Particle identification and analysis in the SciCRT using machine learning tools

Abstract Machine learning is a powerful tool used in many different areas, from image processing to space navigation and high-energy physics. In this paper we present a configuration of different artificial intelligent tools aimed at the extraction of features from data registered in the SciBar Cosmic Ray Telescope (SciCRT). The SciCRT is an array of plastic scintillator bars that work nearly independently as particle detectors. When a particle crosses inside the telescope, scintillation photons are emitted by the plastics. The intensity of photons is directly proportional to the energy deposited in each bar. Taking advantage of the construction of the telescope, the small transverse area of the scintillator bars, it is possible to do particle tracking and analysis. The main purpose of SciCRT is the detection of solar neutrons originated in the violent phenomena taking place at the surface of the Sun. Nonetheless, the SciCRT is capable of detecting different kinds of secondary particles produced by the interactions of primary cosmic rays with the atmospheric nuclei. For this reason, the task of signal classification is essential. Our final goal will be the classification of detected cosmic ray particles, as well as, the unfolding of the neutron energy spectrum and the estimation of the angular distribution. To achieve this our methodology relies of pattern recognition, artificial neural networks, k-means clustering and k-Nearest Neighbors. In addition, our paper presents a Monte Carlo simulation of the SciCRT for the training and evaluation of the machine learning algorithms.