Page 41 - ESAM-1-1

Engineering Science in
Additive Manufacturing Gen-AI for lattice structure design

Figure 3. An inverse design workflow using conditional diffusion: from implicit representation to optimized lattice structures. Reproduced with permission
from Jadhav et al. 35

Table 1. Targeted properties optimized in lattice structure design using Gen‑AI

Targeted property Description Relevance in application References
Compressive strength Enhancing the ability of lattice structures to Crucial for load-bearing applications in aerospace, 53
withstand compressive loads. automotive, and structural parts.
Strength-to-weight ratio Maximizing strength while minimizing Essential in aerospace and automotive sectors to reduce 54
material usage and weight. weight when maintaining strength.
Energy absorption Designing structures that efficiently absorb and Important for protective structures, crash-resistant 55
dissipate energy during impact. components, and biomedical implants.
Stiffness Increasing the rigidity of structures to resist Critical for mechanical stability in structural and 55
deformation under applied loads. load-bearing applications.
Durability Enhancing the structure’s ability to maintain Important for applications exposed to fatigue or cyclic 55
performance over repeated loading cycles. stresses.
Elongation Improving the capacity of structures to undergo Relevant for applications requiring flexibility and 54
deformation before failure. toughness.
Abbreviation: Gen-AI: Generative artificial intelligence.

generated structures were then fabricated using LPBF design approach using GANs to create lattice structures
with AlSi10Mg alloy and subjected to comprehensive with enhanced mechanical properties. The study began
compression and impact tests. The results demonstrated with parametric design and simulated annealing to
that the GAN-generated lattice configurations exhibited generate a dataset of high-performance lattice structures.
superior mechanical properties, including a notable The GAN model was then trained to produce diverse lattice
increase in normalized energy absorption and extension geometries optimized for strength-to-weight ratios. These
capacities. This study highlights the potential of integrating designs were fabricated using material jetting AM and
generative deep learning with AM for designing high- validated through compression, impact tests, and FEM. The
performance, custom lattice structures. results showed that GAN-generated structures achieved up

To optimize the strength-to-weight ratio of lattice to 108% improvement in strength and 150% improvement
structures, Yüksel et al. developed a deep learning-based in elongation compared to standard designs. The study
55

Volume 1 Issue 1 (2025) 6 doi: 10.36922/ESAM025110006