Cheng Qiu, Jinglei Yang

1 Institute of Mechanics, Chinese Academy of Sciences, China

2 Hong Kong University of Science and Technology, Hong Kong SAR, China 

EXTENDED ABSTRACT: Advanced composite materials have gained prominence in the field of aerospace engineering, and their use is also one of the key indications of aircraft improvement. With the ever-changing development needs of engineering applications, structural load-bearing/multi-functional integration, low-cost, and design/characterization integration technologies of composite materials are important trends. The composite effect, the multi-scale feature of material­structure integration, and the anisotropic properties not only provide great performance designability, but also bring great challenges to all aspects of structural design, manufacturing, and characterization. Although the mechanical theory system suitable for composite materialengineering applications is relativelymature, the traditional analysis anddesign methods based on numericalsimulation and experimentalverification have low computationalefficiency when facing high-dimensional problems, which affectsthe iterative design of structuresAnd with the continuous emergenceof new composite materials, thetheoretical system and simulationmethods based on the deductionmodel are still unclear. As a newscientific concept based on induction.data-driven analysis methods are having a transformative impact on many aspects of composite materials engineering dueto their unigue logical and algorithmic characteristics. By combining the phvsical knowledge of composite materials witithe model framework, loss function, optimization iteration process, etc. of deep learning, we can obtain a model with betteigeneralization performance under small sample conditions and solve problems related to composite materials engineering. Inthis report, we will introduce the curent issues in the engineering application of composite materials, and briefly review theresearch progress of data-driven methods in the multi-functional integrated design, low-cost manufacturing, and analyticalcharacterization of composite materials. Some of the speaker's work is included: data-driven composite material design andanalysis methods integrating physical knowledge, which is used for the design of typical engineering structures of compositematerials: Uncertainty analysis of composite materials based on the microstructure-property quantitative relationship: And thematerial damage parameter mining to optimize the design of composite materials using a building block approach methodata-driven methods.combined with theoretica or numerical simulation data. have become a mature paradigm for academi(research on composite materials. Further exploring and expanding the application of machine learning methods in structuralengineering is currently the focus and will surely be a new research growth point.

Keywords: Data-driven: Fiber composites: Structural design