In order to perform a reassessment of existing structures or design new structures in a more efficient and economical way non-linear finite element (NLFE) models are used to take into account all non-linear behavior in reinforced concrete structures. The Model Code 2010 (fib, 2012) suggest several safety formats to perform non-linear finite element analyses with an intended reliability index β_R=3.04. The safety formats seems to be safe for statically determinate structures with a predictable failure mode. For instance a bending failure in a simply supported reinforced concrete beam. However it is unclear if the intended safety level is met for statically indeterminate structures. Since the redundancy of the structural system and the several possibilities to redistribute the internal forces could lead to unpredictable failure modes.

In this thesis three reinforced concrete portal frame designs are used to evaluate the probability of failure of the structural resistance determined with the safety formats i.e. the safety of the safety formats will be assessed. The ductility of the portal frame designs determines the degree of redistribution of the internal forces. Therefore this case study will focus on three portal frame designs with all a different ductile behavior. The global design resistance is determined according to the following safety formats: the global resistance factor method (GRFm), the estimation of a coefficient of variation (ECOV) method and the partial factor method (PFm). A comparison is made with the local design resistance of the portal frame designs according to the Eurocode 2 using partial safety factors.

The reliability level of the structural design resistance according to the least conservative safety formats is determined for each portal frame design. Several response surfaces were constructed and the first order reliability method (FORM) was used to determine the reliability level. In case of relatively low material and geometrical uncertainties the safety formats lead to a structural resistance that can safely be used i.e. the intended reliability level is met. However when the detailing uncertainties in concrete frame corners are relatively large these uncertainties should be implemented in the safety formats. Further research and more experimental results are needed to make a better estimation of the (model) uncertainties possible.

Finally, the used methods are evaluated and several comments are given on the difficulties of using those methods. Constructing a response surface in combination with the FORM in order to find the design point is definitely not an automated process as the mathematical procedures seems to indicate.