Composites are materials that combine two or more materials to achieve new material properties to overcome the limitations of single materials. It can be easily found at almost all kinds of industry, bio-materials and research for advanced materials. In this paper, I conducted a study on particle-reinforced composites, biomimetic composites, and random composites that represent the trend of composite re-search. The particle-reinforced composite material has been studied since the 1980s. It has been studied in a mathematical way based on micro-mechanical theory. In this study, a new homogenization method of hyperelastic-viscoelastic materials has been developed. I applied to ABAQUS USERMATERIAL, and it was verified by comparison with the representative volume element model. The staggered platelet composite, which is a kind of bio-inspired composite, has been actively studied since the 2000s. However, although toughness is the biggest advantage of this composite, it has been difficult to analyze. so it was difficult to interpret the toughness. In this study, the effective elastic modulus prediction and failure mode were predicted by analytical method, and the composite toughness was optimized based on simu-lation and experiment using 3D printer through Gaussian process and Markov Chain Monte Carlo, one of the data driven techniques. The random composite material, which is defined by randomly placing a hard material and a soft material in the form of a checker board, has no restrictions on the shape or orientation of the composite microstructure. In this study, prediction and optimization of nonlinear property values were performed using Convolution Neural Network (CNN) using simulation data until the material was completely destroyed.
본 논문에서는 복합재 연구 동향의 흐름을 대표하는 입자강화복합재, 생체모사 복합재, 무작위 복합재에 대한 연구를 수행하였다. 입자강화복합재 중 그동안 없었던 초탄성-점탄성 재료의 균질화법을 새로 개발하여 유한요소해석 상용 코드에 적용하고, 대표 체적 요소와 비교 검증하였다. 생체모사 복합재의 일종인 엇물린판상 적층 복합재는 본 연구에선 유효 탄성 계수와 파괴 모드를 예측해 보고, 데이터 중심적 기법 중 하나인 크리깅 기법을 통해 시뮬레이션 및 3차원 프린터를 활용한 실험 기반으로 복합재 인성을 최적화하였다.격자판 형태에 무작위로 경질 및 연질 재료를 배치함으로써 정의되는 무작위 복합재는 복합재 미시 구조의형태나 방위에 대한 제약이 없다는 특징이 있다. 본 연구에선 균열 상장 모델을 바탕으로 재료가 파괴될 때까지의 시뮬레이션 데이터를 확보하였고, 이를 통해 비선형 물성치에 대한 예측 및 최적화를 인공 신경망을활용하여 수행하였다.
