Page 26 - ESAM-1-4

Engineering Science in
Additive Manufacturing Machine learning for biomedical metal AM

(ii) Cobalt-chromium alloys possess exceptional wear metallurgical process involving rapid solidification,
resistance and high mechanical strength. Ideal phase transformations, and complex stress evolution.
33
for articular prostheses (orthopedics) and dental The ultimate result, in relation to the performance and
restorations/implant abutments, where wear resistance quality of the biomedical metal, is significantly influenced
and mechanical stability are critical. However, long- by material composition, powder characteristics,
24
term implantation may involve risks of Co/Cr ion process parameters (e.g., laser power, scanning speed,
release and potential toxicity. AM processes require and scanning strategy), equipment condition, and even
25
careful hot cracking control and surface roughness environmental factors. 34,35 These factors exhibit strong
regulation to minimize bacterial adhesion and ion nonlinear interactions, forming a high-dimensional,
release, combining scanning strategy optimization. complex parameter space. Conventional research and
(iii) Medical-grade stainless steels (e.g., 316L) offer cost- production models are predicated on engineers’ experience
effectiveness and good processability. Limitations and extensive trial-and-error experimentation. This not
26
include relatively inferior corrosion resistance, only leads to prolonged development cycles and high costs
potential Ni ion-induced allergies, and high elastic but also hinders the systematic capture and understanding
modulus. Suitable for short-term orthopedic fixators, of underlying patterns. Consequently, this can result
prosthetic sockets, and general medical devices, in limitations on the performance of the product and a
particularly in cost-sensitive, low-load-bearing compromise to batch consistency.
scenarios. AM focuses on preventing solidification
defects by controlling energy density and reducing the 1.2. Introduction to machine learning (ML)
evaporation of elements with a high vapor pressure. 27 ML is a fundamental component of artificial intelligence
(iv) Biodegradable metals represent an emerging frontier (AI) that provides a novel approach to addressing the
in biomaterials research. Magnesium alloys have aforementioned limitations. It possesses the distinct
mechanical properties that match those of human capacity to automatically discern patterns from data,
bone and excellent biocompatibility, but they suffer facilitating precise predictions and decisions. By
33
from excessively fast degradation, which makes them continuously acquiring new knowledge and advanced
prone to premature mechanical failure and hydrogen capabilities, and through self-optimization and
evolution during corrosion. They can be employed updating through specific optimization algorithms
28
in temporary orthopedic fixators, porous bone and other methods, more precise judgment outcomes
scaffolds, and biodegradable cardiovascular stents, can be achieved. 36-38 Against the backdrop of rapid AI
avoiding secondary surgery for implant removal. advancement, ML exhibits increasingly broad applicability.
AM focuses on addressing these issues by employing In biomedical metal AM, its implementation typically
low-energy density to suppress evaporation and follows a systematic workflow comprising four key stages:
balling effects, mitigating challenges associated data acquisition and preprocessing, feature engineering,
with magnesium’s low boiling point and high vapor model selection and training, and model evaluation.
pressure, while rapid solidification refines the With regard to the collection of data, the present corpus
microstructure to regulate degradation rates. Zinc of ML data in AM is primarily derived from the following
29
alloys offer moderate degradation rates and favorable sources: 39,40 Experimental data form the core, encompassing
biocompatibility 30,31 but suffer from inadequate process parameters and mechanical property data obtained
mechanical properties and low as-fabricated density through systematic experiments, real-time molten pool
in AM processes. To enhance performance, AM dynamics captured via in situ monitoring techniques
strategies focus on optimizing scanning strategies (e.g., thermal imaging and acoustic emission [AE]), and
to reduce elemental segregation, thereby improving microstructural/defect data derived from microscopic
mechanical integrity. Iron-based biodegradable characterization (e.g., scanning electron microscope [SEM]
alloys possess high strength and biocompatibility. and micro-computed tomography [micro-CT]). However,
32
However, their degradation is excessively slow, acquiring experimental data is costly and time-consuming.
hindering complete absorption within the desired Second, simulation data, particularly that generated from
timeframe. AM enables the fabrication of tailored multi-physics models such as finite element analysis, have
porous structures to accelerate corrosion, making been demonstrated to effectively supplement experimental
these alloys suitable for load-bearing bone scaffolds data. Databases formed by integrating public datasets
and orthopedic fixators. and academic literature, along with industrial production
However, the actual AM process of biomedical data accumulated by manufacturers, also provide valuable
metals constitutes a thermo-fluid-solid coupled physical resources for model development. Nevertheless, the latter is

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