Prognostic/Diagnostic Health Management (PHM) System for FAB Efficiency

woundcallousΗμιαγωγοί

1 Νοε 2013 (πριν από 3 χρόνια και 5 μήνες)

142 εμφανίσεις

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
1

1

Prognostic/Diagnostic

Health Management (PHM) System

for

FAB Efficiency


Chin Sun




chin
@globalcybersoft.com




IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
2

2

Outline


Introduction


Industry Trend


PHM


What?


Method


Results


Conclusion


IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
3

3

APC/AEC 2005 / Samsung Electronics Co., Ltd.

"An Application
of Multivariate Statistics in Detecting Equipment
Changes"

Presenter: Lee, Seungjun

Industry Trend:

APC/AEC 2005 Presentation from Samsung

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
4

4

Industry Trend:

APC/AEC 2005 Presentation from Samsung

APC/AEC 2005 / Samsung Electronics Co., Ltd.

"An Application
of Multivariate Statistics in Detecting Equipment
Changes"

Presenter: Lee, Seungjun

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
5

5

Industry Trend:

APC/AEC 2005 Presentation from Helix Tech.

APC/AEC 2005 /
Helix Technology Corporation

"Predictive Capability
Enabled by a Deterministic Method of Analysis or Real World Vacuum
System e
-
Diagnostics"

Presenter: Gaudet, Peter



IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
6

6

Industry Trend:

APC/AEC 2005 Presentation from Helix Tech.

APC/AEC 2005 /
Helix Technology Corporation

"Predictive Capability
Enabled by a Deterministic Method of Analysis or Real World Vacuum
System e
-
Diagnostics"

Presenter: Gaudet, Peter





IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
7

7

Industry Trend:

APC/AEC 2005 Presentation from Adventa

APC/AEC 2005 /
"Reaping the Benefits of Heuristic Fault
Modeling"

Presenter: Jared Warren,

Adventa Control Technologies







IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
8

8

Industry Trend: Weighted FDC

APC/AEC 2005 Presentation from Intel

APC/AEC 2005 /
Intel Corporation

"Weighted Fault Detection
and Classification"

Presenter: Mao, John









IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
9

9

APC/AEC 2005 /
Intel Corporation

"Weighted Fault Detection
and Classification"

Presenter: Mao, John









Industry Trend: Weighted FDC

APC/AEC 2005 Presentation from Intel

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
10

10

FDC

PHM
-
Equip

The Evolution of Quality Control

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
11

11


Opportunities of Human Errors:

Labor intensive and Time
consuming


Passive Approach:

No knowledge sharing or self learning,
lacking of predictive capability


Inconsistency:

Analysis results are human dependent


Cost of Resources:

Delay Time
-
to
-
Corrective Actions,
Long training time for new engineers

Host

Slow Trouble Shooting Process

?

Equipment Engineers

CONVENTIONAL e
-
Diagnostic APPROACH

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
12

12

Host +
PHM
-
Equip

Enable Real Time Auto
-
Diagnostic

Management

Equipment Engineers


Reduce or Eliminate potential Human Errors:

Automated, Knowledge
based Analysis


Feed Forward ↔ Feed Backward

Proactive Approach:

Enable
Knowledge Sharing, Self Correction, and providing Predictive Capability


Consistency:

Analysis results are based on Data and Knowledge


Saving Resources:

Fast Time
-
to
-
Corrective Actions, Shorten training
time for new engineers

AUTOMATED e
-
Diagnostic APPROACH

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
13

13

Equipment Manufacturers

PHM
-
Equip Systems


Equip. Engr. A

PHM
-
Equip Client

Equip. Engr. C

PHM
-
Equip Client

PHM DVP

Servers

PHM Production

Servers

Verified Rules
Transfer

NO DTC

Scenarios


PHM
-
Equip will help resolve
NO DTC (Diagnositc Troub
-
shooting Code)

problems


PHM
-
Equip will help resolve
DTC False Alarm

problems


PHM
-
Equip will accumulate
Prognostic Rules

from experienced equipment
engineers

Equip. Engr. B

PHM
-
Equip Client

Prognostic Rules

upload

e.g. failing oxygen

sensors

A global Internet
-
based
collaborative
Knowledge Base
accumulation and
sharing
environment

DTC False

Alarm

Scenarios

Knowledge is power,
but only when it is
shared

PHM
-
Equip Infrastructure

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
14

14

Fab equipment sets

E
-
TEST

Equipment Engineers

populate PHM
-
Equip with
equipment rules based on
their knowledge

Yield/Product
Engineers


populate PHM
-
BE
with feedback

rules based on
previous analysis

Gate

Ox

Vt

implant

Litho

APC

Fab Processes

PHM
-
Equip

PHM
-
FAB

PHM
-
BE

Process Engineers

populate PHM
-
FAB
with APC rules based
on their knowledge

FAB Front End

FAB Back End

Process
Flow

Fab equipment sets

PHM
-
INT

PHM
-
E2

PHM
-
E1

PHM
-
F1

PHM
-
F2

KNOWLEDGE
BASES


Wafer
S
ort

/Final

T
est

D
efect
D
ensity

R
eduction

PHM
-
Etest

PHM
-
DDR

PHM
-
BEST

KB Info
feed
forward/

feed
backward
thru
entire
process
flow


Device Info

Process Info

PHM
-
INT, PHM
-
Equipment, PHM
-
FAB & PHM
-
BE

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
15

15



Western Electrical (WE) control charts with pattern recognition
capability + Multivariate FDC to identify out of control tool parameters



Advanced Real Time
-
Knowledge Management (RT
-
KM) Rule
-
based
methodology automatically determine when an equipment fault occurs,
what caused it, and how to correct it

RT
-
KM

Engine

Equip/Tool

Fault

Classification

Data

Fault

Cause

RT
-
KM Rule
-
Based

Root Causes

Identification


Fast Corrective Action

Tool

Diagnostic

Report

PHM
-
Equip

Multivariate Mahalanobis
Distance Fault Detection
Engine


PHM
-
Equip Architecture

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
16

16


Shorten time

to analyze data to validate decisions


Automatic Equipment diagnostics


Increase Engineers
’ productivity and efficiency.


Resolve equipment malfunction problems faster


Use Knowledge system as a continuous learning tool


Integrated Knowledge base/Database

optimized for Ediag data


Fast, simple access to diagnostic report


Facilitates collaboration among different FAB equipment
engineers


Versatile and interactive
Rule development tool


Worksheet based


Easy to use


Rules are specific for Equipment Diagnostic analysis


Highlighted Features

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
17

17


Improved
Time To Market and
Reduce the
waste of
manpower


Enable effective
Knowledge Sharing



Utilize

Engineering Knowledge in FDC to have more
accurate detection


Enable
Real
-
time feedback
,
Continuous Improvement


Eliminate

False Alarms


Reduce

scrapped/low performance wafers


Enable

24x7

Equipment Process Monitoring


Capable of Supporting

multiple Equipment


Reduce

engineers’ pressure, increase productivity and
efficiency


Permanent repository of
Knowledge and Expertise


Highlighted Benefits

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
18

18



Real time

feedback of
Equipment & Process Status




Automatically Identify

equipment
malfunctions/Process misprocessing



Real time

feedback of
Diagnostic/Prognostic reports




Knowledge

retained in database, never lost


Fast Time
-
to
-
Corrective Actions

and

Enabling Continuous Improvement

PHM
-
Equip Solutions

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
19

19


Knowledge Based

Methodology


On
-
line, Real time


Auto diagnostics/prognostics


Permanent repository for knowledge


24 X 7

Equipment monitoring


Enable Global e
-
Diagnostic

Conclusions


In Summary
:

PHM
-
Equip provides an innovative
methodology for Equipment Control. PHM
-
Equip

enables
continuous improvement in the day
-
to
-
day Operation of
Equipment. As the results, PHM
-
Equip presents numerous
possibilities to improve the Overall Equipment Efficiency (OEE)

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
20

20

Method: Mahalanobis Distance

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
21

21

Method: Mahalanobis Distance

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
22

22

Method: Mahalanobis
-
Taguchi System

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
23

23

Method: Mahalanobis
-
Taguchi System

A Multidimensional diagnosis system

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
24

24

Where S
i
= standard deviations of i


th variable,

C
-
1

= the inverse of correlation matrix,

k = number of variables,

n = number of observations,

T = transpose of the standard vector.


Method: Mahalanobis
-
Taguchi System

A Multidimensional diagnosis system

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
25

25

PHM
-
Equip Examples: Data Source

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
26

26

(OES) Optical Emission

Spectroscopy

wavelength monitored


250 nm


261.8 nm


266.6 nm


272.2 nm


278.3 nm


284.6 nm


288.25 nm


…..


791.5 nm

PHM
-
Equip Examples: Data Source

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
27

27


Step 1: Define the Problem


Results: Fault Detection

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
28

28


Step 2: Define Control/Response Variables


(OES) Optical Emission

Spectroscopy

wavelength monitored


250 nm


261.8 nm


266.6 nm


272.2 nm


278.3 nm


284.6 nm


288.25 nm


…..


791.5 nm

(MD) Mahalanobis
Distance

Results: Fault Detection

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
29

29


Step 3:
Construct the “Full Model MTS
Measurement Scale”

Note: The measurement scale is constructed by training
datasets




Results: Fault Detection

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
30

30


Step 4:
Validate the ability of measurement
scale

Note: the capability of measurement scale is demonstrated by
test datasets.




Results: Fault Detection

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
31

31

Results: Fault Classification

Method: Distinguish the signal pattern shift of each
variable between the test dataset and the model

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
32

32

Results: Fault Classification

Results: Test wafer 2 and test wafer 18 have the same four
machine state variables associated with the RF
-
12 system
fault.


IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
33

33

Create Diagnostic Rule from pattern
signature

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
34

34

PHM Fault Detection and Classification


Real Time
FDC
Monitor
Window

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
35

35

PHM Fault Detection and Classification

Report

Root cause of
equipment
malfunction
and PIDs
associated
with faults

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
36

36

PHM Fault Detection and Classification

Summary

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
37

37

Promote Predictive Maintenance

Example of Prognostic Rule for oxygen sensor


IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
38

38

Normal
process
patterns

PHM
-
Equip Example: Diagnostic Results

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
39

39

Progressive
degrading
Operating
patterns can
be used to
generate
prognostic
pattern
recgonition
rules

PHM
-
Equip Example: Diagnostic Results

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
40

40

1. Normal

2. Predictive

Monitoring
started

3. Recommand
preventive maintenance
(PM) in 48 hr

Monitoring started

4. Do not
commence

processing

5. Stop

Processing


State
-
based Warning System

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
41

41

Fab equipment sets

E
-
TEST

Equipment Engineers

populate PHM
-
Equip with
equipment rules based on
their knowledge

Yield/Product
Engineers


populate PHM
-
BE
with feedback

rules based on
previous analysis

Gate

Ox

Vt

implant

Litho

APC

Fab Processes

PHM
-
Equip

PHM
-
FAB

PHM
-
BE

Process Engineers

populate PHM
-
FAB
with APC rules based
on their knowledge

FAB Front End

FAB Back End

Process
Flow

Fab equipment sets

PHM
-
INT

PHM
-
E2

PHM
-
E1

PHM
-
F1

PHM
-
F2

KNOWLEDGE
BASES


Wafer
S
ort

/Final

T
est

D
efect
D
ensity

R
eduction

PHM
-
Etest

PHM
-
DDR

PHM
-
BEST

KB Info
feed
forward/

feed
backward
thru
entire
process
flow


Device Info

Process Info

PHM
-
INT, PHM
-
Equipment, PHM
-
FAB & PHM
-
BE

IEEE/SEMI Advanced Semiconductor Manufacturing Conference

May 22
-
24, 2006

ASMC 2006


Boston, Massachusetts



C. Sun Slide
42

42

PHM VALUE PROPOSITION


Provide Predictive Equipment Maintenance
& Diagnostics


Correct Problems before failure occurs


Real time process/tool/equipment health
feedback


Pinpoints miss processing/equipment
malfunction steps


Diagnostic report feeds backward


Diagnostic report feeds forward


Knowledge reusable, never lost