Data‐Driven Design of Bimodal Networked Dielectric Elastomers for High‐Performance Artificial Muscles

Abstract

Dielectric elastomers (DEs) are soft, electroactive materials capable of large deformations, fast actuation, and high compliance, making them promising for wide applications such as soft robotics, wearable electronics, and energy harvesting. Bimodal networked DEs have emerged as a class of high‐performance DEs owing to their tunable mechanical properties and electromechanical instability (EMI) suppression capability. However, their development is limited by complex synthesis and empirical trial‐and‐error methods. Here, an AI‐assisted design framework is presented that integrates experimental data with machine learning models and demonstrates it on the design of a typical bimodal networked DE material, processible high‐performance dielectric elastomer (PHDE). The neural network in the framework enabled accurate prediction of mechanical properties, while the support vector machine reliably classified EMI behavior. The framework also achieved reverse design of DEs from user‐defined targets in which experimentally validated formulations matched predictions and demonstrated EMI suppression. A soft gripper fabricated from the optimized PHDE materials lifted 25 times its own weight, demonstrating the practical utility of the framework. This study establishes a data‐efficient and scalable approach for designing advanced DE‐based artificial muscles, accelerating discovery while reducing experimental workload.