Page 42 - ESAM-1-1

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

Figure 4. A typical pipeline of the performance optimization of lattice structure achieved by Gen-AI. Reproduced with permission from Challapalli et al. 53
Abbreviation: Gen-AI: Generative artificial intelligence.
highlights the potential of deep learning in expanding the For simultaneous optimization of lattice structure and
design space and optimizing lattice structures for tailored process parameters, Duan et al. proposed a process-
56
mechanical performance. aware auxiliary classifier generative adversarial network
(ACGAN)-based computational framework for the design
4. Generative AI for process-aware design of lattice structures with target mechanical properties. The
of lattice structures framework consists of two ACGAN models: InverseACGAN
The design of lattice structures in AM is inherently complex, for generating design parameters (line distance, layer
as it involves not only the optimization of geometric height, and infill pattern) and process parameters (print
configurations but also the consideration of process speed and print temperature), and ForwardACGAN
parameters that significantly influence the final properties. for predicting mechanical properties like porosity and
Different materials, whether metallic or polymeric, present elastic modulus. Using a dataset of 675 experimentally
unique processing limitations in AM. fabricated samples, the framework demonstrated the
ability to generate parameters that resulted in fabricated
Metallic lattices (e.g., Ti-6Al-4V and AlSi10Mg): structures closely matching target properties, with average
Processing limitations include high residual stress, mean absolute percentage errors (MAPEs) of 6.481% for
warping, and thermal gradients, which can compromise porosity and 10.208% for compressive elastic modulus. The
the mechanical performance and dimensional accuracy of study highlights the framework’s robustness and efficiency
the final product. Process parameters such as laser power, in optimizing both design and process parameters for
scan speed, and layer thickness need careful optimization achieving desired mechanical performance in AM lattice
to mitigate these issues. structures.
Polymeric lattices (e.g., PLA and TPU): Processing
challenges include layer adhesion, surface finish, and 5. Generative AI for simulation efficiency of
shrinkage, especially in complex geometries. Optimizing lattice structures
parameters like print temperature, cooling rate, and print The integration of Gen-AI models into simulation
speed is essential for achieving high-quality prints. workflows has emerged as a transformative strategy
In this context, Gen-AI offers a transformative to enhance computational efficiency, particularly for
approach to tackle these challenges. By integrating data- complex lattice simulations in AM. Traditional simulation
driven models with traditional design methods, AI can techniques, such as FEM simulations or Monte Carlo
facilitate the generation of optimized lattice geometries methods, often encounter significant challenges, including
when accounting for AM process parameters. high computational costs, long processing times, and

Volume 1 Issue 1 (2025) 7 doi: 10.36922/ESAM025110006