Extended paper for the 13th International Symposium on Process Systems Engineering.

Process Systems Engineering:

Academic and Industrial Perspectives
Ignacio E. Grossmanna and Iiro Harjunkoskib
a)
Center for Advanced Process Decision-making (CAPD), Department of Chemical
Engineering, Carnegie Mellon University, Pittsburgh, PA 15213, U.S.A.
grossmann@cmu.edu
b)
ABB Corporate Research Center Germany, 68526 Ladenburg, Germany
iiro.harjunkoski@de.abb.com

Abstract
In this paper, we present both academic and industrial perspectives on the research and
applications of Process Systems Engineering (PSE). After a brief introduction on the
history of PSE, we describe the major research accomplishments in the areas of process
simulation, conceptual design and synthesis, process control, process operations and
optimization. This is followed by a discussion on the industrial impact and benefits of this
work, which have made it to be industrially relevant. Next, we address the issue of the
current standing of PSE both in academia and in industry, and for which we present results
of a survey conducted by the authors. Finally, we close with a discussion on future
challenges in PSE from both the industrial and academic perspectives.

Keywords: Process systems engineering, process simulation, process synthesis, process

control, scheduling and planning, mathematical programming.

1. Introduction
The field of process systems engineering (PSE) has been around in various forms for over
50 years mostly under the labels of process designs and process control. It was not until
1982 when the 1st International Symposium on Process Systems Engineering took place
in Kyoto, Japan, that the term Process Systems Engineering was adopted as the general
term to be applied to the different areas such as process design, control and operations,
including product design.

The creation of Process Systems Engineering as a discipline can be traced back to two
pioneering works. The first one is the article “Integrated Design and Optimization of
Processes” published by Professor Roger Sargent in Chemical Engineering Progress (Vol.
63, Issue 9 , pp 71-78) in 1967. That article was truly visionary in that it outlined the
areas of process design and integration with control and reliability, it addressed
development of process models (steady state, dynamics), strategies of process
calculations, and computational methods for optimization. Furthermore, the article
advocated for collaborations with researchers in control systems, operations research,
numerical analysis and computer science. This pioneering article outlined the future
research agenda for the area of Process Systems Engineering.

The second major pioneering work was the textbok “Strategy of Process Engineering“ by
Professor Dale Rudd and his co-author Charles Watson, Wiley, 1968. The textbook had
three major parts: a) creation and assessment of alternatives, b) optimization, c)
engineering in the presence of uncertainty. The major contribution of this textbook was
2 I.E. Grossmann and I. Harjunkoski

to present systematic approaches for each of these three topics, which nowadays are still
core to process systems engineering, namely synthesis, optimization and uncertainty.

In the next section we describe some of the major accomplishments that have been
achieved in Process Systems Engineering, both at the academic and at the industrial level.

2. Past accomplishments of process systems engineering

2.1. Impact in academic research
In his section we describe some of the major academic accomplishments in the major
areas: process simulation, conceptual design/process synthesis, process control, process
operations and optimization.

2.1.1. Process Simulation

The initial efforts in academic research in PSE were largely directed to process modeling
and process simulation. The first paper on process simulators was the paper on SPEED-
UP by Sargent and Westerberg (1964) at Imperial College that emphasized the ordering
of computations of units in a flowsheet The concept of sequential modular simulators and
its implementation was proposed by Johnson (1972) with the simulator GEMCs at
McMaster University, and by Seader, Seider and Paules (1974) through their
FLOWTRAN program under the CACHE organization.

The next major step was a major project funded by the Department of Energy directed by
Evans, Britt, Chen, and Boston (Chen et al., 1982) at MIT and that led to the Aspen Plus
simulator using the sequential modular structure, currently a major commercial simulator
(Fig. 1). Next major effort took place at Imperial College through the development of
gPROMS by Barton and Pantelides (1994), which in contrast to Aspen Plus adopted the
equation oriented architecture. These efforts promoted the development of other
commercial simulators such as CHEMCAD, Invensys, PRO-II and UNISIM. A major
reference that captures the major concepts and techniques in process simulation is the
book “Process Flowsheeting” by Westerberg, Hutchinson, Motard and Winter (1979).

Figure 1. Simulation of a process flowsheet

2.1.2. Conceptual Design/Process Synthesis

The first textbook that was published in the area of process synthesis was the book
“Process Synthesis” by Rudd, Powers and Siirola (1973). The emphasis of that book was
Process Systems Engineering: Academic and Industrial
Perspectives 3

on the generation of process flowsheets, which in turn motivated the development of the
program AIDS that was motivated by initial ideas of search and problem solving in AI.

Another major development in process synthesis was the concept of pinch for heat
integration that allowed the targeting of utility consumption prior to the detailed synthesis
of heat exchanger networks (Fig. 2). That work was published first in the open literature
by Linnhof and Flower (1978), although Ed Hohmann (1971) had previously reported a
similar result. The main results were eventually published in the “User Guide on Process
Integration for the Efficient Use of Energy,” by Linnhoff (1994) at UMIST.

Figure 2. Pinch analysis with composite curves.

A major development was also conceptual design of process flowsheets through

hierachical decomposition as proposed by Jim Douglas from the University of
Massachusetts in his textbook „Conceptual Design of Processes.,“ (Douglas, 1988). This
work together with the pinch concept motivated the work by Robin Smith at UMIST
through his book „Chemical Process Design and Integration“ (Smith, 2005) in which he
addresses for instance the synthesis of water networks.

Finally, the concept of superstructure was first introduced by Sargent and Gaminibandara
(1976) in the determination of trays in distillation columns (Fig. 3), and for which the
authors modeled the problem as a nonlinear program (NLP). Subsequently, with the idea
of addressing the synthesis of process flowsheets and utility plants, Grossmann and
Santibanez (1980) proposed the use of mixed-integer nonlinear programming (MINLP)
in which 0-1 variables are used to select potential units. These authors proposed
approximations based on MILP as MINLP codes were not available at that point.

Figure 3. A superstructure for synthesizing a process flowsheet.

4 I.E. Grossmann and I. Harjunkoski

2.1.3. Process Control

The initial pioneers in this area were researchers like Foss (1973) at Berkeley and Luyben
(1973) at Lehigh, and subsequently Shinskey (1981), whose major concern was the tuning
of PID controllers. At about the same time, pioneering work on optimal control took place
at Imperial College with the work by Joffee and Sargent (1973).

The next major development took place at the University of Minnesota with the work by
Morari and Stephanopoulos (1980) in which they addressed the synthesis of control
structures, where the major goal was to determine which variables to measure and which
ones to manipulate, both at the level of units and at the level of the entire plant.

Another major step change in process control research took place with the concept of
Model Predictive Control (MPC) pioneered by Cutler (1983) at his company Dynamic
Matrix Control (Fig. 4). Major idea was to optimize the control actions through a receding
horizon. Morari, Garcia, Prett (1989) documented some of the initial successful industrial
applications of MPC, while Mayne, Rawlings, Rao and Scokaert (2000) established
theoretical conditions for stability.

Figure 4. Model Predictive Control (MPC)

2.1.3. Process Operations

This area was not considered until the late 70’s through the pioneering papers by Reklaitis
(1978) at Purdue on process scheduling, and by Mauderli and Rippin (1979) at the ETH
on batch scheduling motivated by the food and pharma industry in Switzerland. Espuña
and Puigjaner (1989) at the Universidad Polytecnica in Barcelona were among the
pioneers in applying scheduling techniques in industrial operations.

Heating Heater
52 20 52

Reactor 1
80 56
Reaction 1 Reactor 2
50

Reactor 1
Reaction 2 80 80
Reactor 2
50 50 50

Reactor 1
80
Reaction 3 Reactor 2
50

Separation Still
130

1 2 3 4 5 6 7 8 9 10

Figure 5. State-Task-Network for batch scheduling.

Process Systems Engineering: Academic and Industrial
Perspectives 5

A major breakthrough took place at Imperial College with the work by Kondili, Pantelides
and Sargent (1993) with the development of the State-Task-Network (Fig.5),. The
importance of this discrete time model that is formulated as an MILP, is that it represents
a very general framework that goes beyond the traditional flowshop and jobshop models
used in Industrial Engineering, as it allows for complex and arbitrary network
configurations.

Other important development was the area of supply chain optimization as proposed by
Shah (2004) for moving away from the plant or unit level, and instead consider global
process networks.

Finally, the area of operability has also emerged as another area of that is relevant for
process operations. A good example is the work on flexibility and resiliency as was first
presented by Grossmann and Morari (1984).

2.1.4. Optimization
Process Systems Engineering has a rich history in the area of optimization. The first
branch and bound code for solving MILP problems was in fact developed by Martin Beale
at British Petroleum in 1958.

The area of nonlinear programming has also been very active in PSE. The Variable-
Metric Projection method by Sargent and Murtagh (1973) was one of the first methods
for solving NLP problems. This subsequently motivated the development of the reduced
gradient method by Murtagh and Saunders (1978) that was implemented in the code
MINOS. Successive-quadratic programming developed by Han (1976) and Powell (1978)
was implemented in the code SNOPT by Gill, Murray and Saunders (2005), and extended
for large-scale problems by Westerberg and Berna (1980). More recent effort in this area
is by Wächter, Biegler (2002) who proposed an interior-point method for NLP and
implemented it in the well-known code IPOPT.

Figure 6. MINLP problem formulation with special cases

The next major step was in the area of mixed-integer nonlinear programming (MINLP,
Fig. 6) in which Duran and Grossmann (1986) proposed the outer-approximation
algorithm that was subsequently extended by Fletcher and Leyffer (1994). This method
was implemented by Viswanathan and Grossmann (1990) as the computer code DICOPT.
6 I.E. Grossmann and I. Harjunkoski

Another important development was the extended cutting plane (ECP) by Westerlund and
Pettersson (1995) that led to to code alpha-ECP.

In the area of optimization under uncertainty a first major contribution was by Reklaitis
in 1975 (Friedman and Reklaitis, 1975) in the area of robust optimization for linear
programming with uncertain right hand sides and matrix coefficients. This was followed
by the work on flexibility summarized by Grossmann and Straub (1991) that dealt with
characterizing regions of feasibility in terms of uncertain parameters and accounting for
control or recourse variables that could react to the uncertainties. Subsequently the work
on parametric programming by Dua and Pistikopoulos (2000) addressed the solution of
various classes of mathematical optimization problems as function of uncertain
parameters so that one could obtain their explicit solutions in terms of critical regions
where in each of them explicit expressions for the objective function and decision
variables is obtained.

The next major area that has received considerable attention by the PSE community is
global optimization. For instance at Princeton Floudas and his group developed the α-BB
method (Adjiman et al., 1998a; Adjiman et al. 1998b) that relied on the idea of modifying
the Hessian of the Lagrangean in nonconvex NLP problems to yield a convex relaxation.
The next major effort was at the University of Illinois by Ryoo and Sahinidis (1996) and
Tawarlamani and Sahinidis (2002) that led to the development of the code BARON that
relied on a spatial branch and bound search coupled with the use of convex
underestimators in factorable functions. Kesavan, Allgor, Gatzke and Barton (2004) at
MIT developed an Outer Approximation algorithm for nonconvex MINLP. Finally,
Floudas and Misener (2013) developed at Princeton GLOMIQO and ANTIGONE in
which a major aim was to handle bilinear terms through piecewise McCormick
underestimations (Bergamini et al, 2008).

Other important efforts by the PSE community have included Logic-based Optimization
at Carnegie Mellon (Raman and Grossmann, 1994), Optimal Control through orthogonal
collocation at Carnegie Mellon (Cuthrell and Biegler, 1987) and integration of adjoints at
Imperial College (Vassiliadis, Sargent and Pantelides, 1994), and Optimization of Hybrid
Systems at Imperial College and ETH, respectively (Barton and Pantelides, 1994;
Bemporad and Morari, 1999).

Finally, PSE researchers have also made major contributions in decomposition

techniques, most notably Benders (1962) at Shell for decomposing large-scale MILP
problems, and Stephanopoulos and Westerberg (1974) at University of Florida for
Lagrangean decomposition in nonconvex NLP problems.

From the above, it is clear that researchers in Process Systems Engineering have had a
major impact in the area of optimization, probably more than any other engineering
discipline.
2.2. Impact in industrial practice
Process Systems Engineering research has become such a fundamental part of industrial
automation and control systems that its results are often integrated into larger industrial
solutions and systems and it is not anymore that trivial to recognize the role of PSE.
Nevertheless, as can be anticipated based on the previous section PSE has had a
remarkable impact on the way today’s industry is managed. In the area of process design,
Process Systems Engineering: Academic and Industrial
Perspectives 7

e.g. energy recovery networks contribute to vast savings both from economics (USD) and
environmental (CO2) aspects, better ways to handle flowsheets and superstructures
contribute both to safer and more efficient production facilities. In the area of process
control model predictive control (MPC) (Garcia et al., 1989) has become a standard and
most control theories and applications such as statistical process control are a result from
PSE activities. In the last decades increasing attention has been given to process
monitoring and fault analysis and diagnosis. In the area of process operations, almost all
industrially used mathematical models for scheduling and supply chain optimization
come from the PSE domain. Here, also data reconciliation and real-time optimization
(RTO) (Marlin and Hrymak, 1997) have been developed closely with the oil & gas
industry. As a result, many optimization and simulation tools today globally used are de
facto results of process system engineering research. These include almost all algorithms
aimed to solve MINLP and global optimization problems.

This is not at all a co-incidence since process system engineering has from its very early
stages had both a strong focus on industrial problems, as well as, a strong interest from
industrial practitioners. There have been many industrial consortia at the main PSE-
focused universities, such as Carnegie Mellon University 1 (since 1985), University of
Manchester 2, McMaster University 3, Imperial College 4, University of Texas at Austin
and University of Wisconsin 5, to name a few. Many of these are still fully active and
benefit from a membership from up to around 25 companies.

Following these facts, the PSE discipline has always been well aware of the relevant
industrial problems and their significance to the practices of designing, operating and
controlling complex processes. Some methodologies, however, need a strong theoretical
fundament and therefore a significant research phase before being able to be deployed in
the industry. A very well-known example on the design of process networks is the pinch-
analysis (Linnhof and Flower, 1978). The theoretical work began in the early 70’s and
about a decade later the work made a big impact as it could prove 30% energy savings at
chemical and petrochemical operations. After successful commercialization (Linnhoff-
March) and another decade roughly 80% of all the world’s largest oil and petrochemical
companies were either clients or sponsors of the company. Finally KBC acquired
Linnhoff-March, which was a few years later acquired by Yokogawa. This is a very
common pattern, where a brilliant PSE technology is first developed and piloted in small
scale but later shows such a great potential that it ends up as a central part of the product
portfolio of a large automation vendor.

There are also companies that have stayed fully focused on the PSE core domain.
AspenTech 6 is a good example of a company with an extremely strong PSE DNA. It
started in early 80’s and first focused on simulation (AspenPlus), which rapidly became

1 http://capd.cheme.cmu.edu/
2 https://www.ceas.manchester.ac.uk/cpi/
3 http://macc.mcmaster.ca/
4 https://www.imperial.ac.uk/process-systems-engineering
5 http://twccc.che.wisc.edu/
6 https://www.aspentech.com/en/about-aspentech/milestones-and-innovations
8 I.E. Grossmann and I. Harjunkoski

the most used simulator both in academia and industry. After this AspenTech has strongly
extended their portfolio from refinery planning, MPC and supply chain management. This
has also involved numerous acquisitions of small PSE-companies and today the
AspenOne technology integrates solutions across engineering, manufacturing, asset
optimization and supply chain functions. Their process information system InfoPlus.21
is one of the leading solutions in the process industries. Similarly, PSE Limited 7 was
founded in the late 90’s as a spin-out from Imperial College and has based its main
success on gPROMS, which is today world’s foremost advanced process modeling
software, also used for research and teaching in over 200 universities around the world.
It has also received numerous awards, e.g. the MacRobert’s Award in innovation, which
has been also cited as “the Nobel Prize for engineering”. Multiple other success stories
include the ADMC work by Charles Cutler that built the foundation of MPC, which is
today a standard component of all automation vendors, as well as, ProSensus by Prof.
John McGregor bringing big data analytics to the process industries already more than a
decade ago. The list of similar developments is long (Haverly Systems, Chesapeake,
SmartOps), not forgetting about the important developments of optimization algorithms
where the PSE-community has played an important role, many of which are today
available in commercial packages such as GAMS, AMPL and AIMMS.

This just highlights the fact that PSE has often contributed to future-oriented
technologies, which after becoming a fundamental part of larger systems lose their PSE-
roots. This is especially the case where the solution goes through multiple acquisitions
and takeovers. One of the largest risk of PSE solutions is not being fully understood either
technically or from their business value. The most typical setup (Fig. 7) is that a PSE
group does a joint project with an industrial partner that first provides the problem
requirements and some representative data for the development. Often these
methodologies are not only theoretically brilliant but also proven in practice through
successful pilot testing, where the collaborating company plays a key role. The main
challenge occurs during hand-over (Harjunkoski et al., 2014). In order to ensure that a
typical PSE solution can be taken over and maintained for years or decades, the company
needs to adapt the organization such that there is a solution or product owner that can
support and maintain the solution also during larger SW-migration projects. Here, the
main guarantee to keep the solutions available is to make them ”PhD-free“, i.e. ensure
that you need not be an expert to be able to manage their daily use and configuration. If
it is not possible to manage the solution by the local resources, even the best solution will
not survive. Unfortunately, companies do not in general anymore have optimization
groups or resources driving various PSE in-house improvement projects.

7 https://www.psenterprise.com/
Process Systems Engineering: Academic and Industrial
Perspectives 9

Figure 7. The path from innovation to an established industrial solution

Another major challenge apart from the lifecycle of a PSE solution is how to prove the
exact value that an optimization solution contributes to the overall value chain. Often the
focus is on optimizing a sub-process, which could be a control loop of a unit process,
heat-exchanger network or scheduling a part of the plant. As improvements are normally
continuous, the neighboring environment may also have ongoing improvement projects,
both hardware as well as software focused, which makes it extremely difficult to pinpoint
the exact contribution of an individual solution to the “big picture”. Sometimes,
neighboring solutions may even be competing against each other, e.g. when trying to
increase the throughput simultaneously having another optimization solution aiming at
minimizing the energy consumption and cost. Furthermore, almost no production section
can work without manual interaction. Process experts frequently take the solution and
modify something to match with the company practice or to avoid some safety issues that
have not been considered in a model. These changes may work against the logic of an
optimization solution and without clearly defined key performance indicators (KPIs) and
understanding their relationship and roles it may often be impossible to tell how much an
investment in a better optimization actually contributes to a better profit. To conclude, it
is often crucial to analyze the entire value chain, understand the relationships between
different sub-systems and focus on the overall improvement.

3. Currents standing of process systems engineering

While the area of Process Systems Engineering has been around technically since 1982
when the term was coined at the 1st International Symposium of Process Systems
Engineering, the fact is that the area suffers from lack of academic recognition. Few
examples to support this claim:
1. Many academics not familiar with PSE label it simply as “process control”.
2. At a recent study at the National Academy of Engineering where major skill
areas were identified for chemical engineering, Process Systems Engineering
was not mentioned.
3. ABET, the accreditation board of engineering studies, never makes reference to
Process Systems Engineering for chemical engineering studies.
4. The AIChE Division “Computing and Systems Technology Division”
supposedly deals with PSE. What happened with the name of the division?
5. ESCAPE, the European meeting for PSE stands for European Symposium
Computer-Aided Process Engineering (No PSE!!).
10 I.E. Grossmann and I. Harjunkoski

6. The flagship journal of PSE is “Computers and Chemical Engineering”, again it

has no reference to PSE.
The above facts motivated a survey described on the next section to gage the perception
of PSE by academics as well as by industrial practitioners.
3.1. Survey results
We made a survey on the role and perception of PSE and sent it out to both academic and
industrial colleagues. The questions were:
1. How many PhD and how many MSc students have you graduated in your
academic career?
2. What are the major companies or universities that have hired your students?
3. Do you think your academic colleagues appreciate the importance of Process
Systems Engineering?
4. Do you think industry appreciates the importance of Process Systems
Engineering?
5. How do you feel about the future prospects of Process Systems Engineering in
both academia and industry?
6. Do you think your company appreciates the importance of Process Systems
Engineering?
7. Do you think universities attach enough importance to Process Systems
Engineering?
8. How do you feel about the future prospects of Process Systems Engineering?

The questionnaire was answered by 81 attendants of which 81% were from academia and
19% from industry. The majority of the answers came from North America (62 %) and
the remaining answers came from Europe (18 %), Latin America (13 %) and Asia (7 %).
It turned out that the persons who responded had altogether graduated about 1200 PhD
students and 1120 MSc students, i.e. more than 2300 persons in the area of PSE. This is
a significant number and proved that the survey results should represent a relevant part
of the community.

The list of companies and universities were vast and cannot be repeated here. It can only
be said that apart from traditional process industries many graduates have also found their
ways to banking / finance, manufacturing industry, as well as, pure software companies.
There are also a significant number of graduates who have chosen an academic career in
all major continents. The responses on the remaining questions 3-8 are shown in Table 1.

Table 1. Responses on the questionnaire

Question Yes / positive Somewhat / neutral No / negative

3 42,4 % 43,9 % 13,6 %
4 57,6 % 40,9 % 1,5 %
5 71,2 % 16,7 % 12,1 %
6 68,8 % 31,2 % 0%
7 6,2 % 43,8 % 50 %
8 81,3 % 12,5 % 6,2 %

The results show concretely that the appreciation of PSE is clearly higher in the industry
than in the academia, which can also be expected taking into account the long-term
synergies with industry (Fig. 8). As the challenges – both on the theoretical and practical
Process Systems Engineering: Academic and Industrial
Perspectives 11

Figure 8. Some results of PSE survey.

level are increasing, most people were very positive about the future prospects of PSE.
The appreciation of PSE in industry was very high and did not get any negative replies.
However, the most alarming signal was that industry feels that academia is not investing
12 I.E. Grossmann and I. Harjunkoski

enough efforts into PSE, leaving them often alone to face the current and future
challenges. At the same time, industry also had a most positive opinion on the future
prospects of PSE, mainly due to the numerous emerging challenges e.g. renewable
energy, digitalization, including Smart Manufacturing and Industry 4.0, both of which
call for more software-centric operations and control technologies. Simultaneously,
exactly this fact was seen as the major threat for PSE. If the field is not sufficiently well
supported by engineers and domain experts, it may be that SW-companies such as
Google, Microsoft and Amazon take over part of the traditional responsibilities of PSE.
It was also seen that the current view on PSE is not the same in the industry and academia,
which would call for more interactions and joint coordination. Frequently there have been
complaints that academia does not get access to realistic industrial problem data, as well
as, industry does not have a chance to test novel appearing methods early enough. Maybe
it is now a good time to join and solve these issues!

4. Future trends and major challenges in process systems engineering

4.1. Industrial Perspective
The main question is how will a process be controlled and operated in the future? Will
the traditional control room operators be replaced by SW-agents? Is the level of, for
instance, autonomy increasing? Will the importance of optimization increase?

On the one hand, there are driving megatrends, such as aging workforce (baby boom
generation is retiring) and infrastructure (many plants were built in the 70’s), shift in
consumer patterns (more individualized products), environmental concerns (climate
change and global warming) and skills gap (there are very few people still working that
know how to build a plant). These are challenges that call for more training, interactions
between generations and maintaining the first principle knowledge.

On the other hand, the disruption especially in digital technologies is significant and the
main focus is slowly shifting from business-to-consumers (mobile phone apps,
entertainment, banking) to business-to-business apps. Here technologies such as machine
learning, big data, time sensitive networking, blockchains, cloud computing, and 5G-
communication are being introduced sometimes even with partially unrealistic hopes.
Nevertheless, this has opened great opportunities for software vendors and almost every
larger SW-company (e.g. Google, Amazon, Microsoft, IBM) is now offering platforms
that can be used to host process data analytics, partially also trusting in that machine
learning (ML) technologies can at least partially replace domain expertise. This is surely
true especially for routine tasks but as ML mainly builds on existing training data, it is an
open question how reliably it would be able to handle unexpected and new situations that
never occurred in the history. Upcoming trends can also be observed in daily technology
news, which report an increasing use of open source software 8, need for standardized
components, companies building up strategic directions for the digitalization and cases
where traditionally isolated problems are mixed and solved together. Software companies

8 https://www.openapc.com/
Process Systems Engineering: Academic and Industrial
Perspectives 13

also approach the process industries 9 with the aim of proving their capabilities in process
analytics and decision making.

Without doubts, if the current trend continues digitalization will play a central role in
forming the future operations. Here, the above enabling technologies will be creatively
mixed and matched in order to provide collaborative value generation. This is truly
moving towards implementing enterprise-wide optimization in practice and due to this
transition there are also many threats that occur and the impact of which no-one
completely understands yet. One of the major threats is cyber security. Through
digitalization the amount of connected software is increasing exponentially and this may
make earlier completely isolated entities vulnerable. One of the ways to estimate and
avoid some potential caveats is to look at industries that are much higher up on the so
called digital s-curve. These are especially financing, insurance, media and ICT
businesses. Banking for instance has been possible already for more than two decades on
a home PC and new security concepts are developed and introduced even today. It seems
very realistic that much of that infrastructure can be modified to be used in process
industries.

The motivation for digitalization of manufacturing operations is very similar to those of

traditional PSE efforts: To connect different parts of the processes and increase
production throughput, product quality, process yield, equipment and energy utilization
and uptime, as well as, improve regulatory compliance. This should happen
simultaneously while decreasing inventory, lead times, regulatory costs, waste, process
and product variations & errors, cycle times, changeover times and maintenance cost. In
principle nothing new, but this time these should happen simultaneously faster and in a
much more holistic manner as before. This is a golden opportunity for process systems
engineering!

The evolution is very visible in digital equipment. Many earlier “dummy” components
are today enhanced by monitoring capabilities and connectivity. Systems are becoming
faster to integrate through flexible configuration tools and more standardized interfaces.
Power systems is a good example on this, especially when looking at the development of
micro-grids and virtual power plants. In larger process systems, asset management has
become more intelligent and many equipment are aware of their states and can estimate
their remaining life time before necessary maintenance. This, combined with real-time
control and monitoring, allows to combine short and long-term planning with
maintenance ensuring that the productivity of a plant can be driven to a maximum.

The challenge for companies comes from the disruption 10. How to ensure that the existing
process can continue to produce and a business does not go bankrupt after too high
investments and unrealistic payback times? Some guidelines are to:

9 https://www.repsol.com/en/press-room/press-releases/repsol-and-google-cloud-optimize-
refinery-management.cshtml
10 https://www.flowcontrolnetwork.com/6-critical-issues-manufacturers-will-face-2016/
14 I.E. Grossmann and I. Harjunkoski

• Ensure step-by-step introduction of internet of things concepts (protect the

investments)
• IoT and new services shall not risk current production (keep production system
stable also during disruption)
• Ensure transparent and controllable plant upgrades (requires also good control
of data flows)
• Use standards to deploy the full and long-term potential of digitalization (where
possible)
• Keep data protection and data integrity high on the agenda (consider different
aspects of cyber security)

Systems become more complex as the increasing number of individual components also
become more complex. This calls for flexible and adaptable design, not only of processes
but also of process systems. Operations aim at 100% availability, which calls for
collaborative optimization models. Simultaneously, these systems are expected to be easy
to manage and configure (compare it to an iPhone) in order to maximize the operator
effectiveness, as well as, reduce the needed number of highly trained individuals to fight
against the competence gap. Figure 9 shows these aspects and emphasize the need for
identifying and supporting synergies. This also calls for a strong collaboration between
wide set of skills. The thought that the PSE-community can handle this alone is
unrealistic, as we already today are building most of our work on software platforms
partially designed by other domains.

Automation Systems Design Process Operations

flexible & adaptable 100% availability

Synergies
Modeling and Optimization Simplicity and Usability
adaptable & collaborative maximum operator effectiveness

Figure 9. Synergies between the traditional PSE areas

Industries have started a journey driven by environmental awareness, governmental

regulation and global competition that will most likely continue with a continuously
increasing speed. Concepts such as food-energy-water nexus have not yet been
considered other than on a very small scale but these types of aspects will evidently lead
to a stronger regulation and dependency on external factors. The PSE community should
also evaluate and adopt the best and most suitable SW-tools and platforms in order to be
able to keep the pace of evolution. Research should not anymore be categorized as
“industrial” or “academic” but rather by its role in a longer time scale. Novel solutions
should be per design easily adaptable and early tested in industry to both evaluate their
Process Systems Engineering: Academic and Industrial
Perspectives 15

capability, as well as, raise new relevant questions in moving the digitalization and PSE
jointly forward to a successful direction.

4.2. Academic Perspective

While there have been many important research accomplishments in Process Systems
Engineering as discussed in section 2.1, it is clear that a number of major challenges
remain in the areas listed below, and for which we provide several representative
examples of on-going research to reinforce the point that research in Process Systems
Engineering remains a vibrant area.

4.2.1. Mathematical Optimization. Here a major challenge continues to be the effective

solution of MINLP problems. One direction has been to expand the capabilities of MINLP
algorithms as has been the case of incorporating quadratic cuts in the outer-approximation
algorithm (Su et al., 2018). A promising direction is the one by Ruth Misener’s group
(Mistry et al., 2018) who is exploring the use of Satisfability Modulo Theories that have
originated in Computer Science and which have shown to provide promising results in
some scheduling problems. Another important challenge are Decomposition Methods
(Benders and Lagrangean) for Global Optimization under Uncertainty, which has been
addressed for instance by Paul Barton and his group (Li et al., 2012). A related problem
is the global optimization of multilevel optimization problems that are being addressed
by Alexander Mitsos through discretization and relaxation techniques (Mitsos et al.,
2008; Mitsos et al. 2010)

4.2.2. Machine Learning. This is an area closely related to optimization as has been shown
by Sahinidis and his group (Cozad et al., 2014; Wilson and Sahinidis, 2017) with the
development of the system ALAMO that allows the selection of a rich set of functions
through the use of global optimization techniques for MINLP like BARON. A related
issue in machine learning is the development of hybrid models that combine basic
physical principles with data driven models based on neural networks as described by
Venkatasubramanian (2019).

4.2.3. Process and Product Design. A major direction in process design has been the
incorporation of molecular design for the simultaneous design of materials such as
solvents and a corresponding process. Claire Adjiman and her group have been
developing mathematical models based on combinatorial search and optimization to
accomplish this objective. On a related effort Gani and co-workers have been developing
ProCAPD, a chemical product design simulator, based on computer-aided methods for
design-analysis of single molecular products (solvents, refrigerants, etc.); mixtures-
blends (gasoline, jet-fuel, lubricants); and liquid formulated products (cosmetics,
detergents, paints, insect repellents). As for process synthesis Chen et al. (2018) are
developing superstructure optimization methods for process flowsheets using
Generalized Disjunctive Programming algorithms in Python combined with global
optimization methods. Major research efforts are also being undertaken by large research
groups such as the RAPID project of AIChE for process intensification (RAPID, 2018)
and the IDAES project lead by NETL (IDAES, 2018).

4.2.4. Planning, scheduling, Enterprise-wide optimization.. Major directions include

embedding inventory policies in the optimization of supply chain planning, as described
in the work by Grossmann and his group (Brunaud and Grossmann, 2017). Refining
16 I.E. Grossmann and I. Harjunkoski

optimization continues to be relevant problem as shown by Ierapetritou and co-workers

with her work on simultaneous scheduling of production units and blending (Shah et al.,
2010). Demand side management has also emerged as an interesting area for scheduling
the operation of power intensive processes to take advantage of changes of hourly
electricity pricing, which inturn translates into improving the grid performance as
discussed by Zhang and Grossmann (2016). Another interesting direction has been the
development of MILP-based online scheduling as economic MPC as proposed by
Christos Maravelias and Jim Rawlings who have applied this tool to the control of HVAC
systems in commercial buildings (Risbeck et al, 2017). Finally, reinforcement learning
has been proposed by Lee for integrating planning and scheduling in process systems
(Badgwell et al, 2018, Shin and Lee, 2019).

4.2.4. Energy and Sustainability. Energy and sustainability have emerged as areas of great
interest in Process Systems Engineering. A new application has been the design and
operation of infrastructure for shale gas production that involves both design decisions
(selection of pipeline and compressors) and planning decisions (drilling and water supply)
giving rise to large scale MINLP models as described in Cafaro and Grossmann (2014)
and Drouven and Grossmann (2016). Another area that has received significant attention
is design of supply chains and processes for biofuels for which MILP models have been
proposed as described in the work by Yue et. al (2014). Integrated solar energy with
production processes has also been addressed through a systems viewpoint as proposed
by Gencer and Agrawal (2017). Also, long-term planning of electric power systems
comprising coal, gas, nuclear, solar and wind sources is also being addressed through
very large scale-MILP models as proposed by Lara et al. (2018). Finally, in order to assess
the environmental impact through life-cycle analysis, pioneered by Azapagic (1999), is
now being expanded to sustainability issues like climate change mitigation (Galan-Martin
et al. (2018).

4.2.5 Healthcare. This area, which traditionally has not been an area of focus for Process
Systems Engineering, is starting to receives some attention by researchers in PSE. An
example is the work by David Bogle who has been addressing the modeling of the liver
as a flowsheet in order to predict ‘ultradian’ oscillations in the concentration of glucose
produced by the liver (Ashworth et al., 2016). Another example is the work by Reklaitis
to reduce adverse effect of drugs for individualized medicine for which Bayesian
inference methods are combined with stochastic programming models (Jayachandran et
al., 2015; Reklaitis, 2017).

5. Conclusions
In this paper, we have presented academic and industrial perspectives on the research and
applications of Process Systems Engineering (PSE). We have shown that there have been
major research accomplishments in the areas of process simulation, conceptual design
and synthesis, process control, process operations and optimization, which in turn have
translated into industrially useful methodologies and software tools. We have also shown
the disconnect between academia and industry in regards to the appreciation of Process
Systems Engineering. The encouraging result has been that both industry and academic
researchers in PSE are optimistic about the future of the area. Finally, we have discussed
some future challenges in PSE from both the industrial and academic perspectives, which
will hopefully motivate the younger generation of researchers and practitioners to keep
Process Systems Engineering a vibrant research area that is industrially relevant.
Process Systems Engineering: Academic and Industrial
Perspectives 17

Acknowledgments
The authors gratefully acknowledge members of the PSE community, academic and
industrial, who in one way or another provided input into this article. Ignacio Grossmann
would also like to thank companies of the Center of Advanced Process Systems
Engineering who have motivated our research over the years and have made it to be
industrially relevant. He would also like to thank the organizers of ESCAPE-27, who
motivated the study on the historical evolution of PSE.

References
Adjiman, C.S., Androulakis, I.P., Floudas, C.A., 1998a. A global optimization method,
αBB, for general twice-differentiabe constrained NLPs-II. Implementation and
computational results. Comput. Chem. Eng., 22, 1159-1179.

Adjiman, C.S., Dallwig, S., Floudas, C.A., Neumaier, A., 1998b. A global optimization
method, αBB, for general twice-differentiable constrained NLPs — I. Theoretical
advances. Comput. Chem. Eng., 22, 1137–1158.

Adjiman C.S., Harrison N.M., Weider S.Z., 2017, Molecular science and engineering: a
powerful transdisciplinary approach to solving grand challenges, Briefing paper,
https://spiral.imperial.ac.uk:8443/handle/10044/1/53609

Ashworth, W; Perez-Galvan, C; Davies, N; Lockhart Bogle, ID. 2016). Liver function as

an engineering system. AIChE J., 62, 3285-3297.

Azapagic, A. 1999. Life cycle assessment and its application to process selection, design
and optimization. Chemical engineering journal, 73 (1), 1-21

Badgwell, T., J.H. Lee and K.H. Liu. 2018. Reinforcement Learning – Overview of
Recent Progress and Implications for Process Control. Computer Aided Chemical
Engineering. Computer Aided Chemical Engineering, 44, Pages 71-85.

Barton, P. I. and Pantelides, C. C. (1994). Modeling of combined discrete/continuous

processes, AIChE J., 40, pp. 966-979

Barton, P.I., Pantelides, C.C., 1994. Modeling of combined discrete/continuous

processes. AIChE J. 40, 966–979.

Bemporad, A., Morari, M., 1999. Control of systems integrating logic, dynamics, and
constraints. Automatica 35, 407–427.

Benders, J.F., 1962. Partitioning procedures for solving mixed-variables programming

problems. Numer. Math. 4, 238–252.

Bergamini, M.L., I.E. Grossmann, N. Scenna and P. Aguirre, “An Improved Piecewise
Outer-Approximation Algorithm for the Global Optimization of MINLP Models
Involving Concave and Bilinear Terms,” Computers and Chemical Engineering 32, 477–
493 (2008).
18 I.E. Grossmann and I. Harjunkoski

Berna, T.J., Locke, M.H., Westerberg, A.W., 1980. A new approach to optimization of
chemical processes. AIChE J. 26, 37-43.

Brunaud, B. and I.E. Grossmann. 2017. Perspectives in Multilevel Decision-making in

the Process Industry. Frontiers of Engineering Management 4, 256-270.

Cafaro, D.C. and I.E. Grossmann, “Strategic Planning, Design and Development of the
Shale Gas Supply Chain Network,” AIChE J. 60, 2122-2142 (2014).

Chen, C.C., Boston, J.F., Britt, H.I., and Evans, L.B. (1982). Local Composition Model
for the Excess Gibbs Energy of Electrolyte Systems - Part I: Single Solvent, Single
Completely Dissociated Electrolyte Systems, AIChE J., 23, pp. 588-596.

Chen, Q., E.S. Johnson, J.D. Siirola, Ignacio E. Grossmann (2018) Pyomo.GDP:
Disjunctive Models in Python.

Cozad, A., N. V. Sahinidis and D. C. Miller, Learning surrogate models for simulation-
based optimization, AIChE J., 60, 2211-2227, 2014.

Cuthrell, J.E., Biegler, L.T., 1987. On the optimization of differential‐algebraic process

systems. AIChE J. 33, 1257-1270.

Cutler, C. R. (1983). Dynamic matrix control: an optimal multivariable control algorithm

with constraints. PhD thesis, University of Houston

Douglas, J.M., Conceptual design of chemical processes,.McGraw‐Hill, New York, 1988

Drouven, M.G. and I.E. Grossmann, “Multi-Period Planning, Design and Strategic
Models for Long-Term, Quality-Sensitive Shale Gas Development,” AIChE J., 62, 2296-
2323 (2016).

Duran, M.A., Grossmann, I.E., 1986. An outer-approximation algorithm for a class of

mixed-integer nonlinear programs. Math. Program. 36, 307–339.

Espuña, A. and Puigjaner, L. (1989). On the solution of the retrofitting problem for
multiproduct batch/semicontinuous chemical plants.Comput. Chem. Eng., 13(4-5), 483-
490.

Fletcher, R., Leyffer, S., 1994. Solving mixed integer nonlinear programs by outer
approximation. Math. Program. 66, 327–349.
Foss, A. S. (1973). Critique of chemical process control theory. AIChE J., 19(2), 209-
214.

Friedman, Y., Reklaitis, G. V., 1975. Flexible solutions to linear programs under
uncertainty: Equality constraints. AIChE J. 21, 83-90.

Galan-Martin A, Pozo C, Azapagic A, Grossmann IE, Mac Dowell N, Guillen-Gosalbez.

2018. Time for global action: an optimised cooperative approach towards effective
climate change mitigation, Energy & Environmental Science, Vol: 11, Pages: 572-581
Process Systems Engineering: Academic and Industrial
Perspectives 19

Garcia, C. E., Prett, D. M. and Morari, M. (1989). Model predictive control: theory and
practice—a survey. Automatica, 25(3), 335-348.

Gencer, E., R. Agrawal, Strategy to synthesize integrated solar energy coproduction

processes with optimal process intensification. Case study: Efficient solar thermal
hydrogen production, Comput. Chem. Eng. doi:10.1016/j.compchemeng.2017.01.038.

Gill, P.E., W. Murray, and M. A. Saunders (2005), SNOPT: An SQP algorithm for large-
scale constrained optimization, SIAM Review 47(1), 99-131.

Grossmann, I.E. and D.A. Straub, "Recent Developments in the Evaluation and
Optimization of Flexible Chemical Processes," Proceedings of COPE '91 (Eds. Puigjaner
and Espuña), 41, Elsevier (1991).

Grossmann, I.E. and M. Morari, "Operability, Resiliency and Flexibility - Process Design
Objectives for a Changing World," Proceedings of 2nd International Conference on
Foundations of Computer-Aided Process Design (Eds. A.W. Westerberg and H.H.
Chien), 931 (1984).

Grossmann, I.E. and Santibanez, J. (1980). Applications of mixed-integer linear

programming in process synthesis,Comput. Chem. Eng., 4, pp. 205–214.

Han, S.P., "Superlinearly Convergent Variable Metric Algorithms for General Nonlinear
Programming Problems," Math Progr., Vol. 11, p. 263-82, 1976.

Harjunkoski, I., Maravelias, C.T., Bongers, P., Castro, P.M., Engell, S., Grossmann, I.E.,
Hooker, J., Méndez, C., Sand, G. and Wassick, J. (2014). Scope for industrial applications
of production scheduling models and solution methods.Comput. Chem. Eng., 62, pp. 161-
193.

Hohman, E.C. Optimum networks for heat exchange. University of Southern California:
1971.

Jayachandran, Devaraj, José Laínez-Aguirre, Ann Rundell, Terry Vik, Robert

Hannemann, Gintaras Reklaitis, Doraiswami Ramkrishna. 2015. Model-based
individualized treatment of chemotherapeutics: Bayesian population modeling and dose
optimization. PLOS/ONE. https://doi.org/10.1371/journal.pone.0133244

Joffe, B. L. and Sargent, R. W. H. (1973). A naive approach to the optimal control of

nonlinear stochastic systems. Recent mathematical developments in control.(A 74-23558
09-19) London and New York, Academic Press, 1973, 83-96.

Johnson, A.I., 1972. Computer Aided Process Analysis and Design - A Modular
Approach, Brit. Chem. Eng, 7, 28

Kalakul, S., M. Eden, Rafiqul Gani (2017). The Chemical Product Simulator – ProCAPD,
Computer Aided Chemical Engineering, 27th European Symposium on Computer Aided
Process Engineering, Barcelona, Spain, 979-984.
20 I.E. Grossmann and I. Harjunkoski

Kesavan, P., Allgor, R.J., Gatzke, E.P., Barton, P.I., 2004. Outer approximation
algorithms for separable nonconvex mixed-integer nonlinear programs. Math. Program.
100, 517–535.

Kondili, E., Pantelides, C. C. and Sargent, R. W. H. (1993). A general algorithm for short-
term scheduling of batch operations—I. MILP formulation.Comput. Chem. Eng., 17(2),
211-227.

Kronqvist, J., D.E. Bernal, I.E. Grossmann, “Using Regularization and Second Order
Information in Outer Approximation for Convex MINLP” to appear in Mathematical
Programming (2018).

Lara, C. L., D. Mallapragada, D., Papageorgiou, A., Venkatesh, and I.E. Grossmann,
“Electric Power Infrastructure Planning: Mixed-Integer Programming Model and Nested
Decomposition Algorithm,” European Journal of Operational Research 271, 1037–1054
(2018).

Li, X., Chen Y, Barton PI. 2012. Nonconvex Generalized Benders Decomposition with
Piecewise Convex Relaxations for Global Optimization of Integrated Process Design and
Operation Problems. Industrial & Engineering Chemistry Research. 51:7287-7299.

Linnhoff, B. A User Guide on Process Integration for the Efficient Use of Energy.
Institution of Chemical Engineers: 1994.

Linnhoff, B. and Flower, J.R. (1978). Synthesis of heat exchanger networks: I. Systematic
generation of energy optimal networks, AIChE J., 24, pp. 633–642.

Luyben, W. L. (1973). Parallel cascade control. Industrial & Engineering Chemistry

Fundamentals, 12(4), 463-467.

Marlin, T. and Hrymak, A. (1997). Real-time operations optimization of continuous

processes. American Institute of Chemical Engineering Symposium Series - Fifth
International Conference on Chemical Process Control (CPC-5), 93, 156-164.
Mauderli, A. and Rippin, D. W. T. (1979). Production planning and scheduling for multi-
purpose batch chemical plants.Comput. Chem. Eng., 3(1-4), 199-206.

Mayne, D. Q., Rawlings, J. B., Rao, C. V. and Scokaert, P. O. (2000). Constrained model
predictive control: Stability and optimality. Automatica, 36(6), 789-814.

Misener, R., Floudas, C.A., 2013. GloMIQO: Global mixed-integer quadratic optimizer.
J. Glob. Optim. 57, 3–50.

Misener, R., C.A. Floudas ANTIGONE: algorithms for continuous/integer global

optimization of nonlinear equations, J. Glob. Optim. 59, 503-526.

Mistry, M., A. Callia D’Iddio, M. Huth and R. Misener, Satisfiability Modulo Theories
for Process Systems Engineering, Optimization On-line (2018).
http://www.optimization-online.org/DB_HTML/2017/06/6086.html
Process Systems Engineering: Academic and Industrial
Perspectives 21

Mitsos, A., P Lemonidis, PI Barton. 2008. Global solution of bilevel programs with a
nonconvex inner program Journal of Global Optimization 42 (4), 475-513.

Mitsos, A., 2010. Global solution of nonlinear mixed-integer bilevel programs, Journal
of Global Optimization 47 (4), 557-582.

Morari, M. and Stephanopoulos, G. (1980). Studies in the synthesis of Control Structures

for Chemical Processes: Part II: Structural aspects and the synthesis of alternative feasible
control schemes. AIChE J., 26(2), 232-246.

Powell, M.J.D., 1978 (a). "A Fast Algorithm for Nonlinearly Constrained Optimization
Calculations," In Numerical Analysis, Dundee 1977. G.A. Watson (ed.), Lecture Notes
in Mathematics 630, Springer-Verlag, Berlin

Raman, R., Grossmann, I.E., 1994. Modelling and computational techniques for logic
based integer programming. Comput. Chem. Eng. 18, 563–578.

Reklaitis, G.V., 2017. Prospects and Challenges for Process Systems Engineering in
Healthcare. Computer Aided Process Engineering, 3-7.

Reklaitis, G.V. Review of Scheduling of Process Operation. AIChE. Symp. Ser.1978, 78,
119-133.

Risbeck, M.J.; Maravelias, C.T.; Rawlings, J.B.; Turney, R.D. 2017. A Mixed-Integer
Linear Programming Model for Real-Time Cost Optimization of Building Heating,
Ventilation, and Air Conditioning Equipment, Energy and Buildings, 142, 220-235.

Rudd, D.F., Powers, G.J., and Siirola, J.J. Process synthesis. Prentice-Hall: 1973.

Ryoo, H.S., Sahinidis, N. V, 1996. A branch-and-reduce approach to global optimization.

J. Glob. Optim. 8, 107–138.

Sargent, R.W.H. and B.A. Murtagh, “Projection Methods for Non-Linear Programming,”
Mathematical Programming, 4, 245-268 (1973).

Sargent, R.W.H. and Gaminibandara, K. (1976). Optimum Design of Plate Distillation

Columns. In Optimization in Action. Dixon, L.C.W., Ed. Academic Press: London, pp.
267–314.

Sargent, R.W.H. and Westerberg, A.W. (1964). SPEED-UP (Simulation Programme for
the Economic Evaluation and Design of Unsteady-State Processes) in chemical
engineering design, Trans. IChemE. 42:190–97.

Seader, J.D., Seider, W.D., and Pauls, A.C. Flowtran Simulation: An Introduction.
Monsanto Company: 1974.

Shah, N. (2004). Pharmaceutical supply chains: key issues and strategies for
optimisation.Comput. Chem. Eng., 28(6-7), 929-941.
22 I.E. Grossmann and I. Harjunkoski

Shah, N.K., Z Li, MG Ierapetritou. 2010. Petroleum refining operations: Key issues,
advances, and opportunities. Industrial & Engineering Chemistry Research 50 (3), 1161-
1170.

Shin, J., J. Lee. 2019. Multi-Scale Multi-Stage Decision-Making Model Development by

Combining Reinforcement Learning and Mathematical Programming. In preparation

Shinskey, F. G. (1981). Controlling Multivariable Processes: An Independent Learning

Module from the Instrument Society of America. Instrument Society of America.

Smith, R. Chemical Process Design and Integration. Wiley: 2005.

Stephanopoulos, G., Westerberg, A.W., 1975. The use of Hestenes’ method of multipliers
to resolve dual gaps in engineering system optimization. J. Optim. Theory Appl. 15, 285–
309.

Su, L., L. Tang D. E. Bernal, I. E. Grossmann, “Improved quadratic cuts for convex
mixed-integer nonlinear programs,”Comput. Chem. Eng. 109, 77–95 (2017).

Tawarmalani, M., Sahinidis, N. V, 2002. Convexification and Global Optimization in

Continuous and Mixed-Integer Nonlinear Programming, Springer.

Vassiliadis, V.S., Sargent, R.W.H., Pantelides, C.C., 1994. Solution of a Class of

Multistage Dynamic Optimization Problems. 2. Problems with Path Constraints. Ind.
Eng. Chem. Res. 33, 2123–2133.

Venkatasubramanian, V. 2019. The promise of artificial intelligence in chemical

engineering: Is it here, finally? AIChE J., 65, 466-478.

Viswanathan, J., Grossmann, I.E., 1990. A combined penalty function and outer-
approximation method for MINLP optimization. Comput. Chem. Eng. 14, 769–782.

Wächter, A., Biegler, L.T., 2005. On the implementation of an interior-point filter line-
search algorithm for large-scale nonlinear programming. Math. Program. 106, 25–57.

Westerberg, A.W., Hutchison, H.P., Motard, R.L., and Winter, P. Process Flowsheeting.
Cambridge University Press: 1979.

Westerlund, T., Pettersson, F., 1995. An extended cutting plane method for solving
convex MINLP problems. Comput. Chem. Eng. 19, 131–136.

Wilson, Z. T and N. V. Sahinidis, The ALAMO approach to machine learning, Computers

and Chemical Engineering, 106, 785-795, 2017.

Yue, D., F. You, S.W. Snyder. 2014. Biomass-to-bioenergy and biofuel supply chain
optimization: overview, key issues and challenges.Comput. Chem. Eng. 66, 36-56.

Zhang, Q. and I.E. Grossmann. 2016. Enterprise-wide Optimization for industrial

demand side management: Fundamentals, advances and perspectives. Chemical
Engineering Research and Design 116, 114-131.