To contribute to filling this research gap, we propose a unified parametric origami design workflow based on grasshopper combined with a multi-objective optimization process. To this end, first, a parametric model for a ring-shaped fourfold origami structure, called the Miura-oRing metastructure, is...
In a typical device optimization, the goal is to minimize or maximize a figure or meritF(p)F(p)that depends on a set ofNNdesign parametersp=(p1,...,pn)p=(p1,...,pn). The figure of merit can be any quantity used to benchmark the device’s performance and the design parameters c...
a ML workflow for material design where an RL agent (material generator) is assigned rewards based on predicted material properties in a feedback loop. b Inorganic materials generation process. At each step of the generation sequence for both the policy gradient network (PGN) and deep Q-network...
In addition to the high fidelity of the generated candidates with this approach, we find they adequately cover the distribution of the original training data and while maintaining a high degree of diversity in the design parameter space. To demonstrate this, distributions of the training data and ...
LIGENTEC used Photonic Inverse Design (PID) capabilities in Ansys Lumerical FDTD for the design and optimization of its waveguide crossing.
The benefit of combining an automated AI-Chemist with a computational brain has been demonstrated in the design and fabrication of films with optimal chiroptical performance27. AI-based methods have been widely used to accelerate the discovery of new materials, including photonic materials28,29,30,31...
The proposed framework has its advantages in portability and flexibility to naturally incorporate the parameterization, physics simulation, and objective formulation together to build up an effective inverse design workflow. A series of adaptive strategies for smoothing radius and learning rate updating are...
Inverse problem solving (or inverse design) comprises a rather heterogeneous collection of methods that are characterized by first setting the performance requirement(s) and then obtaining the optimal configuration of a material, geometry or process through a search targeting the selected performance. Fro...
In this work, we present a rapid inverse design methodology that can produce designs that replicate tailored mechanical behaviors upon loading via ML and desktop 3D printing. Our approach leverages generative inverse and surrogate forward neural network (NN) models (details of the ML workflow shown ...
aML workflow for material design where an RL agent (material generator) is assigned rewards based on predicted material properties in a feedback loop.bInorganic materials generation process. At each step of the generation sequence for both the policy gradient network (PGN) and deep Q-network (DQN...