Machine Learning-Optimized Design of 3D-Printed Cylindrical Luneburg Lenses

Designing a Luneburg lens is challenging due to its radially graded permittivity, which introduces fabrication complexity, material constraints, bandwidth limitations, and high costs. This work presents a machine learning-optimized 3D-printed multishell Luneburg lens using air voids in polylactic acid (PLA) to emulate ideal performance. A cylindrical prototype with a 100 mm diameter and 30 mm thickness is designed for X-band operation, closely matching the performance of a near-ideal 1000-shell lens. The proposed method provides a cost-effective and practical solution for high-performance antenna lenses, overcoming traditional manufacturing limitations.