In-situ antenna measurements are often impractical or too costly, since they require the fabrication of the antenna and structural modifications of the platform for antenna installation. As the forward EM problems can be numerically solved ever more efficiently, it is possible to perform antenna-platform optimizations in a reasonable time. They provide allow the investigation of alternative antenna sites, material treatments, or antenna modifications to mitigate the platform or nearby structure effects in order to satisfy the performance requirements. In this proposal, it is assumed that the installed antenna is near optimum and only small adjustments or modifications are required to regain its performance characteristics. As such, the limited-memory BFGS with simple bound constraints is chosen to solve the large scale mutli-objective optimization problem. In this approach, the gradient of the vector functional is computed by the adjoint method and the explicit storage of the approximate Hessian computed based on a recursive equation is not required. The function evaluation will be performed by the method-of-moment solver in the commercial FEKO.
Benefit: The main benefit of this Phase I effort is a robust and efficient quasi-Newton optimization approach for general large scale 3-D radiation and scattering problems. In particular, it will provide useful information on necessary modifications of the antenna such as geometry, material properties, feed locations, and antenna site before it is physically installed on the platform in order to restore its intended performance or to reconfigure its radiation characteristics to minimize unwanted interference. This approach can be applied to refine the electromagnetic model of an electrically large scattering target from the limited near field measurements to ensure a high level of confidence in the far field RCS predictions of the resulting target model. BRC anticipates that the software package developed under this effort will become an essential engineering tool for system integration. It will significantly save time and money in reducing the frequency of extensive in-situ or flight testing to verify requirements. BRC intends to demonstrate the optimization software with the Northrop Grummans hybrid FEM/BI code called Switch-H. A successful demonstration will likely generate substantial interests within NG and in the RF community as a whole.
Keywords: limited-memory BFGS with simple bounds (L-BFGS-B), limited-memory BFGS with simple bounds (L-BFGS-B), Frchet differentiable, Inverse Problem, Method of Moments (MOM), adjoint method, Computational Electromagnetics, constrained non-linear optimization, in-situ antenna performance
