R C. Peterson

The Effect of Nonlinear Viscoelasticity on Interfacial Shear Strength Measurements

Experimental evidence demonstrates that diglycidyl ether of bisphenol-A (DGEBA)/meta phenylenediamine (m-PDA) epoxy resin matrix used in the single fiber fragmentation tests exhibits nonlinear stress strain behavior in the region where E-glass fiber fracture occurs. In addition, strain hardening after the onset of yield is observed. Therefore, linear elastic shear-lag models and the Kelly-Tyson model are inappropriate for the determination of the interfacial shear strength for this epoxy resin system. Using a strain-dependent secant modulus in the Cox model, the calculated interfacial shear strength is shown to be relatively lower by at least 15% than the value determined using a linear elastic modulus. This decrease is consistent with numerical simulations which show the linear elastic approximation over predicts the number of fragments in the fragmentation test. In addition, the value obtained by the strain-dependent secant modulus is approximately 300% relatively higher than the value predicted by the Kelly-Tyson model.

Measuring Interface Strength in Single Fiber Composites: The Effect of Stress Concentrations

Fiber-matrix interface strength is known to be a critical factor in controlling the long-term performance of structural composites. This parameter is often obtained by using the average fragment length data generated from the single-fiber fragmentation test (SFFT). The interfacial shear strength is then obtained by using this data in a micro-mechanics model that describes the shear-stress transfer process between the matrix and the fiber. Recently, a non-linear viscoelastic micro-mechanics model was developed to more accurately account for the matrix material properties. This new model indicates that the interface strength is dependent on the testing rate. Experimentally, it has been shown that the final fragment length distribution in some systems is dependent on the testing rate. However, data analysis using the new model indicates that the distribution change with testing rate is promoted by the presence of high stress concentrations at the end of the fiber fragments. From the model, these stress concentrations were found to exist at very low strain values. Experimentally, the fragment distributions obtained from specimens tested by different testing rates were found to be significantly different at strain values well below the strain values required to complete the test. These results are consistent with the research of Jahankhani and Galiotis and finite element calculations performed by Carrara and McGarry. These authors concluded that stress concentrations can promote failure of the fiber-matrix interface on the molecular level. Our results support this conclusion. In addition, our research results suggest that altering the SFFT testing rate can lower the magnitude of these stress concentrations and minimize failure of the fiber-matrix interface.

Viscoelastic properties of a resin commonly used in the single fiber fragmentation test

The fiber-matrix interface can play an important role in the performance of a composite, and consequently, it has been the subject of considerable study. One of the experiments often used to characterize the strength or quality of the interface is the single-fiber fragmentation test. The models used to analyze the data from this test involve a number of assumptions, one of which is the constitutive behavior of the matrix resin. To evaluate this assumption, a fragmentation apparatus was modified to include a load cell so both stress and strain could be measured during the experiment. Surprisingly, the results show that not only is the behavior viscoelastic, but virtually all of the fragmentation takes place in a range where the response is non-linear. To characterize this behavior, single-step, stress-relaxation experiments were conducted on a resin system often used in such tests. The results indicate that a simple power law model with strain-dependent parameters could describe the behavior over a very wide range of conditions. By using this characterization and the strain history, a crude fit to the actual loading curve in a fragmentation test could be obtained. In order to achieve quantitative agreement, however, a modified power law model was required. Such a relationship was shown to describe the loading curve for two quite different loading procedures.