WSC 2008 Final Abstracts |
Wednesday 8:30:00 AM 10:00:00 AM
Statistical Methods
Chair:
Shi-Shang Jang (National Tsing-Hua University)
Demand Forecast of Semiconductor Products based on
Technology Diffusion
Chen-Fu Chien and Yun-Ju Chen (National Tsing
Hua University) and Jin-Tang Peng (Yuanpei University)
Abstract:
Demand forecast plays a critical role to determine
capital investment for capacity planning. Given the involved uncertainties and
long lead-time for capacity expansion, semiconductor companies have to predict
future demands as a basis for related manufacturing strategic decisions. As
semiconductor products in a consumer era become more diversified with
shortening life cycle, demand forecast also becomes more complex and
difficult. This study aims to develop a demand forecast model based on product
life cycle and technology diffusion. While little research has been done to
employ diffusion models to forecast the demands of semiconductor products. The
proposed model modifies a multi-generation diffusion model incorporating
marketing variables into the model for semiconductor product and uses
nonlinear least square method to estimate the parameters. An empirical study
is conducted to validate the proposed model with real data of semiconductor
products. This research concludes with discussion on future research
directions.
A Bayesian Framework to Integrate Knowledge-Based and
Data-Driven Inference Tools for Reliable Yield Diagnoses
Chih-Min
Fan (Yuan Ze University)
Abstract:
This paper studies the issues of designing a Bayesian
framework for the reliable diagnosis of various yield-loss factors induced in
semiconductor manufacturing. The proposed framework integrates both the
results from knowledge-based and data-driven inference tools as input data,
where the former resembles expert's knowledge on diagnosing pre-known
yield-loss factors while the latter serves for exploring new yield-loss
factors. Three modules with specific designs for yield diagnosis applications
are addressed: Pre-Process for generating candidate factors and corresponding
prior distributions, Bayesian Inference for calculating posterior
distributions, and Post-Process for deriving reliable rankings of candidate
factors. The final output, a Bubble Diagram with Pareto Frontier, provides
visual interpretations on the integral results from data-driven,
knowledge-based and Bayesian inference tools. Specific issues addressed in the
proposed Bayesian framework provide directions for implementing a real system.
Systematic Applications of Multivariate Analysis to
Monitoring of Equipment Health in Semiconductor Manufacturing
A.G.
Chao (National Tsing-Hua Unversity), S.P. Lee (Tsing-Yu University) and S.T.
Tseng, David, S.H. Wong, and Shi-Shang Jang (National Tsing-Hua University)
Abstract:
In this work, a systematic procedure of building a
model for monitoring batch processes in semiconductor manufacturing and
visualization of monitoring results will be presented. Semiconductor
manufacturing batch-processing stages usually consist of many steps. Aging
trends likely to be detected only in the steady state period of each step.
Large fluctuations can be found in on-off period of each step. Hence
"step-trend" variables, i.e. mean shifts from reference profiles of each
steps, are defined to track aging trends. Residuals from this shifted profile
are then used to provide a combined health index of each batch through
Hotelling T2 analysis.
Wednesday 10:30:00 AM 12:00:00 PM
Simulation Technologies
Chair: Leon McGinnis (Georgia Institute of
Technology)
Automated Generation and Parameterization of
Throughput Models for Semiconductor Tools
Jan Lange (AMD Saxony LLC
& Co. KG), Oliver Rose (Dresden University of Technology) and Kilian
Schmidt and Roy Boerner (AMD Saxony LLC & Co. KG)
Abstract:
Cluster tools play an important role in modern
semiconductor fabs. Due to their complexity in configuration and their varying
material flow, the creation of accurate throughput models for cluster tools is
a demanding task. Proposed analytical approaches are either quite intricate
and require manual maintenance process, or are inexact due to reliance on
lot-level events. This paper presents an approach for the automated generation
and parameterization of detailed cluster tool models based on bottleneck
analysis. Equipment configuration as well as all throughput model base data is
extracted from recent equipment reporting data.
Toward On-Demand Wafer Fab Simulation Using
Formal Structure & Behavior Models
Edward Huang, KySang Kwon,
and Leon McGinnis (Georgia Institute of Technology)
Abstract:
Contemporary factories in capital intensive industries
such as semiconductor manufacturing are very complex, with many sources of
risk. The highly competitive and global business environment forces companies
to analyze, design, and continuously re-design factories with distributed
multi-disciplinary teams. Traditional factory design approaches using
spreadsheets and stand-alone simulations cannot adequately cope with the
resulting time, cost, and risk requirements. In this paper, we address the
opportunity to achieve support fab design teams by providing on-demand
simulation. The method for achieving this combines for-mal fab descriptive
models with a process for generating fab analysis models from relatively
standard sources of fab data.
Using OMG's SysML to Support
Simulation
Christiaan J.J. Paredis and Thomas Johnson (Georgia
Institute of Technology)
Abstract:
Currently, system engineering problems are solved using
a wide range of domain-specific models and corresponding languages. It is
unlikely that a single unified modeling language will be able to model in
sufficient detail the large number of system aspects addressed by these
domain-specific languages. Instead, a model integration framework is needed
for managing the various modeling languages used to solve systems engineering
problems. The Systems Modeling Language (OMG SysMLTM) can provide an answer to
this need for model integration. Using SysML, a modeler can abstract a
domain-specific language to a level that permits its interaction with other
system models. In addition, graph transformation approach can be use to
accomplishing automated, bidirectional transformation between SysML and the
domain specific language. In this paper, a generic approach for defining such
graph transformations is presented.