Control and Modelling of Bioprocesses

1
Control and Modellingof Bioprocesses
Slides adapted from Dr. Katie Third
2
Lecture Outline

• Purpose of Process Control
• Building blocks of process control
• The bioreactor (modelling)
• Sensors
• Actuators
• Controllers
• Basic control schemes
• Basic Controller Actions
• Case examples

3
Process Control

• Guidance of the process along a certain path to
  produce a product that meets predefined quality
  specifications
• The Aim
• To produce the product of interest at a minimum
  of operating costs (ie. Increase the cost/benefit
  ratio)

4
Process Control

• Involves the use of monitored information to
  make decisions that affect the process in a
  desirable way

Make decision
On the right path?
Process
5
Reasons for Process Control

• Easier optimisation of the process
• More constant product quality
• Detection of problems and their location at an
  early stage
• Greater quality assurance

6
4 Basic Building Blocks of a Controlled Process
3. Actuators
4. Controllers
2. Sensors
1. The plant (bioreactor)
7
(1) Bioreactor

• Batch process
• significant changes of process variables over
  time
• requires more complex control
• requires experience with the process (feed
  forward control)
• Steady state processes (chemostat)
• constant process conditions
• more simple process control
• feedback control often sufficient

8
(2) Sensors (Measuring Devices)

• Enable monitoring of the state of the process
• e.g. temperature, DO concentration, biomass
  conc.
• Measurements can be on-line or off-line.

9
On-line Measurements

• Performed automatically
• Results directly available for control
• Monitored continuously
• Off-line Measurements
• Require human interface
• Less frequent and usually irregular
• Best suited for checking and calibrating

10
Types of On-line Measuring Equipment

• Physical Measurements
• Temperature
• Weight
• Liquid flow rates
• Gaseous flow rates
• Liquid level
• Pressure inside vessel

10.12 kg
11
Sensors (continued)

• Physico-Chemical Measurements
• pH
• Oxidation-reduction potential (ORP, Eh)
• Dissolved oxygen
• Conductivity
• Off-gases (CO2, H2, CH4)
• NH4 (ion-selective electrodes)

12
Sensors (continued)

• Biochemical Measurements
• Respiration rate (OUR, SOUR)
• Volatile fatty acids (VFAs)
• Flourescence (e.g. NADH)
• Turbidity

13
Requirements of a good on-line sensor

• Heat and pressure resistant ? autoclavable
• Mechanically robust
• Resistant to bacterial adhesion
• Stable over a long period
• Fast dynamics in relation to the measured
  variable
• Linear characteristics ? easy in-situ calibration

14
(3) Actuators

• Devices which make the changes to the process,
  e.g.
• Aeration pumps
• Stirrers
• Feed pumps
• Chemical dosing pumps
• Inoculation ports
• Recycle pumps

15
(4) Controllers

• Devices that decide on the appropriate action to
  be taken to keep the process running along the
  desired path
• Computers
• Biocontrollers

16
Basic Control Schemes

• Open-Loop Control (Feedforward)
• Closed-Loop Control (Feedback)
• Inferential control
• Combined feedforward and feedback
  (model-supported control)

17
Feedforward Control (Open-Loop Control)

• The pattern of the manipulable variable is
  predetermined, and directly adjusts the actuator
• There is no feedback from the process to the
  controller
• Requires no measurement of the variable
• Often model-based ? requires reliable model
• Large deviations of the process from the required
  path are not corrected for

18
Feedforward Control (Open-Loop Control)
Input
Output
Feedforward controller
Process
E.g. In fed-batch cultivation, the pattern of the
feed rate profile is used to directly adjust the
feed pump
19
Feedback Control (Closed-Loop Control)

• Conventional and most common type of control
  scheme safest
• Measurements from the process are used to
  calculate a suitable control action
• Appropriate when the accuracy requirement is
  higher
• Deviations between the variable and its setpoint
  are used to change the process
• ? smaller deviations

20
Feedback Control (Closed-Loop Control)
Measured output
Actuator
error
Controller
Process
21
Ideal Feedback Controller
2
DO mg L-1
1
Time
22
Overshooting

• If the input signal does not immediately affect
  the output ? delayed action typical of on/off
  controllers
• Caused by things such as
• feed pump too large for required dosage
• delay in sensor response

2
DO mg L-1
1
Time
23
Combined Feedforward and Feedback Control

• To compensate for small model deviations and
  unpredicted disturbances
• Feedforward control establishes control according
  to process model
• Feedback allows for refinement by correcting for
  deviations

24
Combined Feedforward and Feedback Control
Feedforward controller
Process
Feedbackcontroller
Set point
25
Inferential Control

• When direct feedback of the variable of interest
  is not possible, on-line measurements can be used
  to infer the state of the variables (also
  called State Estimation)
• E.g. DO fluctuations ? SOUR

DO
dcL/dt ? OUR
Time
26
State Estimation

• Measurements give indirect information about
  critical variables in the process (e.g. biomass
  activity, biomass concentration, substrate
  concentration etc.)
• Using the on-line measurements to estimate the
  current state of the biomass ? state estimators
  (e.g. SOUR)
• Advantage enables on-line control of a variable
  that cannot be measured on-line
• Modelling plays important role

27
State Estimation

• Also the Control action itself can be recorded
  and used as an online or offline process analysis
  tool.
• For example the total duration over which the
  alkali dosing pump has been switched on, allows
  to calculate the amount of alkali used to
  counteract the acid produced in the bioprocess ?
  Biological acid production is recorded online.

28
Car steering analogy of PID controller
Current signal

• Setpoint

29
Basic Controller Decision making
Get New Temp.

• Temp lt
• Setp.?

N
Y
Turn Heater On
Turn Heater Off
Wait X sec
30
Basic Controller Actions

• Simplest type digital on-off switching, e.g.
  thermostat
• PID control (very common and important)
• Fuzzy logic control, Adaptive Controllers, Self
  learning systems (not covered in this unit)

31
On-Off controller

• E.g. stop airflow if DO is higher than setpoint ?
  large oscillations of process variable
• can use an acceptable band of values with no
  control action, e.g. If pH gt 8 then run acid
  pump. If pHlt6 then run base pump. ? no precise
  control

32
Proportional Controller

• Multiplies the deviation of the variable from the
  setpoint with a constant, Kp
• The further away the variable from the setpoint,
  the stronger the action
• Control input (Process output Setpoint).Kp ?
  Controller
• signal signal output

33
Proportional controller

• Setpoint

Car steering analogy Check distance from
middle of the lane and correct steering in
proportion to distance from desired position
34
Integral controller

• Setpoint

• Car steering analogy
• Look out through the back window and keep track
  of
• how long the car has been out of desired position
  and
• by how much.
• How long (sec) how much (m) is the integral
  (secm).
• The longer the car was positioned away from the
  setpoint the stronger the signal
• Good to correct for long term and only slight
  deviation from setpoint.

35
Integrating Controllers

• Integration of a curve ? area under the curve
• Integrated input signal is multiplied by a
  factor, Ki

36
Integrating Controllers

• A purely integrating controller is slow and
• Error takes long time to build up
• Action can become too strong ? overshooting
• Int controller is unaware of current position ?
  Generally used combined with P control (looking
  at current position) PI control

37
Differentiating Controller

• Examines the rate of change of the output of the
  process
• The faster the change, the stronger the action
• The derivative of the output (slope) is
  multiplied by a constant, Kd

38
Car steering analogy of Differential controller -

• Setpoint

39
Differentiating Element and PID Controllers

• Differential control is insensitive to slow
  changes
• If the variable is parallel to the setpoint, no
  change is made (slope 0)
• Differential control is very useful when combined
  with P and I control ? PID control

40
Problems with individual PID control elements

• Setpoint

P Alarm strong left turn needed I No problem
Past Right and Left errors are about equal D No
problem Direction is parallel to setpoint
41
Problems with individual PID control elements

• Setpoint

P No problem Signal position is on setpoint D
Alarm Direction is wrong. Left turn needed
42
Conflicting or neutralising advice by PID control
elements

• Setpoint

P Alarm Position too far left. Turn right D
Alarm Direction too far towards right. Turn
Left. position is on setpoint
43
Time Analogy of PID Controllers

• P Present time. Only considers current position.
  Not aware of current direction and of error
  history
• I Past time. Only compiles an error sum of the
  past. Not aware of current distance of signal
  from setpoint and of current direction.
• D Future time. Only considers current direction
  (trend). Now aware of current distance of signal
  from setpoint and of error history.

44
Questions True of False?

• Differentiating elements are capable of detecting
  small changes providing they occur rapidly
• Integrating elements always respond rapidly to
  changes in output signals
• A long delay time in a feedback control system
  may lead to considerable overshoot

- TRUE
- FALSE
- TRUE
45
Questions True of False?

• Time between changes in measured values and
  control action should always be as short as
  possible
• A proportional controller once set up to maintain
  an output of a process at a setpoint will not
  require any re-adjustment to ensure the output
  remains constant
• A state estimator allows us to operate on-line
  control of a variable for which no on-line
  measurements are available

- FALSE
- Usually FALSE
- TRUE
46
Content beyond this point is not examinable
47
Proportional Integral Derivative (PID) Controllers

• Conventional and classical approach of control
  engineering
• Parameters Kc, ?I and ?D can be determined from
  simple experiments

48
Determining the PID values
DO mg L-1
A KA/B gain
B
Time
a
T
Actuating signal
Process response
49
Determining the PID values

• Ziegler/Nicols Procedure
• PID Control
• KC (1.2/K) T/a (proportional)
• ?I 2.0 a (differential)
• ?D 0.5 a (integral)

50
Adaptive Controllers (not examinable)

• The state of the biomass changes continuously
  during the course of a non-steady state
  bioprocess (the car may turn into a boat)
• Required PID values of controller change
• Adaptive controllers continuously adjust control
  parameters during the running process
• Requires finding how to tune the control values
• ? Experimentation and finding linear
  relationships between state of biomass and PID
  values

51
Adaptive Controllers

• Result in significant improvements to the control
• Tuning of control parameters can be easy when
  simple black-box assumptions can be made
• When simple assumptions are not adequate, process
  dynamics must be considered in a process model
• Model-supported control (or combined feedback and
  feedforward control

52
Fuzzy Logic Control

• Useful when concrete knowledge cannot be
  transformed into mathematical equations
• Based on fuzzy logic
• e.g. If happens, take action
• Although very simplified, whole bioprocesses can
  run effectively on fuzzy logic rules

53
Learning Outcomes

• You should be able to
• Explain the range of control schemes that exist
  for controlling a bioprocess
• Understand how the different types of controllers
  work
• Identify which variables will need controlling in
  a bioprocess
• Identify useful features of an on-line measuring
  device
• Recognize applications of process control in the
  food industry