Research Motivation

1
Research Motivation
2
??????????????????????

• ? ? ?
• M.-D. Hu, S.-C. Chang, Translating Overall
  Production Goals into Distributed Flow Control
  Parameters for Semiconductor Manufacturing,
  Journal of ManufacturingSystems, Vol. 22, No. 1,
  46-63, 2003. NSC-89-2212-E-002-040
• ??????????????

3
Outlines

• Research Motivation
• OQN Modeling
• Decomposition-Based OQN Analysis
• Translation Problem Solution BQNA
• Numerical Experiments
• Conclusions Future Research

4
Research Motivation
5
Wafer Fabrication

• Process Flow Highly Re-entrant
• Fabricate an Oxide Layer
• Machine Unreliable
• High Capital Investment
• High Product Add-on Value / Short Life Cycle

6
PFC Activities in a Fab
7
PFC Process Logic in a Fab

• Modeling

• Performance Analysis

• Control

8
Need Translation Methodology
9
Define Generic Translation Problem
10
OQN Modeling
11
OQN Model of a Re-entrant Line

• Node Group of Identical Failure Prone Machines

• Queue Infinite Buffer for each Step

• Job Class Part Type

• Arrival General Independent Processes

• Service General Time Distribution

• Routing Deterministic with Feedback

• Discipline First-Come-First-Serve

12
OQN Modeling Summary

• Considered Features
• Multiple Part types
• Multi-Server Node
• Re-entrant Deterministic Routing
• Failure Prone Machine
• Batch Server/ Batch Arrival
• Not Modeled
• Scrap
• Rework
• Process Yield
• Stability

• Lot Size Change
• Probabilistic routing

13
Decomposition-BasedOQN Analysis
14
Decomposition Approximation

• Two Notions

• Each Network Node as an Independent GI/G/m Queue

• Two Parameters, Mean SCV, to Characterize
  Arrival Service Processes

15
Approximated by 3 Basic Operations
16
Two Traffic Equations
17
Decomposition Analysis Summary
18
Translation Problem Solution BQNA
19
Estimate Performance Measures

Performance Measures
Arrival Parameters
20
Translation Problem Solution BQNA
Flow Control Parameters
Production Goals (d, T)

How to Achieve the Goals (d, T)?
21
General BQNA Results

• Desired Arrival Rate (1st order)
• Bounds on Inter-Arrival Time SCV (2nd order)

22
BQNA Summary

• Translation Problem Solution
• External Arrival Parameters
• Internal Arrival Parameters

• PFC Application
• External Arrival Parameters
• Internal Arrival Parameters

• Wafer Release Control

• Nodal Level WIP/CT Measures

• Dispatching Decision Reference

• Apply to Existing PFC Schemes

23
Numerical Experiments
24
Numerical Validation Flow
Overall System Production Goals
25
FAB1 Experiment Example

• Overall Production Goals
• Desired Output Rate 0.52 lots/hr
• Target Mean CT 383.25 hours
• Model Characteristics
• Lu, Ramaswamy and Kumar (1994)
• Single Product Type
• 12 Machine Groups
• Total 40 Failure Prone Machines
• 60 Steps for a Complete Process
• Exponential Time Distribution
• Average Loading Intensity over 90

26
FAB1 Product Process Flow
27
FAB1 Machine Group Data
28
FAB1 - Nodal Cycle Time Results
29
FAB1 - Nodal WIP Results
30
FAB1 - System Level Results
31
FAB2 Experiment Example

• Overall Production Goals
• Desired Output Rate 10 lots/hr
• Target Mean CT 40.64 hours
• Model Characteristics
• Bitran and Tirupati (1988)
• 10-Product Fab Model
• 13 Machine Groups
• Different Processing Time Distribution

32
FAB2 Product Routing Sequence
33
FAB2 Machine Group Data
34
FAB2 Mean Cycle Time Results
35
FAB2 Cycle Time Standard Deviation
36
FAB2 - System Level Results
37
Experiments Summary

• In the Simulation
• Wafer Release follows the derived external
  parameters.
• FCFS is the service discipline of each node.
• Discussions
• WIP CT mostly within 95 of accuracy at both
  node system
• Less than one millisecond of CPU time to apply
  BQNA to the FAB Model
• Dimension of the linear equations depends only on
  the number of different MGs

performance of individual nodes may serve as
guiding references for real applications
38
Conclusions

• Proposed a Generic Definition of the Translation
  Problem
• Constructed an OQN Model for Failure Prone
  Re-entrant Lines
• Established a Decomposition-Based Analysis
  Procedure for OQN Models
• Developed BQNA for Translating Overall Production
  Goals into Local Control Parameters
• Conducted Numerical Experiments on Two Full-Scale
  Fab Models to Demonstrate the Application
  Potential

39
Future Research

• Enhance the Applicability of BQNA to Include
• Lot Size Change
• Scrapts
• Reworks
• Statistical Production Flow Control
• Statistical Process Control Concept Based
• Developing SPFC Theory
• Developing SPFC Application Methods (SPFC Control
  Charts)

WIP
Upper Bound
Target Level (m)
Lower Bound
Machine