Speaker
Description
The Inner Tracker of the ATLAS experiment requires the optimal performance
of its pixel sensors. To test their efficiency, a reliable track reconstruction and
analysis for testbeam data is necessary to ensure the precise detection of par-
ticles. The quality of data from testbeam campaigns are influenced by many
factors, including high beam densities, which can impair the track reconstruc-
tion.
To analyse and evaluate the data taken at beam tests, the track reconstruction
software Corryvreckan is used. It is now the predominant reconstruction frame-
work for beam tests and was developed with the intention to reduce external
dependencies without reducing the quality and versatility of track reconstruc-
tion in complex environments.
The reconstruction of particle tracks with too many hits becomes increasingly
difficult due to the ambiguity of track fits. In order to differentiate between
false and true reconstructed tracks, a machine learner is implemented, which is
trained on simulated testbeam data, generated by the Allpix Squared software.
This talk presents results of the track reconstruction of high track density using
Corryvreckan and the performance of a machine learner for true track tagging.
Both simulated data and real testbeam data is investigated.
