Course Introduction

• Professor Jay H. Lee, CBE 341 Process Simulation and Control

• This is an Edu 4.0 course. (Watch the video lecture on Tuesday, come
to the lecture room on Thursday for Q/A, recitation, problem solving,
and quizzes.
• Syllabus
– Read Carefully!
• TA: Mr. Mujin Cheon, Jaeuk Lee(?)
• Class Homepage
– KLMS for CBE341 (klms.kaist.ac.kr)
– Important Announcements
– Video lectures, HWs, Exams, Solutions, Samples
• JAVA Applets for interactive learning, visualization (web link)
• MATLAB/Simulink
• Schedule
– Tentative schedule on the syllabus (will be adjusted due to the
delayed Spring semester start.
– Will be updated regularly. (check KLMS)
• Lectures
– Designed to complement the textbook (Read ahead!)
 Course Introduction(2)
• Course Objective (“Measure of Success”)
– Get the best grade that you are capable of
getting.
– Learn “a little” about process dynamics /
control (and beyond)
– Have fun doing it!
 Course Outline
Process Dynamics
Process Dynamic Modeling
• Models: Unsteady State Material / Energy Balances (thermo,
kinetic, transport)
• ODEs, DAEs, PDEs  ODEs (Initial Value Problem),

Process Simulation Models for Control

• Numerical Integration of ODEs • Linearization  Transfer Function
• Plant Testing and Empirical (Data-
based) Modeling

Process Control
Advanced Control Processwide and
Feedback System Design / System Plantwide Control
Analysis • Feed-forward • Basic Loops
• Stability • Cascade, etc. • Production Control
• PID Controller Design • Model-based Control • Optimization
 Lecture 1
Introduction to Process Control

CBE341
©Professor Jay H. Lee
KAIST
 Process Design vs. Operation
• Design
– Flowsheet synthesis
– Detailed design of unit operations
• Operate
– Most processes are meant to be at steady state,
but in reality, very much dynamic
– Control: maintain the operation at intended
steady state(s) despite disturbances
– Optimize: Determine economically favorable
operating condition (it can change!) and drive
the process there.
 Basic Principles of Process Operation

• Operate safely.
• Meet the product quality specifications.
• Meet the required production rate (or maximize).
• Minimize operating costs (energy, raw materials)
 Examples of Continuous Processes
Chapter 1
 Examples of Batch Processes

start end
Chapter11
Chapter
Example: Refinery Fractionator

1. No explosion, other accidents, or long shutdowns

2. Meet the composition spec’s for each stream
3. Meet the intended production rate for each stream
4. Minimize energy consumption
 Operational Issues with Crude Switch

• Light→Heavy
– Less Naptha than before → Reduce its production rate
or lower the product quality spec’s or both.
– The operating condition (e.g., temperatures) needs to be
adjusted accordingly.
Economic Benefits of Control
octane
rating
higher
Economic Benefits of Control
Ensuring plant safety
 Bhopal Disaster
The Bhopal disaster was an industrial disaster
that took place at a Union Carbide pesticide
plant in the Indian city of Bhopal, Madhya
Pradesh. At midnight on 3 December 1984,
methyl isocyanate (MIC) gas was accidentally
released from the plant, exposing more than
500,000 people to MIC and other chemicals.
The first official immediate death toll was
2,259. The government of Madhya Pradesh has
confirmed a total of 3,787 deaths related to the
gas release.[6] Others estimate 8,000-10,000
died within 72 hours and 25,000 have since
died from gas-related diseases. 40,000 more
were permanently disabled, maimed, or
rendered subject to numerous grave illnesses;
521,000 exposed in all.[7][8][9] As of 2010 no
one has yet been prosecuted for the disaster.
“The Bhopal disaster was the result of a combination
of legal, technological, organizational, and human
errors. The immediate cause of the chemical reaction
was the seepage of water (500 liters)into the MIC
storage tank. The results of this reaction were
exacerbated by the failure of containment and
safety measures and by a complete absence of
community information and emergency procedures.
The long term effects were made worse by the
absence of systems to care for and compensate the
victims. Furthermore, safety standards and
maintenance procedures at the plant had been
deteriorating and ignored for months.”
TED Case Studies
 Some of the Found Defects
•Gauges measuring temperature and pressure in the various
parts of the unit, including the crucial MIC storage tanks, were
so notoriously unreliable that workers ignored early signs of
trouble (Weir, pp.41-42).
•The refrigeration unit for keeping MIC at low temperatures
(and therefore less likely to undergo overheating and expansion
should a contaminant enter the tank) had been shut off for some
time (Weir, pp.41-42).
•The gas scrubber, designed to neutralize any escaping MIC,
had been shut off for maintenance. Even had it been operative,
post-disaster inquiries revealed, the maximum pressure it could
handle was only one-quarter that which was actually reached in
the accident (Weir, pp.41-42).
•Etc. Etc. Etc.
 What’s Process Control?
• Measure and monitor the process variables.
• Adjust the process’s (dynamic) degrees-of-
freedom (e.g., valve positions, pump speeds)
– To maintain the relevant variables within acceptable
ranges or close to their desired values (“setpoints”)
– To move the process to a new operating condition fast
but smoothly without violating constraints (as in startup,
shutdowns, and grade transitions)
• Read Textbook Examples:
– Cruise Control
– Blending Process
– Distillation Column
Example of Process Control
Blood pressure regulation
Brief history
 Application of Control Theories
• Chemical Processes
• Mechanical Systems
– Robots, Cars, etc.
• Military Applications
– Fighters, Satellites, Space Structures, Reconnaissance
Vehicles
• Biomedical Systems
• Electronic Circuits
• Power Plants
• Supply Chain

Control is a truly a multi-disciplinary engineering science.

 Closed-loop Artificial Pancreas
glucose u
setpoint
y
r

controller pump patient sensor

measured glucose
 Basic Questions to Ask
• What is the end goal? E.g., What are the variables
to be controlled (CVs)?
• What are the obstacles? E.g., What are the
disturbances (DVs) or other changes?
• What are the degrees of freedom I have to achieve
the end goal? E.g., What are the manipulated
variables (MVs)?
• What information will I have? E.g., What are the
measured variables (MSVs)? Can you measure the
CVs? How well and fast?
Loop Control
Process Control
Plantwide Control
Example: Crude Oil Furnace

Main CV

DVs
MVs
 Manual Control

Operator adjustments of the two valves based on his

observations and prior experience
• Overadjustment: “Myopic” and impatient.
• Inconsistency
 Feedback Control

Perfect end-result (no “offset”) but adjustment takes place only after
the temperature is affected significantly.
 Feedforward Control

• Faster response to disturbances in the feed.

• Adjustment is not perfect and “offset” results. WHY?
 Combined FB/FF Control

• Combines the advantage of both FB (no offset) and FF

control (fast response to disturbances in the feed)
• Still slow to respond to disturbances in the fuel pressure.
 FF/FB + Cascade Control

Slave
Master

•Best configuration
•Responds fast to disturbances in the fuel pressure as well as feed
•No offset.
 Why (When) Do We Need Process
Control?
• Imperfect process design
– Incorrect parameter values
– Unmodelled phenomena
• Disturbances
– Feed condition, utility supply pressure / temperature, catalyst
activity, heat transfer efficiency, etc.
• Changes in the operating condition
– Feed switch
– Grade transitions (as in polymer plants)
– Production rate change
– Real-time optimization (RTP) for economic improvement
• Efficient startup, shutdowns
– Distillation columns, Reactors
Servo vs. Regulatory Control
 Control Hardware Elements

• Sensors (↔ vision, hearing, olfactory, etc.)

• Transmitter (↔ neurons)
• Controllers (↔ brain)
• Final control element (↔ muscles)
 Classification of Variables

• Input variables (independent variables)

– Manipulated inputs: u =[QF, QA]
– Disturbances (measured or unmeasured): d=[F, Ti,, PF, λF]
• Output variables (dependent variables)
– Controlled variables (measured or unmeasured): y=[T, F0]
• State variables(x) ⊇ Output variables:
– Variables determining the internal dynamic condition
Variable Types: Example
 Types of Control Strategies
• Manual vs. Automatic
• Open-loop vs. Closed-loop
• Strategies
– Feedback Control
– Feedforward Control
– Cascade Control
• Linear vs. Nonlinear
Open-Loop Control
Feedforward Control
Feedback Control
 Power of Feedback

• Feedback is very important in control!

• With feedback, you can control despite substantial
uncertainty and incomplete knowledge about the
process.
– Imperfect knowledge of disturbances
– Imperfect knowledge of the dynamic effect of MVs on
CVs.
• Open-loop control or just feedforward control
alone seldom works well in an uncertain plant
environment.
Uncertainty and feedback control
Positive vs negative feedback loop
Effect of positive and negative
feedback
Regulation via negative feedback
Amplification via positive
feedback
PID feedback control
Feedback stabilization
Feedback stabilization of double-
inverted pendulum
 Stabilization of oscillation in a flexible
mechanical structure
• Mechanical Structures

Wind speed tower displacement

0.2
AR (cm/s/N)

uncontrolled

0.1 w/ active
mass damper

0
0.20 0.21 0.22 0.23 0.24 0.25
frequency (Hz)
 Continued
Active mass damper is turned on
mass position here!

PI
+ +
motor DVFB mass displacement
velocity
ref
tower velocity

structure displacement
tower

Response of bridge
tower with wind
disturbance.
Feedback can also destabilize!
Even the negative feedback.
 Feedback Control

Perfect end-result (no “offset”) but adjustment takes place only after
the temperature is affected significantly.
 More Terminologies
• Deadtime
• SISO (Single-Input/Single Output) vs. MIMO
(Multiple-Input/Multiple Output) process
• Continuous vs. Batch / Semibatch Process
• Open-Loop Stable vs. Open-Loop Unstable
Process
• Feedback Error = Setpoint – Measured CV Value
• Offset = Steady-State Error
• Valve or Transmitter Saturation
• Valve Deadband
 Types of Control Algorithm

• On-Off Control
• Proportional Control
p (t ) = p + kc (rm (t ) − ym (t )) ⇒ p ' (t ) = kc e(t )
gain
• Proportional-Integral Control
 1
t

p (t ) = p + kc  e(t ) + ∫ e(t*)dt *
 τI 0 
Integral time or reset time
• Proportional-Integral-Derivative (PID) Control
 1
t
de 
p (t ) = p + kc  e(t ) + ∫ e(t*)dt * +τ D 
 τI 0 dt 
Derivative time constant
• Model-Based Control
 Commonly Found Control Loops
• Flow Loops
– Orifice meter, Proportional Control
• Level Loops
– Differential Pressure Sensor, Proportional Control
• Pressure Loops
– Pressure Sensor, P or PI Control
• Temperature Loops
– Theromocouple or RTD, PID Control
• Composition Loops
– Gas Chromatograph, Spectrophotometer, Statistical
Process Control
 Responsibilities of Control Engineers

• Control Strategy Design or Decision

• Process Modification, Revamping
• Controller Tuning
• Controller Troubleshooting / Maintenance
• Documentation of Process Control Changes
 P&ID
Learn to read the P&IDs (Piping and Instrument Diagram):
Figure 1.7
Complexity of a chemical plant’s
P&I diagram
From analog to digital

analog digital
 Distributed Control System
(DCS)
• Many local control units (handling 1000s of loops)
connected with each other and other workstations /
mainframes through network. Operator display
In the control room

Various I/O
readings are
floating around
Each unit is equipped
with ID tags.
w/ many I/O channels
Typical Control Room
Typical Operator Display
Touch the screen to
get the reading on
any specification
location.

(Re)-configure controllers
directly from the screen.
 Computer Integrated Process Management

Advanced
Process Control
(APC)

What we study in
this class mostly.
Example: Liquid Receiver Level Control

Control the level (h) to a desired value.

 Strategy 1

What type of control is this?

 Strategy 2

What type of control is this?

 Strategy 3

What type of control is this?

 Mathematical Analysis: Strategy 1
•Process Model
dh
Ac = Fi − F atsteady
 → 0 = Fi ss − F ss
state

dt
dh dy
Ac = ( Fi − Fi ) − ( F − F ) ⇒ Ac
ss ss
= d −u
dt dt
where y = h − h ss , d = Fi − Fi ss , and u = F − F ss

•Feedback Controller
F = F ss + K (h − h ss ) ⇒ u = Ky
Plot y vs. t for
Assume constant disturbance d=ζ different K values
dy ζ Ky η =ζ − Ky ζ   K 
= −  
→ y (t ) = 1 − exp − t 
dt Ac Ac K  Ac 
Response Under P-Control
(for Strategy 1)
 Mathematical Analysis: Strategy 3
A Feedforward Control Strategy:

dy
F = Fi ⇒ Ac =0
dt

What would happen to h if F=Fi+δ due to a measurement error

or valve error?

dy
F = Fi + δ ⇒ Ac = −δ ⇒ y = ??
dt
 Difficulties for Control

• Process nonlinearity
– We use linear controllers in most cases.
• Model errors
• Delays (process + measurement)
• Instrumentation problems: Imperfect
measurements, imprecise actuators, inaccurate
transmissions
• Complex process structure (with multiple inputs
and outputs)
 Multiple Control Loops

Level and temperature loops: Can affect each other.

Typical industrial process has tens or even hundreds of
loops.
Distillation Column