Speaker
Description
Modern particle accelerators have been used in a wide range of fields, including scientific research and medical applications that demand a high-quality particle beam. To achieve an excellent beam standard for these applications, an accelerator requires reliable and accurate diagnostic instruments capable of measuring the phase space distribution of the beam in complex experiments. Tomography is a typical technique for measuring the phase-space distribution, in which projections from various angles are integrated to reconstruct an item in higher dimensions via mathematical transformations, without prior assumptions about the shape of the beam distribution. However, these conventional tomographic approaches pose several major drawbacks: in terms of computational resources, they require considerable computing time and resources; the traditional tomography algorithms can be prone to artefacts in the reconstruction, which are defined as any persistent discrepancies between the reconstructed image and the actual object, particularly in a reduced number of projections; and it is difficult to apply in higher-dimensional reconstruction. In this study, we present an approach based on generative machine learning to create a neural network model capable of predicting the six-dimensional macroparticle phase-space distribution, called generative phase space reconstruction (GPSR). This approach achieves a good agreement with the ground truth across five in six phase space dimensions, but in high dispersion, the GPSR method appears less accurate in reconstructing the horizontal distribution.
| Presenting Author | Angga Dwi Saputra |
|---|---|
| Is the Presenting Author a PhD Student or Early Career Scientist ? | Yes |
| Area of research | Beam dynamics, including electrodynamics |