5th ICFA Beam Dynamics Mini-Workshop on Machine Learning for Particle Accelerators

Session

Poster session

9 Apr 2025, 16:00

80/1-001 - Globe of Science and Innovation - 1st Floor (CERN)

104. Accelerator Surrogate Model Based on Diffusion Models: A Case Study of the MEBT Section in CAFe2

Kunxiang Sun (近代物理研究所)

Surrogate Modelling and Digital Twins

Poster

This paper introduces an accelerator surrogate model utilizing Diffusion Models. The proposed model leverages the parameters of the incoming beam bunch as conditional inputs to accurately and efficiently predict the phase space distribution of the beam at the accelerator’s exit. By focusing on the Medium Energy Beam Transport (MEBT) section of the China Accelerator Facility for Superheavy...

88. Analyzing the Impact of Seismic Activities on EuXFEL's Optical Synchronization System Using Data Mining

Arne Grünhagen

Anomaly Detection and Diagnostics

Poster

In pump-probe experiments with free-electron lasers, the arrival time stability between the FEL pump pulse and the probe pulse is of utmost importance. An optical synchronization system is used to synchronize several components of the accelerator and the pump-probe laser. Different seismic activities cause the tunnel length and thus the length of the installed optical fibers to change. In...

32. Anomaly Detection for fault analysis in the Proton Synchrotron using Machine Learning

Joel Axel Wulff

Anomaly Detection and Diagnostics

Poster

The CERN Proton Synchrotron (PS) is equipped with several RF systems covering a wide range of revolution harmonics (7 to 21, 42, 84) with heterogeneous hardware of different age. Despite a good track record of high availability (over 99%), a degradation observed from 2021 to 2023 highlighted a need for enhanced fault diagnosis of these RF systems. Accurate identification of failure sources is...

106. Application of Bayesian Optimization on Booster to AGS Injection at BNL through Xopt – Experiences and Challenges

Georg Hoffstaetter (Cornell University)

Optimisation and Control

Poster

Accelerator control systems consist of very large numbers of parameters, many of which are continually re-tuned to account for different working conditions and drifting optimal points. In the case of the BNL RHIC Injector Complex, many different ion species are accelerated, and while there is a scarcity of diagnostics, there is a surplus of control knobs for optimizing the injection process....

26. Application of Reinforcement Learning on online Coupling Correction for Storage Ring

Yihao GONG (Shanghai Advanced Research Institute)

Optimisation and Control

Poster

In modern synchrotron light sources, maintaining beam stability is critical for ensuring high-quality synchrotron radiation performance. Light source stability is governed by stability of current, beam position and beam size. Beam size stability on the order of several microns need to be improved for future experiments. Reinforcement learning (RL) offers a promising approach for real-time beam...

29. Automation of Device Server creation in TANGO Controls using Large Language Models

Krzysztof Klimczyk

LLMs and AI Assistants

Poster

This project explores methods for automating the creation of device servers in the TANGO Controls environment using Large Language Models (LLMs). The primary goal is to streamline and accelerate the coding process for device servers, reducing the time and effort required by developers. The application features a web-based user interface where users specify device attributes, commands, and...

13. Batch Spacing Optimization via Reinforcement Learning

Matthias Remta (University of Vienna (AT))

Optimisation and Control

Poster

Beams designated for the LHC are injected in multiple batches into the SPS. With a tight spacing of 200 ns between these batches, the injection-kickers have to be precisely synchronised with the injected beam so that injection oscillations are minimized. Due to machine drifts the optimal settings for the kickers vary regularly. In this paper a Reinforcement Learning agent was developed as an...

19. Bayesian Optimization for Model-Based Automated Multi-Turn Extraction Tuning at the Proton Synchrotron.

Levente Alex Foldesi (University of Zurich (CH))

Optimisation and Control

Poster

The beam for CERN's North Area proton physics program is produced through a Multi-Turn Extraction (MTE) scheme at the Proton Synchrotron (PS). Using fourth-order resonant excitation, the beam is split into five beamlets in horizontal phase space, with extraction occurring over five consecutive turns. The quality of the splitting is measured by the uniformity of intensities across the beamlets....

21. Bayesian optimization for the SIS18 injection

Sabrina Maria Appel

Optimisation and Control

Poster

The complexity of the GSI/FAIR accelerator facility demands a high level of automation to maximize the time for physics experiments. Accelerator laboratories across the globe are investigating numerous techniques to achieve this goal, including classical optimization, Bayesian optimization (BO), and reinforcement learning. This presentation will provide an overview of recent activities in...

103. Bayesian optimization in the Karabo SCADA system at the European XFEL

Florian Sohn (European XFEL)

MLOps, Infrastructure and Scalability

Poster

The automation of repetitive tasks at FEL light-sources like the European XFEL allows to allocate staff more efficiently and increases reliability and safety. Recurrent procedures often require adjustments of various control parameters to maximize a measured quantity, for example, repositioning of optical components during beam-alignment to maximize intensity. In many cases an analytic form of...

69. Beam based alignment using a neural network

Dr Wei Xu (University of Science and Technology of China)

Optimisation and Control

Poster

Beams typically do not travel through the magnet centers because of errors in storage rings. The beam deviating from the quadrupole centers is affected by additional dipole fields due to magnetic field feed-down. Beam-based alignment (BBA) is often performed to determine a golden orbit where the beam circulates around the quadrupole center axes. For storage
rings with many quadrupoles, the...

54. Challanges in designing an anomaly detection at ISIS

Mihnea Romanovschi (UKRI/STFC/ISIS)

Anomaly Detection and Diagnostics

Poster

The controls group at ISIS has been exploring anomaly detection and its associated challenges. This overview highlights the challenges faced, unsuccessful attempts, and lessons learned. Initially, the group implemented a machine learning anomaly detection system on the methane moderator for Target Station 1. The anomaly detection work began before the system upgrade, rendering previous...

112. Development of AI-ML and Generative AI Tools for Accelerator Design and Optimization

Brahim Mustapha

Surrogate Modelling and Digital Twins

Poster

Accelerator design, optimization and simulation, are important steps in the construction of future accelerator-based facilities. Typically, multiple alternatives are investigated, involving multiple design iterations, before selecting a final design. Once a design is selected, the focus shifts to the detailed design of individual accelerator components and end-to-end lattice design and...

116. Digital Assistants using LLMs for the Electron Ion Collider Project

Linh Nguyen

LLMs and AI Assistants

Poster

Brookhaven National Laboratory is home to future Electron Ion Collider (EIC). The EIC will collide electrons with protons and nuclei to produce snapshots of particles’ internal structure. This will allow us to study the role of gluons in the matter within and all around us. As the project expands, the growing collection of technical documentation across multiple sub-systems creates challenges...

53. Efficient data-driven model predictive control for online accelerator tuning

Chenran Xu (KIT)

Optimisation and Control

Poster

Reinforcement learning (RL) is a promising approach for the online control of complex, real-world systems, with recent success demonstrated in applications such as particle accelerator control. However, model-free RL algorithms often suffer from sample inefficiency, making training infeasible without access to high-fidelity simulations or extensive measurement data. This limitation poses a...

91. eLog Analysis for Accelerators: Status and Future Outlook

Antonin Sulc (DESY)

LLMs and AI Assistants

Poster

This work presents a systematic analysis of electronic logbook (eLog) systems and their analytical capabilities at the accelerator facilities of DESY and Lawrence Berkeley National Laboratory (LBNL). We evaluate contemporary tools and methodologies for enhanced information retrieval, focusing on extracting operational insights from eLog entries through state-of-the-art natural language...

64. Enhancements of beam diagnostics and simulation environments toward ML-based XFEL improvements at SACLA

Dr Hirokazu Maesaka (RIKEN SPring-8 Center)

MLOps, Infrastructure and Scalability

Poster

An ML-based optimizer has been working on maximizing XFEL performance at SACLA [1]. The spectral brightness of XFEL was successfully optimized by using a new high-resolution inline spectrometer [2]. To improve the XFEL performance further, we plan to enhance beam diagnostics and simulation environments for the ML-based optimizer. As for beam diagnostics, we are developing an X-band RF...

20. Enhancing Beamline Simulation in Particle Linear Accelerators through Advanced Deep Learning Techniques

yaxin hu (Institute of Modern Physics)

Surrogate Modelling and Digital Twins

Poster

This study focuses on the time-series prediction problem in beamline simulation for particle linear accelerators, aiming to improve simulation accuracy and computational efficiency by introducing advanced deep learning techniques. We compare and evaluate the performance of several deep learning models, including Multi-Layer Perceptron (MLP), Long Short-Term Memory (LSTM) networks, Transformer...

31. EPU Coupling Correction by Bayesian Optimization

Mau-Sen Chiu (National Synchrotron Radiation Research Center)

Surrogate Modelling and Digital Twins

Poster

The APPLE-II types of elliptically polarized undulators (EPU) are extensively utilized in synchrotron light sources. Manufacturing imperfections in the EPU inevitably lead to the creation of a residual skew quadrupole component, which couples horizontal betatron oscillation and dispersion to the vertical plane, consequently altering the vertical beam size. To regulate the vertical beam size,...

3. Exploring interpretable deep learning architectures for anomaly detection in SOLARIS synchrotron

Michał Piekarski (NSRC SOLARIS, Jagiellonian University)

Anomaly Detection and Diagnostics

Poster

The National Synchrotron Radiation Center SOLARIS, third generation light source, is the only synchrotron located in Central-Eastern Europe, in Poland. The SOLARIS Center, with seven fully operational beamlines, serves as a hub for research across a diverse range of disciplines. The most important aspect of such research infrastructure is to provide stable working conditions for the users,...

73. Exploring Reinforcement Learning for Optimal Bunch Merge in the AGS

Georg Hoffstaetter

Optimisation and Control

Poster

In BNL’s Booster, the beam bunches can be split into two or three smaller bunches to reduce their space-charge forces. They are then merged back after acceleration in the Alternating Gradient Synchrotron (AGS). This acceleration with decreased space-charge forces can reduce the final emittance, increasing the luminosity in RHIC and improving proton polarization. Parts of this procedure have...

40. Exploring the Adaptation of LLM and the Formation of AI Governance at European XFEL

Mahmoud Ajami (European XFEL GmbH)

LLMs and AI Assistants

Poster

At the European XFEL, Large Language Models provide opportunities to facilitate technical and administrative workflows, with applications such as chatbots for technical support, travel assistance and coding assistance. A task force was formed to evaluate these applications and to develop a European XFEL AI policy, focusing on technical assessment, framework creation, and ensuring ethical and...

41. First Attempts at Tuning the ISIS LEBT with Xopt and Bayesian Optimization

Kathryn Baker, Raunakk Banerjee

Optimisation and Control

Poster

Until now, the majority, if not all, accelerator tuning at the ISIS Accelerators has been done manually. With the migration of the control system to EPICS as well as the development of accessible optimization frameworks such as Xopt, we have made the first attempts to automatically tune the Low Energy Beam Transport (LEBT) at ISIS. In this presentation we will discuss the specifics of the...

25. High-Speed Differentiable Simulations as an Enabler for Reinforcement Learning-based FEL Tuning Agents

Jan Kaiser

Optimisation and Control

Poster

The photon pulse intensity is one of the key performance metrics of Free Electron Laser (FEL) facilities and has a direct impact on their experimental yield. To date, FEL intensity tuning is a time-consuming manual task that requires expert human operators to have significant skill and experience. Autonomous tuning methods have been demonstrated to reduce setup times and improve the attained...

4. Image segmentation using deep neural networks to eliminate shift in laboratory sample images caused by thermal drift

Mateusz Floras

Anomaly Detection and Diagnostics

Poster

The National Synchrotron Radiation Centre SOLARIS, a third-generation light source, is the only synchrotron facility in Central-Eastern Europe, located in Poland. The SOLARIS Centre, equipped with seven fully operational beamlines, serves as a key research hub for a wide array of scientific disciplines. The Centre requires advanced software tools to support the analysis of experimental data...

74. Improve Beam Brightness with Bayesian Optimization at the AGS Booster Injection at BNL

Xiaofeng Gu (Brookhaven National Lab)

Optimisation and Control

Poster

Alternating Gradient Synchrotron (AGS) and its Booster serve as part of the injector compound for RHIC and the future EIC at Brookhaven National Laboratory. Injection and early acceleration processes set maximum beam brightness for the collider rings. Such processes have many control parameters and are traditionally optimized empirically by operators. In an effort to streamline the injection...

119. Integrated Denoising for Improved Stabilization of RF Cavities

Jonathan Edelen, Morgan Henderson

Optimisation and Control

Poster

Typical operational environments for industrial particle accelerators are less controlled than those of research accelerators. This leads to increased levels of noise in electronic systems, including radio frequency (RF) systems, which make control and optimization more difficult. This is compounded by the fact that industrial accelerators are mass-produced with less attention paid to...

77. Integrating Machine Learning in the Brookhaven Control System with Badger

Mr Levente Hajdu (Brookhaven National Laboratory - C-AD Controls)

MLOps, Infrastructure and Scalability

Poster

The Brookhaven Pre-injector Accelerator Facility, which serves RHIC, NSRL, BLIP, and the future EIC, requires occasional tuning of its transfer beam line optics by control room operators to optimize parameters like beam current and emittance. Machine learning (ML) can significantly speed up this tuning process by helping operators quickly identify optimal settings. To facilitate this, ML...

84. Large language Models for Human-Machine Collaborative Particle Accelerator Tuning through Natural Language

Jan Kaiser

LLMs and AI Assistants

Poster

Autonomous tuning of particle accelerators is an active and challenging research field with the goal of enabling advanced accelerator technologies and cutting-edge high-impact applications, such as physics discovery, cancer research and material sciences. A key challenge with autonomous accelerator tuning remains that the most capable algorithms require experts in optimisation and machine...

95. Lattice optimization using ML-assisted MOGA for ultralow- emittance storage ring

Saroj Jena

Optimisation and Control

Poster

The Multi-Objective Genetic Algorithm (MOGA) is a powerful and increasingly adopted method for optimizing both linear and nonlinear beam dynamics in accelerator lattices, particularly for ultralow-emittance storage rings. Key objectives in this optimization include minimizing beam emittance for high brightness, maximizing dynamic aperture to ensure efficient particle injection, and expanding...

83. Machine Learning Applications for CLARA

Amelia Pollard

Surrogate Modelling and Digital Twins

Poster

CLARA is an electron accelerator test facility at STFC’s Daresbury Laboratory in the UK. The first phase has been operated since 2018 for a wide range of accelerator applications. A major upgrade is presently being commissioned that increases the electron beam energy from ~35 MeV to 250 MeV, and adds an experimental area featuring a new high-power laser. Machine learning will play an important...

42. Machine Learning Applied to BGI Profile Measurements

Javier Martinez Samblas

Anomaly Detection and Diagnostics

Poster

A key measurement in CERN accelerators for beam diagnostics is transverse size. The Beam Gas Ionization (BGI) instrument enables non-destructive observation of transverse beam profiles by detecting free electrons produced through beam-gas ionization using a Timepix-family detector. However, BGI profiles often suffer from artifacts, such as beam losses, which degrade profile quality and...

90. Machine Learning Assisted Bayesian Calibration of Accelerator Digital Twin from Orbit Response Data

Georg Hoffstaetter

Surrogate Modelling and Digital Twins

Poster

Digital twins of particle accelerators are used to plan and control operations and design data collection campaigns. However, a digital twin relies on parameters that are hard to measure directly, e.g., magnet alignments, power supply transfer functions, magnet nonlinearities, and stray fields. These parameters can be constrained by beam position and profile measurements. We use Bayesian...

68. Machine Learning based Optical Distortion measurement by Phase Advance in SSRF

liu xinzhong

Optimisation and Control

Poster

Optical distortion measurement and correction are pivotal for the stable operation of accelerators. This study introduces a machine learning-based approach to optical distortion measurement and correction implemented on the Shanghai Synchrotron Radiation Facility (SSRF). We trained models from modulated orbits to phase advance and from phase advance to quadrupole models, establishing a...

120. Machine Learning Based Sample Alignment at ORNL

Jonathan Edelen, Morgan Henderson

Optimisation and Control

Poster

Neutron scattering experiments are a critical tool for the exploration of molecular structure in compounds. The TOPAZ single crystal diffractometer at the Spallation Neutron Source and the Powder Diffractometer at the High Flux Isotope Reactor study these samples by illuminating them with different energy neutron beams and recording the scattered neutrons. Aligning and maintaining the...

65. Machine Learning driven beam emittance optimization at EuXFEL

Alexander Klemps (Hamburg University of Technology (TUHH)), Denis Ilia (DESY)

Surrogate Modelling and Digital Twins

Poster

Planned upgrades of the European X-Ray Free Electron Laser (Eu-
XFEL) target higher photon energy and a high duty-cycle operation up to CW-
operation using a superconducting RF gun with lower gradient. An operation in
this regime though critically depends on improvements of the beam slice emit-
tance of the electron gun. Within the OPAL-FEL project, we are addressing this
challenge by...

96. Machine Learning for Calibration Drift Forecasting in Superconducting RF Cavities

Yue Sun (Deutsches Elektronen-Synchrotron DESY)

Surrogate Modelling and Digital Twins

Poster

Superconducting radio frequency (SRF) cavities are critical components in particle accelerators, where accurately calibrated RF signals are essential for assessing cavity bandwidth and detuning, providing key insights into cavity performance and facilitating optimal accelerator operation. In practice, however, calibration drift due to humidity and temperature fluctuations over time poses a...

82. Machine Learning for Quench Detection in SRF Cavities at the European XFEL

Lynda Boukela

Anomaly Detection and Diagnostics

Poster

Superconducting Radio Frequency (SRF) cavities are essential components in particle accelerators, where quenches, which can cause an abrupt loss of superconductivity, remain a significant challenge. While quench detection has traditionally relied on single-parameter analysis, these methods are often limited in terms of robustness and scalability. In this work, we explore the use of machine...

23. Machine Learning-Based Anomaly Diagnostics for Optimizing SRF Operation at CAFE2

Dr Lijuan Yang (Institute of Modern Physics, Chinese Academy of Sciences)

Anomaly Detection and Diagnostics

Poster

In 2021, the Chinese ADS Front-end demo superconducting radio-frequency (SRF) linac, known as CAFe, successfully conducted a commissioning of a 10 mA, 200 kW continuous wave proton beam. During this commissioning, it was observed that the SRF faults are the leading causes of short machine downtime trips, contributing to approximately 70% of total beam trips. Analyzing fault data and...

105. Machine Learning-Driven Calibration of the Beam Position Monitors in the SPS

Vasiliki Stergiou (University of Oxford (GB))

Surrogate Modelling and Digital Twins

Poster

In the SPS, a flexible machine serving the LHC and a multitude of fixed-target experiments and fast-extraction facilities, reliable monitoring of the transverse beam position across a wide range of different beam structures and intensities is essential for stable and efficient operation. Today, the calibration procedure and signal processing of the beam position monitors (BPMs) of the SPS –...

22. Machine Learning-Driven Optimization of a Silicon Bent Crystal System for Enhanced Slow Extraction Efficiency at CERN-SPS

Francesco Maria Velotti (CERN)

Optimisation and Control

Poster

The CERN-SPS slow extraction has been recently equipped with a silicon bent crystal to reduce losses at the electrostatic septum (ES) wires. Such a concept exploits the coherent deflection that a thin crystal can give to part of the separatrix to avoid the ES wires.
In this contribution, we show how machine learning played a fundamental role in the design and operational deployment of the Si...

45. Model Predictive Control for energy efficient HVAC systems for CERN accelerators

Nikolina Bunijevac

Surrogate Modelling and Digital Twins

Poster

In times of concern over the environmental impact of high-energy physics organizations, our research in CERN's Cooling and Ventilation group (EN/CV) investigates energy-saving strategies for heating, ventilation, and air conditioning (HVAC) systems. Widely used in both residential and industrial settings, HVAC systems contribute up to 40% of residential and 70% of industrial consumption,...

113. Multi-Objective Bayesian Optimization of Beam Emittances and Energy Spread in Linear Accelerators

Prof. Chong Shik Park (Korea University)

Optimisation and Control

Poster

This study explores a multi-objective optimization and modeling approach for enhancing the performance of a linear accelerator (linac) used in Fourth Generation Synchrotron Radiation (4GSR) facilities, focusing on the minimization of horizontal and vertical emittances as well as energy spread at the linac end. Efficient control of these parameters is critical to achieving high beam quality,...

33. Operational Results of Data-Driven Automated Intensity Optimization at CERN’s LEIR.

Borja Rodriguez Mateos (Universitat Politecnica Catalunya (ES))

Optimisation and Control

Poster

Aging of the stripper foil and unexpected machine shutdowns are the primary causes for reduction of the injected intensity from CERN’s linac3 into the Low Energy Ion Ring (LEIR). As a result, the set of optimal control parameters that maximizes beam intensity in the ring tends to drift, requiring daily adjustments to the machine control settings. In this paper, several data-driven methods such...

114. Phase Space Reconstruction of Heavy Ion Linac Beams at RAON Facility Using Machine Learning Techniques

Prof. Chong Shik Park (Korea University)

Anomaly Detection and Diagnostics

Poster

This study investigates the application of machine learning techniques for the phase space reconstruction of heavy ion linac beams at the Rare isotope Accelerator complex for ON-line experiments (RAON) facility in Korea. Phase space analysis is a critical component in understanding and optimizing beam dynamics, enabling precise control of beam quality for advanced nuclear physics experiments....

55. Preliminary Findings on the Application of Reinforcement Learning for Real-Time Optimization of Crystal Channelling in the LHC

Andrea Vella (University of Malta (MT))

Optimisation and Control

Poster

The Large Hadron Collider (LHC) operates with high intensity proton and heavy ion beams that necessitate a robust collimation system to prevent damage to sensitive equipment along the ring. However, the efficiency of cleaning ion beams is approximately 100 times less efficient than with protons. To address this, bent silicon crystals were implemented to enhance collimation efficiency. The...

36. Research on AI-Assisted Operation System for Particle Accelerators

Zhijun wang (IMP,CAS)

MLOps, Infrastructure and Scalability

Poster

The long-term stable operation of high-power accelerators demands high availability, reliability, and safety, requirements that traditional labor-intensive maintenance methods struggle to meet. In recent years, with the rapid advancement of artificial intelligence and machine learning technologies, various efficient algorithms have demonstrated great potential in the field of particle...

43. Simulation package for the Beam Synchrotron Radiation Longitudinal density monitor (BSRL)

Alexander Nicholas Jury (University of Liverpool (GB))

Surrogate Modelling and Digital Twins

Poster

The Beam Synchrotron Radiation Longitudinal density monitor (BSRL) at the LHC leverages time-correlated single-photon counting to provide high-dynamic-range measurements of particle populations within each bunch in the LHC including monitoring of “ghost” and “satellite” bunches, which represent charge captured in nominally empty buckets, thereby enhancing the accuracy of luminosity...

35. Surrogate Model Training and Applications at the S-DALINAC based on FCNNs

Dominic Schneider (Institut für Kernphysik, TU Darmstadt)

Surrogate Modelling and Digital Twins

Poster

Machine learning methods provide a significant potential for the optimized operation of complex facilities, such as particle accelerators. In this contribution, the first training and application of surrogate models to the electron accelerator S-DALINAC based on Fully-Connected Neural Networks (FCNN) will be presented.

An exhaustive data-mining algorithm has been developed to generate the...

48. Surrogate Model-Driven Reinforcement Learning for Efficient Beam Tuning in Accelerators

Chunguang Su (Institute of modern physics, Chinese Academy of Sciences)

Surrogate Modelling and Digital Twins

Poster

Beam tuning in particle accelerators is a complex task, especially when physical modeling is impractical due to the lack of complete beam diagnostics. Traditional methods often rely on iterative manual tuning by operators, which can be inefficient. Reinforcement learning (RL) algorithms offer a promising alternative for automating this process. In this work, we demonstrate the successful...

59. Surrogate Models for Longitudinal Beam Manipulations in the CERN Proton Synchrotron

Stamatina Detsi (National Technical University of Athens, School of Electrical and Computer Engineering)

Surrogate Modelling and Digital Twins

Poster

The Proton Synchrotron (PS) at CERN is equipped with numerous RF systems allowing for evolved longitudinal beam manipulations to adapt the number of bunches and their spacing. The beam produced for the LHC undergoes several bunch splittings, merging and batch compression. Each manipulation must be carefully adjusted to minimize the spread in bunch parameters at PS extraction. The design of...

107. Surrogate Models for X-Ray Pulse Diagnostics at the European XFEL

Fady Bishara (European XFEL GmbH)

Surrogate Modelling and Digital Twins

Poster

The European XFEL is a scientific research facility that produces ultra-short and ultra-brilliant x-ray pulses. The facility is 3.4 kilometers long and comprises, very simplistically, three main sections: a linear electron accelerator, undulators, and photon beamlines. The entire facility is densely instrumented with various diagnostic devices that produce a vast amount of diverse...

30. Time-varying Bayesian optimisation of injection in the Proton Synchrotron Booster

Francisco Huhn (CERN)

Optimisation and Control

Poster

The Proton Synchrotron Booster (PSB) receives 160 MeV H$^-$ ions, which are converted to protons at injection via a charge exchange mechanism, an upgrade that allows the production of high-intensity beams ($> 10^{13}$ per ring). Nevertheless, with the increase in intensity, space-charge losses arise. To mitigate these effects, horizontal phase-space painting is performed with a system of fours...

89. Toward Autonomous Control: Reinforcement Learning for Improving accelerator performance in CLEAR

Antonio Gilardi

Optimisation and Control

Poster

Particle accelerators, such as the CERN Linear Electron Accelerator for Research (CLEAR), play a critical role in various scientific fields.
Ensuring their operation is automatic, stable, and reproducible is vital for the scalability of future large-scale accelerator projects.
This paper presents an initial step toward autonomous control of the CLEAR beamline, beginning with a basic beam...

85. Towards Lattice-Agnostic Reinforcement Learning Agents for Transverse Beam Tuning

Chenran Xu, Jan Kaiser

Optimisation and Control

Poster

Reinforcement learning (RL) has been successfully applied to various online tuning tasks, often outperforming traditional optimization methods. However, model-free RL algorithms typically require a high number of samples, with training processes often involving millions of interactions. As this time-consuming process needs to be repeated to train RL-based controllers for each new task, it...

46. Updates to the Xopt/Badger Ecosystem for Advanced Optimization and Online Control

Ryan Roussel

Optimisation and Control

Poster

The Xopt/Badger ecosystem offers a versatile suite of tools designed to address the growing needs of advanced optimization and online control in scientific applications. The goal of these tools is to standardize the implementation and use of advanced optimization algorithms at arbitrary scientific facilities for the benefit of the wider accelerator community. In this work, we provide a summary...

39. Using Hyper-GAN to improve Uncertainty Quantification for Surrogate Models

Kathryn Baker

Surrogate Modelling and Digital Twins

Poster

Gaussian Processes are the most popular and accurate modelling technique for estimating uncertainty in a system. However, they struggle to scale with high dimensional data and large datasets that are often required to train surrogate models of accelerator components. Typical alternatives include deep ensembles, Monte-Carlo Dropout and quantile regression. However, each of these methods has...

57. Using Surrogate Models to make BO safer for particle accelerators

Kathryn Baker, Raunakk Banerjee

Optimisation and Control

Poster

Tuning of the accelerator at the ISIS Neutron and Muon Source has traditionally been done manually, relying on experts to vary parameters in the control system to achieve optimum beam efficiency and intensity. Bayesian Optimisation (BO) is an effective way to automatically tune the Low Energy Beam Transport (LEBT), a section within the linear accelerator. However, due to losses and radiation...