Experimental testing is an essential part of the design process for composite components and structures, both to ascertain the material properties of the specific composite materials used and to validate the expected structural response in terms of stiffness and strength. Material property tests are carried out on small laboratory scale 'coupons' and scale-model prototypes of components and especially structures are often the only economically viable option to validate full-scale behaviour.
The small-scale properties and responses must of course reflect those at full-scale. The principles of dimensional analysis are often used to scale up results, but this relies on having a complete model of what is often a very complex system. If any part of this model is incomplete or only approximate, then discrepancies are encountered whilst scaling from model-scale to full size, and these errors are referred to as 'size-effects'. Size effects are potentially dangerous because they may well lead to a structure being significantly weaker than was calculated during it's design.
A commonly proposed hypothesis for size effects in composite materials is based on 'weakest link' theory where the failure of the material as a whole is assumed to occur when any one of it's many brittle fibres break; much like one link of a chain leading to the failure of a chain. Since there is a given statistical chance of any given link failing, then as the number of links is increased so too does the chance of one link and hence the whole chain failing. Since this effect is not included in the mathematical models used to scale up laboratory data, this behaviour can lead to a 'size-effect' where a larger structure is weaker than those at smaller prototype or laboratory scale.
During my doctoral studies I studied these 'size effects' for typical marine composites via a large experimental study, by applying the techniques of statistical experimental design. The main results were that any 'weakest link' size effects were too small to be seen above the high statistical variability inherent in hand-laid up marine GRP, but that processing and manufacturing differences between laboratory and shipyard scale marine composites could easily lead to errors if not carefully controlled, and I distinguished between the two by characterising the latter as 'scale-' rather than 'size-effects'.
I have also completed a study on the scaling of impact on marine composites (see 'Publications').