Speaker
Description
In order to consistently operate collider at peak luminosity, one have to know in detail the full magnetic lattice of the colliding rings, and to know in particular the deviation of the real lattice from the model used in the design phase.
Turn-by-turn BPMs surveys are one of the method available to measure this deviation. Based on the n-BPM method developped at CERN, the spectral response of the TbTBPMs all around the rings allow for the reconstruction of the full effective magnetic lattice.
However, the measurement is very dependant on the status and precision of each TbTBPM in the ring.
A method to automatically detect and eliminate problematic BPMs from magnetic lattice reconstruction scripts is presented. It is based on a library called Time2Feat to extract and select automatically the most dissociative features from the TbTBPMs time-measurements, and the Density Based Spectral Clustering Application with Noise (DBSCAN) algorithm to detect potential outliers from a clustering of the BPMs inside the hyperspace of the computed features.
The main asset of this method is its light weight since very little training is needed, the possibility to detect outliers from very scarse measurements as input (the very problematic BPMs are detected with only one measurement from each BPM), and a data-based selection of features, preventing the introduction of any bias and allowing for the best feature-set selection possible for each measurement.
| In which format do you inted to submit your paper? | LaTeX |
|---|