                      -------------------------------------
                       The FC110 Advances Around the Globe
                        Fuzzy Industrial Control in Japan
                      -------------------------------------

In the search for high speed fuzzy capability, the stage is shifting from fuzzy
software to fuzzy hardware.  Fuzzy logic, which has become a household word in
Japan's consumer electronics market, is now being applied in industrial
equipment applications with the advent of the fuzzy chip.  

TIL's FC110 Digital Fuzzy Processor is being used in the metal manufacturing
industry for plasma etching in electro-discharge process machines.  The FC110
DFP(tm) was chosen by the Sodick company in Japan for development of a fuzzy
system that automatically controls process settings for plasma etching
processes.  The FC110 was incorporated because of its speed advantages for
fuzzy logic applications with real time requirements.  Sodick's "Mark 20
Series" of machines which incorporate this fuzzy system are being used in
industry today.

In plasma etching, process effectiveness and process speed have improved
through real time optimal control.  As functionality and capabilities have
advanced, improving factory efficiency continues to be important.  The fuzzy
controller used in this application was devised to automate an operator's
knowledge in adjusting settings during process runs.  

Until now, even in the latest, highly-automated electro-discharge process
machine, the operator decided the optimal settings for the two main input
parameters, servo-voltage (SV) and arc-jump-cycle (AJC.)  Furthermore, during
process runs, adjustments were made to stabilize the discharge, improve process
efficiency, and address changes in the environment, as well as accidents.  The
level of experience of the operator played a key role in the success of process
runs.

For optimal fuzzy control, it is necessary to infer the stability of the
process and to control the optimal SV and AJC.  Inference of the stability of
the process is not simply the level of the process voltage and process current,
but also the pattern of these values.  Experienced operators are able to take
into account past data to make correct decisions as to the required
adjustments.  For inference of the stability, fuzzy logic was employed due to
its suitability for multi-input, nonlinear control.

However, incorporating fuzzy logic alone is not sufficient.  Important factors
are:  1) how to create an appropriate knowledge base;  2) how to process data
at high speed.  Experience showed that fuzzy inference entries (membership
functions, rules, etc.) required significant modifications for stability.  In
addition, software inference on PCs became too slow when the knowledge base
grew large.

With the use of TIL development tools, the time necessary to create a knowledge
base was shortened considerably.  By using the FC110 for fuzzy inferencing,
speed improvement on the order of ten times have been realized compared with
using software alone on a 32-bit 80386 CPU.  The process time has also been
reduced ten percent compared to the process time when using experienced
operators.  The results indicate that even if non-optimal values are set,
optimal fuzzy control yields results equaling or exceeding those of experienced
operators. 

It is believed that industrial applications will increase markedly in
manufacturing processes and factory automation where experts' subjective
decision making is required in real time at high-speeds.
--
Copyright (c) Togai InfraLogic, 1992.  All rights reserved.
Permission to freely distribute this document, provided that it remains
complete and intact, is hereby granted.
