Numerical modeling of 3D woven composite reinforcements: A review

The literature of numerical modeling of 3D woven composite reinforcements shows that a wide range of impressive studies have been carried out in the last two decades. During this period, two distinct strategies have emerged: the predictive approaches that call for a mechanical construction as well as numerical simulations (e.g., FE method), and the descriptive approaches that are devoted to extracting the geometry of a real textile from micro-tomographic images. In the former methods, different geometrical and mechanical strategies have been employed for mimicking the yarn behavior at either the meso- or sub-mesoscales. And in the latter ones, different approaches ranging from ad hoc image processing pipelines up to more advanced machine learning strategies have been used but only at the mesoscale. This paper aims to highlight the advantages and ideal usecases of each method as well as for each analysis scale (meso- or sub-mesoscale). A common terminology is proposed for organizing and discussing the various meso- and sub-mesoscale strategies. It should be noted that this work only covers the modeling strategies for the textile reinforcement (i.e., dry fabric), thus meso- or macroscale analyses of complete composites are not discussed.