Page 11 - IJB-6-1
P. 11
Yu and Jiang
3 Perspective on using machine learning in can be further explored using machine learning.
bioprinting Figure 3 shows an example case of using neural
networks to improve the bioprinting process. The
Machine learning has been integrated into 3D variables are the inputs influencing the objective
printing processes in many ways to improve results (e.g., cell damage, cost, and time). In the
applications, including process optimization, case here, voltage, gas, nozzle size, pressure, etc.,
dimensional accuracy analysis, manufacturing can be fed into the neural network for training.
defect detection, and material property prediction. Corresponding outputs (cell damage, cost, time,
However, machine learning has not been applied in etc.) need to be provided to tune machine learning
3D bioprinting yet. In this section, the perspective parameters. Once the algorithm is done, new input
on how machine learning can help to improve 3D data can be used for performance evaluation.
bioprinting will be illustrated.
3.2 Manufacturing defect detection
3.1 Process optimization
In the traditional 3D printing process, Scime
In traditional 3D printing processes, Aoyagi et al. and Beuth used computer vision techniques
[17]
[26]
proposed a method to construct a process map and unsupervised machine learning to identify
for 3D printing using a support vector machine. in situ melt pool signatures indicative of flaw
This method can predict a process condition that formation in a laser powder bed fusion process.
is effective for manufacturing a product with low Caggiano et al. developed a machine learning
[27]
pore density. Menon et al. used hierarchical method to timely recognize metal material defects
[18]
machine learning to simultaneously optimize in Selective Laser Melting processes. Images
material, process variables, and formulate 3D obtained from the layer-by-layer manufacturing
printing of silicone elastomer through freeform process are analyzed through a bi-stream deep
reversible embedding. He et al. investigated CNN for identifying defects. Zhang et al.
[19]
[28]
using different machine learning techniques described a CNN strategy for monitoring porosity
for modeling and predicting the proper printing in laser additive manufacturing (AM) processes.
speed in a vat photopolymerization process The melt-pool data were gained through a high-
(Continuous Liquid Interface Production). In their speed digital camera for in-process sensing. Then,
study, siamese network model has the highest the data were analyzed by their developed neural
accuracy. In a previous study, the convolutional network.
neural network (CNN) was applied to enable In 3D bioprinting, similarly, machine learning
the angular re-orientation of a projector within a can be used to detect defects such as wrong
fringe projection system in real-time without re-
calibrating the system . In addition, a conceptual
[20]
framework on combining mathematical modeling
and machine learning to evaluate and optimize
parameters in Powder Bed Fusion processes was
proposed by Baturynska et al. [21]
In 3D bioprinting, similarly, machine learning
can be used for improving the fabrication
process, such as predicting process conditions
and optimizing process parameters. Taking
extrusion-based bioprinting as an example, it
is now able to stably fabricate organoids using
low-concentration gelatin-methacryloyl with the
help of electrostatic attraction . However, what Figure 3. Example neural network for process
[37]
are the best values of these parameters? This still optimization in three-dimensional bioprinting.
International Journal of Bioprinting (2020)–Volume 6, Issue 1 7
