NSTO: Neural Synthesizing Topology Optimization for Modulated Structure Generation

Nature evolves structures like honeycombs at optimized performance with limited material. These efficient can be artificially created the collaboration of structural topology optimization and additive manufacturing. However, extensive computation cost causes low mesh resolution, long solving time, rough boundaries that fail to match requirements for meeting growing personal fabrication demands printing capability. Therefore, we propose neural synthesizing leverages a self-supervised coordinate-based network optimize significantly shorter where encodes material layout as an implicit function coordinates. Continuous solution space is further generated from tasks under varying boundary conditions or constraints users' instant inference novel solutions. We demonstrate system's efficacy broad usage scenario through numerical experiments 3D printing.