Taking the T300 carbon fiber/epoxy resin matrix unidirectional composites as an example

an equivalent thermal conductivity prediction method based on microscopic image recognition was established

in consideration of the influence of the interface defect. The microstructure scanning electron microscope photos were processed first through image recognition. Then according to the fiber volume fraction stability criterion

the representative volume element(RVE) were built by geometry reconstruction technique. Meanwhile

the thermal contact resistance was added at the interfaces in the RVEs to represent the influence of the interface defects. Finally

the prediction results of equivalent thermal conductivity (ETC) were obtained by the finite element method. Results show that the randomness of the interface defect's position affects ETC weakly

while ETC decreases significantly with the increasing proportion or the thickness of the interface defect. When the the interface defect proportion is greater than 0.8 and the dimensionless thickness of the interface defect is less than 0.15

the most significant influence on ETC appears. Compared with the simulations of ignoring the interface defect between fiber and matrix

the biggest drop of ETC can reach 52.1% considering the interface defect.