using predictive modeling to increase efficiency and ... · using predictive modeling to increase...

Copyright © 2012 Rockwell Automation, Inc. All rights reserved. Rev 5058-CO900C

Using Predictive Modeling to Increase Efficiency and Effectiveness

Ric Snyder Sr. Product Manager November 7-8, 2012

Copyright © 2012 Rockwell Automation, Inc. All rights reserved.

Automation for Process Industries

2


Pavilion8 Dynamic MPC

PlantPAx Fuzzy Designer

Logix IMC, CC, MMC

Logix PID, PIDE, AOI

PlantPAx Model Builder

PlantPAx Advanced Process Control Portfolio

3

Real-time Optimization

Model Predictive Control

Nonlinear Multivariable

Control

Linear Multivariable Control

Inferential Sensors

Advanced Regulatory Control

Regulatory Control Incr

easi

ng E

ffort

& In

crea

sing

Val

ue


Why Advanced Process Control?

Advanced Regulatory Control Focus is on PROCESS variables – levels, flows, temperatures,

pressures, etc. Improve poorly performing loops and/or automate manual loops Compensate for dead-time and simple process interactions Improve process stability, consistency, reliability

Advanced Supervisory Control Focus is on PRODUCT variables – production rate, product quality,

product specifications (e.g. moisture, color, density, purity, etc.) Sends setpoints to process control loops – good regulatory control is

prerequisite to achieve full MPC benefits Improve product quality Increase throughput and/or yield Reduce energy usage

4


Scalable Portfolio of Advanced Controls

Regulate

Single Loop

Multiple Loops

Process Unit

Multiple Units

Plant-wide

Value-Add

Predict

Control

Optimize

Area

Span of Control

5


Starch Dryer Example PID Temperature Control

• Neither moisture laboratory samples nor moisture soft sensor are available

• Disturbances are not corrected or slowly corrected

• Ambient humidity • Starch consistency • Inlet air flow temperature

Starch feeder

PID

Output limit

Tout Ctrl

Tin

Tin Setpoint

CTRL-valve(s) Heat recovery

Hot Steam

Inlet Temp Ctrl

heat flow

Inlet

temp

2 bar

12 bar

Hot Water

Feeder Speed

Valve Position

PID SP

Tim

e De

lay

SP

Motor

Tout

Hot a

ir &

star

ch m

ixtur

e

Inlet air flow T_Z1 T_Z2

Dried starch out

Heat exchangers

VFD

6


PID Control Performance

Moisture SP

Moisture PV

Feed Rate

Exhaust Temperature PV

Exhaust Temperature SP

7


What is a Model?

A model explains or emulates the behavior of a process...

… using a computational/mathematical representation

y = a3 u3 + a2 u2 + a1 u + a0

monomer

modifier

catalyst

melt index

density

Models Represent “Knowledge”

8


Uses of Models

Analysis Analyze vast amounts of historical

data to discover unknown correlations and relationships to gain better insight into the process

Prediction Build models that can accurately

predict process changes based on changing process inputs and disturbances

Combine prediction models with real-time data to build Soft Sensors to identify product changes before they occur, allowing pro-active operator actions

Control Closed-loop multivariable model

predictive control (MPC) provides faster responses to changing operating conditions and constraints to minimize product variability and improve yields

Optimization Steady-state models can identify

the optimum operating parameters based on current conditions and constraints to meet desired economic objectives (maximize rate, minimize energy, …)

9


You Can’t Control What You Can’t Measure

What if in-line sensors are unavailable, unreliable, too expensive

Analysis

Process Samples

Lab Results

One hour to one day delay !

Process Adjustments

143.0

10


Solution: Soft Sensor

Software model that predicts process values based on real-time process data

143.0

Analysis

Process Samples

Process Adjustments

Soft Sensor Prediction

Real-time measurement!

⌡ ⌠ ( x - υ )2

dυ 4c2 t

exp -1

+1 Uo 2c√π t

∂ u ∂ 2u ∂ t ∂ t2 = c2

Real-time Process Data

Lab Results 0.34 12.01 5.10

0.45 12.34 5.34

0.12 12.67 5.23

0.14 13.03 6.02

0.46 13.67 6.70

0.67 13.69 4.98

rate temp thick

0.34

0.45

0.12

0.14

0.46

0.67

%

11


PlantPAx ModelBuilder PlantPAx ModelBuilder for off-line data

analysis and modeling SoftSensor Designer converts models to AOI’s

143.0

Analysis

Process Samples

Process Adjustments

12


Model Based APC Function Blocks Regulatory control functions based on Internal Model

Control (IMC) algorithm Handles first order plus dead-time (FOPDT)

models Gain, time constant, and delay parameters

Good for processes with dead-time or processes with simple coupling

Different Function Blocks for Specific Problems IMC (Internal Model Control)

Replace single PID loop to address dead-time CC (Coordinated Control)

Additional control outputs (2 or 3) for coordinating multiple actuators for a single process variable

MMC (Modular Multivariable Control) Coordinates 2 loops that interact with each other

13

IMC_01

IMC

...

Internal Model Control

0.0

PV

0.0

SPProg

0.0

SPCascade

0 0

CVEU

0.0

SP

0.0

ProgOper

0

0

CC_01

CC

...

Coordinated Control

0.0

PV

0.0

SPProg

0.0

CV1Prog

0.0

CV2Prog

0

CV3Prog

0

CV1EU

0.0

CV2EU

0.0

CV3EU

0.0

SP

0.0

ProgOper

0

0

MMC_01

MMC

...

Modular Multivariable Control

0.0

PV1

0.0

PV2

0.0

SP1Prog

0.0

SP2Prog

0.0

CV1Prog

0.0

CV2Prog

CV1EU

0.0

CV2EU

0.0

CV3EU

0.0

SP1

0.0

SP2

0.0

ProgOper

0


Improved Starch Dryer Control with APC

14

IMC Moisture Control • Moisture Soft Sensor is used for outlet

moisture control

Starch feeder

IMC

Moisture Ctrl

Tin

Tin Setpoint

CTRL-valve(s) Heat recovery

Hot Steam

Inlet Temp Ctrl

heat flow

Inlet

temp

2 bar

12 bar

Hot Water

Feeder Speed

Valve Position

PID SP

Tim

e De

lay

SP

Motor

Tout

Hot a

ir &

star

ch m

ixtur

e

Inlet air flow T_Z1 T_Z2

Amb Air Humidity

Dried starch out

Moisture SoftSens

Moisture Lab Sensor

Heat exchangers

Starch density

VFD

Sample

PID SP

Exhaust T Ctrl

PV

PV

CV


IMC Control with Soft Sensor – SP Change

15


IMC Control with Soft Sensor

16


Model Predictive Control

The key to effective control is good modeling

Process

Model Predictive Controller

Process Model

Optimizer

Measurements

Control actions

Trial outcomes

Trial actions

Disturbances

A Systematic Approach for Control of Processes with Constrained Nonlinear Dynamics

Targets, Constraints

17


Spray Dryer Challenges

Drop Tank

TIC 315 Natural Gas

Flow FT

804

FIC 314

CV 347

Deaerator

To Cyclones

TT 301 - 304

Avg

FIC 305

TT 310

DT 304

Slurry Flow

Slurry Temperature

Slurry Density

Inlet Air Temperature

Inlet Air Flow

Powder Moisture

Hold Tank PIC 319

Tower Top Temperature

Tower Pressure Controller OP

Slurry Recirculation

Valve

From Mixers

VSD

Dual valve slurry flow control

No inline moisture sensor

• Control powder moisture to target

• Maintain tower temperature and tower under-pressure below limits

• Minimize tower temperature when possible to reduce energy usage

Multiple interacting control loops

18


Pavilion8 Model Predictive Control

PlantPAx Fuzzy Designer

Logix IMC, CC, MMC

Logix PID, PIDE, AOI

PlantPAx Model Builder

Manipulate air flow, slurry flow, and air temperature setpoints to control powder moisture to target and maintain tower temperature & pressure within constraints while compensating for slurry temperature and density variations

Powder Moisture Prediction

Startup Setpoints

Slurry Flow

Inlet air flow and temperature

Spray Dryer Solution from Rockwell Automation

Drop Tank

TIC 315 Natural Gas

Flow FT

804

FIC 314

CV 347

Deaerator

To Cyclones

TT 301 - 304

Avg

FIC 305

TT 310

DT 304

Slurry Flow

Slurry Temperature

Slurry Density

Inlet Air Temperature

Inlet Air Flow

Powder Moisture

Hold Tank PIC 319

Tower Top Temperature

Tower Pressure Controller OP

Slurry Recirculation

Valve

From Mixers

VSD

19


Spray Dryer Model Predictive Control

20


Spray Dryer MPC Performance Metrics

21


Benefits of Model Predictive Control

Faster moisture measurement and tighter control allows average moisture target to be increased Sell more water/less expensive ingredients Reduce energy used for drying Increase throughput through dryer

SPECIFICATION OR LIMIT

BEFORE MPC WITH MPC

TIME

KEY

TAR

GET

WITH OPTIMIZATION

22


Polymer/Chemical – Reactors/Extruders – Distillation Towers – Furnaces Process Types

– PE, PP, PS, PC – Ethylene Plants – Styrene Plants – Crude Refining – Gas Plants

CPG – Spray Dryers – Evaporators – Energy Centers Process Types

– Milk Powder – Coffee – Laundry

Detergent – Conc. juice

CMM – Crushing/Grinding – Kilns & Drying – Stockpile Blending Process Types

– Cement – Minerals – Fertilizer – Ammonia

Bio-fuels – DDGS Evap/Dryer – Water Balance – Fermentation – Distillation Process Types

– Corn Ethanol – Cane Ethanol – Bio-diesel

Typical Benefits – 5 to 8% production increase – 30 to 60 % moisture variability reduction – 20 to 50% off-spec product reduction – 5 to 10% energy consumption reduction

Typical Benefits – 2 to 5% production increase – 2 to 5% energy consumption reduction – 20 to 40% product variability reduction – 10 to 30% off-spec product reduction

Typical Benefits – 4 to 8% prime product yield increase – 35 to 75% product variability reduction – 20-40% transition time reduction – 3 to 7% feed stock wastage reduction

Typical Benefits – 4 to12% ethanol production capacity increase – 2 to 5% ethanol yield increase – 3 to 6% energy use/gallon reduction – 1 to 2% DDGS yield increase

Pavilion MPC Applications

Typical Project Payback: 3 to 9 months!

23


Benefits of PlantPAx Advanced Controls

Smart, Safe & Sustainable Production 24


PlantPAx Advanced Controls & Optimization

Scalable, integrated controls & optimization

Value assessments and engagement methodology

Industry expertise and best practices

Sustained Value Services & Support

Faster time to Value / Lower TCO

25

using predictive modeling to increase efficiency and ... · using predictive modeling to increase...

Documents