TY - CONF

T1 - Detector setup optimization based on training artificial neural-networks.

AU - Галактионов, Кирилл Александрович

AU - Руднев, Владимир Александрович

AU - Валиев, Фархат Фагимович

N1 - Conference code: LXXIV

PY - 2024/7/1

Y1 - 2024/7/1

N2 - We apply artificial neural networks (ANN) to event-wise analysis of simulated data from amicrochannel plate detector (MCP)[1] being considered for installation in future experiments on NICAcollider [2]. We have demonstrated, that neural networks can estimate the parameters of the collision notonly from spatial distribution of particles, but also benefit from high resolution time-of-flight distributionsthat can be obtained from MCP. From this data we estimate the impact parameter and the collision pointof an event. We have performed the analysis based on several Monte-Carlo models of the event. Eventhough the quality of the existing event models is not sufficient for a reliable model-independent estimationof the event parameters, the proposed parameter reconstruction procedure allows us to evaluate - and tooptimize - the technical characteristics of the detector. These characteristics include the geometry of thedevice, its placement, the number of sensors, and the time resolution.In [3, 4, 5] we have demonstrated that – subject to the detector geometry – the collision point and theimpact parameter of each event can be estimated quite accurately only from the raw detector data. Ourapproach exploits Monte-Carlo models of high energy collisions. As we have demonstrated in [3, 4, 5], thedata from QGSM generator [6] allows us to estimate the impact parameter within an uncertainty of about1 fm, and to reconstruct the collision point with uncertainty about 1 cm. This result, however, is model-dependent, and processing data from alternative generators [7, 8] leads to different ANN parameters.Despite this model dependence of the ANNs, the detector parameters providing the best reconstructionof the event parameters do not depend on the Monte-Carlo model of the event.We report the results of ANN training and suggest the optimal MCP configuration which is model-independent and, thus, can be used in future detector specification.1. A.A.Baldin, G.A. Feofilov, P. Har’yuzov, and F.F. Valiev // Nucl. Instrum. Meth.A 2020,V.958,P.162154. https://doi.org/10.1016/j.nima.2019.04.1082. https://nica.jinr.ru/3. K.A. Galaktionov, V.A. Roudnev, and F.F. Valiev, Neural network approach to impact parameterestimation in high-energy collisions using the microchannel plate detector data // Moscow UniversityPhysics Bulletin 2023, V. 78, P. S52-S584. Galaktionov K.A., Roudnev V.A., Valiev F.F. Artificial Neural Networks Application in Estimatingthe Impact Parameter in Heavy Ion Collisions Using the Microchannel Plate Detector Data: Physics ofAtomic Nuclei //Phys. At. Nucl. 2023 V.86(6), P.1426-1432. https://doi.org/10.1134/S10637788230602485. Galaktionov, K., Roudnev, V., Valiev, F., Application of Neural Networks for Event-by-EventEvaluation of the Impact Parameter // Physics of Particles and Nuclei 2023 ,V. 54, P. 446-4486. Amelin N. S., Gudima K. K., Toneev V. D. Ultrarelativistic nucleus-nucleus collisions within adynamical model of independent quark - gluon strings // Sov. J. Nucl. Phys. 1990. V. 51(6), P. 1730-17437. Werner, Klaus and Liu, Fu-Ming and Pierog, Tanguy Parton ladder splitting and the rapiditydependence of transverse momentum spectra in deuteron-gold collisions at the BNL Relativistic HeavyIon Collider // Physical Review C 2006, V. 748. Aichelin, J. and Bratkovskaya, E. and Le F`evre, A. and Kireyeu, V. and Kolesnikov, V. and Leifels,Y. and Voronyuk, V. and Coci, G. Parton-hadron-quantum-molecular dynamics: A novel microscopicn-body transport approach for heavy-ion collisions, dynamical cluster formation, and hypernucleiproduction // Physical Review C 2020, V. 101

AB - We apply artificial neural networks (ANN) to event-wise analysis of simulated data from amicrochannel plate detector (MCP)[1] being considered for installation in future experiments on NICAcollider [2]. We have demonstrated, that neural networks can estimate the parameters of the collision notonly from spatial distribution of particles, but also benefit from high resolution time-of-flight distributionsthat can be obtained from MCP. From this data we estimate the impact parameter and the collision pointof an event. We have performed the analysis based on several Monte-Carlo models of the event. Eventhough the quality of the existing event models is not sufficient for a reliable model-independent estimationof the event parameters, the proposed parameter reconstruction procedure allows us to evaluate - and tooptimize - the technical characteristics of the detector. These characteristics include the geometry of thedevice, its placement, the number of sensors, and the time resolution.In [3, 4, 5] we have demonstrated that – subject to the detector geometry – the collision point and theimpact parameter of each event can be estimated quite accurately only from the raw detector data. Ourapproach exploits Monte-Carlo models of high energy collisions. As we have demonstrated in [3, 4, 5], thedata from QGSM generator [6] allows us to estimate the impact parameter within an uncertainty of about1 fm, and to reconstruct the collision point with uncertainty about 1 cm. This result, however, is model-dependent, and processing data from alternative generators [7, 8] leads to different ANN parameters.Despite this model dependence of the ANNs, the detector parameters providing the best reconstructionof the event parameters do not depend on the Monte-Carlo model of the event.We report the results of ANN training and suggest the optimal MCP configuration which is model-independent and, thus, can be used in future detector specification.1. A.A.Baldin, G.A. Feofilov, P. Har’yuzov, and F.F. Valiev // Nucl. Instrum. Meth.A 2020,V.958,P.162154. https://doi.org/10.1016/j.nima.2019.04.1082. https://nica.jinr.ru/3. K.A. Galaktionov, V.A. Roudnev, and F.F. Valiev, Neural network approach to impact parameterestimation in high-energy collisions using the microchannel plate detector data // Moscow UniversityPhysics Bulletin 2023, V. 78, P. S52-S584. Galaktionov K.A., Roudnev V.A., Valiev F.F. Artificial Neural Networks Application in Estimatingthe Impact Parameter in Heavy Ion Collisions Using the Microchannel Plate Detector Data: Physics ofAtomic Nuclei //Phys. At. Nucl. 2023 V.86(6), P.1426-1432. https://doi.org/10.1134/S10637788230602485. Galaktionov, K., Roudnev, V., Valiev, F., Application of Neural Networks for Event-by-EventEvaluation of the Impact Parameter // Physics of Particles and Nuclei 2023 ,V. 54, P. 446-4486. Amelin N. S., Gudima K. K., Toneev V. D. Ultrarelativistic nucleus-nucleus collisions within adynamical model of independent quark - gluon strings // Sov. J. Nucl. Phys. 1990. V. 51(6), P. 1730-17437. Werner, Klaus and Liu, Fu-Ming and Pierog, Tanguy Parton ladder splitting and the rapiditydependence of transverse momentum spectra in deuteron-gold collisions at the BNL Relativistic HeavyIon Collider // Physical Review C 2006, V. 748. Aichelin, J. and Bratkovskaya, E. and Le F`evre, A. and Kireyeu, V. and Kolesnikov, V. and Leifels,Y. and Voronyuk, V. and Coci, G. Parton-hadron-quantum-molecular dynamics: A novel microscopicn-body transport approach for heavy-ion collisions, dynamical cluster formation, and hypernucleiproduction // Physical Review C 2020, V. 101

KW - Machine learning

KW - Detector modelling and simulations

KW - high-energy collisions

M3 - Abstract

T2 - LXXIV International Conference “Nucleus-2024: Fundamental problems and applications”, Dubna, July 1–5, 2024:

Y2 - 1 July 2024 through 5 July 2024

ER -