Progressive DNN-metaheuristics for programmable broadband mechanical metamaterial absorbers design

Abstract

The rapid advancement of industrial and military mobility platforms necessitates lightweight yet load-bearing metamaterials capable of absorbing electromagnetic waves (EMWs) within target operating bands. However, co-optimizing broadband EMW absorption and mechanical robustness remains a formidable challenge, especially under demanding performance and design requirements, such as those for stealth drones and electronic device enclosures, underscoring the need for a more intelligent exploration of design variables. Herein, we present a new programmable framework for designing broadband mechanical metamaterial absorbers (BMMAs) that combines progressive sampling-based neural networks (NNs) and metaheuristic optimization, overcoming the limitations of conventional single-focus approaches. The proposed method efficiently navigates a mixed-variable design space, enabling rapid and precise performance predictions at a minimal computational cost. The effectiveness of this framework is demonstrated by successfully identifying the optimal BMMA, which employs an octet-truss structure that satisfies specific broadband frequencies spanning 5.8–18 GHz and mechanical stiffness constraints for target performance ranges. The optimal BMMA achieves an average EMW absorption of 95% with a full effective absorption bandwidth (≥90% absorption), and the experimentally observed absorption spectra closely match the simulations and NN predictions, with deviations of less than 4%. These findings highlight the promise of integrating NN-based metaheuristic optimization with additive manufacturing to engineer multifunctional metamaterials tailored for applications in electronics, communications, and aerospace. Furthermore, this comprehensive design strategy emphasizes the significant potential of advanced computational techniques for efficiently solving complex engineering problems.

Topics & Concepts