Combining high resolution NIR furnace vision with Expert System III model based furnace control

GLASS SERVICEErik MuijsenbergThierry Barriant

Glassman Conference 2019Lyon, France

GS GROUP2

gsl.czfic-uk.comflammatec.comasens.hr

Process SimulationDefect analysisExpert System controlSpecialized EngineeringRaw materialsBurnersBoostingNIR Furnace Camera’s

Augmented Reality, IoT

Volvo AutopilotUsing cameras, radars andSoftware

Self flying TaxisVertical takeoffNot possible to fly yourself

4

Additional PC computer with OPC/DDE communication is only necessary HW addition to run ESIII. Original control equipment is used:

ES IIITM INTRODUCTION

SiemensHoneywellABBAllen BradlyYokogawaRosemount FisherEurotherm

• ESIII finds correlations between multiple input and output variables, incl. disturbance variables.

• GS ESIII state of the Art with about 5% of glass furnaces worldwide installed

• Final Goal is automatic glass quality control (melting seeds)

Models are in most cases created automaticaly from Historic data and used for Full Automatic Furnace control 95% of time

MPC Models

Historical data

Courtessy of Cortex

Internal customers: - Temperature for IS machines- Energy for all

Model example 2

Gas Boosting Cullet Ratio Pull

MPC PROCESS MODELS EXAMPLES

PID MPC

Model example 2

Gas Boosting

MPC PROCESS MODELS EXAMPLES

PID MPC

8

ESIIITM ADVANCED CONTROLLERS TOOLBOX

ESIIITM Advanced Controllers

MIMO process control

MPCModel-based Predictive Control

FLCFuzzy Logic Control

RLCRule-based Logic Control

SLCSetpoint-based Logic Control

9

ESIIITM NAVIGATION WITH PROCESS PREDICTION

10

Evaluation period: 05.09.2007 – 09.10.2007 and 18.10.2007 – 31.10.2007

(ESIII pretest and missing data periods excluded)

Total 43 days

Evaluation period: 02.02.2008 – 16.03.2008

Total 43 days

Changes in furnace glass pull (‘Glastrek’)Most frequent pulls 144 tpd (line D closed) and 204 tpd (max. pull)

Operator Control ESIIITM Control

Data from Fibre furnace control

glass temperature

gas

glass pull

11

Data sampling 5 minutes

TB3 tolerance Operator ESIIITM (old furnace) ESIIITM (new furnace)± 2 °C 19.6% 64.8% 74.4%± 4 °C 44.5% 91.6% 97.4%

Percentage of time TB3 is within tolerances ± 2 °C and ± 4 °C during entire process control, including pull changes

Data sample 2 months

ESIIITM was in control for 96% of the time

ESIIITM seoint zone± 1°C for TB3 control

Temperature stability

12

Data sampling 5 minutes

TB3 tolerance Operator ESIIITM (old furnace) ESIIITM (new furnace)± 2 °C 19.6% 64.8% 74.4%± 4 °C 44.5% 91.6% 97.4%

Percentage of time TB3 is within tolerances ± 2 °C and ± 4 °C during entire process control, including pull changes

Data sample 2 months

From 1.4.2010-1.6.2010ESIIITM was in control

for 96% of the time

ESIIITM setpoint zone± 1°C for TB3 control

To safe additional energy SP can be lowered

13

How to Save Energy with ESIIITM Process Control?

Glass temperature stabilization can be used to save energyby reduction of production temperature setpoint

Operate furnace at lower temperature,but keep above critical limit to prevent glass defects

1390

1380

1370

1360

13501350

Example adding E boosting with MPCBottom Temperature control

Manual Control ESIII Control

15

Combustion Controller

ESIIITM Extended Combustion Control Design for Melter

Gas SP’s

Gas Profile Crown Temperatures

BTU

CV’s CV’s

DV

MV

Batch Charger

O2

Air/Gas Ratio

DV

MV

(O2, CO on-line sensors in port, regenerator or stack)

CV’s

DVMin

Max

Min

Max

Air/Gas Ideal

Gas Chromatograph

GasComposition

Ambient Temp.

DV

CV SP: NOx/CO,O2 Stack or Port

ESIIITM NOx and CO minimizer, one stack sensor, 2 furnaces

• GS IP HD Visual & Near Infrared Camera with 1.300.000 virtual TC’s

GS ASENS Near Infrared Camera

Picture: GS New Furnace Camera System

• Monitoring• Input to the ES III

Camera info is forwarded into ES4/ESIII Computer for digital image analysis

ES IV level

Operator level

PLC level

Process instrumentationlevel

Picture: Camera(s) in furnace process control

ES IIITM CAMERA – Overview for Float furnace

Projection

Measured data, statistics, analysis

+

Image Grabbing and Batch Identification

Batch Identification and Monitoring Container Glass

A furnace camera image isgrabbed and analyzed every minute.

Special batch identificationsoftware is able to recognize the islands of batch(yellow) floating on top of theglass (blue). Potential builduparound the camera opening(red) is excluded from theanalysis.

In a calibrated top viewthe batch can becharacterized in termsof distribution of batchcoverage [m2] andmaximum batch length[m].

Bat

ch le

ngth

[m]

Batch cover relation with crown temperaturesExample

Batch coverage in front of furnace above 1.5 m2 has been related toAffect the number of glass defects.

Trend of batch coverage [m2] in front of furnace.

Critical area in front of furnace.

Batch cover correlation with crown temperature

Variation in batch coverage has direct impact on furnace temperatures: crown temperature dropswhen batch moves to front offurnace.

Measured batch coverage area isused as additional input in ESIIIfurnace temperature control.

Next step: ESIII control of batchcoverage by automatic adjustmentof batch charging directions.

Crown temperature

Batch front area [m2]

GS A-Sens new camera solution With 2 video streams, one visual one Infrared

Removed large animations

GS A-Sens NIR cameraWith Multiple Virtual TC’s

Glass quality in relation to batch cover & temperature monitoring (example)

1 day

Batch front (bad) Batch short (good) Batch front (bad)N

see

ds /

100

gr

Virtual NIR Temperature info correlates well with crown temperature

Batch image, history and trend IP Browser

Picture:: Browser

27

Batch image, history and trend IP Browser

28

Batch correction example with automatic burner control

ESIII Integration into DCS

29

30

Complete melting process overview for advanced control, 2 melters and forehearths

31

Additional control can be added depending on available data.

GLASS SERVICE, a.s.Rokytnice 60, 755 01 VsetínCzech RepublicT: +420 571 498 511F: +420 571 498 599

erik.muijsenberg@gsl.czthierry.barriant@glass-service.euwww.gsl.cz

Thank you for your attention

CO2 price: 22.4 €/ton