Author's Accepted Manuscript

Near electrical resonance signal enhancement (NERSE) in eddy-current crack detection

R. Hughes, Y. Fan, S. Dixon

NDT&E international independent nondestructive testing and evaluation

Ul rdsonies Ther Electromagnetics * Radiology • ma] • Signal & Image Praces&ia( Optical w

www.elsevier.com/locate/ndteint

PII: S0963-8695(14)00070-X

DOI: http://dx.doi.org/10.1016/j.ndteint.2014.04.009

Reference: JNDT1611

To appear in: NDT&E International

Received date: 5 September 2013 Revised date: 23 April 2014 Accepted date: 30 April 2014

Cite this article as: R. Hughes, Y. Fan, S. Dixon, Near electrical resonance signal enhancement (NERSE) in eddy-current crack detection, NDT&E International, http://dx.doi.org/10.1016/j.ndteint2014.04009

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Near Electrical Resonance Signal Enhancement (NERSE) in Eddy-Current Crack Detection

R. HUGHES1*, Y. FAN1 and S. DIXON12

Department of Physics1 and School of Engineering2 University of Warwick, Gibbet Hill Road,

Coventry, CV4 7AL, United Kingdom

An investigation was performed into the effects of operating an absolute eddy-current testing (ECT) probe at frequencies close to its electrical resonance. A previously undocumented defect signal enhancement phenomenon, resulting from associated shifts in electrical resonant frequency, was observed and characterised. Experimental validation was performed on three notch defects on a typical aerospace superalloy, Titanium 6Al-4V. A conventional absolute ECT probe was operated by sweeping through a frequency range about the electrical resonance of the system (1 - 5MHz). The phenomenon results in signal-to-noise ratio (SNR) peak enhancements by a factor of up to 3.7, at frequencies approaching resonance, compared to those measured at 1MHz. The defect signal enhancement peaks are shown to be a result of resonant frequency shifts of the system due to the presence of defects within the material. A simple, operational approach for raising the sensitivity of conventional industrial eddy-current testing is proposed, based on the principles of the observed near electrical resonance signal enhancement (NERSE) phenomenon. The simple procedural change of operating within the NERSE frequency band does not require complex probe design, data analysis or, necessarily, identical coils. Therefore, it is a valuable technique for improving sensitivity, which complements other ECT methods.

1 Introduction

Eddy-current testing (ECT) is a well-established non-destructive testing (NDT) technique, routinely implemented in industry for the inspection of safety-critical metallic components, because of its high sensitivity to small surface defects.

High-strength, low density superalloys are used frequently for many industrial applications, particularly in Aerospace [1]. The design and service lifetime of components is based on the assumption that the smallest defect that can be reliably detected by NDT techniques is present in the part. For this reason, research is generally focused on detecting smaller defects. Industrial eddy-current methods can reliably detect 0.75mm long (max 0.38 mm deep) surface-breaking cracks, but achieving greater sensitivity is hampered by poor signal-to-noise ratios (SNR) [2]. Conventional ECT inspections operate in a range between 100Hz and 1MHz [3], so as to avoid the detrimental effects of environmental noise and the instabilities of electrical resonance. However, superalloys typically have very low electrical conductivities, leading to relatively large electromagnetic skin-depths at these frequencies. As a result, conventional

* Corresponding author. Tel: 02476151778:, Email: Robert.Hughes@warwick.ac.uk

operation does not provide sufficient resolution to the smallest defects, for which higher frequencies must be used.

Higher frequency inspections are not without their problems. They suffer from a greater susceptibility to liftoff, and variable surface conditions due to machining features or conductivity changes associated with shot-peening or burnishing (common in many manufacturing processes) [4, 5]. This can lead to higher levels of background noise.

An effective approach for maximizing the SNR of any ECT defect inspection is to use signal averaging, but in an industrial environment this is not always possible or practical, where time and throughput are important financial considerations. Averaging will also have a negligible SNR improvement on materials that have random but coherent noise (i.e. grain structure). Whilst eddy-current arrays (ECAs) are becoming more commonplace in industrial use [6-8] because of their ability to inspect large areas very quickly, these advantages often come at the cost of resolution and sensitivity. Therefore, it is desirable to develop and implement techniques that can maximize the sensitivity of single coil and arrayed eddy current probes.

Many authors have implemented multi-frequency, data fusion techniques [9-11] to cancel out unwanted signals, such as liftoff and temperature variations, so as to improve the SNR. Although techniques such as these have had success in laboratory environments, their application in industry is limited by the longer inspection times, complex operation and sophisticated signal processing algorithms [2, 12, 13] required. As a result such techniques are not commonly used in ECAs.

Authors such as Owston [14], Liu [15, 16] and Ko [17, 18] have recognised the potential of measuring the shifting of electrical resonant frequency, due to changes in its environment, as a highly sensitive means of measuring proximity, surface roughness and surface conductivity variations. The significant power transfer and large rates of change around high quality factor resonance peaks in the electrical impedance of an eddy current coil and cable, offer an extended dynamic range of impedance measurements. Such advantages make operating at frequencies around resonance highly sensitive to even the slightest changes in an electromagnetically coupled system, but currently there is no documented account of authors exploiting resonance effects specifically for defect detection.

An investigation was performed on the effects of operating an absolute ECT probe at frequencies approaching and passing electrical resonance. The research presented in this paper documents the initial findings of the investigation and highlights the implications that the work will have on future defect inspection techniques.

2 Theory

2.1 Electrical Resonance

An eddy-current sensor is an electromagnetic inductor coil, connected to a current source via a coaxial cable. It can be very simply modeled as a parallel inductor-capacitor (LC) circuit with additional series resistive components [19] as shown in Figure 1. The inductive component represents the coil and the capacitive component represents the coaxial cable. Contributions to the capacitance from adjacent coil turns are considered negligible, compared to the dominant capacitive interactions within the coaxial cable.

Figure 1 - Equivalence circuit in air. Simplified electronic circuit for an eddy-current probe in free space with

a capacitive coaxial cable connection

For the case of high frequencies, the resistive component can be considered negligible compared to the inductive reactance component (R0 ^ wL0 ). As a result the total impedance of the system is equivalent

to the impedance between a parallel capacitor and an inductor. An expression can be obtained for Z0 by the summation of component admittance (Y = 1/ Z) and taking the reciprocal, such that [20],

Z = R0 + jwL0 1 0 2 . 1 + jwRoCo -w L0C0

Consequently, there will be a frequency at which the denominator of equation 1 becomes zero, and thus the impedance tends towards a maximum (Figure 2). The system is in a state of parallel LC electrical

resonance at the resonant frequency, w0 = 2nf0.

The impedance to the flow of current is a maximum at resonance, and thus the voltage across the coil required to drive the same current through the system must increase to a maximum. As the system goes through its primary resonance, the phase will change from a positive (predominantly inductive) regime, to a negative (predominantly capacitive) regime (Figure 2). Beyond resonance, the reactance is dominated by the capacitive effects in the coil [21], and the probe will cease to measure changes in the inductive component efficiently. Complex systems have multiple resonances occurring at higher frequencies as the electrical significance of the components resonating decreases.

O 2000

| 1000 Q.

CO I 50

O CL5 0

oq CD -C -50

-1-1-1-1-1-1— -1-1-1-

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6

1 I 1 1 I ! i i <

Inductive

i i i i i i N. Capacitive _

3 3.5 4 Frequency, MHz

Figure 2 - Electrical resonance in air. The magnitude and phase of impedance of an ECT coil as it passes through the resonant frequency, f0.

This model represents an ECT coil in air but when brought into the proximity of an electrically conducting material, electromagnetic coupling occurs between the coil and the material surface. The equivalence circuit can be extended to model the coupling interaction, and has been utilized by many authors [14, 16, 22].

2.2 Mutual Induction Model

When an inductor is brought into proximity of an electrically conducting material, it will electromagnetically couple to the eddy-currents generated in the material surface. Coupling to the surface of a material

changes the electrical properties of the probe such that the system can be approximated to an equivalent circuit containing components whose values change with the amount of coupling. The coupling interaction can be modeled by the transformer circuit diagram shown in Figure 3 where the eddy-currents are modeled are a passive series inductor and resistance (LR) circuit [14].

probe circuit (0) eddy-current circuit (e)

- ■ ^ (( + r )

Figure 3 - Mutual inductance transformer model. Simple transformer circuit model for an eddy-current probe (0) coupled to the surface of an electrically conducting material (e).

The impedance of the inductive arm of the system, Z1L , can be formulated using Kirchoff's laws [23].

Z1,L = R + >Lc(l - k2) + R2 + 2 L

Re + 0 Le

By approximating uniform magnetic field interactions, the resistive and inductive-reactance components of the secondary circuit are assumed to be equal in magnitude (wLe = Re) as shown by Wheeler [22].

The impedance of the inductive arm of the circuit, Z1L , of the circuit in Figure 3 therefore reduces to the

series combination of the probe coil impedance in air, Z0 L , plus additional series resistance and reactance of equal magnitude and frequency dependence.

Zi, l = R + j®Lo + 2 œL0k 2 (1 - J )

Zl, L =

+ 2 ®lok2 (i - j).

The result is consistent with the equivalence circuit proposed by Wheeler [22] for a uniform excitation field as shown in Figure 4.

Figure 4 - Coupled ECT system equivalence circuit, adapted from Wheeler [22]

As a result the circuit diagrams in Figure 3 and Figure 4 can be remodeled as a single equivalence inductive circuit with effective inductive ( L1 ) and resistive ( R1 ) components (Figure 5).

Z1L = Ri + >Li,

R1 = R +ÀR = R + - coL0 k

Figure 5 - Coupled eddy-current probe equivalence circuit with effective resistance and inductance.

In the same way as the free space case in section 2.1, the total impedance of the system is given by,

Ri + jvLi

1 + jœCQ R1 - (D2C0 L1

The effective resistance, or real component of the impedance of the system, now has frequency dependence, so it cannot be eliminated with traditional arguments. Instead the impedance must be fully expanded.

R + aL0 — + j®L0

' k 1 - — 2

k2 k2 1 + R + j®2C0L0 y - o2C0L0 (1 - —)

The general expression for the resonant frequency, (, but can be simplified for the specific case when the probe circuit has a large quality factor [14] i.e. R ^ (L0,

Co Lj1 - k2 + — C0 L0

r k2 1

C0 L0 1- y(l + j ) _

Typical values for k are rarely greater than 0.5 for surface coupling in which case equation 11 can be approximated to,

' k 2^' 1 - — 2,

The resonant frequency of the system is dependent on the coupling between the probe and the surface of the material [14]. An expression for the coupling coefficient can be derived from equations 2 and 11 as a

function of the ratio between the resonant frequency of the probe in air, co0, and of the coupled system,

k2 - 2

The coupling coefficient, k , is dependent on, and very sensitive to, many of the variables of the coupled system, i.e. conductivity, magnetic permeability, liftoff, material surface finish, tilt, frequency and temperature. As with any resonating system, even slight changes can have a large effect on the amplitude of oscillation. A change in k represents changes in the inductive and resistive components of the eddy-current circuit in Figure 3. A reduction in the coupling coefficient, k, from that of the system coupled to the surface of an undamaged material will be referred to as decoupling. This could be due to an increase in liftoff from the sample (liftoff-decoupling) or the presence of a defect (defect-decoupling). Each will cause a shift in the resonant frequency of the system (resonance-shift).

An experimental investigation was carried out to investigate ECT operation at frequencies approaching electrical resonance and determine the effects, if any, of defect-decoupling resonance-shifts on ECT defect signals.

3 Experimental Method

A single coil pen probe was used to investigate how operating at frequencies passing through resonance would affect the sensitivity of absolute ECT crack detection. The probe comprised 38 turns around a 1.02 mm ferrite core, within a ferrite cap (as shown in Figure 6), and was connected to the source using a

RG174 coaxial cable. The probe had a characteristic inductance in air of L0 = 10.34 ± 0.09^H and the coaxial cable had a capacitance of C0 = 101 pFm-1 [24].

Measurement error ±0.01 mm

Ferrite

Figure 6 - Eddy-current probe. X-Ray CT cross-sectional image of pen probe coil with ferrite core and shielding cap. Credit: Rolls-Royce NDE lab, Bristol, UK.

The probe was operated in the absolute mode [3] and was driven by a current source sweeping through frequencies up to and beyond resonance in the MHz frequency range. In order to recreate industrial inspection conditions, all measurements with the probe were performed with a single layer of Kapton® tape between the coil and test material.

Function Generator

Computer/Labview

Howland current source

Co-axial cable ECT Probe

XY Scan

Figure 7 - Experimental set-up. Schematic diagram of the experimental set up for a 2D, frequency sweep scan. A Function generator outputs a voltage sweep which a Howland current source converts into an equivalent current sweep which is sent to the probe. Electrical properties of the probe are monitored via a Labview program as the probe is scanned across the surface of the test specimen.

With reference to Figure 7: A Tektronix 3021B arbitrary function generator was used to create a swept voltage input signal which was converted into an equivalent drive current within a Howland current source.

The Howland current source [25], built in house, converts voltages from the arbitrary function generator into a constant current, which is supplied to the load coil. As a result, the amplitude of the current through the coil never changes and is linearly proportional to the driving voltage amplitude, even as the coil enters

different environments. Instead, the voltage across the coil (Vout) varies to maintain the constant current,

and is thus the measured quantity within the scan. The properties of the Howland current source make it particularly useful in multi-coil probes, when passive measurements are made. The Howland source ensures that passive sensors will only measure changes in the eddy-current flow within the material, and not arbitrary changes in the current through the primary excitation coil.

The current source converts a sinusoidal ±0.5 V signal into a sinusoidal ±50 mA drive current, which is supplied to the probe via a 1.56 m length of RG174 coaxial cable. The voltage in (Vm), from the function

generator, and the voltage across the probe ( Vout ), were monitored and recorded at a rate of 50 MSamples/ s using a National Instruments Labview program (4), which also controlled the XY stage

to scan the probe across the sample. For each step of the scan, Vin and Vout were recorded over one frequency sweep, and saved together in a binary file containing the probe co-ordinates. The ratio of Vout over Vin is proportional to the impedance, with the knowledge that Vin = IinR.

Three large calibration notch defects in a Titanium 6Al - 4V (Ti 6-4) specimen (pictured in Figure 8) were inspected, to investigate the effects of defect-decoupling resonance-shifts on defect signals. Ti 6-4 is one of the most widely used superalloys in the Aerospace industry, and so is an ideal test specimen [1]. The notch defects spanned the width of the test sample, and had depths of 0.20, 0.50 and 1.00 mm (0.008,

0.020 and 0.040 inches respectively). Ti 6-4 typically has a conductivity of ^ „ = 0.60 x106 S • m-1

Ti (6-4)

and a relative magnetic permeability ¡LiTi (6-4) = 1.00005 [26]

^ 40 50 60 70 80 90 100 110

Figure 8 - Calibration block. Image of a Titanium 6Al - 4V test specimen and the three notches of increasing

depth.

3 3.5 4 Frequency, MHz

Figure 9 - Material coupling resonance shift. Impedance frequency profiles of the absolute probe in air and coupled to an undamaged section of a Ti 6-4 test specimen.

A 4294A Agilent impedance analyzer was used to find the impedance of the eddy current probe and cable, as a function of frequency, Z (f), of the probe in different environments. Figure 9 shows the

Z (f) profiles of the probe both on (at zero liftoff with Kapton® layer), and off (in air) an undamaged

section of the Ti 6-4 test piece. The Z (f) profiles were also measured for the probe positioned directly

above the three notch defects where the maximum frequency shift (defect-decoupling) occurred (Figure 10).

Figure 10 - Defect-decoupling resonance shifts in the presence of calibration slots. Impedance frequency profiles of the absolute probe on undamaged Ti 6-4 and above three notch defects of increasing depth.

The resonant frequencies, f0, for each Z (f) profile are shown in Table 1 along with the estimated

coupling coefficient, k, calculated using equation 13, and the equivalent coil inductance, L1, of the system, calculated using equation 8.

Table 1 - Resonant frequency of an ECT probe in air, above undamaged and three defects within Titanium 6-

4 (zero liftoff).

Material Defect f0 (±0.003 MHz) k (±0.001) L1 (±0.01 pH)

Air - 3.938 0.000 10.34

Measurements under Kapton tape

Ti 6-4 Undamaged 4.125 0.421 9.42

0.20 mm 4.113 0.408 9.48

0.50 mm 4.088 0.380 9.59

1.00 mm 4.075 0.364 9.65

The resonant frequency shifts in Table 1 represent the defect-decoupling of the probe circuit with varying defect size. From the table, it is clear that the material condition affects the level of coupling that can be achieved by the system. It is these changes in the coupling, and therefore the resonant frequency, which give rise to a signal enhancing phenomenon.

4 Results

Three notches in the Ti 6-4 sample were tested using a single coil probe operated in the absolute mode and swept through frequencies from 1 - 5 MHz . A 0.52 ms repeating frequency sweep signal was

generated using a 25 MSamples.s-1 arbitrary function generator, to output a driver waveform that satisfies the Nyquist criterion.

The waveform was created to decrease exponentially with frequency (see Figure 7) so that the Vout

signal would not saturate the measurement scale as the frequency approached resonance. Vin and Vout were scaled such that their maxima, when coupled to undamaged Ti 6-4, were equal to 80% of full screen height. The resulting output signals, Vin and Vout, were converted into the frequency domain via Fourier Transform and 'binned' (bin width = 1923 Hz) over the full frequency range.

The ratio of Vcmt over Vin is proportional to the impedance of the system (Z = Vcmt /1 n). In this way a

value proportional to the magnitude of the ECT defect signals, |Z|, was found for each position within the

XY scan to build an image of the surface. The C-scan image was zeroed to an area of undamaged material. Figure 11 shows a high contrast C-scan image of the test piece surface constructed by the linear combination of data from frequency bins within the frequency range of 2.5 -4.0 MHz (avoiding resonance at 4.125 MHz).

-1500 -3.75 750 18.75 30.00

Zeroed Magnitude

Figure 11 - Calibration block C-scan image. Frequency "collapsed" C-scan image of the material surface showing an example of where defect (blue dashed circle) and noise (black solid square) data is measured.

At each frequency, the maximum signal strength magnitude, S(f ) = | V2 / Vl|, of each defect and the

root-mean squared (rms) noise level, Nrms(f) , of a 10 x 10 mm area of undamaged material was

recorded from the scan. The data was then plotted as a function of frequency (Figure 12), and compared to the impedance magnitude profiles of the probe on the defect, and on an undamaged section of the material.

Frequency, MHz

Figure 12 - Signal and noise. Defect signal (dashed blue) and rms noise (solid black) level for the 1.0mm deep notch as a function of frequency. Also showing the associated resonant frequency shift of the probe in the presence of the defect. Probe locations on and off the defect are shown in a schematic diagram.

It is clear from Figure 12 that there occurs a frequency, close to resonance, where the defect signal reaches a maximum. Note that the background noise reaches a peak at a higher frequency, where the defect signal reaches a local minimum. The resulting SNR plot as a function of frequency is shown in Figure 13 for the 1.0mm notch defect.

The broad distribution within the SNR plots could be a result of the background electronic noise produced by the XY scanning table. The SNR data was fitted with spline best fit curves (smoothing factor 0.99) to better display the shifting frequency of the signal peak with changing defect depth (Figure 14). The exact frequencies of the SNR peaks are displayed in Table 2.

2.5 3 3.5

Frequency, MHz

Figure 13 - Signal-to-Noise Ratio (SNR). Plot of the 1.0mm defect as a function of frequency. The plot is shown in relation to the resonance peak of the impedance profile (solid black).

a> <s>

§>10 09

0 2 mm 0.5 mm 1.0 mm

........

- ........

1500 g1

1000 S

500 —

2.5 3 3.5

Frequency, MHz

Figure 14 - Near electrical resonance signal enhancement. Experimental plot showing the SNR trends as a function of frequency for each defect. The plots are shown in relation to the resonance peak of the impedance profile (solid black).

The SNR peaks occur in a region conventionally avoided by probe manufacturers and operators owing to the unpredictable and unstable nature of resonance. Conventional probe operating frequencies finish significantly short of the electrical resonance of the probe, keeping safely within a range where the sensitivity scales linearly with frequency due to the inductive reactance component of impedance (wL).

Table 2 -SNR peak details in experimental data for three defects of increasing depth with reference to SNR at

Defect depth (mm) SNR at 1MHz SNR Peak frequency (±0.01 MHz) SNR at Peak SNR Enhancement from 1MHz

0.20 3.50 3.79 8.79 2.51

0.50 4.46 3.77 16.03 3.59

1.00 5.74 3.76 21.43 3.73

Beyond resonance, the reactive component of the system is dominated by capacitive changes within the cable, such that successful measurements of the inductive changes are impossible so this region must be avoided. Between the conventional limit and the electrical resonance of the system lies a region of probe sensitivity dominated by the effects of defect-decoupling resonance-shifts, where significant SNR enhancement occurs. This has been termed the Near Electrical Resonance Signal Enhancement (NERSE) frequency band, or zone, and will be dependent on the probe, cable length, test material and defect size.

Finding the NERSE Peak Frequency (fNERSE)

The defect signal magnitude, S(f) , shown in Figure 12, peaks as a result of resonance shifts caused by defect-decoupling of the probe as it passes over a defect. It occurs at a frequency where the difference in rate of change between the impedance on and off a defect is greatest. This can be expressed as the

maximum gradient of the ratio between the defect-decoupled, |Zd (f )| , and undamaged materialcoupled, |Z1 (f) , impedance profiles.

S (f)K f df

Zd (f )|

JZi( f

Using the impedance profile s of the three notch defects, equation 14 was calculated and compared to the defect signals, S(f) , for each defect. The scaled, predicted results are shown in Figure 15, to make clear the direct relationship between equation 14 and the defect signals.

There is a strong correlation between peaks in the signal strength and the location of the maxima and minima of equation 14. Table 3 compares the predicted peak position of fNERSE and the signal peak

fNERSE frequency, based on a spline best fit of the data, with a smoothing factor of 0.9998.

Table 3 - Defect NERSE frequency peak positions based on experimental absolute ECT scan data and predicted from impedance profile data calculation.

Defect depth (mm) Signal /NERSe ( ±0.01 MHz ) PrediCted fNERSE ±0.006 MHz ) (

0.20 3.88 3.869

0.50 3.85 3.856

1.00 3.83 3.850

fNERSE is dependent on the size of the defect, and will tend towards the frequency at which the gradient of |Z1(f )|, the coupled impedance, is a maximum as the defect size decreases to zero. This sets an

upper limit for the frequency of NERSE operation equivalent to the noise in resonance shifts on undamaged material. Beyond this frequency limit, the measurements of the electrical properties are complicated by high background noise and capacitive effects in the system.

There occurs a sharp minimum just before 4MHz. This corresponds to the frequency where the rate of change of impedance tends towards a minimum as it reaches resonance. After resonance the system starts to become dominated by its capacitive components but the inductance will still have some effect. This is why there is a smaller secondary signal peak just beyond resonance.

a) Notch 0.2 mm

b) Notch 0.5 mm

c) Notch 1 mm

Defect signal Ratio Gradient Gradient turning points

Frequency, MHz

Frequency, MHz

Frequency, MHz

Figure 15 - Relationship between signal enhancement and defect-decoupling. Normalized 0.20 (a), 0.50 (b) & 1.00 (c) mm deep, defect signal strength compared, on the same scale, to the gradient of the ratio of defect-decoupled impedance. Plots are shown in full (top) and for positive signals (bottom). Dotted lines represent the frequencies of maxima and minima in the predicted gradient plot.

5 Conclusions & Future Work

This paper has highlighted a band of frequencies, outside the conventional operation range, and close to electrical resonance of an eddy current probe, where the magnitude of impedance SNR reaches a peak. The SNR of scans of three slots of varying depth were enhanced by a factor of up to 3.7, from the SNR measured at 1MHz. This is a result of a defect-decoupling resonance-shift effect and is referred to as the near electrical resonance signal enhancement (NERSE) phenomenon. NERSE frequency operation has significant potential for ECT inspection, and opens up a range of investigative possibilities. Within this investigation, only the magnitude of the electrical impedance has been analyzed. An immediate extension of this investigation will be to consider phase information, and determine whether a similar exploitable NERSE effect exists.

In a break from conventional ECT, the identification of the NERSE frequency band has introduced the possibility of operating ECT probes at a single NERSE frequency (that of a target defect), in order to increase the probability of detecting smaller defects. Such an approach could improve the SNR for small defects where, at lower frequencies, the signal from the defect would be below the electrical background noise. This approach will also be investigated for detecting defects within materials of high levels of microstructural background noise. So long as the decoupling resonance-shift caused by the microstructure is less than that caused by any defects then a signal enhancement effect should be observed.

The electrical resonant frequency of an ECT system can be affected by a number of environmental factors including, but not limited to; temperature, liftoff, tilt and the degree of surface roughness. Operating at frequencies close to resonance will therefore lead to a greater sensitivity to these issues. However, for an automated inspection system, i.e. one controlled via robotic motion, serious liftoff and tilt variations can be suppressed such that resonance shifts from these factors are negligible compared to shifts resulting from the presence of defects. In all of the experiments carried out, no attempts to limit the effects of temperature were made. Variations in the temperature and surface roughness can be filtered out using a band-pass filter. So long as these factors do not vary in the extreme, so that the level and spatial distribution of resonance shifting they produce is comparable to that of defect shifts, the SNR

improvement will still be observed. These factors become more of an issue the smaller the target defect becomes. Future work will investigate the limits of this.

In addition to the use of NERSE frequencies in single coil inspections stated above, there are other promising applications for the additional information sweeping through resonance offers. Sizing and profiling may be possible, by carefully locating the peak frequency of swept defect signals, and ascertaining the level of defect-decoupling associated with a given defect in order to determine its approximate or relative size. Finally, the complex resonance interactions that occur in densely populated eddy-current arrays (ECAs) will be examined. The focus will be on exploiting the NERSE phenomenon in transmit-receive and arrayed probes in order to achieve greater sensitivity.

The prospect of improving sensitivity through a simple procedural change, without the need for complex probe design or data analysis, is a potentially valuable complementary technique to any ECT method.

Acknowledgements

Many thanks go to Rolls-Royce plc. and the Research Centre for Non-Destructive Evaluation (RCNDE)

for their support, funding and encouragement. Credit also goes to Rolls-Royce NDE lab, Bristol UK, for

the use of the X-Ray CT system.

References

1. Boyer, R.R., Titanium for aerospace: Rationale and applications. Advanced Performance Materials, 1995. 2(4): p. 349-368.

2. Gros, X.E., J. Bousigue, and K. Takahashi, NDT data fusion at pixel level. NDT & E International, 1999. 32(5): p. 283-292.

3. García-Martín, J., J. Gómez-Gil, and E. Vázquez-Sánchez, Non-Destructive Techniques Based on Eddy Current Testing. Sensors, 2011. 11(3): p. 2525-2565.

4. Abu-Nabah, B., W. Hassan, D. Ryan, M. Blodgett, and P. Nagy, The Effect of Hardness on Eddy Current Residual Stress Profiling in Shot-Peened Nickel Alloys. Journal of Nondestructive Evaluation, 2010. 29(3): p. 143-153.

5. Blodgett, M.P. and P.B. Nagy, Eddy current assessment of near-surface residual stress in shot-peened nickel-base superalloys. Journal of Nondestructive Evaluation, 2004. 23(3): p. 107-123.

6. Grimberg, R., L. Udpa, A. Savin, R. Steigmann, V. Palihovici, and S.S. Udpa, 2D Eddy current sensor array. NDT & E International, 2006. 39(4): p. 264-271.

7. Huang, H., N. Sakurai, T. Takagi, and T. Uchimoto, Design of an eddy-current array probe for crack sizing in steam generator tubes. NDT & E International, 2003. 36(7): p. 515-522.

8. Zilberstein, V., K. Walrath, D. Grundy, D. Schlicker, N. Goldfine, E. Abramovici, et al., MWM eddy-current arrays for crack initiation and growth monitoring. International Journal of Fatigue, 2003. 25(9-11): p. 1147-1155.

9. Bartels, K.A. and J.L. Fisher. Multifrequency eddy current image processing techniques for nondestructive evaluation. in Image Processing, 1995. Proceedings., International Conference on. 1995.

10. Liu, Z., K. Tsukada, K. Hanasaki, Y.K. Ho, and Y.P. Dai, Image fusion by using steerable pyramid. Pattern Recognition Letters, 2001. 22(9): p. 929-939.

11. Chady, T. and R. Sikora, Optimization of eddy-current sensor for multifrequency systems. Magnetics, IEEE Transactions on, 2003. 39(3): p. 1313-1316.

12. Liu, Z., K. Tsukada, K. Hanasaki, and M. Kurisu, Two-Dimensional Eddy Current Signal Enhancement via Multifrequency Data Fusion. Research in Nondestructive Evaluation, 1999. 11(3): p. 165-177.

13. Zheng, L., D.S. Forsyth, J.P. Komorowski, K. Hanasaki, and T. Kirubarajan, Survey: State of the Art in NDE Data Fusion Techniques. Instrumentation and Measurement, IEEE Transactions on, 2007. 56(6): p. 2435-2451.

14. Owston, C.N., A high frequency eddy-current, non-destructive testing apparatus with automatic probe positioning suitable for scanning applications. Journal of Physics E: Scientific Instruments, 1970. 3(10): p. 814.

15. Liu, C. and Y. Dong, Resonant enhancement of a passive coil-capacitance loop in eddy current sensing path. Measurement, 2012. 45(3): p. 622-626.

16. Liu, C.Y. and Y.G. Dong, Resonant Coupling of a Passive Inductance-Capacitance-Resistor Loop in Coil-Based Sensing Systems. Ieee Sensors Journal, 2012. 12(12): p. 3417-3423.

17. Ko, R.T., M.P. Blodgett, S. Sathish, and T.R. Boehnlein, Resonant Frequency Eddy Current Liftoff Measurements for Shot Peening Intensity Assessment in Materials (Preprint), 2007, DTIC Document.

18. Ko, R.T., M.P. Blodgett, S. Sathish, and T.R. Boehnlein, Application of resonant frequency eddy current technique on a shot-peened nickel-based engine-grade material, in Review of Progress in Quantitative Nondestructive Evaluation, Vols 26A and 26B, D.O. Thompson and D.E. Chimenti, Editors. 2007, Amer Inst Physics: Melville. p. 1608-1615.

19. Blitz, J., Electrical and magnetic methods of non-destructive testing, 1997: Chapman & Hall.

20. Horowitz, P. and W. Hill, The Art of Electronics1989: Cambridge University Press. 1152.

21. White, J.F., High Frequency Techniques: An Introduction to RF and Microwave Engineering2004: Wiley.

22. Wheeler, H.A., Formulas for the Skin Effect. Proceedings of the IRE, 1942. 30(9): p. 412-424.

23. Bleaney, B.I. and B. Bleaney, Electricity and magnetism 1989: Oxford University Press.

24. RS Online. 13/08/13]; Available from: http://uk.rs-online.com/web/p/coaxial-cable/5218206/.

25. Anudev, J. and I.J. Raglend. Analytical study of howland current source model. in Computing, Electronics and Electrical Technologies (ICCEET), 2012 International Conference on. 2012.

26. Welsch, G., R.F. Boyer, and E.W. Collings, Materials Properties Handbook: Titanium Alloys, 1994: Asm International.

Table 1 - Resonant frequency of an ECT probe in air, above undamaged and three defects within Titanium 6-

4 (zero liftoff).

Material Defect /„ (-v1.):1* MHz) If (=!Ï,W1) ¿1 (=Mi

Air - 0.000 10.34

Measurements under Kapton tape

Ti 6-4 Undamaged 4, 0.421 9.42

0.20 mm 4.11S 0.408 9.48

0.50 mm 0.380 9.59

1.00 mm 0.364 9.65

Table 2 -SNR peak details in experimental data for three defects of increasing depth with reference to SNR at

Defect depth (mm) SNR at 1MHz SNR Peak frequency HHË) SNR at Peak SNR Enhancement from 1MHz

0.20 3.50 3.79 8.79 2.51

0.50 4.46 3.77 16.03 3.59

1.00 5.74 3.76 21.43 3.73

Table 3 - Defect NERSE frequency peak positions based on experimental absolute ECT scan data and predicted from impedance profile data calculation.

Defect depth (mm) Signal (^. UMHt)

0.20 3.88 3.869

0.50 3.85 3.856

1.00 3.83 3.850

Figure Captions

1 - Equivalence circuit in air. Simplified electronic circuit for an eddy-current probe in free space with a capacitive coaxial cable connection

2 - Electrical resonance in air. The magnitude and phase of impedance of an ECT coil as it passes through the resonant frequency, f0.

3 - Mutual inductance transformer model. Simple transformer circuit model for an eddy-current probe (0) coupled to the surface of an electrically conducting material (e).

4 -Coupled ECT system equivalence circuit, adapted from Wheeler [22]

5 - Coupled eddy-current probe equivalence circuit with effective resistance and inductance.

6 - Eddy-current probe. X-Ray CT cross-sectional image of pen probe coil with ferrite core and shielding cap. Credit: Rolls-Royce NDE lab, Bristol, UK.

7 - Experimental set-up. Schematic diagram of the experimental set up for a 2D, frequency sweep scan. A Function generator outputs a voltage sweep which a Howland current source converts into an equivalent current sweep which is sent to the probe. Electrical properties of the probe are monitored via a Labview program as the probe is scanned across the surface of the test specimen.

8 - Calibration block. Picture of a Titanium 6Al - 4V test specimen and the three notches of increasing depth.

9 - Material coupling resonance shift. Impedance frequency profiles of the absolute probe in air and coupled to an undamaged section of a Ti 6-4 test specimen.

10 - Defect-decoupling resonance shifts in the presence of calibration slots. Impedance frequency profiles of the absolute probe on undamaged Ti 6-4 and above three notch defects of increasing depth.

11 - Calibration block C-scan image. Frequency "collapsed" C-scan image of the material surface showing an example of where defect (blue dashed circle) and noise (black solid square) data is measured.

12 - Signal and noise. Defect signal (solid blue) and rms noise (dotted black) level for the 1.0mm deep notch as a function of frequency. Showing the associated resonant frequency shift, of the probe, in the presence of the defect. Probe locations on and off the defect are shown in a schematic diagram.

13 - Signal-to-Noise Ratio (SNR). Plot of the 1.0mm defect as a function of frequency. The plot is shown in relation to the resonance peak of the impedance profile (solid black).

14 - Near electrical resonance signal enhancement. Experimental plot showing the SNR trends as a function of frequency for each defect. The plots are shown in relation to the resonance peak of the impedance profile (solid black).

15 - Relationship between signal enhancement and resonance decoupling. Normalized 0.20 (a), 0.50 (b) & 1.00 (c) mm deep, defect signal strength compared, on the same scale, to the gradient of the ratio of decoupled impedance. Plots are shown in full (top) and for positive signals (bottom). Dotted lines represent the frequencies of maxima and minima in the predicted gradient plot.

• Defect signal-to-noise enhancement at frequencies approaching electrical resonance

• Enhancement is due, and proportional to, resonance decoupling above a defect

• The signal enhancement peak frequency is dependent on the defect size.