Verification of shrinkage curvature code prediction models

An attempt is made to theoretically and experimentally verify the shrinkage curvature models presented in Eurocode 2 and BS 8110. These codes claim that the models originally derived and proven for uncracked sections are suitable, with modification, for predicting the behaviour of cracked sections, although this claim has never been proven experimentally. To achieve verification, an alternative theoretical approach is initially proposed in this paper. In this theoretical model, the effect of shrinkage, creep and the variation in the neutral axis position of the section are taken into account. The stresses developed in the steel and concrete at a cracked section according to this theoretical model are then applied to a finite-element (FE) model representing a portion of the beam from the crack to mid-way between the crack and an adjacent crack. Ultimately, the mean curvature is determined. Experimentally, pairs of beams were cast and subjected to a level of flexural loading to produce a stabilised crack pattern in the constant-moment zone. The behaviour of the beams was monitored for up to 180 days. For any pair of beams, one beam was cast using a high-shrinkage concrete and the other with a low-shrinkage concrete. Each concrete type, however, exhibits similar creep. Therefore, shrinkage curvature can be obtained by subtracting the long-term movements of one beam from the other. These experimentally defined curvatures were compared with the mean curvatures obtained from the FE analysis. The comparison showed reasonable agreement. The curvatures were also compared with uncracked and cracked curvatures predicted by the codes. The curvatures derived in this investigation fell within the boundaries of the uncracked and cracked curvatures predicted by the codes and, for the fully cracked case, the curvatures were closer to the uncracked boundary.