Abstract
Electromagnetic devices often feature components that are in relative motion to each other. In numerical models, this is commonly implemented via dynamical coupling. Hence, two independent model parts are coupled via coupling matrices for each motion increment. This concept has been combined with model order reduction (MOR) to create highly efficient yet accurate surrogate models. A common approach is to sample already coupled models, which are then reduced and interpolated. This work proposes to interpolate the coupling matrices instead of the already coupled ones. This physically-motivated concept achieves superior accuracy and is fully compatible to MOR at just slightly higher computational cost. A numerical case study of a moving magnet demonstrates the claimed properties.