Fiber/matrix interface cracking plays an important role in determining the final failureof unidirectional (UD) composites. When subjected to longitudinally tensile loading,fiber/matrix interface debonds originate from fiber breaks or initial defects propagatealong loading direction. Depending on the quality of fiber/matrix interface, debondscould keep growing longitudinally which leads to the degradation of compositestiffness or kink out of interface and connect with neighboring debonds or fiberbreaks that forms a so called critical fracture plane which leads to the final failure ofUD composite. For UD composite subjected to transversely tensile loading, theinitiation, growth and coalesce of arc-shape fiber/matrix interface debonds result inthe formation of macro-size transverse cracks, the propagation and multiplication ofthese transverse cracks, although would not directly lead to the final failure ofcomposite, could cause significant stiffness degradation of composite structures.In the presence thesis, the growth of a fiber/matrix interface debond of a UDcomposite with hexagonal fiber packing under longitudinal and transverse tensileloading was investigated numerically, with the special focus on the influence ofneighboring fibers. In the current study, energy release rate (ERR) is considered as thedriving force for the debond growth and was calculated based on J Integral andVirtual Crack Closure Technique (VCCT) using finite element software ANSSY.Papers A – C in the present thesis deal with the influence of neighboring fibers on theERR of a debond emanating from a fiber break under longitudinal loading condition.In longitudinal loading case, debond growth is mode II dominated. In paper A, anaxisymmetric model consisting 5 concentric cylinders that represent broken fiber withdebond, surrounding matrix, neighboring fibers, surrounding matrix and effectivecomposite was generated. It’s found that there are two stages of debond growth, thefirst stage is when debond length is short, the ERR decreases with increasing debondlength, and the presence of neighboring fibers significantly increase the ERR ofdebond. For relatively long debond, the debond growth is steady when ERR is almostconstant regardless of debond length. In steady state of debond growth, the presenceof neighboring fibers have little effect on the ERR. In papers B and C, a 3-D modelwas generated with broken fiber and its 6 nearest fibers in a hexagonal packed UDcomposite were modelled explicitly, surrounded by the homogenized composite.

Based on the obtained results, it’s shown that ERR is varying along debond front, andhas its maximum at the circumferential location where the distance between two fibercenter is the smallest. This indicates that the debond front is not a circle. For steadystate debond, the presence of neighboring fibers have little effect on averaged ERR(averages of ERR along debond front). For short debond, the presences ofneighboring fibers increases the averaged ERR, and that increase is more significantwhen inter-fiber distance is the smallest. Paper D investigates the growth of afiber/matrix debond along fiber circumference under transverse loading. It’s foundthat debond growth in this case is mixed-mode, and both mode I and mode II ERRcomponents increase with increasing debond angle and then decreases. Debondgrowth is mode I dominated for small debond angle and then switch to mode IIdominated. The presence of neighboring fibers have an enhancement effect on debondgrowth up to certain small debond angle and then changes to a protective effect. InPaper E, the interaction between two arc-size debond under transverse loading isinvestigated. It’s found that when two debonds are close to each other, the interactionbetween two debond becomes much stronger, and that interaction leads to the increaseof ERR of each debond significantly, which facilitates further growth for bothdebond.