Size effect in resin/glass composite flexure strengths

The objective of this study was to test the hypothesis that composite restorative materials possess an elastic-brittle nature and therefore will exhibit a size effect for flexure strength data. The experimental material consisted of 20 wt% 60:40 BISGMA:TEGDMA, 10 wt% colloidal silica, and 70 wt% Sr glass and was cured by light irradiation. Two sizes of flexure specimens were fabricated: 3.2 x 1.6 x 35 mm, and 6.25 x 3.1 x 35 mm. Half of the specimens made were soaked to equilibrium weight gain in 50:50 ethanol:water. The fracture strengths were measured in four-point bending tests. The beams under load were modelled by the finite element package ABAQUS. A statistical fracture mechanics methodology embodied in a public domain computer program called CARES/LIFE, developed by NASA, utilized the ABAQUS input and the fracture strengths of the smaller specimens to predict the fracture strengths of the larger specimens. In making the computation it used an approach that combines a Weibull distribution of flaw size with Batdorf's fracture mechanical model for failure at a material flaw. Both the soaked and unsoaked specimens exhibited Weibull behaviour, with shape parameters ranging from 4.04 to 8.15. Soaking had a clearly detrimental effect on the strengths of specimens of both sizes, and produced a comparable percentage reduction in the estimated scale parameter of the fracture strength distribution. Both the soaked and unsoaked specimens also exhibited a clear and comparable size effect, i.e. the larger specimens had a fracture strength that was lower than that of the smaller specimens by roughly the same percentage. Moreover, the magnitude of the size effect was well predicted by the CARES/LIFE methodology for both the soaked and the dry specimens. The elastic-brittle character of both soaked and unsoaked composite specimens was validated by load-deflection data, the magnitude of the Weibull shape parameters of the observed-fracture strength data (<10), and the observed effect of specimen size. The accuracy of CARES/LIFE in predicting the magnitude of the observed size effect in beams of two different sizes strongly suggests that CARES/LIFE will be useful for computation of failure probabilities for clinically relevant structures.