Answers are grounded in this paper and cite their sources.
Microstructures enable programmable effective properties, but their fabrication by layer-wise additive manufacturing is often limited by overhang-induced support requirements. Existing overhang-aware topology optimization methods can generate self-supporting designs, yet their iterative, forward-design nature limits the rapid inverse generation of designs with prescribed properties. In contrast, property-conditioned generative AI models enable efficient synthesis but typically lack hard supportability enforcement during decoding. To address this gap, we propose PEAR , a P rinting- E nforced A uto R egressive framework for generating self-supporting 2D rasterized microstructures under prescribed elastic tensors and a discrete layer-wise support rule. PEAR formulates layer-wise fabrication as a bottom-to-top conditional sequence prediction, in which each new row is restricted to the supportable region induced by the previously generated geometry. During decoding, a dynamically updated support mask projects the predicted probabilities onto the admissible distribution, assigning zero probability to unsupported material placements, thereby guaranteeing compliance with the adopted discrete layer-wise support rule. Finite-element re-evaluation of generated test samples demonstrates that PEAR accurately matches the prescribed properties.
No takes yet. Share an insight, caveat, or question.
Hu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: