Page 26 - IJAMD-2-3

International Journal of AI for
Materials and Design AI applications in composite materials

A

B C

Figure 11. Manufacturing methods and applications utilizing automation models, and research utilizing automation models in the field of composites.
(A) Manufacturing methods and applications utilizing automation models; (B) Defect detection results using automation models in the automated fiber
141
placement process. Reprinted with permission from Tang et al. Copyright © 2024 John Wiley and Sons; (C) Study on predicting mold filling patterns
141
in the resin transfer molding process. 142
Abbreviations: RNN: Recurrent neural network; LSTM: Long short-term memory.
deviations in the properties of composite materials. In CNNs to develop an automatic system for void content
particular, the performance of composite materials can assessment in composite laminates. Voids were detected
be highly sensitive to external environments or specific with high accuracy, and the approach delivered significant
conditions, making it crucial to analyze and manage these improvements over traditional optical microscopy
factors precisely. Therefore, the automation of composite methods by offering faster and more reliable results, even
material analysis using automation models for quality in challenging lighting conditions or with small void sizes.
control is becoming increasingly important. 141-145 Furthermore, Tang et al. applied a CNN-based DL
141
In quality control applications, CNN-based DL model to the automated fiber placement (AFP) process,
models are extensively employed due to their optimized successfully detecting lay-up defects, such as bridging,
characteristics for processing image and spatial data wrinkles, and puckers, using 3D scanning data. The model
through the convolution and pooling of nonlinear high- demonstrated high accuracy, particularly for defects with
dimensional feature representations. Machado et al. used significant height differences compared to the normal
143

Volume 2 Issue 3 (2025) 20 doi: 10.36922/IJAMD025210016