Building Digital Twins for Particle Accelerators: The Role of Architecture and Patterns

Abstract

Particle accelerators are complex machines consisting of hundreds of interdependent devices. Control systems and commissioning applications are used to steer, control, and optimize them. Online models—now increasingly implemented as digital twins—enable real-time derivation of critical beam parameters during operation.

This paper presents our experience building reusable and configurable digital twins for particle accelerators using architectural layering and software design patterns. By separating domain logic from control infrastructure and applying both established and custom patterns, we have developed a flexible twin framework capable of supporting multiple backends, simulation engines, and control systems.

We demonstrate how this architecture has enabled rapid deployment across two accelerator facilities (BESSY II and MLS) with minimal changes, while also improving testing and packaging via containerized delivery. Our findings highlight that intentional architecture and pattern-based design are essential for sustainable, adaptable scientific software in complex engineering domains.