GLS strength prediction of glass-fiber-reinforced polypropylene

Abstract

Discontinuous-glass-fiber-reinforced plastic (GFRP) composite material is widely used in industrial fields, mainly because glass fiber improves the strength and stiffness of polymer and because it is much less expensive than carbon fiber. It is thought that the use of long fiber is important in more efficiently improving the strength and stiffness of composites. In our previous study [1], we reported that the fracture mode of the discontinuous-fiber-reinforced composite changes from the matrix-cracking mode to the fiberbreaking mode (Fig. 1), when the aspect ratio for the fiber length to the fiber radius exceeds about 150. We also demonstrated that the strength is dramatically improved compared to the strength of the short fiber-reinforced composites and that the global load-sharing (GLS) model can roughly predict the strength (Fig. 2). Recently, Thomason [2, 3] produced the long discontinuous-fiberreinforced composites where the aspect ratio was about 250 and reported the stiffness and strength of the composites. In this article, we applied the GLS model to his experiments and discuss the validity of our models. First, we discuss the strength of unidirectional (UD) discontinuous-fiber-reinforced composites, rUD, based on the GLS assumption [4, 5]. We applied two types of GLS approaches to predict the composite strength. The GLS model focuses on one fragmented fiber (i.e., discontinuous fibers), aligned in the fiber axial direction, and neglects the interaction among fibers in the fiber cross-sectional direction. It predicts the composite’s strength by simulating the fiber damage evolution in such a fiber. One approach is based on Monte Carlo simulation [6] for fragmentation in a fiber in the composites. The other is based on the analytical model by Duva et al. [5]. (Hereafter, we refer to this as the DCW model.) Monte Carlo simulation deals with a detailed fiber stress distribution and fragment distribution, though multiple calculations are required for the prediction because it is a probabilistic approach. In contrast, the DCW model assumes an approximate stress distribution and fragment distribution, but it predicts the composite strength analytically. In simulating the fiber-damage evolution, the first approach utilized Monte Carlo simulation with the elastic– plastic hardening shear-lag model given by Okabe and Takeda [7]. The schematic of the elastic–plastic shear-lag model is illustrated in Fig. 3. The axial length of the model was set to 25 9 lf (lf is the length of discontinuous fiber), and the axial length was divided into 10,000 segments. The fiber ends in discontinuous-fiber-reinforced composites were represented by setting some random segments to the initially broken segments. Thus, the averaged length lf of the discontinuous fibers was related to the density of the initially broken segments introduced in the model. The transverse length of the matrix shear region in the model