BLIND SOURCE SEPARATION TECHNIQUES FOR MODAL ANALYSIS: EXPERIMENTAL DEMONSTRATION
Fabien Poncelet, Gaëtan Kerschen, Jean‐Claude Golinval, D. Verlest · Open Repository and Bibliography (University of Liège) · 2007
The present study carries out output-only modal analysis using two blind source separation (BSS) techniques, namely independent component analysis and second-order based identification. It is shown under which circumstances the normal coordinates of the vibration modes may be interpreted as virtual sources, which renders the application of these BSS techniques possible. To support the theoretical findings, numerical experiments are first carried out using a discrete system of three masses. Secondly the method is applied to a real-life structure which is a blade extracted from a turbojet engine stator. The results are compared with the stochastic subspace identification method.