Speaker
Description
The MWPC-tracking based MOS (Muography Observation System), consisting of
tracking chambers and lead absorbers has been simulated in a very detailed
way in Geant4, with emphasis on all relevant physics phenomena and
including the effect of the read out electronics. The simulation results
were processed with the tracking algorithm dedicated and used for the
actual detector system.
It was found that the tracking efficiently suppresses low energy (< 0.5
GeV) muons while offering ( >95 %) tracking efficiency for the medium
energy range (5-100 GeV). The efficiency drops for >100 GeV due to
showers. The presentation will discuss these effects.
To further suppress low energy background muons, a neural network based
machine learning algorithm is currently being developed. The
machine learning algorithm is trained on the output of the simulation.
The talk will demonstrate how such algorithm can aid background
suppression in muography, and which are the requirements on the
sufficiently detailed simulation output. The ML algorithm is applied to
real data and the reliability of such a physics result is critically
assessed.
