Page 8 - IJAMD-2-3

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

lightweight characteristics and outstanding mechanical microstructure design, novel material discovery, and
performance of composite materials make them essential data augmentation 22
in various industries, including aerospace, automotive, (iii) Automation: ML is applied to optimize composite
energy harvesting, robotics, and construction, as shown manufacturing processes, automate defect detection
in Figure 1. 13-19 The properties of composite materials are and quality control, and analyze real-time data to
influenced by multiple factors, including the types of fiber enhance manufacturing efficiency. 23
and resin, as well as the manufacturing process. In addition, AI thus serves as a powerful tool for addressing various
their inherent anisotropy, nonlinearity, and complex challenges in the field of composite materials, significantly
microstructure make design and analysis significantly
more challenging than conventional materials. enhancing research and design efficiency. However, the use
of insufficient or unreliable data, as well as the application
Traditionally, the field of composite materials has of unsuitable ML techniques, can lead to suboptimal results
relied heavily on the expertise of skilled professionals for or computational inefficiencies.
conducting experiments and developing physics-based
models. In addition, simplifying assumptions have often Therefore, a structured and systematic understanding
been used to simulate the complex behavior of composite of ML applications in the field of composite materials
materials. However, these conventional approaches have is essential. To address this need, this review provides a
clear limitations in fully capturing the intricate and diverse comprehensive and systematic analysis of ML-driven research
characteristics of composite materials. As a result, machine on composite materials. The primary contributions of this
learning (ML) and deep learning (DL) methodologies paper are threefold. First, we propose a clear and structured
have recently received significant attention in the field of classification of ML applications for composite materials
composite materials. ML is increasingly being explored into three categories: predictive, generative, and automation
to overcome the challenges associated with traditional models. Second, we provide an in-depth review of the state-
methods, offering new possibilities for improving efficiency of-the-art ML techniques within each category, focusing on
and accuracy in the design, analysis, and manufacturing of their practical applications in composite material design,
composite materials. manufacturing, and analysis. Third, we critically discuss the
current limitations and challenges of these ML approaches
The applications of ML in the field of composite
materials, as illustrated in Figure 2, can be broadly and propose future research directions to address them.
categorized into three aspects: The remainder of this paper is organized as follows:
(i) Prediction: ML enables the analysis of large-scale sections 2, 3, and 4 provide detailed discussions on the
datasets to uncover complex patterns in areas such applications of predictive, generative, and automation
as mechanical property and behavior prediction, and models, respectively, in the field of composite materials.
design parameter estimations 20,21 Finally, Section 5 presents the key findings, highlights the
(ii) Generation: ML leverages data-driven learning current limitations, and outlines future research directions
to construct new information, such as for for advancing AI-driven developments in this field.

Figure 1. Components made of composite materials and the proportions (by weight) of different materials used in the aerospace, automobile, and wind
turbine industries 17-19

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