Hybrid Reduced Order Modeling for the Inverse Design of Dielectric Elastomer Actuators
Please login to view abstract download link
Dielectric Elastomer Actuators (DEAs) are at the forefront of soft robotics, yet their optimization is often hindered by the high computational cost of high-fidelity, coupled electromechanical finite element (FE) simulations. To address this, we present a framework of Reduced Order Models (ROMs) designed to accelerate both the inverse design and material selection processes for DEAs. The framework is built upon data generated from high-fidelity FE simulations. Initially, we demonstrate a data-driven ROM utilizing nonlinear dimensionality reduction techniques. This approach effectively solves the inverse problem for bending DEAs, identifying the required magnitude and location of electric stimuli to achieve specific target shapes [1]. In a distinct yet related development, we introduce a projection-based hyper-reduced model [2] specifically tailored to accelerate the material selection process, allowing for rapid iteration across parametric material models for polymers. By leveraging data-driven manifold learning to improve the projection basis, this hybrid approach aims to solve complex inverse shape-response problems with greater accuracy and efficiency than traditional linear methods. This research highlights the synergy between computational mechanics and reduced order modelling, providing a pathway for the real-time design and control of sophisticated soft-actuated electromechanical systems. Acknowledgements: The group acknowledges the financial support received via project POTENTIAL (PID2022-141957OB-C21) funded by MICIN/AEI/10.13039/501100011033/ FEDER, UE. The first author also acknowledges the funding PREP2022-000220 provided by MCIN/AEI/10.13039/ 501100011033 and the FSE+.
