Orbit Correction With Machine Learning

Orbit correction is usually an important task in the operation of accelerators. In practice, due to various errors, many devices can not operate in ideal state. By correcting the errors of magnets with corrector magnets, the beam can return to the correct position to ensure the stable operation of the accelerator. In the process of orbit correction, inaccurate BPM output will lead to incorrect correction magnet strength setting, so that the orbit correction will be impacted. BPM may make mistakes in the process of signal acquisition and current conversion. A BPM anomaly detection and predict method based on machine learning and its using in orbit correction optimization is reported in this paper. This method does not need to observe the details of BPM system, electronics technology and so on. It can monitor and predict the BPM status directly by machine learning with the information of the beam inferred from BPM and others, and optimize the orbit correction.