Optimization of Vierendeel-type steel structure by means of nonlinear numerical analyses

The development of technology combined with the improvement and upgrade of the software algorithms used by engineers, provides the opportunity of investigating the behavior of a construction under any conditions with very good accuracy. In the present thesis, the object is to empirically investigate the margin for optimizing the plate thickness of the steel parts and the connections of a Vierendeel beam (upper chord, lower chord and vertical members). The results of this work can form the basis for the creation and application of optimization algorithms that will have the advantage of automation.

Initially the model is designed with beam elements, in order to select the dimensions and thickness of the cross sections based on the deflection displacement of the structure. The cross sections are hollow square, and after checking the results at the level of stresses and deflection displacement, then a second model of the same structure with shell elements is drawn. In the shell-element model, cross sections are designed with corners and not with curved sections. This simplification implies greater concetration of tensions at connection points of vertical members with chords. However, this small difference does not significantly affect the conclusions in the end. For this purpose, initially the analyses carried out are linear and the areas-zones that are candidates for optimization are being selected. An overview of the strain and stress distribution on the structure is taken by the results of the linear analysis. Furthermore, some zones can be accordingly defined around the connections, which may be selected to optimize their thickness. Eventually, by performing non-linear analyses, some models get to emerge which have a different thickness in some zones compared to the original model. The results of the non-linear analyses lead to conclusions about the type of optimization that can be implemented. The optimization of cross-sections aims to lead to a model that behaves similar to the designed one, using as much less material as possible.