Across the Netherlands close to 70 prestressed concrete T-beam bridges with cast-in-between slabs and transverse prestressing built in the 60's and are still in service. The current code NEN-EN 1992-1-1+C2:2011 assesses the shear capacity more conservatively. In addition, the NEN-EN 1991-2+C1:2015 prescribes an increased traffic load. This results in this type of bridge not complying with the current codes and the shear capacities are considered insufficient. However, upon inspection these bridges do not show signs of distress. This suggests the presence of additional load-carrying capacity, which is not considered in the current Eurocode.

Non-linear finite element analysis (NLFEA) can be used to accurately approximate the structural behaviour. This includes yielding of steel, cracking and crushing of concrete, the development of alternative load paths as well as snap-back and snap-through behaviour. However, performing each load step requires a great amount of computational effort depending on the amount of degrees of freedom of the model. In FEA a continuous shape is divided into discrete elements which together form a mesh. These meshes can be volumes, surfaces or lines.

To describe the geometry of the prestressed concrete T-beam bridges with cast-in-between slabs either volume elements or multiple surface meshes in different planes are required. Using a mesh of solids would result in system with such a high number of degrees of freedom which might even exceed the available computational capabilities or result in a very long duration of the analysis at best. Using shell elements to construct the mesh reduces the number of degrees of freedom by at least two thirds.

In this thesis I investigate to which extent we can simulate the structural behaviour of a prestressed T-beam slab bridge deck using a non-linear finite element model with a 3D non-planar mesh of shell elements. The Vechtbrug bridge near Muiden was a bridge of this type. A team of researchers from TU Delft have performed several collapse tests on this bridge. This includes extensive measurements of all the experiments as well as material testing on concrete and steel samples. For my own research, a single case study is conducted by recreating collapse tests performed on the Vechtbrug in which both isolated beams and unmodified spans have been loaded past failure. The results of the material tests provide accurate material properties as input for my numerical models. The results of the collapse tests allow for verification and validation of the outcome of the performed finite element analyses.\\

The results of the numerical analyses show a close approximation of the true collapse load with an overestimation of 15\% for the isolated beam model and 12\% for the cooperative beams model. The deflection again is overestimated with 18 and 56\%. The deflection of the adjacent beams relative to the loaded beam is too low. The numerical model is underestimating the transverse load distribution by $\pm$ 25\% for the adjacent beams and $\pm$ 34\% for the beams adjacent to those. The Guyon Massonnet method was applied to estimate the transverse load distribution with the supplied material properties and including the two cross beams. By contrast, the results were an overestimation of approximately 70% for the immediate adjacent beams. In the third numerical analysis the complete bridge deck and ultimate limit state verification is performed by applying the prescribed traffic load with all safety factors applied. The bridge can withstand 234\% of the prescribed load which agrees with the lack of damage present on the Vechtbrug after experiencing over 50 years of traffic load.

The results show that a non-planar shell mesh can generate a realistic structural response considering the collapse load approximates the actual one found in the collapse tests. However, this is somewhat limited for decks consisting of multiple beams since the implementation of the transverse load distribution in the numerical model was inaccurate. The structural response of the structure was too ductile in the numerical analysis with the deflection being overestimated and the strain under the loading plate double the value of the collapse test. Both Mustafa and Ensink have performed a numerical analysis of the isolated beam model using a mesh of solid elements prior to my thesis work. The results of the isolated beam model match closely in both the results of the NLFEA performed by Mustafa and Ensink. The solid mesh does yield more realistic cracking patterns. The isolated beam model showed the required evidence to demonstrate the activation of arching action: An increase in the horizontal reaction force required to lateral restrain the beam with the bending crack occurring under the loading plate so the arch action phenomenon could be activated. In the complete deck model evidence of compressive membrane action in the transverse direction was detected. In both the complete deck models, evidence of lateral confinement was demonstrated, increasing the maximum compressive stress of the concrete.

Finally we can conclude that a NLFEA with a 3D non-planar mesh of shell elements yields accurate results when considering a single strip of the bridge deck. However, the model with a the mesh representing the complete bridge deck, the capacity of the transverse load distribution is underestimated and the structure shows overly ductile behaviour. The model is capable of including the load-carrying mechanisms arch-action, compressive membrane action and fixed boundary action as well as the effect of lateral confinement.