학술논문

Multifrequency Excitation and Support Vector Machine Regressor for ECT Defect Characterization
Document Type
Periodical
Source
IEEE Transactions on Instrumentation and Measurement IEEE Trans. Instrum. Meas. Instrumentation and Measurement, IEEE Transactions on. 63(5):1272-1280 May, 2014
Subject
Power, Energy and Industry Applications
Components, Circuits, Devices and Systems
Probes
Coils
Software
Data mining
Feature extraction
Reliability
Materials
Defect characterization
giant magnetoresistance (GMR) sensor
multifrequency eddy current testing (ECT)
nondestructive testing (NDT)
signal processing
support vector machine
Language
ISSN
0018-9456
1557-9662
Abstract
Eddy current testing (ECT) has three main tasks: detection, location, and characterization of defects. The characterization task, which means the ability to find the geometrical characteristics of the defect, is still in the research domain, although in many industrial applications this task has to be carried out with good accuracy to allow reliable acceptance or rejection decision indispensable to save costs and even human lives. This paper proposes an ECT measurement method that allows the reliable estimation of the geometrical characteristics of thin defects (i.e., length, height, and depth) by using a combination of a multifrequency excitation and an optimized support vector machine for regression. The proposed solution is tested on real specimens with known cracks by using a suitable measurement setup comprising a giant magnetoresistance-based biaxial ECT probe.