Ch-2

Modelling
Aspects
Modelling
• A model is an abstract representation of a system, process, or
concept, built to understand, analyze, and predict its behavior.
• Models can be physical, mathematical, or computational.
• The process of creating a mathematical or physical representation of a
real-world system to analyze and predict its behavior.
• Example: Representing a distillation column as a set of material and
energy balance equations.
Introduction
• Most of the engineering processes that demand accurate product
quality, and proceed at high rates, high temperatures, and high
pressures, are distinct for their utmost complexity.
• A simple change in one of the variables may bring about complex and
non-linear changes in other variables.
• The external potential of information about any engineering process
is extremely high.
• This complex situation can be handled diligently with very narrow
channels of perception by gaining an insight into a particular process
using models.
• A model is a simplified representation of those aspects of an actual
process.
Deterministic and Stochastic
Processes
Deterministic Process
• In this process the observables take on a continuous set of values in a
well-defined (or definable) manner, while the output variable most
representative of the process is uniquely determined by the input
variable.
• These processes can be adequately described by classical analysis and
numerical methods.
• An example is the process that takes place in a simple continuous
stirred tank reactor (CSTR).
• A deterministic process is one whose behaviour can be predicted
exactly.
Stochastic Process
• This is a process in which observables change in a random manner
and often discontinuously.
• The output variable is not directly related to the input variable.
• These processes are described in terms of statistics and probabilistic
theory.
• Examples are the contact-catalytic process (packed beds) in which the
yield of the product diminishes with decrease in the activity in the
catalyst as it ages with time and the pulse properties (such as pulse
frequency, pulse velocity, pulse height, base hold-up, and pulse hold-
up) in trickle bed reactors.
• For stochastic processes, we can predict its response only
approximately.
Physical Modelling
• In physical modelling, the experiment is carried out directly on the
real process.
• The process of interest is reproduced on different scales, and the
effect of physical features and linear dimensions is analysed.
• The experimental data are reduced to relationships involving
dimensionless groups made up of various combinations of physical
quantities and linear dimensions.
• The relationships determined with this dimensionless presentation
can be generalized to classes of events having these dimensionless
groups, or similarity criteria.
• The resulting models are also known as ‘empirical models’.
• Physical modelling consists in seeking the same or nearly the same
similarity criteria for the model and the real process.
• The real process is modelled on a progressively increasing scale, with
the principal linear dimensions scaled up in proportion (the similarity
principle).
• Thus, a physical model is restrained directly within the system where
the real process of interest takes place.
• This approach requires that the process be modelled up to the
commercial scale, along with the complex systems that one has to
deal with in chemical engineering.
Mathematical Modelling
• A mathematical model of a real chemical process is a mathematical
description which combines experimental facts and establishes
relationships among the process variables.
• Mathematical modelling is an activity in which qualitative and
quantitative representations or abstractions of the real process are
carried out using mathematical symbols.
• In building a mathematical model, a real process is reduced to its bare
essentials, and the resultant scheme is described by a mathematical
formalism (formulation) selected according to the complexity of the
process.
• The resulting models could be either analytical or numerical in nature
depending upon the method used for obtaining the solution.
• The objective of a mathematical model is to predict the behaviour of
a process and to work out ways to control its course.
• The choice of a model and whether or not it represents the typical
features of the process in question may well decide the success or
failure of an investigation.
• A good model should reflect the important factors affecting a process,
but must not be crowded with minor, secondary factors that will
complicate the mathematical analysis and might render the
investigation difficult to evaluate.
• Depending on the process under investigation, a mathematical model
may be a system of algebraic or differential equations or a mixture of
both.
Mathematical modelling involves three steps:
• formalization-the mathematical description of the process under
investigation (mathematical formulation)
• development of an algorithm for the process
• testing of the model and the solution derived from it
Process to Build a Model
Basis:
• The bases for mathematical models are the fundamental physical and
chemical laws, such as the laws of conservation of mass, energy, and
momentum.
Assumptions:
• Probably the most vital role an engineer plays in modelling is in
exercising his engineering judgment as to what assumptions can be
validly made.
Mathematical consistency of model:
• Once all the equations of the mathematical model have been written,
it is usually a good idea, particularly with big, complex systems of
equations, to make sure that the number of variables equals the
number of equations.
Solution of the model equations:
• The available solution techniques and tools must be kept in mind as a
mathematical model is developed. An equation without any way to
solve it is not worth much.
Verification:
• An important but often neglected part of developing a mathematical
model is proving that the model describes the real-world situation.
Degrees of Freedom Analysis for
Model
• Degrees of freedom of the system must be zero in order to obtain a
solution.
• If this is not true, the system is underspecified or over specified and
something is wrong with the formulation of the problem.
• This kind of consistency check may seem trivial, but experience shows
that it can save many hours of frustration, confusion, and wasted
computer time.
• This is required for a simulation exercise.
• For an optimization exercise, there should be some degrees of
freedom available for optimizing, that is, an optimization problem is
an underspecified problem.
• Checking to see that the units of all terms in all equations are
consistent is perhaps another trivial and obvious step, but one that is
often forgotten.
• It is essential to be particularly careful of the time units of parameters
in dynamic models. Any unit can be used (seconds, minutes, hours,
etc.), but these cannot be mixed.
Independent and Dependent
Variables,
and Parameters
• Independent variables: These are quantities describing a system that
can be varied by choice during a particular experiment, independent
of one another.
• Examples include time and coordinate variables.
• Dependent variables: These are the properties of a system which
change when the independent variables are altered in value. There is
no direct control over a dependent variable during an experiment.
• The relationship between independent and dependent variables is
one of cause and effect; independent variables measure the cause
and dependent variables measure the effect. Examples for dependent
variables include temperature, concentration and efficiency.
• Parameters: This is by far the largest group, consisting mainly of the
characteristic properties of the apparatus and physical properties of
the materials.
• The term parameters is used because the properties remain constant
during an individual experiment, but take different constant values
during different experiments.
• Examples include overall dimensions of the apparatus, flow rates,
heat transfer coefficients, thermal conductivity, heat capacity, density,
initial or boundary values of dependent variables.
Single-stage extraction
• The equilibrium relationship is

• The mole fraction of the solute in the raffinate is

• Eliminating x from Eq. 1 using Eq. 2, we get

• The symbols used in the above example can be classified as follows:
1. Independent variable: θ
2. Dependent variables: x, y
3. Parameters: m, R, S, c, V1, V2,
• Parameters are all fixed in value during the experiment, but can be
varied between experiments for comparison purposes.
• θ is the only variable whose value can be altered during an
experiment.
• x and y vary during an experiment and between experiments; their
values depend upon the choice of values for both the parameters and
the independent variable.
Classification of Models
• Mathematical modelling is classified under three different bases:
1. Variation of various independent variables
2. State of the process
3. Type of the process
Classification of mathematical modelling
Classification based on Variation
of Independent Variables
• This classification of mathematical modelling is done based on
whether the process variables vary with time as an independent
variable or with both time and space coordinates as independent
variables.
1. Distributed Parameter Models
2. Lumped Parameter Models
Distributed Parameter Models
• If the basic process variables vary with both time and space, or if
these changes occur only with space of dimension exceeding unity
(i.e., more than one space coordinate), such processes are
represented by distributed parameter (DP) models, which are
formulated as partial differential equations.
Lumped Parameter Models
• Processes in which the basic process variables vary only with time are
represented by lumped parameter (LP) models, which are formulated
as ordinary differential equations.
Classification Based on the
State of the Process
• The primary objective of cybernetics is to control a given system or
process.
• As a consequence, a complete mathematical model is expected to
describe relationships between the basic process variables under
steady-state conditions (a static model) and transient conditions (a
dynamic model).
• A static or steady-state model ignores the changes in process
variables with time.
• The construction of a static model involves the following steps:
• Step 1: Analysis of the process to establish its physical and chemical
nature, its objective, the governing equations describing a given class
of processes and also its specific features as a unit process.
• Step 2 Identification and ascertaining of the input and output
variables of the process. The variables include Fig. the following:
1. Control variables: Variables that change with the progress of the
process (yi)
2. Manipulated variables: Variables that directly affect the course
of the process and that can be measured and changed
purposefully (xi).
 3. Disturbances: Variables that directly affect the course of the
process but that cannot be changed purposefully (zi).
4. Intermediate variables: Variables related to the course of the
process only indirectly (xni).
• Step 3 Identification and establishing of the relationships, constraints,
and boundary conditions of the process. The static model of a unit
process should allow for all possible modes of operation of the typical
reactor.
• The construction of a dynamic or unsteady-state model reduces to
obtaining the dynamic characteristics of the process, i.e., establishing
relationships between its main variables as they change with time.
• Dynamic characteristics can be obtained by theory, experiment, or
both.
• In obtaining dynamic characteristics experimentally, disturbances are
applied to the input of a system and the time response to the
disturbance is noted.
• Such experiments are based on the laws of signal flow dealt with in
information arid control theory.
• The dynamic model of a process may take the form of any of the
following:
1. A set of transfer functions relating the selected dependent variable
to one or several independent variables.
2. Ordinary or partial differential equations derived theoretically and
containing all the necessary dependent and independent variables.
3. Equations derived for the various elements of the unit process that
may be analysed independently of one another.
Classification Based on the Type
of the Process
• Depending on whether a given process is deterministic or stochastic,
it may be represented by any one of the following mathematical
models
1. An analytical rigid model
2. A numerical rigid model
3. An analytical probabilistic model
4. A numerical probabilistic model
Rigid or Deterministic Models
• These models usually describe deterministic processes without the
use of probability distributions.
• As mentioned above, the subclassification of rigid (or deterministic)
models is analytical rigid models and numerical rigid models,
depending on whether the solution is obtained analytically or
numerically.
• This is not to say that the underlying phenomena, especially when a
numerical rigid model is used, are not statistical in character.
• This implies that only the averages and not the whole distributions
are dealt with here.
Stochastic or Probabilistic Models
• These models usually represent stochastic (random) processes.
• Here again, the sub classification of probabilistic models is analytical
probabilistic (or stochastic) and numerical probabilistic (or stochastic)
models, depending on whether the solution is obtained analytically or
numerically.
• It may be noted that simulation corresponding to a numerical
stochastic model is known as Monte Car10 simulation.
The Black Box Principle
• The black box principle is employed when, in an analysis of complex
processes, one cannot establish internal relationships within the
system.
• As a result, the object under study appears to be enclosed in a
container, the so-called black box, to which we, as observers, have
incomplete access.
• By this principle, the mathematical description is concerned only with
the dependence of outputs on the inputs and ignores the interior
structure of the process.
• The concept of the black box is a pivotal, as it provides a useful
mathematical approach to the behaviour of complex systems, i.e., to
their responses to various external stimuli, without making any
attempt to understand their internal structure.
• Many systems, especially large ones, are so complex that even with
perfect information about the state of their components, it cannot
practically be related to the behaviour of the system as a whole.
• In such cases, the matter can be greatly simplified by representing a
complex system as a black box model which functions in the same
manner as the system.
• Analysing the behavior of the model and comparing it with that of the
system, one can derive a number of conclusions as regards the
properties of the system, and if the model adequately fits the
prototype system, one can select a working hypothesis about the
internal structure of the system.
• The black box principle proves extremely useful when one system has
to be replaced with another functioning in a similar way.
• For example, in the automatic control of hazardous chemical
manufacture the human operator needs to be replaced with an
automatic device capable of performing the same functions.