Affiliations: Structural Dynamics Research Group, Aerospace &
Mechanical Engineering Department, University of Liege, Chemin des Chevreuils,
Belgium
Note: [] Corresponding author: J.-C. Golinval, Structural Dynamics
Research Group, Aerospace & Mechanical Engineering Department, University
of Liege, Chemin des Chevreuils, 1 B52/3, B-4000 Liège 1, Belgium.
E-mail: JC.Golinval@ulg.ac.be
Abstract: In the field of structural health monitoring or machine condition
monitoring, most vibration based methods reported in the literature require to
measure responses at several locations on the structure. In machine condition
monitoring, the number of available vibration sensors is often small and it is
not unusual that only one single sensor is used to monitor a machine. The aim
of this paper is to propose an extension of fault detection techniques that may
be used when a reduced set of sensors or even one single sensor is available.
Fault detection techniques considered here are based on output-only methods
coming from the Blind Source Separation (BSS) family, namely Principal
Component Analysis (PCA) and Second Order Blind Identification (SOBI). The
advantages of PCA or SOBI rely on their rapidity of use and their reliability.
Based on these methods, subspace identification may be performed by using the
concept of block Hankel matrices which make possible the use of only one single
measurement signal. Thus, the problem of fault detection in mechanical systems
can be solved by using subspaces built from active principal components or
modal vectors. It consists in comparing subspace features between the reference
(undamaged) state and a current state. The angular coherence between subspaces
is a good indicator of a dynamic change in the system due to the occurrence of
faults or damages. The robustness of the methods is illustrated on industrial
examples.
