Download as pdf or txt
Download as pdf or txt
You are on page 1of 18

Honeywell Forge APC-4516

Process Controller Implementation


Process Controller Design

Honeywell Confidential
Copyright © 2012 Honeywell International Inc.
Honeywell Restricted
1 September 26, 2022
Controller Design – Project Steps Honeywell
• Benefit analysis or Feasibility study
• Functional design
• Preliminary plant test /PID tuning
• Application design
• Final plant tests
• Identification - modeling
• Off-line controller build/simulation
• Commissioning
• Trouble Shooting
• Operator & engineer training
• Post Audit
• Appendix 04A covers the Benefit Analysis concept, 04B covers the CV
Transformations and 04C Preliminary –Step Testing.

Honeywell Confidential
Copyright © 2012 Honeywell International Inc.
Honeywell Restricted 2 September 26, 2022
Functional Design Honeywell
• Define Process Control Objectives
– Economics
– Operation
• Specify:
– Manipulated Variables (MV’s)
– Disturbance Variables (DV’s)
– Controlled Variables (CV’s)
• Identify and Assess:
– Missing Instrumentation
– Basic Regulatory Control Performance
• Process Control Objectives
– Understand the process and its operation in detail
– Understand the problems
– Understand operating objectives of the process (unit)
– Define the control objectives

Honeywell Confidential
Copyright © 2012 Honeywell International Inc.
Honeywell Restricted 3 September 26, 2022
Control Expectations Honeywell
• What should we expect the controller to do for us ?
– Key process parameters within limits subject to noise level
– Minimize movement in MV’s
– Steady State optimization
– Stabilize process (downstream units)
– Allow operator to perform other tasks
• What should we not expect:
– Cannot solve physical equipment limitations
– Control of Model Based Control Variables depends on accuracy of model
(Inferential or Engineering Model)
– Control of Analyzer based Control variables depends on analyzer reliability
• Set / Revise the control expectations accordingly
• Understanding the economics of the unit is essential to designing good
controls
• Sensors should function correctly
• Valves in working order

Honeywell Confidential
Copyright © 2012 Honeywell International Inc.
Honeywell Restricted 4 September 26, 2022
Regulatory Control Strategy Honeywell
• Break regulatory PID control cascades ? (e.g. top temperature to reflux
flow)
– does the loop reject disturbances
– does the loop work well
• Compare to minimum variance performance
• Use a commercially available package for evaluation
– does the loop respond fast relative to RMPCT
• Regulatory controllers can reduce model matrix density
• Break regulatory level control loops ?
– Is there sufficient level control inventory ?
– Is the level control valve adjusting the feed to another unit ?
– Do we need to stabilize the feed to the downstream unit ?
• for quality control reasons
• pushing the unit closer to constraints
– Step testing more difficult
– Can calculate models for many level configurations
Honeywell Confidential
Copyright © 2012 Honeywell International Inc.
Honeywell Restricted 5 September 26, 2022
Regulatory Controller Tuning Honeywell
• Tune regulatory controls preferably before pre-step tests and certainly
before main step test
• Dynamics of regulatory controls are embedded in matrix model
• Poor regulatory controller tuning leads to poor performance of
multivariable controllers
• Consistent regulatory controller tuning leads to optimum performance

Honeywell Confidential
Copyright © 2012 Honeywell International Inc.
Honeywell Restricted 6 September 26, 2022
Options for Tuning Honeywell
• Trial and error
• “1st Generation” Computer based methods
– Loop-tune, CUPID etc.
• “Robust” Computer based methods
– R-PID (Profit PID), next generation tools
• “2nd Generation” Computer based methods
– Oper-tune (Experion)
– Tai-Ji

Honeywell Confidential
Copyright © 2012 Honeywell International Inc.
Honeywell Restricted 7 September 26, 2022
Controller Scope Honeywell
• Decide on boundaries between controllers
– One big controller covering multiple units
– Separate controllers for
• Reactor/regenerator
• Main fractionator
• Light-ends columns
• One large controller
– Enables unit wide optimization in control matrix
– Handles common constraints between sub-units
• Individual controllers
– Less diverse range of CV response times
– Cannot handle common constraints between sub-units
– More control over possible MV/CV pairings
– Easier to step test - less time
– Easier to commission

Honeywell Confidential
Copyright © 2012 Honeywell International Inc.
Honeywell Restricted 8 September 26, 2022
Choosing MV's (control issues) Honeywell
• Independent of other MV's and DV's (exceptions)
– Not only its moves, but also its effects on CV's
• Independent of other control schemes
– Minimum requirement: one-way interaction
– Tuning for inactive MIMO controllers can be risky
• A good MV should have
– Significant effects on some CV's
– Predictable effect on all CV's (if none that’s OK too)
• Use for control vs for optimization
– Control
• MV's with a fast, strong and predictable effect, with a good range to
move
– Optimization usage
• MV's with small effect and MV's with a less predictable effect

Honeywell Confidential
Copyright © 2012 Honeywell International Inc.
Honeywell Restricted 9 September 26, 2022
Examples of MV's Honeywell
• Reflux flow
• Reboiler flow or duty
• Tray Temperature
• Overhead Pressure
• Feed Temperature
• Feed Flow
• Compressor Speed
• Heater Fuel Gas Pressure

Honeywell Confidential
Copyright © 2012 Honeywell International Inc.
Honeywell Restricted 10 September 26, 2022
Choosing DV's (control issues) Honeywell
• Must be independent of other MV’s and DV’s
– Significant effects on CV's
– Reliable measurement
– DV change must be independent of MV's and other DV's
– Sometimes desirable to specify as DV for selected CV’s and as MV for
remaining CV’s (be aware of consequences in terms of sub-optimality)
• DV's often cannot be moved independently
– Does that violate the rule of independence?
– What if some of them always come together?
• Use only one of them

Honeywell Confidential
Copyright © 2012 Honeywell International Inc.
Honeywell Restricted 11 September 26, 2022
Examples of DV's Honeywell
• Feed flow
• Feed temperature
• Feed composition (analyzer)

Honeywell Confidential
Copyright © 2012 Honeywell International Inc.
Honeywell Restricted 12 September 26, 2022
Choosing CV's (control issues) Honeywell
• Dependent on at least one MV
– Controllability of that CV under all operations and all year around
– Build the relationship between the CV and the control objective
• For measured CV's
– Consider measurement noise, dead-time, and dynamics
• For inferred CV's
– Replaces/complements on-line analyzer
– First principle engineering model

Honeywell Confidential
Copyright © 2012 Honeywell International Inc.
Honeywell Restricted 13 September 26, 2022
Inferential Calculations1 Honeywell
• If analyzer is not available, is unreliable or dead-time is prohibitive
consider use of inferential calculation
• Use a statistical modeling tool which includes various techniques such
as:
– Linear models
– Partial least squares
– Neural Networks
• Use a first principles engineering model

Honeywell Confidential
Copyright © 2012 Honeywell International Inc.
Honeywell Restricted 14 September 26, 2022
Inferential Calculations2 Honeywell

Dynamic Compensation

• where f1, f2 ... fm filters (possibly with delays) are chosen so that the
inputs to the calculation block are synchronized

• Advantage: the inferred CVs have a cleaner dynamics
• A Note: over-compensation can cause a sluggish CV

Honeywell Confidential
Copyright © 2012 Honeywell International Inc.
Honeywell Restricted 15 September 26, 2022
Choosing Objective Function1 (LP) Honeywell

• Linear Program Optimization

J = min ∑ $ icost cvi + ∑ $ cost


j mv j
xss i j

– Try to push to a corner of operating regime


– Minimize utilities or other operating variables
– Maximize most valuable product
• Product Value Optimization (PVO)

Honeywell Confidential
Copyright © 2012 Honeywell International Inc.
Honeywell Restricted 16 September 26, 2022
Choosing Objective Function2 (QP) Honeywell

• Quadratic Program Optimization

J = min ∑ $δi cos t (cvi − yiss ) 2 + ∑ $δj cos t (mv j − u ssj ) 2


xss
i
– Use to keep CV's/MV's around a desired optimal soft target
• Assumes the optimum MV/ CV values are known
• Can be downloaded from rigorous model/optimizer
– Difficult to prioritize CV QP terms against other LP costs. Alternatively use
soft bounds

Honeywell Confidential
Copyright © 2012 Honeywell International Inc.
Honeywell Restricted 17 September 26, 2022
Product Value Optimization (PVO) Honeywell
• Define the objective function



- ∑ Product flows × Product prices 

J = + ∑ Feed flows × Feed costs 
 
+ ∑ Energy/utility flows × Energy/Utility costs

! Remember to convert the flows into


proper eng. units so that by multiplying
with the LP coefficient it translates in to $

Honeywell Confidential
Copyright © 2012 Honeywell International Inc.
Honeywell Restricted 18 September 26, 2022

You might also like