de
en
Schliessen
Detailsuche
Bibliotheken
Projekt
Impressum
Datenschutz
de
en
Schliessen
Impressum
Datenschutz
zum Inhalt
Detailsuche
Schnellsuche:
OK
Ergebnisliste
Titel
Titel
Inhalt
Inhalt
Seite
Seite
Im Dokument suchen
Study of the cosmic ray composition sensitivity of AugerPrime : probing the prospects of the upgrade to the Pierre Auger Observatory with a deep learning approach / vorgelegt von M. Sc. Sonja Mayotte. Wuppertal, June 2021
Inhalt
ABSTRACT
ZUSAMMENFASSUNG
TABLE OF CONTENTS
LIST OF FIGURES
LIST OF TABLES
LIST OF ABBREVIATIONS
INTRODUCTION: ASTROPARTICLE PHYSICS AT THE HIGHEST ENERGIES
COSMIC RAYS AND EXTENSIVE AIR SHOWERS
The Flux of Cosmic Rays
Features in the Cosmic Ray Energy Spectrum
The Cutoff of the Cosmic Ray Energy Spectrum
Detection of Cosmic Rays
Extensive Air Showers
Characterizing an EAS
The Composition of Ultra-High Energy Cosmic Rays
THE PIERRE AUGER OBSERVATORY
The Fluorescence Detector
The Surface Detector
The SD Signal Calibration
The SD Trigger System
The SD Shower Reconstruction
The AugerPrime Upgrade
The Surface Scintillator Detector
The Upgraded Unified Board
Deployment of AugerPrime
The CDAS Software
The Off line Framework
Advanced Data Summary Trees
The Python tool PyIK
FIRST ANALYSES OF ENGINEERING ARRAY DATA
Engineering Array Layout
SD433 Reconstruction and HV Stability
Charge Estimation of SSD Data and Correlation between Twin Stations
Estimated Charge vs@汥瑀瑯步渠. Distance to Shower Core/Zenith Angle
Timing Resolution of the SSD
Timing Analysis without Shower Direction Correction
Timing Analysis with Shower Direction Correction
Timing Analysis during Full-Bandwidth Trigger Period
Timing Analysis after switch of GPS module
Cut on Signal Size and Analysis of the Signal Start Time
Annual Variation of SSD PMT Temperature
MIP Peak Stability vs@汥瑀瑯步渠. Time/Temperature
Trace Baseline Stability
Conclusions from the EA Analysis
ARTIFICIAL NEURAL NETWORKS & DEEP LEARNING
Basic Components of a Neural Network
Initialization and Training of a Neural Network
Data Set Splitting
Global Parameters, Metric, and Weight/Bias Initialization
Forward Propagation and Cost Calculation
Back Propagation and Network Evaluation
Combating Under- and Overtraining
Convolutional Neural Networks
Striding and Padding
Cartesian Convolution
Hexagonal Convolution
General Layer Types
Pooling Layers
Dense, Flatten, and Reshape Layers
SIMULATION SETTINGS AND PRE-PROCESSING
CORSIKA Simulation Setup
AugerPrime Simulation and Reconstruction
Pre-processing Detector Simulations with Python
Parameter Selections and Read-out
Data Augmentation and Elongation Rate Correction
Axial Layout and Padding
Removal of Geometric Degeneracy
RECONSTRUCTION OF Xmax FROM AUGERPRIME SIMULATIONS
Network Architecture
3D Cartesian Trace Convolution
2D Hexagonal Group Convolution
Final Layers
In-training Data Augmentation
Re-weighting of the Xmax,MC Distribution
Non-functioning Station Simulation
Xmax Reconstruction Results with AugerPrime Simulations
Evaluation Criteria
Choosing the Epoch of Evaluation
Evaluating the Model with SSD Trace Information
Evaluating the Model without SSD Trace Information
Zenith Angle Quality Cuts and Energy-dependent Corrections
Improvement with the Addition of SSD Data
CONCLUSION AND FUTURE PROSPECTS
EA ANALYSIS: ADDITIONAL PLOTS
SD-433 Reconstruction and HV Stability
Timing Analysis without Shower Direction Correction
Timing Analysis during Full-Bandwidth Trigger Period
Timing Analysis after switch of GPS module
MIP Peak Stability vs@汥瑀瑯步渠. Time/Temperature
Trace Baseline Stability
COMPARISON OF TWO AUGERPRIME SIMULATION SETS
selectEvents FILE
REFERENCES
ACKNOWLEDGEMENTS
DECLARATION