0
Effects of Macrocycle Time and Sampling Rates on Control Loop Performance Dan Daugherty – Sr. Engineer – Product Engineeri...
Presenters <ul><li>Dan Daugherty </li></ul><ul><li>Mark Coughran </li></ul><ul><li>Ferrill Ford </li></ul>
Why <ul><li>F OUNDATION  Fieldbus perceived as slow by some </li></ul><ul><li>Control Response specifications by end user ...
What <ul><li>Safe place to do a controlled test on a real process </li></ul><ul><li>Availability of both F OUNDATION  Fiel...
Lab setup for hydraulic pressure control <ul><ul><li>Fluid process dynamics are negligible </li></ul></ul><ul><ul><li>Sign...
3 rd  Loop – Marshalltown Flow Lab PV PT Disturbance EnTech Toolkit
Timing – 4-20mA PID D/A Conversion DVC  4-20/HART Pneumatic Actuator DVC dead time and  time constant Load Valve Motion Hy...
Timing – FF CIF FF PID FF AO   Pneumatic Actuator DVC dead time and  time constant Load Valve Motion Hydraulic Pressure (P...
Control Response Period by subtraction 4-20 mA / HART 0.05 sec Load Valve Motion Hydraulic Pressure (Process) 3051C Dead T...
Control Response Period by subtraction Foundation Fieldbus Control-In-the-Field (CIF) 0.10 sec Load Valve Motion Hydraulic...
Load step tests for Control Response Period <ul><li>Step the output to the load valve </li></ul><ul><li>The PID control lo...
Sample Control Response Period measurement CIC, module execution = 1.0, macrocycle = 0.5 1.37 – 0.10 – 0.07 = 1.20 seconds
Sample Control Response Period measurement 4-20 mA, module execution = 0.2 0.30 – 0.05 – 0.07 = 0.18 seconds
Sample histogram from 21 measurements CIC, module execution = 1.0, macrocycle = 0.5 Mean value of raw dead time = 1.39 sec...
Control Response Period results overview 4-20 mA, DeltaV Control in DVC (CIF) Control in DeltaV (CIC) 2:1 Control in Delta...
Lambda Tuning for self-regulating processes <ul><li>Closed Loop (Auto) </li></ul><ul><ul><li>No oscillation </li></ul></ul...
Lambda Tuning for self-regulating process sample Manual step 5% on controller output
Average process dynamics and recommended tuning
Controller tuning philosophy <ul><li>Only needed for sine wave load disturbance and setpoint response tests </li></ul><ul>...
Theoretical setpoint step response
Theoretical load frequency response
Load Frequency Response Tests—Introduction and Notation <ul><li>Sinusoidal output to the load valve </li></ul><ul><li>Most...
Load Frequency Response, period 100, CIC, module execution = 1.0, macrocycle = 1.0 AR = 0.41
Load Frequency Response, period 100, CIC, module execution = 0.5, macrocycle = 0.5 AR = 0.26
Load Frequency Response, period 100, CIC, module execution = 1.0, macrocycle = 0.5 AR = 0.38
Load Frequency Response, period 100, CIF, macrocycle = 0.15 AR = 0.18
What if 8 loops on the FF segment? CIC (DeltaV) theoretical
What if 8 loops on the FF segment? CIF (DVC) theoretical
Conclusions with more loops on the segment <ul><li>Shows even more reason to use CIF </li></ul><ul><li>CIF should be fast ...
Business Results Achieved <ul><li>Density on Fieldbus segments </li></ul><ul><li>Identifying latency ‘opportunities’ </li>...
Acknowledgements <ul><li>In the Marshalltown lab, thanks to </li></ul><ul><ul><li>Rick Osborne </li></ul></ul><ul><ul><li>...
Summary <ul><li>Foundation Fieldbus Control-In-the-Field </li></ul><ul><ul><li>proved Control Response Period equal to mac...
Where To Get More Information <ul><li>[email_address] </li></ul><ul><li>Ferrill. [email_address] </li></ul><ul><li>[email_...
Appendix—Setpoint Step Response
Setpoint Step Tests—Introduction and Notation <ul><li>Timing of the setpoint steps was not automated </li></ul><ul><li>Sam...
Setpoint step test sample data
Setpoint step test sample data
Setpoint step test sample data
Setpoint step test conclusions <ul><li>Did not attempt to optimize PID tuning for each case </li></ul><ul><li>All SP respo...
Upcoming SlideShare
Loading in...5
×

Effects of Macrocycle Time and Sampling Rates on Control Loop Performance

3,869

Published on

Emerson Exchange 2009 presentation by consultants Dan Daugherty, Ferrill Ford, and Mark Coughran.

Published in: Technology, Business
0 Comments
3 Likes
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total Views
3,869
On Slideshare
0
From Embeds
0
Number of Embeds
3
Actions
Shares
0
Downloads
84
Comments
0
Likes
3
Embeds 0
No embeds

No notes for slide
  • Unlike older tuning methods such as Ziegler-Nichols, Lambda tuning gives a smooth non-oscillatory response. But equally important is the ability to design Lambda for the loop requirements. Lambda can be selected based on the performance requirements of the particular loop to separate the dynamics of interacting loops to establish fast vs. slow for inner vs. outer (slave vs. master) loops The testing that we use to determine the process dynamics, required for Lambda tuning, also identifies problems with the control equipment.
  • Unlike older tuning methods such as Ziegler-Nichols, Lambda tuning gives a smooth non-oscillatory response. But equally important is the ability to design Lambda for the loop requirements. Lambda can be selected based on the performance requirements of the particular loop to separate the dynamics of interacting loops to establish fast vs. slow for inner vs. outer (slave vs. master) loops The testing that we use to determine the process dynamics, required for Lambda tuning, also identifies problems with the control equipment.
  • Add experimental data to support the Bode plot
  • Transcript of "Effects of Macrocycle Time and Sampling Rates on Control Loop Performance"

    1. 1. Effects of Macrocycle Time and Sampling Rates on Control Loop Performance Dan Daugherty – Sr. Engineer – Product Engineering Ferrill Ford – Sr. Engineer – Product Engineering Mark Coughran – Sr. Industry Consultant – Industry Solutions Group
    2. 2. Presenters <ul><li>Dan Daugherty </li></ul><ul><li>Mark Coughran </li></ul><ul><li>Ferrill Ford </li></ul>
    3. 3. Why <ul><li>F OUNDATION Fieldbus perceived as slow by some </li></ul><ul><li>Control Response specifications by end user or process licensor </li></ul><ul><li>Lack of actual field data </li></ul><ul><li>Questionable recommendation for oversampling (module execution = 2x macrocycle) </li></ul>
    4. 4. What <ul><li>Safe place to do a controlled test on a real process </li></ul><ul><li>Availability of both F OUNDATION Fieldbus and 4-20 mA loops </li></ul><ul><li>Ability to test </li></ul><ul><ul><li>Control Response period </li></ul></ul><ul><ul><li>load frequency response </li></ul></ul><ul><ul><li>setpoint step response </li></ul></ul>
    5. 5. Lab setup for hydraulic pressure control <ul><ul><li>Fluid process dynamics are negligible </li></ul></ul><ul><ul><li>Significant dynamics are in the sensor/transmitter, control valve, controller, communications </li></ul></ul><ul><ul><li>Control valve first with DVC6010f, then DVC6010 </li></ul></ul><ul><ul><li>PT FF, 4-20 were Rosemount 3051C </li></ul></ul><ul><ul><li>PT FAST were Toolkit, 100 Hz </li></ul></ul><ul><ul><li>All signals recorded with Emerson’s EnTech ™ Toolkit </li></ul></ul>
    6. 6. 3 rd Loop – Marshalltown Flow Lab PV PT Disturbance EnTech Toolkit
    7. 7. Timing – 4-20mA PID D/A Conversion DVC 4-20/HART Pneumatic Actuator DVC dead time and time constant Load Valve Motion Hydraulic Pressure (Process) Change 3051 4-20/HART output 3051C Dead Time and Time Constant A/D Conversion
    8. 8. Timing – FF CIF FF PID FF AO Pneumatic Actuator DVC dead time and time constant Load Valve Motion Hydraulic Pressure (Process) Change 3051 FF AI 3051C Dead Time and Time Constant FF Compel Data
    9. 9. Control Response Period by subtraction 4-20 mA / HART 0.05 sec Load Valve Motion Hydraulic Pressure (Process) 3051C Dead Time and Time Constant 3051C 4-20 output PID A/D DVC 4-20 input D/A DVC6000 Dead Time and Time Constant Pneumatic Actuator Fast Reference Pressure Sensor 0-750 psig Fast Reference Pressure Sensor 0-50 psig Control Response Period Typical Customer Spec. 0.07 sec Measured Loop Dead Time In Load Step Test
    10. 10. Control Response Period by subtraction Foundation Fieldbus Control-In-the-Field (CIF) 0.10 sec Load Valve Motion Hydraulic Pressure (Process) 3051 Dead Time and Time Constant 3051 FF AI FF PID FF Compel Data FF AO DVC6000f Dead Time and Time Constant Pneumatic Actuator Fast Reference Pressure Sensor 0-750 psig Fast Reference Pressure Sensor 0-50 psig Control Response Period Typical Customer Spec. 0.07 sec Measured Loop Dead Time In Load Step Test
    11. 11. Load step tests for Control Response Period <ul><li>Step the output to the load valve </li></ul><ul><li>The PID control loop approximates proportional-only action </li></ul><ul><ul><li>Gain = 0.5 </li></ul></ul><ul><ul><li>Reset = 100000 </li></ul></ul><ul><li>Fit the responses in Emerson’s EnTech ™ Toolkit </li></ul><ul><li>Only the “dead time” part of the measurement is significant </li></ul><ul><li>Subtract the response times of transmitter and control valve that are not defined as part of Control Response Period </li></ul><ul><li>Average the results from at least 10 measurements </li></ul>
    12. 12. Sample Control Response Period measurement CIC, module execution = 1.0, macrocycle = 0.5 1.37 – 0.10 – 0.07 = 1.20 seconds
    13. 13. Sample Control Response Period measurement 4-20 mA, module execution = 0.2 0.30 – 0.05 – 0.07 = 0.18 seconds
    14. 14. Sample histogram from 21 measurements CIC, module execution = 1.0, macrocycle = 0.5 Mean value of raw dead time = 1.39 seconds Corrected value (Control Response Period) = 1.22 seconds
    15. 15. Control Response Period results overview 4-20 mA, DeltaV Control in DVC (CIF) Control in DeltaV (CIC) 2:1 Control in DeltaV (CIC) 4:1 Control in DeltaV (CIC) 1:1 Ratio for Fieldbus Control in DeltaV is Module Execution : Macrocycle
    16. 16. Lambda Tuning for self-regulating processes <ul><li>Closed Loop (Auto) </li></ul><ul><ul><li>No oscillation </li></ul></ul><ul><ul><li> is the closed-loop time constant </li></ul></ul><ul><ul><li>Choose the speed </li></ul></ul><ul><li>Open loop (Manual) </li></ul><ul><ul><li> is the open-loop time constant </li></ul></ul>SETPOINT PV  63% 63% PV OUT 
    17. 17. Lambda Tuning for self-regulating process sample Manual step 5% on controller output
    18. 18. Average process dynamics and recommended tuning
    19. 19. Controller tuning philosophy <ul><li>Only needed for sine wave load disturbance and setpoint response tests </li></ul><ul><ul><li>Does not apply to a Control Response Period specification </li></ul></ul><ul><li>Lambda = 1.5 seconds is fast relative to typical tuning of flow and pressure loops in the field </li></ul><ul><li>Is based on fast controller execution </li></ul><ul><li>In principle, this should be changed (detuned) as we increase either module execution time or macrocycle </li></ul><ul><li>In practice, we didn’t have time to customize tuning for each combination of communication method, module execution, and macrocycle </li></ul>
    20. 20. Theoretical setpoint step response
    21. 21. Theoretical load frequency response
    22. 22. Load Frequency Response Tests—Introduction and Notation <ul><li>Sinusoidal output to the load valve </li></ul><ul><li>Most tests used disturbance period = 100 seconds </li></ul><ul><ul><li>This period gives the feedback loop a chance to attenuate a significant amount of the variability </li></ul></ul><ul><li>Same PID tuning for all: Gain = 0.35, Reset = 0.48 </li></ul><ul><li>CIC ≡ Fieldbus, DeltaV, DVC6000f </li></ul><ul><li>CIF ≡ Fieldbus, DVC6000f </li></ul><ul><li>Analog ≡ 4-20 mA, DeltaV, DVC6000 </li></ul><ul><li>AR ≡Amplitude Ratio </li></ul><ul><ul><li>Auto Amplitude / Manual Amplitude </li></ul></ul>
    23. 23. Load Frequency Response, period 100, CIC, module execution = 1.0, macrocycle = 1.0 AR = 0.41
    24. 24. Load Frequency Response, period 100, CIC, module execution = 0.5, macrocycle = 0.5 AR = 0.26
    25. 25. Load Frequency Response, period 100, CIC, module execution = 1.0, macrocycle = 0.5 AR = 0.38
    26. 26. Load Frequency Response, period 100, CIF, macrocycle = 0.15 AR = 0.18
    27. 27. What if 8 loops on the FF segment? CIC (DeltaV) theoretical
    28. 28. What if 8 loops on the FF segment? CIF (DVC) theoretical
    29. 29. Conclusions with more loops on the segment <ul><li>Shows even more reason to use CIF </li></ul><ul><li>CIF should be fast enough for nearly all loops in the plant </li></ul><ul><li>Exceptional loops already have dedicated controllers; e.g. surge control, compressor lube oil </li></ul><ul><ul><li>Even these applications can be handled in some cases with CIF; see Rezabek and Peluso, EGUE2008 </li></ul></ul>
    30. 30. Business Results Achieved <ul><li>Density on Fieldbus segments </li></ul><ul><li>Identifying latency ‘opportunities’ </li></ul><ul><li>Avoid slow responses </li></ul>
    31. 31. Acknowledgements <ul><li>In the Marshalltown lab, thanks to </li></ul><ul><ul><li>Rick Osborne </li></ul></ul><ul><ul><li>Mike Himes </li></ul></ul><ul><ul><li>Kyle Hokanson </li></ul></ul><ul><ul><li>Others </li></ul></ul><ul><li>Other Emerson sponsors </li></ul><ul><ul><li>Advanced Applied Technologies in PS&S </li></ul></ul>
    32. 32. Summary <ul><li>Foundation Fieldbus Control-In-the-Field </li></ul><ul><ul><li>proved Control Response Period equal to macrocycle </li></ul></ul><ul><ul><li>Can get 0.18 seconds, adequate for almost all loops </li></ul></ul><ul><li>Foundation Fieldbus Control-In-the-Controller/Host </li></ul><ul><ul><li>Control Response Period can be much greater than expected </li></ul></ul><ul><ul><li>C-I-C not recommended to get full benefit from Fieldbus </li></ul></ul><ul><ul><li>Oversampling (Module Execution>Macrocycle) did not show any benefit </li></ul></ul><ul><li>Your comments and questions are welcome </li></ul>
    33. 33. Where To Get More Information <ul><li>[email_address] </li></ul><ul><li>Ferrill. [email_address] </li></ul><ul><li>[email_address] </li></ul><ul><li>John Rezabek in Control Magazine (www.controlglobal.com); July 2008 “Ready for Control in the Field?”; November 2007 “Load ‘Em Up!” </li></ul><ul><li>John Rezabek and Marcos Peluso, EGUE2008, “Field- based control for compressor anti-surge” </li></ul><ul><li>Pang et al., “Analysis of control interval for foundation fieldbus-based control systems”, ISA Transactions , Volume 45, Number 3, July 2006, pages 447-458. </li></ul>
    34. 34. Appendix—Setpoint Step Response
    35. 35. Setpoint Step Tests—Introduction and Notation <ul><li>Timing of the setpoint steps was not automated </li></ul><ul><li>Same PID tuning for all: Gain = 0.35, Reset = 0.48 </li></ul><ul><li>CIC ≡ Fieldbus, DeltaV, DVC6000f </li></ul><ul><li>CIF ≡ Fieldbus, DVC6000f </li></ul><ul><li>Analog ≡ 4-20 mA, DeltaV, DVC6000 </li></ul><ul><li>AST ≡ Average Settling Time </li></ul><ul><li>Settling time ≡ dead time plus four time constants from first-order curve fit </li></ul>
    36. 36. Setpoint step test sample data
    37. 37. Setpoint step test sample data
    38. 38. Setpoint step test sample data
    39. 39. Setpoint step test conclusions <ul><li>Did not attempt to optimize PID tuning for each case </li></ul><ul><li>All SP responses were stable and quick, with settling time on the order of 5*  as per theory </li></ul><ul><li>Settling times generally faster with smaller module execution time and/or macrocycle </li></ul><ul><li>The limit cycle caused by control valve nonlinearity makes it difficult to measure or compare the responses </li></ul>
    1. A particular slide catching your eye?

      Clipping is a handy way to collect important slides you want to go back to later.

    ×