Inverse Topological Design Method Based on NURBS Modeling for Frequency Selective Surfaces

To enhance design flexibility for frequency selective surfaces (FSSs), a machine learning-based inverse topological design method (ITDM) incorporating non-uniform rational B-splines (NURBSs), termed NURBS-ITDM, is proposed in this work. The proposed method utilizes NURBS curves to model the metallic pattern geometry on the substrate of the FSS unit cell. This method significantly expands the flexibility in defining patch or aperture shapes while reducing the number of modeling variables, thereby mitigating computational complexity and training burden. Furthermore, the inherent smoothness and continuity of NURBS curves eliminate jagged boundaries and point connection artifacts common in pixelated representations. The effectiveness of the proposed NURBS-ITDM is validated through two numerical examples, including a single-layer FSS with a rectangular unit cell and a double-layer FSS with a regular hexagonal unit cell. The designed double-layer FSS structures is fabricated and measured to further verify the performance of the proposed model.