Closing the loop: deploying auto-generating digital twins for particle accelerators

Abstract Digital twins provide a powerful means of interrogating complex physical systems and enhancing their performance across the full operational lifecycle. This article presents an auto-generating digital twin architecture for particle accelerators, where high system complexity and limited, costly beam time make digital twinning vital for improving performance and increasing throughput. A virtual control system mirrors the physical accelerator hardware, driving updates to a simulation model that returns results through virtual diagnostics. All of the accelerator lattice information is cascaded from a single ground source of truth, eliminating ambiguity between hardware and model representations. The modular, extensible design allows researchers to integrate their own simulation tools, machine-learning models, and lattice definitions while preserving structural coherence across the digital twin. This architecture has been demonstrated for three accelerator facilities—the CLARA electron beam facility, the ISIS neutron and muon source, and the proposed UK XFEL—establishing a foundation for a collaborative, community-driven effort toward a generic digital-twin solution for accelerator operation.