Prediction of failure probability of large unidirectional fibre reinforced composites

Tensile strength data obtained from measurement of small test specimens cannot be easily extrapolated to large composite structures. A statistical model is proposed which predicts the strength of composites from the properties of their constituents. It is assumed that with increasing load fibers fail at their most severe flaws and overload adjacent fibers and larger clusters of broken fibers develop locally. At a certain size cluster growth gets critical, i.e. the composite fails. The probability of formation of one cluster of the critical size characterizes the failure probability of the composite. Three different local load sharing rules are compared. The predicted composite strength is insensitive to the chosen local load sharing rule. A parametric study for different Weibull moduli of the fibers, different sizes and failure probabilities of the composite is conducted. The critical cluster size and the tensile strength of the composite are predicted.