Machine Learning Applications in Composites: Manufacturing, Design, and Characterization
Cheng Qiu, Jinglei Yang
Abstract
Composite material, which refers to a combination of two or more components, is generally delicately designed to take the full advantages of its constituent materials having notably distinct chemical or physical properties. With the complex multi-scale architecture and the multiple target properties brought by the tailored combination, enormous possibilities and challenges are being faced in the engineering applications of composites. Machine learning (ML), as a data-driven approach, is emerging as not only an efficient surrogate model for conventional experimental or simulation methods but also a revolutionary way to dig out material-related information from various kinds of data and to serve for the goal of discovering the next generation of the composite with unprecedented properties. In the most recent five years, with the rapid advances in high-performance computing techniques and contributions from the open-source community for easy-to-implement ML libraries, a booming growth rate is found in the literature about the research on machine learning in the composite field. In this chapter, an overview of how advanced machine learning algorithms can be applied in every aspect of composite science is presented. Special emphasis is given to the manufacturing, design, and property characterization of fiber-reinforced plastic (FRP). The advantages of the ML applications over the conventional methods are discussed and our perspectives are given on the future perspectives of further breakthroughs in ML on composite materials.