研究目的
Investigating the use of machine learning techniques for an efficient design of shaped-beam reflectarrays to accelerate the design process while providing accurate results.
研究成果
Machine learning techniques, specifically SVMs, have been successfully employed for the acceleration of reflectarray antennas design, providing high accuracy and considerable acceleration in the design process. The technique shows promise for applications requiring efficient design of shaped-beam reflectarrays.
研究不足
Discrepancies for low levels of the crosspolar pattern were identified, which are produced at very low levels where errors due to tolerances in manufacturing and measurement processes may play a more important role.
1：Experimental Design and Method Selection:
The study employs Support Vector Machines (SVMs) for the characterization of the reflection coefficient matrix to derive scattering parameters associated with unit cell dimensions, replacing Full-Wave analysis based on Local Periodicity (FW-LP).
2：Sample Selection and Data Sources:
The reflectarray unit cell is characterized, focusing on dipole lengths as the main factor contributing to phase-shift, with other parameters fixed.
3：List of Experimental Equipment and Materials:
The study uses a laptop with an Intel Core i7-4712MQ CPU for simulations.
4：Experimental Procedures and Operational Workflow:
The design process involves obtaining phase-shift distributions for dual-polarized reflectarrays, using SVMs within the design process to accelerate computations.
5：Data Analysis Methods:
The accuracy of SVMs is assessed by comparing designs obtained with SVMs and MoM-LP, including the influence of discretizing angles of incidence.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
