INDUSTRIAL EXPERIENCE WITH OBJECT-ORIENTED
MODELLING
FCC Case Study
R. MADHUSUDANA RAO1, R. RENGASWAMY1,*, A. K. SURESH2 and K. S. BALARAMAN3
1
Department of Chemical Engineering, Clarkson University, Potsdam, New York, USA
Department of Chemical Engineering, Indian Institute of Technology, Bombay, India
3
Research and Development, Chennai Petroleum Corporation Limited, Chennai, India
2
P
rocess modelling and simulation have emerged as important tools for detailed study and
analysis of chemical processes. In activities such as design, optimization and control of
processes, realistic process models, which incorporate physics and chemistry of the
process in adequate detail, are becoming almost indispensable. Simulation studies also provide
guidance in the development of new processes and can reduce both time and capital
investment. A difficulty with process models is that they are based on the state of knowledge
and simulation objectives defined at the time of their formulation. In addition, it is not easy to
modify process models to incorporate new knowledge as it becomes available and as new needs
arise. There is a need, therefore, to use advanced modelling and simulation strategies such that
refinements and additional capabilities can be incorporated in the model without disproportionate additional effort. This work presents the framework of one such multipurpose process
simulator, MPROSIM, an object-oriented process modelling and simulation environment.
Though considerable literature is available on process modelling from a subjective or
theoretical viewpoint, very little has been published on application of these ideas on complex
industrial-scale processes. This being the focus of the paper, a case study of an object-oriented
model for automatic generation of a fluid catalytic cracking unit (FCCU) reactor=regenerator is
presented. The utility of the framework is illustrated by demonstrating how the model for
FCCU could be fine-tuned both structurally and parametrically to represent the behaviour under
changing process operating conditions.
Keywords: FCCU; FCC model; object-oriented model; process modelling; MPROSIM.
INTRODUCTION
There is a greater need today than ever before to design and
operate chemical processes with a high degree of understanding. This need arises from several factors: an incessant
demand for higher quality yields and better products; an
increasingly competitive environment that forces plants to
be operated in an optimal manner; variable raw material
quality; stringent environmental and safety regulations; an
increased level of automation arising out of the above
factors, and so on. This need for a greater level of understanding has to be seen in context of the fact that chemical
processes are usually complex to understand and operate.
The complexity arises at different levels: at the level of
physicochemical phenomena involved; at the equipment
level; and finally at the level of the plant where topological
factors resulting from recycles and other connectivities
impose additional demands. All these factors mean that
mathematical models of increasing complexity have to be
used in order to design and operate processes profitably.
Good models can also guide the development of new
processes and can substantially reduce both time and capital
requirements in process development. The utility of a
process model strongly depends on its predictive capabilities.
The predictions should be reliable over wide ranges of feed
composition and process conditions. The availability of such
a model, in conjunction with a simulation tool, can then
facilitate better insight into plant behaviour through simulations instead of the more expensive and time consuming
route involving actual experimentation.
The advantages resulting from efforts invested in the
activities of process modelling and simulation have to be
seen against the fact that any mathematical model is
necessarily based on a set of assumptions. Typically, the
assumptions are governed by several factors: objectives
the model is meant to serve at the time of its formulation;
528
the need to keep the model tractable, consistent with these
objectives; state of knowledge about the processes and
equipment at the time of formulation of model, and so on.
Any or all of these are subject to substantial changes during
the life of a plant. It would be therefore desirable to choose a
framework for modelling and simulation that allows the
flexibility of incorporating changes in parts of the model
without having to undertake the entire modelling exercise
anew. The present work concerns the development of
a process modelling and simulation environment that
attempts to address such concerns. The framework, called
MPROSIM, is based on the concepts of object-oriented
modelling and simulation.
In the next section, we first describe the idea and concept
of MPROSIM. Some of the recent contributions in process
modelling, and methodology of process modelling and hierarchy of modelling objects in MPROSIM are also described. To demonstrate the utility and merits of MPROSIM, we
present a case study that concerns the modelling and
simulation of FCCU, which in many ways typifies the
types of challenges in process modelling and simulation
that MPROSIM seeks to address. In MPROSIM, FCCU is
modelled as a flowsheet using the fundamental chemical
process units. The object-oriented design and representation
of the model as a flowsheet are also described. Various
studies conducted with the model such as reactor tuning,
regenerator tuning, parametric studies on integrated reactor=
regenerator model and yield optimization studies, are also
presented. We conclude this work with some comments on
the utility of an object-oriented framework such as MPROSIM and extension of the framework to form a single
environment wherein various process engineering activities
can be conducted.
MPROSIM—MULTIPURPOSE
PROCESS SIMULATOR
Concept
Typically, in a chemical production unit, process models
are used offline instead of online. Some of the reasons for
this may have to do with issues of reliability of the process
model, inadequate knowledge of the model input parameters, and presence of noise in the data collected from
plant. Model reliability is established through extensive
validation, after the model has been tuned by estimating the
model parameters. Hence, steady-state simulation, data
reconciliation (for error detection in the plant data) and
parameter estimation are some of the process engineering
activities that are very closely related. Typically, these studies
would be carried out in different environments, and would
also often be based on different models. The main aim of
MPROSIM is to develop an integrated framework wherein
various process engineering activities can be carried out
within the same environment and with the same model.
Review of Process Modelling
Traditionally, modelling involves identifying the key
phenomena occurring in the system that influence the
features of interest, developing mathematical equations to
describe the phenomena, and solving these equations, most
often using numerical techniques with the help of computers.
Model development as an activity is expensive, requiring as
it does highly skilled manpower with a good amount of
process knowledge. Considerable work is also involved in
validating the models. Because of the high costs associated
with model development and validation, it is desirable that
the models retain their relevance and usefulness over a
reasonably long period of time.
In recent years, a variety of process modelling and
simulation tools have been developed. As for the latter,
there are several tools, some of which have become
commercially successful e.g., ASPEN PLUS, PRO II, etc.
These tools are widely used in the petroleum industry as
steady state simulators. Some of the reasons for their
extensive use are: fluid data handling capabilities, physical
property and component data base, and an easy to use
graphical user interface (GUI) for unit operations and
flowsheet simulations. Extending these tools to model
more specific or complex processes is difficult. Some of
the typical examples that can be cited are mineral and solids
processes that involve particulate systems, membrane
processes, polymer reaction systems and multiphase reactors. To illustrate this, consider a particulate process, where
it would be necessary to describe the properties of
the dispersed phase (say solids), such as particle size,
particle shape and=or porosity, as a part of the stream
input. General purpose simulators allow only for the definition of a discretized particle size distribution by mass for
each solids stream (Gruhn et al., 1997). Though the particle
size distribution may be easily added as a user defined
variable for the process stream, the user must also supply the
code to handle the additional data at each unit model, i.e.,
even at a simple splitter (Toebermann et al., 2000). Also, the
parameters for a process unit (e.g., size reduction equipment) depend on the type of apparatus, process-relevant
design, operating variables (gap size) and materials being
processed (Gruhn et al., 1997). Apart from the extensibility
constraints, one of the main drawbacks of these simulation
tools is that the models and problem solver are tightly
coupled in software modules, i.e., FORTRAN subroutines
(Nilsson, 1993).
Because of the difficulties in adaptability and extensibility
of process simulation tools, process modelling tools based
on equation-oriented approach have gained importance.
Examples of commercial simulators are SPEEDUP and
gProms. They support the implementation of unit models
and their incorporation in a model library, i.e., the user can
specify all model equations. Modelling of a process unit
would require a profound knowledge not only of chemical
engineering, but also of such diverse areas as modelling and
simulation, numerical mathematics and computer science.
Hence, extending the standard models and further development of specific unit models can be undertaken only by a
small group of modelling experts. Successful attempts have
been made on extending SPEEDUP and gProms for modelling specific units of mineral processing problems (Barton
and Perkins, 1988) and crystallization processes (Pantelides
and Oh, 1996). However, the large effort for the set-up and
evaluation of the equation system, as well as the expert user
knowledge, is a disadvantage (Toebermann et al., 2000).
Equation-oriented languages sometimes result in redundant
modelling as they do not assist the user in developing
models using engineering concepts and thus, reuse of even
529
well-designed and validated models is sometimes compromised (Marquardt, 1996).
While both the above approaches have their advantages
and disadvantages, the experience of early researchers in
these two approaches towards process modelling has
triggered considerable interest over the last decade. Efforts
have been directed towards addressing issues like model
formulation, reusability of process models, extensions
and adaptability to changing conditions. As a result of the
efforts of various researchers, some of the developments
in advanced process modelling tools and environments
are: ASCEND (Piela et al., 1991), gPROMS (Barton and
Pantelides, 1994), MODEL.LA (Stephanopoulos et al.,
1990), and OMOLA (Nilsson, 1993). One of the common
features of all modelling tools is multi-level modularization.
The idea of modularization has been inherited from the
concepts of object-orientation, which in turn has developed
as a branch of computer science and engineering. The main
idea of object-oriented design is to break a large and
complex system into smaller fragments or modules so as
to attain abstraction at each modular level. The objectoriented approach directly supports reusability of common
functions in various parts of the program. Further, it also
allows extensibility of the whole system by the implementation of modules. The reader is referred to the original
literature cited above for a detailed summary on the development of each modelling tool. State-of-the-art reviews on
developments in process modelling are also available
(Marquardt, 1991, 1996; Biegler, 1989; Boston et al.,
1993; Pantelides and Barton, 1993).
Some of the recent advances in the area of process
modelling have been towards a systematic approach to the
development of process models (process modelling methodology) and formal representation of process model equations (Marquardt, 1994, 1996; Bogusch and Marquardt,
1995, 1997; Lohmann and Marquardt, 1996). Marquardt
and his group have laid the foundations for systematic
process model development and at the same time have
proposed an object-oriented methodology for the task of
computer-aided process modelling. Extending the ideas of
system theoretic approach and object-orientation, a process
unit model can be decomposed into smaller modules based
on its structure (for example, reactor wall, catalyst, heat,
material, etc., for a tubular reactor) and behaviour (conservation, constitutive equations, etc.). Similarly, the behavioural
objects can further be decomposed into more fundamental
entities (for example, holdup, transport, transfer, etc., for a
species conservation equation). Following this methodology,
Marquardt and his group have proposed a structure for
hierarchy of elementary process quantities for developing
a unit model. A real unit model is defined by aggregating a
set of these canonical=fundamental modelling objects.
The main aim of adopting such a modelling methodology
is to address the following two issues: (1) at the fundamental
level, any chemical process, be it petroleum, petrochemical,
metallurgical or biological, is the same i.e., conservation
of species, components or particles and conservation=
conversion of energy. Hence, with a well-structured and
generic modelling tool, it should be possible (ideally) to
define any model by aggregating the elementary modelling
objects; and (2) to be able to develop a unit model from a
fundamental approach and also reuse and extend the existing library of models to fit the changing requirements.
However, in spite of all the developments in the area of
object-oriented structured process modelling over the last
decade, there still remain several important issues that need
to be addressed. They can be classified into two categories:
Generic issues. Some of the issues are:
— the amount of effort and expert knowledge that would
be required in various fields of science and engineering for developing such a fundamental and generic
modelling tool.
— the capabilities and conduciveness of the existing
modelling and software tools for encoding and representing the vast amount of diversified scientific and
engineering knowledge in a structured format.
— the extent of granularity that one should build to
model any given process system. One of the practical
issues in this context is that a given hierarchy of
process model structuring, however fine it may be in
its granularity, might still be not complete. For
example, in mineral processing, the liberation state,
in gravel and sand industry processes, the fractional
density and in environmental processes like soilwashing, the fractional contamination must be considered along with the particle size distribution to
characterize a process stream (Toebermann et al.,
2000). Similarly, processes from different fields of
engineering and technology may require additional
input and parameter specifications for both process
streams and process models.
— uniqueness of a decomposition strategy. A chemical
process can be represented by an aggregation of
several possible combinations of fundamental
modelling objects of varying degrees of complexities. Thus, the decomposition strategy for defining
the model objects gives rise to many degrees of
freedom. The problem of selection of a particular
strategy for process representation can be a
difficult task when a large scale industrial process
is considered.
— extension and adaptability of a set of fundamental
modelling objects. A generic process modelling
framework must be capable and flexible to be able
to extend and incorporate additional information=
knowledge.
Implementation issues:
— Though some general guidelines for structured
process modelling have been stated (Marquardt,
1992, 1996), the utility of the formalism can only
be shown by the implementation and evaluation of
modelling tools along with experimentation.
While there is considerable literature on structured
process modelling from a subjective or theoretical viewpoint, very little has been published on application of
these ideas to complex real-life modelling problems. To
the authors’ best knowledge, there has been no direct
evaluation of the formalism or the guidelines for structured process modelling through an implementation on a
large scale industrial process. This is the main focus of
the paper and FCCU is used as a case study. The project
was undertaken in joint collaboration with a refinery,
Chennai Petroleum Corporation Limited (CPCL), India.
Though there are few commercially available simulation
530
tools for FCCU like FCC Plus, it is necessary to reiterate
that the concepts of object-oriented modelling and results
presented in this contribution are not to undertake an
evaluation of the simulation tools. The modelling and
simulation of FCCU is chosen to demonstrate the merits
of MPROSIM and at the same time present an evaluation
of structured object-oriented process modelling guidelines.
The reasons for the choice of FCCU are manifold:
FCCU is a critical unit of a refinery, and is thus important
in its own right.
FCCU is a large system with complexities at several
levels: heterogeneous operations, solids handling, hydrodynamic complexities in the riser and fluid-bed regenerators, complex kinetics resulting from the complexity of
feedstock, topological complexities due to recycles and
interconnected flows, etc. It is therefore a good test case
to test the implementation and evaluation of the ideas of
modularization and structural decomposition.
Various configurations of reactor and regenerator are in
operation in the refineries. Adaptability of the modelling
framework to different configurations can therefore be
tested.
Inputs to the unit often change in terms of both quality
(composition of feed) and quantity (product demands in
the market). Often, a new or improved catalyst becomes
available. This is precisely the situation where a simulation tool can prove its utility, since fine tuning of the
operating conditions in response to such changes is often
needed.
Various assumptions have been made in literature for
modelling the physics and chemistry of the cracking
process. As a result, several approaches are available for
modelling different sections of FCCU. There is thus an
opportunity to test the flexibility of the framework to
model the system using different approaches.
In the next section, we briefly describe the object-oriented
modelling framework developed in MPROSIM. The framework is based on the principles of systematic process
modelling as enumerated in many of the works discussed
above.
Modelling in MPROSIM
MPROSIM is a framework for multipurpose process
modelling and simulation. The simulator is based on the
equation-oriented approach. It has been developed using
the concepts of object-oriented programming (OOP). The
modelling environment and the graphical user interface
(GUI) are developed using JAVA. Even though some of
the simulator framework features are under development and
several program module implementations are rudimentary,
the overall software architecture of MPROSIM is illustrated in Figure 1. The key idea that is being explored in
MPROSIM is to make every activity of process engineering
a class or an object. It is also designed to make all the
important and essential features of process engineering
completely accessible from the GUI. Hence, the overall
architecture of MPROSIM is flat, allowing the user to
carry out various process engineering activities with the
same model built using the GUI.
Object-oriented modelling is a modern approach to
handle complexities. The method involves modularization
Figure 1. Software architecture of MPROSIM.
and characterization of data into an abstraction called class.
A class is like a blueprint for an application or part of an
application. An object, which is an instance or an incarnation of a class, has both substance and behaviour by
comprising data and functions that perform operations on
this data. A well abstracted class can be tested and implemented independently. Abstraction helps in the reuse of a
class=object in various parts of the program. Functions
common to many classes can reside in one class and can
be implemented in other parts of the program with the help
of inheritance. The common functionalities can be adapted
to meet specific requirements with the help of polymorphism. To protect parts of the program from unintentional
changes or side effects, variables and elements of an object
can be encapsulated. The flow of control in the program can
be threaded with the help of multithreading. Multithreading
provides a way for an application to handle many different
tasks at the same time (Niemeyer and Knudsen, 2000). Even
though JAVA supports only single inheritance, it allows
multiple implementation of interfaces. All these features
make JAVA a conducive language to design the GUI and the
modelling framework.
The methodology of process modelling in MPROSIM has
been designed and developed based on the above concepts
of object-oriented programming and modelling. The hierarchy of process modelling in MPROSIM, as of present, is
depicted in Figure 2. The first step is a topological fragmentation where a process is broken into model fragments
or unit modules based on structure of the process. The
representation of the original process is achieved by the
aggregation of these model fragments with the help of
streams in the form of a flowsheet. Each of the model
fragments resulting from the first step of decomposition is
modular and well abstracted. The model fragments can be
tested and implemented independent of other fragments,
thus ensuring their reuse. At the second step, a model
fragment or unit module is further decomposed into molecular fragments based on its internal structure and behaviour. A model fragment can be instantiated by a collection
of the molecular fragments.
In the present development stage of MPROSIM, the
hierarchial modelling structure lies at the molecular fragment
level. Further decomposition of the molecular fragments
into atomic fragments and issues relating to its testing and
implementation are some of the developments presently
under progress. For the FCCU case study, further
531
Figure 2. Hierarchy of process modelling in MPROSIM.
decomposition of molecular fragments could not be undertaken due to limitations such as project schedule, data
collection, etc. Though the modelling hierarchy built
currently is neither suitable nor extensible to model any
given process system, it is appropriate to address various
issues relating to the modelling of an industrial-scale FCCU.
In addition, some of the model fragments resulting from the
decomposition are generic in nature and can be extended to
model other specific units=processes as highlighted in
section on Discussion and Clarifications, below.
For the implementation of the FCCU model, a node class
is defined as the molecular fragment, which contains all the
typical model equations such as mass balance, energy
balance, etc. Every unit module or model fragment inherits
the node class and can modify or add additional information
to the inherited model equations to reflect its own behaviour.
The FCCU model is realized in the form of a flowsheet by
an aggregation of unit modules by connecting them with the
help of streams. A detailed description of the object-oriented
methodology for FCCU modelling in MPROSIM is given in
a subsequent section. As a precursor, a brief review of the
literature on FCCU modelling is presented in the next
section.
Review of FCCU Modelling
Over the years, many models have been proposed for
FCCU. These models have been based on different sets of
assumptions with respect to the kinetics of cracking reactions and hydrodynamics of the equipment involved, such as
riser and regenerator. Some models concern themselves only
with the regenerator (Ford et al., 1976; Errazu et al., 1979;
de Lasa et al., 1981; Guigon and Large, 1984; Krishna and
Parkin, 1985; Lee et al., 1989a). Some have only reactor or
cracking models (Weekman and Nace, 1970; Paraskos et al.,
1976; Jacob et al., 1976; Shah et al., 1977; Lee et al.,
532
1989b; Larocca et al., 1990; Takatsuka et al., 1987). There
also exist integrated models coupling both the regenerator
and reactor (Kumar et al., 1995; Lee and Kugelman,
1973; McGreavy and Isles-Smith, 1986; Bozicevic and
Lukec, 1987; Arandes and de Lasa, 1992; McFarlane et al.,
1993; Arbel et al., 1995; Arandes et al., 2000). Table 1
presents a list of literature which typify the approaches that
have been considered for FCCU modelling.
The variety of approaches present in the literature can be
analysed with reference to Figures 3 and 4. The reactor riser
is often modelled as being in plug flow (Arbel et al., 1995)
with uniform temperatures across a cross section and a
temperature gradient along the height of the riser. Other
assumptions, such as quasi steady state, no slip, adiabatic
operation, uniform temperature of the two phases at any
point and constant heat capacities of oil vapour and catalyst
are usually made in the analysis. On the other hand,
McFarlane et al. (1993) assume isothermal conditions (i.e.,
CSTR type) in the riser, though the bottom of the riser
involves a non-isothermal zone due to a finite mixing time.
Apart from this, riser models differ mainly in their considerations of reaction kinetics. Weekman and Nace (1970)
developed a three lump model which was used by many
authors. Lee and Groves (1985) used the three lump model in
their integrated model. Other models based on the WeekmanNace three-lump model include those of McGreavy and IslesSmith (1986), Kraemer and de Lasa (1988), Arandes and
de Lasa (1992). Other authors expanded the three lump
system into more lumps. Jacob et al. (1976) developed a
ten lump system. A four lump model was developed by Lee
et al. (1989b). Bozicevic and Lukec (1987) published a five
lump model. Takatsuka et al. (1987) developed a Weekman
type six lump model. Kraemer et al. (1990) extended their
model to eight lumps. Two papers appeared in 1995 using ten
lump kinetics with an integrated model. Kumar et al. (1995)
used an isothermal reactor riser and Arbel et al. (1995)
modelled the riser assuming plug flow. Arandes et al.
(2000) have considered the ten lump kinetics for dynamic
and steady-state model of FCCU. Plug flow conditions were
assumed in the riser and for gas flow in the regenerator.
Among the models proposed for the regenerator, most
focus on the dense bed that is characterized by bubbles
rising through an emulsion phase. The earliest models were
Table 1. A brief summary of literature on FCCU modelling.
Author
Year
Model type
Kinetic model
Weekman and Nace
Jacob et al.
Ford et al.
Errazu et al.
Lee and Groves
Kunii and Levenspiel
McFarlane et al.
Arbel et al.
Kumar et al.
Sriramulu et al.
Arandes et al.
1970
1976
1976
1979
1985
1990
1993
1995
1995
1996
2000
C
C
R
R
I
R
I
I
I
R
I
3-Lump model
10-Lump model
NA
NA
3-Lump model
NA
None
10-Lump model
10-Lump model
NA
10-Lump model
C: Cracking model; R: Regenerator model; I: Integrated model.
single phase, simple contacting models with plug flow,
mixed flow, dispersion and tanks in series. The early
researchers, Arthur (1951), Rowe and Partridge (1965) and
Weisz and Goodwin (1966) simply divided the bed into
regions known as dense phase and dilute phase. Later on,
more complete models were developed: the grid-effect
model by Behie and Kehoe (1973) and Errazu et al.
(1979); the two-region model by de Lasa and Grace (1979)
and de Lasa et al. (1981); and the bubbling-bed model by
Kunii and Levenspiel (1968, 1990).
A summary of the assumptions made by various researchers for modelling the physics and chemistry of the processes
in FCCU is illustrated in Figure 3. In the next section, the
FCCU case study is presented. The case study describes in
detail the process modelling methodology adopted for a
system like FCCU and its implementation in MPROSIM.
Model tuning and comparison of model predictions to the
plant data are also presented. The case study is concluded
with a discussion on some merits of developing FCCU
model in an object-oriented framework like MPROSIM.
CASE STUDY: MODELLING OF FCCU
Process Description
The schematic of a current generation FCCU with standpipes and slide valves is represented in Figure 4. The reactor
consists of a vertical section called the riser. The preheated
Figure 3. Summary of various assumptions for modelling FCCU.
533
Figure 4. Schematic of a FCCU reactor=regenerator system.
feed is brought in contact with hot regenerated catalyst at the
bottom of the riser. Feed flashes at the bottom of the riser and
is vapourized. As the catalyst–vapour mixture rises in near
plug flow, cracking reactions take place within the riser. Coke
is deposited on the catalyst surface during cracking reactions.
Heat required for endothermic cracking reactions is supplied
by the hot catalyst. The bulk of the catalyst is separated from
product vapour in the riser termination device (RTD) at the
top of the riser and falls into the stripper section. Catalyst
fines that are entrained along with the product vapour are
separated in the reactor cyclones and returned to the stripper.
The product vapour is fed to the main fractionator where
it is separated into various components such as light gases
(C1–C4), liquified petroleum gas (LPG), gasoline, cycle oil in
the diesel boiling range and heavy bottoms.
In the stripper section, steam is used to strip off hydrocarbons trapped within the bulk of the catalyst and catalyst
pores. Stripped catalyst is regenerated in the regenerator
section by burning coke in a fluidized bed using hot air.
Coke combustion reactions occurring in the dense bed
produce CO, CO2 and H2O. Combustion reactions occur
further in the dilute phase due to entrainment of catalyst
particles from the dense bed by the flow of air. The entrained
catalyst is separated from stack gases in regenerator cyclones
and returned to the dense bed. The heat due to combustion
reactions raises the temperature of the regenerated catalyst
which is recycled to the reactor through the stand pipe. As
shown in Figure 4, spent and regenerated catalyst circulation
is controlled by slide valves in most of the modern units.
Kinetic Lumping Scheme
The reactions that occur in the riser when hot regenerated
catalyst comes into contact with the feed are described by
considering a lumping scheme. The lumps on the feed side
are characterized based on the boiling point range of gas oil
feed and its chemical composition, mainly in terms of
paraffins, naphthenes and aromatics. The lumps on the
product side are characterized by the boiling point range
of main fractionator side draws. The various lumps considered for building the reaction kinetics are listed in Table 2.
The lumps on the feed and product side are chosen based on
the theoretical basis provided by Jacob et al. (1976) and the
experimental support from CPCL for feed and product
characterization.
The network of reactions using the 11 lumps given in
Table 2 are shown in Figure 5 along with their boiling point
ranges. The salient points of the reaction network are:
534
Table 2. Kinetic lumping scheme: feed and product lumps.
Feed lumps
Ph
Nh
Ah
Sh
Pl
Nl
Al
Sl
BP range ( C)
Product
BP range ( C)
370–FBP
370–FBP
370–FBP
370–FBP
IBP–370
IBP–370
IBP–370
IBP–370
Gasoline
Light gases
Coke
C5–IBP
C1–C4
IBP: Initial boiling point of feed; FBP: Final boiling point of feed;
P: Paraffins; N: Naphthenes; A: Aromatic bare rings; S: Substituent groups
on aromatics. Subscripts: l – Light fraction, h – Heavy fraction.
The feed is mainly characterized as consisting of heavy
and light fractions based on the boiling point range,
370 C–FBP and IBP–370 C respectively. Each of the
heavy and light fractions are further divided into four
lumps—paraffins, naphthenes, aromatics and side chains.
The product consists of three lumps—gasoline (C5–IBP),
light gases (C1–C4) and coke.
It is assumed that there are no cross reactions between
heavy and light fractions. For example, heavy naphthenes
(Nh) crack to give only light naphthenes (Nl) and not light
paraffins (Pl). One exception to this assumption is the
cracking of heavy aromatic substituent groups (Sh) going
to light aromatic bare rings (Al).
Bare aromatic rings are mainly responsible for coke
formation. Hence (Ah) and (Al) are shown to form
only coke.
Figure 5 shows the reactions mainly from Ph and Pl and
other typical reactions. Apart from the reactions shown in
the figure, naphthenes (Nh and Nl) and substituent groups
on aromatics (Sh and Sl) will have reactions similar to that
indicated for paraffins (Ph and Pl)
Based on the above set of assumptions, the total number
of cracking reactions in the network is 27.
As described in the review on FCCU modelling literature,
researchers have considered modelling the reaction kinetics
in the riser based on various assumptions and parameters
i.e., number of lumps the feed is characterized into, rate
laws governing the reaction kinetics, kinetic parameters, and
various empirical catalyst deactivation functions. Some of
the recent literature proposes the use of structure-oriented
lumps for modelling the reaction kinetics of complex feedstocks (Quann and Jaffe, 1992, 1996). In the event of so
many factors influencing reaction kinetics in the riser, it has
been modelled as a class=object with all the influencing
factors featuring as attributes of reaction class. Thus, the
model is not restricted to a single lumping strategy. The user
can choose any number of lumps, build a reaction network
between various lumps, input the various kinetic parameters,
define any type of rate laws governing the formation=
depletion of various lumps and also include different types
of catalyst deactivation functions. Modelling the reaction
kinetics as a class=object does not restrict the model to
a particular lumping strategy. Moreover, it imparts the
important characteristic—ease of extension and adaptability
to model other complicated reaction mechanisms, which is
one of the desired essential features of a structured process
modelling.
Object-Oriented Modelling of FCCU
The object-oriented model for FCCU is developed
following the object-oriented and process modelling methodology described in section on Modelling in MPROSIM,
above. Catalytic reactions in the riser, regenerator fluidized bed hydrodynamics and coke combustion are
considered in the model. Detailed momentum calculations
at various sections of the unit have also been taken into
Figure 5. 11-lump reaction kinetic scheme.
535
consideration. Based on the physicochemical phenomena
occurring in various sections of the reactor=regenerator
and geometry of the unit, a structural decomposition was
performed. The unit is divided into smaller fragments of
systems and subsystems. The decomposition is progressively done until each model fragment represents a
fundamental chemical unit operation. The hierarchy of
model fragments as a result of decomposition performed
on the commercial unit (see Figure 4) is illustrated in
Figure 6. Different sections of FCCU which contribute to
the pressure drop in the system are also considered in the
decomposition and model hierarchy. A model fragment, or
an aggregation of some model fragments, represents a
part or the whole of FCCU. For example, the riser bottom
is emulated by an adiabatic mixer model wherein catalyst
and liquid feed combine and produce a single stream
consisting of catalyst and feed vapour. Following this
methodology, various sections of reactor and regenerator
are decomposed to fundamental chemical engineering unit
operation models as described below:
Reactor
The reactor can be divided into various sections which
can be modelled using the following units:
Mixer: The mixer model emulates mixing of hot regenerated catalyst and liquid feed at the bottom of the riser. It
is modelled considering adiabatic conditions. In addition,
the mixer model represents other mixing processes within
the reactor and regenerator, for example, mixing of steam
from stripper bed and product vapour separated from the
riser termination device.
Riser: The riser can be modelled either as a single continuous stirred tank reactor (CSTR), representing complete back mixing with uniform temperature (McFarlane
et al., 1993), or as a plug flow reactor with temperature
gradient along the riser height (Arbel et al., 1995). The
above two assumptions represent two extremes of modelling the riser. The actual conditions within the riser can be
Figure 6. Structural decomposition of FCC reactor=regenerator.
assumed to represent a non-ideal mixing zone. This is due
to the presence of slip between vapour and catalyst
particles, axial dispersion of catalyst due to turbulence
and temperature difference between the riser inlet and
outlet in actual conditions. Hence, the riser is modelled as
a composite object characterized by series of CSTRs. Slip
ratio, catalyst circulation, space velocity and void fraction
are important factors which determine hydrodynamics of
the riser. The chemistry of reactions is accounted for by a
lumped kinetic scheme as described earlier.
Riser Termination Device (RTD): The RTD is modelled
as a hydrodynamic unit which contributes to pressure
drop due to splitting of the vapour-catalyst stream into
two (a ‘TEE’ junction) and a sudden expansion of the
vapour through an opening into the reactor vessel.
Splitter: The separation of catalyst from vapour at the
RTD is modelled using a splitter, which divides the inlet
stream into two and has a splitting efficiency for each
component. It is modelled with the assumptions of no
pressure loss and the three streams, one inlet and two
outlet, are in thermal equilibrium.
Cyclone: Separation of the catalyst fines from the reactor
vapours is modelled using a cyclone. Separation efficiency for each component and pressure drop are taken
into consideration.
Stripper: The stripper is modelled to represent stripping of
hydrocarbons. It is modelled assuming a counter current
operation with mass transfer. The amount of unstripped
hydrocarbons in the stripper is a function of the steam to
catalyst ratio and is given by an empirical function (Arbel
et al., 1995).
A schematic of the reactor configuration in an objectoriented framework using the units described above is illustrated in Figure 7. The riser is represented as a composite
object in the object-oriented framework. The number of
CSTRs for modelling the riser is a model input parameter.
The user can specify any number of CSTRs instead of
physically assembling that number of units on the flowsheet.
As a result, the riser can be configured to represent different
FCC riser models reported in literature.
The riser termination device at the top of the riser (see
Figure 4) is modelled as a combination of the following units:
the RTD, with stream splitting into two halves along with
pressure drop; and a splitter, which accounts for separation of
the catalyst from the product vapour. This methodology of
modelling the pressure drop as a separate model fragment was
adopted in order to retain the generic nature of a component=
stream splitting unit. This results in a well defined and abstract
splitter unit that can be reused. The stripper is modelled in
two sections. This is to account for recycling of the entrained
catalyst through the dip leg of the reactor cyclones into the
stripper bed. The stripper bed, Stripper0, represents the stripper
from the top of the bed to the dip leg exit inside the bed. The
lower portion of the stripper bed, from the exit of the dip leg to
the spent catalyst exit, is represented by Stripper1.
Regenerator
The regenerator is mainly divided into two regions: dense
bed and dilute region. It also consists of hydrodynamic units
such as air distributor and cyclones for catalyst separation
from flue gases. Cyclones are modelled similarly to those
described in the reactor section.
536
moving in near plug flow. In actual conditions, there exists
some amount of back mixing due to heavier particles
falling off before they reach the cyclones. Hence,
this phase is also modelled as a series of CSTRs. It is
assumed that all the entrained catalyst particles return to
the dense bed through the cyclone dip legs.
Figure 7. Configuration of the FCC reactor system in object-oriented
framework.
Air distributor: The air distributor is modelled as a
pressure drop unit for inlet air. The pressure drop in the
unit depends on the type of distributor. For a pipe grid type,
the pressure drop depends on the number of open holes in
the grid, the diameter of each hole and air inlet pressure.
Dense region: This region is characterized by a bed of
solid catalyst particles and the air which flows through the
bed for catalyst regeneration. The dense bed is further
divided into two sections
— Emulsion phase: This phase mainly consists of solid
catalyst particles which are assumed to be completely
mixed with the flow rate of air corresponding to
minimum fluidization velocity. Hence, this phase is
modelled as a single CSTR with uniform temperature
and gas composition.
— Bubble phase: The portion of air with flow rate
exceeding minimum fluidization velocity is assumed
to flow in the form of bubbles. These bubbles are considered to be rising through the dense bed in near plug
flow fashion. The effect of expansion of bubbles due
to pressure gradient along the height of the dense bed,
coalescence of bubbles and increase in volumetric flow
rate of bubbles due to pressure gradient are not considered. The near plug flow of the bubble phase is
modelled as a series of CSTRs. The bubble phase is
assumed to be completely free of the catalyst particles.
Dilute region: This region consists of combustion gases
from the top of the dense bed and the entrained catalyst
particles. Due to superficial velocity of the air and bursting
action of the bubbles at the top of the dense bed, catalyst
particles get entrained along with combustion gases and
form a dilute phase above the dense bed. In this region,
both gas and solid catalyst particles are assumed to be
Various reactions occurring as a result of coke combustion
in different regions of the fluidized bed are considered in
both the dense bed and dilute phase. Overall conversion of
coke depends on flow rate of air, its pressure and temperature,
amount of catalyst in the bed, catalyst circulation rate, and
bed temperature. The combustion reactions are also strongly influenced by characteristics of the fluidized bed. A
detailed hydrodynamic calculation of the fluidization process
is considered to determine various parameters such as bed
density, void fraction, etc. The reader is referred to the
appendix for a detailed account of all the hydrodynamic
calculations and combustion reactions that were taken into
consideration.
The schematic of different regions considered for modelling the regenerator and its flowsheet representation in
an object-oriented framework is shown in Figure 8. The
regenerator is a complex unit characterized by hydrodynamics of the fluidized bed and coke combustion reactions.
In addition, inlet air flow rate and its conditions strongly
influence the physical and chemical phenomena, thus giving
rise to a coupled transfer and exchange of mass, momentum
and energy between various regenerator model fragments.
Hence, dense bed and dilute phase are modelled as a single
composite object similar to the riser. The number of CSTRs
associated with both the bubble phase and regenerator dilute
phase are part of input parameters to the composite object.
Figure 8. Configuration of a FCC regenerator system in object-oriented
framework.
537
The complete flowsheet schematic of the integrated
FCC reactor=regenerator is shown in Figure 9. The representation of FCCU as a flowsheet in MPROSIM is illustrated in Figure 10. The figure also shows the input
template for the riser. In the next section, we present the
results of simulation studies conducted using the FCCU
model.
commercial unit. If the results are not satisfactory, the user
can change the modelling assumptions by changing the
parameters. Thus, the model can be fine tuned to represent
the configuration of the commercial unit.
The model was tuned with industrial data provided by
CPCL to represent their FCCU configuration. Tuning the
model involved carrying out simulation studies by adjusting
various unit parameters and comparing model predictions
with that of the plant data.
Model Tuning
The FCCU model has been developed as a process
flowsheet in MPROSIM. Various units are provided in the
library, with which a user can select the units and connect
them to represent various configurations of FCCU. The
FCCU model has several input parameters. In addition to
those associated with input streams i.e., flow rates, composition, temperature, pressure, etc., there are various unit
parameters. Many of these unit parameters are essentially
the assumptions that are part of a FCCU model. The user
can specify these parameters and thereby make a set of
assumptions for the model. Simulation studies can be
carried out by providing necessary inputs in order to
check whether the model assumptions that the user has
made are valid for the respective configuration of the
Reactor tuning
Object-oriented flowsheet representation of the reactor in
MPROSIM is shown in Figure 11. Inputs to the reactor
flowsheet consisted of feed flow rates, composition of feed
in terms of various lumps, regenerated catalyst and steam for
the stripper section. Inputs to the reactor flowsheet are
shown in Table 3. One of the important tuning parameters
for the reactor flowsheet is number of CSTRs for the riser.
Studies were conducted using the reactor model by varying
number of CSTRs. The model predictions were compared
with process conditions and product yield of the commercial
unit. All other parameters and inputs such as catalyst
circulation rate, feed conditions, etc., are kept constant.
Results of the simulation studies are given in Figures 12–16.
Figure 9. FCC reactor=regenerator model in an object-oriented framework.
538
Figure 10. FCC reactor=regenerator model in MPROSIM.
Table 3. Reactor inputs.
Parameter (units)
Feed conditions
Total feed flow rate (Kg h1)
Composition (Wt. fraction)
Ph
Nh
Ah
Sh
Pl
Nl
Al
Sl
Feed temperature ( C)
Feed pressure (Kg cm2)
Catalyst
Regenerated catalyst flow rate (Kg h1)
Coke on regenerated catalyst (Wt. percent)
Temperature ( C)
Steam
Flow rate (T h1)
Temperature ( C)
Pressure (Kg cm2)
Other inputs
Kinetic parameters
Catalyst deactivation
Value=source
100,482.40
0.2866
0.1061
0.0836
0.1905
0.1151
0.0426
0.0535
0.1220
351.0
2.74
594,000.0
0.25
660.0
1.50
220.0
2.80
Arandes et al., 2000
Arbel et al., 1995
The trends of variables shown in the above figures prove
that the assumption of approximating the reactor riser by a
finite number of CSTRs-in-series is appropriate. A small
number of CSTRs would mean the riser is close to ideal
mixing conditions. At the same time, a large number of
CSTRs would indicate a proximity to near plug flow
condition. It is evident from the above figures that beyond
a certain number of CSTRs (say, 10), change in the model
prediction is not significant. Once the model is tuned with
respect to the number of CSTRs for the riser, an appropriate
number can be selected for carrying out further simulation
studies on the reactor and integrated reactor=regenerator
flowsheet. We have chosen 10 CSTRs to represent CPCL’s
reactor riser for conducting further simulation and optimization studies.
Regenerator tuning
The regenerator modelled as a flowsheet in MPROSIM is
illustrated in Figure 17. The figure also shows the input template for the regenerator composite object. Typical parameters
used for tuning the regenerator model are: (1) number of
CSTRs for bubble phase; (2) number of CSTRs for dilute
phase; and (3) entrained catalyst as weight percent of
539
Figure 11. Reactor flowsheet in MPROSIM.
catalyst in the regenerator. Inputs to the regenerator flowsheet are shown in Table 4. Tuning of the regenerator
flowsheet is done by varying the number of CSTRs of
both bubble phase and dilute region for a fixed value of
catalyst entrainment. After each simulation, model predictions are checked with process data from the commercial
unit. The combinations of number of CSTRs and catalyst
entrainment that match the plant data are chosen as representative values. Regenerator model tuning results are given
in Table 5 and parameters chosen for CPCL’s regenerator
configuration are highlighted.
Parametric=Optimization Studies
Figure 12. Wt. fraction of feed lumps (heavy) vs. no. of CSTRs.
The parameters chosen from tuning of the reactor
and regenerator are used in the integrated FCC reactor=
regenerator flowsheet to conduct further simulation studies.
A comparison of results of model prediction and corresponding process data obtained from the commercial unit
are shown in Table 6. Some of the data corresponding to
model predictions could not be derived from the commercial
unit due to reasons of instrumentation and lack of sampling
points.
A process model can also be used for various optimization studies. In the case of the FCCU model, yield optimization studies were carried out with the model. The effect of
feed composition and temperature on gasoline yield is
shown in Figure 18. Three different feed compositions
considered for this study are listed in Table 7. The behaviour
shown in the plots is typically expected of gasoline. Referring to Figure 5, it can be seen that gasoline not only forms
from heavy and light fractions of the feed, but also under-
540
Figure 13. Wt. fraction of feed lumps (light) vs. no. of CSTRs.
Figure 15. Total riser yield (Wt. fraction) vs. no. of CSTRs.
goes a cracking reaction to produce light components and
coke. Due to the secondary cracking nature of gasoline, its
yield exhibits a maximum as a function of feed temperature.
It is evident from Figure 18 that not only does the gasoline
yield change with change in feed composition, but the
optimum temperature also changes. This result is in good
agreement with procedures typically followed in the operation of the commercial unit. Stable operation of FCCU is
challenged by frequent changes in feed composition due to
change of crude oil input to the refinery. To maximize the
yield from FCCU, operating parameters such as catalyst
circulation rate and feed inlet temperature are regularly
monitored and manipulated whenever feed composition
changes.
The next section briefly summarizes the authors’ experiences in the implementation of object-oriented framework
for modelling a large scale industrial process such as FCCU.
The section also highlights and clarifies some of the lessons
learnt during model development stage and implementation
period, and in hindsight, vis-a-vis the practical issues listed
in the section on Review of Process Modelling, above.
Discussion and Clarifications
To begin with, the FCCU model development project
was undertaken in an academic institution with the support
of CPCL. The scope of the project was to develop a steadystate model for CPCL’s FCC reactor=regenerator configuration. The model was proposed to be developed in an
object-oriented framework to support the refinery’s research
and development team in carrying out off-line simulation
studies for trouble-shooting due to frequent changes in feed
composition, catalyst evaluation and independent studies of
internal components of the FCCU such as cyclone, stripper,
etc. From a detailed review of the literature on FCCU
modelling, the model development was focussed towards
developing a general framework for modelling various
FCCU configurations proposed in the literature. Though
the outcome of this development effort was more oriented
towards FCCU modelling and addressing some of the
refinery-based modelling issues, there are some general
principles that can be culled from this effort towards
understanding the object-oriented modelling of chemical
processes. Based on the modelling effort, the following
Figure 14. Wt. fraction of product lumps vs. no. of CSTRs.
Figure 16. Riser outlet temperature ( C) vs. no. of CSTRs.
541
Figure 17. Regenerator flowsheet in MPROSIM.
comments could be made regarding generic and implementation issues laid down in section on Review of Process
Modelling, above.
the development team consisted of chemical engineers
from academia and industry, specifically from a refinery.
So, the knowledge domain of the group of developers is
restricted to a few areas of chemical engineering. Typically, this would be the scenario in any model development activity undertaken in collaboration with an
industrial partner. As a result, the model development
activity is influenced by the knowledge of the modelling
experts. In addition, several practical issues have to be
taken into consideration as highlighted in the following
Table 4. Regenerator inputs.
Parameter (units)
Air
Air flow rate (Kn m3 h1)
Temperature ( C)
Pressure (Kg cm2)
Catalyst
Spent catalyst (Kg h1)
Temperature ( C)
Coke on spent catalyst (Wt percent)
Other inputs
Fluidization parameters
Value=source
43.0
237.0
2.60
594,000.0
500.5
0.943
Arbel et al., 1995
discussion. Even if a prototype of a generic framework is
developed independent of any industrial input, it would
need to be critically evaluated. Implementation of various
representative large scale processes from diverse engineering fields such as petroleum, petrochemical, biological,
solids-based, etc., have to be considered for the evaluation.
The amount of time and expertise needed would always be
an important limiting factor.
the development team would have to judiciously choose
a decomposition strategy to model the process in an
object-oriented framework. Some of the choices available
for developing an object-oriented model for FCCU are
enumerated in Table 8. The choice of the authors for the
modelling of FCCU in this contribution is also
highlighted. The selection of a particular option is not
only influenced by the modelling experts from academia,
but also by the industrial partner. In addition, the duration
and time-frame of the project have to be taken into
consideration.
the granularity to which the given process is decomposed
and design of the framework directly depend on the
choice made from the options listed in Table 8. Typically,
at each level of decomposition, some assumptions may be
introduced into some or all the model fragments. Essentially, a higher degree of fragmentation would imply a
larger number of assumptions in the model. These
assumptions are verified as a result of experimentation
542
Table 5. Regenerator tuning results.
No CSTRs
BP
Te ( C)
Tbubble ( C)
DP
Model
Plant
Model
Plant
I. Wt. fraction of regenerated catalyst entrained ¼ 0.05
1
1
749.16
—
664.36
2
1
601.87
—
671.52
3
1
624.29
—
677.44
4
1
678.46
—
674.40
5
1
707.22
—
678.43
653.00
653.00
653.00
653.00
653.00
835.34
1003.59
1024.64
1046.46
1022.34
780.00
780.00
780.00
780.00
780.00
170.98
332.07
347.20
372.06
343.91
127.00
127.00
127.00
127.00
127.00
1
2
3
4
5
1
2
3
4
5
669.11
674.84
680.94
682.62
687.73
675.26
674.84
680.96
682.74
696.08
653.00
653.00
653.00
653.00
653.00
653.00
653.00
653.00
653.00
653.00
857.31
1061.90
1085.55
1091.52
1042.38
861.57
1061.95
1086.01
1070.29
1034.96
780.00
780.00
780.00
780.00
780.00
780.00
780.00
780.00
780.00
780.00
188.20
387.06
404.61
408.90
354.65
186.32
387.10
405.05
387.54
338.88
127.00
127.00
127.00
127.00
127.00
127.00
127.00
127.00
127.00
127.00
I. Wt. fraction of regenerated catalyst entrained ¼ 0.10
1
1
748.70
—
664.68
2
1
612.51
—
684.04
3
1
635.60
—
688.86
4
1
648.00
—
690.25
5
1
709.32
—
678.35
653.00
653.00
653.00
653.00
653.00
809.35
940.29
938.64
927.98
941.58
780.00
780.00
780.00
780.00
780.00
144.68
256.25
249.78
237.73
263.23
127.00
127.00
127.00
127.00
127.00
1
2
3
4
5
2
2
2
2
2
748.72
617.84
640.10
652.42
661.39
—
—
—
—
—
671.48
690.11
693.10
693.89
683.75
653.00
653.00
653.00
653.00
653.00
846.13
996.85
977.38
961.03
971.80
780.00
780.00
780.00
780.00
780.00
174.65
306.74
284.28
267.14
288.05
127.00
127.00
127.00
127.00
127.00
1
2
3
4
5
3
3
3
3
3
748.47
617.97
639.78
651.99
691.47
—
—
—
—
—
676.61
690.26
692.83
693.57
695.75
653.00
653.00
653.00
653.00
653.00
850.39
998.39
976.03
959.56
942.62
780.00
780.00
780.00
780.00
780.00
173.78
308.13
283.21
265.99
246.87
127.00
127.00
127.00
127.00
127.00
II. Wt. fraction of regenerated catalyst entrained ¼ 0.14*
1*
1*
749.30
—
663.61
2
1
617.54
—
689.76
3
1
641.07
—
693.95
4
1
653.85
—
694.96
5
1
693.01
—
680.62
653.00
653.00
653.00
653.00
653.00
783.28
895.15
891.18
881.86
861.84
780.00
780.00
780.00
780.00
780.00
119.67
205.39
197.23
186.90
181.22
127.00
127.00
127.00
127.00
127.00
1
2
3
4
5
2
2
2
2
2
748.84
624.37
646.36
659.46
658.47
—
—
—
—
—
668.67
697.28
698.61
699.05
684.70
653.00
653.00
653.00
653.00
653.00
831.72
945.79
922.24
908.73
911.87
780.00
780.00
780.00
780.00
780.00
163.05
248.51
223.63
209.68
227.17
127.00
127.00
127.00
127.00
127.00
1
2
3
4
5
3
3
3
3
3
748.77
624.22
645.82
658.74
682.58
—
—
—
—
—
676.53
697.13
698.19
698.61
692.82
653.00
653.00
653.00
653.00
653.00
840.69
945.30
920.47
906.97
893.03
780.00
780.00
780.00
780.00
780.00
164.17
248.16
222.28
208.36
200.22
127.00
127.00
127.00
127.00
127.00
II. Wt. fraction of regenerated catalyst entrained ¼ 0.15
1
1
749.14
—
663.48
2
1
618.48
—
690.81
3
1
642.13
—
694.91
4
1
678.49
—
676.10
5
1
718.80
—
682.35
653.00
653.00
653.00
653.00
653.00
775.96
885.48
881.59
885.12
851.43
780.00
780.00
780.00
780.00
780.00
112.48
194.67
186.68
209.02
169.09
127.00
127.00
127.00
127.00
127.00
1
2
3
4
5
1
2
3
4
5
653.00
653.00
653.00
653.00
653.00
653.00
653.00
653.00
653.00
653.00
829.72
934.03
911.36
895.08
885.26
839.06
933.29
909.52
896.64
884.74
780.00
780.00
780.00
780.00
780.00
780.00
780.00
780.00
780.00
780.00
159.94
235.52
211.68
210.25
195.97
161.22
234.97
210.29
197.06
188.41
127.00
127.00
127.00
127.00
127.00
127.00
127.00
127.00
127.00
127.00
2
2
2
2
2
3
3
3
3
3
748.68
604.65
627.65
677.56
692.73
748.41
604.66
627.67
639.47
692.55
749.00
625.52
647.65
686.75
666.54
748.13
625.32
647.05
660.16
711.85
Plant
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
Model
Tdelta ( C)
Plant
2
2
2
2
2
3
3
3
3
3
Model
Tdilute ( C)
669.78
698.51
699.68
684.83
689.30
677.84
698.31
699.23
699.58
696.33
BP: Bubble phase; DP: dilute phase. *Parameters chosen for CPCL’s regenerator configuration.
543
Table 6. Comparison of simulated and plant data.
Process variable
Reactor
Riser bottom temperature ( C)
Riser outlet temperature ( C)
Reactor yield
Catalyst
Catalyst circulation rate (T min1)
Spent catalyst temperature ( C)
Regenerated catalyst temperature ( C)
Coke on spent catalyst (Wt.%)
Coke on regenerated catalyst (Wt.%)
Regenerator
Bubble phase outlet temperature ( C)
Emulsion phase outlet temperature ( C)
Dilute phase outlet temperature ( C)
Table 7. Feed composition inputs for gasoline optimization.
Model prediction
Plant data
550.18
501.78
69.3
—
498.50
68.0
9.71
500.4
663.61
0.943
0.45
9.9
469.50
653.0
—
0.25
749.3
663.61
783.28
—
653.0
780.0
and=or by matching the model predictions with the
industrial data. However, the quantity and quality of
data that can be derived from a running industrial plant
are certainly limited. Some of the reasons for this are:
instrumentation for the process is fixed and no additional
process data can be extracted; test runs that an industrial
partner is willing to conduct on the commercial unit;
inaccuracies of the process data; and the availability of
sophisticated instrumentation for carrying out lab-scale
experimental studies. Hence, hierarchy of the granular
structure designed for a process model is limited to a
large extent by the above practical issues. To cite an
example, the stripper section was modelled as an ideal
counter-current mass and energy exchange operation;
though, in the actual design of the unit, the operation is
not so. Since no specific tests could be conducted on the
stripper of the commercial unit and due to unavailability
of any lab-scale experimental setup, the assumptions
could not be validated. Hence, an empirical function for
mass exchange, i.e., stripping of hydrocarbons, was taken
from the literature (Arbel et al., 1995) without any change
of function parameters.
uniqueness of decomposition: decomposition of the
FCCU model proposed earlier is not unique. The decomposition strategy adopted resulted in more than one
degree of freedom in terms of the number of CSTRs for
reactor riser, regenerator bubble and dilute phases. If a
PFR model had been chosen for the same regions, the
degrees of freedom would have been fixed and also
unique to some extent. But the choice of CSTR model
Wt. fraction
Component
Feed 1
Feed 2
Feed 3
Ph
Nh
Ah
Sh
Pl
Nl
Al
Sl
0.2866
0.1061
0.0836
0.1905
0.1151
0.0426
0.0535
0.1220
0.1791
0.0707
0.0557
0.1190
0.1957
0.0823
0.0900
0.2073
0.1592
0.0707
0.0557
0.1058
0.2072
0.0852
0.0963
0.2195
fragment allows the model to be flexible and it can be
configured to represent many commercial FCC reactor=
regenerator units.
as described earlier, many FCCU models have been
proposed in the literature. One of the important aspects of
this model development has been to conceptualize a framework in which many FCCU models can be configured. The
hierarchy of structural decomposition illustrated in Figure 6
is general enough to represent many FCCU models
proposed in the literature. An aggregation of the model
fragments along with appropriate model assumptions would
result in a particular configuration. The following two
examples are highlighted to illustrate the idea:
— if the riser is assumed to be a single CSTR,
regenerator bubble and dilute phases are modelled
by a large number of CSTRs, and empirical kinetics
are assumed for overall reactor conversion, then the
model would represent that proposed by McFarlane
et al. (1993) in its steady-state form.
— on the other hand, if the reactor riser, regenerator
bubble and dilute phases are approximated by large
number of CSTRs, and a 10-lump kinetic model is
used to describe the catalytic reactions, then the
model would be that proposed by Arbel et al.
(1995).
similarly, by considering different combination of CSTRs
for the bubble and dilute phases, various models proposed
for the regenerator can be realized. The object-oriented
modelling of FCCU as a flowsheet consisting of various
parts of the process as model fragments has resulted in
the conceptualization of a general framework which can be
easily extended and adapted to model various configurations of FCC reactor=regenerator. However, the issues
concerning the reuse, extension and adaptability of each
of the model fragments need to be critically evaluated. The
success and failure issues concerning the modelling of riser
and regenerator composite objects are evaluated below:
Table 8. Degrees of freedom for modelling FCCU.
Option
1
2
3
4*
5
Figure 18. Optimization of gasoline yield.
Description
FCC reactor=regenerator
Reactor; regenerator
Reactor kinetics; reactor
hydrodynamics; regenerator kinetics;
regenerator hydrodynamics
Decomposition as given in Figure 6
Decomposition based on Figure 2
*Option chosen in this contribution.
Evaluation
A monolithic block
Two model blocks
Four model blocks
General framework*
Fundamental approach
544
— the riser is modelled as a composite object. It can
function as a single CSTR or as a non-ideally mixed
tubular reactor by a series of CSTRs. It can also be
approximated as a PFR by specifying a large number
of CSTRs, ideally infinite. The riser is one of the
model fragments which truly reflects the behaviour of
a composite object—one that can be completely
reused, and can be easily extended and adapted.
— the regenerator composite object is the model fragment for a fluidized bed. As highlighted earlier, the
regenerator fluidized bed is a complex and coupled
process. In its present state, the regenerator composite
object can be used to represent a single stage regenerator. It is also possible that a two-stage regenerator
can be modelled using a stack of two composite
objects. If the regenerator composite object is further
decomposed resulting in more fundamental model
fragments, then it is possible to model many industrial-scale fluidized bed processes. The issues
concerning this problem are currently being studied.
As indicated in the issues listed above, the development
and implementation of a formal structure, guidelines and
principles of object-oriented modelling on a large-scale
industrial process are strongly influenced not only by the
expertise of the development team as a whole, but also by
various factors which comprise the needs and requirements
of the industry on a commercial scale. Many of the extension and adaptability issues related to MPROSIM as a
simulation and modelling tool are currently being evaluated.
For this purpose, we are undertaking a detailed study of
object-oriented modelling in other processes such as fuel
cells, biochemical and biological processes, etc.
UTILITY AND MERITS
The representation of the FCCU model as a flowsheet has
several advantages. It is possible to carry out simulation
studies with an integrated reactor=regenerator flowsheet,
with individual reactor and regenerator flowsheets and
study the performance of individual units such as reactor
or regenerator cyclones, stripper, air distributor, etc. Several
studies on sensitivity of reactor yield and other important
process parameters to input conditions can be carried out
with the model. Typically, carrying out sensitivity, parametric and optimization studies involves a lot of effort. But
with the above model developed in an object-oriented
framework, studies like these can be conducted easily.
The representation of modelling assumptions in terms of
unit parameters is essentially an extension of the objectoriented concept of encapsulation as described in the section
on Modelling in MPROSIM, above. Some of the assumptions that are encapsulated as unit parameters in the FCCU
model are:
the number of CSTRs for modelling the reactor riser,
regenerator bubble and dilute phases
catalyst deactivation function for catalytic cracking in the
riser, and
splitting efficiencies (either direct numerical values or in
terms of empirical functional forms) for all components in
units such as splitter, cyclone and stripper.
One of the important problems that units in a refinery,
typically FCCU, have to face is that the feed composition
changes regularly. With evolution of technologies and needs
of particular markets in terms of product requirements and
specifications, it becomes necessary for the unit configuration also to change with time. Hence, it is an inevitable
exercise for any process model to be regularly fine-tuned to
represent the changing conditions in the plant. In such a
scenario, a process model developed in an object-oriented
framework can provide better support to the process engineer. In the case of FCCU, various reactor and regenerator
configurations are in operation in the refineries. Some of the
variations in structural configurations of the reactor and
regenerator are:
various types of disengagement section for separation of
product vapour from catalyst at the top of the riser
different designs for the stripper section
single-stage and two-stage regenerators
single-stage and two-stage cyclone separation for catalyst
fines.
In addition to the structural variations based on process
design, operating conditions of the unit also vary widely.
Some of the reasons for this could be: the spectrum of
products the unit is commissioned to produce, frequent
changes in the quality of feed, and fluctuations in the
market demand affecting the quantity of throughput. Such
variations in the configurations and operating conditions can
be modelled very easily and quickly in MPROSIM. A
change in catalyst circulation rate and=or quantity of feed
to the riser will change the hydrodynamics inside the riser.
Similarly, air flow rate and catalyst circulation rate will
affect hydrodynamics of the catalyst bed in the regenerator.
The model can be fine tuned by simply changing the number
of CSTRs to reflect the new process operating conditions.
The key factor responsible for flexibility of the FCCU
model in MPROSIM is its modularization and representation
in the form of a flowsheet. Unlike the traditional models,
which might be difficult to configure to represent various
configurations of FCCU, the object-oriented model can be
easily configured to represent various configurations.
CONCLUSIONS
An evaluation of the principles of structured process
modelling was undertaken in this work. This was done by
highlighting the critical issues and problems encountered
while implementing the guidelines for structured objectoriented process modelling. As a case study for the process
model, an industrial-scale FCCU was chosen. The FCCU
model served as an appropriate example to bring forth the
important object-oriented modelling issues such as granularity, adaptability and extensibility.
The model fragments proposed for the FCCU model are
general enough to be able to represent various models
proposed in the literature and other configurations of
FCCU. However, issues relating to the reusability and
extensibility of these model fragments to other chemical
process systems are still open and work is under progress in
this regard. As for the main aim of MPROSIM and development of a general purpose modelling tool, it is proposed to
study processes and systems from various other science and
engineering fields to address issues such as generality of
process models and their implementational details.
545
APPENDIX: FCCU MODEL EQUATIONS
The model equations for the various sections of the reactor
and regenerator are summarized below. Typically, each unit
comprises the mass, energy and momentum balances. The
reactor riser incorporates the 11-lump kinetic model and
pressure drop equations based on the pneumatic transport
of the solids. On the regenerator side, the detailed hydrodynamic equations are written characteristic of a fluidized
bed taking into consideration the coke combustion kinetics.
Reactor
Mixer
The mixer is modelled under adiabatic conditions with the
typical mass and energy balance equations. The mixer
does not contribute any pressure drop. The equations
pertaining to the pressure drop due to mixing, flow across
nozzles, and so on are modelled as separate hydrodynamic units. For the case of riser bottom, the pressure
drop in the liquid feed across the distributor is modelled
as a hydrodynamic unit.
Riser
The riser is modelled as a composite object. The composite object contains mass and energy balance equations
incorporating reaction kinetics. In addition, detailed
momentum balance equations are taken into consideration. The model equations for the riser are typified by
CSTR model equations. The effect of coking and catalyst
deactivation are taken into consideration by including
appropriate terms in the model equations (McFarlane
et al., 1993; Arbel et al., 1995).
— Mass balance: The mass balance equation for a
component j (either a reactant or product) within
any CSTR inside the riser for the 11-lump kinetic
model can be written as
Fj,out
DPr1 ¼
2
8Fcat
2
4
(3:6 105 ) p2 rpart gdriser
(3)
* static head of the catalyst-vapour mixture
!
Fcat þ Foil
hriser
DPr2 ¼
(Fcat =rpart ) þ (Foil =rv ) 104
(4)
where rv , feed vapour density at the bottom of the
riser, is calculated by the equation
Pmixer Mfeed
RTmixer
(5)
* friction between fluid media and wall
2
driser 0:079
Foil =rv
r
DPr3 ¼ 2
2
hriser Re0:25 v (p=4)driser
(6)
assuming turbulent conditions. The Reynolds number
(Re) is given by the following equation
Re ¼
driser Vvapour rv
mv
(7)
The feed oil vapour velocity (Vvapour ) inside the riser
can be calculated by the following equation
Vvapour ¼
F
¼ Fj,in rv cat f(t)
Foil
11
X
Fk,out
1
n K e(Ek =RT )
1 þ Kd yAh k¼1 k k0
Foil
Foil =rv
2
(p=4)driser
(8)
The viscosity of the vapour at the mixer outlet
conditions is given by
(1)
— Energy balance: A CSTR heat balance gives the
temperature profile in the riser based on the number
of CSTRs considered to model the riser. The energy
balance equation for any CSTR in the riser composite
object can be written as
Fj,in Cpfv Tin þ Fcat Cpcat Tin
¼ Fj,out Cpfv Tout þ Fcat Cpcat Tout
F
1
þ rv cat f(t)
1 þ Kd yAh
Foil
11
X
F
nk Kk0 e(Ek =RT ) k,out DHrk
Foil
k¼1
pressure drop within the riser (Jones et al., 1967;
Yang, 1973, 1974).
* acceleration of solid particles
(2)
— Pressure drop: The pressure drop in the riser is
modelled similarly to the pneumatic transport of
solid particles in the vertical direction. The following
terms are considered as the contributing factors to the
mv ¼
1:0319 106 (Tmixer þ 273:15)0:4896
1:0 þ (3:4737 102 )=(Tmixer þ 273:15)
(9)
* acceleration of the fluid media
1:0 rv Vvapour
DPr4 ¼
2 104
g
(10)
Riser Termination Device (RTD)
The inlet stream of the RTD is assumed to be split into
two equal halves taking into consideration the geometry
of the disengaging section at the top of the riser. Further,
all the three streams are assumed to be in thermal
equilibrium. The pressure drop in the RTD is modelled
by considering the following terms (Bird et al., 1960).
— a ‘TEE’ junction which is approximated by two
90 bends
DPrtd1
!
rmix 2 2:3
V
¼
1:0 104
2 g rtd b21
(11)
546
where b1, rmix and Vrtd are given by the following
equations
driser
b1 ¼
darm
rmix
Vrtd
din ¼
(12)
Fcat þ Foil
¼
(Fcat =rpart ) þ (Foil =rv )
(13)
(Fcat =rpart ) þ (Foil =rv )
¼
2
(p=4)driser
(14)
rmix 2
V b (1 b2 ) 104
g arm 2
(15)
where b2 is the area ratio of the opening in the side
arm of the RTD and the annular area in the reactor
portion. It is given by the following equation
b2 ¼
pwhw
2 d2 )
p=4(dreac
riser
(16)
and Varm is given by the equation
Varm
(Fcat =rpart ) þ (Foil =rv )
¼
2
(p=4)darm
and mgmix, the viscosity of the vapour-steam mixture
can be calculated by the empirical equation
1:0319 106 T 0:4896
1:0 þ (3:4737 102 )=T
(25)
— gas flow reversal
rgmix Vin2
2
— exit contraction
DPc4 ¼
(26)
2
2
DPc5 ¼ 0:5rgmix (Vexit
Vc2 þ KVexit
)
(27)
Splitter
The splitter is modelled to separate the inlet stream
into two streams with an efficiency of splitting for
each component present in the inlet. The model
equations consist of the mass balance for each of the
components
X
Fj,out j ¼ 1, 2, . . . , 11
(28)
Fj,in ¼
and
(17)
(18)
— particle acceleration
DPc2 ¼ LVin (Vpin Vpvessel )
(24)
streams
Cyclones
The cyclone separators used for the recovery of catalyst
fines are modelled for the pressure drop. The total
pressure drop in the cyclone is a sum of the following
individual pressure drops (Perry and Green, 1997)
— inlet contraction
2
þ KVin2 )
DPc1 ¼ 0:5rgmix (Vin2 Vvessel
4(inlet area)
inlet perimeter
mgmix ¼
— sudden expansion of the catalyst and product vapour
mixture into the reactor vessel
DPrtd2 ¼
and din is calculated as
(19)
For small particles, we can consider
Vpin ¼ Vin
(20)
Vpvessel ¼ Vvessel
(21)
Fj,out ¼ Zj Fj,in
j ¼ 1, 2, . . . , 11
(29)
where Zj is the efficiency of splitting for the component
j. The splitter is modelled such that the inlet and
outlet streams are in thermal and hydrodynamic
equilibrium.
Stripper
The stripper is modelled as a counter current operation
with steam stripping the hydrocarbons from the voids in
the stripper bed as well as within the catalyst pores. The
efficiency of stripping depends on the steam to catalyst
flow rate and is assumed to be a linear function of this
ratio (Arbel et al., 1995). Since the stripping is a counter
current operation with steam injected at the bottom and
catalyst flowing down, the pressure of the streams which
lie on the same side of the stripper unit are considered to
be equal and the two outlet streams are assumed to be in
thermal equilibrium.
and
Regenerator
— barrel friction
DPc3 ¼
2f rgmix Vin2 pDc Ns
din
(22)
The friction factor f is based on the Reynolds number
calculated at the inlet area. The Reynolds number is
calculated by the following equation
Re ¼
din Vin rgmix
mgmix
(23)
Air distributor
Assuming a uniform distribution of air without any
channelling effects, the following equation models the
pressure drop for the pipe grid type air distributor (CPCL,
1999)
DPad ¼
Pair Vair
3960Tair
(30)
Dense bed
As described earlier the regenerator dense bed is divided
into two phases, emulsion phase and bubble phase. The
emulsion phase consists of the catalyst particles and air
flow equivalent to the minimum fluidization velocity. The
547
bubble phase consists of the air flow rate which is in
excess over the minimum fluidization requirement. The
air entering from the distributor, stream U0, (see Figure 8)
is divided into two streams, (U0 7 Umf) corresponding to
the bubble phase flow and Umf, equivalent to the
minimum fluidization velocity (Kunii and Levenspiel,
1968, 1990).
— Splitter 1
The flow in stream Umf will correspond to minimum fluidization velocity which is given by the
equation
9:0 104 d p1:8 b(rpart rair )gc0:934
(31)
Umf ¼
0:87
r0:066
air mair
where rair can be calculated from the ideal gas
equation written at the air distributor exit conditions
and the viscosity of air can be calculated from the
empirical correlation
mair ¼ (0:016931 þ 4:9863 105 Tair 3:1481
2
3
þ 1:2682 1011 Tair
)
108 Tair
103
g
(32)
The average particle size can be calculated given the
particle size distribution (PSD) of the equilibrium
catalyst using the following equation
1
106
dp ¼ P
(x
i i =dpi )
(33)
A balance across the Splitter 1 (see Figure 8) will give
the equation for calculating the air flow rate in the
form of bubbles
p
2
Fair,in ¼ Fair,bubble þ Umf dreg
(34)
4
— Emulsion phase
The emulsion phase is assumed to be a completely
mixed zone. One of the attributes of the CSTR that
should be known is the volume occupied by the CSTR.
Hence, before the balance equations for the individual
combustion components are presented, detailed hydrodynamic calculations for the fluidized bed are listed
below.
The fraction of the total dense bed occupied by the
bubbles is given by the equation
U Umf
d¼ o
Ub
Therefore, the volume occupied by the bubble phase
is given by
Vb ¼ dZbed Areg
(
Zbed ¼ min Zcyc ,
Wreg rdil Areg Zcyc
Areg (rdense rdil )
)
rdil ¼ max{0:0,(aUo b)}
rdense ¼ (1 edense )rpart
U þc
edense ¼ o
Uo þ d
(39)
(40)
(41)
(42)
(43)
and the volume occupied by the emulsion phase is
Ve ¼ Zbed Areg Vb ¼ Zbed Areg (1 d)
(44)
The emulsion phase is considered as a single CSTR
and the bubble phase is approximated by CSTRsin-series, see Figure 8. The following combustion
reactions are considered in the regenerator (Arbel
et al., 1995; Weisz and Goodwin, 1966)
* r1 : C þ 0:5O2 ! CO
* r2 : C þ O2 ! CO2
* r3 : CO þ 0:5O2 ! CO2 (Heterogeneous)
* r3h : CO þ 0:5O2 ! CO2 (Homogeneous)
* r4 : H2 þ 0:5O2 ! H2 O
It is assumed that the complete hydrogen in the coke
is burnt in the emulsion CSTR itself and hence, no
reaction kinetics are taken into consideration for the
hydrogen balance equation (Arbel et al., 1995). All
the other reactions (r1, r2, r3c and r3h) occur in the
emulsion phase. Writing the balance equation for
each of the components in the emulsion phase
* Mass Balance:
Balance equation for catalyst
X
streams
Fcat,in ¼
X
Fcat,out
(45)
streams
Balance equation for carbon
Coke is assumed to be of the formula CHn.
Therefore the balance can be written as
Fcoke,in
12 ¼ r1 þ r2 þ Fcoke,out
12 þ n
(46)
(35)
where the bubble rise velocity (Ub) is given by
Ub ¼ Uo þ Umf þ Ubr
(36)
Ubr ¼ Uo Umf
(37)
pffiffiffiffiffiffiffiffiffiffi
þ 0:711 (gdb )
The product stream from the emulsion CSTR
does not contain any hydrogen in the coke as all
the hydrogen is converted to H2O in the emulsion
phase itself. The reaction rates for the carbon in
the first two reactions described above, r1 and r2
are given by
where db, the effective bubble diameter is calculated
from the following equation
db ¼
0:667q0:375
o
(38)
r1 ¼ Ve (1 edense )rpart K1
Fcoke,out
P
Fcat,out O2
(47)
548
where the rate constant K1 and the partial pressure of O2, PO2 are defined by the following
equations
2:6 1011 exp(12926=Te )
2512 exp(3420=Te ) þ 1
(FO2 ,out =32)
¼
Preg
Ftotal,out
K1 ¼
(48)
PO2
(49)
Ftotal,out is total sum of the molar flow rates of the
gaseous components in the emulsion phase
outlet stream and is given by
FO2 FCO FCO2
þ
þ
Ftotal,out ¼
32
28
44
FN2 FH2 O
þ
þ
(50)
28
18 out
r2 ¼ Ve (1 edense )rpart K2
Fcoke,out
P
Fcat,out O2
(51)
The mass transfer between bubble and emulsion
phase CSTRs is given by
(
)
k
e
(FCO
=44) (FCO
=44)
k
2 ,out
2 ,out
MCO2 ¼ Kbe
Fair,bubble
Fair,emulsion
1:035 108 exp(9506=Te )
2512 exp(3420=Te ) þ 1
(52)
Balance equation for CO2
The total CO2 produced in the emulsion phase is
contributed by three different reactions, r2, r3c,
r3h taking place simultaneously. Therefore a
balance on CO2 gives us
b
X
44
44
k
r2 þ ðr3c þ r3h Þ MCO
2
12
28
k¼1
N
FCO2 ,out ¼
(53)
where, the last term on the right hand side of the
above equation represents the mass transfer of
CO2 between emulsion phase and bubble phase
CSTRs. The reaction terms r3c and r3h are
combined and given by
r3 ¼ Ve K3 PO2 PCO
The balance equation for CO is similar to that
written above for CO2
N
FCO,out ¼
b
X
28
k
r1 r3 MCO
12
k¼1
(59)
The reaction terms have already been described
above. Writing the mass transfer equations
(
)
k
e
(FCO,out
=28) (FCO,out
=28)
k
MCO ¼ Kbe
Fair,bubble
Fair,emulsion
dZbed Areg
2
(60)
Balance equation for O2
The O2 supply to the emulsion phase takes place
from both the flow stream as well as the mass
transfer lines from the bubble phase. Writing a
balance would give
1 32
32
1 32
FO2 ,out ¼ FO2 ,in r1 r2 r3
2 12
12
2 28
Fcoke;in
1
n 32
4
12 þ n
Nb
X
þ
MOk 2
(61)
k¼1
FO2 ,in
Fair,emulsion (Pair
21
þ 1:033)=1:033
32
¼
100
82:057 103
(Tair þ 273:15)
(54)
(
K3 ¼ xpt (1 edense )rpart K3c
þ edense K3h
Ek3c
K3c ¼ K3c0 exp RTe
E
K3h ¼ K3h0 exp k3h
RTe
(58)
Balance equation for CO
where the rate constant K2 is defined by the
following equation
K2 ¼
dZbed Areg
2
MOk 2 ¼ Kbe
(55)
(56)
(FOk 2 ,out =32)
Fair,bubble
dZbed Areg
2
(FOe 2, out =32)
)
(62)
Fair,emulsion
(63)
Balance equation for N2
(57)
The balance equations are similar to that written
for O2 with one exception being that there will
549
be no reaction terms.
FN2 ,out ¼ FN2 ,in þ
The heat input due to the entrained catalyst falling
into the dense bed is given by
Nb
X
MNk 2
(64)
k¼1
FN2 ,in
Fair,emulsion (Pair
79
þ 1:033)=1:033
¼
28
100
82:057 103
(Tair þ 273:15)
(
MNk 2
¼ Kbe
(FNk 2 ,out =28)
Fair,bubble
(65)
)
(FNe 2 ,out =28)
Fair,emulsion
dZbed Areg
2
(66)
ent
Cpcat (Tent Tref )
Qent ¼ Fcat
The heat input from the spent catalyst entering the
regenerator dense bed is given by
Qsc ¼ Fcat Cpcat (Tsc Tref )
(75)
Heat input due to coke is calculated as
Qcoke ¼ Fcoke Cpcoke (Tsc Tref )
(76)
The heat inputs due to the mass transfer from bubble
phase CSTRs for O2 and N2 are given by
Qmt
O2 ¼
Nb
X
MOk 2 CpO2 (Tbubble,k Tref )
(77)
MNk 2 CpN2 (Tbubble,k Tref )
(78)
k¼1
Qmt
N2 ¼
Balance equation for H2O
(74)
Nb
X
k¼1
The balance equation for H2O can be written as
Fcoke,in
1
n 18
2
12 þ n
¼ FH2 O,out þ
Nb
X
MHk 2 O
k¼1
MHk 2 O ¼ Kbe
(
(FHk 2 O,out =18)
Fair,bubble
(67)
)
(FHe 2 O,out =18)
Fair,emulsion
dZbed Areg
2
(68)
* Energy Balance: The energy balance equation for
the emulsion phase can be written as
Qinput ¼ Qoutput þ Qgen
(69)
(70)
The last two terms of input heat are due to the mass
transfer of O2 and N2 from the bubble phase
CSTRs to the emulsion phase CSTR. The heat
input due to air (Qair) can be calculated by
Qair ¼ Mair Cpair (Te Tref )
((Pad þ 1:033)=1:033)Fair,emulsion 106
82:057 Te
(72)
Pad is the pressure at the exit of the air distributor
and is calculated by
Pad ¼ Pair DPad
þ QCO2 þ QH2 O þ Qmt
CO
mt
þ Qmt
CO2 þ QH2 O
(79)
Qcat ¼ Fcat Cpcat (Te Tref )
(80)
QO2 ¼ FO2 CpO2 (Te Tref )
(81)
QN2 ¼ FN2 CpN2 (Te Tref )
(82)
QCO ¼ FCO CpCO (Te Tref )
(83)
QCO2 ¼ FCO2 CpCO2 (Te Tref )
(84)
QH2 O ¼ FH2 O CpH2 O (Te Tref )
(85)
Nb
X
Qmt
CO2 ¼
Qmt
H2 O ¼
k
MCO
CpCO (Te Tref )
k¼1
Nb
X
k¼1
Nb
X
(86)
k
MCO
CpCO2 (Te Tref )
2
(87)
MHk 2 O CpH2 O (Te Tref )
(88)
k¼1
(71)
where the Mair is the flow rate of air in mol s1 and
is given by
Mair ¼
Qoutput ¼ Qcat þ QO2 þ QN2 þ QCO
Qmt
CO ¼
Qinput ¼ Qair þ Qent þ Qsc þ Qcoke
mt
þ Qmt
O2 þ QN2
Similarly, the following equations give the expressions
for the heat output from the emulsion phase
(73)
The heat generation in the emulsion phase is due to
the coke combustion reactions, which are exothermic in nature. Hence, the total heat generated can
be written as
T
T
T
e
e
e
Qgen ¼ r1 DHf ,CO
þ r2 DHf 2,CO
þ r3 DHf 3,CO
2
2
T
e
þ r4 DHf ,H
2O
(89)
The heats of formation at the temperature conditions of the emulsion phase CSTR are given by
550
(Perry and Green, 1997)
Te
DHf ,CO
¼ 763168 þ 86:604Te þ 8:456
458920
103 Te2 (90)
Te
T
e
DHf 2,CO
¼ 4160500:82 þ 158:928Te
2
þ 0:0236Te2 8130804
Te
in the emulsion bed, the corresponding mass and energy
balance equations are not taken into consideration.
Regenerator Cyclones
The regenerator cyclone equations are similar to the
reactor cyclone model equations. To model a two-stage
cyclone separator, two single-stage cyclones are connected
in series and model equations are written accordingly.
(91)
NOMENCLATURE
T
e
¼ 2961244 þ 22:836Te þ 1:0274
DHf 3,CO
2
10
2
Te2
7409644
Te
(92)
T
e
¼ 1011938:58 63:665Te
DHf ,H
2O
þ 1:387 102 Te2 542305:545
Te
(93)
The specific heats of the gaseous components
which are required for calculating the enthalpy of
the streams can be written as functions of temperature (Smith and Van Ness, 1987)
CpCO ¼ 0:2357 þ 4:286 105 T
(94)
5
CpCO2 ¼ 0:235 þ 6:23 10 T
4443:18
T2
CpH2 O ¼ 0:457 þ 8:33 106 T
þ 7:44 108 T 2
Cpcoke ¼ 0:22275 þ 1:454
6494:44
104 T T2
CpN2 ¼ 0:232 þ 3:57 105 T
a, b, c, d
Areg
Cpair
Cpfv
Cpcat
Cpcoke
Cpi
darm
db
d̄p
dpi
dreac
dreg
driser
Dc
Ek
Ek3c
Ek3h
(95)
(96)
(97)
(98)
CpO2 ¼ 0:2584 þ 8:0625
5865:625
106 T (99)
T2
— Bubble Phase
The bubble phase is considered to be free of catalyst
particles. Hence the only reaction that takes place
within each of the bubble phase CSTRs is the CO
combustion reaction in the gas phase i.e., the homogeneous reaction (r3h). The rate equation for each
CSTR can be accordingly written as
Fair;bubble
Fair;emulsion
Fcat
ent
Fcat
Fj,in
Fj,out
e
Fj;out
k
Fj;out
Foil
g
hriser
hw
K
Kbe
K1
K2
K3
K3c
K3c0
rCO ¼ Vbk K3h PO2 PCO
(100)
where the partial pressures have to be calculated as
described earlier but based on the outlet conditions of
the corresponding bubble phase CSTR. The balance
equations for other components are similar to that
written for the emulsion CSTR.
Dilute phase
The dilute phase is modelled as a series of CSTRs. The mass,
momentum and energy balance equations for the CSTRs are
similar to those written for the emulsion phase. Since
hydrogen in coke is assumed to be completely combusted
K3h
K3h0
Kd
Kk0
L
Mfeed
Mik
Nb
Ns
Pair
Pmixer
empirical constants in the regenerator
hydrodynamic equations
cross sectional area of the regenerator, m2
heat capacity of air, kcal mol1 C1
heat capacity of the oil vapour, kcal kg1 C1
heat capacity of the catalyst, kcal kg1 C1
heat capacity of the coke on catalyst, kcal kg1 C1
specific heat of component i, kcal kg1 K1
inside diameter of the side arm of the RTD, m
effective bubble diameter, m
average particle size, m
particle size, mm
internal diameter of the reactor, m
internal diameter of the regenerator, m
inside diameter of the riser, m
cyclone barrel diameter, m
activation energy for cracking reaction of lump k, kcal mol1
activation energy for the heterogeneous CO
combustion reaction, 13888.55 kcal kmol1
activation energy for the homogeneous CO
combustion reaction, 35555 kcal kmol1
flow rate of air through the bubble phase, m3 s1
flow rate of air through the emulsion phase, m3 s1
catalyst circulation rate, kg h1
entrained catalyst in regenerator cyclones, kg h1
inlet flow rate of component j, kg h1
outlet flow rate of component j, kg h1
outlet flow rate of component j from emulsion
phase CSTR, kg h1
outlet flow rate of component j from bubble phase
CSTR k, kg h1
oil feed flow rate, kg h1
acceleration due to gravity, 9.81 m s2
total height of the riser, m
height of the opening in the RTD arm, m
empirical proportionality constant for cyclone
pressure drop
mass transfer coefficient between emulsion and
bubble phase CSTRs, kg m3 h1
rate constant for carbon combustion reaction to
CO, cm2 kg1 h1
rate constant for carbon combustion reaction to
CO2, cm2 kg1 h1
rate constant for CO combustion reaction to
CO2, kg CO m3 h1(kg cm2)2
rate constant for heterogeneous CO combustion
reaction, kg CO (kg cat)1 h1 (kg cm2)2
intrinsic rate constant for heterogeneous CO combustion
reaction, 3060 kg CO (kg cat)1 h1 (kg cm2)2
rate constant for homogeneous CO combustion
reaction, kg CO m3 h1 (kg cm2)2
intrinsic rate constant for homogeneous CO combustion
reaction, 1.33 1016 kg CO m3 h1 (kg cm2)2
aromatic adsorption coefficient, (Wt frac of aromatics)1
intrinsic rate constant for cracking of component k,
m3 (kg cat)1 h1
loading of the entrained catalyst, kg m3
molecular weight of the oil feed
mass transfer of component i between bubble phase
CSTR k and emulsion phase CSTR, kg h1
number of bubble phase CSTRs
number of spirals made by the solids inside the barrel
blower discharge pressure, psi
pressure at the riser bottom, kg cm2
551
PCO
PO2
Preg
qo
R
RCSV
Re
SCSV
T
Tair
Tbubble,k
Tdelta
Tdilute
Te
Tent
Tin
Tmixer
Tout
Tref
Tsc
Uo
Umf
Ub
Ubr
Vair
Varm
Vkb
Vb
Vc
Ve
Vexit
Vin
Vpin
Vpvessel
Vrtd
Vvapour
Vvessel
w
Wreg
xi
xpt
yAh
Zbed
Zcyc
Greek symbols
d
edense
Zj
nk
mgmix
mair
mv
f(t)
rair
rgmix
rdense
rdil
rpart
rmix
rv
t
DHfT;i
DHrk
partial pressure of CO in the emulsion phase, kg cm2
partial pressure of O2 in the emulsion phase, kg cm2
pressure in the regenerator dilute phase, kg cm2
volumetric flow rate of air per hole in the air grid at
the operating temperature and pressure, m3 s1
universal gas constant, 8.314 Pa m3 mol1 K1;
1.987 kcal kmol1 K1; 82.057 atm cm3 mol1 K1
regenerated catalyst slide valve
Reynolds number
spent catalyst slide valve
temperature of interest of any stream or unit, K
temperature of air at the inlet, K
outlet temperature of the bubble phase CSTR k, K
temperature difference between emulsion and dilute
phase, K
dilute phase outlet or regenerator cyclone inlet
temperature, K
temperature of the emulsion phase CSTR, K
temperature of the entrained catalyst, K
riser CSTR inlet oil vapour temperature, C
temperature at the riser bottom, C
riser CSTR outlet oil vapour temperature, C
reference temperature for enthalpy calculations, 273.15 K
temperature of the spent catalyst, K
superficial velocity of inlet air, m s1
minimum fluidization velocity of air, m s1
velocity of the rise of bubbles, m s1
velocity of the rise of bubbles with respect to
emulsion solids, m s1
velocity of air through jets, ft s1
velocity of the vapour-catalyst mixture in the side arm
of RTD, m s1
volume of bubble phase CSTR k, m3
volume occupied by the bubble phase, m3
velocity of the vapour-steam mixture in cyclone
barrel, m s1
volume occupied by the emulsion phase, m3
exit velocity of the vapour-steam mixture from the
cyclone, m s1
velocity of the vapour-steam mixture at the inlet of the
cyclone, m s1
actual particle velocity at inlet of the cyclone, m s1
superficial velocity of the particles inside the cyclone, m s1
velocity of the vapour-catalyst mixture at the inlet to
the RTD, m s1
velocity of the vapour inside the riser, m s1
velocity of the vapour-steam mixture inside the
cyclone, m s1
width of the opening in the RTD arm, m
catalyst loading in the regenerator bed, kg
fraction of particles in size range i
relative catalytic CO combustion rate
mass fraction of heavy aromatic rings within the riser CSTR
height of the fluidized bed of catalyst, m
height of the cyclone inlet from the air distributor, m
fraction of the dense bed occupied by the bubble phase
void fraction in the dense bed
efficiency of split for the component j
stoichiometric coefficient for component k, (1)
viscosity of the vapour-steam mixture, kg m1 s1
viscosity of air at the exit of air distributor, kg m1 s1
viscosity of the oil vapour at the riser bottom
conditions, kg m1 s1
catalyst deactivation function
density of air at the exit of air distributor, kg m3
mixture density of catalyst and steam, kg m3
density of the dense bed medium, kg m3
density of the dilute phase medium, kg m3
particle density of the catalyst, kg m3
mixture density of catalyst and vapour, kg m3
oil vapour density at the bottom of the riser, kg m3
catalyst residence time within each riser CSTR, s
heat of formation of component i at
temperature T, kcal kg1
heat of cracking of component k, kcal kg1
DPad
DPri
DPrtd1
DPrtd2
DPci
pressure
pressure
pressure
pressure
pressure
drop
drop
drop
drop
drop
across the distributor, psi
terms inside the riser, kg cm2
term inside the RTD, kg cm2
term inside the RTD, kg cm2
terms inside the cyclone, kg cm2
References
Arandes, J.M., Azkoiti, M.J., Bilboa, J. and de Lasa, H.I., 2000, Modelling
fcc units under steady and unsteady state conditions, Can J Chem Eng, 78:
111–123.
Arandes, J.M. and de Lasa, H.I., 1992, Simulation and multiplicity of steady
states in fluidized FCCUs, Chem Eng Sci, 47: 2535–2540.
Arbel, A., Huang, Z., Rinard, I.H., Shinnar, R. and Sapre, A.V., 1995,
Dynamic and control of fluidized catalytic crackers. l. modeling of the
current generation of FCCs, Ind Eng Chem Res, 34: 1128–1243.
Arthur, J.R., 1951, Reactions between carbon and oxygen, Trans Faraday
Soc, 47: 164–178.
Barton, P.I. and Pantelides, C.C., 1994, The modeling of combined
discrete=continuous processes, AIChE J, 40: 966–979.
Barton, G.W. and Perkins, J.D., 1988, Experiences with speedup in mineral
processing industries, Chemical Engineering Research Design, 66:
408–418.
Behie, L.A. and Kehoe, P., 1973, The grid region in a fluidized bed reactor,
AIChE J, 19(5): 1070–1072.
Biegler, L.T., 1989, Chemical process simulation, Chem Eng Progr, Oct:
50–61.
Bird, R., Stewart, W. and Lightfoot, E., 1960, Transport Phenomena (John
Wiley & Sons Inc., New York, USA).
Bogusch, R. and Marquardt, W., 1995, A formal representation of process
model equations, Comput Chem Eng, 19(Suppl): S211–S216.
Bogusch, R. and Marquardt, W., 1997, A formal representation of process
model equations, Comput Chem Eng, 21(10): 1105–1115.
Boston, J.F., Britt, H.I. and Tayyabkhan, M.T., 1993, Software: tackling
tougher tasks, Chem Eng Progr, Nov: 38–49.
Bozicevic, J. and Lukec, D., 1987, Dynamic mathematical model for a fluid
catalytic cracking process, Trans Inst Meas Control, 9(1): 8–12.
CPCL, 1999, FCC Operating Manual.
de Lasa, H.I., Errazu, A.F., Barreiro, E. and Solioz, S., 1981, Analysis of
fluidized bed catalytic cracking regenerator models in an industrial scale
unit, Can J Chem Eng, 59: 549–553.
de Lasa, H.I. and Grace, J.R., 1979, The influence of a freeboard region in a
fluidized bed catalytic cracking regenerator, AIChE J, 25: 984.
Errazu, A.F., de Lasa, H.I. and Sarti, F., 1979, A fluidized bed catalytic
cracking regenerator model—grid effects, Can J Chem Eng, 57: 191–197.
Ford, W.D., Reineman, R.C., Vasalos, I.A. and Fahrig, R.J., 1976, Modeling
catalytic cracking regenerators, NPRA Annual Meeting,
Gruhn, G., Rosenkranz, J., Werther, J. and Toebermann, J.C., 1997, Development of an object-oriented simulation system for complex solids
processes, Comput Chem Eng, 21(Suppl): S187–S192.
Guigon, P. and Large, J.F., 1984, Application of the Kunii-Levenspiel model
to a multi-stage baffled catalytic cracking regenerator, Chem Eng J, 28:
131–138.
Jacob, S.M., Gross, B., Voltz, S.E. and Weekman, V.M., 1976, A lumping
and reaction scheme for catalytic cracking, AIChE J, 22(4): 701–713.
Jones, J.H., Braun, W.G., Daubert, T.E. and Allendorf, H.D., 1967, Estimation of pressure drop for vertical pneumatic transport of solids, AIChE J,
13: 608.
Kraemer, D.W. and de Lasa, H.I., 1988, Catalytic cracking of hydrocarbons
in a riser simulator, Ind Eng Chem Res, 27(11): 2002–2008.
Kraemer, D.W., Sedran, U. and de Lasa, H.I., 1990, Catalytic
cracking kinetics in a novel riser simulator, Chem Eng Sci, 45(8):
2447–2452.
Krishna, A.S. and Parkin, E.S., 1985, Modeling the regenerator in commercial fluid catalytic cracking units, Chem Eng Prog, 81(4) 57–62.
Kumar, S., Chadha, A., Gupta, R. and Sharma, R., 1995, A process simulator
for an integrated FCC-regenerator system, Ind Eng Chem Res, 34:
3737–3748.
Kunii, D. and Levenspiel, O., 1968, Bubbling bed model for kinetic process
in fluidized beds, gas-solid mass and heat transfer and catalytic reactions,
Ind Eng Chem Proc Des Dev, 7(4): 481–492.
Kunii, D. and Levenspiel, O., 1990, Fluidized reactor models. l. for bubbling
beds of fine, intermediate and large particles. 2. for the lean phase: Free
board and fast fluidization, Ind Eng Chem Res, 29: 1226–1234.
Larocca, M., Ng, S. and Lasa, H., 1990, Fast catalytic cracking of heavy gas
oils: modeling coke deactivation, Ind Eng Chem Res, 29: 171–180.
552
Lee, L., Chen, Y., Huang, T. and Pan, W., 1989b, Four-lump kinetic model
for fluid catalytic cracking process, Can J Chem Eng, 67: 615–619.
Lee, E. and Groves, F., 1985, Mathematical model of the fluidized bed
catalytic cracking plant, Trans Soc Comp Sim, 2(3): 219–236.
Lee, W. and Kugelman, A.M., 1973, Number of steady-state operating points
and local stability of open-loop fluid catalytic cracker, Ind Eng Chem
process Des Dev, 12(2): 197–204.
Lee, L., Yu, S., Cheng, C. and Pan, W., 1989a, Fluidized-bed catalyst
cracking regenerator modeling and analysis, Chem Eng J, 40: 71–82.
Lohmann, B. and Marquardt, W., 1996, On the systematization of process
model develoment, Comput Chem Eng, 20(Suppl): S213–S218.
Marquardt, W., 1991, Dynamic process simulation — recent progress and
future challenges, Chemical Process Control IV, 131.
Marquardt, W., 1992, An object-oriented representation of structured
process models, Comput Chem Eng, 16(Suppl): S329–S336.
Marquardt, W., 1994, Computer-aided generation of chemical engineering
process models, Int Chem Eng, 34(1): 28–46.
Marquardt, W., 1996, Trends in computer-aided process modeling, Comput
Chem Eng, 20(6=7): 591–609.
McFarlane, R.C., Reineman, R.C., Bartee, J.F. and Georgakis, C., 1993,
Dynamic simulation for model IV FCCU, Comput Chem Eng, 17(3):
275–300.
McGreavy, C. and Isles-Smith, P.C., 1986, Modeling of fluid catalytic
cracker, Trans Inst Meas Control, 8(3): 130–136.
Niemeyer, P. and Knudsen, J. 2000, Learning JAVA (O’Reilly & Associates
Inc., California, USA).
Nilsson, B., 1993, Object-oriented modeling of chemical processes,
Doctoral Dissertation, Lund Institute of Technology, Sweden.
Pantelides, C.C. and Barton, P.I., 1993, Equation-oriented dynamic simulation — current status and future perspectives, Comput Chem Eng,
17(Suppl): S263–S285.
Pantelides, C.C. and Oh, M., 1996, Process modelling tools and their
applications to particulate processes, Powder Tech, 87: 13–20.
Paraskos, J.A., Shah, Y.T., McKinney, J.D. and Carr, N.L., 1976, A
kinematic model for catalytic cracking in a transfer line reactor, Ind
Eng Chem Process Des Dev, 15(1): 165–169.
Perry, R.H. and Green, D.W. (eds) 1997, Perry’s Chemical Engineers’
Handbook (McGraw-Hill, New York, USA).
Piela, P.C., Epperly, T.G., Westerberg, K.M. and Westerberg, A.W., 1991,
Ascend: an object-oriented environment for modeling and analysis: the
modeling language, Comput Chem Eng, 15: 53–72.
Quann, R.J. and Jaffe, S.B., 1992, Structure-oriented lumping: describing
the chemistry of complex hydrocarbon mixtures, Ind Eng Chem Res, 31:
2483–2497.
Quann, R.J. and Jaffe, S.B., 1996, Building useful models of complex
reaction systems in petroleum refining, Chem Eng Sci, 51(10):
1615–1653.
Rowe, P.N. and Partridge, B.A., 1965, An x-ray study of bubbles in fluidized
beds, Trans Inst Chem Eng, 43: 157.
Shah, Y.T., Huling, G.T., Paraskos, J.A. and McKinney, J.D., 1977, A
kinematic model for an adiabatic transfer line catalytic cracking reactor,
Ind Eng Chem Process Des Dev, 16(1): 89–94.
Smith, J.M. and Van Ness, H.C., 1987, Introduction to Chemical Engineering Thermodynamics (McGraw-Hill, New York, USA).
Sriramulu, S., Sane, S., Agarwal, P.K. and Mathews, T., 1996, Mathematical
modelling of fluidized bed combustion. 1. Combustion of carbon in
bubbling beds, Fuel, 75: 1351–1362.
Stephanopoulos, G., Henning, G. and Leone, H., 1990, MODEL.LA. A
language for process engineering. Part I and II, Comput Chem Eng, 14:
813–869.
Takatsuka, T., Sato, S., Morimoto, Y. and Hashimoto, H., 1987, A reaction
model for fluidized-bed catalytic cracking of residual oil, Int Chem Eng,
27: 107–116.
Toebermann, J.C., Rosenkranz, J., Werther, J. and Gruhn, G., 2000, Blockoriented process simulation of solids processes, Comput Chem Eng, 23:
1773–1782.
Weekman, V.W. and Nace, D.M., 1970, Kinetics of catalytic cracking
selectivity in fixed, moving and fluid bed reactors, AIChE J, 16: 397–404.
Weisz, P.B. and Goodwin, R.B., 1966, Combusion of carbonaceous deposits
within porous catalyst particles. II. Intrinsic burning rate, J Catal, 6:
227–236.
Yang, W.C., 1973, Estimating the solid particle velocity in vertical pneumatic
conveying lines, Ind Eng Chem Fund, 12(3): 349–352.
Yang, W.C., 1974, Correlations for solid friction factors in vertical and
horizontal pneumatic conveying, AIChE J, 20(3): 605–607.
ACKNOWLEDGEMENTS
The authors wish to acknowledge the support provided by Invensys India
Private Limited (IIPL), Foxboro Division, Chennai, India throughout the
duration of the project. We wish to specially acknowledge and thank
B. Jayaram (Manager, Advanced Applications Group) for extending his support
and cooperation in various model development activities during the project.
The manuscript was received 14 February 2003 and accepted for
publication after revision 3 September 2003.