Abstract
A comprehensive investigation of the strength distribution of unsupported single fibres (in air) and supported single fibres (embedded in epoxy-resin) in the context of the weak-link theory was conducted. Strength measurements of impregnated bundles were also carried out in order to compare the strength data with the unsupported and supported fibres. The entire experimental work was performed under similar experimental conditions, where a long length of a single fibre, or a bundle of fibres, was cut from the same spool of fibre (Celion fibre) into four specimens of 5, 12, 30 and 75mm gauge lengths respectively, thus ensuring a consistent base for the comparison. All the data produced were characterised using the Weibull model and other statistical tests, such as the non-parametric test, which can test the weak-link property without recourse to Weibull statistics. Unsupported single fibres were found to deviate strongly from the Weibull model and the weak-link behaviour. This was due to the diameter variation found along the lengths of the tested specimens, which violated the implicit assumption in the Weibull model and the weak-link property that specimens should be similar in all respect except length. Another carbon fibre (XAS) was used to compare the Celion fibre data in relation to its behaviour to the Weibull model. The XAS fibre, however, was found to comply well to the Weibull model and the weak-link property, this was due to the mean diameter being uniform across the four gauge lengths. The Celion fibre was shown to have an increasingly strong dependence of strength on the fibre diameter as the gauge length decreases. This was due to the different effect that diameter produced on the four gauge lengths. The overall behaviour of the Celion fibre data was found to be complex due to the presence of two effects: firstly, the presence of two classes of failure flaws which may be represented by surface and volume flaws, and secondly, within the volume flaw mode itself there was a strong diameter dependence. The use of epoxy-resin to support the fibres provided a means to separate these two classes of flaws and to understand the complex behaviour of the unsupported fibres. There was an apparent diameter uniformity found in the supported fibres, and the weak-link property was restored by the use of the resin. The use of the resin has also increased the strength of the fibres. Failure of the fibres was found to be affected by length only, which explained the compliance of the data to the Weibull model. It was found that surface flaws were eliminated in all the gauge lengths allowing volume flaws to dominate the fibre's failure. Impregnated bundles were found to have similar mean diameters across the set of gauge lengths. The weak-link property was found to be present and the data were found to comply with the Weibull model. The strength of the impregnated bundles were found to be much stronger than even that of the supported single fibres. In fact their values for the four lengths were found to be relatively uniform suggesting that impregnated bundles must fail through the interaction of a group of fibres rather than an isolated first fibre failure.