Control Using Two Manipulated Parameters
Terry Blevins – Principal Technologist and Greg McMillan – Principal Consultant
Presenters
§ Terry Blevins § Greg McMillan
Introduction § Overview – Typical Examples § Split-Range Control § § §
Concept, variations in implementation Setup in field vs. Splitter Block and IO for each valve. Using Splitter Block, Example.
§ § §
Concept and typical implementation Setup of I-only control in implementation Impact of mode/status, Example.
§ §
How to implement in DeltaV Example
§ § §
Advantage if process has large deadtime, difference in dynamics Setup of MPC and MPC-Pro Blocks Example Applications
§ Valve Position Control
§ Combining Split Range and Valve Position Control § Utilizing Predict/PredictPro for Control Using Two Manipulated Parameters
§ Summary § References
Control Using Two Manipulated Parameters
One(1) Controlled Parameter
Controller SP Unmeasured Disturbance
§ Under specified problem that has multiple solutions for unlimited operation. Two(2) Manipulated § Extra degree of freedom Parameters is used to achieve unique solution that satisfied specific control objective. § Most common techniques are: split Process range, valve position control. § Combination of these technique and MPC offer new capability to address this class of problems
Split Range – Traditional Implementation Temperature Example
TIC 101
4-20ma 3-15PSI
§ §
IP 101
§
A/C
Cooling A/O
Process
TT 101
Heating 100 Valve Position (% of Span) 0
§ § § §
Cooling Heating 3
IP Output ( PSI )
15
§
Sequencing of valve accomplished through calibration of positioner, selection of actuator (A/O or A/C) Pro – Less expensive installation (1 pair of wires to field and 1 I/P) Con – You are not using the best technology for valve performance (e.g. digital positioners). Con -Difficult to initially calibrate and continuously improve to get best gap and most constant gain. Con -Individual valves not accessible for trouble shooting loop and actuator/valve problem. Con – The actuator, pneumatic positioner, and I/P performance shift with time and field conditions Con – I/P failure disables 2 valves Con - Replacements in the night may not have the special settings
Split Range – Traditional Implementation pH Example
AIC 102
4-20ma 3-15PSI
IP 102
§ § §
PS 102
§ pH
A/O
A/O
Fine Valve
Process
AT 102
§ §
Coarse Valve §
100 Valve Position (% of Span)
Fine Valve Coarse Valve
0 3
I/P Output ( PSI ) 15
§
Sequencing of fine and coarse valve requires pressure switch, two solenoid valves and associated wiring and tubing Con – Complex installation Con – You are not using the best technology for valve performance (e.g. digital positioners). Con -Difficult to initially calibrate and continuously improve to get best gap and most constant gain. Con -Individual valves not accessible for trouble shooting loop and actuator/valve problem. Con – The switch, actuator, pneumatic positioner, and I/P performance shift with time and field conditions Con – I/P failure disables 2 valves Con - Replacements in the night may not have the special settings
Split Range – DeltaV Implementation AI
PID
SPLT
AO AO
AI
PID
SPLT
AO AO
§ Splitter bock is used to implement split range control. § When using traditional valves, split range control may implemented in DeltaV Controller using two(2) current outputs § Split range control may be partially or fully assigned to fieldbus devices.
Split Range Control in DeltaV
Splitter Block Calculation
IN_ARRAY Parameter § The SP range associated with each output is defined by IN_ARRAY. § SP range of outputs may be defined to overlap SP range associated with OUT1 SP range associated with OUT2
§ The SP upper end of range must be greater that lower end of range for each output
OUT_ARRAY Parameter § When SP is outside defined range, then the value at the end of range is used to determine the output.
OUT1 Range for associated SP range
§ LOCKVAL determines if OUT1 value is held if SP is greater that the upper end of range defined for OUT1. § No restrictions are placed on the output range.
Splitter Block OUT_1
100
0
OUT_2
OUT_ARRAY 0 100 0 100 IN_ARRAY 0 100 0 100
0
100 OUT_ARRAY 100 0 0 100 IN_ARRAY 0 40 35 100
100
0
100
0 100 LOCK_VAL “holds ”
HYSTVAL 0
LOCK_VAL “is zero ”
0
100
SP
OUT_ARRAY 0 100 0 100 IN_ARRAY 0 40 35 100
Heating-Cooing Example AI
PID
TT103
TIC103
SPLT FY103
AO IP103A
AO TIC 103 FY 103 IP 103A
HEATER
IP 103B COOLER
TT 103
IP103B
Split Range Output (FY103)
100
Valve Position (% of Span)
Cooling (IP103B) Heating (IP103A)
0 0
TIC103 Output (% of Span)
100
Steam Header Example AI
PID
PT104
PIC104
SPLT FY104
AO IP104A
AO IP104B
1475# Header
Boiler IP 104A
FY 104 IP 104B
400# Header
PIC 104
PT 104
Turbo Generato r
Split Range Output (FY104) - Capacity
100
Valve Position (% of Span)
Valve 104A Valve 104B
0 0
PIC104 Output (% of Span)
100
Basic Neutralizer Example AI
PID
AT105
AIC105
SPLT FY105
AO IP105A
AO IP105B
Reagent
IP 105A
Coarse Valve Fine Valve
IP 105B
FY 105 AIC 105
Neutralizer
AT 105
Discharge
pH Nonlinearity and Sensitivity pH
8 6
Reagent Flow Influent Flow
Split Range Output – Valve Sequencing
100
HYSTVAL Valve Position (% of Span)
Fine Valve (IP105B) Coarse Valve (IP105A)
0 0
AIC105 Output (% of Span)
100
Calculating Splitter SP Ranges Example: Steam flow to Header, splitter interfacing directly to PRV’s, no overlap Valve 1 rating = 50kph Valve2 rating = 150kph Desired Splitter Span valve 1 = 100*(50/ (150+50)) = 25% SP range for valve 1 = 0-25% SP range for valve 2 = 25-100%
§ A 1% change in controller output to the splitter should have the same impact on control parameter when operating with either valve. § When manipulating the same or similar material e.g. steam flow to header, then the range may be calculated based on valve rating. § Tests may be performed to determine impact of each valve on the controlled parameter.
Testing Process to Determine Splitter SP Ranges Example: Slaker feed temperature controlled using heating and cooling valves Controlled Temperature
Heating
1.1degF
1%
Cooling
1% Time
Desired Splitter Span cooling valve = 100*(1.1/(1.1+2.2)) = 33% SP range for cooling valve = 0-33% SP range for heating valve = 33-100%
2.2degF
§ With the process at steady state and AO’s in Auto mode, determine the magnitude of change in the controlled parameter for a 1 percent change in each valve. § Calculate the splitter SP span and range for each output based on the observed response
Example – Split Range
Response to SP Change – Split Range Output To Large Valve/Small Valve
SP Small Valve
PID OUT PV
Large Valve
Split Range – Strengths and Weaknesses § Pro - Process operation in simplified since two actuators are treated as one control manipulated parameter. § Pro – immediate change in target actuator position can be achieved over the entire operating range independent of the size of change in the splitter SP § Con – To achieve stable control over the entire operating range, Controller tuning must be established based on the slower responding manipulated parameter. § Con- Does not take advantage of difference in resolution of actuator e.g. fine vs. coarse valve. § Valve position control may be used in place of split range control when there are differences in dynamic response or resolution in actuators.
Valve Position Control – Traditional Implementation pH Example
AIC 106
Mode Target Valve Position
IP 106A ZC 106
IP 106B
I-Only Controller
A/O Coarse Valve
§ PID control is implemented using the actuator with finer resolution or fastest impact on controlled parameter Fine Valve
Process
AT 106
pH Fine Valve
Coarse Valve Time
§ The actuator with coarse resolution or slower impact on the controlled parameter is adjusted by an I-only controller to maintain the long term output of the PID controller at a given target § I-Only controller must be disabled when the PID controller is not in an Automatic mode. Target Valve Position
Valve Position Control – DeltaV Implementation AI
PID
AO I-Only
AI
PID
Fieldbus devices
Traditional field devices
AO I-Only
AO
AO
§ I-Only control is achieved by configuration of the PID Block STRUCTURE, GAIN and RESET parameters. § It is possible to implement valve position control in the DeltaV controller or for this function to be distributed to fieldbus devices.
Valve Position Control in DeltaV § Actuator with fastest impact or highest resolution is used to maintain the controlled parameter at setpoint. § The OUT of the PID used for control is wired to IN on the PID block used for IOnly regulation of slower responding or coarse resolution. PID configured for I-Only control
Configuring PID for I-Only Control § The STRUCTURE parameter should be configured for “I action on Error, D action on PV” § The GAIN should be set to 1 to allow normal tuning of RESET (even though proportional action is not implemented. § RESET should be set significantly slower than that the product of the PID gain and reset time used for control e.g. 5X slower
Precise Flow Using Big/Small Valve AI
PID
FT107
FIC107
AO IP107A
I-Only ZC107
IP 107A
FIC 107 FT 107 ZC 107
IP 107B
AO IP107B
Neutralizer Using Valve Position Control AI
PID
AT108
AIC108
AO IP108A
IZC108 Only Reagent
IP 108B
Coarse Valve Fine Valve
IP 108A
AO IP108B
ZC 108
AIC 108 Neutralizer
AT 108
Discharge
Example -Boiler BTU Demand AI FT109B
AI
SUM
PID
FY109
FIC109
AO IP109A
IZC109 Only
FT109A
BTU Demand FY 109
FIC 109
IP 109A
FT 109B
FT 109A
HI BTU Fuel ZC 109
IP 109B
Low BTU – Waste Fuel
Boiler
AO IP109B
Example –Reformer Air Demand AI
PID
FT110
FIC110
AO IP110
ZC 110 SC 110
Air Machine
I-Only
AO
ZC110
SC110
FIC 110
Total Air Demand
IP 110 FT 110
Secondary Reformer
Example – Valve Position Control
Response to SP Change - Valve Position Control with Large Valve/Small Valve § Target position for fine valve is 30%. PV SP Coarse Valve
Limite d
Fine Valve
§ When the fine valve saturates, then response is limited to be reset of the I-Only control
Valve Position Control – Strengths and Weaknesses § Pro – Immediate control response is based on actuator with finest resolution and/or faster impact on controlled parameter. § Pro – Actuator with coarse resolution or slower impact on controlled parameter is automatically adjusted to maintain the output of the controller output long term at a specified operating point. § Con – The controller output may become limited in response to a large disturbance or setpoint change. For this case, the dynamic response becomes limited by the slower tuning of the I-only controller. § Con – Since stick-slip or resolution limits are a % of stroke, the big valve will go into a slow limit cycle § The features of split range control and valve position control may be combined to provide immediate response to large changes in demand while retaining the features of valve position control for normal changes.
Combining the Best Features of Split Range and Valve Position Control § A composite Block can be created that combines the features of split range and valve position control § Support for BKCAL_IN and BKCAL_OUT can be implemented to provide bumpless transfer
Composite Algorithm NORMAL SP
CAS_IN
FILTER_TC
T
-
+
x
OUT_1
Filter
-
x x
MODE
RANGE SPAN
OUT_2
Scaling
BKCAL_IN1 Balance Calculation BKCAL_IN2
BKCAL_OUT
Composite Implementation §
§
Parameters that must be configure are: FILTER_TC, SPAN (of SP), RANGE (of OUT1), and NORMAL (desired position of) The FILTER_TC should be configured similar to the reset time of the I-Only Controller that would be used for valve position control.
Demo – Composite Combining Valve Position and Split Range Control
Example: Response to SP Change § Small change
Large change
SP, PV OUT of PID
Fine Valve Coarse Valve
§
For small changes in SP or load disturbance, the response is similar to that provided by valve position control For large changes in SP or load disturbance, the immediate response is similar to split range control
Composite for Valve Position/Split Range Control – Strengths and Weaknesses § Pro – All the advantage of valve position control without the dynamic limitations on large setpoint change or load disturbance. § Con – If there is a significant delay in the control parameter response to changes in the two valves, then this limits the response that can be achieved using PID for the control . § Model Predictive control automatically compensates for process dynamic and may be configured to provide the best features of valve position and split range control and can also address operating constraints.
Example of Different Dynamic Response – Waste Fuel Boiler Control Steam Flow
Constraints ooo
MPC
Bark
Gas
Steam SP Steam Flow
20 Lo Cost Slow Waste Bark Hi Cost Fast Fuel Gas
Desired Response to unmeasured disturbance
§ Objective: Maximize use of bark, only use gas when required to maintain Steam SP. § Steam response to change in bark is much slower than for a change in gas. § Bark alone may not be sufficient to address a sudden increase in steam demand.
Example of Different Dynamic Response – Bleach Plant Control AT
MPC Hi Cost Fast Chemical 2 20 minutes
Lo Cost Slow Chemical 1
60 minutes
KAPPA SP KAPPA Lo Cost Slow Chemical 1 Hi Cost Fast Chemical 2
Desired response to unmeasured disturbance
§ Objective: Maintain KAPPA target though the addition of Chemical 1 and Chemical 2. Minimize the use of Chemical 2. § Desired operation is for Chemical 2 to be used for short term correction in KAPPA to replace Chemical 2 with Chemical 1 in the longer term.
Utilizing MPC for Control § Both Predict and PredictPro can be configured and tuned for maintaining the critical controlled variable (CV), such as steam or composition, at its target and maximizing the low cost slow MV set point as an optimization variable. manipulated variables
Maximize
optimization controlled variable variable
MPC
High Cost Fast Feed SP
Critical PV (normal PE)
Low Cost Slow Feed SP (lowered PE)
null
Low Cost Slow Feed SP
MPC Guidelines for This Application § The best load and set point response for the critical CV is obtained with a short term tradeoff in efficiency by reducing the penalty on error (PE) for the optimization variable. § When riding the low cost MV maximum set point, this PE lets both the slow and fast MV to move to improve the load and set point response of the critical CV. § When riding the high cost MV low set point limit, it does not slow down the response of the other MV to upsets and set point changes to the critical CV. Only the response of the optimization variable is slowed down. This is consistent with the general theme that disturbance rejection must be fast while optimization can be slow. § For coarse and fine valve control, the small valve is a low cost (low stickslip) fast MV and the big valve is a high cost (high stick-slip) slow MV. The optimization variable is fine valve set point with a strategy of keeping it within limits (mid range throttle position). The PE for the optimization variable is reduced rather then the PM increased for the coarse valve so that both are available for load disturbance rejection. § For the following examples, the slow MV has a lower cost, so its optimization strategy is maximization.
DeltaV Predict Configuration § MPC block should be configured for two control and two manipulate parameters. § The controlled measurement is wired to CNTRL1
DeltaV Predict Configuration (Cont)
§ CNTRL2 is configured as an optimized parameter Maximize (not wired)
Control Generation - DeltaV Predict §
§
In Predict, the Penalty on Error (PE) is significantly decreased on the “Controller Generation” screen as shown in this example. The PE was lowered form 1.0 to 0.1 to make the optimization of the slow MV much less important than the control of the critical PV at its target
MPC Response to Disturbance and Set Point Changes §
Riding Max SP on Lo Cost MV Load Upsets
Riding Min SP on Hi Cost MV
Critical CV
Set Point Changes
Load Upsets Low Cost MV Maximum SP Increased
Critical CV
Set Point Changes
Low Cost MV Maximum SP Decreased
Lo Cost Slow MV Hi Cost Fast MV
§
In this example, low cost MV initially is riding its maximum set point, which leaves the fast cost MV free to respond Later, the maximum for the low cost MV has been increased to the point where it is no longer achievable, which drives the high cost MV to its low set point limit.
DeltaV PredictPro Configuration §
When configuring the MPC-Pro block, selects “Target” in the optimize column for the critical PV, and “Maximize” for the low cost MV.
§
Browse to specify the RCAS_IN of the low cost slow MV (FC1-2) to specify the measurement associated with the low cost slow MV..
Control Parameter - MPC-Pro Block
Control Generation - DeltaV PredictPro §
§
The Penalty on Error (PE) is significantly decreased on the “Controller Generation” screen In this example, the PE was lowered form 1.0 to 0.1 to make the optimization of the slow MV much less important than the control of the critical PV at its target.
MPC-Pro Response to Disturbance and Set Point Changes
Riding Max SP on Lo Cost MV Load Upsets
Critical CV
Set Point Changes
§
In this example, the low cost MV initially is riding its maximum set point, which leaves the fast cost MV free to respond
§
Later, the maximum for the low cost MV has been increased so it is no longer achievable, which drives the high cost MV to its low set point limit.
Riding Min SP on Hi Cost MV Load Upsets
Low Cost MV Maximum SP Increased
Critical CV
Set Point Changes
Low Cost MV Maximum SP Decreased
Lo Cost Slow MV
Hi Cost Fast MV
Summary § Split range control allows fully dynamic response to major setpoint of load disturbance changes. Valve position control may be used to takes advantage of any difference in control response or resolution in the manipulated parameters. A composite block has been demonstrated that combines the best features of split range and valve position control. § DeltaV Predict and PredictPro and the associated MPC and MPC-Pro blocks may be effective used to address control using two manipulated parameters. Improved performance over PID is expected if the process has significant dead time or the manipulated variables have significantly different dynamics. Also, using this approach allow operating constraints and feedforward to be easily incorporated into the control strategy. § Please direct questions or comments on this presentation to Terry Blevins (
[email protected]) or Greg McMillan (
[email protected] ).
Where To Get More Information § “Effectively Addressing Control Applications”, Terry Blevins, Emerson Exchange, 2004. § “Addressing Multi-variable Process Control Applications”, Dirk Thiele, Willy Wojsznis, Pete Sharpe, Emerson Exchange, 2004 § “Advanced Control Unleashed, Plant Performance Management for Optimum Benefit”. Terry Blevins, Gregory McMillan, Willy Wojsznis, Mike Brown, ISA Publication, ISBN 1-55617-815-8, 2003.