Fig. 1: System components of the DL-assisted nonparameterized MEMS design.
From: Deep learning for non-parameterized MEMS structural design

a The major components for the proposed system architecture, which include the training process and the testing process. b An example disk resonator pattern generated from Brownian motion, followed by folding along the two axes of symmetry. The pattern is represented as a 100 × 100 pixel matrix, where 0 is a void element and 1 is a solid element. Inset: representative finite element mesh for a 2 × 2 solid region. c The structure of the deep residual neural network model used in this work, featuring input images, 2D features, residual blocks, the neural network, and the output target.