• Automated Rack-Supported Warehouse
• Push-over analysis
• Reduced-order model
• Seismic porformance

The present thesis is part of a research activity carried out mostly at the National Technical University of Athens in the context of the Erasmus European exchange programme and concluded at the University of Pisa. The main objective of this research was the assessment of the seismic performance of an Automated Rack-Supported Warehouses, in particularly one of the study cases to be examined in the European research project “STEELWARE”.
ARSWs represent the evolution of the traditional racking system. Such structures are designed to provide the necessity to store a huge amount of goods improving the efficiency in storing and therefore, providing substantial savings in terms of cost. ARSWs, despite their lightness, can reach considerable heights bearing very high loads, larger than their dead load and opposite to what happens in usual civil engineering structures. Furthermore, they are also designed to support the weight of all the non-structural elements (clad, roof, technological facilities…). Thus, standard design approaches are not applicable to this type of structures, especially when we are considering seismic and wind actions. In this context, the final aim of the STEELWARE project is to propose the guidelines to design such structures increasing ARSW safety, reliability and economy.
Moreover, due to their structural configuration (cold formed profile, open sections, non-linear behaviour of the main connections, …), and due also to the fact that highly sophisticated machines need to run along loading-unloading isles to store and retrieve goods in a completely automatic way, for the design of the ARSWs no margin of mistake is allowed. By examining all of those problems and considering that the ARSWs consist of hundreds or thousands of steel members and nodes connected to each other, it became clear that the finite element analysis for these structures is a demanding work; the problem arises considering nonlinear phenomena i.e. material and geometric nonlinearity, that may lead to prohibitive analysis costs in terms of time and CPU or even convergence and stability problems. The purpose of the thesis was to develop a computationally simpler model called reduced-order model, in terms of degrees of freedom, in order to check their nonlinear behaviour under seismic actions but checking to reach a balance between costs and accuracy.
The reduced-order model was developed by substituting the upright frames and the roof truss of the detailed model with simpler beams. Assigning to the beams the same properties of the detailed model, it must be considered the shear deformation of the uprights in contrast to what happens in typical columns; Euler-Bernoulli beam theory cannot be used and Timoshenko beam theory must be considered, determining suitably linear properties (A, I, Aeff) depending on the different behaviour that various type of the bracing elements configuration, of the upright frames, have.
Apart from the modelling, first of all a modal analysis has been performed for both models and the reliability of the reduced-order model in the elastic region was assessed; using the results of the modal analysis, the stresses on all the elements of each upright frame has been estimated. Examining these results, it is clear that a good accuracy (with differences of about 10% maximum) has been reached gaining a reduction of about 55% in terms of DOFs. Later, a push-over analysis was performed, taking into account the material and geometric nonlinearity in order to check the reliability of the reduced-order model also in the inelastic region; the push-over curves of both models, show a similar behaviour.
Some considerations about the 3D model, have been moved ahead showing a much bigger reduction in terms of DOFs. In this context, one of the intentions about desirable future works, is to do better evaluation of the 3D model, checking if it is possible to have the same accuracy, with a better reduction in terms of CPU costs and including much more information in the 3D model instead of a two dimensions.