Page 26 - IJAMD-2-2

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

7.1. AI-driven high-throughput material (iii) Data-driven augmentation of physics-based models:
development for AM By learning from simulation outputs and experimental
Traditional material development for AM relies on validation, AI improves the predictive power of
time-consuming experimental trials and computational physical models, correcting deviations and refining
simulations, limiting the speed of new material discovery. underlying assumptions. The synergy between AI and
AI-driven high-throughput material development offers physics-based simulations will enable more accurate
a transformative approach by leveraging automated predictions of AM material behavior, improving
workflows that integrate ML, combinatorial synthesis, and process optimization and material performance
rapid characterization techniques: forecasting.
(i) AI-guided experimental design: ML models 7.3. Integration of digital twins and closed-loop
can predict optimal material compositions and design and manufacturing
processing conditions, reducing the need for
extensive experimental iterations. By incorporating The concept of digital twins – real-time virtual replicas
experimental data in real time, AI refines its of physical AM systems – offers new opportunities for
predictions to enhance material discovery efficiency. AI-driven material design and process control. The
(ii) Automated high-throughput screening: AI has the integration of digital twins with AI enables closed-loop
potential to accelerate the evaluation of new alloys, adaptive manufacturing, where real-time data informs
polymers, and ceramics through high-throughput continuous optimization:
experimentation and combinatorial approaches, (i) Real-time process monitoring and defect prediction:
enabling the rapid assessment of microstructural AI-powered digital twins integrate sensor data, in situ
stability and mechanical properties. monitoring, and ML models to predict and mitigate
(iii) Inverse materials design: The prediction capacity defects such as porosity, cracking, and residual stress
of AI facilitates inverse materials design, allowing in AM-fabricated components.
researchers to specify desired properties and identify (ii) Adaptive process control: AI-driven control systems
compositions that meet these criteria efficiently. This dynamically adjust process parameters (e.g., laser
approach has the potential to significantly shorten the power and scan speed) based on real-time feedback
material development lifecycle. to optimize microstructure formation and mechanical
properties.
7.2. Integration with multiscale high-fidelity (iii) Virtual prototyping and predictive maintenance:
simulation with AI Digital twins enable virtual prototyping of materials
While AI has demonstrated strong predictive capabilities, and components, allowing for iterative design
its full potential in AM material development requires refinements before fabrication. In addition, predictive
integration with multiscale, high-fidelity simulations to maintenance models help extend machine lifespan
improve physical accuracy and interpretability. AI can and reduce production downtime.
enhance and accelerate simulations in the following ways: 7.4. Active learning-driven material composition
(i) Surrogate modeling for computational efficiency: optimization for AM
AI-based surrogate models can approximate high-
fidelity simulations (e.g., FEA, molecular dynamics, One of the major challenges in AI-driven material
and CALPHAD) with significantly reduced development for AM is the scarcity of high-quality
computational cost, enabling rapid exploration of experimental data. Active learning, a branch of ML that
process–composition–microstructure relationships. selects the most informative data points for labeling,
For complex-shaped parts, accelerating simulations provides an efficient solution by minimizing experimental
through AI should be a key research and development efforts while maximizing predictive performance:
focus. A promising approach is to first identify defect- (i) Efficient exploration of composition space: Active
sensitive regions and then strategically adjust process learning strategies guide experimental design by
parameters in these susceptible areas to mitigate focusing on unexplored or high-uncertainty regions
defects. of material composition space, accelerating the
(ii) Multiscale modeling for process optimization: discovery of optimized AM-compatible materials.
AI can bridge different lengths and time scales in (ii) Iterative AI-experimental feedback loops: AI models
AM simulations, integrating macro-scale thermal dynamically update as new experimental data is
simulations with microstructural evolution models to acquired, refining their predictions and continuously
predict final part properties more accurately. improving the efficiency of material optimization.

Volume 2 Issue 2 (2025) 20 doi: 10.36922/IJAMD025100007