Page 12 - IJAMD-2-3
P. 12
International Journal of AI for
Materials and Design AI applications in composite materials
A
B C
D
Figure 4. CNN architecture and research utilizing CNN in the field of composite materials. (A) Structure of CNN; (B) Image data and data classification
process for predicting stacking angle based on cross-sectional images; (C) CNN architecture and circuit analogy configuration used for estimating the
45
principal permeability values of the considered 3D fibrous structures; (D) Ground truth of defects for each layer of the composite laminate and the
46
predicted results. 47
Abbreviations: CNN: Convolutional neural network; FEM: Finite element method.
To address these challenges, data augmentation even with a limited dataset. In this process, each image was
techniques, such as rotation, distortion, and noise addition, rotated by 90°, 180°, and 270°, and then horizontally flipped
can be used to enhance the training dataset. 53,54 For to generate eight images per single specimen. Generative
example, Shim et al. applied CNN and data augmentation models, discussed in the next section, can also be used
55
to predict the stress-strain curves of composites fabricated for data augmentation to obtain more diverse images and
through mechanical recycling, enabling effective learning further enhance training efficiency. 56-58
Volume 2 Issue 3 (2025) 6 doi: 10.36922/IJAMD025210016
