Nonlinear XFEM modeling of delamination in fiber reinforced composites considering uncertain fracture properties and effect of fiber bridging

• Main Author: Motamedi, Damoon
• Language: English
• Published: University of British Columbia 2013
• Online Access: http://hdl.handle.net/2429/44630

Initiation and propagation of a crack in composite materials can affect their global mechanical properties severely. From a numerical modeling perspective, most conventional macro-level methods reported for composite laminates are based on the assumption that a Representative Volume Element (RVE) of the material is periodically repeated over the entire sample. However, a considerable amount of spatial non-uniformity in material and geometrical parameters can exist in both unidirectional (UD) and woven fabric composites. The scattered distribution of fibers, fibers penetration between composite layers, voids within the matrix, human errors during sample preparation, and imperfect thickness distribution can be among the most common sources of such non-uniformity. In turn, these non-uniformities can make the numerical simulation of composites under the assumption of a periodic RVE unreliable, and thereby, the stochastic modeling of effective material properties becomes essential for a more precise assessment of composites’ mechanical behaviour. In the present work, a new three-dimensional (3D) stochastic extended finite element method (XFEM) is proposed and implemented to model the delamination surface in composite samples by integrating the capabilities of the finite element method (FEM) commercial software (ABAQUS) into a user-defined FORTRAN code and MATLAB package. XFEM is known to offer significant advantages over conventional FEM by enabling optimal convergence rates in the presence of pronounced discontinuities/singularities such as cracks. The effect of nonlinear modeling parameters such as cohesive zone length, penalty stiffness factor and large deformation are also considered in the proposed approach to add to the accuracy of simulations. The XFEM model is first tested and validated against previously reported data in the literature. Next, a statistical distribution is sought from data non-repeatability during a set of double cantilever beam (DCB) and end-notched flexure (ENF) tests conducted on Poly (phenylene Sulfide) PPS/Glass thermoplastic composite samples. Results from the experiments and XFEM are compared and demonstrate the capability of the new numerical approach in capturing non-repeatable material response, often seen during the fracture testing of UD composites to characterize their mode I and mode II fracture properties. === Applied Science, Faculty of === Engineering, School of (Okanagan) === Graduate