|
|
Sensitivity analysis of stability problems of steel structures
Valeš, Jan ; Vičan,, Josef (referee) ; Kučerová,, Anna (referee) ; Melcher, Jindřich (referee) ; Kala, Zdeněk (advisor)
The doctoral thesis is focused on evaluation of global sensitivity analysis of load-carrying capacity of steel hot-rolled beams. These beams are subjected to lateral-torsional buckling, weak axis buckling and strong axis buckling. Very comprehensive computational models which were both geometrically and materially nonlinear were created in Ansys software using solid finite elements to calculate the load-carrying capacity. The computational models allowed modelling of random initial imperfections such as initial curvature, deviations of cross-section dimensions and steel properties. Sensitivity analysis quantified their influence on the load-carrying capacity. Simulation runs of random imperfections were generated using the Latin Hypercube Sampling method. Since the evaluation of sensitivity analysis of load-carrying capacity of all finite element models would cost an extreme amount of computer time, the thesis aimed at developing a meta-model (also known as surrogate model) based on approximation of FEM model. The approximation polynomial then facilitated the evaluation of sensitivity indices using a high number of simulation runs. At the end, the relationships between the slenderness and the first and second-order sensitivity indices are plotted in graphs. Those random input imperfections that influence the variability of load-carrying capacity the most are pointed out.
|
|
|
Sensitivity analysis of stability problems of steel structures
Valeš, Jan ; Vičan,, Josef (referee) ; Kučerová,, Anna (referee) ; Melcher, Jindřich (referee) ; Kala, Zdeněk (advisor)
The doctoral thesis is focused on evaluation of global sensitivity analysis of load-carrying capacity of steel hot-rolled beams. These beams are subjected to lateral-torsional buckling, weak axis buckling and strong axis buckling. Very comprehensive computational models which were both geometrically and materially nonlinear were created in Ansys software using solid finite elements to calculate the load-carrying capacity. The computational models allowed modelling of random initial imperfections such as initial curvature, deviations of cross-section dimensions and steel properties. Sensitivity analysis quantified their influence on the load-carrying capacity. Simulation runs of random imperfections were generated using the Latin Hypercube Sampling method. Since the evaluation of sensitivity analysis of load-carrying capacity of all finite element models would cost an extreme amount of computer time, the thesis aimed at developing a meta-model (also known as surrogate model) based on approximation of FEM model. The approximation polynomial then facilitated the evaluation of sensitivity indices using a high number of simulation runs. At the end, the relationships between the slenderness and the first and second-order sensitivity indices are plotted in graphs. Those random input imperfections that influence the variability of load-carrying capacity the most are pointed out.
|
guest ::
