Page 50 - IJAMD-2-2

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

One of the key applications of ML in ceramic AM is steps such as sintering. Supervised learning models have
the prediction of viscoelastic and printability properties been trained to correlate process parameters and binder
of ceramic pastes, particularly in binder-free DIW content with sintering outcomes–including shrinkage,
systems. Since the rheological behavior of ceramic pastes densification, and grain coarsening–enabling optimization
strongly influences their printability and structural of thermal profiles to minimize residual stress and improve
integrity, researchers have utilized ML models to correlate final part integrity. Other studies have applied ML
159
formulation parameters with viscoelastic properties. techniques to tailor ceramic microstructures by predicting
Recent studies have demonstrated that ML-driven grain growth and porosity as functions of printing
predictive models can effectively optimize ceramic ink conditions and sintering temperature profiles. However,
160
compositions, ensuring improved extrusion stability and modeling ceramic microstructures remains particularly
defect-free printing. 154 challenging due to the complex multiscale phenomena
Beyond material formulation, ML has played a crucial involved, including powder packing, binder removal, and
157
role in enhancing defect detection and surface quality thermally driven phase transitions. Progress in ML for
evaluation in 3D-printed ceramic components. Due to ceramic AM is further constrained by the scarcity of high-
the inherent brittleness of ceramics, even minor surface quality datasets, as experimental studies are limited and
154
imperfections can lead to mechanical failure, making involve high variability across materials and processes.
precise defect detection essential. Deep learning-based Moreover, the brittle nature of ceramics demands
image processing techniques have been employed to extremely high prediction accuracy since minor defects
160
identify surface defects in ceramic AM, mitigating or heterogeneities can lead to catastrophic failure. These
interference from background noise and improving factors highlight the need for specialized ML strategies and
detection accuracy. Furthermore, low-contrast defects, interdisciplinary domain knowledge tailored to the unique
155
particularly in curved ceramic surfaces, pose significant challenges of ceramic AM. 161
challenges for conventional inspection methods. Deep Overall, the integration of ML into ceramic AM has
learning models trained on high-resolution imaging significantly enhanced material formulation, process
datasets have demonstrated superior performance in optimization, defect detection, and quality control.
identifying and classifying such defects, improving the By leveraging advanced ML models, researchers and
reliability of ceramic AM processes. 156 manufacturers can overcome the limitations of conventional
Another major challenge in ceramic AM is process ceramic printing techniques, paving the way for improved
stability and consistency across deposition lines in mechanical performance, reduced production costs, and
extrusion-based printing techniques. Variability in increased industrial adoption. However, challenges such
extrusion pressure, layer adhesion, and drying-induced as data scarcity, model generalization across different
deformations can compromise final part quality. To address ceramic materials, and real-time implementation of ML
these issues, ML-based quality optimization frameworks algorithms in AM workflows remain key areas for future
have been developed to analyze real-time process data and research. Continued advancements in ML-driven ceramic
adjust deposition parameters dynamically. These models AM optimization will further strengthen the role of AI in
enable adaptive control of layer deposition, reducing print next-generation AM.
failures and enhancing mechanical uniformity in ceramic
structures. 157 3.4. Carbon-based materials
In addition to process control, ML has facilitated the Carbon-based materials have garnered significant
material development and optimization of lithography- interest in AM due to their exceptional mechanical,
based ceramic AM, particularly for porous alumina electrical, and thermal properties. Carbon fiber-reinforced
ceramics. The microstructure of porous ceramics is polymers, carbon nanotube composites, and graphene-
highly dependent on photopolymerization kinetics, based materials exhibit high strength-to-weight ratios,
resin composition, and post-processing conditions. By electrical conductivity, and flexibility, making them ideal
leveraging ML algorithms to predict porosity, shrinkage for applications in aerospace, energy storage, structural
behavior, and mechanical performance, researchers reinforcements, and wearable electronics.
have improved the material design process for high- One of the most critical applications of ML in carbon AM
performance ceramic components. 158 is the optimization of mechanical properties in continuous
Beyond material design and process monitoring, another carbon fiber-reinforced composites. The mechanical behavior
critical application of ML in ceramic AM is the prediction of these materials is influenced by fiber alignment, resin
of microstructural evolution during post-processing infiltration, and interfacial bonding. Deep learning models
162

Volume 2 Issue 2 (2025) 44 doi: 10.36922/IJAMD025130010