LAUSR

Abstract Accurate prediction of unidirectional (UD) fiber reinforced plastics' (FRP) mechanical properties can significantly shorten the time from material design to structural application. Data mining, as a promising alternative approach to the traditional analytical micromechanics model and finite element model, can overcome their limitations of lack of certain constituent parameters and failure to differentiate processings. In order to establish a reliable predictive model for UD FRP properties based on fiber and resin properties by means of data mining, this paper created a mechanical property database of 203 kinds of FRPs and their constituents. Statistical tools were then applied to build and test structure-property correlation models. Results suggest that only longitudinal tensile modulus and longitudinal tensile strength can be credibly predicted, whereas the prediction of other six properties was quite unreliable or even infeasible. Overall, the study revealed that data mining, limited by incompleteness of data and complicated failure mechanisms, can only partially replace the actual testing of UD FRP properties.