Author:
Xie X.Y.,Xu H.L.,Li Q.Y.,Sun Y.J.
Abstract
Abstract
A data-based machine learning approach is proposed to study
the properties of time resolution of RPC detectors by measuring the
time of flight of cosmic muons. This method utilises a multi-layer
perceptron and a type of recurrent neural network called long
short-term memory. The neural network is trained with the waveforms
of RPC signals digitized by an oscilloscope at a sampling frequency
of 10 GHz and a 2 GHz bandwidth. A data augmentation approach is
implemented for labelling. Compared to the results from
conventional waveform analysis, this approach achieves a better time
resolution of 1-mm gap RPCs. Based on the data, the approach has a
generalisation capacity for performance studies of other timing
detectors.
Subject
Mathematical Physics,Instrumentation