Lift-off nulling and internal state inspection of multi-layer conductive structures by combined signal features in pulsed eddy current testing

Pulsed eddy current (PEC) testing, as an emerging technique of eddy current testing, has been used for detection and analysis of internal states of multi-layer conductive structures. However, the lift-off noise, caused by irregular sample surface, varying coating thicknesses or movement of probes, m...

Full description

Saved in:
Bibliographic Details
Published in:Nondestructive testing and evaluation 2018-07, Vol.33 (3), p.272-289
Main Authors: Huang, Pingjie, Luo, Xuwei, Hou, Dibo, Yang, Zhaohe, Zhao, Ling, Zhang, Guangxin
Format: Article
Language:English
Subjects:
Accuracy
Bimetals
Collinearity
combined signal features
Conductivity
Eddy current testing
Eddy currents
Inspection
internal states of multi-layer structures
Lift
lift-off noise
Liftoff
Mathematical models
Multilayers
Noise
Pulsed eddy current
Support vector machines
Variation
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Pulsed eddy current (PEC) testing, as an emerging technique of eddy current testing, has been used for detection and analysis of internal states of multi-layer conductive structures. However, the lift-off noise, caused by irregular sample surface, varying coating thicknesses or movement of probes, may seriously influence the accuracy. Therefore, in order to improve the detection accuracy under lift-off noise, an algorithm is proposed. Firstly, the response signals with lift-off noise are analysed theoretically and experimentally; secondly, according to analysis of the multi-collinearity between the lift-off and the response signal, a model to predict the lift-off is constructed by Partial Least Square (PLS); then, based on the prediction value of the lift-off by the PLS model, an approach to detect the internal states of multi-layer conductive structures is proposed. In this approach, the internal states are determined by the classification model, which is constructed by Support Vector Machine (SVM) based on combined features. The combined signal features are composed of the lift-off feature and principal component features. Finally, a bimetallic thermostat is taken as our experimental subject to validate the proposed method. The experimental results indicate that the proposed method can reduce the effect of lift-off fluctuations and improve detection accuracy.
ISSN:1058-9759
1477-2671
DOI:10.1080/10589759.2018.1428321