Page 112 - IJB-5-1

Electrohydrodynamic printing process monitoring by microscopic image identification
A B C

Figure 4. Effect of the applied voltage on the deposited fiber patterns (65 wt/v% polycaprolactone, stage speed = 150 mm/s, D = 3 mm).
(A) 2.6–3 kV. (B) 3.2 kV. (C) 3.4 kV

A B C D E F

Figure 5. Electrohydrodynamic printing jet under varied stage speed (65 wt/v% polycaprolactone, V = 3 kV) (A) 50 mm/s; (B)100 mm/s;
(C) 150 mm/s; (D) 200 mm/s; (E) 250 mm/s; (F) 300 mm/s

is essential to decide which features should be used as the precipitation, variation of process parameters, and
inputs of the learning algorithms . environmental factors.
[14]
In the past decade, CNN is making major advances by Under the same fabrication process parameters, the
allowing a learning algorithm to be fed with natural cone shape would shift from standard to meniscus if
data and to automatically discover the representations the PCL ink concentration increases from 70 wt/v% to
needed for detection or classification . CNN is widely 80 wt/v% and above. The larger cone in the meniscus
[15]
used in image classifications and considered as a mode is due to higher surface tension caused by higher
dominant approach for many recognition and detection PCL concentration. For 70 wt/v% PCL solution, the cone
applications . A CNN model gets its output by shape can evolve from meniscus, standard, and tiny to
[16]
conducting convolutional on an input image data and a multijet when the applied voltage increases from 2 to
wt/v, updating the wt/v iteratively by back propagation, 5 kV while keeping other EHDP process parameters
and finally ending up with an optimal network for constant. Besides, the cone shape would be tiny or huge
classification applications [17,18] . when the solution FR is low (FR <0.5 μL/min) or high
In this study, CNN is used to classify the microscopic (FR >1.2 μL/min). Moreover, the cone would break when
images into eight categories as shown in Table 2, including the solution feeding rate is too low or the SS is too fast
a standard cone mode (stable cone-jet mode) and seven (SS >250 mm/s). Environmental parameters such as
other categories. The EHDP cone categories are usually temperature and humidity have distinct effects on the
characterized by the geometrical form of the droplets at cone modes. The cone jet may become dry or continuous
the nozzle tip, the breakup mechanisms, and the type discharge when the relative humidity is <50%, since the
of instabilities . Majority of researchers use the stable lower humidity speedup the evaporation rate of PCL/zein
[19]
[20]
cone-jet mode for biopolymer scaffold fabrication. Such composite solutions .
mode is known to exist only for a limited window, where 4.1 Applying CNN for EHDP Cone Modes’
the applied voltage is just above the lowest voltage for
stable printing and the FR is above a minimum threshold Classification
necessary to sustain a steady jet. The proposed CNN model can perform feature extraction
It is quite hard to describe the physical conditions for and classification in a unified framework. Compared with
each mode using EHDP process parameters, since they traditional machine learning methods (such as support
are significantly affected by the properties of biopolymer vector machines), the proposed CNN is a better option
solutions. The causes for spontaneous mode switch for the mentioned application.
mainly come from the change of the electric field strength We have tried different CNN layer structures in the
in fiber layer stacking process, solution impurities or training and explored a tradeoff between the training

6 International Journal of Bioprinting (2019)–Volume 5, Issue 1