Page 21 - IJAMD-2-2

International Journal of AI for
Materials and Design AI-driven material development for AM

5.2.1. Material design for polymer AM negative Poisson’s ratios, covering a broad range of elastic

In polymer AM, material design primarily revolves properties. This scalable framework extends to multiphysics
around structural engineering rather than composition metamaterials, advancing the AI-driven automated design
of high-performance materials.
optimization, as modifying polymer composition is far
82
less feasible compared to metals. To date, very few studies Similarly, Xue et al. proposed an AI-driven
have employed AI for the targeted optimization of polymer optimization framework for the automated design of
feedstock composition or molecular-level properties. composite mechanical metamaterials. Their approach
Unlike metals, where alloying elements can be systematically utilized a variational autoencoder to encode representative
adjusted to tailor phase stability, mechanical properties, volume element images into a latent space, enabling
and processing behavior, polymer properties are largely efficient exploration of material distributions. Bayesian
dictated by their intrinsic chemical structures, molecular optimization (BO) was then employed to identify optimal
weight distributions, and polymerization mechanisms, representative volume element configurations that achieve
making composition-based optimization significantly target macroscopic elastic moduli. The optimized designs
more constrained. In addition, synthesizing new printable were fabricated using multi-material 3D printing and
polymers often requires extensive chemical modifications experimentally validated, demonstrating the framework’s
and rigorous processing validation, further limiting rapid reliability in generating high-performance metamaterials.
material innovation. As a result, this review focuses on the Bessa et al. extended AI-driven metamaterial design
84
AI-assisted design of mechanical metamaterials in polymer to brittle polymers, employing a Bayesian ML framework
AM – an emerging direction where structural architecture, to develop super-compressible metamaterials. As seen
rather than chemistry, defines material performance. in Figure 9, their approach employed sparse Gaussian
These metamaterials achieve properties such as auxetic processes to model uncertainty and identify recoverable
behavior (negative Poisson’s ratio), programmable structures with extreme deformation capabilities. By
93
mechanical responses, controlled buckling behavior, systematically adapting designs to different length scales
95
94
97
shape morphing, and acoustic band gap through precise and material constraints, they demonstrated the potential
96
structural design rather than relying solely on material of AI-driven approaches in creating lightweight, tunable,
composition. By integrating computational modeling, ML, and highly deformable metamaterials.
and multi-objective optimization, AI facilitates the design
of next-generation metamaterials tailored for specific AM 5.2.2. Performance optimization in polymer AM
applications. Beyond metamaterial design, AI-driven approaches
In polymer AM, VPP techniques are particularly well- play a central role in optimizing process parameters and
suited for fabricating metamaterials due to their high predicting material performance for polymer AM. These
resolution, surface quality, and processing speed. However, methods enable data-driven decision-making, significantly
traditional design methodologies based on prior knowledge improving the efficiency and accuracy of mechanical
89
and intuition are increasingly inadequate for achieving performance predictions. 88,98 For example, Zhang et al.
next-generation metamaterial designs with optimized employed a long short-term memory neural network,
performance. In addition, the computational cost associated a DL algorithm adept at handling sequential data, to
with exploring extensive lattice configurations using FEA predict the tensile strength of polylactic acid components
presents a significant bottleneck, further restricting design based on in-process sensor data. The model captured
innovation. To overcome these limitations, AI-driven layer-wise temporal dependencies in FDM and achieved
approaches have been increasingly employed to automate higher predictive accuracy than traditional ML models,
the design process and optimize metamaterial architectures. underscoring the advantage of DL in time-series analysis
By efficiently navigating the vast design space, AI enables for polymer AM.
the discovery of novel lattice structures with optimized In a generative design context, Lee et al. applied
90
properties while reducing reliance on computationally AI-based optimization using Bézier curve manipulation
expensive simulations. This integration of AI with polymer and finite element simulations to design polymer lattice
AM facilitates the rapid development of high-performance structures with superior mechanical performance.
metamaterials tailored for advanced applications. Chen Their approach involved iteratively adjusting lattice
et al. developed an AI-driven computational approach beam geometries and evaluating mechanical responses,
81
for the automated discovery of mechanical metamaterials effectively shortening design cycles and outperforming
with extreme properties. Experimental validation confirms human-guided design in terms of both modulus and
its effectiveness in discovering auxetic structures with strength.

Volume 2 Issue 2 (2025) 15 doi: 10.36922/IJAMD025100007