Application of Frequency Domain Decomposition Identification Technique to Half Spectral Densities

Because of its simplicity and robustness, the Frequency Domain Decomposition (FDD) identification technique have become very popular in the operational modal analysis community. The basic idea behind this technique consists of computing the singular value decomposition of the power spectral densities estimated with the periodogram (also known as “Welch’s” periodogram) approach to identify the natural frequencies and mode shape vectors. In this paper, the benefits of the application of the FDD technique to half spectral densities - the power spectral densities estimated from the positive part of the correlation functions - are investigated. In order to illustrate such benefits from a practical perspective, the FDD identification results obtained from the half spectral densities, of both simulated and real structures, are compared to those from the classical periodogram-driven FDD.