Generation of two‐level representative volume element model for uncertainty analysis of composite materials

Abstract This work presents a new two‐level representative volume element model for generating random fiber distributions in continuous fiber composite materials. A representative volume element (RVE) model with random geometric boundaries is generated, and the adaptive connections between neighboring RVEs are realized to eliminate unreasonable connections that may occur at the boundaries of RVEs. Subsequently, a novel adaptive generation and construction algorithm for larger RVE (LRVE) is proposed to consider more micro‐scale geometric and material uncertainties. A computational framework is established to implement the adaptive construction of LRVE and determine the statistical uncertainties of continuous fiber composite materials. The uncertain statistical results of the proposed algorithm are compared with those of the random sequential expansion algorithm and the complete spatial random pattern, showing good agreements among them. Finally, the uncertainty of macro‐scale linear elastic anisotropic material parameters is evaluated by using the integrated software framework. The maximum relative error of the predicted macro‐scale elastic properties is only 8.70%, which outperforms other numerical analysis results in literatures. The proposed algorithm is helpful for further research on multi‐scale linear elastic uncertainty analysis and reliability design of continuous fiber composite materials.