Speaker
Description
In modern accelerator modeling, many lattice components can be accurately described using established first-principles physics. However, certain intricate effects, such as complex boundary conditions, collective interactions, and self-fields, remain difficult to model reliably and efficiently from theory alone. At the same time, high-resolution beam position and profile measurements provide rich information about these poorly understood dynamics. In this paper, we present a hybrid modeling framework with built-in automatic differentiation, designed to seamlessly integrate physics-based lattice models with data-driven representations of complex effects. This approach improves predictive accuracy, enables gradient-based optimization, and offers a practical path toward more faithful digital twins of accelerator systems.
| Paper status | Resubmitted proceeding files received and assigned to an editor. Accepted by Submitter. |
|---|