Finite element modelling and experimentation of plain weave glass/epoxy composites under high strain-rate compression loading for estimation of Johnson-Cook model parameters

Abstract

Split Hopkinson pressure bar (SHPB) is a widely used experimental technique for testing materials at a high strain-rate. SHPB testing of fiber-reinforced polymer (FRP) composite samples have become common in recent decades. Specimen geometry is one of the essential parameters that can ease sample fabrication and help in-situ analysis when altered as per convenience. Further, developing a computationally accurate model would reduce the need for multiple experimental iterations, thus saving cost and time. As part of the current study, high strain-rate tests using a compressive SHPB setup were conducted on glass/epoxy (GE) laminated composite samples of both square prismatic and cylindrical shape along the through-thickness direction experimentally. Negligible differences were observed due to the change in specimen geometry. Further, we know that the woven fabric consists of yarns, which are a bundle of fibers. The strain-rate dependency of the epoxy's material property is well known, but very limited information is available about the strain-rate dependency of yarn's properties. The effect of strain-rate dependency of the yarn was analysed numerically on Abaqus software using a square prismatic shaped yarn-level finite element composite model. The homogenised properties of the yarn were taken to be reflective of a high fiber volume fraction composite. The bulk epoxy's and yarn's material property were defined using the Johnson-Cook (JC) constitutive model, and the failure mechanism was defined using the JC damage model. One of the primary reasons for using this model is its availability and ease of implementation in almost every finite element analysis (FEA) simulation software. Three cases were analysed numerically, first keeping both yarn and bulk matrix property strain-rate dependent, second keeping the strain-rate dependency only in the bulk matrix, and in third case both yarn and matrix were kept strain-rate independent. It was observed that completely strain-rate dependent model showed comparable peak stress and strain values with the experimental result than other models.