Lucie Mill

  • Uploaded by: Joko Dewoto
  • 0
  • 0
  • January 2021
  • PDF

This document was uploaded by user and they confirmed that they have the permission to share it. If you are author or own the copyright of this book, please report to us by using this DMCA report form. Report DMCA


Overview

Download & View Lucie Mill as PDF for free.

More details

  • Words: 1,376
  • Pages: 50
Loading documents preview...
LUCIE Introduction to the Expert System Theory

La f a r g e Un i v e r s a l Co n t r o l &

?

I nference

En g i n e 2

Expert System 

Basic form of artificial intelligence



Decisions equivalent to those of the human bean



Developed by interviewing an experienced person



Consolidates process operating know how into a standard product easy portable to any plant



Two key components: … 3

1. The Knowledge base 

A set of rules, information, facts about a certain subject





Stored in an organized structure

Populated with both questions and answers 4

2. The Inference Engine 

Rule-based algorithm that interacts with a Knowledge Base to draw conclusions about a set of

inputs 

Emulates the human capability to arrive at a conclusion by reasoning

5

LUCIE Mill Process Principles

What do you wish as Mill Operator?

The highest production of very good quality cement/raw mix under stable conditions 

Is this all ? 7

What do you need?

Sensors

Actuators

8

What do we use … SENSOR Amps Elev

ACTUATOR Separator

Sep Speed

Rejects

Quality – Blaine, SO3 Finish Product

Mkw

Fresh Feed

Nl1

Feed Rate

Temp

Nl2 Mill

Gypsum %

9

Control Limitations 

LUCIE changes set-points ONLY!



No actual equipment control (motor starts/stops, alarm acknowledgement)



Lucie is not hiding mechanical/process problems.

On the contrary! 10

Principles 

1st Stabilize Mill Throughput



2nd Increase Production Level by Optimizing Throughput



3rd Optimize Quality

11

Mill Strategy Organization Sensor 1 Set-points

Sensor 2

Sensor 3

Virtual Sensor (Estimates) Normalized values

Short term Potential

Long term Potential LT-Action

STActions

Time constant

 Lucie Actuators Set-points 12

Treatment of sensors

WHY? 

To allow Lucie to continue to operate when a sensor signal is no longer significant



To enable the strategy to always work with a plausible signal value



To provide the most representative information of the real state of the kiln / mill

13

Treatment of sensors

HOW? 

By filtering - eliminate the signal noise



By defining inside Lucie of four possible sensor “states” and two “validity” values

14

FILTERS - Example Sensor

Field

Value Set-point

State

Validity

The field-value of the sensor is not enough filtered.

The Lucie filtered value

15

Sensor

Signal Treatment

Normal

Doubtful

Frozen

Abnormal

Valid

Valid

Valid

Invalid

Invalid

16

LUCIE Mill The Estimates

The Estimates (Virtual Sensors) 

Evaluate and forecast continuously how a particular control parameter (mill throughput, material level, etc.) will vary



Are the



All actions are determined from the estimate results

of Lucie

18

The Mill Estimates 

Estimates with impact on production  The Mill Throughput Estimate  The Material Level Estimate  The Drying Estimate



Estimates with impact on quality  The Quality Estimates

19

The Mill Throughput Estimate 

Goal:

Calculate the mill throughput deviation

from the set point 

Sensors: Elevator Amps, ((Rejects, Feed))



To each sensor a mono-estimate is connected



The mono-estimate converts the value from the sensor into a common reference unit (t/h of MTP)

20

The Mono-Estimate 

Mathematically expressed: Mono-Estimation = Gain x (PV - Set Point) + Offset



The gain can be calculated:

 Reference Sensor Gain =

 Mono„s Sensor 21

The Multi-Estimate 

The Mono-Estimate which is Estimating the Smallest Margin is Chosen



The output of the multi-estimation are the State and the Tendency in Normalized Values

22

Normalization 

Converts a particular value within a predefined range [-4 , +4]



Brings all the signal on the same “playing field”



Enables reasoning with symbolic states



error = Value - Set Point 23

Normalized State 30 t/h

+4

very high

24 t/h

+3

20 t/h

high

17 t/h

slightly high Multi Estimate

9 t/h

+2 +1

normal -9 t/h

slightly low

-17 t/h

-2

low -24 t/h

very low -30 t/h

-1

-3 -4

N O R M A L I Z A T I O N

24

Normalized Tendency 

How quickly and in what direction the error is changing



Based upon 2 errors compared ~8 minutes apart 

Norm. Tendency = Norm. State (t) - Norm. State (t-)



Value between (-4 to +4)



i.e., fast filling, slow emptying

25

The Material Level Estimate 

Goal:

Calculate the material level of the mill (security)



Sensor: Electrical ears (C1 / C2) Mill power / Amps (DP)



Same treatment as done by the mill throughput estimate

26

The Drying Estimate 

Goal:

Qualify the margin of available heat in the mill



Sensor: Gas temperature at mill exit Material temperature at mill exit (Gas temperature at mill inlet)



This estimate is reducing the feed if the minimum temperature is not achieved 27

Potentials



From each multi-estimate a potential of feed is determined 

A Short Term Potential



A Long Term Potential

28

Potential Calculation

Sum of Normalized Mill Tend. and State from Estimate [- 4; +4 ]

ST/LT Action Fuzzy Logic Table

Short/Long Term Action Potential

in tons of mill feed

29

Potential Selection Major vs. Minor 

Major  Continuous control  Potential used



Minor  Security control - SAFEGUARD  Potential Used IF (State, Tendency) Exceeds Threshold

30

The Min-Action Object 

The Minimum of the short- and long term potentials is chosen



These potentials are piloting the mill



They are called the short- and long term Pilot

31

The Short Term Actions



Used to stabilize the mill



They Are:    

Proportional to the set point deviation Of big amplitude Temporary Superimposed on the long term actions

32

The Long Term Actions 

Used to maintain the long term stability



They are:   



Of low amplitude Cumulative Permanent

Weighted by a factor which takes into account the past

33

Who Is The Pilot? Major MTP estimator

Minor ML estimator which has not exceeded the threshold

Proposes

Proposes

+ 1 ton per hour

- 3 tons per hour

Pilot estimator = Mill Throughput Result = + 1 ton per hour

34

Who Is The Pilot? Major MTP estimator

Minor ML estimator which has exceeded the threshold

Proposes

Proposes

+ 1 ton per hour

- 3 tons per hour

Pilot estimator = Material Level Result = - 3 tons per hour

35

LUCIE Mill Optimization Of Mill Throughput

Relationship Feed / Mill throughput Feed



LUCIE Calculates the Feed and MTP Set Point Variation



Same Sign -> MTP Set Point Increases



Different Sign -> MTP Set Point Decreases

Max Feed

Positive Increment

Negative Increment

D Feed >0 D MTP

D Feed <0 D MTP

Opt. Set Point

Mill Throughput

37

LUCIE Mill The Quality Estimates

The Quality Estimates 

Fineness, SO3 ...



Input: Sensor or Manually



Quality Target is the Set Point in LUCIE



Designed to mimic SPC control

39

The Quality Estimates 

Calculation: Quality Level = Input Value - Set Point



A normalized value is then calculated from this quality level



Actions triggered by control card

40

Normalization 350 300 200 90

3750 – 3500 Blaine

-90 -200 -300 -350

Very High High Slightly High Normal Normal Slightly Low Low Very Low

+4 +3 +2 +1

-1 -2 -3

-4

N O R M A L I Z A T I O N

41

Calculation Of Action

Gain State of the Quality Estimate

LT-Fuzzy Table

X

Long-term Increment for separator speed

42

LUCIE Mill The Product Table

The Product Table



Add / Remove Products



Define individual recipe for each product  Set Points for Mono Estimators  Scale Factors for Actions  Quality set points

44

Recipe Files

45

LUCIE 

Is a tool for the plant improvement



Duplicates the Operator behaviour based on fundamental process principles



Can yield higher production rates (~3%) and lower standard deviation for quality parameters



Is dedicated to both Process and Production

46

Do you know that Lucie controls   



109 cement mills 34 raw mills 5 coal mills 7 vertical mills

in more than 50 plants all over the world ?

47

LUCIE Mill The Operator Screen

49

50

Related Documents

Lucie Mill
January 2021 1
Als Lucie Operation
February 2021 1
Ball Mill
February 2021 0
Ok Mill
January 2021 3
Bowl Mill
February 2021 2

More Documents from "TELEGBIASIA"

Coal Mill Safety
January 2021 1
Lucie Mill
January 2021 1
Rayap
January 2021 4