Composites Science and Technology vol:114 pages:42-49
Three issues are investigated that may influence the accuracy of strength models for unidirectional composites. Firstly, most authors limit themselves to 25–100 single fibre tests to determine the Weibull distribution. This is insufficient to accurately determine this crucial input parameter, leading to significant errors in the predicted composite failure strains. Secondly, random, square and hexagonal fibre packings are shown to lead to a similar predicted failure strain and cluster development. This is the first strong indication that the type of fibre packing has little influence on modelling predictions. Finally, boundary fibres introduced at the model perimeter were shown to prevent preferential cluster formation near the model perimeter. This makes the models less sensitive to the number of fibres. Boundary fibres do not influence the early cluster development, but do increase the critical cluster size. The results provide guidelines for when such boundary fibres should be included.