Directional Latent Hybridization: Beyond Random Noise in Physics‐Informed Generative Inverse Design of Nonlinear Metamaterials
Source abstract
Abstract
ABSTRACT Conventional metamaterial inverse design is hindered by stochastic generative models that lack physical grounding and often fail to reach optimal performance. Here, we introduce a framework coupling a physics‐informed neural operator (DeepONet) and a 3D geometry generator (3D‐cVAE) with a Directional Latent Hybridization (DLH) strategy. By merging dominant traits from parent geometries in the latent space, our approach enables deterministic latent optimization, yielding a high prediction accuracy for energy ( R 2 = 0.957) and force ( R 2 = 0.981). Unlike random noise‐based generation, which suffers from a 0% success rate at high volume fractions ( V r = 0.8), the DLH strategy maintains a 73% success rate and achieves significantly lower volume errors (< 4.81%). Experimental validation using additively manufactured thermoplastic polyurethane (TPU) lattices confirms that DLH‐optimized architectures exceed the performance of their base designs, achieving up to 24.03 J of absorbed energy and a peak force of 9.80 kN. This framework establishes a novel physics‐informed generative paradigm for discovering next‐generation metamaterials with physically interpretable and predictable performance.
Source: crossref · Quality: baseline
Affiliations
Department of Mechanical Engineering Seoul National University Seoul Republic of Korea
References (44)
Show source-provided references
1. 10.1007/978-981-15-6446-810.1007/978-981-15-6446-8 ↗
2. 10.1515/nanoph-2022-067110.1515/nanoph-2022-0671 ↗
3. 10.1177/108128651773569510.1177/1081286517735695 ↗
4. 10.3389/fmats.2023.127396110.3389/fmats.2023.1273961 ↗
5. 10.1016/j.cjmeam.2023.10009110.1016/j.cjmeam.2023.100091 ↗
6. Prediction and Experimental Validation Approach to Improve Performance of Novel Hybrid bio‐Inspired 3D Printed Lattice Structures Using Artificial Neural Networks10.1038/s41598-023-33935-0 ↗
7. Numerical/Experimental Investigation of Energy Absorption and Compressive Properties in 3D‐Printed bi/Multi‐Material Polymeric Metamaterials10.1016/j.heliyon.2025.e43416 ↗
8. 10.1038/s41563-020-0798-110.1038/s41563-020-0798-1 ↗
9. 10.1016/j.eml.2023.10201310.1016/j.eml.2023.102013 ↗
10. 10.1016/j.addma.2023.10345110.1016/j.addma.2023.103451 ↗
11. 10.1016/j.addma.2023.10375110.1016/j.addma.2023.103751 ↗
12. 10.1002/adma.20230253010.1002/adma.202302530 ↗
13. 10.1002/adma.20220623810.1002/adma.202206238 ↗
14. Inverting the Structure–Property Map of Truss Metamaterials by Deep Learning10.1073/pnas.2111505119 ↗
15. 10.1016/j.ijmecsci.2022.10753110.1016/j.ijmecsci.2022.107531 ↗
16. 10.1002/adma.20190111110.1002/adma.201901111 ↗
17. Designing Complex Architectured Materials with Generative Adversarial Networks10.1126/sciadv.aaz4169 ↗
18. 10.1038/s42256‐021‐00322‐110.1038/s42256‐021‐00322‐1 ↗
19. 10.1021/acs.nanolett.8b0317110.1021/acs.nanolett.8b03171 ↗
20. 10.1016/j.matdes.2024.11337710.1016/j.matdes.2024.113377 ↗
21. 10.1002/adfm.20240416510.1002/adfm.202404165 ↗
22. 10.1002/admt.20250029310.1002/admt.202500293 ↗
23. A Novel Approach for Identifying Rotational Stiffness of Semirigid Joints by Physics‐Informed Neural Networks10.1061/JCCEE5.CPENG-6151 ↗
24. 10.1016/j.cma.2020.11337710.1016/j.cma.2020.113377 ↗
25. 10.1016/j.ijmecsci.2025.11087510.1016/j.ijmecsci.2025.110875 ↗
26. 10.1016/j.ijmecsci.2025.11071210.1016/j.ijmecsci.2025.110712 ↗
27. 10.1016/j.ijmecsci.2025.11012310.1016/j.ijmecsci.2025.110123 ↗
28. 10.1038/s41467-023-40854-110.1038/s41467-023-40854-1 ↗
29. 10.1002/adma.20230348110.1002/adma.202303481 ↗
30. 10.1038/s41524‐020‐0309‐610.1038/s41524‐020‐0309‐6 ↗
31. 10.1002/adma.20242006310.1002/adma.202420063 ↗
32. 10.1016/j.ijmecsci.2022.10802910.1016/j.ijmecsci.2022.108029 ↗
33. 10.1016/j.mtphys.2022.10061610.1016/j.mtphys.2022.100616 ↗
34. Inverse Design of Truss Lattice Materials with Superior Buckling Resistance10.1038/s41524-022-00938-w ↗
35. 10.1002/adfm.20190903310.1002/adfm.201909033 ↗
36. 10.1016/j.actamat.2024.12070010.1016/j.actamat.2024.120700 ↗
37. 10.1016/j.ijmecsci.2022.10792010.1016/j.ijmecsci.2022.107920 ↗
38. 10.1016/j.matdes.2021.10993710.1016/j.matdes.2021.109937 ↗
39. 10.1002/adma.20231230310.1002/adma.202312303 ↗
40. L.Lu P.Jin andG. E.Karniadakis Deeponet: Learning nonlinear operators for identifying differential equations based on the universal approximation theorem of operators. arXiv preprint(2019): arXiv:1910.03193.
41. 10.1038/s42254-021-00314-510.1038/s42254-021-00314-5 ↗
42. 10.1016/j.ijmecsci.2025.11048910.1016/j.ijmecsci.2025.110489 ↗
43. 10.1016/j.cma.2022.11458710.1016/j.cma.2022.114587 ↗
44. Z.Li N.Kovachki K.Azizzadenesheli et al. Neural operator: Graph kernel network for partial differential equations. arXiv preprint(2020): arXiv:2010.08895.