PACMAN HIPA

Particle Accelerators and Machine Learning

Started
January 2, 2019
Status
In Progress
Share this project

Abstract

The High Intensity Proton Accelerator (HIPA) at Paul Scherrer Institute (PSI) provides the primary beams to PSI’s versatile experimental facilities which in turn provide high intensity beams for research. In an accelerator control room several hundred of continuous sensor data are displayed in order to aid the operators in running the accelerator with maximal performance. We propose to bring Machine Learning (ML) to particle accelerator operation, in order to increase the performance. A more accurate parameter control based on the surrogate modelling will contribute to reliable and safe operation, and increase the accelerator efficiency. The immediate benefits will be: reducing the risks related to the high beam power by reducing the activation and beam losses, an action that will in turn, lead to fewer machine interruptions and possibly higher beam intensities. The project is likely to have a game-changing impact in howwe model and operate charged particle accelerators in the near future.

People

Collaborators

SDSC Team:
Fernando Perez-Cruz

PI | Partners:

Cyclotron Development and Beam Dynamics:

  • Dr. Jochem Snuverink
  • Dr. Nicole Hiller

More info

Laboratory for Scientific Computing and Modelling:

  • Dr. Andreas  Adelmann
  • Dr. Jaime Coello de Portugal
  • Sichen Li

More info

Learning & Adaptive Systems Group:

  • Prof. Andreas Krause
  • Johannes Kirschner
  • Mojmir Mutny

More info

description

Goals:

  1. Minimise beam losses: To be able to predict the reaction of a knob, especially those at the first sections of the accelerator, a reliable machine model needs to be available.
  2. Better control of accelerator parameters: we will establish a fast on-line enhancement of the machine protection system with beam diagnostics data. The accelerator parameters will be predicted from the beam diagnostics data. Appropriate changes to the machine / beamline settings will be proposed as operational enhancements by the surrogate model.
  3. Prevent unnecessary machine interruptions: A surrogate model that captures fast responses from the machine can be used in the forecasting of machine interruptions. Here we would expand from research goal 1 and use results from the fusion community [10] (and the references therein).

Gallery

Annexe

Additionnal resources

Bibliography

Publications

Related Pages

More projects

ML4FCC

In Progress
Machine Learning for the Future Circular Collider Design
Big Science Data

CLIMIS4AVAL

In Progress
Real-time cleansing of snow and weather data for operational avalanche forecasting
Energy, Climate & Environment

SEMIRAMIS

Completed
AI-augmented architectural design
Energy, Climate & Environment

4D-Brains

In Progress
Extracting activity from large 4D whole-brain image datasets
Biomedical Data Science

News

Latest news

Efficient and scalable graph generation through iterative local expansion
March 20, 2024

Efficient and scalable graph generation through iterative local expansion

Efficient and scalable graph generation through iterative local expansion

Have you ever considered the complexity of generating large-scale, intricate graphs akin to those that represent the vast relational structures of our world? Our research introduces a pioneering approach to graph generation that tackles the scalability and complexity of creating such expansive, real-world graphs.
RAvaFcast | Automating regional avalanche danger prediction in Switzerland
March 6, 2024

RAvaFcast | Automating regional avalanche danger prediction in Switzerland

RAvaFcast | Automating regional avalanche danger prediction in Switzerland

RAvaFcast is a data-driven model pipeline developed for automated regional avalanche danger forecasting in Switzerland. It combines a recently proposed classifier for avalanche danger prediction at weather stations with a spatial interpolation model and a novel aggregation strategy to estimate the danger levels in predefined wider warning regions, ultimately assembled as an avalanche bulletin.
PassGPT | Using language models to enhance password security
February 6, 2024

PassGPT | Using language models to enhance password security

PassGPT | Using language models to enhance password security

PassGPT is a Large Language Model for password generation trained on leaked passwords, which can outperform existing methods based on generative adversarial networks by guessing twice as many unseen passwords.

Contact us

Let’s talk Data Science

Do you need our services or expertise?
Contact us for your next Data Science project!