Page 51 - IJAMD-2-2

International Journal of AI for
Materials and Design ML-driven optimization in additive manufacturing

such as CNNs and multi-objective optimization techniques The intricate microstructure of these lattices directly affects
like NSGA-II have been employed to fine-tune printing their electrochemical properties, including charge storage
parameters, ensuring improved tensile strength, flexural capacity and ion transport efficiency. ML models have
performance, and reduced void formation. In addition, enabled the design of 3D-printed carbon microlattices with
163
ML-based predictive modeling has been applied to estimate tailored properties, enhancing supercapacitor performance
the flexural strength of additively manufactured carbon fiber for next-generation energy storage solutions. 166
composites, aiding in material selection and process design. 164
Strain sensing and structural health monitoring
In this review, continuous fiber-reinforced composites represent another important area where ML has advanced
and nano-carbon composites–including those utilizing carbon-based AM. 3D-printed carbon nanotube/
carbon nanotubes and graphene–are collectively discussed polypyrrole/UV-curable composites have been studied for
under the category of carbon-based materials, as they share their strain-sensing capabilities, with ML models facilitating
common goals such as enhancing mechanical strength, accurate prediction of electromechanical responses and
electrical conductivity, and structural performance real-time sensor calibration. Similarly, graphene-based
167
through carbon-based reinforcements. Despite this shared self-powered strain sensors for smart tires in autonomous
objective, these materials differ significantly in terms of vehicles have been developed, utilizing ML for real-time
scale, structure, and manufacturing challenges. Continuous performance optimization and fault detection. 168
carbon fibers provide macroscopic reinforcement and
often induce anisotropic mechanical behavior in printed Graphene-based materials have also found applications
parts, requiring ML models that can account for fiber in gas sensing, Internet of Things (IoT)-enabled energy
orientation, continuity, and alignment during process harvesting, and biocompatible electronic interfaces.
monitoring or property prediction. In contrast, nano- ML has been employed to enhance the sensitivity
carbon composites employ dispersed carbon nanotubes and selectivity of textile-based graphene gas sensors,
or graphene as nanoscale fillers that influence rheological improving their performance in energy harvesting-
169
behavior, filler dispersion, and electrical percolation assisted IoT applications. In addition, graphene-based
networks. Accordingly, ML approaches in nano-carbon nanomembrane bioelectronics have been optimized using
systems tend to focus on optimizing material formulations ML to achieve multimodal human-machine interfaces
and predicting bulk properties – such as strength, elasticity, with improved flexibility and durability (Figure 7A). 170
or conductivity – based on composition and processing Furthermore, ML has contributed to the development
conditions. of conformal, wearable carbon-based sensors for health
While both material systems rely on carbon monitoring. 3D-printed graphene-based humidity and
reinforcements, their distinct physical characteristics strain sensors have been designed for human motion
demand tailored ML strategies. Continuous fiber-based prediction, leveraging ML to refine sensor calibration and
composites benefit from spatially aware models capable response accuracy (Figure 7B). 171
of capturing fiber trajectories and interfacial features, In summary, the integration of ML into carbon-based
whereas nano-carbon systems often utilize ML to uncover AM has significantly advanced the fabrication, optimization,
correlations between nano-filler morphology, distribution and application of these materials. By harnessing ML-driven
uniformity, and macroscopic functional properties. This insights, researchers have improved mechanical properties,
distinction highlights the necessity of material-specific ML enhanced structural accuracy, optimized energy storage
pipelines, even within a unified classification of carbon- devices, and enabled next-generation sensor technologies.
based materials in AM.
However, challenges such as real-time ML implementation,
Beyond mechanical optimization, ML has been leveraged multi-material process optimization, and the generalization
to enhance the dimensional accuracy of 3D-printed carbon- of predictive models across different carbon-based materials
based structures. Variability in the curing process, thermal remain key areas for further investigation. Continued
expansion, and shrinkage can lead to deviations from advancements in ML-driven optimization will expand the
intended geometries, particularly in polydimethylsiloxane- capabilities of carbon-based AM, driving innovations in
carbon nanotube (PDMS-CNT) composites. ML-driven high-performance materials for industrial and biomedical
models have been used to predict and correct dimensional applications.
inaccuracies, improving precision in printed components. 165
In energy storage applications, ML-guided optimization 4. Conclusions and future perspectives
has played a crucial role in tuning the architecture and The integration of ML with AM has demonstrated significant
performance of carbon microlattices for supercapacitors. potential in optimizing 3D printing processes, improving

Volume 2 Issue 2 (2025) 45 doi: 10.36922/IJAMD025130010