Diseno optimo de uniones semirrigidas mediante simulacion numerica y modelos Kriging.

摘要

In the last few years an important amount of research works have been carried out to develop strategies for the optimum design of steel structures with semirigid joints, where joints are modeled like elastic elements, through a moment-rotation curve obtained by available simplified models in the literature (e.g. empiric models). The main trouble of these models is that strength and rotational stiffness of the joint do not match exactly with obtained values from a real joint. One way of solving this trouble is to use more detailed models of the joint (e.g. the components method and numerical models), what allows a better relationship between the properties of the resistant elements (end plates, bolts, etc.) and the strength and stiffness of the joint.;In this thesis, a numerical model of detail is introduced, based on the finite element method, to model bolted extended end-plate beam-to-column joint. This model has been checked through experimental tests collected in the literature, showing that one accurately reproduces the real behaviour of the joint.;In order to reduce the computational cost that involves to use numerical techniques of optimization coupled directly to an analysis by finite elements, a methodology for optimum design of semirigid joints is introduced, by means of integration of the analysis by finite elements, metamodels and optimization techniques. The metamodel is built starting from of a group of points of the design space and a Kriging model. Those points are generated through a Latin Hypercube sampling procedure and analyzed by the developed numerical model.;The proposed methodology provides designer a tool to obtain the optimum design of a joint in a quicker way than using the numerical model. Design variables are location and diameter of bolt and dimensions of the end plate. The objective function is to maximize either strength or stiffness or minimize cost of the joint. The constraints can be maximum and minimum values of the geometric parameters of the joint (according to EN 1993-1-8:2005), the strength and/or stiffness of the joint. The used optimization algorithms are mathematical programming and genetic.;The methodology has been proven through several examples. The results demonstrate the efficiency of the procedure, allowing to obtain a detailed optimum design in a reasonable time.;Finally, design abaci are introduced which relate resistant moment and stiffness of the joint. Those abaci allow the designer to obtain, for different design criteria, the optimum configuration of the joint in a quick and simple way.