Page 24 - IJAMD-2-2

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

Xu et al. introduced an ML framework to predict parameters using AI-driven sensors, as well as enabling
106
the viscosity of heterogeneous bioink compositions. rapid optimization of print parameters. For example,
Traditional models, such as the Cross model; fall short in Huang et al. introduced an ML-based model to predict
110
this domain due to the non-Newtonian nature of bioinks. cell numbers in bioprinted droplets by analyzing droplet
Their approach leveraged BO to work with sparse datasets, velocity. Utilizing a non-destructive optical system, the
employing a mask technique to define feasible parameter study accurately detected the presence and quantity
spaces based on domain expertise. By balancing the of cells in droplets. Among the evaluated models, RF
exploration of new possibilities and exploitation of existing regression achieved 80% accuracy for single droplet
data, their AI-guided BO framework effectively reduced cell presence, while extra tree regression had the lowest
experimental workload and streamlined the building of error at 12% for cell number predictions across multiple
the surrogate model for viscosity prediction. droplets. Besides, the gel fraction of hydrogel during

Qiao et al. explored the application of AI in bioink bioprinting can be measured with the ML or DL model
107
development for cryobioprinting, which integrates through in situ measurement of the hydrogel’s ultraviolet
extrusion bioprinting and cryopreservation to enhance transmissivity. 111
bioink shelf availability without using dimethyl sulfoxide Moreover, reinforcement learning-based rapid print
due to its potential toxicity. They developed a gelatin parameter optimization can increase sample count and
methacryloyl (GelMA)-based bioink incorporating efficiency. Bonatti et al. proposed a DL-based control
112
cryoprotective agents (CPAs) and assessed two CPA system to reduce the trial-and-error in EBB. The authors
formulations, finding that ethylene glycol outperformed collected a high-resolution video dataset of various
glycerol. Using this dataset, they established four ML EBB parameters and trained a CNN to optimize print
models, with the ANN showing the highest predictive parameters and monitor the process in real time. This
accuracy. This ANN model was successfully applied to control loop halted erroneous prints to conserve resources
predict outcomes for various CPA-based formulations, and time while integrating the ML model with existing
showcasing the potential of ML in the development of mathematical models, demonstrating the potential of
effective cryoprotective bioinks for cryobioprinting. ML to automate and ensure quality in EBB. Chen et al.
113
In reinforcement learning applications, Ruberu et al. introduced an AI-assisted high-throughput printing-
108
demonstrated how BO can quantitatively evaluate and condition-screening system (AI-HTPCSS) to enhance the
optimize the printability of biomaterial inks, such as optimization of printing conditions for 3D bioprinting.
GelMA and hyaluronic acid methacrylate, by adjusting This system integrated a programmable pneumatic
GelMA composition, ink reservoir temperature, pressure, extrusion printer with an AI-driven image-analysis
speed, and platform temperature. They significantly algorithm to efficiently screen conditions for printing
reduced the number of experiments required for 3D uniformly structured hydrogel scaffolds. The optimized
bioprinting optimization from 10,000 to an order of conditions achieved through AI-HTPCSS led to scaffolds
109
10. Hashemi et al. also explored the development of a with favorable mechanical properties, improved in vitro
chitosan-gelatin-agarose biomaterial ink optimized for biological performance, and effectiveness in enhancing
extrusion-based 3D bioprinting through BO. The study diabetic wound healing in vivo.
focused on achieving desirable mechanical properties, The integration of AI into bioprinting has advanced
biocompatibility, and precise printability. The optimized bioink design by optimizing formulation properties, such
ink composition (27% agarose, 53% chitosan, and 20% as printability and biocompatibility. However, the field is
gelatin) showed promise for fabricating complex 3D still emerging, challenged by the lack of extensive datasets
tissue constructs. The optimized biomaterial ink exhibited necessary for robust AI models. In the future, AI has the
sufficient viscosity for reliable printing while maintaining potential to revolutionize bioprinting further by enabling
shape integrity in 3D structures. Biological evaluations real-time process monitoring and adaptive control, thereby
involving bone marrow mesenchymal stem/stromal cells enhancing the precision and functionality of bioink
revealed that the ink supported favorable cell adhesion, development.
growth, and viability.
However, a major challenge in applying AI to 6. Summary
bioprinting is the slow collection of biologically relevant As summarized in Figure 12, this review examines how
datasets, such as cell viability. A small sample size can AI is transforming material development for AM, with a
hinder the development of robust models. To mitigate this focus on material design and performance optimization.
issue, AI can be applied for fast measurement of crucial The integration of AI with AM processes has facilitated

Volume 2 Issue 2 (2025) 18 doi: 10.36922/IJAMD025100007