Statistical Characterization of Unidirectional Tensile Strength of FRP Composites

Fiber reinforced polymer (FRP) has been extensively used in civil infrastructure in recent decades. The uncertainty of its unidirectional tensile strength is of great importance for analysis and design with respect to structural safety. In this work, a large database comprising 40 data sets of carbon- and glass-fiber reinforced polymer tensile tests is fitted by the Normal, Lognormal and Weibull distributions. The observed significance level (OSL) which is based on the Anderson-Darling statistic is used for determining the goodness-of-fit. Fitting results show that all three distributions can be used for the tensile strength of FRP composites from the perspective of experimental justification. However, the Weibull distribution is preferred as it reveals the weakest link hypothesis of failure and is the most commonly used distribution for composites.