11/10/2022

University of Khartoum
Faculty of Engineering 2 Outlines:
Chemical Engineering Department
 Input-output Models and Transfer Functions
 Block Diagram
 Frequency Response
 Dynamic behavior of process Systems
 First order systems

1 Modelling and Analysis

for Process Control

3 Input-output Models and Transfer 4

Functions
 A fundamental model solving for all dependent variables is often not
required for process control, because the control system is principally
involved with all input variables but only one or a few output variables.
Thus, we need a method for "compressing" the model, which can be
achieved by first grouping variables into three categories: input (causes),
output (effects), and intermediate.
 For linear dynamic models used in process control, it is possible to eliminate
intermediate variables analytically to yield an input-output model, so that
intermediate variables are considered in the model even though they are
not explicitly calculated. Thus, no further assumptions or simplifications are
involved in input-output modelling of linear systems.
 A very common manner for presenting input-output models, which finds
considerable application in process control, is the transfer function.
 The transfer function of a system
is defined as the Laplace transform of the output variable (response), Y(t), divided
by the Laplace transform of the input variable (forcing function), X(t), with all initial
conditions equal to zero.
 Transfer functions will be represented by G(s), with subscripts to denote the
particular input-output relationship when more than one input-output
relationship exists.

5 6 Example

 The assumptions of Y (0) = 0 and X(0) = 0 are easily achieved by expressing  The continuous stirred-tank mixing model formulated in a previous lecture is
the variables in the transfer function as deviations from the initial conditions. solved here.
Thus, all transfer functions involve variables that are expressed as deviations
from an initial steady state.
 All derivatives are zero if the initial conditions are at steady state. (Systems
having all zero initial conditions are sometimes referred to as "relaxed.")
 The fundamental model in deviation variables is
 The prime symbol “ ‘ " for deviation variables is redundant and is not used
here when dealing with transfer functions.

 The Laplace transform is taken of each term in the model:

1

7  The initial value of the tank concentration, expressed as a deviation
variable, is zero, and the deviation of the inlet concentration is constant at
the step value for t > 0; that is, . defined for each input-output relationship. Since the transfer function is a
 Substituting these values and rearranging above equation gives:
 The inverse transform of the expression can be determined from the table
of Laplace transform to give the same expression as derived in previous
lecture.
 This can be rearranged to give the transfer function for this system:
 The model could be used to form transfer function, because it was in terms
of deviation variables with zero initial conditions.
9 10
 A zero is a root of the numerator of the transfer function.
 Physical systems conform to a specific limitation between the orders of the
numerator and denominator; that is, the order of the denominator must be
larger than the order of the numerator. This limitation results from the
observation that real physical systems do not contain pure differentiation,
as would be required for a system with a numerator order greater than the
denominator order.
 The steady-state gain is the steady-state value of ∆Y/ ∆X for all systems
whose outputs attain steady state after an input perturbation ∆X. The
steady-state gain is normally represented by K, often with a subscript, and
can be evaluated by setting s = 0 in the (stable) transfer function. This is
exact for linear systems and gives the linearized approximation for nonlinear
systems.
11 Block Diagrams
 The transfer function introduced in the previous section describes the
behavior of the individual input-output system on which it is based. Often,
several different individual systems are combined, and the behavior of the
combined system is to be determined.
 The overall model could be derived by writing all equations in a large set,
taking the Laplace transforms, and combining into one transfer function.
Another approach retains the distinct transfer functions of the individual
systems and combines these transfer functions into an overall model. This
second approach is usually preferred because
1. It retains individual systems, thereby simplifying model changes (e.g., a different
sensor model).
2. It provides a helpful visual representation of the cause-effect relationships in the
overall system.
3. It gives insight into how different components of the system influence the overall
behavior (e.g., stability).
11/10/2022
8  Note that the transfer function relates one output to one input variable. If
more than one input or output exists, an individual transfer function is
linear operator (as a result of the zero initial conditions), the effects of
several inputs can be summed to determine the net effect on the output.
 The transfer function clearly shows some important properties of the system
as:
 The order of the system is the highest derivative of the output variable in the
defining differential equation, when expressed as a combination of all
individual equations. For transfer functions of physical systems, the order
can be easily determined to be the highest power of (s) in the
denominator.
 A pole is defined as a root of the denominator of the transfer function; thus,
it is the same as a root of the characteristic polynomial. Important
information on the dynamic behavior of the system can be obtained by
analyzing the poles, such as
1. The stability of the system
2. The potential for periodic transients.
 In a specific situation the behavior of an output variable, from time 0 to
completion of the response, depends on its initial conditions, input forcing,
and transfer function (input-output) model. However, some very important
properties of linear dynamic systems depend only on the transfer function,
because the properties are independent of initial conditions and type of
(bounded) forcing functions.
 To summarize the procedure for determining the transfer function for a
process:
Step 1. Write the appropriate balance equations (usually mass or energy
balances for a chemical process).
Step 2. Linearize terms if necessary.
Step 3. Place balance equations in deviation variable form.
Step 4. Laplace-transform the linear balance equations.
Step 5. Solve the resulting transformed equations for the transfer function, the
output divided by the input.
12  The block diagram provides the method for combining individual transfer
functions into an overall transfer function.
2

13  The block diagram can be prepared based on linearized models (transfer
functions) of individual units and the knowledge of their interconnections.
Then an input-output model can be derived through the application of
block diagram algebra, which uses the three operations in figures (a
through c). The model reduction steps normally followed are
1. Define the input and output variables desired for the overall transfer
function.
2. Express the output variable as a function of all variables directly affecting it
in the block diagram. This amounts to working in the direction opposite to the
cause-effect relationships (arrows) in the diagram.
3. Eliminate intermediate variables by this procedure until only the output and
one or more inputs appear in the equation. This is the input-output equation for
the system.
4. If a transfer function is desired, set all but one input to zero in the equation
from step 3 and solve for the output divided by the single remaining input. This
step may be repeated to form a transfer function for each input.
15 Frequency Response
 An important aspect of process (and control system) dynamic behavior is
the response to periodic input changes, most often disturbances. The range
of possible dynamic behavior can be determined by considering cases at
different input frequencies.
 If an input variation is slow, with a period of once per year, the output
response would be essentially at its steady-state value (the same as the
input), with the transient response being insignificant.
 If the input changed very rapidly, say every nanosecond, the output would
not be significantly influenced; that is, its output amplitude would be
insignificant.
 Finally, if the input varies at some intermediate frequency near the response
time of the process, the output will fluctuate continuously at values
significantly different from its mean value.
 The behavior at extreme frequencies is easily determined in this thought
experiment, but the method for determining the system behavior at
intermediate frequencies is not obvious and is useful for the design process
equipment, selection of operating conditions, and formulation of control
algorithms to give desired performance.

17
11/10/2022
14  The overall transfer functions can be used to determine some important
properties of the system without solving the defining differential equations.
These properties include
1. The final value of the output variable
2. The stability of the response
3. The response of the output to a sine input
 Naturally, information about the entire transient is not obtained by
analyzing the poles of the transfer function or by the frequency response
calculations. The complete transient response can be obtained if needed
from analytical or numerical solution of the algebraic and differential
equations.
 See Figure 4.11 in Process Control – Designing Processes and Control
Systems for Dynamics Performance.
16  For general frequency response analysis, periodic inputs will be limited to
sine inputs, which will be a mathematically manageable problem. Also,
only the "long-time" response (i.e., after the initial transient, when the output
is periodic) is considered.
 The frequency response defines the output behavior of a system to a sine
input after a long enough time that the output is periodic. The output ( Y')
of a linear system will be a sine with the same frequency as the input ( X'),
and the relationship between input and output can be characterized by
 Phase angle = phase difference between the input and output
18 Example
 The feed composition to a reactor varies with an amplitude larger than
acceptable for the reactor. It is not possible to alter the upstream process
to reduce the oscillation in the feed; therefore, a drum is located before
the reactor to reduce the feed composition variation, as shown in Figure
What is the minimum volume of the tank required to maintain the variation
at the inlet to the reactor (outlet of the tank) less than or equal to ± 20 g/m3
3
 11/10/2022


19 20  Since V is a constant to be determined, the equation is linear, and we can
proceed without linearization. This could be solved by using either the
integrating factor or Laplace transforms. Here, the Laplace transform of
that equation is taken to give, after some rearrangement,

 The dynamic behavior of the concentration can be determined by

evaluating the inverse of the Laplace transform.
 The resulting expression for the time behavior is given in the following
equation:

21  The first term in previous equation tends to zero as time increases; thus, the 22 
response of the process after a long time of operation (about four time
constants) is not affected by this term. The second term describes the "long-
time" behavior of the concentration in response to a sine input. It is
periodic, with the same frequency as the input forcing and an amplitude
that depends on the input amplitude and frequency, as well as process
design parameters. For this example, the output amplitude must be less
than or equal to 20; by setting the amplitude equal to the limit, the time
constant, and thus the volume, can be calculated.

23 Example

Dynamic Behavior of Typical

24
Process Systems

4
 11/10/2022

25 Introduction 26 First-Order System



 physical systems, which involve very different physical principles, can have
similar dynamic behavior. Thus, it is possible to acquire understanding of a
large number of systems from a thorough study of a much smaller number
of basic models.
 Here we study some fundamental model structures that occur frequently in
process plants, along with their effects on dynamic behavior. This
experience will enable us to recognize the effects of process designs on
dynamic behavior.

27 Transient Response 28  STEP RESPONSE

 If a step change of magnitude A is introduced into a first-order system, the
 Now that the transfer function of a first-order system has been established, transform of X ( t ) is
we can easily obtain its transient response to any forcing function. Since this
type of system occurs so frequently in practice, it is worthwhile to study its
response to several common forcing functions: step, impulse, ramp, and
sinusoidal.  The transfer function is

 STEP FUNCTION. Mathematically, the step

function of magnitude A can be
expressed as  Combining these equations gives:

 A step function can be approximated

very closely in practice. For example, a  This can be expanded by partial fractions to give
step change in flow rate can be
obtained by the sudden opening of a
valve.

29  Solving for the constants C 1 and C 2 by the techniques covered before,
gives C 1 A and C 2 A. Inserting these constants into equation of Y(s) and
taking the inverse transform give the time response for Y:
30  Several features of this response are worth
remembering:
1. The value of Y ( t ) reaches 63.2 percent of its ultimate
value when the time elapsed is equal to one time
constant t . When the time elapsed is 2 t , 3 t , and 4 t ,
the percent response is 86.5, 95, and 98, respectively.
From these facts, one can consider the response
essentially completed in three to four time constants.
2. One can show from equation that the slope of the
response curve at the origin in figure is 1. This means that
if the initial rate of change of Y ( t ) were maintained,
the response would be complete in one time constant.
(See the dotted line in figure.)
3. A consequence of the principle of superposition is that
the response to a step input of any magnitude A may
be obtained directly from figure by multiplying the
ordinate by A. Figure actually gives the response to a
unit-step function input, from which all other step
responses are derived by superposition.

5
 11/10/2022

31 Example: Liquid Level 32  a transient mass balance around the tank:

 A time-varying volumetric flow q of liquid of constant

density ρ enters the tank.
 Determine the transfer function that relates head to
flow.  We will introduce deviation variables into the analysis before proceeding to
the transfer function. Initially, the process is operating at steady state, which
means that dh/dt = 0

33  Subtracting 34  Rearranging into the standard form of the first-order lag to give:

 If we define the deviation variables as

 The term R is simply the conversion factor that relates h(t) to q(t) when the
system is at steady state, this value is the steady-state gain.
 We can verify the physical significance of this value by applying the final-
 Then
value theorem to the determination of the steady-state value of H when
the flow rate Q ( t ) changes according to a unit-step change; thus

 Taking the transform  The transform of Q ( t ) is

 Applying the final-value theorem to H ( s ) gives

 Notice that H (0) is zero, and therefore the transform of dH/dt is simply sH (s).

35  36 Liquid-Level Process with Constant-Flow

Outlet


6
 11/10/2022

37  Taking the Laplace transform 38 

 the solution of this equation is

 Clearly, if we increase the inlet flow to the tank, the level will increase
because the outlet flow remains constant. The excess volumetric flow rate
into the tank accumulates, and the level rises.
 For instance, if a step change Q ( t ) = u ( t ) were applied to the system,
the result would be

 The step response given by this equation is a ramp function that grows
without limit. Such a system that grows without limit for a sustained change
in input is said to have nonregulation. Systems that have a limited change
in output for a sustained change in input are said to have regulation.

39 Mixing Process 40  Assume the density of the solution to be constant, the flow rate in must
equal the flow rate out, since the holdup volume is fixed.
 Consider the mixing process shown in the figure in which a stream of  We may analyze this system by writing a transient mass balance for the salt;
solution containing dissolved salt flows at a constant volumetric flow rate q thus
into a tank of constant holdup volume V.
 The concentration of the salt in the entering stream x (mass of salt/volume)
varies with time. It is desired to determine the transfer function relating the
outlet concentration y to the inlet concentration x.

 At steady state, this becomes

41  Introducing deviation variables. 42 Using MATLAB to Obtain the Response of a

First-Order System to a Step Function
 give
 The transform of the response is 10/s(0.1s + 1). We can simulate that in
MATLAB by defining a system using the numerator and denominator of the
response:

 Taking the Laplace transform of this expression and rearranging the result
give

7
 11/10/2022

43
 >> plot(t,temp)
44 Linearization
 Most physical systems of practical importance are nonlinear.
 Characterization of a dynamic system by a transfer function can be done
only for linear systems. The convenience of using transfer functions for
dynamic analysis, which we have already seen in applications, provides
significant motivation for approximating nonlinear systems by linear ones.
 Assume that the flow out of the tank (example of liquid level with
resistance) follows a square root relationship

 where C is a constant.
 For a liquid of constant density and a tank of uniform cross-sectional area
A, a material balance around the tank gives

45  46  Where and

 Substituting in material balance equation gives

 At steady state the flow entering the tank equals the flow leaving the tank;
thus

 After rearrangement

 Introducing deviation variables and transforming give

where

47 48 Using MATLAB to Compare Nonlinear

(Exact) Solutions and Linearized Solutions

8
 11/10/2022

49