Page 39 - IJAMD-2-2

International Journal of AI for
Materials and Design ML-driven optimization in additive manufacturing

Table 1. Summary of ML models for polymer‑based 3D printing
Motivation Materials AM method ML model Model inputs Remarks References
Process ABS, PLA FDM SVM Image data from Real-time defect detection 84
optimization semi-finished printed using image processing and
parts at checkpoints ML
ABS, PLA FDM CNN Top-view images from Detecting infill pattern in 85
a static camera during real-time
printing
PLA FDM CGAN, DANN Top-view grayscale Fault diagnosis under 81
images of printed layers process parameter drift
PLA FDM CNN (ResNet-50) Real-time images from a Autonomous in situ 87
nozzle-mounted camera correction of under- and
over- extrusion
PLA FDM YOLOv3-Tiny, Real-time nozzle-near Automated defect detection 54
YOLOv4-Tiny, images and G-code correction for
ONNX-optimized in situ extrusion
YOLO compensation
ABS FDM Reinforcement Printing speed, flow rate Online-learning based defect 57
learning multiplier, cooling fan, mitigation
surface quality images
PLA FDM XceptionTime Temperature, humidity, Real-time classification of 88
air pressure, gas particle FDM process states using
concentration environmental sensor data
GelMA and alginate DIW CNN (ResNeXt-50) Layer images, interlayer Real-time anomaly detection 90
(bioprinting) continuity, uniformity in bioprinting
metrics
16 biomaterials DIW DT, RF, ANN Bioink composition Predicting the printability of 91
(bioprinting) ratios, printability labels bioink formulations
GelMA with DLP U-Net-based Light scattering Generating optimized 97
encapsulated cells (bioprinting) master-slavee neural patterns, corrected correction masks to mitigate
network exposure masks cell-induced light scattering
Epoxy acrylate DLP RNN with LSTM Per-pixel grayscale Predicting deformation 55
resin+commercial (grayscale) layers, EA values, deformed and optimizing grayscale
photopolymer resin structure data distribution for enhanced
print accuracy
Acrylate DLP LSTM Temperature data, UV Optimizing DLP printing 99
photopolymer resin exposure time, layer via real-time temperature
thickness prediction
Three commercial DLP U-Net, CGAN Grayscale pixel data, Improving printing 100
photopolymer resins boundary images resolution and reducing
jagged edges
Commercial DLP RF with EWMA Strain gauge data, UV Real-time detection of part 104
photopolymer resin p-control chart exposure levels detachment and automatic
process halting
Commercial DLP MLP Prediction feature Predicting optimal idle time 107
photopolymer resin region, layer geometry for resin drainage
Photopolymer resin SLA CNN, Two-stream FEA-generated stress Predicting layer-wise stress 56
CNN distributions, geometry distribution to improve print
data reliability
Commercial DLP RF UV exposure time, light Predicting print accuracy 106
photopolymer resin intensity, layer thickness and optimizing printing
parameters
Property PLA FDM RF, SVM, K-NN Layer height, printing Optimizing mechanical 108
optimization speed, printing properties by predicting
temperature tensile strength
(Cont’d...)
Volume 2 Issue 2 (2025) 33 doi: 10.36922/IJAMD025130010