Change detection under environmental variability
OPEN ACCESS
Loading...
Author / Producer
Date
2014-09
Publication Type
Conference Paper, Conference Paper
ETH Bibliography
no
Citations
Altmetric
OPEN ACCESS
Data
Rights / License
Abstract
Change detection is the process of announcing, from inspection of the sequence of measured signals, if changes in a system have taken place. Change detection is tantamount to a hypothesis test on the likelihood of the aggregate of the data given a statistical model of reference state. A fundamental assumption in change detection is that the data points are independent and identically distributed (i.i.d) under the null hypothesis. When the data depends on environmental parameters, the i.i.d premise for a sequential stream of points is not justified. However, operations on the data can be performed, so the premise is reasonably satisfied, and the change detection scheme can then be applied. Among alternatives that do not require that the environmental parameters be measured, the one that has been most widely examined is Principal Component Analysis. In this case, the data is projected into a subspace where the variance from the environment is slight. The price paid is that the damage-related information on the principal directions is then discarded. There are (at least) two other possibilities: one is to decorrelate the data using an autoregressive model, and the other is to break the correlation by scrambling. In the latter case, the price is that detection must be carried out with some irreducible delay. This paper examines the relative merit of the alternatives noted.
Permanent link
Publication status
published
External links
Editor
Book title
5th International Operational Modal Analysis Conference (IOMAC 2013). Proceedings of a meeting held 13-15 May 2013, Guimaraes, Portugal
Journal / series
Volume
868
Pages / Article No.
1 - 8
Publisher
International Operational Modal Analysis Conference (IOMAC); Curran
Event
5th International Operational Modal Analysis Conference (IOMAC 2013)
Edition / version
Methods
Software
Geographic location
Date collected
Date created
Subject
Organisational unit
09469 - Kaufmann, Walter / Kaufmann, Walter