Information-divergence based methods for acoustic micro-defect identification in materials

Result code in IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F68407700%3A21340%2F11%3A00329602" target="_blank" >RIV/68407700:21340/11:00329602 - isvavai.cz</a>

Alternative codes found

RIV/68407700:21340/11:00187626

Result on the web

—

DOI - Digital Object Identifier

—

Result language

angličtina

Original language name

Information-divergence based methods for acoustic micro-defect identification in materials

Original language description

We deal with the diversity of acoustic emission sources in materials through signal processing methodology for the data sets detected by the measurement device Dakel Xedo 5. Applying the following methods of Fuzzy Clustering (FC), Model-Based Clustering (MBC) and Support Vector Machines (SVM) in combination with empirical nonstandard signal and spectrum attributes, we arrive to the efficient source separation technique. These methods belong to fundamentally different groups. The FC is based on the optimization of an objective function. The MBC consists of two parts - the Agglomerat.ive Clustering and the iterative EM algorithm [1] minimizing likelihood function of the statistical model under consideration. Finally, the SVM [2] searches for optimal separating liyperplanes between clusters. The signals are compared by means of suitable parameters obtained directly from detected signals or their normalized spectral density estimates. We also use distinctive (Φ-divergence dista nee measures [3] between signal spectra as the additional attribute for the signal cluster separation. The structure of Φ-divergeuce parameter is given in discrete form by DΦ(S̃ refer, S̃)= ∑ T-1 f=0 S̃(f)Φ (S̃ refer(f)/ S̃(f)), where S̃ refer is a normed reference spectrum from the number m of samples of a given signal type and S̃ stands for the normalized spectrum of individual signals.

Czech name

—

Czech description

—

Type

D - Article in proceedings

CEP classification

—

OECD FORD branch

20301 - Mechanical engineering

Project

—

Continuities

S - Specificky vyzkum na vysokych skolach

Publication year

2011

Confidentiality

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Article name in the collection

Proceedings of Forum Acusticum 2011

ISBN

978-84-694-1520-7

ISSN

—

e-ISSN

2221-3767

Number of pages

6

Pages from-to

991-996

Publisher name

European Acoustics Association

Place of publication

Madrid

Event location

AALBORG

Event date

Jun 26, 2011

Type of event by nationality

WRD - Celosvětová akce

UT code for WoS article

—