A multiscale simulation method incorporating fiber distribution characteristics for off-axis tensile analysis of 2D woven SiCf/SiC

• A new multiscale modeling method of 2D CMCs was proposed. • Latin hypercube sampling was used to determine the micro-RVE size. • The damage mechanisms of unidirectional CMCs under four loads were studied. • A coupled damage constitutive model for 2D CMCs was proposed. Ceramic matrix composites (CMCs) often experience complex loading conditions in practical applications, making it essential to study their mechanical behavior under off-axis tensile loads. This paper develops a multiscale modeling method to explore the nonlinear tensile behavior of 2D woven CMCs, considering the random distribution characteristics of fibers and pores within the CMCs. The fundamental mechanical properties of the fibers, interface, and matrix were determined through nanoindentation and fiber push-out experiments. The model accounted for the spatial randomness of fibers, employing Latin hypercube sampling on 87 SEM images to derive the smallest possible microscale representative volume element (RVE). A Weibull distribution was used to represent the pore in the matrix, and the effect of mesh density was investigated. CT scanning of the 2D woven CMC provided structural details of the composite, leading to the creation of a mesoscale RVE that includes fiber bundles, matrix, and structural pores. The nonlinear tensile behavior of the mesoscale RVE was studied, integrating the homogenized mechanical responses from the microscale RVE. The findings reveal that the matrix modulus is 1.8 times that of the fibers, the interface shear strength is approximately 16.1 ± 2.4 MPa, and the mode I energy release rate is about 1.9 ± 0.7 J/m 2 . The microscale RVE effectively captures the random distribution characteristics when its edge length is approximately 7.2 times the average fiber diameter. The elastic modulus of the mesoscale RVE shows less than 8 % deviation from the off-axis tensile test results at four different off-axis angles, and the error in predicting fracture strength remains within 20 %.