Optimize Antenna Designs using Machine Learning

Smit Baua_20523 · August 2021

The Challenge

A complex design problem involves design parameters that interact with one another and have several design goals. Solving such complex design problems with an approach based on intuition can be challenging. Machine learning (ML) is a method of data analysis that automates building analytical models.

As antennas are becoming more and more complex, their designers can take advantage of machine learning to generate mathematical models for their physical antenna designs and perform fast, intelligent optimization on these mathematical models. With Altair as simulation partner, customers can achieve this by coupling Altair Feko and Altair Hyperstudy. This blog focuses on demonstrating the workflow that helps in plugging Feko models and responses in Hyperstudy with ease.

Altair Feko is a well-known and trusted numerical analysis tool for a wide range of problems in electromagnetics. Its efficient solvers make it a very good tool to utilize as part of a process that explores solution spaces or performs advanced optimization tasks in electromagnetics.

Altair HyperStudy is a multidisciplinary design study software that enables engineers to explore and optimize their product performance and robustness. By using automatic processes combining state-of-the-art mathematical methods, predictive modeling and data mining, HyperStudy explores the design space of any system model smartly and efficiently. Users are guided to understand data trends, perform trade-off studies, and optimize design performance and reliability, while considering multiphysics constraints.

Patch Antenna Example

Here, we take an example of patch antenna and optimize the design with two simple design variables with Hyperstudy. The image below shows the high level workflow of optimizing a patch antenna design using Feko and Hyperstudy.