Page 45 - ESAM-1-4

Engineering Science in
Additive Manufacturing Machine learning for biomedical metal AM

5.3. Outlooks In summary, the development of an integrated intelligent

Although ML has achieved significant results in individual technology system represents an inevitable trend. In the
stages, the future development of biomedical metals future, driven by clinical needs, the integration of structural
AM inevitably requires breaking down barriers between design, intelligent manufacturing, and performance
these stages to build a comprehensive, multi-functional regulation within a unified framework will be achieved
intelligent technology chain spanning front-end design to to achieve full-process closed-loop optimization. The
156
back-end manufacturing. Future research will focus on integration of design, prediction, optimization, and control
the following key directions: into a unified framework has the potential to enhance the
(i) Overcoming data constraints: Future efforts should capabilities of AM in high-end medical applications. This
concentrate on the development of small-sample approach will deliver efficient, cost-effective, and highly
learning, zero-shot learning, and meta-learning/ reliable personalized medical solutions for patients.
transfer learning frameworks across materials and Acknowledgments
devices. The transfer of knowledge from data-rich
domains to data-scarce domains is a significant aspect None.
of this approach, as it reduces reliance on the volume
of data in the target domain while enhancing the Funding
model’s generalization capabilities. The authors acknowledge the financial supports from
(ii) Enhancing model credibility: The incorporation of the National Key Research and Development Program
physical laws as soft or hard constraints within models, of China (Grant No. 2024YFE0109000), the National
for instance through the construction of physics- Natural Science Foundation of China (Grant Nos.
informed neural networks, ensures that predictions 52274387, 52311530772), the Medical-Engineering Cross
are aligned with physical principles. Concurrently, the Foundation of Shanghai Jiao Tong University (Grant No.
widespread implementation of interpretability tools YG2024LC04), and the Fundamental Research Funds for
such as SHAP and local interpretable model-agnostic the Central Universities (Grant No. YG2023QNA21).
explanations (LIME) serves to transform opaque
systems into comprehensible ones. Conflict of interest
(iii) Prospective smart alloy design and manufacturability The authors declare that they have no competing interests.
prediction: The starting point for the future should
be further advanced to the material design itself. Author contributions
Generative models and active learning should be
leveraged to reverse-engineer novel alloys that Conceptualization: Yi Mao, Liqiang Wang, Uglov Vladimir,
simultaneously exhibit ideal biological functionality Zhou Jing
and superior printability, 157-159 establishing a Visualization: Yi Mao, Deyu Jiang
quadruple-loop design paradigm of composition- Writing–original draft: Yi Mao, Deyu Jiang
structure-property-manufacturability to achieve Writing–review & editing: All authors
synergistic design of materials and processes from the
outset. Ethics approval and consent to participate
(iv) Establishing digital archives for AM process: AM is Not applicable.
a process with strong temporal dependencies, where
the quality of each layer is cumulatively influenced by Consent for publication
the thermal history and physical state of preceding Not applicable.
layers. By preserving layer-by-layer data throughout
the manufacturing process for each component, the Availability of data
construction of digital archives holds immeasurable
value for product performance traceability and data- Not applicable.
driven certification systems. References
(v) Further combination of digital twins: Enhance real-
time interaction and online decision-making between 1. Cui Y, Wang L, Zhang L. Towards load-bearing biomedical
digital twins and physical production lines to drive titanium-based alloys: From essential requirements to future
adaptive adjustments to process parameters, thereby developments. Prog Mater Sci. 2024;144:101277.
achieving precise closed-loop control. 160 doi: 10.1016/j.pmatsci.2024.101277

Volume 1 Issue 4 (2025) 23 doi: 10.36922/ESAM025440031