A closed form relation for the strain energy density in the vicinity of a macroscopic mode I crack in a random fiber network is derived using an implicit gradient nonlocal continuum field theory. An expression for the characteristic length, used in the nonlocal formulations, in terms of microstructural properties is derived and it is found that the characteristic length is proportional to the average fiber segment length to the power of two. It is illustrated that the crack-tip singularity vanishes for a characteristic length greater than zero. An open fiber structure exhibits a distributed strain energy field in the crack tip vicinity. As the network becomes relatively denser, the characteristic length decreases and the networks mechanical behavior approaches the behavior of a classic elastic continuum. Only for an infinitely dense network is the r −1-singularity in strain energy field achieved. The theory explains why open network structures have difficulties in localizing failure to macroscopic cracks. It is found that there is a one-to-one relation between characteristic length and size of the smallest crack that can initiate macroscopic failure.

A directional crack growth prediction in a compressed homogenous elastic isotropic material under plane strain conditions is considered. The conditions at the parent crack tip are evaluated for a straight stationary crack. Remote load is a combined biaxial compressive normal stress and pure shear. Crack surfaces are assumed to be frictionless and to remain closed during the kink formation wherefore the mode I stress intensity factor KI is vanishing. Hence the mode II stress intensity factor KII remains as the single stress intensity variable for the kinked crack. An expression for the local mode II stress intensity factor k2 at the tip of a straight kink has been calculated numerically with an integral equation using the solution scheme proposed by Lo (1978) and refined by He and Hutchinson (1989). The confidence of the solution is strengthened by verifications with a boundary element method and by particular analytical solutions. The expression has been found as a function of the mode II stress intensity factor KII of the parent crack, the direction and length of the kink, and the difference between the remote compressive normal stresses perpendicular to, and parallel with, the plane of the parent crack. Based on the expression, initial crack growth directions have been suggested. At a sufficiently high non-isotropic compressive normal stress, so that the crack remains closed, the crack is predicted to extend along a curved path that maximizes the mode II stress intensity factor k2. Only at an isotropic remote compressive normal stress the crack will continue straight ahead without change of the direction. Further, an analysis of the shape of the crack path has revealed that the propagation path is, according the model, required to be described by a function y = cxγ , where the exponent γ is equal to 3/2. In that case, when γ = 3/2, predicts the analytical model a propagation path that is selfsimilar (i.e. the curvature c is independent of any length of a crack extension), and which can be described by a function of only the mode II stress intensity factor KII at the parent crack tip and the difference between the remote compressive normal stress perpendicular to, and parallel with, the parent crack plane. Comparisons with curved shear cracks in brittle materials reported in literature provide limited support for the model discussed.

Directional crack growth criteria in compressed elastic–plastic materials are considered. The conditions at the crack tip are evaluated for a straight stationary crack. Remote load is a combined hydrostatic stress and pure shear, applied via a boundary layer assuming small scale yielding. Strains and deformations are assumed to be small. Different candidates for crack path criteria are examined. Maximum non-negative hoop stress to judge the risk of mode I and maximum shear stress for mode II extension of the crack are examined in some detail. Crack surfaces in contact are assumed to develop Coulumb friction from the very beginning. Hence, a condition of slip occurs throughout the crack faces. The plane in which the crack extends is calculated using a finite element method. Slip-line solutions are derived for comparison with the numerically computed asymptotic field. An excellent agreement between numerical and analytical solutions is found. The agreement is good in the region from the crack tip to around halfway to the elastic–plastic boundary. The relation between friction stress and yield stress is varied. The crack is found to extend in a direction straight ahead in shear mode for sufficiently high compressive pressure. At a limit pressure a kink is formed at a finite angle to the crack plane. For lower pressures the crack extends via a kink forming an angle to the parent crack plane that increases with decreasing pressure.

Dynamic fracture behavior in both fairly continuous materials and discontinuous cellular materials is analyzed using a hybrid particle model. It is illustrated that the model remarkably well captures the fracture behavior observed in experiments on fast growing cracks reported elsewhere. The material's microstructure is described through the configuration and connectivity of the particles and the model's sensitivity to a perturbation of the particle configuration is judged. In models describing a fairly homogeneous continuous material, the microstructure is represented by particles ordered in rectangular grids, while for models describing a discontinuous cellular material, the microstructure is represented by particles ordered in honeycomb grids having open cells. It is demonstrated that small random perturbations of the grid representing the microstructure results in scatter in the crack growth velocity. In materials with a continuous microstructure, the scatter in the global crack growth velocity is observable, but limited, and may explain the small scattering phenomenon observed in experiments on high-speed cracks in e.g. metals. A random perturbation of the initially ordered rectangular grid does however not change the average macroscopic crack growth velocity estimated from a set of models having different grid perturbations and imply that the microstructural discretization is of limited importance when predicting the global crack behavior in fairly continuous materials. On the other hand, it is shown that a similar perturbation of honeycomb grids, representing a material with a discontinuous cellular microstructure, result in a considerably larger scatter effect and there is also a clear shift towards higher crack growth velocities as the perturbation of the initially ordered grid become larger. Thus, capturing the discontinuous microstructure well is important when analyzing growing cracks in cellular or porous materials such as solid foams or wood.