Page 115 - MSAM-3-2

Materials Science in Additive Manufacturing Hybrid lattice structures design with AI

of lattice structures. Efforts were made by researchers costs. Particularly beneficial for addressing high-complexity
13
to investigate different cell topologies to improve the challenges characterized by a multitude of design
mechanical responses of lattice structures. 14-19 parameters, such as those encountered in the architecture
Lattice structures inspired by triply periodic design of lattice structures, AI algorithms excel due to their
39
minimal surfaces (TPMS) have captured increasing rapid and reliable nature. Research indicates that machine
attention within the research community due to their learning algorithms have significantly accelerated the
promising combination of lightweight properties and process of lattice structure design compared to conventional
high mechanical performance. 20-22 Mathematically, methods, such as trial-and-error or analytical physics-based
39
TPMS is characterized by zero mean curvature, three approaches. Moreover, the advance of new technologies
dimensionally periodic, and smooth topology, which has not only facilitated quicker solutions but has also
is achieved through the minimization of an area within enabled the discernment of previously unseen correlations
40
defined surface boundaries. In practical terms, structures between design parameters and mechanical performance.
with TPMS-like topologies can be modeled using level- The growing demand for novel structures with distinctive
set approximation methods. Two primary strategies functionalities serves as a driving force for researchers to
23
are employed in generating a TPMS-based lattice. In explore innovative approaches aimed at accelerating the
the network phase TPMS structures, one sub-domain development and discovery of novel architectures. Efforts
of the space isolated by the surface is filled with a solid toward automating the design process and enhancing the
material, while the matrix phase is formed by assigning evaluation rate of new structures have been underscored in
41,42
thickness to the TPMS surface. Evaluation of the recent literature. This necessitates the establishment of
20
mechanical properties of TPMS-based lattices has been frameworks that utilize various AI techniques to streamline
carried out using various materials such as stainless these processes. Indeed, the pivotal role of AI in advancing
steel, aluminum alloy, polymer 26,27 and concrete, 28,29 the field of metamaterials is becoming increasingly evident;
24
25
illustrating superior properties to conventional strut- as it facilitates the automation of labor-intensive discovery
43-46
based structures. Furthermore, the geometry of TPMS- processes and unveils previously unnoticed patterns.
based structures can be tailored locally, as it is controlled With the aim of accelerating the design of hybrid
by mathematical equations. 30,31 For instance, Maskery lattice structures inspired by TPMS, this work explores
et al. demonstrated the feasibility of designing TPMS- the potential of predicting the mechanical properties of
based lattices with graded relative density, resulting in complex hybrid lattice structures using machine learning
customized collapse behaviors and enhanced energy methods. First, the strategy to design hybrid lattices
absorption under compression. Al-Ketan et al. proposed with TPMS-like unit cells, the homogenization method
32
a novel class of stochastic cellular structures inspired to generate datasets for training, and the architecture of
by TPMS, presenting superior mechanical responses at artificial neural networks are introduced in the paper.
high relative densities. Recently, Maskery and Ashcroft Then, the dataset evaluation, performance of the model,
31
proposed the concept of designing honeycomb-like and further validation by finite element (FE) modeling
structures based on TPMS. Peng et al. investigated the are discussed. Finally, the key findings of this work are
33
multifunctional properties of hybrid honeycomb-like presented.
lattices based on TPMS, revealing exceptional mechanical
properties and tunable responses. These TPMS-based 2. Materials and methods
34
honeycomb-like structures presented superior mechanical 2.1. Design of TPMS honeycomb cell
properties compared to traditional honeycomb structures.
Meanwhile, hybrid structures can be designed based on Mathematically, minimal surfaces are surfaces with zero
different unit cells to exhibit distinct properties that are mean curvature at a given point H = 0 (I)
not possible with monolithic structures. However, the where H=(k + k )/2; and k and k represent the
2
2
1
1
complexity of these structures presents challenges in their principal curvatures in orthogonal planes. 47
design and property evaluations.
The normal unit vector of the surface maintains a zero
In recent years, there has been a notable surge in interest divergence across all points, and the Gaussian curvature of
regarding the application of artificial intelligence (AI) the surface is expressed as follows:
and machine learning in the designs of metamaterials. 35-38 K = kk (II)
Notably, neural networks and evolutionary algorithms 12
have emerged as prominent examples, offering expedited When a minimal surface is replicated infinitely and
problem-solving capabilities with reduced computational periodically throughout three-dimensional (3D) space, it

Volume 3 Issue 2 (2024) 2 doi: 10.36922/msam.3430