Application of Adapt Technology for Intelligent Control

1
Application of Adapt Technology for Intelligent
Control
2
Presenters

• Terry Blevins
• Willy Wojsznis

3
Introducing DeltaV Insight

• It all comes together with the DeltaV Insight - A
  revolutionary product.
• One suite with seamless transition between loop
  diagnostics and tuning
• Advanced features are possible in all areas based
  on automatic identification of process models

Process Learning
Duncan just got Smarter!
4
Enabling Process Learning System Wide

• As simple as enabling a block on-line in control
  studio.
• All controller resident PID blocks may be
  enabled/disabled on-line from in any Process
  Area, unit, or cell in DeltaV Insight
  Application
• Learning may be enabled when configuring a
  controller resident PID block
• Models are utilized throughout DeltaV Insight

5
DeltaV Insight Architecture
UI from any workstation
Data Analysis and Reporting at Server

• Embedded Learning Runs in the Controller
• Fast Data Capture
• Small Communications Load
• Runs in background

6
Basis of Model Identification

• Identification algorithm based on Model switching
  with interpolation and re-centering
• Patented technique unique to Emerson Process
  Management Patent No. 6,577,908 plus four
  pending patents
• Field proven at 4 beta sites, on over 700 loops.

Initial Model Gain G1
Multiple iterations per adaptation cycle
G2-? G2 G2?
G3-? G3 G3?
7
The Value of Process Models

• There is a real value in saving and trending
  models identified over time
• Peformance based indicies
• Identify sources of variability
• Impact of Noise and Unmeasured Disturbances on
  model identification may be minimized through the
  application of non-linear filtering.

8
Non-linear Filtering of Models

• A filter Factor (FF) is calculated for every
  adapted parameter. Two criteria are used for the
  FF calculation.
• Ratio of the maximum residual error to the
  minimum residual error (RE). If the ratio is
  small, it implies high noise level or a
  significant distance from the true model value.
• Test if the model with a middle value parameter
  value has the smallest RE. Satisfaction of these
  criteria indicates that true model value lies
  between low and high range of the parameter.

9
Non-linear Filtering (Cont)
10
Model Quality

• Model quality is calculated based on the
  following criteria
• Number of adaptations
• Average filter factor
• Model main parameter (gain) average value and
  variability of the last n adaptations of that
  parameter

11
Tuning Index

• Provides a direct indication of the desired
  control performance vs. current control
  performance
• Identifies loops that would benefit from more
  aggressive tuning or by less aggressive tuning.
• Calculation is based on the identified process
  model, selected tuning rule and current tuning,

12
Tuning Index Calculation
Process Model
Selected Tuning Rule
Tuning Index
Current Tuning
13
DeltaV Insight Tuning Index

• Tuning index is defined as the ratio of the
  potential residual PID variability reduction to
  the actual PID residual variability.
• More meaningful measure than the Harris index
  which is based on minimum variance tuning in
  statistical sense

closed loop time constant for the current PID
tuning
closed loop time constant for recommended
tuning
sampling period
14
DeltaV InSightPerformance Monitoring

• Control Statistics for
• Incorrect mode
• Limited control output
• Bad/Uncertain input
• High variability
• Explorer Tree allows easy navigation
• New Control Diagnostics
• Model based tuning index pinpoints tuning
  opportunities

New
15
Insight Performance Monitoring Block Level

• Automatic trending of parameter utilizing
  historian if assigned in module.
• Information contained on one screen for easy
  access

16
DeltaV InSight - Adaptive Tuning

• Process model is displayed and saved in the model
  database
• Model Quality and Identification Status
• Tuning criteria and desired speed of response
• Tuning Recommendation

17
DeltaV InSight - Model Analysis

• Models automatically stored in a model database.
• Various plot options to analyze impact of
  operating conditions on process models
• Average of selected models may be utilized to
  establish the recommended tuning

18
DeltaV InSight - Control Simulation

• Closed loop simulation of setpoint and
  disturbance changes for recommended tuning
• Simulation for both Adaptive and Manual Tuning.

19
DeltaV Insight Navigation

• Launch application from Start Button or in
  context from DeltaV Explorer or Control Studio
• Select Inspect and Tuning View From the same
  window

20
Summary

• DeltaV Insight is a revolutionary control
  technology breakthrough in process learning and
  intelligent control that provide unique
  capability for loop diagnostics and tuning.
• Process models are automatically developed during
  normal plant operations when Process Learning is
  enabled.
• Models that have been identified are saved in a
  model database to support analysis of the impact
  that operating conditions have on process
  dynamics
• Loops that would benefit from application of the
  recommended tuning are automatically identified
  and improvement in performance is quantified by a
  model based Tuning Index.

21
Where To Get More Information

• Visit the exhibit area and see DeltaV Insight in
  action.
• Improving Batch Operations with DeltaV Insight
  Bruce Johnson, Efren Hernandez, Terry Blevins,
  Emerson Exchange 2006
• Field Experience with Advanced Monitoring and
  Tuning Applications - Saul Mtakula, James Beall,
  John Caldwell Emerson Exchange 2006
• Product Update DeltaV InSight for Intelligent
  Process Control John Caldwell Emerson
  Exchange 2006
• The Next Generation Adaptive Control Takes a
  Leap Forward. Gregory McMillan, Mark Sowell,
  Peter Wojsznis, Chemical Processing, September,
  2004
• Theoretical Analysis of a Class of Multiple Model
  Interpolation Controllers (Presentation), Dale
  Seborg and Joao Hespanha, AIChE Annual Meeting,
  San Francisco, November 21, 2003