Process ControlTitolo presentazione

and Instrumentation Laboratory (PC&IL)

A.Y. 2022-2023, Lecture 2
sottotitolo
• Milano, XX mese 20XX
Prof. Carlo Giorgio Visconti
Politecnico di Milano, Dipartimento di Energia,
Laboratory of Catalysis and Catalytic Processes
LCCP Laboratory of Catalysis and Catalytic September
Processes21th, 2022
 2
Monitoring and controlling a process: example

Control requirements:
- Measuring the level (h)

- Comparing the level

reading with a set-point
value (hs) and computing
an error (e=hs-h)

- Generating a command
(c) depending on the error

Open tank: feedback level control - Trasmitting the control

(disturbance first, then) action to a regulating

device (final control elem.)

LCCP Laboratory of Catalysis and Catalytic Processes

 3
Dynamic behaviour of first-order systems

A 1st order system is one whose output is modelled by a 1st order differential eq.:

𝑑𝑦 𝑡 𝜏𝑝 Time constant
𝜏𝑝 + 𝑦 𝑡 = 𝐾𝑝 𝑓 𝑡
𝑑𝑡 𝐾𝑝 (Static) gain
First order processes are characterized by their capacity to store material or energy and the resistance
associated with the flow of mass or energy in reaching the capacity

𝑑ℎ 𝑡 𝜌𝑔ℎ 𝑡 8𝜂𝐿
𝐹 𝑡 𝐴 =𝐹 𝑡 − 𝑅=
𝑑𝑡 𝑅 𝑆𝐸𝐶 2

𝒉(𝒕) A 𝐴𝑅 𝑑ℎ 𝑡 𝑅
R +ℎ 𝑡 = 𝐹 𝑡
𝜌𝑔 𝑑𝑡 𝜌𝑔

𝑑𝑇 𝑡
𝑻(𝒕) 𝑚𝑐𝑝 = 𝑈𝐴 𝑇𝑣𝑎𝑝 − 𝑇(𝑡)
𝑑𝑡
UA
𝑚
𝑚𝑐𝑝 𝑑𝑇 𝑡
+ 𝑇(𝑡) = 𝑇𝑠𝑡
steam
𝑈𝐴 𝑑𝑡
LCCP Laboratory of Catalysis and Catalytic Processes
 4
Dynamic behaviour of first-order systems

Let’s imagine that a 1° order system is at steady-state. Intuitively, what

happens to 𝒚 𝒕 if, starting at t=0, 𝒇 𝒕 is suddenly changed by a factor X?
A
𝑇 𝑡
𝑋
steam
𝑚𝑐𝑝 𝑑𝑇 𝑡
+ 𝑇(𝑡) = 𝑇𝑠𝑡
𝑈𝐴 𝑑𝑡

𝑇 𝑡 reaches a new steady state for t going

to infinity (i.e. it is autoregulating, but is not
stable): the system needs to be controlled
to keep T at the desired value.

𝑡
𝑚𝑐𝑝
𝑈𝐴
LCCP Laboratory of Catalysis and Catalytic Processes
 5
Dynamic behaviour of pure-capacitive system

A pure-capacitive system is one whose output is modelled by a 1st order differential

equation missing the linear term:
𝜏𝑝 Time constant
𝑑𝑦 𝑡
𝜏𝑝 = 𝐾𝑝 𝑓 𝑡
𝑑𝑡 𝐾𝑝 (Static) gain

Let’s imagine that a pure capacitive system is at

𝐹 𝑡 steady-state (i.e. F(t)=F0). What happens to 𝒚 𝒕 if,

starting at t=0, 𝒇 𝒕 is suddenly changed by a factor X?

𝒉(𝒕) A 𝐹𝑜 ℎ 𝑡 ℎ 𝑡 never reaches a new

𝑋 steady state: the system is
not autoregulating and
needs to be controlled to
𝑑ℎ 𝑡 remain stable.
𝐴 = 𝐹 𝑡 − 𝐹𝑜
𝑑𝑡
𝑡
𝐴
LCCP Laboratory of Catalysis and Catalytic Processes
 6
Dynamic behaviour of pure-capacitive system

The majority of the processes relevant for chemical

engineers are first order or pure-capacitive!

LCCP Laboratory of Catalysis and Catalytic Processes

 7
Dynamic behaviour of 2nd order systems

A 2nd order system is one whose output is modelled by a 2nd order differential eq.:

𝑑 2
𝑦 𝑡 𝑑𝑦 𝑡 𝜏 Natural period of oscillation
2
𝜏 + 2𝜉𝜏 + 𝑦 𝑡 = 𝐾𝑝 𝑓 𝑡 𝜉 Damping factor
𝑑𝑡 2 𝑑𝑡
𝐾𝑝 (Static) gain
Multicapacitive processes: Inherently second
𝐹 𝑡 2nd-order systems obtained order systems
by the in-series combination
of two 1st order capacitive
ℎ1(𝑡) A1
systems

R1

𝒉𝟐(𝒕) A2
R2

1st order systems,

once controlled,
may become 2nd
order
LCCP Laboratory of Catalysis and Catalytic Processes
 8
Dynamic behaviour of multicapacitive 2nd order systems

𝐹 𝑡 Let’s imagine that a 2° order system is at

steady-state. Intuitively, what happens to
ℎ1(𝑡) A1
𝒉𝟐 𝒕 if, starting at t=0, 𝑭 𝒕 is suddenly
R1
changed by a factor X?

𝒉𝟐(𝒕) A2
R2

The response resembles a little the

response of a 1° order system, but
the system initially delays to
respond and then its response is
rather sluggish. This response is
defined OVERDAMPED

LCCP Laboratory of Catalysis and Catalytic Processes

 Dynamic behaviour of inherently 2nd order systems, or 9
2nd order caused by the process-controller interaction

These systems can bring to OVERDAMPED responses (if the dumping factor is > 1) or
UNDERDAMPED responses (if the dumping factor is response is < 1).

UNDERDAMPED responses are faster than OVERDAMPED responses at the beginning, reaches
the ultimate value quickly, but then it does not stay there, but it starts oscillating with progressively
decreasing amplitude.

LCCP Laboratory of Catalysis and Catalytic Processes

 10
How to bring a perturbed system back to its initial state?

Example: we want to keep 𝑻 𝒕 = 𝒄𝒐𝒏𝒔𝒕.

Assumptions: 𝐹𝑖 = 𝑐𝑜𝑛𝑠𝑡. = 𝐹 (volumetric flowrates) and 𝑇𝑠𝑡 = 𝑐𝑜𝑛𝑠𝑡.. 𝑻𝒊 varies along with time.

𝑑𝑇 𝑡
𝑚𝑐𝑝 = 𝐹𝜌𝑐𝑝 𝑇𝑖 𝑡 − 𝑇 𝑡 +𝑄 (1)
𝑑𝑡

At steady state (𝑇(𝑡) = 𝑐𝑜𝑛𝑠𝑡. ≝ 𝑇𝑠 ):

0 = 𝐹𝜌𝑐𝑝 𝑇𝑖,𝑠 − 𝑇𝑠 + 𝑄𝑠 (2)

Let’s subtract (2) from (1):

𝑑 𝑇 𝑡 − 𝑇𝑠
𝑚𝑐𝑝 = 𝐹𝜌𝑐𝑝 𝑇𝑖 𝑡 − 𝑇𝑖,𝑠 − 𝑇 𝑡 − 𝑇𝑠 + 𝑄 − 𝑄𝑠 (3)
𝑑𝑡

𝑻 𝒕 − 𝑻𝒔 is the error we want to drive to zero by introducing

a control loop.

LCCP Laboratory of Catalysis and Catalytic Processes

 11
Introducing a control action!

How to define the control action 𝒄 𝒕 ?

LCCP Laboratory of Catalysis and Catalytic Processes

 12
Proportional controller

P Controller 𝑐 𝑡 = 𝛼 𝜀 𝑡 + 𝑐𝑠

Proportional control: 𝑸 = −𝜶 𝑇 𝑡 − 𝑇𝑠 + 𝑸𝒔 (4)

Replacing (4) in (3):

𝑑 𝑇 𝑡 − 𝑇𝑠
𝑚𝑐𝑝 = 𝐹𝜌𝑐𝑝 𝑇𝑖 − 𝑇𝑖,𝑠 − 𝑇 𝑡 − 𝑇𝑠 − 𝛼 𝑇 𝑡 − 𝑇𝑠
𝑑𝑡

Solving this equation in 𝑇 𝑡 − 𝑇𝑠 for different values of 𝛼:

None of the solutions

is effective in achieving
𝑇 𝑡 − 𝑇𝑠 = 0 , but the
higher 𝛼 is, the smaller
the error will be. The
ultimate DT is named
offset.

LCCP Laboratory of Catalysis and Catalytic Processes

 13

Let’s try to make this result “more general”

LCCP Laboratory of Catalysis and Catalytic Processes

 Dynamic behaviour of controlled processes: 14
First order process – P controller.

𝑐 𝑡 = 𝐾𝑐 𝜀 𝑡 + 𝑐𝑠 𝐾 c Proportional gain

Imagine that one disturbance (aka load, e.g. Ti in

the example on the right) has a step-change (at
constant set-point). How does a first-order
system, equipped with a P controller, react?

An offset is observed at steady-state, i.e. y(t)

does not reach ysp even after an infinite time,
which decreases by increasing Kc

LCCP Laboratory of Catalysis and Catalytic Processes

 Dynamic behaviour of controlled processes: 15
First order process – P controller.

A similar behaviour is observed when we consider how a first order reacts to a step change in
the set-point.

Imagine the set-point for our variable (y) is initially 0. At t = 0 we suddenly change the set-point
from 0 to 2. How does a first-order system, equipped with a P controller, react?

a) An offset is observed at steady-state,

i.e. y(t) does not reach ysp even after an
infinite time.

b) The offset decreases by increasing Kc

c) The response rate also grows with Kc

From this moment on, we will only consider the responses to a step change in the set-point.
The system response to a step change in the load is always equivalent.

LCCP Laboratory of Catalysis and Catalytic Processes

 Dynamic behaviour of controlled processes: 16
Second order process – P controller.

Imagine the set-point for our variable (y) is initially 0. At t=0 we suddenly change the set-point
from 0 to 2. How does a second-order system, equipped with a P controller, react?

a) An offset is observed at steady-state,

i.e. y(t) does not reach ysp even after an
infinite time.

b) The response rate grows with Kc

c) The offset decreases by increasing Kc

d) When Kc is high, the system may show

an UNDERDAMPED response, with
oscillations growing with Kc

A similar conclusion applies for a step change in the load.

LCCP Laboratory of Catalysis and Catalytic Processes

 17
Feedback proportional control: summary

• The order of the (uncontrolled) process is not changed by introducing a P control

• Inaccurate control: an off-set exists, which decreases for increasing Kc

• P control is fast: Kc controls the speed

• For high values of Kc, second order systems may become unstable (oscillating)

LCCP Laboratory of Catalysis and Catalytic Processes

 18

Do we have alternatives to P control?

LCCP Laboratory of Catalysis and Catalytic Processes

 Let’s go back to the example of slide 10 19
Integral controller
𝑡

I Controller 𝑐 𝑡 = 𝛽 න 𝜀 𝑡 𝑑𝑡 + 𝑐𝑠
0
𝑡
Integral control: 𝑸 = −𝛽 ‫׬‬0 𝑻 − 𝑻𝒔 𝑑𝑡 + 𝑸𝒔 (4)

Replacing (4) in (3):

𝑡
𝑑 𝑇 𝑡 − 𝑇𝑠
𝑚𝑐𝑝 = 𝐹𝜌𝑐𝑝 𝑇𝑖 − 𝑇𝑖,𝑠 − 𝑇 𝑡 − 𝑇𝑠 − 𝛽 න 𝑇 𝑡 − 𝑇𝑠 𝑑𝑡
𝑑𝑡
0

Solving this equation in 𝑇 𝑡 − 𝑇𝑠 for different values of 𝛽:

Some of the solutions

are effective in
achieving 𝑇 𝑡 − 𝑇𝑠 =
0, i.e. in eliminating the
offset. The quality of
control depends on the
value of 𝛽.

LCCP Laboratory of Catalysis and Catalytic Processes

 20

Let’s try to make this result “more general”

LCCP Laboratory of Catalysis and Catalytic Processes

 Dynamic behaviour of controlled processes: 21
First order process – I controller.

𝑡
𝐾c 𝐾 c Proportional gain
𝑐 𝑡 = න 𝜀 𝑡 𝑑𝑡 + 𝑐𝑠
𝜏𝑖 𝜏i Integral time
0

a) The response rate grows with increasing Kc and/or decreasing 𝜏𝑖

b) The order of the response increases (+1)
c) The off-set is eliminated
d) The response is slower than with a P control

LCCP Laboratory of Catalysis and Catalytic Processes

 22
Feedback integral control: summary

• The order of the (uncontrolled) process is changed (+1) by introducing a I control

• Accurate control: the targeted set-point is achieved at steady state

• I control is slow: the weaker the integral action is (small ti), the slower the response
is.

• For high values of Kc or low values of ti, even first order systems may become
unstable (oscillating)

LCCP Laboratory of Catalysis and Catalytic Processes

 Dynamic behaviour of controlled processes: 23
First order process – PI controller.

𝑡
𝐾c 𝐾 c Proportional gain
𝑐 𝑡 = 𝐾c 𝜀 𝑡 + න 𝜀 𝑡 𝑑𝑡 + 𝑐𝑠
𝜏𝑖 𝜏i Integral time
0

a) The response is fast (P) – no sluggish

behaviour at the beginning

b) The order of the response is increased

(I)

c) The off-set is eliminated (I)

d) The response rate increases with

increasing Kc/𝜏𝑖

LCCP Laboratory of Catalysis and Catalytic Processes

 Dynamic behaviour of controlled processes: 24
D controller.

𝑑𝜀 𝑡 𝐾 c Proportional gain
𝑐 𝑡 = 𝐾 c𝜏 𝑑 + 𝑐𝑠
𝑑𝑡 𝜏d Derivative time

The purpose of derivative action is to anticipate where the process is going by looking at
the rate of change of error, de/dt.

Advantage: it does not change the order of the response

Disadvantage: it does not do anything when the error is constant → it makes sense only if
used in combination with P or PI actions

LCCP Laboratory of Catalysis and Catalytic Processes

 Dynamic behaviour of controlled processes: 25
PD controller.

𝑑𝜀 𝑡 𝐾 c Proportional gain
𝑐 𝑡 = 𝐾 c 𝜀 𝑡 + 𝐾 c𝜏 𝑑 + 𝑐𝑠
𝑑𝑡 𝜏d Derivative time

The presence of the D action, does not help in eliminating the main drawabacks of the P
control: an off-set exists

For this reason, PD controllers are not used!

LCCP Laboratory of Catalysis and Catalytic Processes

 Dynamic behaviour of controlled processes: 26
PID controller.

𝑡 𝐾 c Proportional gain
𝐾c 𝑑𝜀 𝑡
PID control 𝑐 𝑡 = 𝐾c 𝜀 𝑡 + න 𝜀 𝑡 𝑑𝑡 + 𝐾c𝜏𝑑 + 𝑐𝑠 𝜏i Integral time
𝜏𝑖 𝑑𝑡
0 𝜏d Derivative time

a) The response is fast (P) – no sluggish

PI PID behaviour at the beginning

b) The order of the response is increased (I)

c) The off-set is eliminated (I)

d) The response is more robust (D): even for

fast responses (high Kc/𝜏𝑖 ) the D action
(which is stronger when 𝜏𝑑 is large) limits
the oscillations.

LCCP Laboratory of Catalysis and Catalytic Processes

 Dynamic behaviour of controlled processes: 27
PID controller.

Note 1:

Superimposing the responses of the same

systems controlled with a PI or with a PID control,
we notice that the presence of the D action slows
𝑑𝜀 𝑡
down the system ( is usually < 0)
𝑑𝑡

Note 2:

The D action is more effective on slow processes,

i.e. when the error does not changes frequently its
derivative over time

LCCP Laboratory of Catalysis and Catalytic Processes

 28
Summary: types of feedback controllers

1. PROPORTIONAL (P)

𝑐 𝑡 = 𝛼 𝜀 𝑡 + 𝑐𝑠

2. PROPORTIONAL-INTEGRAL (PI)
𝑡

𝑐 𝑡 = 𝛼 𝜀 𝑡 + 𝛽 න 𝜀 𝑡 𝑑𝑡 + 𝑐𝑠
0

3. PROPORTIONAL-INTEGRAL-DERIVATIVE (PID)
𝑡
𝑑𝜀 𝑡
𝑐 𝑡 = 𝛼 𝜀 𝑡 + 𝛽 න 𝜀 𝑡 𝑑𝑡 + 𝛾 + 𝑐𝑠
𝑑𝑡
0
Video P vs PID
LCCP Laboratory of Catalysis and Catalytic Processes
 29
How to select the most appropriate control type?

1. What is the aim of the control action?

- Readiness (P)

- Accuracy (I)

- Stability (D)
Textbook plot used to show the
response of a system to a unit 2. What about the dynamics of the process to
step change in the load with no control?
control and P, PI and PID control.
- Velocity

- Noiseness

Are PID controllers always the

best choice?

LCCP Laboratory of Catalysis and Catalytic Processes

 30
Level control (in a “tank”)

Case 1: Large cross section of the tank Case 2: Small cross section of the tank
The level evolves slowly The level evolves quickly
The accuracy is not fundamental; The accuracy is fundamental;
→P → PI

Note:
D action is usually not adopted for level control, especially when the free surface is
rough/agitated (for example because the liquid is boiling)

LCCP Laboratory of Catalysis and Catalytic Processes

 31
Flow control (in a “pipe”)

The flowrate evolves very rapidly

It is a noisy variable (due to turbulence)

→ PI

D action is not used, not to damage the valve (quick opening and closing)

LCCP Laboratory of Catalysis and Catalytic Processes

 32
Pressure control (in a “vessel”)

Case 1 - Gas Case 2 - Vapour

Slow process (large vessels are typically Rather fast / fast processes
used compared to inlet flowrare) Accuracy is needed to prevent
Accuracy is not needed condensation

→P → PI or PID

Case 3 – Liquids: the pressure varies quickly and is noisiy → PI

LCCP Laboratory of Catalysis and Catalytic Processes

 33
Temperature control

Temperature evolves slowly (involving slow heat

transfer processes)
Accuracy is neeed

→ PID

LCCP Laboratory of Catalysis and Catalytic Processes

 34
Composition control

Controlling the composition often requires long analyses → composition is not used as
measured varibale in designing control loops.

LCCP Laboratory of Catalysis and Catalytic Processes

 35
Conventional feedback control Loop (SISO)

Consider the generalized process shown in the figure. It

has an output y, a potential disturbance d (load), and an
available manipulated variable m. What is shown in the
figure is called open-loop representation of the process.

The closed loop representation of the process represents the same process, when:
1) a measuring device is used to measure the output variable (ym)
2) a comparison is made between ym and the desired value (ysp), calculating the error (aka
deviation)
3) the value of the error is used by the controller to generate a command c, which is sent to
the final control element /usually a control valve) to affect the manipulated variable

LCCP Laboratory of Catalysis and Catalytic Processes

 36
Feedback control with multiple loops (MIMO)

Example: Exothermic reaction occurring in a CSTR, cooled with a coolant flowing in the jacket.
GOAL: Keeping T constant

SISO MIMO

SISO reacts well to changes in Tfeed, but it’s rather slow to compensate changes in Tcoolant

MIMO (CASCADE CONTROL) improve the response to changes in Tc, because it doesn’t wait
that the effect of a variation of the coolant temperature has been felt by the reacting mixture. The
loop measuring T is dominant (primary), the loop measuring Tc is secondary (slave)

LCCP Laboratory of Catalysis and Catalytic Processes

 37
CASCADE Control

Consider a process consisting of 2 parts: process I and process II.

Process I (primary) has as its output the variable we want to control (in the previous example,
the reaction in the tank and the temperature T).

Process II (secondary) has an output we are not interested in controlling but which affects the
output we want to control (in the previous example, the reactor jacket and its output Tc).

LCCP Laboratory of Catalysis and Catalytic Processes

 38
CASCADE Control

SISO

This is the typical simple (SISO) feedback control system

LCCP Laboratory of Catalysis and Catalytic Processes

 39
CASCADE Control

SISO

MIMO
CASCADE

LCCP Laboratory of Catalysis and Catalytic Processes

 40
CASCADE Control

This is the typical cascade control system,

where disturbances arising within the
secondary loop are corrected by the
secondary controller before they can affect
the value of the primary controlled output.

MIMO
CASCADE:
Single Input Multiple Output
From a process perspective
Multiple Input Single Output
From a controller perspective

LCCP Laboratory of Catalysis and Catalytic Processes

 41
Typical examples of CASCADE Control

Heat exchanger

Control objective: exit T of stream 2 constant

Secondary loop objective: compensate for changes

in the flow rate of stream 1

Furnace

Control objective: exit T of the process stream

leaving the furnace

Secondary loop objective: compensate for

changes in the fuel flow rate

LCCP Laboratory of Catalysis and Catalytic Processes

 42
Typical examples of CASCADE Control

Distillation column - bottom

Control objective: T (and thus concentration) at the

bottom of the column

Secondary loop objective: compensate for changes

in the flow rate of steam to the reboiler

Distillation column - top

Control objective: T (and thus concentration)

at the top of the column

Secondary loop objective: compensate for

changes in the flow rate of recycle to the
column

In chemical
LCCP processes,
Laboratory flow-rate
of Catalysis control Processes
and Catalytic loops are very often cascaded with other control loops.
 43
OVERRIDE Control (Single Input, Multiple Output)

During the normal operation of a plant or during its startup/shutdown, dangerous situations
may arise which may lead to destruction of equipment. In such cases it is necessary to change
from the normal control action. This can be achieved through the use of special switches.

LSS = Low Switch Selector

OVERRIDE:
Single Input Multiple Output
From a process perspective
Multiple Input Single Output
From a controller perspective

Example 1: protection of a boiler system. The LSS prevents that the liquid level in the boiler
falls below the allowable limit (the heating coil must remain immersed in water not to burn out)

LCCP Laboratory of Catalysis and Catalytic Processes

 44
OVERRIDE Control (Single Input, Multiple Output)

Example 2: protection of a steam distribution system. The HSS prevents that the high
pressure steam line is excessively pressurized.

HSS = High Switch Selector

LCCP Laboratory of Catalysis and Catalytic Processes

 45
OVERRIDE Control (+ CASCADE Control)

Example 3: protection of a compressor system. The HSS prevents that discharge pressure
from the compressor exceeds an upper limit. FC and PC are cascaded to the speed control of
the compressor’s motor.

HSS = High Switch Selector

SC = Speed Control

ST = Speed Transducer

LCCP Laboratory of Catalysis and Catalytic Processes

 46
AUCTIONEERING Control (Single Input, Multiple Output)

Such a control configuration select among several similar measurements the one with the
highest value and feed it to the controller (several measured outputs, one manipulated input).
AUCTIONEERING:
Single Input Multiple Output
From a process perspective
Multiple Input Single Output
From a controller perspective

Example: catalytic tubular reactor with highly exothermic reactions. The Auctioneering
system allows to control the coolant flowrate depending on the hot-spot temperature.

LCCP Laboratory of Catalysis and Catalytic Processes

 47
SPLIT-RANGE Control (Multiple Input, Single Output)

Such a control splits the controlled output into several parts, each affecting one of the available
manipulated variables affecting this output

Example: control of the pressure in chemical reactor. The two

SPLIT-RANGE:
valves are coordinated by the controller, so that the pressure in Multiple Input Single Output
the reactor remains at the set-point value. From a process perspective
Single Input Multiple Output
From a controller perspective
LCCP Laboratory of Catalysis and Catalytic Processes
 48
SPLIT-RANGE Control (Multiple Input, Single Output)

Example: control of the pressure in a steam header (collecting steam from several parallel
boilers). The control must keep constant the pressure in the steam header when the steam
demand changes.

LCCP Laboratory of Catalysis and Catalytic Processes

 [email protected]
Politecnico di Milano, Dipartimento di Energia
LCCP – Laboratory of Catalysis and Catalytic Processes
Via La Masa 34, 20156 - Milano, Italy
LCCP Laboratory of Catalysis and Catalytic Processes