Orbit correction -moving the orbit to a desired orbit, orbit stability -keeping the orbit on the desired orbit using feedback to filter out unwanted noise, and orbit analysis -to learn more about the model of the machine, are strongly interrelated. They are three facets of the same problem.The better we know the model of the machine, the better predictions we can make on the behavior of the machine (inverse modeling) and the more accurately we can control the machine. On the other hand, one of the tools to learn more about the machine (modeling) is to study and analyze the orbit response to 'kicks'.
Orbit correctionWith smaller and smaller vertical beamsizes ( 5 50p) in the new generation of Synchrotron Radiation facilities and with the advent of small angle and crab crossing colliding beams in the B and C$ factories, there is a need for very accurate ( 5 lop) orbit control.
ApproachesThere axe different approaches t o the orbit correction. They axe not all exclusive to each other, some methods can fall into more then one categories.The corrective step is calculated based on model predictions. It needs a very good knowledge of the model of the machine (including model-magnet calibrationsee Section 4). Since these models are almost all linear, orbit correction needs several iterative steps. All early methods and many of the. more recent ones (111-[14], [lS], [26]-[27]) fall into this category.They consist of a knowledge base (data) and a separate reasoning mechanism (inference machine). The knowledge base contains the model of the machine and rules what the most expert accelerator physicist/engineer/operator would apply. Thus an Expert System represents a model based system. Expert Systems are capable of treating large amount of information in an empirical way.There were attempts in mid and late 80's by CERN -using commercial ES packages, primarily for fault diagnosis [15]b, but in [27]h it was concluded, that a simple algorithmic process is more efficient for the analysis and correction of closed orbits.Basically, this was the approach taken by SLAC in collaboration with the Stanford Knowledge Systems Laboratory ([15]c, [27]b,c) A Fortran coded in-house 'expert', Able was incorporated into the system €or finding errors and misalignments in the machines and for controling the machines.