Enhanced mode shape estimation in multi-dataset OMA using frequency domain decomposition

The identification of the global mode shapes in multi-dataset vibration tests with the classic Frequency Domain Decomposition is carried out by identifying each dataset individually and by merging the individual mode shape parts with aid of the reference mode shape components. It turns out that, in case of closely spaced modes, the mode shape parts in the different datasets are computed from subspaces that are affected differently by the closest modes. In this circumstance, the global mode shape vectors formed by merging all the mode shape parts might not yield well-defined global shapes. In other to overpass this issue, another strategy is proposed in this paper to force the mode shape parts corresponding to each dataset to be in the same subspace, and thus, yield global mode shapes with clearer configuration. The efficiency of the proposed strategy is demonstrated using an application example in the final part of the paper.