Using submodels for a probabilistic nonlinear analysis of corroded RC-structures

Probabilistic analysis is best suited to cover a wide range of concrete and steel properties including corrosion effects and its influence on structural behaviour. If a reduction of the computational time to calculate sufficient samples is needed, a use of submodels might be very useful. The general idea is to compute very accurately only a relatively small section that is of interest to capture the nonlinear steel properties including any localized corrosion and its bond with the concrete. The rest of the structure is considered either with simple FE elements or with an analytical solution. The procedure was verified by large scale tests on cantilever walls with artificially induced corrosion and with a conventional nonlinear FE analysis. The presented method combines the advantages of a detailed and complex nonlinear FE analysis with the applicability and performance of an analytical solution or simple FE calculation. The advantages, accuracy and limitations of the method are broadly discussed, and the field of application is described.