| One of the challenges in modeling
the performance of semi-conductor equipment is modeling the impact of
maintenance activities on equipment performance. Performing too much maintenance
can have just as negative an impact on equipment performance as the wrong
or insufficient maintenance.
Semiconductor Equipment Maintenance
Working with the staff of a Digital Equipment Corporation, one of our
staff developed a method for representing maintenance events as input
to a neural network. This technique was specifically applied to a plasma
etch machine. It enabled the successful modeling of key parameters of
Etch Rate, Standard Deviation, and Selectivity. The resulting neural
network models were integrated into a Dynamic Hill Climbing optimizer
to suggest different tactics for maintenance activities that would restore
the equipment to an acceptable level of operation.
Our key contribution was conceptualizing a method for representing
maintenance events. This led to the ability to effectively model the
process. This is an example of the kinds of innovative solutions that
our staff can provide.
Behavoral Synthesis
Behavioral Synthesis is technology that allows a chip designer to describe
an algorithm in C++ using a class library (such as SystemC) and translate that
into an intermediate form that can be used to create a gate-level net list for
fabrication. The class library provides a mechanism to implement
hardware specific constructs such as ports, registers, and memories.
At Forte Design Systems, one of
our staff developed a series of synthesizable random number generators
that provide trade-offs in area, latency, and quality of the random
numbers generated. Other work included developing a series of optimizations
that reduced area and latency for certain algorithmic constructs. These
optimizations include enhanced expression balancing (including loop
carry dependencies), global switch optimization, and numerous peep-hole
optimizations.
Other Applications
In other applications, our staff were key contributors in the development
of a high-density optical disk storage system, micro-processor development
systems, and automated printed circuit board (PCB) layout systems.
Selected Publications
- The following two publications describe the specific modeling solution
and the broader application.
- Klimasauskas, C. (2000), Analyzer for modeling and optimizing
maintenance operations, US Patent Number 6,110,214. Assigned to
Aspen Technology.
- Card, Jill P., Sniderman, Debbie L., Klimasauskas, Casimir
(1997), Dynamic
Neural Control for a Plasma Etch Process, IEEE Transactions
on Neural Networks. Volume 8, Number 4. May 1997.
|
