A feature extraction technique based on principal component analysis for pulsed Eddy current NDT
Introduction
Eddy current sensors are generally robust and small in size [1]. Due to robust to the medium, non-contact Eddy sensor can be used for precision engineering to displacement and geometric measurement by using novel sensor design and blind signal separation [1], [2], [3], [4], [5], [6]. By employing advanced signal processing techniques, the measurement resolution for displacement can be down to 0.1 μm [2], [5]. The resolution varies for different target materials, and lift-off can be compensated by the nominal gap. Eddy current sensors have been widely used for non-destructive evaluation (NDE) [7], crack detection in particular. However, the measurement accuracy and resolution is much behind than the surface geometric measurement because of the feature of subsurface ‘blinding’ measurement [4].
Conventional Eddy current techniques use single frequency sinusoidal excitation and detect flaws as impedance or voltage changes on an impedance plane display with inspectors interpreting magnitude and phase changes. However, these techniques are sensitive to a variety of parameters that are inherent in the flaws [3]. Additionally multiple frequency measurements have been combined to provide a more rigorous assessment of structural integrity by reducing signal anomalies that may otherwise mask the flaws [8].
Pulsed Eddy current (PEC) sensing is a new and emerging technique [9] that has been particularly developed and devised for subsurface crack measurements, with some success at Iowa State University in USA [10],1 DERA in UK [11] and the Cegely Laboratorium in France [8]. PEC techniques excite the probe's driving coil with a repetitive broadband pulse, usually a square wave. The resulting transient current through the coil induces transient Eddy currents in the test piece, which are associated with highly attenuated magnetic pulses propagating through the material. The probe provides a series of voltage–time data pairs as the induced field decays, and since the produced pulses consist of a broad frequency spectrum, the reflected signal contains important depth information. Physically, the pulse is broadened and delayed as it travels deeper into the highly dispersive material, and flaws or other anomalies close to the surface affect the Eddy current response earlier than deeper flaws. Peak values and peak times have been used for flaw detection and identification. However, these systems cannot currently be readily used because of difficulties in calibration and the lack of suitable response signal processing algorithms [12].
PCA is a useful statistical signal processing technique to reduce the dimensionality of datasets for compression, pattern recognition and data interpretation [13]. Other reported uses of PCA include classification of petroleum products and fruit based on the near-infrared spectra [14]. With the increase of computing power, advanced signal processing has been more widely used for computer science and engineering, particularly pattern recognition and artificial intelligence [15]. In this paper, we will apply PCA-based signal processing for PEC sensor systems to extract robust features, which is extended from our other work in computer vision, for defect detection and identification. The rest of the paper is organized as follows. Section 2 discusses feature extraction for PEC sensors by using PCA; Section 3 introduces the experimental set-up; 4 Results, 5 Conclusions present the experiment results and the conclusions.
Section snippets
Feature extraction for pulsed Eddy current sensors
In non-destructive testing (NDT), it is desired to be able to detect, classify and quantify defects that may occur in metal structures, such as aircraft lap joints. In such structures, defects mainly take place as surface cracks, sub-surface cracks and hidden corrosion. To answer this requirement, we have developed a PEC system based on Hall-effect device and a new feature extraction technique. In theory, PEC has the potentials to give more information than single frequency sinusoidal Eddy
Experimental set-up
For testing purposes we have prepared two aluminum samples. One is to test thickness variation, which can be used for metal loss simulation, and the other for surface and subsurface defects detection and quantification. The samples can be seen in Fig. 2, Fig. 3. For sub-surface crack simulation, we place the probe next to the surface of the sample shown in Fig. 3, and for surface-crack we place it next to the other side.
Results
In this test, we are evaluating the performance of both feature extraction techniques, PCA and peak characteristics in time domain. The samples mentioned in Section 3 are used. The samples provide three classes of defects, namely surface cracks (four different sizes), sub-surface cracks (four different sizes) and sub-surface metal loss (nine different sizes), with assumption that the normal, defect-free Al slab has a depth of 10 mm. For each defect, eight measurements are taken with slightly
Conclusions
A new PCA-based feature extraction method for PEC NDT has been developed and investigated. The method reduces the dimensionality of the response signals and extract relevant features, which allow effective classification of defects. The performance of classification has proven to be better than the conventional method using the response peak characteristics. By extracting principal components with highest variances, the effect of noise has also been reduced as data that contained in the n first
Acknowledgements
The authors would like to thank TWI Ltd and DfES and the University of Huddersfield for funding the research work.
References (15)
- et al.
A miniaturized displacement sensor for deep hole measurement
J Precision Engng Am Soc
(1999) - et al.
Real-time classification of petroleum products using near-infrared
Comput Chemical Engng
(2000) - et al.
The research of inhomogeniety in Eddy current sensors
Sensors Actuators A
(1998) Frequency output Eddy current sensors for precision engineering
INSIGHT
(2001)- et al.
Computational algorithms for linear variable differential transformers (LVDTs)
IEE Proc: Sci, Measurement Technol
(1997) - et al.
Blind sensing
IEE Manufact Engr
(1997) - Tian GY, Zhao ZX, Baines RW. Precision measurement using an Eddy current sensor device. Proceedings of Twelfth National...
Cited by (302)
Optimal feature subset deduction based on possibilistic feature quality classification and feature complementarity
2024, Expert Systems with ApplicationsDevelopment and fusion of NDT classifiers for defect detection on underwater structures
2024, NDT and E InternationalAn improved POD-Galerkin method for rapid prediction of three-dimensional temperature field for an IGBT module
2024, International Communications in Heat and Mass TransferDetection of fatigue degradation in austenitic stainless steel with eddy current probe and machine learning
2023, Journal of Materials Research and TechnologyEnhancing the accuracy of metocean hindcasts with machine learning models
2023, Ocean Engineering