Computational modeling strategy of steel connections

Sammanfattning: When designing large complex steel structures numerical calculation methods are always used to some extent. The simulation technology of today is advanced and can with proper modeling technique provide detailed analyses which captures the structural response accurately. For most projects it is not reasonable to analyze a complete structure in exact detail and therefore simplified methods are used based on linear elastic theory since they provide conservative results. There is however a need for more detailed analyses for special parts of a structure that differs from the elementary cases described in Eurocode. A technique of how to establish a finite element model of the critical parts would save significant computational time and still present accurate results for many cases and will, therefore, contribute to a more efficient and careful design of structures. The goal for this study is to optimize the modeling process by creating a link between traditional beam models and more detailed modeling of steel connections. The modeling strategy includes whether force controlled or displacement controlled loading should be used, how boundary conditions should be set and how the extracted forces or displacements from the traditional beam model should be applied. The main focus for the model is to predict an accurate behavior with a conservative approach. In order to evaluate a modeling strategy two case studies were performed where several detailed models of the connection were created and benchmarked to a reference model, which consists of the whole structure combined with a detailed part of the connection. In order to evaluate the different models of the connection, output acquired from the reference model was extracted as well and inserted into the separate detailed models of the connection. If the model is reliable for the specific case, the simulations with the output from the reference model should provide an almost precise behavior. The use of output from the traditional beam model will decide if it will serve its purpose and whether it is possible to combine these two simulation methods or not. During the study it was found that the output extracted from the traditional beam model and the reference model differed significantly. The cause for this was deemed to be due to a difference in the stiffness of the joint between the beam model and the reference model. The section forces, displacements and rotations extracted from the beam model therefore do not provide identical behavior when inserted into a detailed analysis of a single joint. The use of force controlled loading generally generated more stable and conservative results compared to the models with displacement controlled loading. Even if the models with displacement control were found to provide almost perfect results with the correct input, they proved to be sensitive to differences in the input. The output extracted from the traditional beam model led to inaccurate bending compared to the reference model, if not boundary conditions that eliminated shear forces were present. The best models proved to be the ones where the forces were applied at a distance from the area of interest. Preferably with an external beam length applied to the model.