Condensation of the correlation functions in modal testing. (1st March 2019)

• Record Type: Journal Article
• Authors: Olsen, Peter; Juul, Martin; Brincker, Rune
• Is Part Of: Mechanical systems and signal processing. Volume 118(2019)
• Page Start: 377
• Page End: 387
• Publication Date: 2019-03-01
• Subjects: Operational modal analysis -- Condensation -- Correlation function -- Time domain identification -- Singular value decomposition -- Principle component analysis; Structural dynamics -- Periodicals; Vibration -- Periodicals; Constructions -- Dynamique -- Périodiques; Vibration -- Périodiques; Structural dynamics; Vibration; Periodicals; 621
• DOI: 10.1016/j.ymssp.2018.08.036
• Languages: English
• ISSNs: 0888-3270

Abstract

Highlights: The theory of condensation based on the principle component analysis is presented. The issue of over fitting is overcome by transforming to a new set of pseudo coordinates. The effect of noise is reduced by removing the null-space of the singular values. Normalization by the use of the singular values enhance the physical information of the signal. Abstract: A well known issue when performing Operational modal analysis in the time domain is the challenge of choosing the right order of the system matrix. Setting the order higher than the rank of the system results in over fitting, i.e. having more poles to place than modes of the system. This results in fitting poles to noise or non linearities of the system. In This paper a method to avoid over fitting by reducing the physical channels to a reduced number of pseudo channels condensing the relevant physical information and leveling the level of energy in each channel is suggested.