基于数据驱动的纤维增强复合材料高效多尺度损伤分析方法

In order to realize the damage analysis of fiber reinforced composites， an efficient multiscale damage analysis method is developed. Firstly， based on the generalized method of cells， a multiscale damage analysis framework is constructed for laminate and plain weave composites， and the damage processes at the mesoscale and microscale under uniaxial tension are analyzed， The results show that the complex weave structure of plain weave composites leads to a more complex mesoscale and microscale damage evolution process， which is significantly different from the damage process of laminates. Based on this， neural networks are introduced， and a data-driven multiscale damage analysis strategy is proposed to realize the efficient damage simulation of plain weave composites. Compared with the experimental and simulant results， the error of predicting tensile strength by the efficient multiscale damage analysis method is less than 7%; and compared with traditional multiscale damage analysis method，the efficiency of macroscale calculations can be improved by about 12.47 times.