Compression Failure Mechanics in FRP Composites

Doctoral research into a long-standing problem in composite testing: standard compression specimens for FRP materials tend to fail prematurely, near their grips or at sharp geometric transitions, rather than in the zone of interest — systematically underestimating true compressive strength. This work developed a new flat "hourglass" specimen geometry, with a smoothly curved waisted section, specifically to force failure into the centre of the gauge section where it can be properly characterised.

The geometry was validated experimentally on two structurally different materials — an unbalanced E-glass/epoxy woven and a balanced carbon/PEEK woven — using full-field 3D digital image correlation to track strain right up to the point of fracture, and the resulting data was used to build a progressive, non-linear damage model of compressive behaviour.

The hourglass geometry was adapted from a tensile specimen shape used elsewhere in the literature, redesigned specifically for end-loading compression (conforming to ASTM D695) and four-point bending (conforming to ASTM D6272). Rather than the sharp width transition of a standard dog-bone specimen, the hourglass shape uses a large, constant curvature radius through its waisted section, chosen to be as large as the test fixture's dimensional constraints would allow — because a larger radius means a smaller in-plane stress concentration. FEM simulation quantified the resulting stress concentration factors at 1.045 in compression and 1.035 in bending, compared to 1.20 for the standard ASTM dog-bone geometry it replaced. The practical effect was decisive: hourglass specimens consistently failed near the centre of the gauge section, while dog-bone specimens failed prematurely near the sharp width transition close to the loaded end.

Full-field strain was measured throughout using 2D and 3D digital image correlation, cross-validated against strain gauges and a video-extensometer to confirm measurement accuracy. In four-point bending, DIC-derived through-thickness strain distributions confirmed linear, Kirchhoff-consistent behaviour and the absence of interlaminar delamination, allowing skin stresses to be extracted via classical beam theory even under large deflections — for which a dedicated geometrical protocol was derived and validated against FEM to correctly account for the moving contact points between specimen and loading rollers. In compression, DIC confirmed a homogeneous strain field across the curved gauge section, validating the geometry's core design assumption.

On the E-glass/epoxy woven, hourglass specimens in pure compression failed at 765 ± 23 MPa and −2.72 ± 0.04% strain, against 688 ± 18 MPa and −2.24 ± 0.08% strain for standard dog-bone specimens under the same test conditions — an 11% gain in failure stress and 24% gain in failure strain, simply from relocating failure away from the stress-concentrated zone. On the carbon/PEEK woven, tested in four-point bending, compressive failure occurred at 730 MPa and −1.6% strain via visible kink-band formation, with a markedly non-linear stress-strain response.

That non-linearity was captured with a progressive damage constitutive model: a power-law relationship between stress and strain, fitted to the DIC-derived compressive response by error minimisation, identifying a non-linearity parameter that reproduced the carbon/PEEK's progressive stiffness evolution under load all the way to fracture. Together, the specimen geometry, full-field strain measurement, and damage model gave a more accurate and more complete picture of compressive failure behaviour than standard test methods allow — directly relevant to avoiding overly conservative composite structural designs.