Köln 2025 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
HK: Fachverband Physik der Hadronen und Kerne
HK 23: Poster
HK 23.27: Poster
Tuesday, March 11, 2025, 17:30–19:00, Foyer Physik
Neural network approach for energy estimation of the digital calorimeter EPICAL-2. — •Jan Scharf — Institut für Kernphysik, Goethe Universität Frankfurt, Frankfurt am Main, Deutschland
The EPICAL-2 prototype has been designed and constructed to study a concept of electromagnetic digital pixel calorimeters. The detector is based on a SiW sampling design using 24 layers, each composed of a W absorber and two ALPIDE chips featuring ∼ 30 × 30 µ m2 sized pixels. This results in a high spatial resolution of the detector, making it possible to measure three-dimensional shapes of electromagnetic showers. To estimate the energy of a particle depositing energy in the detector, pixel hits or clusters of pixel hits can be counted as a proxy. The energy resolution of the detector is thereby affected by the energy estimation capability of the proxy used.
In this poster, we present the current status of an effort to employ neural networks to estimate the energy of single initial particles from the three-dimensional pattern of hits or clusters that they generate in the detector. Features and patterns in data used to train the neural network, the network’s architecture and its design will be discussed. The energy estimate and effect of the new approach on the energy resolution of the detector for simulated data will be presented. Finally, the potential of neural networks for fast and efficient simulations of electromagnetic showers in digital calorimeters will be addressed.
Supported by BMBF and the Helmholtz Association.
Keywords: EPICAL-2; Electromagnetic shower; Digital calorimeter; Neural network; Energy estimation