Intelligent Automated Tuning Systems Based On Hybrid Neural Networks

Intelligent Automated Tuning Systems Based On Hybrid Neural Networks--Physical Optics Corporation, 20600 Gramercy Place, Suite 103, Torrance, CA 90501-1821; (310) 320-3088 Dr. Wayne Wang, Principal Investigator Ms. Patty Shaw, Business Official DOE Grant No. DE-FG03-97ER82453 Amount: $75,000 Control and tuning of accelerator systems used in nuclear physics research presents an extremely challenging problem because of the large number of sophisticated monitoring and controlling devices involved and various systematic and environmental instabilities. Using ¿intelligent¿ software, rather than current manual methods with low efficiency, to improve accelerator tuning is of great interest to the Department of Energy . The purpose of this project is to design and develop an efficient auto-tuning system using hybrid neural network techniques. The goal is to quickly find the optimal set of control parameters that induce the best quality beam. The resulting system quickly adapts to the overall behavior of an accelerator system, ¿learns¿ the intrinsic relationships among accelerator parameters, and then automatically predicts an optimal set of control parameters. The Phase I research will cover three major areas: understanding the parameters of specific accelerator instruments; designing and optimizing the auto-tuning system; and integrating the software tuning system on a personal computer and conducting a proof-of-concept demonstration. The Phase II effort, in collaboration with the Los Alamos National Laboratory, will focus on developing a prototype control system for a real accelerator system. Practical issues such as software and hardware interfaces to various accelerator instruments will be investigated. The final prototype system will be fully tested in terms of tuning quality and speed. Commercial Applications and Other Benefits as described by the awardee: The success of this project would be manifested in an automated control software system that greatly improves the control and optimization of particle accelerators and associated components, increases efficiency, and saves valuable personnel resources for accelerator operations and maintenance. The auto-tuning technology would form the basis of spin-off applications for the optimization of a wide variety of complex dynamic systems, ranging from calibrating to tuning high-performance engines._