1. Chapter 9 – Part 1
Higher Level Automation
Techniques
CHEE413: Chemical Process Control
Instructor: Dr. Monica Titus
2. 2
Spring Semester: 16 weeks (4 days left!!)
1/11 (w) Syllabus Review; Ch1 2/24 f Ch4 quiz 5 4/17 m CH7
1/13 (F) Ch1 2/27 m Ch4 4/19 w Exam 3, Ch5, 6
1/16 M holiday 3/1 w Ch4 4/21 f Ch7
1/18 w Ch1 quiz1 hw1 3/3 f CH5 hw 6 4/24 m Ch7, HW9, quiz 8
1/20 f Ch2 3/13 m CH5 4/26 w Ch9
1/23 m Ch2 3/15 w CH5 4/28 f CH9
1/25 w Ch2 hw2 3/17 f Ch5 5/1 m Class cancelled
1/27 f Ch2 3/20 m Exam 2 5/3 w Review, HW10, Quiz9
1/30m Ch2 quiz 2 3/22 w CH5 5/11 10:30-12:30 Exam 4
2/1 w Ch8 hw3 3/24 f CH5
2/3 f Ch8 3/27 m CH5, HW7, quiz6
2/6 m Ch8 quiz 3 3/29 w CH5
2/8 w Exam 1 3/31 f CH5
2/10 f Ch3 4/3 m Ch6
2/13 m Ch3 hw 4 4/5 w CH6
2/15 w Ch3 quiz 4 4/7 f Ch6
2/17 f Ch4 4/10 m Ch6
2/20 m Ch4 hw 5 4/12 w Ch6, HW8, quiz7
2/22 w CH4 4/14 f CH7,
Design day
Review day
• Final Exam
• Mostly Chapter 7
• Partial (concept) Ch 9
• Exam will be written to complete in 50 min
(although will have the full exam block)
• Exam 4 is non-cumulative but will be weighted
at 20%
3. In this chapter…
We will gain an introductory understanding of more advanced
automation functions that can be layered on top of the basic
regulatory controls to increase process efficiency and reliability.
3
9.1 Plant automation concepts
9.2 Advanced process controls
9.3 Plant-wide and process area automation*
9.4 Higher level automation of batch and semi-batch processes*
4. What are the general goals of every process
facility?
4
5. Plant Automation Concepts
5
The general overall goals of every process facility are:
• Maximizing profit
• Meeting acceptable safety & health criteria
• Minimizing environmental impact
These are too general to be directly useful.
• Must build specific objectives for each actual process facility
6. How do we automate a plant?
6
Entire Plant
Indiv
Process
Area
7. A functional model of the PAS
7
Facility Functions
Process Optimization
Functions
Supervisory-Level Functions
Regulatory-Level Functions
Process/Field Instruments/Human Actions
The highest level functions
set the direction and review
progress for the facility
Optimize overall operation
Occur at the process unit level,
manipulates measurements
Monitors & Controls using
process the direct process
using
9.1.3
9.1.2
9.1.1
8. A data / information transfer model of the PAS
8
success
Profits
Corporate
directives
opportunity
Scheduling
Performance
Meets
setpoints
Operations
Constraints
Production
Scheduling
Process
/ Engr
regulation
Operations
Setpoints/
Constraints
Instrument
ation
Engr/Tech
High
accuracy
Regulatory
outputs
Setpoints/
Constraints
Measureme
nts
When data flows through a facility – the “audience” will determine how “high”
or “low” level the data is and what the “relevant” data/content is
9. Online Models
9
Are used to infer properties
for APCs
• These models are especially useful for
predicting compositions, since
analytical measurements are expensive
and slow.
On-line models are also used for
performance assessment such as heat
exchanger fouling, reactor
performance/catalyst activity monitoring,
and rotating equipment efficiency.
• Models must be simple enough to
achieve acceptable response times (in
general, update times should at least
be as fast as an on-line analyzer –
order of seconds to minutes).
10. Supervisory Controls
10
• Objective: implement control
strategies that cannot be implemented
without intermediate manipulation of
direct measurement data used to
determine changes to independent
variable setpoints.
• Global description: A control strategy
that involves the manipulation of the
direct measurements rather than their
direct use. Supervisory controls are
indirect. Measurements are assessed
by the control scheme and then actions
are taken to change the setpoints of
regulatory controllers.
11. Supervisory Controls
11
• A control strategy that uses a control
computer to convert one or more
inputs into one or more outputs
• Regulatory controls use controllers
• Sequential logic control may be
regulatory and still use a control
computer – or a PLC
• Best used to provide ESPs to regulatory
loops instead of direct outputs to CVs
• The code used for an APC within the
control computer is known as the
control script
• Many different higher level languages
can be used (Fortran, basic, C, C++)
• More info on control script is provided
in Appendix B
12. Supervisory Controls (MISO)
12
Note – the APC is different than the CALC
blocks (Ch7)
• MISO: multiple input/single output
• We’ve seen some of these at the
regulatory level
• Using calculation blocks
• MISO APCs may be more efficient than
execution in regulatory controller
blocks
• MISO APCs can execute more complex
controls, and is often more
computational efficient
13. CALC-block based regulatory control
13
We can replace this series of
calculation blocks that results in
a complicated regulatory control
design… with…
Monitoring Heat Exchangers for
Fouling & Scaling
14. Converted to APC - MISO
14
Monitoring Heat Exchangers for
Fouling & Scaling
TI
103
TE
103
Process
Fluid
Inlet
Process Fluid
Outlet
Utility
Fluid
Inlet
Utility
Fluid
Outlet
TE
101
TI
101
FE
101
FI
101
TE
102
TE
104
TI
102
TI
104
XC-101 Calculates the Overall Heat Transfer Coefficient. This
Coefficient is Compared to an Externally Determined Minimum
Acceptable Value. When the Value Drops Below this Value, an
Alarm, XA-104, is Activated to Alert the Operator to Take the
Heat Exchanger Out of Service for Cleaning.
XA
104
XC
101
XC
101
XA
104
A MISO to calculate the overall heat transfer coefficient
16. Fuzzy Logic Controllers
16
• Used when an exact condition isn’t necessary
• Also known as a decision tree controller
• Inputs are evaluated against one or more criteria
using inequalities
17. Example: Fuzzy Logic Controllers
17
• Example:
• The value of a reference parameter, MV1, is used to determine
how the controller will address the error between MV2 and
MV2,sp.
• If MV1 is greater than its setpoint, then 50% of the error between MV2
and its setpoint will be used in the control algorithm to calculate the
output from the controller to the control variable.
• If MV1 is less than or equal to its setpoint, then 25% of the negative
value of the error between MV2 and its setpoint will be used in the
control algorithm.
1. if MV1<MV1,sp go to 4
2. e(t) = 0.5*(MV2-MV2,sp)
3. go to 5
4. e(t) = -0.25*( MV2-MV2,sp)
5. CVo=f(e)
18. Fuzzy Logic Controller Applications
18
• Most commonly used in manufacturing
applications where multiple replicates of the
same item are processes
• No two replicates are exactly the same, so
inequalities based on tolerances are the best
way to do control
19. Consider this scenario…
19
Fig 6.5
• Reactant A & B react to form Product C
• Reactants are pre-treated to an ideal (hot) temperature
• Exothermic reaction is regulated w/ BFW utility stream
20. What are the ideal/optimum operating
conditions?
20
As the engineer, or experimentalist, what are trying to define or control in order to get the desired product to yield?
21. What are the ideal/optimum operation
conditions?
21
Fig 6.5
As engineers, we collect or
generate data… to determine
the best operating condition.
• Reaction Temperature
• Concentration of reactants
• Residence time
• Reaction Pressure
22. What are the ideal/optimum operation
conditions?
22
Fig 6.5
• Reaction Temperature
• Concentration of reactants
• Residence time
• Reaction Pressure
23. What are the ideal/optimum operation
conditions?
23
Fig 6.5
• Reaction Temperature
• Concentration of reactants
• Residence time
• Reaction Pressure
24. MIMO APCs
24
MIMO: multiple input/multiple output
• Simple MIMO APCs may be developed in-house for specific applications
• Example: generate the optimum conditions for all of the CVs of a
reaction system based on a 4-D plot of the key parameters
Now, can use this data (with MIMO-
APC) to specify the best position or
setting for the CVs in the process
25. Consider this scenario…
25
• Removing solute from feed gas
via liquid solvent absorption
process
Solute concentration
defines steam rate to
reboiler
Solute concentration defines
flowrate of solvent to
absorber
The concentration of the solvent is
based on a manually defined setpoint
and is based on vendor recommendation
or lab data
26. What we considered vs reality…
26
What we considered and designed regulatory
controls for…
Other required and critical considerations for
facility/plant operating success…
• Solvent:Solute Ratio
• Gas purity/recovery product
• Solvent regeneration/recovery
• Utility use
• Cost of regeneration
• Rate of solvent degradation due to regen process
• Pumping costs in transfer pump
• Trim cooler costs
• Loss of solvent in solute-lean gas outlet and
recovered solute gas streams
• Solvent chemical concentration in lean solvent
stream (within specified range)
27. What we considered vs reality…
27
Other required and critical considerations for
facility/plant operating success…
• Cost of regeneration
• Rate of solvent degradation due to regen process
• Pumping costs in transfer pump
• Trim cooler costs
• Loss of solvent in solute-lean gas outlet and
recovered solute gas streams
• Solvent chemical concentration in lean solvent
stream (within specified range)
How do we account for all of these
components in the control
scheme?!?
MVC Control
28. 28
• MVC: multi-variable control
• Also known as Model Predictive Control (MPC) or Dynamic Matrix
Control (DMC)
• An empirical (or semi-empirical model-based) control system
that assess the dynamics of all relevant variables simultaneously
• Usually created by a specialty vendor and implemented for the
specific application
MVCs
29. 29
Parts of an MVC
An Objective
Function
• May be economic or performance based
• Uses the Process and Controller modules to determine the optimum values for the CVs that maximizes
the value of the objective function
• A vector of ESPs for the controllers is generated and used to adjust the regulatory control layer
Process/
Predictor
Module
• Process-based models similar to process simulators
• Dynamic models that replicate the behavior of the process and its regulatory control system
• Purely empirical mathematical matrix-based models
Controller
Module
• May simulate the regulatory control system using PID algorithms, empirical algorithms or a combination
of both
Feedback/
Residual
Module
• Compares actual values to predicted values to determine if the prediction is accurate
• Generates an error vector that is then used to improve the accuracy of the predictor module
30. 30
Benefits of an MVC
• The MVC calculates every step, at the specific time
resolution, to go from the current state to the optimum
state.
• The optimum changes at each step to get there are also
defined
• This is known as the event or prediction horizon
• The steps themselves are known as the control horizon
• Only a portion of the entire horizon is used before the MVC
recalculates the optimum
31. 31
Development of an MVC
• The most common strategy is the mathematical matrix-
based process model with an empirical control model.
1. The designer specifies all of the MVs, CVs, and DVs
and inputs their regulatory tags into the MVC database
• The system can be over specified – MVs ≠ CVs
• The MVC will build internal dependencies to generate pseudo
variables that satisfy the math equations
32. 32
Development of an MVC
2. After the regulatory controls are commissioned, the MVC is
put into learning/tuning mode.
3. The MVC makes small (<5%) perturbations to the CVs and
DVs and evaluates the MV responses
4. These data are used to determine the tuning/weighting
factors used in the simulated adaptive controllers in the
MVC
5. These changes are made until the dynamics are fully
explored.
33. 33
Development of an MVC
• The learning mode can require thousands of tests and
takes a few weeks to complete
• The MVC must be able to operate if a MV or CV is out of
service.
• Most MVC’s are coupled with a data reconciliation
program to help with this
Important Notes
34. 34
Data Reconciliation
• Used to
• identify bad measurements automatically,
• predict the true value of a bad measurement,
• and allow generation of mass and energy balances
• Can be used by both regulatory and supervisory layers
• Can be used by optimizers and data generation
programs
35. 35
Benefits of Data Reconciliation
• When included with a fully integrated PAS, maintenance
work tickets can be automatically generated for repair
of bad instruments.
• By replacing bad measurement values with predictions,
unit data reconciliators allow supervisory controls to
remain on-line even when one or more measurement(s)
is/are not available.
36. 36
Features of Data Reconciliation
Two basic features:
1. Comparing current MVs to historical values
• If there is too great of a difference, an alarm can be activated and a
substitute value can be recommended based on historical data
• The operator decides whether to use the original or substitute values
2. Calculate mass and energy balances
• Requires over specification of instruments to close the balance
• By over specifying the measurements, the system can determine if one of
the measurements is “bad” and activate an alarm. A substitute value
based on the mass/energy balance can be recommended.
• If the system isn’t over specified, the system may be able to run an off-line
model of the system to provide a substitute value
37. 37
Optimizers
• Used to set process objective targets for Supervisory control
strategies
• Plant wide optimizers manipulate key selected controllers
and control strategies across the entire plant
• Unit optimizers focus on a specific process unit or complex
suite of unit operations
• Optimizers are often closely coupled with multi-variable
controllers.
38. 38
Unit Optimizers
• Unit optimizers are best justified for processes that change
process objectives frequently
• The optimizer can be used to determine the most efficient
path to move from one set of operating objectives to
another
• The optimizer can also help to efficiently recover from a
major process upset and can help to optimize catalyst life,
minimize energy consumption, and minimize environmental
impacts
40. RTO Scripts
A typical optimizer script might look something like this:
Obj(T) = a1Ob1(T) +a2Ob2+ . . .
Ob1(T) = maximize(FC-100 * ProdVal1 ) [example of a revenue]
Ob2(T) = minimize(FC-200 * HPStmVal ) [example of a cost]
.
.
.
C(1,t) = AC-100 .LE. Prod1HKConc
C(2,t) = SC-303 .GT. Pmp1Low .and. SC-303.LE.Pmp1High
.
.
[examples of constraints]
41. RTO Scripts
Provide the specification information necessary to program a
real-time optimization controller script for the mixer/settler
system shown below.
42. Planning and Scheduling Programs
Planning: Used to determine the overall inputs and outputs
of the process facility
Inputs: raw material supply schedules/inventories
Outputs: product delivery commitments/inventories
Cost Data: utility, raw materials, products, chemicals, other
operating cost data
Constraints: emission limits, capacity limits
43. Planning and Scheduling Programs
Planning: Used to determine the overall inputs and
outputs of the process facility
The program provides information used to:
• Schedule shipments to/from the plant
• Predict inventory
• Identify the target quantities of products and wastes
generated
• Identify the predicted quantities of raw materials, utilities,
and chemicals used
44. Planning and Scheduling Programs
Scheduling: Used to determine the throughput for
each process unit in order to meet the plant product
targets
These two programs can be used separately, but are
often coupled together in a single coordinated system.
45. Enterprise and Supply Chain Management
ERP’s are the corporate level extension of the Plant’s
planning and scheduling system.
Also known as business management systems.
ERP’s are often used to provide centralized accounting
functions and resource allocation.
46. Higher Level Automation
of Batch Sequences
• Not as well developed as for continuous processes
• May need to custom design APC and higher
applications
• Easier to implement for cyclical batch sequences since
information from multiple iterations of the sequence can be
used for optimization
47. Higher Level Automation
of Batch Sequences
• Can often apply MISO and MIMO type control
concepts to regulatory control performed during
the individual steps within the overall sequence.
Example: Optimize the flow and temperature of a hot inert gas used
to regenerate a fixed bed solvent or sorbent to minimize
regeneration costs. These conditions may change as the bed ages
and the length of time the bed is in process mode decreases.