Author(s)
| Di Croce, Davide (Ecole Polytechnique, Lausanne) ; Giovannozzi, Massimo (CERN) ; Krymova, Ekaterina (Ecole Polytechnique, Lausanne ; Zurich, ETH) ; Pieloni, Tatiana (Ecole Polytechnique, Lausanne) ; Seidel, Mike (PSI, Villigen ; Ecole Polytechnique, Lausanne) ; Van der Veken, Frederik (CERN) |
Abstract
| The Dynamic Aperture (DA) is an important concept for the study of non-linear beam dynamics in a circular accelerator. It refers to the region in phase space where a particle’s motion remains bounded over a given number of turns. Understanding the features of DA is crucial for operating circular accelerators as it provides insights on non-linear beam dynamics and the phenomena affecting beam lifetime. The standard approach to calculate the DA is computationally very intensive. In our study, we aim at determining an optimal set of parameters that affect DA, like betatron tune, chromaticity, and Landau octupole strengths, using a Deep Neural Network (DNN) model. The DNN model predicts the so-called angular DA, based on simulated LHC data. To enhance its performance, we integrated the DNN model into an innovative Active Learning (AL) framework. This framework not only enables retraining and updating of the model, but also facilitates efficient data generation through smart sampling. The results demonstrate that the use of the Active Learning (AL) framework allows faster scanning of LHC ring configuration parameters without compromising the accuracy of the DA calculations. |