Damage Detection of a large space structure using a model obtained by Modal Coupling

In the first part of the work, the model of a large space structure is created, starting from the finite element models of its components. The presented results refer to an experiment carried out in 1992 at the Politecnico di Milano, but the same method has been used with great success in many other instances in which building the entire full scale model of a structure was not feasible. The matter at hand fits into the broader cathegory of substructuring, and a brief overview of other methods is also given. Having chosen to implement modal coupling through the use of fictitious masses, the numerical version of the boundary masses used in the experiments, the technique is explained in detail. Before entering the specifics, and the results, it is appropriate to remind that the connection between the substructures, or each substructure and the fictitious mass, is always modeled as isostatic – this is not the only way. Then, examples are provided in order to give the taste of the power of performing modal coupling with this architecture. In this sense, a parametric study is carried out on the fictitious mass “shape”, in order to show the benefits coming from the use of this device, and on the size of the whole process, depending on which the results can differ.
The second part of the work deals with a problem which is becoming more and more important as structures increase in size, and operational times increase: damage detection. This can be undertaken resorting to different techniques, lately non-destructive are improving the level of performance. Dynamic tests, although, can still be a valuable help, and we will concentrate on these. The hypothesis we make is that the damage is located in a precise area of the structure, and the algorithm we want to use is a least-squares approximation based on two different indicators. The first kind of indicators we use are the changes in the natural frequencies values brought about by the damaging of the structure, while the second group are the transfer functions values close to the resonances that show more evident differences with respect to the ones obtained from the undamaged structure. The obtained results highlight the difficulty of the analysis we set out for. First, using the model devised in the first part constitutes a problem, in that even if the approximation is very accurate, the differences brought about by the damages are not always distinctively significant. Second, depending on the parameters we use for damage detection, the results can differ quite significantly, because the damage introduced alters the particular parameter in a specific way.