Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

AI-Ready Control System for the Fermilab Accelerator Complex

Tia Miceli, Erik Gottschalk, Donovan Tooke, Evan Milton, Robert Santucci, Hayden Hoschouer, Michael Balcewicz, Jennifer Case, Abhishek Deshpande, Kit Fieldhouse, Sudeshna Ganguly, Beau Harrison, Aisha Ibrahim, Thomas Kobilarcik, Michael Olander, Abhishek Pathak, Jason St. John, Aaron Sauers

Abstract

Reliable, high-intensity operation of the Fermilab Accelerator Complex is critical to the success of the Long-Baseline Neutrino Facility and Deep Underground Neutrino Experiment. We describe the requirements and infrastructure necessary to support routine use of artificial intelligence and machine learning (AI/ML) in the accelerator control system. Three capabilities are identified: a machine learning operations (MLOps) framework standardizing the lifecycle of AI/ML automation from data management through deployment and monitoring; a data quality framework defining and enforcing standards required to build trustworthy AI/ML applications; and workflow integration with large language models to assist physicists, engineers, and operators with information retrieval, code development, and routine analysis. Use cases spanning beam diagnostics, beam control, and support system automation illustrate the technical requirements across the complex.

AI-Ready Control System for the Fermilab Accelerator Complex

Abstract

Reliable, high-intensity operation of the Fermilab Accelerator Complex is critical to the success of the Long-Baseline Neutrino Facility and Deep Underground Neutrino Experiment. We describe the requirements and infrastructure necessary to support routine use of artificial intelligence and machine learning (AI/ML) in the accelerator control system. Three capabilities are identified: a machine learning operations (MLOps) framework standardizing the lifecycle of AI/ML automation from data management through deployment and monitoring; a data quality framework defining and enforcing standards required to build trustworthy AI/ML applications; and workflow integration with large language models to assist physicists, engineers, and operators with information retrieval, code development, and routine analysis. Use cases spanning beam diagnostics, beam control, and support system automation illustrate the technical requirements across the complex.
Paper Structure (52 sections, 11 figures)

This paper contains 52 sections, 11 figures.

Table of Contents

  1. Purpose
  2. Executive Summary
  3. Background and Motivation
  4. Document Project Context
  5. Fermilab Accelerator Complex
  6. Control System
  7. System Automation
  8. Relationship of Automations and Current Accelerator Control System
  9. AI/ML
  10. Vision for an AI/ML-Enhanced Accelerator Control System
  11. Lifecycle Management of Automations: MLOps
  12. Data Management
  13. Data Management Steps
  14. Data Management Tooling
  15. Data Management Risk Mitigation
  16. ...and 37 more sections

Figures (11)

  • Figure 1: Layout of the current Fermilab Accelerator Complex showing the major accelerators and beam transfer lines.
  • Figure 2: (a) Manual or open-loop controller, (b) feedback controller, and (c) two examples of feedforward controllers.
  • Figure 3: The "Autonomy Levels for Unmanned Systems (ALFUS) Framework" defines five autonomy levels based on the required amount of human interaction to achieve a system's mission. Level 0 requires constant human involvement, similar to how a remote control depends on a human to push its buttons. At the other end of the spectrum, Level 4 represents a system capable of functioning with minimal human interaction.
  • Figure 4: A system must transmute data into information, information into knowledge, and add experience and situational awareness to create the wisdom necessary to make trustworthy decisions.
  • Figure 5: Methodologies for unsupervised anomaly detection. Presently the most rudimentary method is primarily used: setting warn and alarm thresholds on parameters.
  • ...and 6 more figures