An Efficient Adjoint-Based Approach for Computational Electromagnetic Design

Antennas optimized for free space environments often have sub-optimal performance when placed in an in-situ environment. Nearby structures can alter the radiation pattern of the antenna or conversely the antenna can alter the radiation pattern of nearby antenna. In either case, there is a need to optimize these antenna for the in-situ environment. Furthermore, these optimization tools must be efficient and find ways to avoid the high computational costs associated with forward solves. Typically, for antenna calculations, forward solves can require hours or even days thus implying that any optimization tool requiring hundreds of forward solves is infeasible. In this project, we propose to examine modern optimization algorithms, referred to as adjoint methods, that minimize the number of forward solves. In addition to the adjoint methods, we will explore algorithms that have strong convergence and thus require only a small number of iterations to find a converged optimal solution. To ensure that these tools are easy to use, we will explore their usage with a commercial computational antenna code called WIPL-D. We will show that an optimization toolkit can be built that easily interfaces with WIPL-D or any other commercial antenna code.