Developing an ML-based model for RF tuning of the DTL machine at ESS

Sammanfattning: The European Spallation Source (ESS) infrastructure is being constructed in Lund, and will be one of the most powerful research facilities of its type in the world. The ESS linear accelerator (linac) utilizes different accelerating sections where a wide variety of techniques should be employed to accelerate a beam of protons to 2 GeV kinetic energy through Radio Frequency (RF) cavities before being collided with a tungsten target for the final production of neutrons, through the process of spallation. This master’s thesis is a continuation on another master’s thesis project, (Lundquist, 2022), in which the focus was more on particle accelerator physics and to introduce some ML models using single-shot measurement scenario, which all failed to meet the requirements. This thesis, however, is focused on developing more ANN- models for both single-shot and multi-shot measurement scenarios, which succeed in meeting the requirements. Another focus of the project is to make the models small and feasible to be deployed in the ESS control system. The data is the simulated response of the Beam Position Monitor (BPM) sensors for the first Drift Tube Linas (DTL) tank, DTL1, at EES. DTL tanks are of great importance due to their influence on the overall performance. This will give the ESS physicists a powerful tool to direct the proton beam within the whole set of the DTL tanks properly, leading to a better control and thus, fewer beam losses once they start with the power ramp-up of the linac. The deployment of ML-based models in the ESS control system will be a step towards more automated and intelligent particle accelerators in this infrastructure and in similar future facilities.