Co-authors​ :

 David MOLLENHAUER (UNITED STATES), Mark FLORES (UNITED STATES), Robert WHEELER , Jeremiah LIPP , Keith BALLARD (UNITED STATES), Eric ZHOU (UNITED STATES), Kevin HOOS (UNITED STATES), Endel IARVE (UNITED STATES), Mathew SCHEY (UNITED STATES) 

As part of a wider micro-scale composite test program, an off-axis IM7/977-3 carbon fiber/epoxy micro-pillar was mechanically tested in compression to failure under observation within an SEM. The pillar dimensions were approximately 17 by 19 by 61 microns with the fibers at ~20 degrees off-axis. Fabrication of the specimen was accomplished through a series of micro-machining operations, followed by focused ion beam (FIB) etching. The machining process left one end of the pillar free for load application and the other end as bulk composite for gripping. X-ray CT scans and SEM observations were performed to guide the FIB machining process in the production of a desired final geometry, while also recording the initial, untested geometric configuration (Fig 1a). While testing, the micro-pillar was imaged in an SEM at 20 applied loads. The volume averaged axial strain was calculated using the relative displacement between local features at the ends of the pillar, with the features acting as fiducial marks. Failure, as expected, occurred through a combination of matrix cracking and fiber-matrix debonding. In addition to the experiment, a modeling campaign is being conducted. A discrete damage modeling research code developed by the US Air Force, called BSAM, was used to model the progressive failure of the micro-pillar. The tool allows for discrete damage within constituents in a mesh-independent manner using a regularized extended finite element method (RX-FEM) and along material interfaces using a cohesive surface formulation, which were used to account for discrete cracking in the matrix and fiber-matrix disbonding, respectively. Initial modeling consisted of using idealized specimen dimensions, average fiber volume fraction, and a fixed boundary condition instead of explicitly modeling the composite bulk (“Model 1” in Fig 2). An additional model was built that had more precise fiber locations based on SEM imaging and includes the bulk composite grip region (Fig 1b and Fig 2 “Model 2”). Simulated stress-strain curves and failure mechanisms compare favorably to experiment, with the simulated compliance being approximately 13% too stiff in the linear region (see Fig 2). Model 1 over-predicts the nonlinear behavior, while showing qualitative similarity to the experiment. Model 2 tracks the experiment remarkably well considering the material properties used in the model were determined via inverse methods from macro specimen testing. The experimentally determined micro-pillar load drop is not represented properly by either Model 1 or Model 2. It is likely that this is due to additional damage mechanisms not being modeled and/or geometric nonlinear effects such as fiber buckling. Ongoing simulation investigations are pursuing more accurate specimen representations through extraction of geometric features from CT imaging, insertion of directly measured constituent properties, and the addition of geometric and material nonlinearity.