Exploring implementations for online control & optimisation applications in the operation of a particle accelerator

This bachelor thesis has been conducted at the European Organization for Nuclear Research (CERN) in the framework of the technical student programme. The focus of this work was the implementation of a control and optimisation algorithm to counteract a recurring drift in a specific process in a particle accelerator: the CERN Proton Syn- chrotron. The goal was to increase the overall efficiency of said process by achieving an increase in performance and increase in autonomy. The work presented in this thesis covers the following domains: learning the funda- mentals in accelerator physics to understand the problem at hand, incorporating the computational infrastructure at CERN to implement a solution and studying certain control and optimisation algorithms to leverage their capabilities. Subsequently, the technicalities of the general setup will be clarified before explaining the design of the cost function for the algorithm. As the performance of this specific process is influenced by different factors, the cost function must cover multiple objectives simultaneously. The complexity induced by the different features of the cost function as well as the methodology adopted for testing were studied in detail. In the final observations the achievements reached in this study are highlighted and the caveats are summarised. The observations were based on findings made in large categorical time series that are difficult to process. A specialised tool is presented, which was developed specifically for this task and greatly facilitated the analysis of the time series. Finally, an outlook is given based on the discoveries made in this study and the next actions in this project will be presented. Additionally, possibilities to improve the cost function in the future thanks to planned hardware upgrades are discussed.