Equipment failure forecast in a semiconductor production line

GENUA Caterina Maria, PAPPALARDO Maria Vittoria STMicroelectronics

0

Purpose ƒ Improve the data model of the mathematical simulator used in STMicroelectronics fabs, predicting equipment failures (unscheduled down or preventative maintenance), through the application of non-parametric inferential statistics techniques on historical equipment data. ƒ Obtainable benefits: Equipment failure prediction can improve STAP simulator data model accuracy, thus providing better indications to the shop floor for dispatching, and giving more reliable commitment to customers. 1

Catania STMicroelectronics production lines

M5

Catania site aerial view

2

Catania STMicroelectronics shop floor

Diffusion Furnace

Grey area equipment Diffusion area

25-wafers lot

Photolithography area

Clean room operator

Etching area

Production software tools COBOL extracts

Workstream DB

STAP toolbox

Production data extracts

STAP engine

Simulator input

Fab Performance Viewer framework

Simulator output

STAP toolbox

APF reporting system

FPV repository

Reports 4

Production software tools used in this project Production control Group uses Shop Floor Scheduling tools for data analysis and for production process dispatching indications. Among those tools, the following have been used in this project: ƒ STAP (ST Autosched Accelerated Processing) : mathematical simulator produced by Amat, which, through fab data model, reproduces the whole production process and generates reports, such as production targets. ƒ FPV (Fab Performance Viewer): framework for graphical and statistical analysis of production data related to process flows, shop floor equipment, operators. 5

Reliability theory (1/2) ƒ Availability: proportion of time a system is in a functioning condition ƒ Availability indicators ƒ MTBF – Mean Time Between Fail ƒ MTTF – Mean Time To Fail ƒ MTTR – Mean Time To Repair

up

down MTTR

MTTF MTBF

Reliability theory (2/2) Probability distributions mainly used in reliability theory ƒ Exponential distribution, if the system has a constant failure rate, i.e. the rate does not vary over the life cycle of the system with aging ƒ Weibull distribution, if the failure rate of the system grows as the system grows older, due to aging and use, so the older the system is, the more it tends to fail

7

STAP data model Product file

Route files Generic resources file

Order files

Calendar files STAP simulator engine

Stations file

Reports

STAP calendar files (1/3) ƒ Equipment down (unscheduled failure) calendar ƒ Association of down calendar file to single station ƒ Equipment PM (Preventative Maintenance) calendar ƒ Association of PM calendar file to single station

9

STAP calendar files

(2/3)

up

down MTTF

MTTR

Equipment down (unscheduled failure) calendar DOWNCALNAME

DOWNCALTYPE

MTTFDIST

MTTF

DN_LAM4520

mttf_by_cal

exponential

144.86

MTTF2

MTTF3

MTTFUNITS

MTTRDIST

MTTR

hr

exponential

15.5

MTTR2

MTTR3

hrs

Association of down calendar file to single station RESTYPE

RESNAME

CALTYPE

CALNAME

FOADIST

FOA

FOA2

stn

LAM4520O207

down

DN_LAM4520

weibull

0.571429

10.26469

FOA3

MTTRUNITS

FOAUNITS hr

10

STAP calendar files (3/3)

up

down MTTR MTBPM

Equipment PM (Preventative Maintenance) calendar PMCALNAME

PMCALTYPE

MTBPMDIST

MTBPM

PM_LAM4520O207

mtbpm_by_cal

exponential

27.06

MTBPM2

MTBPMUNITS

MTTRDIST

MTTR

MTTR2

MTTRUNITS

hr

weibull

0.425532

1.009563

hr

Association of PM calendar file to single station RESTYPE

RESNAME

CALTYPE

CALNAME

FOADIST

FOA

FOA2

stn

LAM4520O207

pm

PM_LAM4520O207

weibull

0.571429

10.26469

FOA3

FOAUNITS hr

11

Exponential distribution (1/2) Probability density function: ⎧ 1 − x β se x ≥ 0 ⎪ e f ( x) = ⎨ β altrimenti ⎪⎩ 0

Cumulative distribution function: ⎧⎪1 − e − x β se x ≥ 0 F ( x) = ⎨ altrimenti ⎪⎩ 0

Mean:

μ=β Variance:

σ2 = β2

12

Exponential distribution (2/2) Properties: ƒ Skewed distribution ƒ Used for events with high variability (e.g. equipment MTTF)

STAP exponential distribution data model ƒ MTTFDIST Æ exponential ƒ MTTF Æ 10 ƒ MTTFUNIT Æ hr 13

Weibull distribution (1/2) Probability density function: ⎧ −α α −1 − ⎛⎜⎝ x β ⎞⎟⎠ se x ≥ 0 ⎪ f ( x ) = ⎨αβ x e altrimenti ⎪⎩ 0 α

Cumulative distribution function: ⎛ x ⎞α ⎧ ⎪1 − e −⎜⎝ β ⎟⎠ se x ≥ 0 F ( x) = ⎨ altrimenti ⎪⎩ 0

Mean:

μ=

β ⎛1⎞ Γ⎜ ⎟ α ⎝α ⎠

Variance: β2 σ = α 2

⎧⎪ ⎛ 2 ⎞ 1 ⎨2Γ⎜ ⎟ − ⎪⎩ ⎝ α ⎠ α

⎡ ⎛ 1 ⎞⎤ ⎢Γ ⎜ α ⎟ ⎥ ⎣ ⎝ ⎠⎦

∞

2

⎫⎪ ⎬ ⎪⎭

where: Γ( z ) = t z −1e −t dt (gamma function)

∫ 0

14

Weibull distribution (2/2) Properties: ƒ Distribution family including exponential ƒ Widely used in reliability theory to analyze system life cycle ƒ Unimodal and skewed ƒ when α=1, it’s an exponential with mean β ƒ when α 0 ⎪ f ( x) = ⎨ Γ(α ) ⎪⎩

altrimenti

0

Cumulative distribution function: α −1 ⎧ ( βx ) j se x > 0 − βx ⎪ 1− e F ( x) = ⎨ ⎪ ⎩

∑ j =1

0

j!

altrimenti

Mean:

μ=

α β

Variance: σ2 =

α β2 ∞

where: Γ( z ) = t z −1e −t dt (gamma function)

∫ 0

16

Gamma distribution (2/2) Properties: ƒ Distribution family including exponential ƒ Unimodal, can be skewed or almost symmetric ƒ when α=1, it’s an exponential with mean 1/β ƒ when α>10, can be approximated by a normal with mean μ and variance σ2

Gamma distribution STAP data model ƒ MTTFDIST Æ gamma ƒ MTTF Æ 1 ƒ MTTF2 Æ 17 ƒ MTTFUNIT Æ hr 17

STAP data model improvement ƒ golden tools selection ƒ PM and down calendar files update, applying inferential statistics to historical data of golden tools ƒ what-if analysis: compare two simulation runs ƒ sim_before: calendar files updated with deterministic approach ƒ sim_after: calendar files updated with probabilistic approach

ƒ Simulation results comparison ƒ Benefits 18

Calendar files update Deterministic approach ƒ Calendar manual update ƒ Periodical (e.g. every three months) ƒ Exponential distribution ƒ MTTF, MTBPM, MTTR values extracted from production reports and communicated by production people to production control people in charge of updating STAP data model

19

Deterministic approach sim_before (1/2) Equipment down calendar

DOWNCALNAME

DOWNCALTYPE

MTTFDIST

MTTF

MTTFUNITS

MTTRDIST

MTTR

MTTRUNITS

DN_LAM4520

mttf_by_cal

exponential

144.86

hrs

exponential

15.5

hrs

Association of down calendar file to single station

RESTYPE

RESNAME

CALTYPE

CALNAME

FOA

FOAUNITS

stn

LAM4520O207

down

DN_LAM4520

122676

sec

stn

LAM4520O303

down

DN_LAM4520

57458.9

sec

stn

LAM4520O208

down

DN_LAM4520

418386

sec

stn

LAM4520O213

down

DN_LAM4520

284171

sec

stn

LAM4520O209

down

DN_LAM4520

499481

sec

stn

LAM4520O210

down

DN_LAM4520

27500.5

sec

stn

LAM4520O306

down

DN_LAM4520

481997

sec

stn

ALLIAN112

down

DN_LAM4520

245064

sec

NOTE: The same calendar is associated to more than one station.

20

Deterministic approach sim_before (2/2) Equipment PM calendar

PMCALNAME

PMCALTYPE

MTBPMDIST

PM_LAM4520

mtbpm_by_cal

exponential

MTBPM 300

MTBPMUNITS

MTTRDIST

MTTR

hrs

exponential

11.7

Association of PM calendar file to single station

RESTYPE

RESNAME

CALTYPE

CALNAME

FOA

stn

LAM4520O207

pm

PM_LAM4520

254057

sec

stn

LAM4520O303

pm

PM_LAM4520

118995

sec

stn

LAM4520O208

pm

PM_LAM4520

866463

sec

stn

LAM4520O213

pm

PM_LAM4520

588507

sec

stn

LAM4520O209

pm

PM_LAM4520

1.03E+06

sec

stn

LAM4520O210

pm

PM_LAM4520

56952.7

sec

stn

LAM4520O306

pm

PM_LAM4520

998199

sec

stn

ALLIAN112

pm

PM_LAM4520

507520

sec

NOTE: The same calendar is associated to more than one station.

FOAUNITS

MTTRUNITS hrs

Calendar files update Probabilistic approach Calendar update based on inferential statistics techniques on equipment data found in hystorical repository ƒ Extraction of golden tools data related to failures and preventative maintenance, from Fab Performance Viewer, FPV framework archive ƒ Application of Kolmogorov- Smirnov test to golden tools data, to determine probability distribution related and associated parameters ƒ Update STAP calendars with the calculated parameters

22

Probabilistic approach Data extraction from FPV archive EQP_ID

EQP_NAME

EQP_ID

OLD_STATE

NEW_STATE

TRANSACTION_INSTANT

STATUS

1070

ALLIAN112

482

DOWN

UP

27-Nov-2010 03:26:45 PM

STAND-BY

482

LAM4520O207

482

UP

DOWN

27-Nov-2010 01:54:49 PM

UNSCHEDULED

715

LAM4520O208

485

UP

DOWN

28-Nov-2010 06:25:50 PM

SCHEDULED

1069

LAM4520O209

485

DOWN

UP

27-Nov-2010 04:47:22 PM

STAND-BY

624

LAM4520O210

485

UP

DOWN

27-Nov-2010 12:46:16 PM

UNSCHEDULED

1068

LAM4520O213

485

DOWN

UP

27-Nov-2010 09:43:47 AM

STAND-BY

892

LAM4520O303

895

LAM4520O306

Stations table

up_down_transaction table EQP_ID

DAY

MTBF

MTTR

DOWN_PERCENT

482

25-Nov-2010 12:00:12 AM

33.2769

1.5875

.0477

UNSCHEDULED

876

25-Nov-2010 12:12:51 AM

11.5636

3.1258

.2703

SCHEDULED

626

25-Nov-2010 12:35:28 AM

18.7222

4.3786

.2339

SCHEDULED

728

25-Nov-2010 12:44:10 AM

7.1517

.1236

.0173

UNSCHEDULED

1030

25-Nov-2010 12:48:53 AM

83.9725

2.0039

.0239

UNSCHEDULED

987

25-Nov-2010 12:59:56 AM

16.5936

.465

.028

UNSCHEDULED

820

25-Nov-2010 01:09:35 AM

48.0644

1.2867

.0268

SCHEDULED

1140

25-Nov-2010 01:23:17 AM

54.0175

1.4156

.0262

SCHEDULED

down_data table

STATUS

Probabilistic approach Kolmogorov-Smirnov test (1/4) ƒ “goodness-of-fit” test, proposed by Kolmogorov in 1933 and developed by Smirnov ƒ compare a sample with a reference probability distribution ƒ The Kolmogorov–Smirnov statistic quantifies the distance between the empirical distribution function of the sample and the cumulative distribution function of the reference distribution ƒ The computed distance will be compared to a threshold value to verify the null hypothesis that the samples are drawn from the reference distribution. 24

Probabilistic approach Kolmogorov-Smirnov test (2/4) ƒ X1,...,XN – N Independent and identically-distributed random variables ƒ Empirical distribution function SN for N iid observations Xi is defined as SN (X i ) =

1 Fi N

Where Fi is the number of observations ≤ Xi

ƒ F0(.) - completely specified cumulative distribution function ƒ F0(Xi) – expresses the expected value of samples observed ) = QKS

([

])

N + 0.12 + 0.11 / N D

Where QKS is defined as: ∞

Q KS (λ ) = 2∑ (− 1)

j −1

e

− 2 j 2 λ2

j =1

is a monotonic function with: QKS(0)=1

QKS(∞)=0

27

Probabilistic approach Application of K-S test to data extracted from FPV (1/4) MTTR results for golden tools failures Test significance

Distribution type

Distribution parameters

Station

# data

Avg

Standard deviation

D gamma

D exponential

D Weibull

threshold

Gamma

Exponentia l

Weibull

Gamma

Exponentia l

Weibull

α

β

LAM4520O207

78

3.49

8.88

0.987179

0.433173

0.20363

0.153763

0

0

0.002552

no

no

no

0.454545

1.439969

LAM4520O303

108

6.32

34.78

0.990741

0.529611

0.401984

0.130674

0

0

0

no

no

no

0.298507

0.667273

LAM4520O208

102

5.46

9.94

0.990196

0.317283

0.167916

0.134462

0

0

0.005468

no

no

no

0.588235

3.535006

LAM4520O213

123

5.55

33.07

0.99187

0.548473

0.351509

0.122447

0

0

0

no

no

no

0.285714

0.476918

LAM4520O209

116

2.91

14.12

0.991379

0.507595

0.435371

0.126087

0

0

0

no

no

no

0.31746

0.40053

LAM4520O210

135

3.57

9.16

0.992593

0.328891

0.229257

0.116878

0

0

0.000001

no

no

no

0.454545

1.472084

LAM4520O306

111

7

54.42

0.990991

0.676216

0.495541

0.128896

0

0

0

no

no

no

0.25641

0.339019

ALLIAN112

59

1.75

2.77

0.983051

0.28751

0.253574

0.176797

0

0.000082

0.000776

no

no

no

0.645161

1.268947

Lowest K-S test

28

Probabilistic approach Application of K-S test to data extracted from FPV (2/4) MTTF results for golden tools failures

Test significance

Distribution type

Distribution parameters

Station

# data

Avg

Standard deviation

D gamma

D exponential

D Weibull

threshold

Gamma

Exponential

Weibull

Gamma

Exponential

Weibull

α

β

LAM4520O207

78

16.51

31.38

0.987179

0.190428

0.083977

0.153763

0

0.005888

0.622476

no

no

yes

0.571429

10.26469

LAM4520O303

108

17.07

53.34

0.990741

0.178551

0.280952

0.130674

0

0.001717

0

no

no

no

-

17.07

LAM4520O208

102

28.54

53.24

0.990196

0.186742

0.094751

0.134462

0

0.001352

0.304325

no

no

yes

0.571429

17.74648

LAM4520O213

123

12.65

21.51

0.99187

0.169067

0.086951

0.122447

0

0.001497

0.297043

no

no

yes

0.625

8.847901

LAM4520O209

116

11.46

33.31

0.991379

0.140527

0.26001

0.126087

0

0.01832

0

no

no

no

-

11.46

LAM4520O210

135

9.28

10.44

0.140484

0.106748

0.09012

0.116878

0

0.086074

0.212239

no

no

yes

0.909091

8.867484

LAM4520O306

111

11.31

20.61

0.990991

0.109366

0.136451

0.128896

0

0.131517

0.028919

no

yes

no

-

11.31

ALLIAN112

59

21.36

54.01

0.983051

0.221482

0.21944

0.176797

0

0.004994

0.005575

no

no

no

0.454545

0.81333

Lowest K-S test

29

Probabilistic approach Application of K-S test to data extracted from FPV (3/4) MTTR results for golden tools PMs Test significance Station

# data

Avg

Standard deviation

D gamma

Distribution type

Distribution parameters

D exponential

D Weibull

threshold

Gamma

Exponential

Weibull

Gamma

Exponential

Weibull

α

β

LAM4520O207

194

2.85

8.22

0.994845

0.464547

0.280585

0.097499

0

0

0

no

no

no

0.425532

1.009563

LAM4520O303

197

3

9.95

0.994924

0.472284

0.320832

0.096753

0

0

0

no

no

no

0.384615

0.809296

LAM4520O208

118

3.64

6.98

0.991525

0.360658

0.216121

0.125014

0

0

0.000025

no

no

no

0.555556

2.169273

LAM4520O213

276

2.73

7.64

0.996377

0.454566

0.178707

0.081742

0

0

0

no

no

no

0.425532

0.964051

LAM4520O209

285

2.55

6.78

0.996491

0.495555

0.204094

0.080441

0

0

0

no

no

no

0.444444

0.999823

LAM4520O210

227

4.65

26.78

0.995595

0.456195

0.456734

0.090134

0

0

0

no

no

no

-

4.65

LAM4520O306

233

2.68

7.27

0.995708

0.380725

0.251248

0.088966

0

0

0

no

no

no

0.434783

0.998243

ALLIAN112

184

3.76

28.83

0.994565

0.52887

0.622604

0.100113

0

0

0

no

no

no

-

3.76

Lowest K-S test

30

Probabilistic approach Application of K-S test to data extracted from FPV (4/4) MTBPM results for golden tools PMs

Test significance

Distribution parameters

Distribution type

Station

# data

Avg

Standard deviation

D gamma

D exponential

D Weibull

threshold

Gamma

Exponenti al

Weibull

Gamma

Exponential

Weibull

α

β

LAM4520O207

194

27.06

33.74

0.310447

0.165252

0.243772

0.097499

0

0.000041

0

no

no

no

-

27.06

LAM4520O303

197

21.30

20.8

0.163803

0.149983

0.149983

0.096753

0.00042

0.000241

0.000241

no

no

no

1

21.296

LAM4520O208

118

28.3

30.57

0.161027

0.104668

0.118599

0.125014

0.003796

0.141556

0.066777

no

yes

no

-

28.3

LAM4520O213

276

16.5

18.78

0.186568

0.155111

0.141913

0.081742

0

0.000003

0.00025

no

no

no

0.869565

15.382

LAM4520O209

285

17.44

16.96

0.170783

0.151519

0.148993

0.080441

0

0.000003

0.000005

no

no

no

1.052632

17.794

LAM4520O210

227

22.31

29.35

0.366045

0.167081

0.21617

0.090134

0

0.000005

0

no

no

no

-

22.31

LAM4520O306

233

20.33

18.86

0.211389

0.156893

0.150686

0.088966

0

0.000017

0.000043

no

no

no

1.052632

20.750

ALLIAN112

184

29.47

32.5

0.22515

0.250833

0.277687

0.100113

0

0

0

no

no

no

0.822394

0.027

Lowest K-S test

31

Probabilistic approach STAP calendars update (1/2) Equipment failures calendar DOWNCALNAME

DOWNCALTYPE

MTTFDIST

MTTF

MTTF2

MTTFUNITS

MTTRDIST

MTTR

MTTR2

MTTRUNITS

DN_LAM4520O207

mttf_by_cal

weibull

0.571429

10.26469

hr

weibull

0.454545

1.439969

hr

DN_LAM4520O303

mttf_by_cal

exponential

17.07

hr

weibull

0.298507

0.667273

hr

DN_LAM4520O208

mttf_by_cal

weibull

0.571429

17.74648

hr

weibull

0.588235

3.535006

hr

DN_LAM4520O213

mttf_by_cal

weibull

0.625

8.847901

hr

weibull

0.285714

0.476918

hr

DN_LAM4520O209

mttf_by_cal

exponential

11.46

hr

weibull

0.31746

0.40053

hr

DN_LAM4520O210

mttf_by_cal

weibull

0.909091

hr

weibull

0.454545

1.472084

hr

DN_LAM4520O306

mttf_by_cal

exponential

11.31

hr

weibull

0.25641

0.339019

hr

DN_ALLIAN112

mttf_by_cal

weibull

0.454545

hr

weibull

0.645161

1.268947

hr

8.867484

0.81333

Association of calendar file to single stations RESTYPE

RESNAME

CALTYPE

CALNAME

FOADIST

FOA

FOA2

FOAUNITS

stn

LAM4520O207

down

DN_LAM4520O207

weibull

0.571429

10.26469

hr

stn

LAM4520O303

down

DN_LAM4520O303

exponential

stn

LAM4520O208

down

DN_LAM4520O208

weibull

0.571429

17.74648

hr

stn

LAM4520O213

down

DN_LAM4520O213

weibull

0.625

8.847901

hr

stn

LAM4520O209

down

DN_LAM4520O209

exponential

11.46

stn

LAM4520O210

down

DN_LAM4520O210

weibull

stn

LAM4520O306

down

DN_LAM4520O306

exponential

stn

ALLIAN112

down

DN_ALLIAN112

weibull

17.07

0.909091

hr

hr 8.867484

11.31 0.454545

hr hr

0.81333

hr

32 NOTE: Each station has its own calendar

Probabilistic approach STAP calendars update (2/2) Equipment PM calendar PMCALNAME

PMCALTYPE

MTBPMDIST

MTBPM

PM_LAM4520O207

mtbpm_by_cal

exponential

27.06

PM_LAM4520O303

mtbpm_by_cal

weibull

1

PM_LAM4520O208

mtbpm_by_cal

exponential

28.3

PM_LAM4520O213

mtbpm_by_cal

weibull

0.869565

PM_LAM4520O209

mtbpm_by_cal

weibull

1.052632

PM_LAM4520O210

mtbpm_by_cal

exponential

22.31

PM_LAM4520O306

mtbpm_by_cal

weibull

1.052632

PM_ALLIAN112

mtbpm_by_cal

gamma

0.822394

MTBPM2

MTBPMUNITS

MTTRDIST

MTTR

MTTR2

MTTRUNITS

hr

weibull

0.425532

1.009563

hr

hr

weibull

0.384615

0.809296

hr

hr

weibull

0.555556

2.169273

hr

15.38174

hr

weibull

0.425532

0.964051

hr

17.79451

hr

weibull

0.444444

0.999823

hr

hr

exponential

4.65

20.75067

hr

weibull

0.434783

0.027905

hr

exponential

3.76

21.296

hr 0.998243

hr

Association of calendar file to single stations RESTYPE

RESNAME

CALTYPE

CALNAME

FOADIST

stn

LAM4520O207

pm

PM_LAM4520O207

exponential

stn

LAM4520O303

pm

PM_LAM4520O303

weibull

stn

LAM4520O208

pm

PM_LAM4520O208

exponential

stn

LAM4520O213

pm

PM_LAM4520O213

weibull

0.869565

15.38174

hr

stn

LAM4520O209

pm

PM_LAM4520O209

weibull

1.052632

17.79451

hr

stn

LAM4520O210

pm

PM_LAM4520O210

exponential

stn

LAM4520O306

pm

PM_LAM4520O306

weibull

1.052632

20.75067

hr

stn

ALLIAN112

pm

PM_ALLIAN112

gamma

0.822394

0.027905

hr

NOTE: Each station has its own calendar

FOA

FOA2

27.06 1

FOAUNITS hr

21.296

28.3

hr hr

22.31

hr

hr

33

Results – application of KS-TEST to golden tools ƒ 4 days run horizon ƒ 8 Golden tools in ETCHING area ƒ Parameters used as reference for comparison ƒ Transactions up-down and viceversa (occurrence and duration) ƒ Moves – transition of a wafer from one operation to the next one

34

Results – golden tools (1/2) 30 25 20 15 10 1600

5

1400

0

1200

down %

1000

fails #

800 600

6000

400 200

5000

0 Average MTTR (hrs)

4000

Actual 3000

Sim_before Sim_after

2000 1000 0 Total MTTR (hrs)

Total MTTF (hrs)

Results – golden tools (2/2) 1500

18000

1450

17500

1400

17000

1350

16500

Moves simul_beforemoves actual

1300 1250

Moves simul_aftermoves actual

1200

Actual

16000

sim_before sim_after

15500

1150

15000

1100

14500

1050 1000

14000 1

1

140

95

120

94.5

100 80 60

94 93.5

92.5

40

92

20

91.5

0

91

25 /0 1/ 20 25 08 /0 1 1/ 20 25 08 /0 2 1/ 20 26 08 /0 3 1/ 20 26 08 /0 1 1/ 20 26 08 /0 2 1/ 27 200 8 /0 3 1/ 20 27 08 /0 1 1/ 20 28 08 /0 2 1/ 20 28 08 /0 1 1/ 20 28 08 /0 2 1/ 20 08 3

Adherence sim_before

93

Adherence sim_after

90.5 90 1

Results – application of KS-TEST to whole simulation model ƒ 4 days run horizon ƒ ~ 500 stations in 10 homogeneous areas ƒ Parameters used as reference for comparison ƒ By station ƒ PCCOMPS – number of wafers processed by station ƒ Down and PM % per shift

ƒ Moves by area

37

Results – whole model % stations adherent to reality, by area

FOTOATT

FOTOSVI

DIFF

METAL

Sim_before

36%

27%

8%

36%

Sim_after

64%

73%

92%

64%

Results – whole model moves by area and by shift

Shift 1

Shift 2

FOTOATT

FOTOSVI

METAL

FOTOATT

FOTOSVI

METAL

Sim_before

14640

8370

8730

15112

8193

9341

Sim_after

14625

8330

8551

13754

7856

7932

Actual

17052

9491

10887

12902

7824

7629

Results – whole model PCCOMPS by station - figures

Results – whole model DOWN and PM % by station - figures

Benefits ƒ Simulation nearer to reality ƒ Number and frequency of transitions and up and down times are next to reality ƒ Moves target better estimated (not over-estimated) ƒ Better Adherence

ƒ Deterministic approach drawbacks ƒ Data manual update is not always based on correct data and executed at right times ƒ Does not consider products mix variability

ƒ Probabilistic approach advantages ƒ Weekly data update based on historical equipment behavior ƒ Real-time data ƒ Better usage of simulator potential 42