Interfacial debonding and fibre pull-out stresses - Part I Critical comparison of existing theories with experiments

Two current theories [11, 17] of interfacial debonding and fibre pull-out, which have been developed on the basis of fracture mechanics and shear strength criteria, respectively, are critically compared with experimental results of several composite systems. From the plots of partial debond stress, σ_d^p, as a function of debond length, three different cases of the interfacial debond process can be identified, i.e. totally unstable, partially stable and totally stable. The stability of the debond process is governed not only by elastic constants, relative volume of fibre and matrix but more importantly by the nature of bonding at the interface and embedded fibre length, L. It is found that for the epoxy-based matrix composite systems, Gao et al.'s model [17] predicts the trend of maximum debond stress, σ_d^*, very well for long L, but it always overestimates σ_d^* for very short L. In contrast, Hsueh's model [11] has the capability to predict σ_d^* for short L, but it often needs significant adjustment to the bond shear strength for a better fit of the experimental results for long L. For a ceramic-based matrix composite, σ_d^* predicted by the two models agree exceptionally well with experiment over almost the whole range of L, a reflection that the assumed stable debond process in theory is actually achieved in practice. With respect to the initial frictional pull-out stress, σ_f, the agreement between the two theories and experiments is excellent for all range of L and all composite systems, suggesting that the solutions for σ_f proposed by the two models are essentially identical. Although Gao et al.'s model has the advantage to determine accurately the important interfacial properties such as residual clamping stress, q_o, and coefficient of friction, μ, it needs some modifications if accurate predictions of σ_d^* are sought for very short L. These include varying interfacial fracture toughness, G_ic with debond crack growth, unstable debonding for very short L and inclusion of shear deformation in the matrix for the evaluation of G_ic and fibre stress distribution. Hsueh's model may also be improved to obtain a better solution by including the effect of matrix axial stress existing at the debonded region on the frictionless debond stress, σ_o.