Semantic Inverse Design of Cellular Structures in Implicit Field Space

• Paper: IDETC26

Highlights

• We propose a semantic inverse design framework for cellular structures that grounds qualitative engineering intent into explicit quantitative targets and hard constraints.

• We introduce a retrieval-and-verification pipeline in implicit field space that retrieves candidate unit-cell geometries and ranks them using surrogate-assisted property prediction.

• We demonstrate that the proposed framework achieves strong target-matching performance and physically consistent behavior, highlighting its potential as an alternative to unconstrained text-to-geometry generation.