Graded multiscale topology optimization using neural networks

In this paper, we propose a novel graded multiscale topology optimization framework by exploiting the unique classification capacity of neural networks. The salient features of this framework include: (1) the number of design variables is only weakly dependent on the number of pre-selected microstructures, (2) it guarantees partition of unity while discouraging microstructure mixing, (3) it supports automatic differentiation, thereby eliminating manual sensitivity analysis, and (4) it supports high-resolution re-sampling, leading to smoother variation of microstructure topologies. The proposed framework is illustrated through several examples.