decoupling control loops - expertune · 1 decoupling control loops using expertune software theory,...

1

Decoupling control loops using ExperTune software

Theory, diagnosis, and practical considerations

Bernardo Soares Torres, M.Sc. (ATAN)Lívia Camargos R. de Oliveira (ATAN)

Topics to be covered in this presentation

Brief ATAN’s presentation

PlantTriage tools to find coupled loops

How to decouple control loops:

• Detuning / tuning tighter

• Static and dynamic decouplers

• Control strategy

How to define variables’ pairing in a multiloop system using the RGA

technique;

Examples

2

9290-103-25 (Rev.:003) 3

+ Mission

+ Business Profile

+ History

+ CompetitiveAdvantages

+ Partnership

+ Metric

+ MarketSegments

+ AT

+ CustomersGeographic Distribution

+ IT

+ SpecialSolutions

ATAN´s PRESENTATION

9290-103-25 (Rev.:003) 4

+ Mission

+ Business Profile

+ History


+ Partnership

+ Metric

+ MarketSegments

+ AT


+ IT

+ SpecialSolutions

MetricsCollaborators (Base Jan./2007)

Areas of Concentration

Level of Education

3

9290-103-25 (Rev.:003) 5

+ Mission

+ Business Profile

+ History


+ Partnership

+ Metric

+ MarketSegments

+ AT


+ IT

+ SpecialSolutions

Customers Geographical Distribution

9290-103-25 (Rev.:003) 6

+ Mission

+ Business Profile

+ History


+ Partnership

+ Metric

+ MarketSegments

+ AT


+ IT

+ SpecialSolutions

Customers Market Segments

Food and Beverage

Aluminum

Cement

Railroad

Mining

Oil and Gas

Metals

Petrochemicals

4

9290-103-25 (Rev.:003) 7

+ Mission

+ Business Profile

+ History


+ Partnership

+ Metric

+ MarketSegments

+ AT


+ IT

+ SpecialSolutions

Business

Automation TechnologyAutomation Technology

Information TechnologyInformation Technology

Special SolutionsSpecial Solutions

9290-103-25 (Rev.:003) 8

+ Mission

+ Business Profile

+ History


+ Partnership

+ Metric

+ MarketSegments

+ AT


+ IT

+ SpecialSolutions

Automation TechnologyIndustrial Automation

INSTRUMENTATION, CONTROL E SUPERVISION IN ON SHORE E OFF SHORE Environments(focus in turn-key project)

Systems specification and configuration (SCADA/PLC, UTR, DCS, radio frequency, data communication, etc.)

Communication drivers development

Start up and commissioning

Training

Technical assistance

5

9290-103-25 (Rev.:003) 9

+ Mission

+ Business Profile

+ History


+ Partnership

+ Metric

+ MarketSegments

+ AT


+ IT

+ SpecialSolutions

Automation Technology PIMS – Plant Information Management System

Production Information and Management Systems (Historian Systems – PI, Infoplus.21, iHistorian, others):

Interface with control / Supervision systems (PLCs, DCS and

SCADA)

Temporal Databank / Real-time process data

Interface with Corporate Systems / Relational databank (ERP,

Oracle, MS-SQL Server, etc…)

Tools for data manipulation and analysis

Development of customized applets

Training

Continuous Help Desk

9290-103-25 (Rev.:003) 10

+ Mission

+ Business Profile

+ History


+ Partnership

+ Metric

+ MarketSegments

+ AT


+ IT

+ SpecialSolutions

Automation Technology Consulting

Master Business Plan

Technical and economic feasibility study

Business plan

Conception, ROI assesment

Variability reduction studies

Basic EngineeringField information collection, Functional SpecificationSolution definition System configurationExecution Planning

6

9290-103-25 (Rev.:003) 11

+ Mission

+ Business Profile

+ History


+ Partnership

+ Metric

+ MarketSegments

+ AT


+ IT

+ SpecialSolutions

Information TechnologyProduction Management - MES

Features:

Process TrackingDowntime ManagementProduction PlanningProduct TrackingProduction Mectrics (KPIs)Asset ManagementActivity Based CostingTPMQuality ManagementShop floor integrationLabor accountingProduction accounting

9290-103-25 (Rev.:003) 12

+ Mission

+ Business Profile

+ History


+ Partnership

+ Metric

+ MarketSegments

+ AT


+ IT

+ SpecialSolutions

Special SolutionsSCORE

SCORE is a Supervisory and Control solution for the primary aluminum production industry:

SCORE 7 – Control and supervision semi distributed system

SCORE 8 – Control and supervision semi distributed system

SCORE 9 - Control and supervision distributed system

SCORE Ultra Vision – supervision system (HMI- SCORE)

Top-SCORE – Web based information system

7

Topics to be covered in this presentation

Brief ATAN’s presentation

PlantTriage tools to find coupled loops

How to decouple control loops:

• Detuning / tuning tighter

• Static and dynamic decouplers

• Control strategy

How to define variables’ pairing in a multiloop system using the RGA

technique;

Examples

Which loops are coupled?

8

Treemap Oscillation Detection –Coupled Loops generally cycle together

7.11810.7337.39102.5148.6278.420Pressão de Ar na Zona 4PIC1424

6.93628.0655.04189294.7801.370Temperatura da Zona 5TIC1521

12.9128.1429.67151.2753.8257.390Vazão de GM da Zona 5FIC1522

5.98726.4439.53148.9256.9746.6100Vazão de GM da Zona 6FIC1622

2.36139.8456.96195.9294.2917.20.001†Pressão de Ar de CombustãoPIC1014

10.914.1238.6893.08146.8277.9100*Vazão de GM da Zona 8FIC1822

5.60610.8340.731024188.2324.8100*Vazão de AR da Zona 8FIC1824

12.5116.5225.79187.8295.8693.1100*Vazão de AR da Zona 7FIC1724

4.43418.9741.5160268.8753.8100*Vazão de AR da Zona 6FIC1624

12.8919.7235.24708.5142.9269.4100*Vazão de GM da Zona 7FIC1722

Oscillation strength 3



Oscillation period 3

Oscillation period 2

Oscillation period 1OscillatingDescriptionLoop name

Many loops oscillate at the same frequency ≈ 280 sec

Oscillations in the temperatures of a furnace

9

Finding possible couplings with PlantTriage Process Interaction Mapping

Process Interaction Mapping diagnosis: No coupled loops (Metallurgical Furnace)

10

Simple strategy to decouple control loops: detune the controller

Making the tuning of a loop more conservative by changing the PID parameters – robustness change.

Kp Ti

Example 1 – Decoupling level and temperature loops

Kp Ti

Detuning a level loop reduces oscillation in a temperature loop

11

Example 2 - PIC-321 – Control of the combustion air pressure in a furnace

Common Oscillation periods!!

What is the best tuning?

PIC 321 FIC 307.1

RRTs of the loops PIC321 and FIC 307.1 became closer after optimal tuning of these loops• RRT of PIC 321 changed from 20 sec to 63

sec after tuning• RRT of FIC 307.1 was 6 times bigger and

became 2 times bigger only

Coupling between loops !!!

The implementation of the best tuning for all loops can decrease the performance of the entire system !

12

PIC-321Tuning synchronism - RRT

FIC 307.1

RRT of the loop PIC321 became 3 times smaller than that of FIC 307.1

The loops were tuned again to adjust the RRTs and reduce the coupling.

PIC 321

Decoupled Loops

Spectral AnalysisPIC-321 x FIC 307.1

Before

FIC-307.1Spectra of PIC-321

and FIC-307.1 became different

PIC-321 Coupling removed !!!!!!!

After

13

PIC-321 and FIC 307.1 after decoupling

Performance after decoupling

Example 3 – Metallurgical Plant –Initial General Statistics

14

Initial AssessmentCouplings among loops

Fan control loops: PIC104, PIC204, PIC305 and PIC405

Coke (carbon fuel) production Process Chart

Coupling Coupling

Coupling

Coupling

15

Is making the tuning more agressive a good way to decouple loops?

PIC104

PIC101

Strategy for tuning: RRT is the key !!!

Goals:

PIC104 Relative Response Time at least 3 times bigger

Steps of the job:

Tests for process identificationAdjustment of the Relative

Response Time (RRT)

16

Decoupling: smaller oscillation strength

Results – Response to disturbances

4 minutes 25 sec

17

PlantTriage Assessments

Final AssessmentGeneral Statistics

Before

After

18

Decoupling loops based on tuningChange tuning (robustness) of the coupled loops in order to have RRT (Relative Response time) as far as possible from each other (at least 3 times).

PIC104

PIC101

Decoupling

However, in some cases, detuning can really harm the performance of the detuned loop to unacceptable levels.

Alternative solution

Decouplers design and implementation

19

Decouplers DesignMIMO System

Level and temperature of a tank

Types of decouplers1. Dynamic

)()()(

22

2121 sGp

sGpsD −=

2. Static (simpler)

22

2121 Kp

KpD −=

MIMO SystemLevel and temperature of a tank

20

Option Advanced of the Loop Setup:• Configuration of extra pens – allows for monitoring other

variables simultaneously (Multiloop Analysis)

Finding the models in PID Tuner

Option Advanced of the Loop Setup:• Configuration of extra loops – After configuring the extra

trends, create a new loop for multiloop analysis. E.g.: Modelingcouplings


21

Configuration of extra loops:• The data acquisition of all variables is done simultaneously.• Through the off-line analysis option, it is possible to choose another

PV-CO pairing and then model the interaction between these variables (couplings).


Example: Finding the models in PID Tuner

Coupling between flow and temperature in DemoMMI

22


Step in the flow loop CO while the temperature loop is in manual

TemperatureFlow


Temperature

Model between the flow loop CO and the temperature

Flow CO

23

Example: 5x5 multiloop system

Block Diagram

Decoupling - exampleMatrix of the system’s transfer functions with the

relevant interactions

⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢

⎣

⎡

+−

+−

+−

++−

−

−−

−−

−−−

−

1489590.00

115407.000

1692185.0

1175175.00

185124.00

0030.0192.000

0025.0016.0110.0180.0

0000210.1

495

2002831

6

se

se

se

se

se

ss

ssss

e

ss

sss

s

TF =

24

Decoupling - example

Simulation - Implementation of the TFs in MATLAB

STEC Simulator


Decouplers Design

25


The only decoupler for this system that is not feasible was:

s

s

s

TTANTA ess

es

essD 44

49

5

1151489690.0

1489590.0

115407.0

)(++

=

+

+=−

−

→

690.0)( =→ sD TTANTA

Hence, the following was adopted instead


Implementation of the decouplers in MATLAB

STEC Simulator

26


Some results achieved by decoupling the system:

Influence of a change in the TTP flow

Without decoupler

With static decoupler

With dynamic decoupler

Decoupling example - furnace

Global diagram of the PID control loops of a 4x4 system and the couplings among them

27

Example of the use of FBD

Conventional PID with decouplers and feedforward structure

Implementation of dynamic decouplers

1)(

+=

−

TsKesD

sτ

28

Example of static decoupling

Decoupling between the differential pressure loop and the temperature in a LPG combustion chamber

Without decoupler With decoupler

Other strategies

Description of the process: Pulp sump at an iron ore processing plant

Control loops:• Level – actuates in a pump regulating the pulp outflow to

flotation columns in order to keep the level constant • Density – actuates in a valve that adds water into the

sump to reduce the density

Disturbance: pulp feed to the sump (it depends on the upstream stages of the process)

29

Old control strategy

Process behavior

Disturbance in the foam level of the flotation column downstream

Level of the pulp sump

Foam level of the flotation column

30

New control strategy

Results of the change in the control strategy

Level of the pulp sump

Foam level of flotation columns

31

Results of the change in the control strategyReduction in the variability of the foam level of the flotation

column

BEFORE

AFTER

Control System Design

What should we do if the disturbances are so strong that our decoupling strategies cannot handle them?

Study the control system design and variable’s pairing (CO x PV)

32

RGA

RGA: Relative Gain Array

Goal: Through the steady-state gains of the system, determine the degrees of interaction among the processvariables

Result:• Definition of the best variables’ pairing

Find the Steady State gains based on the matrix of the system’s transfer functions already obtained with PID tuner

⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢

⎣

⎡

+−

+−

+−

++−

−

−−

−−

−−−

−

1489590.00

115407.000

1692185.0

1175175.00

185124.00

0030.0192.000

0025.0016.0110.0180.0

0000210.1

495

2002831

6

se

se

se

se

se

ss

ssss

e

ss

sss

s

TF =

RGA – First Step

33

RGA

The elements of the RGA can be calculated by the following expression:

= steady-state gain between andelement of H

Steps to determine the RGA:• Computation of K (gains’ matrix)• Calculation of• Calculation of

ijK iC jMth

ij jiH ),(=

( )TKH 1−=

RGA analysisThe elements in a row of column always sums to 1λ > 1 : and interact and the degree of interaction grows as λincreases.λ < 0 : The sign of the open-loop gain and of the closed-loop gain are different; thus, and should not be paired.λ = 1 : ideal pair – there is no interaction with other loops.λ = 0 : does not affect ; therefore they should not be paired.0 < λ < 1 : There is interaction between the loops.

iCV jMV

iCV jMV

jMV iCV

34

Example of RGA application

K (Gains’ matrix)

FCV1 LCV2 LCV1 TCV1 ÂNG.

VZ

NTP

NTAQ

TTP

TTAQ

Example of RGA application

Matrix Λ (RGA)

FCV1 LCV2 LCV1 TCV1 ÂNG.

VZ

NTP

NTAQ

TTP

TTAQ ⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

−−

−

8980.001020.0001020.07960.00164.01183.0000788.09384.00172.0001252.00240.08989.0000001

35

ConclusionsCouplings among loops are harmful to process performance.

This is quite common in industry;

PlantTriage has some tools to help you find the coupled loops;

Once detected, the loops need to be decoupled. Possible approaches:

In order for the control system to reach a great performance level, the

variables’ pairing needs to be carefully chosen. The RGA technique can

help you to define the best variables’ pairing in a multiloop system and

PID Tuner can be used to find the models.

Change control strategyChange Tuning Design decouplers

BBúúzios zios –– Rio de Janeiro Rio de Janeiro -- BrazilBrazil

Bernardo Soares TorresBernardo Soares [email protected]@atan.com.br

www.atan.com.brwww.atan.com.br

decoupling control loops - expertune · 1 decoupling control loops using expertune software theory,...

Documents