Opin vísindi

Supervised-learning-enabled EM-driven development of low scattering metasurfaces

Show simple item record

dc.contributor Reykjavik University
dc.contributor Háskólinn í Reykjavík
dc.contributor.advisor Slawomir Koziel
dc.contributor.author Abdullah, Muhammad
dc.date.accessioned 2023-03-01T11:45:11Z
dc.date.available 2023-03-01T11:45:11Z
dc.date.issued 2022-05-13
dc.identifier.isbn 978-9935-9655-0-9
dc.identifier.uri https://hdl.handle.net/20.500.11815/4040
dc.description.abstract The recent advances in the development of coding metasurfaces created new opportunities to elevate the stealthiness of combat aircrafts. Metasurfaces, composed of optimized geometries of meta-atoms arranged as periodic lattices, are devised to obtain desired electromagnetic (EM) scattering characteristics, and have been extensively exploited in stealth applications to reduce radar cross section (RCS). They rely on the manipulation of backward scattering of electromagnetic (EM) waves into various oblique angles. Despite potential benefits, a practical obstacle hindering widespread metasurface utilization is the lack of systematic design procedures. Conventional approaches are largely intuition-inspired and demand heavy designer’s interaction while exploring the parameter space and pursuing optimum unit cell geometries. Another practical obstacle that hampers efficient design of metasurfaces is implicit handling of RCS performance. To achieve essential RCS reduction, the design task is normally formulated in terms of phase reflection characteristics of the unit cells, whereas their reflection amplitudes—although contributing to the overall performance of the structure—is largely ignored. A further practical issue is insufficiency of the existing performance metrics, specifically, monostatic and bistatic evaluation of the reflectivity, especially at the design stage of metasurfaces. Both provide a limited insight into the RCS reduction properties, with the latter being dependent on the selection of the planes over which the evaluation takes place. As a consequence of raised concerns, the existing design methodologies are still insufficient, especially in the context of controlling the EM wavefront through parameter tuning of unit cells. Furthermore, they are unable to determine truly optimum solutions. Therefore, we have introduced a novel machine-learning-based framework for automated and computationally efficient design of metasurfaces realizing broadband RCS reduction. We have employed a three-stage design procedure involving global surrogate-assisted optimization of the unit cells, followed by their local refinement. In its final stage, a direct EM-driven maximization of the RCS reduction bandwidth has been performed, facilitated by appropriate formulation of the objective function involving regularization terms. Moreover, to handle the combinatorial explosion in the design closure of multi-bit coding metasurfaces, a sequential-search strategy has been developed that enabled global search capability at the concurrent unit cell optimization stage. Latterly, the metasurface design task with explicit handling of RCS reduction at the level of unit cells has been introduced that has accounted for both the phase and reflection amplitudes of the unit cells. The design objective has been defined so as to directly optimize the RCS reduction bandwidth at the specified level (e.g., 10 dB) w.r.t. the metallic surface. The appealing feature of the said framework has consisted in its ability to optimize the RCS reduction bandwidth directly at the level of the entire metasurface as opposed to merely optimizing unit cell geometries. Besides, the obtained design has required minimum amount of tuning at the level of the entire metasurface. Lastly, a new performance metric for evaluating scattering characteristics of a metasurface, referred to as Normalized Partial Scattering Cross Section (NPSCS), has been proposed. The metric involved integration of the scattered energy over a specific solid angle, which allows for a comprehensive assessment of the structure performance in a format largely independent of the particular arrangement of the scattering lobes. Our design methodologies have been utilized to design several instances of novel scattering metasurface structures with the focus on RCS reduction bandwidth enhancement and the level of RCS reduction. Experimental validations confirming the numerical findings have been also provided. To the best of the author’s knowledge, the presented study is the first systematic investigation of this kind in the literature and can be considered a step towards the development of efficient, low-cost, and more high performing scattering structures.
dc.format.extent 129
dc.language.iso en
dc.publisher Reykjavik University
dc.rights info:eu-repo/semantics/openAccess
dc.subject Metasurfaces
dc.subject Optimization
dc.subject Surrogate modeling
dc.subject Supervised learning (Machine learning)
dc.subject Radar cross sections
dc.subject Scattering (Mathematics)
dc.subject Electromagnetic waves
dc.subject Bestun
dc.subject Líkanagerð
dc.subject Vélrænt nám
dc.subject Ratsjár
dc.subject Rafsegulbylgjur
dc.subject Doktorsritgerðir
dc.title Supervised-learning-enabled EM-driven development of low scattering metasurfaces
dc.type info:eu-repo/semantics/doctoralThesis
dc.contributor.department Department of Engineering (RU)
dc.contributor.department Verkfræðideild (HR)
dc.contributor.school School of Technology (RU)
dc.contributor.school Tæknisvið (HR)

Files in this item

This item appears in the following Collection(s)

Show simple item record