iFind
AIFind Project
- Category:
Research project
- Duration: 2006
- 2010
- Funded by:
ELMOS Central IT Services GmbH
- Project
holder: Institute of Knowledge Based Systems &
Knowledge Management
Integrated circuits (ICs,
chips) are produced by the semiconductor industry on circle
shaped disks called wafers. The highly complex production
is therefore following multiply process steps from which
each could result in defects in the micro structures
through changes in process parameters.
To ensure the quality the chips are tested within and at
the end of the production cycle through electrical
measurements and environment and technology tests. The
information gained through this steps are potential sources
to identify and remove real and potential sources of
defects, but the pure mass of production related data
following the high requirement of monitoring is ranging in
regions that render manual analysis and detection a time
critical and often impossible task.
This already critical data amount is steadily growing and
supplemented by additional but not optional product related
sources. While this is a continues and present problem it
becomes immensely critical if the delivering deadline is
near or if an expert involved in the analyzing process is
leaving the company or became equivalent unavailable.
The ifind (Intelligent Framework for Interrelation
Discovery) system aims to face this problems to aid,
accelerate and enable a contemporary evaluation of
production data. The system provides the possibility to
list and select case related production data and reduce
them to a human manageable size of factors with relations
to the misled factor of the problem occurrence.
Within the discovery process the software is following the
three main processing steps:
- Extraction: To ensure a solid basement for the analyzing steps the related parameter sources are chosen and fused. The user then defines problem classes based on the selected parameters.
- Reduction: Based on a application fit feature selection technique a limited set reduced and related parameter sets is built and created.
- Classification & Evaluation: Based on the problem classes we are now able to classify and evaluate the created sets and present the most fitting selection representing the overall event of the problem occurrence.
Fig. 1 : The ifind system process loop.
Projektpartner: ELMOS Central IT Services GmbH
Contakt
Dr. Christian Weber, christian.weber{at}uni-siegen.de