Framework for a general purpose, intelligent control system for particle accelerators

Tuning and controlling particle accelerators is time consuming and expensive. Inherently nonlinear, the control problem is one to which conventional methods cannot satisfactorily be applied. Advanced information technologies such as expert systems and neural networks have been applied separately to the problem, with isolated success. Few, if any, of these advanced information technologies have been applied for general use or in a manner useful to multiple accelerator installations. We discuss results of coupling neural network and expert systems technology to solve several standard accelerator tuning problems based on realistic simulations. We also examine the effectiveness of additional heuristic search techniques such as genetic algorithms. Finally, we show the integration of this hybrid AI system with an existing general-purpose control system.