Polymer matrix composites are finding increased use in structural applications, in particular for aerospace and automotive purposes. Mechanical fatigue is the most common type of failure of structures in service. The relative importance of fatigue has yet to be reflected in design where static conditions still prevail. The fatigue behavior of composite materials is conventionally characterized by a Wöhler or S-N curve. For every new material with a new lay-up, altered constituents or different processing procedure, a whole new set of fatigue life tests has to be repeated for such a characterization. If the active fatigue damage micromechanisms and the influence of the constituent properties and interface were known, it would be possible, at least qualitatively, to predict the macroscopic fatigue behavior. A study of the fatigue damage mechanisms would also give indications of the weakest microstructural element, which is useful information in materials selection for improvement in service properties. In tensile fatigue of a multidirectional laminate, the critical elements are the longitudinal plies which are the last to fail. Although failure of neighboring off-axis plies as well as delamination will influence the fatigue process, an understanding of the behavior of the longitudinal plies forms an important foundation. Effects of plies of other directions may then be interpreted based on this foundation. Fatigue of longitudinal plies is therefore focused on in the present study. The underlying fatigue damage mechanisms were investigated for unidirectional O' carbon fiber reinforced plastics (CFRP) and glass fiber reinforced polypropylene (GF/PP) in tension-tension fatigue. By use of a surface replication technique the evolution of fatigue damage could intermittently be monitored during the course of fatigue testing. In the CFRPS, the matrix was an epoxy resin or polyetheretherketone (PEEK). In the GF/PP system, the matrix was modified with maleic anhydride (MA) to achieve a stronger fibermatrix interface. The macroscopic fatigue behavior was characterized by fatigue life diagrams. A statistical method has been devised to systematically characterize fatigue life data in terms of fatigue life diagrams. On the microscopic level, the CF/epoxy and GF/MA-PP composites have relatively strong interfaces and showed localized and scarce fiber breaks from which matrix cracks propagated perpendicular to the fiber direction. In CF/epoxy, fiber bridged cracks with squeezed fiber tips appeared. Conversely, CF/PEEK and GF/PP have weaker interfaces, and the principal mechanisms were extensive and distributed debonding or longitudinal matrix cracking followed by further fiber breakage. Macroscopically, the weak interface composites showed shorter fatigue lives and more rapid fatigue degradation. This suggests that higher interfacial strengths lead to improved fatigue performance. Modeling studies were undertaken for the two observed mechanisms; debonding from a fiber break, and fiber bridged cracking. The stochastic breakage of fibers next to a growing debond was parametrically investigated with a shear lag model. The stress profile in the surviving fibers becomes attenuated and more distributed as the debonds grow. This results in longer axial distances between fiber breaks, and hence a more jagged and uneven crack propagation. A larger variability in strength along the fibers has basically the same break distributing effect. With a more homogeneous stress distribution caused by long debonds, the variability in fiber stress at failure of the intact fibers decreases. This can explain the experimentally observed lower scatter in fatigue life of composites exhibiting a more homogeneous distribution of damage caused by debonding. Furthermore, the experimental results of fiber bridged cracking was modeled with a fracture mechanics approach. The crack growth curve can be plotted in terms of the effective stress intensity factor where the contribution of the cohesive crack surface forces from the bridging fibers are taken into account. This curve falls somewhat closer to that of the neat matrix material compared to the unbridged crack, but the difference is still considerable. Besides the fiber bridging, there should therefore be other active toughening mechanisms that slows the crack propagation down to account for the fatigue resistant behavior of the tested material. In fatigue of multidirectional laminates, tension-compression loading has shown to be more detrimental than tension-tension loading. The reason for this behavior has not been entirely clarified. The adverse effect of the compressive load excursions is partly caused by the formation of transverse cracks. This was verified by counting transverse cracks in cross-ply laminates. Since debonding is the subcritical mechanism which leads to transverse cracking and eventually influences ultimate failure, the debonding was studied in low cycle fatigue of a single transverse fiber. In tension, contact zones developed at the crack tips for sufficiently large debonds. Due to the inherent geometry and the mismatch in elastic properties of the constituents, an opening zone appeared at the crack tips of the debond in compression. This was also verified by finite element analysis. Since debond propagation is more suseptible to mode I loading, the sensitivity to tension-compression loading is explained by the effective opening zone in compression.
Luleå: Luleå tekniska universitet, 1997. , 109 p.