Efficient vibro-acoustic identification of boundary conditions by low-rank parametric model order reduction

A novel method is presented that detects the proper boundary conditions of a test setup in a short time period by combing numerical models with experimental data. This allows for detection and localization of possible anomalies in the assumed boundary conditions of the system. The method works by combining a low-rank parametric model order reduction technique with a model updating strategy, where the boundary conditions of a numerical finite element model are updated by using frequency response function data. This combination makes it possible to update a large amount of parameters, because the assumed low-rank nature of the changes enables the use of non-parametric model order reduction techniques for the calculation of the reduced basis. This is possible, because the system can be rewritten in such a way that the parameter dependencies only show up in the feedforward matrix of the system, thus no a priori sampling of the parameter space is required. Thus, the resulting model can identify a large amount of parameters, including the identification of local changes in the boundary conditions. The method is validated with a test-setup in which an aluminum plate is attached to an acoustic cavity and the boundary conditions are varied gradually, by removing the bolts that are clamping the plate. By applying the proposed model updating scheme to the rotational stiffness along the edge in combination with an additional damping term, it is shown that the proposed method can detect which bolts are removed and also leads to a good match in the frequency response functions. Moreover, it is shown that these results are achieved in only a few minutes, in contrast to the same procedure with full order models. (C) 2018 Elsevier Ltd. All rights reserved.