半导体之解决方案
STATISTICA Solutions for Semiconductor and Automated Manufacturing
One of the most complex as well as expensive automated manufacturing environments is that required for the
manufacture of semiconductors. The typical process involves the nearly fully automated application of hundreds of
processing steps to lots ("stacks") of silicon wafers, each containing a large number of microchips. Creating a high-
yield process, where most (e.g., 90% or more) of all chips pass final acceptance testing is extremely difficult and time
consuming. At the same time, the cost of failure in this environment is significant, as each wafer can be many times
more valuable than even the most precious metal by weight. Moreover, unexpectedly lengthy ramp-up times (to create a
reliable production process) may significantly undercut the commercial value of the final product, hence jeopardizing
the huge investment in the semiconductor Fab, which may well reach $2 Billion dollars or more!
STATISTICA and the Engineering Process
The
STATISTICA system provides a huge set of tools for engineers, to study processes. First, the
STATISTICA system
will quickly and seamlessly integrate into the existing information infrastructure,
querying directly the relevant
databases (practically all industry standard database formats are supported). There is no need to laboriously import
the data into, for example, a limited spreadsheet format for further analyses; instead
STATISTICA connects directly to your data.
Next,
STATISTICA interactive
graphics are extremely fast and flexible, so meaningful views and graphical summaries
of key processes, variables, measurements, outcomes, etc. can be created very quickly.
Complete Customizability and Programmability
Each process is unique, and the techniques for automated manufacturing are constantly evolving in this highly
competitive environment.
STATISTICA is fully customizable and
programmable, down to all aspects of graphs, data
handling, and so on. Hence, in addition to providing an extremely sophisticated and flexible off-the-shelf tool, the
system also serves as a toolbox that will enable engineers to develop custom analyses and processes quickly, to
support the critical ramp-up of new manufacturing processes, and the specialized analytic tools to support them.
Advanced Data Mining and Predictive Quality Control
STATISTICA Data Miner provides an extremely comprehensive set of knowledge discovery algorithms that can be
applied to support the manufacturing process. In addition to commonly used advanced neural network architectures,
STATISTICA Data Miner implements the most cutting edge tools in a single integrated platform. For example, the
system includes algorithms such as
stochastic gradient boosting, random forests,
support vector machines,
multivariate adaptive regression splines (MARSplines),
independent components analysis, to name a few. These
techniques can be used to build robust and reliable predictive models for quality or failure, even in high-dimensional
environments with large numbers of variables (but few "cases" or "rows") and significant interactions between them
(e.g., interactions between tools). All of these methods are implemented in the same efficient and programmable
STATISTICA platform, yielding the most advanced set of tools for tackling difficult root-cause analysis and
predictive QC problems.
STATISTICA Data Miner and KLA-Tencor
STATISTICA Data Miner and other statistical analysis algorithms are used to provide the core support in
KLA-Tencor's
yield analysis and management systems. Because StatSoft is the recognized leader in the application of advanced,
cutting-edge data mining techniques, KLA-Tencor has chosen
STATISTICA and StatSoft as the partner, to provide
critical advanced data analysis and data mining support for dedicated yield management solutions for the
semiconductor industries. Indeed, the complete programmability and customizability of the
STATISTICA system make
it the ideal toolkit for these types of custom solution systems.
STATISTICA 10 试用版下载