This paper investigates the modeling capabilities of computational homogenization for the mechanical behavior of short fiber reinforced polyamide. Simulations on a representative volume element (RVE) with elastic fibers and an elastic–plastic matrix are compared to tensile experiments on specimens taken in parallel and transversal direction from injection molded plates. In view of the semi-crystalline polymer structure, focus is put on identifying the matrix parameters through two alternative methods:First, the matrix parameters are identified directly using tensile experiments on a non-nucleated and a nucleated unreinforced polyamide. In the RVE computations based on the non-nucleated grade, the composite stress–strain behavior is somewhat underestimated, and with the nucleated grade, the behavior is slightly overestimated. To explain this, the semi-crystalline polymer structure is studied. Polarized light microscopy images reveal that the non-nucleated grade has a coarser and the nucleated grade a finer spherulite structure, compared to the matrix present in the composite. However, the degree of crystallinity measured by differential scanning calorimetry is in a similar range.As an alternative, the matrix parameters are identified inversely by fitting the RVE model to the composite tensile experiments. As uncertainties with respect to the matrix material as well as possibly remaining simplifications in the micromechanical model are compensated for, this reverse engineering approach allows for a very good fit.
Read full abstract