Reduction of large frequency response function data sets using a robust singular value decomposition

In several mechanical engineering applications high spatial resolution frequency response function (FRF) measurements are required. Adapted optical measurement instruments like the laser scanning Doppler vibrometer (SLDV) exist to perform this task. The result of this high spatial resolution measurement is that a large amount of data is available. The processing of this data – i.e. extracting modal parameters from the FRFs – can be a time consuming task. On the other hand, when measuring on real-life structures, a portion of the FRFs is corrupted with high levels of noise (this is caused by laser drop outs which result in outliers in the measurements). In this article a data reduction method will be introduced which can be used to reduce the amount of data in order to limit the computation time. The method is based on a robust singular value decomposition (SVD). In contrast to existing SVD based techniques, outliers in the data can be handled efficiently. A validation of the technique is performed both on a simulation and on scanning laser vibrometer measurements of a car door and a circuit board.