Design of acoustic Cloak using generative modeling and gradient-based optimization

Publication Date


Document Type

Conference Proceeding

Publication Title

Proceedings of INTER-NOISE 2021 - 2021 International Congress and Exposition of Noise Control Engineering




This talk presents a gradient-based global optimization method with generative neural networks that has the potential to search for the globally optimized cloaking devices over a wide range of parameters. A generative model, 2D-GLOnets, is developed to design two-dimensional broadband acoustic cloaking devices by perturbing the positions of each scatterer in planar configuration of cylindrical scatterers. Such optimized cloaking devices can efficiently suppress the total scattering cross section to the minimum at certain parameters over the range of wavenumbers. During each iteration, 2D-GLOnets generates a batch of metamaterials and computes the total scattering cross section and its gradients using an in-house built multiple scattering solver [1, 2]. The model is trained to optimize the probability of generating the globally optimized meta-device in the design space. The geometric constrains of cylinders' positions are enforced by reparameterization, which maps all infeasible devices into the constrained design space. Our obtained results show that our model can yield devices with a better performance than fmincon optimization solver at broad range of wavenumbers.

Funding Sponsor

San José State University


Mechanical Engineering