The Integrated Reverse Design of Load-Bearing Metamaterial Absorbers Based on Deep Learning
摘要
To address the issues of poor load-bearing capacity, low design efficiency, and the difficulty of integrating structural and functional design in current absorbing structures, a study on the integrated reverse design of load-bearing metamaterial absorbers was conducted. A parametric model of the load-bearing metamaterial absorber was developed using a combination of composite core structures and electromagnetic resonant layers. A dataset of electromagnetic and structural properties of the absorber was established. A deep learning-based forward prediction network for the absorber’s absorbance and a design parameter reverse prediction network were developed and trained, achieving high-accuracy predictions of both the absorbance and design parameters. Based on the forward and reverse prediction networks, an integrated reverse design method for metamaterial absorbers was proposed. This method enables the reverse design of the metamaterial absorber on the basis of specified electromagnetic and structural properties while addressing the issue of unattainable design goals through polar coordinate transformation. The method was applied to reverse design tasks for single-frequency, multi-frequency, and broadband load-bearing metamaterial absorbers. The proposed integrated reverse design method holds significant potential for applications in radar stealth material design for military targets, such as naval vessels, and offers broad generalizability and engineering applicability.