Composites play an increasingly important role in contemporary aircraft. Delaminations are of significant concern for composites and have been observed in aircraft in operational use. Compliance with criteria defined in JSSG2006 requires growth of delaminations to be either avoided, or to be predictable, slow and stable. This thesis examines historical approaches to delamination modelling, and proposes methods for accounting for the extensive data scatter seen in carbon/epoxy composites to provide indicative thresholds for airframe design. The growth of delaminations from naturally occurring material defects is also studied, and a novel approach of modelling delaminations and non-linearity in composites is proposed.