Application Exchange

Tutorial files for COMSOL Webinar "Optical Antenna Optimization Using Genetic Algorithms"

Bryan Adomanis, USAF

The genetic algorithm (GA) optimization routine is directly applicable to a wide variety of physics and engineering problems, especially where the parameter space is large and the solution space has many extrema. In this webinar we demonstrated how to implement the GA routine in COMSOL Multiphysics, using the LiveLink for MATLAB module. We examined a multivalued objective function to identify the optimal geometry of an optical antenna--the "Omega particle"--such that it behaves as a Huygens source with maximal forward scattering and minimal backward scattering. Included are the MATLAB functions used to automatically build and solve each population of models, feed the results into a GA routine for evolution of the next generation, and monitor the process. The "LiveLink for MATLAB" module is required for use. "Webinar.m" is main script, with other supporting functions provided. "Webinar2.m" was used for showing a way to add more parameters ("genes") to the GA routine, adding the angle of the dipole arms of the particle. A model template is also included.

User Comments

莉波 武
Sep 27, 2019 at 6:48am UTC

Hi, your code is very intereting. It is known that there are may seeds in GA process. I am wandering how to fullfill the parallel jobs (GA seeds) to save the cost time. I appriate your suggestion.

异 王
Jul 26, 2020 at 9:06am UTC

Hello! I am also using this method of comsol with matlab to optimize the parameters of an electromagnetic model. I have the same question as to whether each solution can be calculated in parallel to save time when completing an iteration. I would like to ask if there is a solution.

Login to comment on this entry.