Image Based Statistical Analysis of Composite Microstructure for RVE Simulations

Abstract Micromechanical simulation of fiber reinforced unidirectional composites typically relies on representative volume elements comprising fibers and matrix. In order to optimize the validity of the predictions at minimum computational effort, it is important to have an understanding on the required level of detail of the micromechanical simulation. Whilst it is clear, that for predicting the longitudinal properties of composite materials based on micromechanical simulation, it is sufficient to model a single fiber and the surrounding matrix with the appropriate fiber volume fraction and matrix volume fraction, for transverse properties, this is not sufficient. The distribution of fibers and their closest distance are relevant for the failure of composite materials under transverse loading. In order to assess the required number of fibers in the micromechanical simulation, this paper proposes image based statistical analysis of composite microstructures. A micrograph of a carbon fiber/ epoxy composite was digitized and subsets of variable size were analyzed and compared to the statistical fiber volume fraction and fiber distribution of the total micrograph. it was found that a subset including 36 fibers has the same statistical characteristics in terms of fiber distance and fiber orientation as a subset including 25,000 fibers. It is therefore concluded that a minimum of 36 statistically distributed fibers should be included in a representative volume element of a composite microstructure if transverse mechanical properties are to be predicted.