ML-Based Acceleration Strategies for FSI Simulation of Deformable Microcapsule Dynamics
Please login to view abstract download link
The synthesis of deformable microcapsules is an alternative for transporting active ingredients in blood vessels for therapeutic targeting. Their numerical modelling and simulation are complicated by the fluid-structure interaction involving the carrying fluid, the capsule membrane, the time evolution, and potentially complex three-dimensional geometries. To achieve less computationally intensive FSI digital twins, model order reduction (MOR) of full-order models (FOM) appears to be the most appropriate approach. In this work, we discuss a reduction strategy using an empirical interpolation method (EIM) that simplifies the calculation of loads on the capsule as well as the surface integrals in the fluid solver, leading to either FOM-ROM or ROM-ROM FSI coupling. An online ‘safety’ procedure tests the accuracy of the reduced-order models and updates them if necessary; it makes use of ‘validation’ empirical interpolation points. Performance analysis allows us to get an estimation of the potential speedup.
