An automated finite element program for micromechanics modeling of random-wavy fiber composites.

摘要

Micromechanics modeling provides a powerful tool for the efficient and accurate analysis of composite materials. Many micromechanics models for unidirectional composites assume idealized microstructures such as periodic arrangements and straight fibers. Real composite microstructures exhibit defects such as randomness in cross-sectional fiber arrangement and longitudinal waviness.;An automated finite element micromechanics modeling program was created to simulate random arrangement and longitudinal waviness in fibers of unidirectional composites. Emphasis was placed on the ease of use, adaptability and computational efficiency of the program for general use. A numerical experiment was conducted to validate the model and to establish the effect of user input parameters on accuracy of results. Recommended values of selected user input values are presented.;The program was used to study domain size considerations for representative volume elements of infinite random fiber domains. Statistical results for the variation of composite elastic constants due to randomness in fibers exhibiting no waviness are presented.;The program was then used to study the composite elastic behavior when both randomness and waviness in fibers are present. The computational expense of modeling a representative volume element of a random fiber domain with fiber waviness is addressed. Statistical variation of composite elastic constants for wavy-random fiber models are compared with straight-random fiber models. These results are also compared with periodic models, both with and without fiber waviness.;Both randomness and waviness have a significant influence on the composite elastic response. A random domain of uniformly sized fibers should contain 50 to 150 fibers to be nearly a representative volume element without excessive computational expense. Together, randomness and waviness exhibit a coupled effect on the composite elastic properties, hence the two effects should not be considered independent. In some cases, the computational expense of running large models might be reduced by instead averaging the results from numerous smaller models, with minimal compromise in accuracy.