Page 23 - IJAMD-2-2

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

Figure 11. Process of computational modeling for the generation of novel biomaterials. Reproduced from Magennis et al. 100

screening a large material library in combination with AI approaches commonly employed in bioink and
computational modeling. As a result, they identified a class biomaterial ink design include supervised learning and
of polymers with outstanding antimicrobial properties reinforcement learning. In supervised learning, models
that can inhibit biofilm formation, thereby enhancing are trained to predict the properties of inks in advance,
the safety of medical devices and bioprinted scaffolds. allowing researchers to evaluate potential formulations
This study highlights the potential of AI and data-driven quickly. Meanwhile, reinforcement learning enables
approaches in the optimization of multifunctional AM models to explore the search space and identify optimal
materials, demonstrating their applicability not only for ink formulations in minimal steps, effectively learning
mechanical enhancement but also for the improvement of from iterative experimentation.
biofunctional properties.
On the prediction side, Qavi et al. examined the
104
5.3. Bioink and biomaterial ink for AM relationship between rheological properties and printability
in multi-material bioinks for extrusion-based bioprinting
The design of bioinks and biomaterial inks is fundamental (EBB). Using a design of experiment approach coupled
in bioprinting, as these materials form the basis for creating with response surface methodology, the study optimized
functional, 3D biological structures. Bioinks consist of bioink formulations containing sodium alginate, gelatin, and
living cells embedded within a biocompatible matrix, laponite, focusing on parameters, such as zero shear viscosity
whereas biomaterial inks may not contain living cells but and storage modulus. By training an ANN on the obtained
are used to construct scaffolds or structures that support dataset, the relationships between the parameters were
cellular activities. Both types of inks must be carefully generalized, achieving a maximum mean absolute error of
101
formulated so that their biological and mechanical 6.3% in predicting the printability of the bioink formulations.
properties are compatible with bioprinting and suitable for
their intended purpose after the printing process. Crafting Lee et al. addressed the challenges in designing
105
an ideal bioink is a multi-objective problem, requiring a biocompatible 3D-printable bioinks by developing an
delicate balance of various properties, such as printability, ML-based method to create viable bioinks using collagen,
biocompatibility, biomimicry, mechanical integrity and hyaluronic acid, and fibrin. They established a relationship
stability, and biodegradability. 102,103 For instance, enhancing between ink mechanical properties and printability,
mechanical integrity and stability often results in reduced highlighting that a high elastic modulus enhances shape
biocompatibility, and vice versa. To discover an optimal fidelity while extrusion remains feasible below the critical
formulation, a large number of different compositions yield stress. Using multiple regression analysis, they
must be tested, each representing a different trade-off developed a model to predict whether a composition
between competing objectives. This is where AI becomes has a high elastic modulus and low yield stress. Various
invaluable, significantly improving the efficiency of bioink formulations were designed to maximize shape
material search and development by helping navigate the fidelity, leading to successful 3D constructs with viable and
complex multidimensional space of possible formulations. proliferative cells.

Volume 2 Issue 2 (2025) 17 doi: 10.36922/IJAMD025100007