple endeavor given the large amount of data
methodologies are described in detail in the needed from different sources. The more mate-
literature [182–188]. LCIA methodologies are rials are involved in the bioprocesses will re-
in a way an extension of systems engineering quire more life cycle inventory data to be col-
and provide a directly applicable framework to lected, verified and analyzed. On the other hand,
assess the sustainability of processes. the life cycle inventory data for biomaterials is
In the area of bioprocesses, the application of not always available. There have been efforts to
LCIA is still not a widespread practice. There increase the body of knowledge of life cycle
are however, examples on how several practi- inventories and impacts of bioprocesses and
tioners have applied LCA metrics primarily materials either derived from biomass or needed
using case studies to better understand the wider in bioprocesses. These challenges have influ-
environmental implications of bioprocesses and enced the development and use of streamlined
to compare them with chemical routes. This life cycle assessment methodologies and ab
type of assessment has provided some key in- initio modeling approaches to estimate
sights, such as the role of separations, a more the life-cycle impacts of bioprocesses and
 Process Systems Engineering, 9. Domain Engineering 35

bio-derived materials. This is precisely the sort . More routine application of multiobjective
of opportunity where system engineering can optimization techniques for sustainability as-
add value, as the development of reliable, con- sessments of bioprocesses.
sistent, transparent, accurate and easy-to-use . Enhanced understanding of the interactions of
modeling and streamlined techniques for LCIA the environmental, social and economic as-
will continue to be an important need to be able pects of bioprocesses for a holistic sustain-
to routinely assess the sustainability of ability view.
bioprocesses.
The development of true sustainability Addressing these modeling and process un-
assessments, with an embedded LCIA ap- derstanding needs will make it possible to inte-
proach will be necessary aligned with the grate sustainability principles into process de-
early modeling needs highlighted in this arti- sign and development in a far more rigorous
cle. In order to routinely assess sustainability manner.
of bioprocesses and to embed sustainability
principles into the bioprocesses design and
development, the following modeling needs 4.6. Future Outlook and Perspectives
can be highlighted:
The development of new bioprocesses as a
. Better deterministic models of unit operations complement to existing chemical and fuel
that are part bioprocesses, such as fermenta- production is an exciting endeavor that will
tion, biocatalysis, etc. This would need to occupy many process engineers in the future.
include fundamental design parameters to There will be a particular role for process
design more resource efficient bioprocesses. systems engineers in this developing sector
. Development and enhancement of property with the advantages of quantitative decision-
prediction packages that would facilitate making tools and rapid simulation that this
estimations of resources (e.g., energy require- brings, including process design and sustain-
ment) and the utilization of optimization ability principles. In the future suitable models
techniques. will inform developments at the infrastructur-
. More extensive use of process integration al level (evaluation of biorefineries, feed-
techniques on bioprocesses, especially at the stocks and integration), the process level
development phase. (evaluation of alternative technologies and
. Increased use of software sensors on biopro- process integration) and the catalyst level
cesses, in order to maximize the information (alternatives for protein and metabolic engi-
content that is available on-line. Closely neering). In addition, these models will allow
related to that, increased use of advanced the integration of sustainability principles into
control and monitoring methods such that process design and development.
bioprocesses can be operated as close as The further development of PSE tools
possible to the optimum. (including property prediction packages and the
. Better understanding of life cycle inventory development of a database for bio-based mole-
and impacts of bioprocesses and bio-derived cules) will be required. To routinely assess
materials. sustainability of bioprocesses will require as
. Increased understanding of the uncertainties well more robust and transparent environmental
in modeling bioprocesses, both from the pro- life cycle inventory databases of bio-derived
cess design and the sustainability assessment materials; as well as better modeling and under-
viewpoints. standing of the social and economic aspects of
. Improved consistency and transparency of sustainability and their relationships. Finally, an
LCIA methodologies as applied to increasing dialogue amongst the biochemical
bioprocesses. engineers, biologists and other related areas of
. Improved streamlined LCIA methodologies expertise will be necessary to enable the vision
that are easy to use by academia and industry of sustainable industrial biotechnology to be
alike. fully exploited.
36 Process Systems Engineering, 9. Domain Engineering

5. Policies and Policy Making . A policy is a purposeful connection of ends

with means [193]
5.1. Introduction
In the classical point of view a policy is the
Process engineers are becoming increasingly product of rational choice; this assumption is
aware that there is a significant role for them also going to be adopted for the rest of this
to play in the arena of policy making. This is not discussion, although the adequacy of this model
to say that there have not been past contributions is in dispute [194].
of PSE in the policy area nor that engineers have Policies are constituted by combinations of
been insensible to the effects on society of the policy measures (also known as policy instru-
systems they design and operate. However, an ments). According to [192] policy measures can
ideological divide between the technical and be of different types:
social sides of a given problem has tended to
reinforce the perception that our exclusive role . Exhortation (e.g., education)
as engineers is to improve product quality and . Economic incentives/disincentives (e.g., sub-
production efficiency and to minimize costs and sidies and taxes)
risks, and that these contributions in themselves . Government provision
will automatically result in a net benefit to . Legislation/regulation
society. Such a perception is a result of a
historical tendency to compartmentalize pro- However, in all cases policy measures have a
blems, i.e., to dissociate the technical and social set of properties such as a degree of effective-
sides of a system, leaving the decision-making ness and a range of implementation costs and
and analysis of the social side to the social times. Other properties, more difficult to mea-
scientists and politicians. sure and quantify, but not less important, are
The split between the technical and social related to issues such as equity, legitimacy, and
camps has deep and old roots, as discussed in [, public support.
190]. Yet, it is increasingly accepted that socio- A policy cycle is a sequence of steps through
technical systems, i.e., systems composed by tech- which a problem is defined; alternative policies
nical artefacts and social arrangements (agents to address it are proposed, analyzed, and refined;
with a purpose), are closely interdependent and, and a proposed policy is selected, implemented,
as such, need to be designed and analyzed as a total and constantly challenged and revised [195].
system [191]. In recent years the trend has begunto
reverse towards the closing of the gap between
technical and social systems, because only by 5.3. Policy Making and the Systems
taking into account the characteristics of the social Approach
subsystems it will be feasible to develop effective
and sustainable engineering systems. The first author to use systems theory to explain
To investigate the relations–past, current, political processes was [196]. In his portrayal,
and possible–between process systems engi- political systems convert inputs, such as politi-
neering (PSE) and policy making it is conve- cal demands and public support, into outputs,
nient to start by providing some definitions for i.e. a group of resulting decisions and actions
policy and its components, and a description of (Fig. 24 A).
how policies are conceived, developed, and Policies themselves can be understood as
implemented. blocks having policy measures as inputs and a
set of desired outcomes as outputs [197], (Fig. 24
5.2. Policies and Policy Measures B). Thus, policies are systems composed of
interacting parts that are the means to reach a
The following two complementary definitions final objective (the ends), as a result they can be
of policy are adopted: represented through a causal model.
The concept of policy cycle, in particular, has
. A policy is a set of effective and acceptable its foundations in systems theory and is analo-
courses of action to reach explicit goals [192] gous to the concept of life cycle in engineering
 Process Systems Engineering, 9. Domain Engineering 37

Figure 24. Systems theory view of political and policy systems

A) Political system; B) Policy system

systems. The schematic policy cycle in Fig- application of formal analytical methods
ure 25, based on a model proposed in [198], such as simulation and optimization
depicts the process of policy decisions, their
implementation, enforcement, and evaluation.
The policy cycle in Figure 25 is an idealiza- The output from the policy formulation stage
tion, as some stages in the process are sometimes is the selection of a policy considered to be the
merged or altogether eliminated. The feedback most appropriate and thus the recommended
arrows intend to express that the process is one for implementation [190]. Policies are cre-
cyclical with backtracking and revision steps. ated during the policy formulation step as new
The rest of this analysis will focus on the third components of a sociotechnical system. For this
stage in the cycle (policy formulation) because reason, policy formulation has been character-
of its many similarities with conceptual process ized as a design activity [].
design in PSE. These similarities explain why During the synthesis step of policy formula-
most of the past efforts and the potential future tion a set of policy measures (the building
advances of PSE in policy making are to be blocks) are combined to configure alternative
found in the policy formulation area. policies; thus the policy maker has to decide
which policy measures to select taking into
account their intrinsic characteristics (such as
5.4. Similarities between Policy effectiveness, cost, etc.) and their interactions
Formulation and Conceptual Process with other policy measures, as in the left-hand
Design part of Figure 26 A. Process synthesis, in turn, is
the invention of a structure and its associated
Policy formulation is a procedure with two operating conditions for a new chemical
components (Fig. 26 A): manufacturing process [199]. Inventing the
structure involves finding the best process con-
1. Synthesis (generation) of alternative courses figuration (which building blocks to include and
of action (alternative policies) how to interconnect them) among a very large
2. Analysis of the alternative policies, i.e., the number of possible alternatives, as in the left-
estimation of their consequences to help in hand side of Figure 26 B.
the selection of the best policy alternatives. In both cases synthesis and simulation steps
This step is generally performed through the are applied in tandem and iteratively: a synthe-

Figure 25. Simplified policy cycle, based on [198]

38 Process Systems Engineering, 9. Domain Engineering

Figure 26. Analogy between

A) Policy formulation; B) Conceptual process design [200]

sis step generating alternative policies/flow- precise predictions are not feasible [190]. Sim-
sheets, and a simulation step evaluating each ulation can also be used to discover the rela-
of the alternatives so that they can be compared, tionship between states through time, i.e., the
and also informing the application of the next dynamics of the system [].
synthesis step in the loop (Fig. 26). In any of the above cases, simulation requires
Policy formulation and conceptual process the development of models that relate policy
design belong to the type of problems where alternatives to their impacts and the application
there is neither an a priori agreement about of such models to predict the impacts of the
which criteria to use nor a prior articulation of policies being considered [204], however this is
preferences [201]. Both are processes of coevo- easier said than done given the complexity of the
lution between what is possible and what is sociotechnical systems that have to be predicted
desirable and proceed by generating potential as will be discussed in Chapter 6.
solutions and evaluating them in a generation–
evaluation cycle. Thus the goals, criteria, and
the artefact being designed (policy or chemical 5.5. The Nature of Policy Formulation
process) evolve in a single front according to the
exploration model proposed by [202]. As a It is widely recognized that a one-for-all and
result there are no unique optimum policies, static policy is unlikely to achieve the desired
just satisficing ones [203]. goals. This is because [205]:
It has already been mentioned that alterna-
tive policies are evaluated during the analysis . A good policy groups together (packages) a
step of policy formulation (Fig. 26 A). This set of policy measures such that synergies are
evaluation entails the exploration of their im- achieved, negative impacts are mitigated, and
plications in terms of what they can accomplish, conflicts are resolved, thus ensuring that the
alongside any desired or undesired side effects. policy will address the problem effectively
Such a task can only be done by means of and equitably over the long term [204]. Alter-
simulation, whereby either a point prediction native policies are not equivalent in their
(the forecast of the state of the sociotechnical effectiveness, implementation costs, public
system at a particular point in time in the future) acceptance, risk, etc. All of these properties
or a set of event frequency distributions is are determined by the properties of its con-
produced. The analysis of policies is often stituent policy measures and their
useful for the insights it provides even when interactions.
 Process Systems Engineering, 9. Domain Engineering 39

. The precise nature and scope of policies are society as a whole, may take actions to subvert
geographically and culturally dependent giv- or circumvent it, rendering the policy
en the variability of resources, of access to ineffective [204].
technology, and of political constraints at
different locations and times.
. Even for a fixed time and place, the identifi- 5.7. Challenges for Modelers of
cation of a suite of alternative policies (rather Sociotechnical Systems
than a single ‘‘optimal’’ one), together with
clear indications of their trade-offs, is crucial Traditional policy analysis assumes that socio-
to accommodate the diversity of stakeholders’ technical systems reach static equilibrium, and
preferences because, after all, decisions about that they can be properly characterized and
desirable futures, and the policies to attain controlled. It has been argued that each of these
them, are essentially a question of social assumptions is true only in special circum-
values and political choice [206]. It is ac- stances [208]. As a result, there are a number
knowledged that no set of values or framings of challenges that modelers of sociotechnical
can definitively be ruled more rational, well systems have to face due to the unique nature of
informed or better than all others [207]. such systems.

5.6. The Nature of Sociotechnical

Systems 5.7.1. Multiple Stakeholders

Sociotechnical systems are systems that involve Policies involve multiple stakeholders with
the interaction of human beings with physical their own preferences, objectives, expectations,
infrastructure and, as in the case of purely and beliefs [209]. Policies also have implica-
technical artefacts, can also be designed. They tions for groups with no or little influence in the
are characterized for their sensitivity to initial decision-making and even for people that do not
conditions and for the complexity of the inter- currently exist, as is the case of future
actions between human actors (possibly mil- generations.
lions of them) and with heterogeneous physical Because stakeholders can be individual peo-
infrastructure. Because of their human compo- ple and organizations, the issue of individual vs.
nent, it is either impossible or prohibitively institutional behavior must also be taken into
expensive (in terms of time and cost) to perform account and included in the overall model; for
in vivo experiments on sociotechnical systems example, in the case of policies to reduce trans-
in order to inform the decisions taken during the port emissions there are many actors such as
formulation of policies. The sensitivity to initial vehicle users, vehicle and fuel manufacturers,
conditions and the complexity of interaction government agencies, and environmental
are two of the reasons that render accurate groups [204].
point predictions impossible, thus requiring Furthermore, the costs and the benefits of
to resort to mapping out a space of possible policies are not evenly distributed between the
states and estimate the relative frequency that stakeholders, as a result
any particular state will occur in a future time
window []. . Stakeholders have contradictory interests and
Policies are courses of action to design, plan, their interaction will often result in conflict
manage, or control complex sociotechnical sys- . The impact of a policy on a stakeholder con-
tems. In this sense the implementation of a ditions how they react to the policy
policy is an experiment but with unknown out-
comes; furthermore, and because of there is 5.7.2. Incommensurable Values
often a lack of time and resources, policy out-
comes are rarely formally monitored. The reconciliation of multiple incommensura-
A policy is also influenced by how the human ble values (values that are not measurable) is
actors respond to it. Individual actors, or the present in all public policy decisions [204].
40 Process Systems Engineering, 9. Domain Engineering

Values that are intangible have often been ally expect them to reduce uncertainty, but
monetized by assuming a hypothetical market policy analysis often increases uncertainty by
and then stating how much society would be identifying and raising new issues [190].
willing to pay either (i) to secure an improve-
ment, (ii) to prevent a loss, or (iii) how much
society would be willing to accept as compen- 5.7.5. Emergent Behavior
sation [210]. The monetized values can then be
merged during a cost-benefit analysis. Howev- An emergent property is one that cannot be
er, despite the fact that there are ethical pro- predicted from the knowledge of the system
blems involved in monetizing health, welfare, components, i.e., it is the product of many local
and environmental values [211], these issues are effects [212]. In practical terms this means that
not always suitably considered. the system is computationally irreducible (there
is no simple set of equations to represent it) and
that the only way to figure out its evolution is by
5.7.3. Externalities running the system itself [213].

Externalities are secondary or unintended con-

sequences, generally a nonmonetary cost or 5.7.6. Complexity of Causation
benefit, incurred by a stakeholder who did not
have a choice and whose interests were not The complexity of sociotechnical systems re-
taken into account [209]. There are different sults in difficulties to establish and represent
types of externalities that may have to be mod- causation. There are several reasons for
eled: spatial (consequences in different re- this [190]:
gions), intertemporal (consequences in different
generations), and social (consequences for dif- . There is a very large number of variables and
ferent social groups). relationships at different levels of granularity
and time dynamics [214]; in particular, the
behavior of complex systems is dominated by
5.7.4. Uncertainty interconnected positive and negative feed-
back loops [212].
Uncertainties are due to the long-term nature of . The effects of a policy are also shaped by how it
policies and the existence of unexpected im- is implemented and not only by its substance.
pacts, for example, when a sociotechnical sys- . The effects of a policy are modified as it
tem undergoes structural change. They are pres- receives feedback, i.e., its content has an
ent in three different forms [209]: interaction with its effects. For example, there
may be unintended effects due to the stake-
. Uncertainty about the model parameters and holders reactions to a policy by circumventing
initial conditions (which is usually addressed or subverting it.
through sensitivity analyses)
. Uncertainty about the model structure (rela- In fact, it has been argued that policies are
tions between variables) more often facilitative than causative, i.e., con-
. Uncertainty about the applicability of the sequences are as likely to depend on actions and
model, i.e., its level of granularity and time influences other than the policy itself [215]. To
scale, and about the selection of variables the extent that this is true, it can be said that the
effects of policies can only be influenced but not
Identifying and managing uncertainty is im- controlled [208].
portant because the risk of wide-range and long-
term hazards is proportional to uncertainty. At
the same time, uncertainties limit the applica- 5.7.7. Objectivity in Policy Analysis
bility for long-term forecasting; looking farther
into the future (a large horizon of analysis) Policy must be interpreted in order to be analyzed.
increases uncertainties. Users of models gener- For example, its objectives may need to be re-
 Process Systems Engineering, 9. Domain Engineering 41

defined in more concrete terms, decisions about Several studies on the application of models
the types of effects that will be examined have to during policy development have been pub-
be made, etc. However, this analysis is not value- lished; the list below is mainly based on [209]
free, as the purposes and mechanisms of analysis and [216] unless stated otherwise. The follow-
are conditioned by ideology and its conduct ing descriptors are common to many of the
shaped by the analyst’s concerns about the con- different categories of models in the list:
sequences of the analysis for society [190].
These issues raise ethical questions because . Mainstream: a well known and widely used
the idealism of analysis as a pure scientific activi- type of method.
ty gets mixed with the impulses of the analyst as a . Descriptive vs. normative: a descriptive mod-
citizen, reflecting ideology and values. el is one that given some inputs will produce
some outputs, i.e., models used for simula-
tion. A normative model suggests how things
5.8. Types of Models Used in the ought to be (as opposed to a descriptive
Analysis of Policies model, which describes how things are).
. Aggregated: the behavior of a system is taken
Models can be used to explore, describe, ex- as an average of the individual component
plain, and predict the behavior of a system [190]. behavior; although it is possible to avoid it,
Effective policy formulation, in particular, de- mathematical representations tend to use ag-
pends on the understanding and modeling of gregation. Averaging is inappropriate for the
sociotechnical systems to forecast their future representation of emergent behavior.
behavior, evaluate the likely impacts (econom- . Mechanistic: mechanistic models use mathe-
ic, environmental, social) of alternative poli- matical equations to simulate a system and
cies, and inform the decision of which policy predict its future state; they are useful in the
measures to adopt. All of these analysis tasks are understanding of the workings of a system but
part of the so-called policy assessment, i.e., the not very reliable for prediction. Mechanistic
comparison of alternative policies using policy models tend to be aggregated.
instruments as inputs and measurable indicators
(such as CO2 emissions, cost, etc.) as outputs. 5.8.1. Macroeconomic Models (Main-
Unfortunately, while it is relatively easy to fit stream, Descriptive, Aggregated,
models to historical data, models are not as good Mechanistic)
in predicting the future.
Many types of models have been used in The general equilibrium models advocated by
policy making due to two factors: neoclassical economic theory are the most pop-
ular type of economic models in use. They
. The development of models used during poli- consist of systems of equations and are based
cy making has attracted the participation of on two premises:
modelers from many different disciplines,
each one bringing their own methods and tools . The economy behaves as a linear mechanical
. The systems to be modeled are complex and system that tends to a stable equilibrium
present a number of challenges, as have been . Human behavior is assumed to be rational and
described in Chapter 7. These challenges can independent, and can be represented by aver-
only be reasonably overcome through the aging the behavior of consumers and
concurrent application of many methods or producers
by restricting the modeling to parts of the
overall system This type of models are used to forecast the
economic impact of policies on the equilibrium
Most modeling methodologies are quantita- of the system; however, it has been argued that
tive, but qualitative analysis is also useful for there is little evidence to suggest that they have
certain aspects of policy making such as much predictive value [217], perhaps because
scenario building. both of the basic assumptions are unrealistic.
42 Process Systems Engineering, 9. Domain Engineering

The parameters in macroeconomic models 5.8.5. Game Theory (Descriptive)

tend to be calibrated rather than validated em-
pirically. While macroeconomic models are not Game theory attempts to model the behavior of
within the direct interest of PSE, they require the individuals when confronted with a choice that
same sort of solution methods as the traditional depends on the choices of others and as such, it
PSE applications. is helpful for the development of a strategy in
situations of competition and/or cooperation.
The recommended strategy is based on the
5.8.2. Optimization Models (Mainstream,
so-called Nash equilibrium, which assumes a
Normative, Aggregated, Mechanistic)
self-interest behavior and perfectly rational
agents (players) [221]. However, agents are not
Optimization models are used to help in the
always fully rational.
selection of alternatives by minimizing capital
and operating costs under constraints of avail-
ability, prices, etc. System optimality assumes 5.8.6. System Dynamics (Aggregated,
that decisions are taken centrally; however, Mechanistic)
society is composed of individuals and groups
making their own decisions. System dynamics is a methodology based on the
Optimization is one of the prime interest areas general systems theory.
in PSE, with several types of applications [218]. It uses the concepts of stocks (levels), flows
(rates), feedback relationships, and time delay
in order to model dynamic behavior.
5.8.3. Control Models (Mainstream, No- System dynamics models are appropriate for
rmative, Aggregated, Mechanistic) the (qualitative) identification of the important
variables and causal links in a system, and for
Process control is one of the mainstream PSE the representation of the nonlinearities arising
areas. A specialized branch of control is model from feedback loops and time delays. This
predictive control (MPC) (! Process Systems methodology is popular in the modeling of
Engineering, 5. Process Dynamics, Control, environmental systems, and to a lesser extent
Monitoring, and Identification, Chap. 5). for economic systems (! Ecology and Envi-
There is, however, considerable uncertainty ronmental Chemistry).
associated to the model parameters and possible
disturbances when the prediction of the effects
5.8.7. Network Theory (Descriptive)
is done over a horizon of many years. Exten-
sions to MPC are required to address this issue.
Network theory groups a family of methods that
For example, stochastic predictive control [219]
enable the representation of nodes and their
and a reformulation introducing multiple objec-
interdependencies, and its subsequent analysis.
tives and dynamics [220] have also been pro-
As a result these methods are suitable to model
posed to model the effects of policy over a
the behavior of networks, such as those resulting
horizon of many years.
from transportation systems or electricity gen-
eration and distribution. Networks have nonlin-
5.8.4. Data-Based Models ear feedback loops, exhibit dynamic behavior,
and their local effects cannot be averaged.
These models fit equations to data without Several specialized types of networks can be
trying to simulate the system in a mechanistic defined according to the focus that the modeler
fashion, e.g., in the discovery of statistical pat- wants the model to have. For example, if the
terns in historical data. Their predictions assume focus is on the management of uncertainty,
that the future will resemble the past, but one Bayesian Networks are risk assessment models
property of sociotechnical systems is that they expressed in terms of influence diagrams that
evolve and change. As a result, this type of are used to infer through causal links the prob-
models by itself is not adequate for policy ability of an effect from the probabilities of its
support [212]. associated causes.
 Process Systems Engineering, 9. Domain Engineering 43

5.8.8. Agent-Based Approaches 4. Models may also have to be multilevel, i.e.,

using various time spans (just ensuring that
An agent is an autonomous entity with an the time spans are long enough to allow the
internal state. In turn, a multiagent system is complete unfolding of the dynamics of the
a collection of agents that interact with other system).
agents and the environment within which they 5. Model development should ideally allow the
operate. The interactions between agents, which inclusion of values and objectives of multi-
are governed by rules, modify the internal state ple stakeholders and facilitate the communi-
of the agents. This type of model has the ability cation of the results.
to capture individual heterogeneous actors as
autonomous decision-makers, with bounded ra- 5.9. Synthesis of Policies
tionality (i.e., limited information), attempting
to maximize their own utility function. The A large portion of the space of alternative poli-
approach is catalogued as bottom-up and has cies is currently left unexplored because the
the property of exhibiting emergent behavior. synthesis of policies is performed manually. A
As in the case of networks, it is possible to systematic approach that aims to automate the
extend the basic multiagent representation in a generation of alternative policies has the poten-
number of ways, for example: tial to accelerate the policy making process and
improve the effectiveness of the resulting poli-
. Incorporating a stochastic approach within an cies; the development of such an approach may
agent-based system to facilitate the manage- benefit from the experience gained in the area of
ment of uncertainties. process synthesis in the last three decades. How-
. Embedding the agents in a spatial setting and ever, the differences between process synthesis
thus creating artificial societies [222] or arti- and policy formulation, in particular the perva-
ficial ecosystems [223], for example, through siveness of qualitative factors in the latter, re-
the creation of synthetic micropopulations to quire a substantial adaptation of the methods
explore policy impacts []. used in process synthesis.
A good policy ‘‘packages’’ a collection of
5.8.9. Some Conclusions on Models for policy measures aiming to ensure the effective-
the Analysis of Policies ness and acceptability of the policy while main-
taining a reasonable cost and a tolerable risk.
There are a few general conclusions that can be The properties of a policy are determined by the
drawn from the breadth of methods used for properties of its constituent policy measures and
policy analysis: their interactions, thus policy makers try to take
advantage of the synergies between policy mea-
1. It is possible to develop partial models to sures and also try to avoid, or at least mitigate,
capture specific aspects of the sociotechnical the negative interactions between them.
system when only one type of modeling A systematic approach for the synthesis of
methodology is used. policies is being developed in the area of trans-
2. To develop more inclusive models it is nec- port policy to reduce CO2 emissions [225]:
essary to combine information from diverse This systematic approach may constitute the
sources and integrate models. Hybrid models first step towards the development of a family of
integrate different methods, e.g., mainstream computer-based systems that support the design
mathematical simulation models with differ- of policies for different sectors, such as trans-
ent tools to account for location (geographi- port, energy, food, and water aiming to achieve
cal information systems), economics (mac- environmental, security, health, and safety tar-
roeconomic models), etc. gets. The output from such systems is a set of
3. In either case, partial or hybrid, models may promising alternative policies, each annotated
have to be multiscale, i.e., spanning indivi- with its associated advantages and disadvan-
duals and organizations. This is the nature of tages. The final decision on which policy
some of the models being developed for to implement rests with the decision makers
PSE [224]. who may decide to include additional policy
44 Process Systems Engineering, 9. Domain Engineering

measures or remove some of the recommended 11 N.M. Garrido, M. Jorge, A.J. Queimada, I.G. Economou, E.A.
ones. Macedo, Fluid Phase Equilibr. 289 (2010) 148–155.
12 N.M. Garrido, A.J. Queimada, M. Jorge, E.A. Macedo, I.G.
Economou, J. Chem. Theory Comput. 5 (2009) 2436–2446.
5.10. Future Directions 13 N.M. Garrido, A.J. Queimada, M. Jorge, I.G. Economou, E.A.
Macedo, Fluid Phase Equilibr. 296 (2010) no. 2, 110–115.
14 Z.A. Makrodimitri, R. Dohrn, I.G. Economou, Macromolecules
The integrated design and analysis of socio- 40 (2007) 1720–1729.
technical systems remains a challenge for both, 15 R.D. Raharjo, B.D. Freeman, E.S. Sanders, J. Membr. Sci. 292
policy developers and engineers; there is still a (2007) 45–61.
16 Z.A. Makrodimitri, I.G. Economou, Macromolecules 41 (2008)
large gap between the objectives of both com- 5899–5907.
munities and reality. 17 J. Pinnau, Z.J. He, Membr. Sci. 244 (2004) 227–233.
Engineers cannot expect to be able to formu- 18 P. Wasserscheid, T. Welton: Ionic Liquids in Synthesis, 2nd ed.,
late, analyze, and predict the behavior of poli- Wiley-VCH Verlag, Weinheim 2007.
19 G.-E. Logotheti, J. Ramos, I.G. Economou, J. Phys. Chem. B
cies in the same manner as they have done it for 113 (2009) 7211–7224.
industrial processes because sociotechnical sys- 20 Sandia National Laboratories, www.lammps.sandia.gov
tems are different and more complex than tech- (accessed 26 Jan. 2010).
nical systems. There is, however, a large scope 21 Groningen Machine for Chemical Simulations, www.gromacs.
org (accessed 26 Jan. 2010).
for the application and/or adaptation of main- 22 Collaborative Computational Projects, www.ccp5.ac.uk/librar.
stream PSE methods and tools for the support of shtml (accessed 26 Jan. 2010).
policy making, particularly in the analysis and 23 Daresbury Laboratory, www.cse.scitech.ac.uk/ccg/software/
synthesis of policies. DL_POLY/index.shtml (accessed 26 Jan. 2010).
24 Accelrys, www.accelrys.com (accessed 26 Jan. 2010).
A word of warning: there is a risk in having 25 Scienomics, www.scienomics.com (accessed 26 Jan. 2010).
excessive confidence on models that may be 26 E.N. Pistikopoulos: ‘‘Energy systems engineering - an integrat-
inaccurate or incomplete because they may give ed approach for the energy systems of the future’’, European
a delusion of control on systems that may be Congress of Chemical Engineering-6 Copenhagen, Denmark
2007.
intrinsically unpredictable. Yet even in the case 27 M.C. Georgiadis, E.S. Kikkinides, E.N. Pistikopoulos: ‘‘Pro-
of systems that cannot be properly characterized cess systems engineering’’, Energy systems engineering, vol.
and controlled, models can be useful if we ask 5, Wiley-VCH Verlag, Weinheim 2008.
the right question, i.e., to provide insights and 28 P. Liu, E.N. Pistikopoulo, Z. Li: ‘‘An energy systems engineer-
ing approach to polygeneration and hdrogen infrastructure
further understanding of the workings of a systems analysis & design’’, 12th International Conference on
sociotechnical system rather than to predict its Process Integration, Modeling and Optimisation for Energy
future behavior. Saving and Pollution Reduction, Rome 2009.
29 H. Yeomans, I.E. Grossmann: ‘‘A systematic modeling frame-
work of superstructure optimization in process synthesis’’,
Comput. Chem. Eng. 23 (1999) no. 6, 709–731.
References 30 T.F. Yee, I.E. Grossmann, Z. Kravanja: ‘‘Simultaneous optimi-
zation models for heat integration. 3. Process and heat-exchang-
1 N. Metropolis, A.W. Rosenbluth, M.N. Rosenbluth, A.N. Tell- er network optimization’’, Comput. Chem. Eng. 14 (1990) no.
er, E. Teller, J. Chem. Phys. 21 (1953) 1087–1092. 11, 1185–1200.
2 B.J. Alder, T.E. Wainwright, J. Chem. Phys. 27 (1957) 1208– 31 I.E. Grossmann, A.W. Westerberg: ‘‘Research challenges in
1209. Process Systems Engineering’’ AIChE J. 46 (2000) no. 9, 1700–
3 D.A. McQuarrie: Statistical Mechanics, Harper and Row, New 1703.
York 1976. 32 A.C. Kokossis, C.A. Floudas: ‘‘Synthesis of isothermal reactor
4 T.M. Reed, K.E. Gubbins: Applied Statistical Mechanics, separator recycle systems’’, Chem. Eng. Sci. 46 (1991) no. 5–6,
Butterworth, London 1973. 1361–1683.
5 D. Frenkel, B. Smit: Understanding Molecular Simulation, 33 R. Agrawal: ‘‘Synthesis of distillation column configurations
Academic Press, New York 1996. for a multicomponent separation’’, Ind. Eng. Chem. Res. 35
6 M.P. Allen, D.J. Tildesley: ‘‘Computer Simulation of Liquids, (1996) no. 4, 1059–1071.
Oxford University Press, Oxford 1987. 34 S.R. Ismail, P. Proios, E.N. Pistikopoulos: ‘‘Modular synthesis
7 M. Kotelyanskii, D.N. Theodorou: Simulation Methods for framework for combined separation/reaction systems’’, AIChE
Polymers, Marcel Dekker, New York 2004. J. 47 (2001) no. 3, 629–649.
8 A.Z. Panagiotopoulos, Molec. Phys. 61 (1987) 813–826. 35 V. Hatzimanikatis, C.A. Floudas, J.E. Bailey: ‘‘Analysis and
9 A.Z. Panagiotopoulos, J. Phys.: Condens. Matter 12 (2000) design of metabolic reaction networks via mixed-integer linear
R25–R52. optimization’’, AIChE J. 42 (1996) no. 5, 1277–1292.
10 I.G. Economou in G.M. Kontogeorgis, R. Gani (eds.): ‘‘Com- 36 C.A. Schweiger, C.A. Floudas: ‘‘Optimization framework for
puter Aided Property Estimation for Process and Product the synthesis of chemical reactor networks’’, Ind. Eng. Chem.
Design’’, Elsevier, Amsterdam 2004, pp. 279–307. Res. 38 (1999) no. 3, 744–766.
 Process Systems Engineering, 9. Domain Engineering 45

37 M.J. Tsai, C.T. Chang: ‘‘Water usage and treatment network 57 P. Liu, E.N. Pistikopoulos, Z. Li: ‘‘An energy systems engi-
design using genetic algorithms’’, Ind. Eng. Chem. Res. 40 neering approach to the optimal design of energy systems in
(2001) no. 22, 4874–4888. commercial buildings’’, Energy Policy 38 (2010) no. 8, 4224–
38 P. Liu: ‘‘Modeling and optimization of polygeneration energy 4231
systems’’ Ph.D. Thesis, Imperial College London, London 58 D. Kuhn, P. Parpas, B. Rustem: Stochastic optimization of
2010. investment planning problems in the electric power industry,
39 A. Hugo: Environmentally conscious process selection, design in Process Systems Engineering, vol. 5, Energy Systems
and optimization. Ph.D. Thesis, Imperial College London, Engineering, Wiley-VCH Verlag, Weinheim 2008.
London 2005. 59 M. Dabros, M. Perrier, F. Forbes, M. Fairbank, P.R. Stuart:
40 C.A. Floudas: Nonlinear and mixed-integer optimization: ‘‘Model-based direct search optimization of the broke recircu-
Fundamentals and applications, Oxford University Press, lation system in a newsprint mill’’, J. Cleaner Production 13
Oxford 2005. (2005) 1416–1423.
41 A.H. Land, A.G. Doig: ‘‘An automatic method of solving 60 E. Salazar, R. Ramson, K. Munnoch, P. Stuart: ‘‘Identifying
discrete programming problems’’, Econometrica 28 (1960) environmental improvement opportunities for newsprint pro-
no. 3, 497–520. duction using life cycle assessment (LCA)’’, Paper Pulp
42 A.M. Geoffrion: ‘‘Generalized benders decomposition’’, Canada 107 (2006) 32–38.
Journal of Optimization Theory and Applications 10 (1972) 61 C. Gaudreault, R. Samson, P.R. Stuart: ‘‘Life-cycle thinking in
no. 4, 237. the pulp and paper industry, part 2: LCA studies and opportu-
43 M.A. Duran, I.E. Grossmann: ‘‘An outer-approximation algo- nities for development’’, Tappi J. 6 (2007) no. 8, 3–10.
rithm for a class of mixed-integer nonlinear programs’’, Math- 62 C. Gaudreault, R. Samson, P.R. Stuart: ‘‘Life-cycle thinking in
ematical Programming 36 (1986) no. 3, 307–339. the pulp and paper industry, Part I: Current practices and most
44 R.E. Steuer: Multiple criteria optimization: Theory, computa- promising avenues’’, Tappi J. 6 (2007) no. 7, 25–31.
tion, and application, J. Wiley & Sons, New York 1986. 63 M. Laflamme-Mayer, J. Linkewich, P.R. Stuart: ‘‘Methodology
45 K.P. Papalexandri, T.I. Dimkou: ‘‘A parametric mixed-integer for using real-time process data for cost modeling and supply
optimization algorithm for multiobjective engineering chain decision support in the pulp and paper industry’’, Pulp
problems involving discrete decisions’’, Ind. Eng. Chem. Res. Paper Canada 109 (2008) no. 3, 44–51.
37 (1998) no. 5, 1866–1882. 64 M.C. Georgiadis, E.N. Pistikopoulos: ‘‘Integrated optimization
46 V. Dua, E.N. Pistikopoulos: ‘‘Algorithms for the solution of of oil and gas production’’, in Process Systems Engineering,
multiparametric mixed-integer nonlinear optimization vol. 5, Energy Systems Engineering, Wiley-VCH Verlag, Wein-
problems’’, Ind. Eng. Chem. Res. 38 (1999) no. 10, 3976– heim 2008.
3987. 65 K. Kouramas, E.N. Pistikopoulos: ‘‘Wind turbines modeling
47 V. Dua, E.N. Pistikopoulos: ‘‘An algorithm for the solution of and control’’, in Process Systems Engineering, vol. 5, Energy
multiparametric mixed integer linear programming problems’’, Systems Engineering, Wiley-VCH Verlag, Weinheim 2008.
Annals of Operations Research 99 (2000) 123–139. 66 J.J. Klemes, I. Bulatov, S.J. Perry: ‘‘Energy efficiency and
48 P. Liu, D.I. Gerogiorgis, E.N. Pistikopoulos: ‘‘Modeling and carbon footprint reduction’’, in Process Systems Engineering,
optimization of polygeneration energy systems’’, Catal. Today vol. 5, Energy Systems Engineering, Wiley-VCH Verlag, Wein-
127 (2007) no. 1–4, 347–359. heim 2008.
49 P. Liu, E.N. Pistikopoulos, Z. Li: ‘‘A mixed-integer optimiza- 67 F.E. Pereira et al.: Integrated design of CO2 capture processes
tion approach for polygeneration energy systems design’’, from natural gas Energy Systems Engineering, Wiley-VCH
Comput. Chem. Eng. 33 (2009) no. 3, 759–768. Verlag, Weinheim 2008.
50 P. Liu, E.N. Pistikopoulos, Z. Li: ‘‘A multi-objective optimiza- 68 M. Markowski, K. Urbaniec: ‘‘Optimization of structure and
tion approach to polygeneration energy systems design’’, operating parameters of a sequence of distillation columns for
AIChE J. 56 (2009) no. 6. thermal separation of hydrogen’’, in Process Systems Engineer-
51 P. Liu, E.N. Pistikopoulos, Z. Li: ‘‘Decomposition based sto- ing, vol. 5, Energy Systems Engineering, Wiley-VCH Verlag,
chastic programming approach for polygeneration energy sys- Weinheim 2008.
tems design under uncertainty’’, Ind. Eng. Chem. Res. 49 (2010) 69 A. Hugo, E.N. Pistikopoulos: ‘‘Environmentally conscious
no. 7, 3295–3305. long-range planning and design of supply chain networks’’,
52 F. Marechal, C. Weber, D. Favrat: ‘‘Multiobjective design and J. Cleaner Prod. 13 (2005) no. 15, 1471–1491.
optimization of urban energy systems’’, in Process Systems 70 J. DiMasi, R. Hansen, H. Grabowski: ‘‘The price of innovation:
Engineering, vol. 5, Energy Systems Engineering, Wiley-VCH new estimates of drug development costs’’, Journal of Health
Verlag, Weinheim 2008. Economics 22 (2003) no. 2, 151–185.
53 A. Hugo, P. Ruther, S. Pistikopoulos, A. Amorelli, G. Zoia: 71 B. Bennett, G. Cole: Pharmaceutical Production: An Engineer-
‘‘Hydrogen infrastructure strategic planning using multi-objec- ing Guide, Institution of Chemical Engineers (IChemE),
tive optimization’’, Int. J. Hydrogen Energy 30 (2005) no. 15, Rugby 2002.
1523–1534. 72 K.V. Gernaey, R. Gani: ‘‘A model-based systems approach to
54 E.S. Kikkinides: ‘‘Hydrogen-based energy systems: The stor- pharmaceutical product-process design and analysis’’, Chem.
age challenge’’, in Process Systems Engineering, vol. 5, Eng. Sci. 65 (2010) no. 21, 5757–5769.
Energy Systems Engineering, Wiley-VCH Verlag, Weinheim 73 I. Grossmann, A. Westerberg: ‘‘Research challenges in
2008. process systems engineering’’, AIChE J. 46 (2000) no. 9,
55 Z. Li et al.: ‘‘Hydrogen infrastructure design and optimization: 1700–1703.
A case study of China’’, International Journal of Hydrogen 74 G. Reklaitis: ‘‘Perspectives on Process Systems Engineering
Energy 33 (2008) no. 20, 5275–5286. R&D in Support of Pharmaceutical Product/Process Develop-
56 A. Giannopoulos: ‘‘Energy systems optimization of a shopping ment and Manufacturing’’, 17th European Symposium on
mall’’, MSc Thesis, Imperial College London, London 2008. Computer Aided Process Engineering 24 (2007) 35.
46 Process Systems Engineering, 9. Domain Engineering

75 M. Rogers, A. Gupta, C. Maranas: ‘‘Real options based 91 T. Kourti: ‘‘The Process Analytical Technology initiative and
analysis of optimal pharmaceutical research and develop- multivariate process analysis, monitoring and control’’, Anal.
ment portfolios’’, Ind. Eng. Chem. Res 41 (2002) no. 25, Bioanal. Chem. 384 (2006) no. 5, 1043–1048.
6607–6620. 92 D. Petrides, A. Koulouris, P. Lagonikos: ‘‘The role of process
76 V. Venkatasubramanian, C. Zhao, G. Joglekar, A. Jain, L. simulation in pharmaceutical process development and product
Hailemariam, P. Suresh, P. Akkisetty, K. Morris, G.V. Reklai- commercialization’’, Pharm. Eng. 22 (2002) no. 1, 56–65.
tis: ‘‘Ontological informatics infrastructure for pharmaceutical 93 J. Worlitschek, T. Hocker, M. Mazzotti: ‘‘Restoration of PSD
product development and manufacturing’’, Comput. Chem. from chord length distribution data using the method of projec-
Eng. 30 (2006) no. 10–12, 1482–1496. tions onto convex sets’’, Part. Part. Syst. Charact. 22 (2005) no.
77 N. Shah: ‘‘Pharmaceutical supply chains: key issues and strat- 2, 81–98.
egies for optimization’’, Comput. Chem. Eng. 28 (2004) no. 94 M.A. Grover, S.C. Barthe, R.W. Rousseau: ‘‘Principal compo-
6–7, 929–941. nent analysis for estimating population density from chord-
78 R. Singh, K.V. Gernaey, R. Gani: ‘‘Model-based computer- length density’’, AIChE J. 55 (2009) no. 9, 2260–2270.
aided framework for design of process monitoring and analysis 95 N. Kail, H. Briesen, W. Marquardt: ‘‘Advanced Geometrical
systems’’, Comput. Chem. Eng. 33 (2009) no. 1, 22–42. Modeling of Focused Beam Reflectance Measurements
79 E. Kloppenburg, E.D. Gilles: ‘‘A new concept for operating (FBRM)’’, Part. Part. Syst. Charact. 4 (2007) no. 3, 184–192.
simulated moving-bed processes’’, Chem. Eng. Technol. 22 96 S.C. Barthe: Investigation and modeling of the mechanisms
(1999) no. 10, 813–817. involved in batch cooling cyrstallization and polymorphism
80 L. Yu: ‘‘Pharmaceutical quality by design: product and process through efficient use of the FBRM, Ph.D. Thesis, Georgia
development, understanding, and control’’, Pharm. Res. 25 Institute of Technology, Georgia 2008.
(2008) no. 4, 781–791. 97 M.P. Elsner et al.: ‘‘Experimental study and simplified mathe-
81 A. Rathore, H. Winkle: ‘‘Quality by design for biopharmaceu- matical description of preferential crystallization’’, Chirality 17
ticals’’, Nature Biotechnology 27 (2009) no. 1, 26–34. (2005) S183–S195.
82 H. Wu, M. Khan, A. Hussain: ‘‘Process control perspective for 98 S. Qamar, M.P. Elsner, I.A. Angelov, G. Warnecke, A. Seidel-
process analytical technology: Integration of chemical engi- Morgenstein: ‘‘A comparative study of high resolution schemes
neering practice into semiconductor and pharmaceutical indus- for solving population balances in crystallization’’, Comput.
tries’’, Chem. Eng. Commun. 194 (2007) no. 6, 760–779. Chem. Eng. 30 (2006) 1119–1131.
83 US Food and Drug Administration: ‘‘Guidance for industry: Q8 99 S. Motz, A. Mitrovi, E. Gilles: ‘‘Comparison of numerical
phamaceutical development’’, May 2006, http://www.fda.gov/ methods for the simulation of dispersed phase systems’’, Chem.
downloads/RegulatoryInformation/Guidances/ucm128029.pdf Eng. Sci. 57 (2002) no. 20, 4329–4344.
(accessed 6 Feb. 2012). 100 S.H. Chung, D.L. Ma, R.D. Braatz: ‘‘Optimal model-based
84 P. Barrett et al.: ‘‘A review of the use of process analytical experimental design in batch crystallization’’, Chemom. Intell.
technology for the understanding and optimization of produc- Lab. Syst. 50 (2000) 83–90.
tion batch crystallization processes’’, Org. Process Res. Dev. 9 101 F. Czapla, H. Haida, M.P. Elsner, H. Lorenz, A. Seidel-Mor-
(2005) no. 3, 348–355. genstein: ‘‘Parameterization of population balance models for
85 T. De Beer, M. Allesø, F. Goethals, A. Coppens et al.: ‘‘Imple- polythermal auto seeded preferential crystallization of enantio-
mentation of a Process Analytical Technology system in a mers’’, Chemical Engineering Science 64 (2009) no. 4, 753–
freeze-drying process using Raman spectroscopy for in-line 763.
process monitoring’’, Anal. Chem. 79 (2007) no. 21, 7992– 102 T. Togkalidou, H.-H. Tung, Y. Sun, A.T. Andrews, R.D. Braatz:
8003. ‘‘Parameter estimation and optimization of a loosely bound
86 S. Gnoth, M. Jenzsch, R. Simutis, A. Lübber: ‘‘Process aggregating pharmaceutical crystallization using in situ infra-
Analytical Technology (PAT): Batch-to-batch reproducibil- red and laser backscattering measurements’’, Ind. Eng. Chem.
ity of fermentation processes by robust process operational Res. 43 (2004) , 6168–6181.
design and control’’, J. Biotechnol. 132 (2007) no. 2, 180– 103 S.M. Mille, J.B. Rawlings: ‘‘Model identification and control
186. strategies for batch cooling crystallizers’’, AIChE J. 40 (1994)
87 J. Lopes, P. Costa, T. Alves, J. Menezes: ‘‘Chemometrics in no. 8, 1312–1327.
bioprocess engineering: process analytical technology (PAT) 104 I. Angelov et al.: ‘‘Optimization of initial conditions for pref-
applications’’, Chemom. Intell. Lab. Syst. 74 (2004) no. 2, erential crystallization’’, Ind. Eng. Chem. Res. (2006) 759–766.
269–275. 105 M. Fujiwara, P.S. Chow, D.L. Ma, R.D. Braatz: ‘‘Paracetamol
88 R. Singh, K.V. Gernaey, R. Gani: ‘‘ICAS-PAT: A software for crystallization using laser backscatternig and ATR-FTIR spec-
design, analysis and validation of PAT systems’’, Comput. troscopy: meta stability, agglomeration, and control’’, Cryst.
Chem. Eng. 34 (2010) no. 7, 1108–1136. Growth Des. 2 (2002) no. 5, 363–370.
89 US Food and Drug Administration: ‘‘Guidance for industry: 106 N.A.F. Abdul Samad, R. Sing, G. Sin, K.V. Gernavy, R. Gani:
PAT—a framework for innovative pharmaceutical develop- ‘‘A generic multi-dimensional model-based system for batch
ment, manufacturing, and quality assurance’’ September 2004, cooling crystallization processes’’, Comput. Chem. Eng. 35
http://www.fda.gov/downloads/Drugs/GuidanceCompliance (2011) no. 5, 828–843.
RegulatoryInformation/Guidances/UCM070305 (accessed 6 107 G. Guiochon, A. Felinger, D.G.G. Shirazi: Fundamentals of
February 2012). Preparative and Nonlinear Chromatography, Academic Press,
90 A. Rathore, M. Yu, S. Yeboah, A. Sharma: ‘‘Case study and New York 2006.
application of process analytical technology (PAT) towards 108 H. Schmidt-Traub: Preparative chromatography of fine che-
bioprocessing: Use of on-line high-performance liquid chro- micals and pharmaceutical agents, Wiley-VCH Verlag, Wein-
matography (HPLC) for making real-time pooling decisions for heim 2005.
process chromatography’’, Biotechnol. Bioeng. 100 (2007) no. 109 G. Dünnebier, K.-U. Klatt: ‘‘Modeling and simulation of non-
2, 306–316. linear chromatographic separation processes: a comparison of
 Process Systems Engineering, 9. Domain Engineering 47

different modeling approaches’’, Chem. Eng. Sci. 55 (2000), 130 H.E.M. Law, C.V.F. Baldwin, B.H. Chen, M. Woodley: ‘‘Pro-
373–380. cess limitations in a whole-cell catalysed oxidation: sensitivity
110 V. Natarajan, B.W. Bequette, S.M. Cramer: ‘‘Optimization of analysis’’, Chem. Eng. Sci. 61 (2006) 6646–6652.
ion-exchange displacement separations i. validation of an 131 B.H. Chen, F. Baganz, J.M. Woodley: ‘‘Modelling and optimi-
iterative scheme and its use as a methods development tool’’, sation of a transketolase mediated carbon-carbon bond forma-
J. Chromatogr. A 876 (2000) 51–62. tion reaction’’, Chem. Eng. Sci. 62 (2007) 3178–3184.
111 A. Rajendran, G. Paredes, M. Mazzotti: ‘‘Simulated moving 132 P.A. Santacoloma, G. Sin, K.V. Gernaey, J.M. Woodley:
bed chromatography for the separation of enantiomers’’, ‘‘Multi-enzyme Catalyzed Processes: Next-generation Bioca-
J. Chromatogr. A 1216 (2009) no. 4, 709–738. talysis’’, Org. Process Res. Dev. 15 (2011) 203–212.
112 J. Hakanen et al.: ‘‘Interactive multi-objective optimization for 133 D. Petrides, C.L. Cooney, L.B. Evans, R.P. Field, M. Snoswell:
simulated moving bed processes’’, Control and Cybernetics 36 ‘‘Bioprocess simulation: an integrated approach to process
(2007) no. 2, 283. development’’, Comput. Chem. Eng. 13 (1989) 553–561.
113 Z. Zhang, K. Hidajat, A.K. Ray, M. Morbidelli: ‘‘Multiobjec- 134 G. Sin, P. Ödman, N. Petersen, A. Eliasson Lantz, K.V.
tive optimization of SMB and Varicol processes for chiral Gernaey: ‘‘Matrix notation for efficient development of first-
separation’’, AIChE J. 48 (2002) no. 12, 2800–2816. principles models within PAT applications: Integrated model-
114 Y. Kawajiri, L.T. Biegler: ‘‘Optimization strategies for simu- ing of antibiotic production with Streptomyces coelicolor’’,
lated moving bed and PowerFeed processes’’, AIChE J. 52 Biotechnol. Bioeng. 101 (2008) 153–171.
(2006) no. 4, 1343–1350. 135 M. Degerman, K. Westerberg, B. Nilsson: ‘‘Determining criti-
115 A. Toumi, F. Hanisch, S. Engell: ‘‘Optimal operation of con- cal process parameters and process robustness in preparative
tinuous chromatographic processes: Mathematical optimiza- chromatography – A model-based approach’’, Chem. Eng.
tion of the VARICOL process’’, Ind. Eng. Chem. Res. 41 (2002) Technol. 32 (2009) 903–911.
4328–4337. 136 G. Sin, A. Eliasson Lantz, K.V. Gernaey: ‘‘Sensitivity analysis
116 A. Seidel-Morgenstern, L. Keßler, M. Kaspereit: ‘‘New devel- of non-linear dynamic models: prioritizing experimental re-
opments in simulated moving bed chromatography’’, Chem. search’’, Comput.-Aided Chem. Eng. 27A (2009) 267–272.
Eng. Technol. 31 (2008) no. 6, 826–837. 137 R. Singh, K.V. Gernaey, R. Gani: ‘‘Model-based computer
117 Y. Kawajiri, L.T. Biegler: ‘‘A nonlinear programming super- aided framework for design of process monitoring and analysis
structure for optimal dynamic operations of simulated moving systems’’, Comput. Chem. Eng. 33 (2009) 22–42.
bed processes’’, Ind. Eng. Chem. Res. 45 (2006) no. 25, 8503– 138 K.V. Gernaey, A. Eliasson Lantz, P. Tufvesson, J.M. Woodley,
8513. G. Sing: ‘‘Application of mechanistic models to fermentation
118 G. Dünnebier, K.U. Klatt: ‘‘Model-based control of batch and biocatalysis for next generation processes’’, Trends
chromatography’’, AIChE J. 47 (2004) no. 11, 2493–2502. Biotechnol. 28 (2010) 346–354.
119 W. Gao, S. Engell: ‘‘Iterative set-point optimization of 139 A.G. Fredrickson, R.D. Megee, H.M. Tsuchiya: ‘‘Mathematical
batch chromatography’’, Comput. Chem. Eng. 29 (2005) models in fermentation processes’’, Adv. Appl. Microbiol. 13
1401–1409. (1970) 419–465.
120 D. Nagrath, B.W. Bequette, S. Cramer: ‘‘Evolutionary opera- 140 M.A. Henson: ‘‘Dynamic modeling of microbial cell popula-
tion and control of chromatographic processes’’, AIChE J. 49 tions’’, Curr. Opin. Biotechnol. 14 (2003) 460–467.
(2003) 82–95. 141 J.J. Nielsen, J. Villadsen: ‘‘Modeling of microbial kinetics’’.
121 S. Abel, G. Erdem, M. Mazzotti, M. Morari, M. Morbidelli. Chem. Eng. Sci. 47 (1992) 4225–4270.
Optimizing control of simulated moving beds—linear iso- 142 J.A. Roels: ‘‘Application of macroscopic principles to micro-
therm. J. Chromatogr. A 1033 (2004) 229–239. bial metabolism’’, Biotechnol. Bioeng. 22 (1980) 2457–2514.
122 G. Erdem, S. Abel, M. Morari, M. Mazzotti, M. Morbidelli, J.H. 143 M. Thilakavathi, T. Basak, T. Panda: ‘‘Modeling of enzyme
Lee: ‘‘Automatic control of simulated moving beds’’, Ind. Eng. production kinetics’’, Appl. Microbiol. Biotechnol. 73 (2007)
Chem. Res. 43 (2004) 405–421. 991–1007.
123 K. Klatt, F. Hanisch, G. Dünnebier, S. Engell: ‘‘Model-based 144 G. Sin, J.M. Woodley, K.V. Gernaey: ‘‘Application of model-
optimization and control of chromatographic processes’’, Com- ing and simulation tools for the evaluation of biocatalytic
put. Chem. Eng. 24 (2000) no. 2–7, 1119–1126. processes: a future perspective’’, Biotechnol. Prog. 25 (2009)
124 C. Grossmann, M. Amanullah, G. Erdem, M. Mazzotti, M. 1529–1538.
Morbidelli, M. Morari: ‘‘Cycle to cycleoptimizing control 145 P. Bansal, M. Hall, M.J. Realff, J.H. Lee, A.S. Bommarius:
of simulated moving beds’’, AIChE J. 54 (2008) no. 1, 194–208. ‘‘Modeling cellulase kinetics on lignocellulosic substrates’’,
125 D.J. Pollard, J.M. Woodley: ‘‘Biocatalysis for pharmaceutical Biotechnol. Adv. 27 (2009) 833–848.
intermediates: the future is now’’, Trends Biotechnol. 25 (2007) 146 J.E. Bailey: ‘‘Mathematical modeling and analysis in biochem-
66–73. ical engineering: past accomplishments and future opportu-
126 J.M. Woodley: ‘‘New opportunities for biocatalysis: making nities’’, Biotechnol. Prog. 14 (1998) 8–20.
pharmaceutical processes greener’’, Trends Biotechnol. 26 147 A.A. Esener, J.A. Roels, N.W.F. Kossen: ‘‘Theory and applica-
(2008) 321–327. tions of unstructured growth models: Kinetic and energetic
127 B.E. Dale: ‘‘Greening’ the chemical industry: research and aspects’’, Biotechnol. Bioeng. 25 (1983) 2803–2841.
development priorities for bio-based industrial products’’, J. 148 T. Agger, A.B. Spohr, M. Carlsen, J. Nielsen: ‘‘Growth and
Chem. Technol. Biotechnol. 78 (2003) 1093–1103. product formation of Aspergillus oryzae during submerged
128 J.M. Woodley: ‘‘Microbial biocatalytic processes and cultivations: verification of a morphologically structured model
their development’’, Adv. Appl. Microbiol. 60 (2006) using fluorescent probes’’, Biotechnol. Bioeng. 57 (1998)
1–15. 321–329.
129 C.V.F. Baldwin, J.M. Woodley: ‘‘On oxygen limitation in a 149 A.M. Zamamiri, Y. Zhang, M.A. Henson, M.A. Hjortsø:
whole-cell biocatalytic Baeyer-Villiger oxidation process’’, ‘‘Dynamics analysis of an age distribution model of oscillating
Biotechnol. Bioeng. 95 (2006) 362–369. yeast cultures’’, Chem. Eng. Sci. 57 (2002) , 2169–2181.
48 Process Systems Engineering, 9. Domain Engineering

150 M.M. Domach, M.L. Shuler: ‘‘A finite representation model for 168 R.A. Sheldon: ‘‘The E factor: fifteen years on’’, Green Chem. 9
an asynchronous culture of E. coli’’, Biotechnol. Bioeng. 26 (2007) 1273–1283.
(1984) 877–884. 169 D.J.C. Constable, A.D. Curzons, V.L. Cunningham: ‘‘Metrics
151 R. Lencastre Fernandes et al.: ‘‘Experimental methods and to ‘Green Chemistry’ – which are the best?’’, Green Chem. 4
modeling techniques for description of cell population hetero- (2002) 521–527.
geneity’’, Biotechnol. Adv. 29 (2011) no. 6, 575–599. 170 A.C.S./G.C.I.P.R.: American Chemical Society, Green Chem-
152 K. Kiviharju, K. Salonen, U. Moilanen, T. Erikäinen: ‘‘Biomass istry Institute, Pharmaceutical Roundtable website, www.acs.
measurement online: the performance of in situ measurements org/greenchemistry (accessed 10 Feb. 2012).
and software sensors’’, J. Ind. Microbiol. Biotechnol. 35 (2008) 171 D.J.C. Constable: ‘‘Green chemistry measures for process
657–665. research and development’’, Green Chem. 3 (2001) 7–9.
153 M. Vanek, P. Hrn
ci
rik, J. Vosik, J. Náhlik: ‘‘On-line estimation 172 G. Jödicke, O. Zenklusen, A. Weidenhaupt, K. Hungerbuhler:
of biomass concentration using a neural network and informa- ‘‘Developing environmentally-sound processes in the chemical
tion about metabolic state’’, Bioprocess Biosyst. Eng. 27 (2004) industry: a case study on pharmaceutical intermediates’’,
9–15. Journal of Cleaner Production 7 (1990) 159–166.
154 G. Acuña, E. Latrille, C. Béal, G. Corrieu: ‘‘Static and dynamic 173 S. Kim, B. Dale: ‘‘Life cycle assessment study of biopolymers
neural network models for estimating biomass concentration (polyhydroxyalkanoates) – Derived from non-tilled corn’’.
during thermophilic lactic acid bacteria batch cultures’’, J. International Journal of LCA 10 (2005) 200–210.
Ferment. Bioeng. 85 (1998) 615–622. 174 S. Kim, B. Dale: ‘‘Ethanol fuels: E10 or E85 – Life cycle
155 P. Ödman, C. Lindvaid Johansen, L. Olsson, K.V. Gernaey, A. perspectives’’, International Journal of LCA 11 (2006) 117–
Eliasson Lantz: ‘‘On-line estimation of biomass, glucose and 121.
ethanol in Saccharomyces cerevisiae cultivations using in-situ 175 S. Kim, C. Jimenez-Gonzalez, B. Dale: ‘‘Enzymes for pharma-
multi-wavelength fluorescence and software sensors’’, J. Bio- ceutical applications—a cradle-to-gate life cycle assessment’’,
technol. 144 (2009) 102–112. International Journal of LCA 14 (2009) no. 5, 392–400.
156 G.A. Montague, E.B. Martin, C.J. O’Malley: ‘‘Forecasting for 176 O.E.C.D.: The application of biotechnology to industrial sustain-
fermentation operational decision making’’, Biotechnol. Prog. ability – a primer, Organization for Economic Cooperation and
24 (2008) 1033–1041. Development, http://www.oecd.org/dataoecd/61/13/1947629pdf.
157 S.K. Maiti, R.K. Srivastava, M. Bhushan, P.P. Wangikar: ‘‘Real 2002
time phase detection based online monitoring of batch fermen- 177 M. Patel: ‘‘Medium and long term opportunities and risks of the
tation processes’’, Process Biochemistry 44 (2009) 799–811. biotechnological production of bulk chemicals from renewable
158 R.K. Henderson, C. Jimenez Gonzalez, C. Preston, D.J.C. resources’’, The BREW project – Final report – The Potential of
Constable, J.M. Woodley: ‘‘Comparison of biocatalytic and White Biotechnology, Europeon Commisson, Utrecht 2006.
chemical synthesis: EHS and LCA comparison for 7-ACA 178 D. Cowan, K.M. Oxenbøll, H.C. Holm, Oil Mill Gazeeter 113
synthesis’’, Ind. Biotechnol. 4 (2008) 180–192. (2008) 10–13.
159 A. Boisen et al.: ‘‘Process integration for the conversion of 179 O. Wolf: Techno-economic feasibility of large-scale produc-
glucose to 2,5-furandicarboxylic acid’’, Chem. Eng. Res. Des. tion of bio-based polymers in Europe, European Commission,
87 (2009) 1318–1327. Sevilla 2005.
160 J. Rass-Hansen, H. Falsig, B. Jørgensen, C.H. Christenson: 180 C. Jimenez-Gonzalez, D.J.C. Constable, A.D. Cuizons, V.L.
‘‘Bioethanol: fuel or feedstock’’, J. Chem. Technol. Biotechnol. Cunningham: ‘‘Developing GSK’s green technology guidance:
82 (2007) 329–333. methodology for case-scenario comparison of technologies’’,
161 E.G. Hibbert et al.: ‘‘Directed evolution of biocatalytic pro- Clean Technology and Environmental Pollution 4 (2001)
cesses’’, Biomol. Eng. 22 (2005) , 11–19. 44–53.
162 B.H. Chen, A. Sayar, U. Kaulmann, P.A. Dalby, J.M. Ward, J.H. 181 C. Jimenez-Gonzalez, D.J.C. Constable, A.D. Cuizons, V.L.
Woodley: ‘‘Reaction modelling and simulation to assess the Cunningham: ‘‘How do you select the ‘greenest’ technology?
integrated use of transketolase and v-transaminase for the Development of Guidance for the Pharmaceutical Industry’’,
synthesis of an aminotriol’’, Biocatal. Biotransform. 24 Clean Products and Processes 3 (2002) 35–41.
(2006) 449–457. 182 International Organization of Standardization: ISO 14040–
163 P.T. Anastas, J.C. Warner: Green chemistry: theory and prac- Environmental management–Life cycle assessment-Principles
tice, Oxford University Press, Oxford 2000. and framework, Geneva 1997.
164 A. Carvalho, R. Gani, H. Matos in W. Marquardt, C. Pantelides 183 International Organization of Standardization: ISO 14040–
(eds.): ‘‘Design of sustainable processes: Systematic generation Environmental management–Life cycle assessment-Require-
& evaluation of alternatives’’, 16th European Symposium on ments and Guidelines, Geneva 2006.
Computer-Aided Process Engineering and 9th International 184 Jeroen Guné (ed.): Life Cycle Assessment. An Operational
Symposium on Process Systems Engineering 21 (2006) 817– Guide to the ISO standards, Centrum voor Milieukunde
822. Leinde, Leiden University, 2002.
165 I.E. Grossmann, M.M. Daichendt: ‘‘New Trends in Optimiza- 185 Environmental Protection Agency: Tool for the Reduction and
tion-based Approaches for Process Synthesis’’, Comput. Chem. Assessment of Chemical and Other Environmental Impacts
Eng. 20 (1996) 665–683. (TRACI): User’s Guide and System Documentation. EPA
166 E. Uerdingen, R. Gani, U. F€ısher, K. Hungerbuhler: ‘‘A New 600/R-02/052. National Risk Management Research Laborato-
Screening Methodology for the Identification of Economically ry. Cincinnati, OH 2003.
Beneficial Retrofit Options for Chemical Processes’’, AIChE J. 186 PRé Consultants. Eco-Indicator 99. A Damage Oriented Meth-
49 (2003) 2400–2418. od for Life Cycle Impact Assessment, Amersfoort 2001.
167 R.A. Sheldon: ‘‘Organic synthesis–past, present and future’’, 187 C.T. Hendrickson, L. Lave, H.S. Matthews, Environmental Life
Chem. Ind. 23 (1992) 903–906. Cycle Assessment of Goods and Services: An Inpur-Output
 Process Systems Engineering, 9. Domain Engineering 49

Approach. Resources for the Fugure Press, Washington, DC, 207 A. Stirling: ‘‘Renewables, sustainability and precaution:
2006. beyond environmental cost-benefit and risk analysis’’, in Issues
188 R. Heijungs: Environmental life cycle assessment of products, in Environmental Science and Technology, no. 19: Sustain-
Guide and Backgrounds, NOH report 9266 and 9267, commis- ability and Environmental Impact of Renewable Sources, RSC
sioned by the National Reuse of Waste Research Programme Royal Society of Chemistry, London 2003, 113–134
(NOH), in collaboration with CML, TNO and B&G. Leiden; 208 W.E. Walker, S.A. Rahman, J. Cave: ‘‘Adaptive policies, policy
1992. analysis, and policy making’’, European Journal of Operation-
189 J. Johnson: ‘‘Science and policy in designing complex futures’’, al Research 128 (2001) 282–289.
Futures 40 (2008) 520–536. 209 P.M. Boulanger, T. Brechet: ‘‘Models for policy-making in
190 T.J. Wilbanks, R. Lee: ‘‘Policy analysis in theory and practice’’, sustainable development: The state of the art and perspectives
in T.R. Lakshmanan, B. Johansson (eds.): Large-scale energy for research’’, Ecological Economics 55 (2005) 337–350.
projects: assessment of regional consequences, Elsevier, 210 D.W. Pearce, R.K. Turner: Economics of natural resources and
Amsterdam 1985. the environment, Harvester Wheatsheaf, Brighton 1990.
191 J.M. Bauer, P.M. Herder: ‘‘Designing socio-technical sys- 211 J. O’Neill: ‘‘Markets and the environment: the solution is the
tems’’, in D.M. Gabbay, P. Thagard, J. Woods (eds.): Hand- problem’’, Economic and Political Weekly 36 (2001) no. 21,
book of the Philosophy of Science, vol. 9: Philosophy of 1865–1873.
Technology and Engineering Sciences, Elsevier, Amster- 212 D. Orrell, P. McSharry: ‘‘System economics: overcoming the
dam 2009, pp. 601–632. pitfalls of forecasting models via a multidisciplinary
192 P. Bridgman, G. Davis: Australian Policy Handbook, Allen approach’’, International Journal of Forecasting 25 (2009)
and Unwin Academic, Sydney 1998. 734–743.
193 A. Fenna: Introduction to Australian public policy, Addison 213 S. Wolfram: A new kind of science, Wolfram Media, Cham-
Wesley Longman, Melbourne 1998. paign 2002.
194 D. Stone: Policy paradox. The art of political decision making, 214 P.M. Gregg: Problems of theory in policy analysis, Heath-
W.W. Norton & Company, New York 2002. Lexington, London 1976.
195 D.A. Stone: Policy paradox and political reason, Harper- 215 T.R. Dye, V. Gray: The determinants of public policy, Heath-
Collins, New York 1988. Lexington, London 1960.
196 D. Easton: ‘‘A systems analysis of political life’’, J. Wiley & 216 S. Pathak, M. McDonald, S. Mahadevan: ‘‘A framework for
Sons, New York 1965. designing policies for networked systems with uncertainty’’,
197 B. Palmer: Beyond program performance indicators: perfor- Decision Support Systems 49 (2010) 121–131.
mance information in a national system of health and family 217 J.L. McCauley: Dynamics of markets: econophysics and
services, Department of Health and Family Services, Canberra finance, Cambridge University Press, Cambridge 2004.
1997. 218 L.T. Biegler, I.E. Grossmann: ‘‘Retrospective on optimiza-
198 M. Howlett, M. Ramesh: Studying public policy: policy cycles tion’’, Comput. Chem. Eng. 28 (2004) no. 8, 1169–1192.
and policy subsystems, Oxford University Press, Oxford 1995. 219 B. Kouvaritakis, M. Cannon, V. Tsachouridis: ‘‘Recent devel-
199 A.W. Westerberg: ‘‘A retrospective on design process synthe- opments in stochastic MPC and sustainable development’’,
sis’’, Comput. Chem. Eng. 28 (2004) 447–458. Annual Reviews in Control 28 (2004) 23–35.
200 A.W. Westerberg: Course Notes for Process Synthesis, Car- 220 B. Kouvaritakis, M. Cannon, P. Couchman: ‘‘MPC as a tool for
negie Mellon University, Pittsburgh 1981. sustainable development integrated policy assessment’’, IEEE
201 R. Clift: ‘‘Sustainable development and its implications for Transactions on Automatic Control 51 (2006) no. 1, 145–149.
chemical engineering’’, Chem. Eng. Sci. 61 (2006) 4179–4187. 221 K. Binmore: Game theory: a very short introduction, Oxford
202 T. Smithers, W. Troxell: ‘‘Design is intelligent behaviour but University Press, Oxford 2007.
what’s the formalism?’’, AI EDAM 4 (1990) no. 2, 89–98. 222 J.M. Epstein, R. Axtell: Growing artificial societies: social
203 H.A. Simon: The sciences of the artificial, 3rd ed., MIT Press, science from the bottom up, Brookings Institution Press and the
Cambridge 1996. MIT Press, Washington D.C. 1996.
204 M.G. Badami: ‘‘Environmental policy-making in a difficult 223 C.G. Langton: Artificial life: an overview (complex adaptive
context: motorized twowheeled vehicle emissions in India’’, systems), MIT Press, Cambridge 1997.
Energy Policy 32 (2004) 1861–1877. 224 R. Morales-Rodriguez, R. Gani: ‘‘Computer-aided multiscale
205 R. Bañares-Alcántara: ‘‘Perspectives on the potential roles of modelling for chemical process engineering’’, Comput.-Aided
engineers in the formulation, implementation and enforcement Chem. Eng. 24 (2007) 207–212.
of policies’’, Comput. Chem. Eng. 34 (2010) no. 3, 267–276. 225 A. Taeihagh, R. Bañares-Alcántara, C. Millican: ‘‘Develop-
206 J. Robinson, M. Bradley, P. Busby, D. Connor, A. Murray, B. ment of a novel framework for the design of transport policies to
Sampson, W. Soper: ‘‘Climate change and sustainable develop- achieve environmental targets’’, Comput. Chem. Eng. 33 (2009)
ment: realizing the opportunity’’, AMBIO 35 (2006) no. 1, 2–8. no. 10, 1531–1545.