JACoW is a publisher in Geneva, Switzerland that publishes the proceedings of accelerator conferences held around the world by an international collaboration of editors.
@inproceedings{engel:ibic2022-wep42,
author = {S.E. Engel and P. Boutachkov and R. Singh},
title = {{Application of Machine Learning towards Particle Counting and Identification}},
& booktitle = {Proc. IBIC'22},
booktitle = {Proc. 11th Int. Beam Instrum. Conf. (IBIC'22)},
pages = {508--511},
eid = {WEP42},
language = {english},
keywords = {Windows, network, extraction, detector, experiment},
venue = {Kraków, Poland},
series = {International Beam Instrumentation Conference},
number = {11},
publisher = {JACoW Publishing, Geneva, Switzerland},
month = {12},
year = {2022},
issn = {2673-5350},
isbn = {978-3-95450-241-7},
doi = {10.18429/JACoW-IBIC2022-WEP42},
url = {https://jacow.org/ibic2022/papers/wep42.pdf},
abstract = {{An exploration into the application of three machine learning (ML) approaches to identify and separate events in the detectors used for particle counting at the GSI Helmholtz Centre for Heavy Ion Research. A convolutional neural network (CNN), a shape-based template matching algorithm (STMF) and Peak Property-based Counting Algorithm (PPCA) were developed to accurately count the number of particles without domain-specific knowledge required to run the currently used algorithm. The three domain-agnostic ML algorithms are based on data from scintillation counters commonly used in beam instrumentation and represent proof-of-work for an automated particle counting system. The algorithms were trained on a labelled set of over 150 000 experimental particle data. The results of the three classification approaches were compared to find a solution that best mitigates the effects of particle pile-ups. The two best-achieving algorithms were the CNN and PPCA, achieving an accuracy of 99.8\%.}},
}