Designing the futuristic dielectric elastomer minimum energy structures using artificial neural networks (ANN)

Abstract

Dielectric elastomer minimum energy structures (DEMES) have gained significant attention for their ability to switch between multiple equilibrium states. These structures are formed when a pre-stretched elastomer film adheres to an inextensible frame and achieves equilibrium through energy minimization. Traditional methods for analyzing DEMES mechanics-numerical, theoretical, and experimental are often labor-intensive and time-consuming. This paper introduces the application of artificial neural network (ANN) techniques to predict the behavior of DEMES-based actuators efficiently. Using the Levenberg–Marquardt and Bayesian Regularization algorithms, the performance of two prototypes: the four-arm gripper and the flapping-wing actuator previously studied experimentally and numerically in Khurana et al. (2024a), is predicted. The ANN-based approach demonstrates excellent agreement with the numerical results while significantly reducing computation time. This study highlights the potential of ANN techniques as a fast and reliable tool for the parametric evaluation of DEMES structures, streamlining the design and analysis process. Future applications of DEMES, enhanced by ANN-based predictive models, include the development of adaptive soft robotics, bio-inspired actuators, and energy-efficient morphing structures. These advancements could lead to intelligent material systems with real-time control capabilities for biomedical devices, aerospace engineering, and wearable technologies. © 2026