CFRP Composite Research
My undergraduate research focused on understanding how manufacturing-induced defects impact the performance of carbon fiber-reinforced polymer (CFRP) composites made from repurposed aerospace prepreg scraps. This page highlights the specimen manufacturing process, tensile testing, and defect analysis through microscopy, supported by finite element modeling and computational simulations. The work is driven by the need to reduce composite waste and to explore sustainable upcycling approaches for advanced materials.
Research Advisor: Dr. Paulina Diaz-Montiel and I at the American Society of Composites 2024 Conference
Experimental Work
Composite Coupon Preparation
To explore sustainable upcycling methods, I manufactured CFRP laminates using expired Cycom 5320-1 T650-35 3K carbon/epoxy woven prepreg. The scrap material was cut into three different sizes (1x0.5 in, 1x1 in, 2x0.5 in) and arranged in both structured brick-style and parallel layups, each built to a 6-ply thickness. The laminates were cured through the out-of-autoclave (vacuum bag only) curing method, then tabbed with fiberglass and cut into 1 in × 12 in coupons.
The experimental setup followed ASTM D3039 standards for polymer matrix composites. This controlled approach isolated scrap size and layup as the key variables, allowing us to study how these parameters influence defect formation and tensile properties.
Figure 1. Scrap layup configurations for CFRP composites
Figure 2. Manufactured CFRP coupons with fiberglass tabs, grouped by scrap size
Microscopy and Defect Characterization
After manufacturing, each coupon was carefully examined using optical microscopy and scanning electron microscopy (SEM). Over 1,000+ high-resolution images were captured, and 30 coupon cross-sections where collaged together for the post-microscopy image analysis. Using Adobe Photoshop for manual inspection and CellProfiler an open source AI image processing software, we quantified defects such as:
-
Resin-rich regions (epoxy build-up between scraps)
-
Circular voids (air entrapments up to 400 μm)
-
Inter-laminar voids (gaps between adjacent scraps)
-
Complete voids (visible holes in the microstructure)
-
Micro-voids (<20 μm, not quantified but observed via SEM)
Quantitative analysis revealed that brick-style layups averaged only 0.7% void content, while parallel layups reached over 4.4% void content, despite similar resin-rich volume fractions of about 11%. This difference directly correlated with reduced tensile performance in parallel layups, confirming that scrap arrangement governs defect nucleation and distribution.
Figure 3. Optical imaging and processing setup used for composite coupon analysis.
Figure 4. Characterization of defects observed in the cross-sectional area of the composite coupons: (a) resin-rich region, (b) circular void, (c) inter-laminar void, (d) complete void, and (e) micro void. The optical imaging corresponds to composite coupon 2B1 (Plate 2).
Figure 5. Optical imaging of (a) Plate 1 specimen 1B2 and (b) Plate 2 specimen 2B2, comparing void dimensions. Plate 1 (brick layup) exhibits a higher content of inter-laminar voids, while Plate 2 (parallel layup) predominantly has larger circular voids between the ends of scraps.
Figure 6. Quantification of defects results highlighting the variability between Plate 1 (brick layup) and Plate 2 (parallel layup) in terms of Resin Rich Regions (RRR), void content, and scrap percentages. Plate 2 exhibits notably higher void levels and greater variability compared to Plate 1.
Figure 7. Scanning Electron Microscope (SEM) images showing circular and micro-voids within the cross-section of the composite coupons
Tensile Testing
Mechanical testing was conducted on an Instron 3382 electromechanical universal testing system, with a 2 in extensometer measuring local strain under controlled displacement (2 mm/min). All coupons were tested under identical conditions to ensure repeatability.
Results demonstrated a stark contrast:
-
Brick-style laminates averaged 43.8 ksi tensile strength with stiffness near 8.8 Msi, comparable to continuous-fiber laminates in stiffness but reduced in strength.
-
Parallel laminates dropped to ~8.6 ksi tensile strength, representing an 80% knockdown in performance despite similar stiffness (~7.9 Msi).
Failure analysis showed that brick layups failed via fiber fracture and longer crack propagation paths, while parallel layups failed prematurely at resin-rich regions and void clusters, where cracks initiated and propagated rapidly.
Figure 8. Image of the experimental setup (left). An extensometer (right) was utilized for measuring the local deformations during testing.
Figure 9. Tensile stress–strain results showing ~80% strength knockdown and contrasting behavior of brick vs. parallel layup composites.
Post-Mortem Analysis
Post-mortem imaging of the coupons that were tested to ultimate tensile failure revealed how scrap layup dictated the nucleation and propagation of cracks. In the brick layup, defects were primarily small inter-laminar voids and resin-rich regions, which produced longer crack paths. As a result, crack propagation distances were more than twice as long as those observed in parallel laminates, often involving fiber fracture and tow pull-out. By contrast, the parallel layup was dominated by larger circular voids at scrap gap regions, which acted as preferential crack initiation sites. In this case, cracks followed the shortest distance path between resin-rich pockets, leading to premature separation at approximately 80% lower tensile strength.
Post-mortem imaging highlighted two distinct failure cases:
-
Case 1 – Scrap/Matrix Separation: Cracks initiated at resin-rich regions, causing delamination, void growth, and scrap debonding.
-
Case 2 – Reinforcement Fracture: Cracks cut through fibers and plies, indicating higher energy absorption prior to catastrophic failure.
Together, these analyses confirmed that layup strategy, not stiffness, was the controlling factor in strength degradation, and that optimized brick-style stacking distributes tensile loads more evenly across scraps.
Figure 10. Failed coupons from tensile testing, showing specific failure locations for each specimen
Figure 11. Progressive failure theories: Case 1: Matrix/scrap separation. Case 2: Reinforcement material failure (fiber fracture).
Figure 12. Postmortem imaging of (a) specimen 1B (Plate 1) and (b) specimen 2D (Plate 2). Specimen 1B shows delamination crack propagation of 0.1925 inches due to voids and resin-rich regions, while specimen 2D shows crack initiation in void clusters with a shorter distance of 0.0748 inches.
Computational Analysis
Introduction
Upcycled laminates made from woven CFRP scraps can fail in messy, progressive ways because of resin-rich gaps, scrap edges, and voids. Physical testing shows big strength differences across scrap layouts, but it’s hard to isolate which defect drives which part of the failure. To complement experiments, we automated a meso-scale FEA pipeline in Python (Abaqus/CAE) to parametrize scrap size/layout and void content, then observe where cracks initiate and how they propagate through the laminate.
Modeling Approach
We generate a 3D representative volume (2 in × 2 in × 2 plies) that explicitly includes discrete woven scraps embedded in an epoxy matrix. Each scrap is created as its own solid part, meshed with C3D8R hexahedra (seed size 0.04 in), then boolean-cut into the matrix so that matching matrix faces exist for cohesive interactions. Materials are assigned as an orthotropic CFRP material property and an isotropic epoxy resin material. The Python script automates the following tasks: scrap creation/placement, matrix cutting via InstanceFromBooleanCut, material assignment & orientation, scrap and matrix meshing, and surface definition for 6 faces per scrap, allowing contacts to be applied programmatically.
The C3D8R element choice and meshing scale are consistent for all composite models (150k–200k nodes), and an implicit nonlinear solution analysis was performed, taking approximately 12 hours per trial.
Three models were created to match the three experimental plates studied in the real world. Each model replicated the real-world scrap sizes we studied (0.5x1 in, 1x1 in, 0.5x2 in) during the composite geometry modeling process. The Python scripting managed to save hours of painstaking user-defined cohesion interactions, geometry modeling, and meshing. Alongside the three models, an additional three models were created with a 5% average defect volume fraction so that we could study the difference between defective composite coupons and coupons with 0% defects. Defects were modeled by resin-rich gaps arising naturally from the boolean cuts between scraps, creating the matrix ligaments where cohesive failure can start. Voids were implemented by randomly selecting ~5% of matrix elements and assigning a void material with ~1% of the epoxy stiffness (stiffness penalty), following a research group's previous approach cited in our ASC slides, preliminary research.
Figure 13. Representative Volume Element (RVE) meso-scale FEA model (2 in × 2 in × 2 ply) illustrating CFRP scraps within an epoxy matrix under a Brick 1×1 layup configuration.
Figure 14. FEM of CFRP scrap composite with epoxy matrix and voids, showing C3D8R elements (~150k–200k nodes), nonlinear implicit integration, and unidirectional tensile loading with fixed bottom boundary conditions
Cohesive Interactions
Delamination and debonding are modeled by assigning a cohesive traction–separation law to every scrap–matrix surface which interact. In this framework, the interface behaves elastically at first, with penalty stiffness values controlling the slope in both the normal and shear directions. This ensures the scrap and resin remain bonded until stresses reach critical levels.
To avoid artificial constraints, the model assumes frictionless tangential sliding, allowing the scrap to move freely once separation occurs. In the normal direction, a hard contact condition is enforced: surfaces cannot penetrate one another, but they are free to separate when damage initiates.
Fracture is governed by a damage initiation criterion followed by a Benzeggagh–Kenane (BK) mixed-mode energy evolution law with exponent 1.45, which accounts for the nonlinear interaction between opening and shear failure modes. A small amount of viscosity stabilization is included to help the solver converge during rapid softening.
By systematically applying these cohesive interactions to the proper faces of every scrap, the model reproduces observed failure mechanisms such as interface debonding, shearing, and crack deflection around voids. This provides a physically consistent way to bridge interface-level mechanics with the macroscopic fracture behavior of upcycled laminates.
Figure 15. Cohesive interaction modeling between CFRP scraps and epoxy matrix, using traction–separation laws to capture damage initiation and evolution at the scrap–matrix interface.
Boundary Conditions & Loads
The model is loaded in uniaxial tension, applied as a prescribed vertical displacement on the top surface. The bottom surface is fully constrained in translation (U1 = U2 = U3 = 0) so the specimen stretches against a fixed base. This setup mirrors a standard tensile test, ensuring the simulation results can be compared directly with experimental coupons.
Loading is applied through a quasi-static step. To capture the initiation and growth of cohesive damage, the step uses many small increments, with tight convergence tolerances that prevent the solver from skipping over rapid softening events.
In practice, the script applies a displacement of u2 = 0.01 in for quick demonstration runs, while a commented option applies a larger pull of u2 = 0.25 in for full-field simulations. The smaller displacement is useful for debugging and verifying interface behavior, while the larger value allows the simulation to run through complete fracture. This staged approach balances computational efficiency with the ability to show progressive delamination and failure.
Results & Discussion
Experimental Findings
The experiments showed that scrap layup strongly influenced tensile strength. Brick-style layups (Plate 1, 0.5 × 1 in scraps) averaged 43.8 ksi ultimate tensile strength (UTS), nearly five times higher than parallel layups (Plate 2, 1 × 1 in scraps at 8.6 ksi). Overlapping scraps helped distribute load, while parallel gaps concentrated stresses and triggered early failure.
Defect analysis explained these differences. Both plates had ~11% resin-rich content, but Plate 2 contained ~6× more voids (4.4% vs 0.7%). These larger voids clustered at scrap gap regions, acting as crack initiation sites. Plate 1’s overlapping layout reduced circular voids, instead producing smaller interlaminar ones.
Failure modes reflected this defect distribution. Plate 2 fractured cleanly at void clusters, while Plate 1 showed mixed failure, including fiber fracture and longer crack propagation distances, allowing higher loads before failure.
Stiffness (~8–9 Msi) remained similar across all coupons and matched manufacturer data, showing stiffness is less sensitive to layup than strength. Overall, the experiments confirmed that brick-style overlap mitigates defect severity and enables progressive failure, while parallel layups concentrate damage into dominant weak points.
Computational Findings
The simulations confirmed that voids act as stress concentrators, initiating cracks at localized clusters in a manner consistent with the “hot spot” fracture origins seen under post-mortem microscopy. By isolating void placement in the model, the computational work highlighted how even small distributions reduce the margin to cohesive failure, an effect difficult to quantify directly in experiments.
Scrap size was shown to amplify defect sensitivity. Smaller scraps produced sharper knockdowns in strength at the same void content, matching coupon-level results but also revealing the mechanism: damage accelerated because cracks encountered more frequent scrap–matrix boundaries.
Layup geometry dictated the crack path. Parallel layouts funneled cracks directly through resin-rich ligaments, while brick layouts deflected and lengthened the crack path, sometimes requiring fiber fracture. These behaviors reproduced experimental fracture surfaces but also clarified why the two layups diverged.
This computational lens not only validated the experiments but also provided mechanism-level insight into stress redistribution, crack deflection, and defect interactions that physical tests could not directly resolve.
Figure 16. Stress–strain response comparing 1×1 in and 0.5×1 in CFRP scraps, showing higher strain to failure for smaller scraps, with elastic moduli of 82.36 GPa and 48.44 GPa, respectively, and a 50.5% increase in ultimate tensile strength (UTS).
Figure 17. Results highlighting damage progression: (1) initial model before testing, (2) cohesive damage contours, (3) S22 stress distribution in the y-direction, (4) localized crack growth, and (5) focal stress concentration region.
Figure 18. Stress–strain curves comparing layup configurations (left: Brick vs. Parallel) and scrap sizes (right: 1×1 in vs. 0.5×1 in) under 0% (dotted lines) and 5% (solid lines) void content. Results show that voids reduce strength and stiffness, with a more pronounced effect on scrap size than layup, as indicated by δ₁ and δ₂.
Impact and Future Research
This research demonstrates that scrap geometry and layup design are critical for improving the performance of upcycled CFRP composites. By integrating experimental testing, microscopy, and computational modeling, we built a framework for predicting how defects shape stiffness, strength, and failure mechanisms. Together, the work shows that thoughtful layup strategies can mitigate defect severity, while simulations reveal how voids and cohesive interactions drive crack initiation and growth.
Future work will expand in the following two areas:
-
Broader experimental testing, exploring new scrap geometries and randomized orientations to capture a wider range of manufacturing scenarios.
-
Advanced FEA modeling, incorporating more realistic void distributions and inter-laminar damage to simulate progressive failure under multiple load cases.
These efforts aim to move upcycled CFRPs closer to structural applications in aerospace and automotive industries, transforming what is currently waste into high-value, sustainable materials.