Optimisation of the SHiP Beam Dump Facility with generative surrogate models
File(s)
Author(s)
Shirobokov, Sergey
Type
Thesis
Abstract
The SHiP experiment is a proposed fixed target experiment at the CERN SPS to search for new particles. To operate optimally, the experiment should feature a zero background environment. The residual muons flying from the target are one of the largest sources of the background. To remove them from the detector acceptance, a dedicated muon shield magnet is introduced in the experiment. The shield should be optimised to deliver the best physics performance at the lowest cost.
The optimisation procedure is very computationally costly and, thus, requires ded- icated methods. This thesis comprises of a detailed description of a new machine learning method for the optimisation, comparisons to existing techniques, and the application of the method to optimising the muon shield magnet. In addition, the set of technological and simulation problems affecting the optimisation is discussed in details. Finally, the set of requirements for the muon shield prototype design and verification is presented.
The optimisation procedure is very computationally costly and, thus, requires ded- icated methods. This thesis comprises of a detailed description of a new machine learning method for the optimisation, comparisons to existing techniques, and the application of the method to optimising the muon shield magnet. In addition, the set of technological and simulation problems affecting the optimisation is discussed in details. Finally, the set of requirements for the muon shield prototype design and verification is presented.
Version
Open Access
Date Issued
2021-07
Date Awarded
2022-02
Copyright Statement
Creative Commons Attribution-Non Commercial 4.0 International Licence
License URL
Advisor
Golutvin, Andrey
Publisher Department
Physics
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)