1. Data ‐Driven Long‐Term Energy Efficiency Prediction of Dielectric Elastomer Artificial Muscles.
- Author
-
Li, Ang, Cuvin, Phil, Lee, Siyoung, Gu, Jiahao, Tugui, Codrin, and Duduta, Mihai
- Subjects
- *
ENERGY consumption , *SOFT robotics , *ENERGY conversion , *LEARNING strategies , *MACHINE learning - Abstract
The widespread adoption of dielectric elastomer actuators (DEAs) as compliant artificial muscle is on the horizon after a wide range of prototype demonstrations. The next step is to accelerate the material optimization for specific design requirements. This work proposes a data‐driven framework to predict long‐term DEA energy consumption and efficiency using measurements of initial electrical properties. DEA datasets are generated from 242 pre‐stretched single‐layer actuators and 53 multi‐layer bending actuators actuated for 30–180min. Devices are made with different elastomer and electrode materials, and tested with a novel instrument that can measure multiple aspects of DEA actuation. First, experiments are conducted to develop an empirical understanding of the electro‐mechanical energy conversion mechanism and the impact of material choices during DEA actuation. Second, data‐driven models are applied to the datasets to predict energy consumption and efficiency. Third, the potential generalization of this approach is investigated by using transfer learning strategies to predict energy efficiency at 100 min with as little as 1 min of input data. Moreover, it is found that linear regression can extend the prediction up to 180 min. Additionally, transfer learning is applied to predict energy‐related properties of multi‐layer DEAs using a small dataset. The proposed data‐driven framework can evolve into an intelligent system for accelerating new material discovery by predicting performance and providing device optimization strategies. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF