NDT&E International 51 (2012) 85–93
Contents lists available at SciVerse ScienceDirect
NDT&E International journal homepage: www.elsevier.com/locate/ndteint
A differential planar eddy currents probe: Fundamentals, modeling and experimental evaluation Luis S. Rosado a,b,c,n, Telmo G. Santos d, Pedro M. Ramos a,b, Pedro Vilac- a a,e, Moise´s Piedade a,c a
Instituto Superior Te´cnico (IST), Universidade Te´cnica de Lisboa, 1049-001 Lisboa, Portugal ~ (IT), Instituto Superior Te´cnico, Universidade Te´cnica de Lisboa, 1049-001 Lisboa, Portugal Instituto de Telecomunicac- oes c ~ e Desenvolvimento, 1000-029 Lisboa, Portugal Instituto de Engenharia de Sistemas e Computadores (INESC), Investigac- ao d UNIDEMI, Departamento de Engenharia Mecˆ anica e Industrial, Faculdade de Ciˆencias e Tecnologia, Universidade Nova de Lisboa, 2829-516 Caparica, Portugal e Instituto de Engenharia Mecˆ anica (IDMEC), 1049-001 Lisboa, Portugal b
a r t i c l e i n f o
abstract
Article history: Received 3 February 2012 Received in revised form 26 June 2012 Accepted 28 June 2012 Available online 7 July 2012
This paper presents a new eddy current probe specifically designed to inspect imperfections along Friction Stir Welding joints. The proposed probe has a planar design and a differential operation introducing several innovative aspects in eddy currents generation and sensing. The fundamentals on the probe operation are presented and explained in the presence of a metallic part to be inspected with and without defects. A finite element model was used to detail the probe operation and to assess the influence of several operational parameters on the probe response. Finally, the simulations were validated experimentally using a prototype probe produced in printed circuit board and an aluminum alloy block with standard defects produced using electro-discharge machining. & 2012 Elsevier Ltd. All rights reserved.
Keywords: Eddy currents Planar probe Differential probe Non-destructive testing Friction Stir Welding.
1. Introduction The detection and characterization of small imperfections and defects in metallic parts remain a complex task even for the latest Non-Destructive Testing (NDT) methods. The typical approach, when the part to be analyzed is already deployed, is the application of eddy currents or ultrasonic based methods because of their sensitivities and because they can be used in portable instrumentation. Due to its operation principle, eddy currents are often selected to detect superficial or sub-surface defects. One basic eddy current probe is a cylindrical coil used to simultaneously generate and sense the electrical currents in the metallic part. When a defect interacts with the induced eddy currents, the magnetic field is modified causing an impedance change on the coil [1]. Over the last years, several improvements to the probe itself have been proposed taking advantage of the separation of the eddy current generation and sensing elements. By making such modification, it is possible to individually optimize the elements for their intended purpose. A common approach is to use a coil to produce a strong and uniform magnetic field together with small,
n Corresponding author at: Universidade Te´cnica de Lisboa, Instituto Superior Te´cnico, Av. Rovisco Pais, 1, 1049-001 Lisboa, Portugal. Tel.: þ 351 218417665. E-mail address:
[email protected] (L.S. Rosado).
0963-8695/$ - see front matter & 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.ndteint.2012.06.010
very sensitive sensing coils or other sensing devices. Another advantage of this type of solution is the spatial resolution enhancement. In the work reported in [2] the authors studied the response of a probe composed by a driver coil wired around a small coil with a considerable number of windings to increase sensitivity. Other works propose the introduction of a second sensing coil to achieve differential based operation [3]. Another approach along these lines is to replace the sensing coil by other magnetic field sensors. Latest developments on sensor technology provided a wide range of magnetic sensor types, some of them commercially available. Instead of being sensitive to the magnetic flux as coils, this type of sensors is generally magnetic field driven. Because of such operation, its sensitivity is ensured even for low frequencies where the use of coils may be unpractical. Thus, they are preferred when designing eddy current probes for deep buried defects requiring the use of very low frequencies. The use of Superconducting Quantum Interference Devices (SQUID) for analyzing the resulting magnetic field while a coil is used to generate eddy currents is reported in [4]. In [5], a Hall Effect sensor together with a flat coil for eddy currents generation was used. In [6], the same type of sensors was evaluated on the detection of defects in riveted structures. Several papers reported the use of Giant Magneto-Resistance (GMR) sensors with driver coils with different shapes [7–9]. Other sensors such as the Spin-Dependent Tunneling (SDT) were also used as reported in [10].
86
L.S. Rosado et al. / NDT&E International 51 (2012) 85–93
This paper begins with the description of the probe structure and principle of operation making use of Finite Element Modeling (FEM) examples. The same model was used to assess the probe response for several defective conditions and the influence of the several operational parameters. The validation of the simulated results using an experimental prototype is also reported.
2. Probe design fundamentals
Fig. 1. Probe design. Driver trace and sensing coils representation in the probe layer.
The use of planar probes has also been proposed by several authors. In [11], the authors modeled the use of a spiral planar coil in the detection of defects. Later in [12], the authors studied the behavior of rectangular planar coils in the defect detection and the influence of the coil aspect ratio in its behavior. The typical way to produce these probes is the use of photolithographic processing to reproduce the desired patterns in a support substrate allowing for a simple and cheap production process. Another advantage of the planar probes is that they can be produced in flexible substrates allowing contact with corners and complex or low accessibility surfaces. This advantage was used in [13] where a polyamide flexible substrate probe allowed the inspection of non-flat surfaces. Depending on the selected photolithographic process, it is possible to achieve good resolutions enabling the production of micro-dimensional coils. In [14], the authors produced and tested eddy current probes with rectangular coils featuring 40 windings in a surface with area as small as 1 mm2. Other planar geometries were also evaluated either to generate the eddy currents or to sense them. In [15] a probe composed of two overlapping rectangular coils is described. The work reported in [16] describes a probe composed of a sensing mesh coil overlapping a meandering shaped coil to generate the eddy currents. Another approach was introduced in [17] where a probe concept based on a meandering shaped eddy current generator trace and multiple sensing windings placed beneath the meander gaps was patented. Later, in the work reported in [18], the applicability of this probe concept to the inspection of in service critical components was demonstrated. In this work a new differential planar eddy current probe optimized for the detection of defects following a specific orientation is studied. This type of defects can be typically found at the vicinity of the root and top surfaces of Friction Stir Welding (FSW) joints whose effective NDT is still a challenge. Although this welding process has good reliability, under industrial conditions, these defects may appear as a consequence of small variations of the welding parameters and aging of the welding tool. These defects are superficial cracks with depths ranging from 50 mm to 500 mm which can seriously reduce the joint expected life under fatigue loads and thus compromise its application to critical components. The FSW process causes conductivity changes on the joint that can reach up to 10% from the original material conductivity which results in increased difficulties in detecting the small defects. Probes for this application should have (i) high sensitivity to superficial defects, (ii) confined induced eddy currents, (iii) improved immunity to lift-off and slowly varying conductivity changes, (iv) high repeatability and reliable probe manufacturing process. The application of a preliminary version of the proposed probe to FSW joints with several defective conditions was reported in [19].
The proposed probe was designed to overcome some limitations found using conventional coil probes when inspecting small dimension defects. In this case, low sensitivity, small conductivity changes in the metallic part and the probe lift-off strongly compromise successful defects detection. Examples of such small defects are present in the root of FSW joints as lack of penetration cracks and particle alignments interfaces with depths as low as 50 mm. The specific orientation of such defects may be used as an advantage when designing eddy current probes for such applications. In the proposed probe design, the sensitivity was improved to defects extended in the direction of a symmetry axis defined in a differential design. Also, the differential based operation and the ability of generating very confined eddy currents allows reducing the probe sensitivity to slowly varying conductivity changes and lift-off. In this new probe, eddy currents are generated by a driver trace and sensed in the two surrounding planar coils as shown in Fig. 1. The driver trace element is a planar metallic trace responsible for carrying an alternating current that induces the eddy currents in the material. When an alternating current Iin is made to flow in the driver trace, the generated magnetic field lines are contained in an orthogonal surface S, concentric and governed by (neglecting the displacement current term in the Maxwell equations) I ! ! Hd d l ¼ Iin ð1Þ S
Considering a surface P defined in a metallic part placed underneath the probe, the magnetic flux due to the driver trace in this surface is Z f ! ¼ Bd dA, Bd ¼ mHd ð2Þ Bd ,P P where A is the area element on the defined surface. This flux generates an electromotive force
ed ¼
df ! Bd ,P dt
ð3Þ
which in turn will originate the eddy currents in the metallic part. These eddy currents modify the magnetic field disposition and ! intensity from the original Bd to ! ! ! ð4Þ B ¼ Bd þ Be ! where Be refers to the magnetic field generated by the eddy currents. Any interaction between eddy currents and defects ! shows up as a perturbation of the magnetic field Be and conse! quently on the overall magnetic field B . As shown in Fig. 1, the sensing coils around the driver trace are symmetrical with respect to the driver trace and share a common terminal. The induced voltage sensed by each coil is
ec ¼
N X i¼0
df
B ,Pi
dt
ð5Þ
where N is the number of coil windings and Pi the surface defined inside each winding. Taking into account the coils shared terminal, the voltage across the probe output terminals is the sum of the induced voltages in the two sensing coils denoted by Uout.
L.S. Rosado et al. / NDT&E International 51 (2012) 85–93
87
situations described before can be considerably reduced using adequate equipment to handle the probe positioning. Another advantage resulting from the probe structure is the possibility to manufacture it using flexible substrates for use in inspecting nonplanar and complex geometry surfaces.
3. Finite Element Modeling
Fig. 2. Probe operation principle illustration, magnetic fields and eddy currents. Metallic part free of defects (a) and with a defect underneath one of the sensing coils (b).
To understand the probe operation, three situations are now described. In the first one, the probe is considered to operate without the presence of the metallic part. As eddy currents will ! not be generated, the magnetic field is reduced to Bd , the driver trace current contribution which will be necessarily symmetric in the two sensing coils. Therefore, the induced voltages on each coil will be symmetrical and consequently cancel each others contribution on the probe output voltage, accordingly to its differential operation configuration. In the second case, a homogeneous metallic part is placed underneath and parallel to the probe surface. In this case, eddy currents will be induced describing loops through the material passing underneath the driver trace as shown schematically in Fig. 2(a). As both the eddy currents ! ! magnetic field Be and Bd will be symmetric, it is a similar situation to the previous one and, once again, the probe output voltage will be 0. In the third case, a defect is considered underneath one of the sensing coils affecting the symmetry with respect to the probe axis, Fig. 2(b). The defect will modify the eddy currents disposition and consequently change the original magnetic field leading to different contributions in the two coils. The presence of the defects shows up as an increase in the probe output voltage resulting from the unbalanced induced voltage in each of the sensing coils as long as the defect is not symmetric with the probe symmetry axis. As described, the probe design ensures a differential operation with respect to the symmetry axis defined by the driver trace. The modification of the symmetry condition may be caused either by defects or conductivity variations inside the metallic part to be inspected. In both the cases, the symmetry condition modification causes an increase on the probe output voltage amplitude that can be detected by characterizing the ratio between the complex quantities U out and I in . The main advantage of this differential based operation is the improved signal to noise ratio in the defects response and consequently a sensitivity enhancement is obtained. Although the probe differential operation reduces its sensitivity to the lift-off effect, its influence may still be noted in two specific situations. In the first one, the sensing coils are not parallel to the metallic part surface, i.e., there is an angular liftoff. The symmetry modification in this situation leads to a nonzero output voltage even without defects or conductivity variations. In the second case, the probe remains parallel to the metallic part surface but the distance between them is modified leading to a change in the probe sensitivity and the consequent change on the response to defects and conductivity variations if they are present. As the probe is completely planar, the lift-off
The Finite Element Modeling (FEM) software CST EM Studio [20] was used to confirm the projected characteristics of the probe and to predict the probe responses for several defect patterns. Moreover, the results from these simulations allowed the definition of requirements on the electronic devices used to drive the probe and measure its output. The dimensional parameters of the probe are presented in Fig. 3 and the corresponding values on the designed model are registered in Table 1. Some of the values in Table 1 were chosen in order to meet technical specifications required to produce a prototype probe using printed circuit board technology. In the designed probe model, both the driver trace and the sensing coils are made of copper with thickness 35 mm surrounded by vacuum. The metallic part is a 5 mm thickness aluminum block which is placed 200 mm below the probe plane as shown in Fig. 4. The probe layer thickness and distance between the probe plane and the material surface were selected to match as closely as possible the conditions later used on the experimental validation. Some initial experiments were made to evaluate the spatial grid resolution necessary to achieve acceptable results in the simulated magnetic fields. A hexahedral mesh with about 1.5 million elements (corresponding to a total of 9 million degrees of freedom) was defined and used to compute the simulation results. To validate the operation described in Section 2, the simulation software was used to verify the generated eddy currents in the aluminum block and the magnetic field disposition in the presence of a defect. A sinusoidal waveform with 1 A amplitude and frequency of 1 MHz was applied in the driver trace and the software was set to register the electrical currents and the magnetic field. In
A B C D
E
Fig. 3. Probe dimensional parameters representation. See Table 1 for description and values.
Table 1 Probe characteristics and dimensions in agreement with Fig. 3.
A B C D E –
Description
Value
Sensing coils windings width Sensing coils windings clearance Sensing coils external dimension Driver trace width Driver trace length Number of windings on each sensing coil
100 mm 100 mm 10 mm 1 mm 11 mm 12
88
L.S. Rosado et al. / NDT&E International 51 (2012) 85–93
Z Y
X
Mesh Representation
Aluminum part
Sensing Coils
Driver Trace
Defect
Fig. 4. Probe model and hexahedral mesh representation.
Z
A sweep across the defect number 2, in the direction shown in Fig. 6, was simulated. The probe was placed so that the driver trace remains parallel to the defect extension and the relative positioning between the probe and the defect was successively modified. A sinusoidal waveform with 1 A amplitude and 1 MHz frequency is used in the driver trace while the probe moves in the direction of the defect. The selection of this operating condition was done after performing some preliminary simulations to estimate the expected output voltage amplitude while testing the defect. Due to the inductive nature of the sensitive element, this output voltage amplitude increases with growing operating frequencies. The 1 MHz testing frequency was selected as it provides appropriate signal levels for the experimental validation of the probe described in Section 4. Fig. 7 shows the probe response amplitude evolution for a 14 mm sweep starting 7 mm before the defect location with resolution of 100 mm. The phase of the output voltage referred to the driver trace current is also shown in Fig. 7. As shown in Fig. 7, when the probe is away from the defect location, the output amplitude response is roughly 0. Slightly before the external winding of the sensing coils overlaps the defect, the output amplitude starts increasing as is shown in position 5 mm (notice that the probe representation below is not in scale with the horizontal axis). The output voltage amplitude continues to increase until the probe is positioned about 1 mm before the defect where it reaches its maximum. This maximum output amplitude occurs when the defect strongly reduces the eddy currents beneath the driver trace, i.e., when the defect is extremely near the driver trace. On this situation, the highest eddy currents asymmetry leads to maximum magnetic
X
Sweep Direction
Z Y 0
X
8×10 A/m
Fig. 6. Notch defects dimensions.
Z X
180
90
Fig. 5. Induced eddy currents intensity (a) and the magnetic field vertical component (b) in a cut-plane transversal to the driver trace which was centered with the probe (y ¼0). On both the figures, data corresponds to when the driver trace current reaches its maximum.
Fig. 5(a) the induced eddy currents are shown in a cut plane transversal to the driver trace orientation which is centered with the probe. It is clear that eddy currents are concentrated under the driver trace and the influence of the defect on their path. The same plane was used to obtain the representation of the magnetic field component across the sensing coils in Fig. 5(b) where the perturbation from the defect is also clear. A set of notch defects with different depths was defined in the metallic part surface next to the probe plane. In all the defects, the notch has 400 mm width and depths as shown in Fig. 6.
¯
Phase{ UI¯ } [°]
0.0006 V.s/m
x 10
¯
-0.0006
Amplitude{ U I¯ } [Ω]
2
1
0
-90
0
-8
-6
Defect
-4
-2
0
2
4
6
8
-180
X [mm]
Sweep Direction Fig. 7. Probe positioning sequence and response signal evolution to defect number 2 with 0.5 mm depth and 400 mm wide (notice that the probe representation is not in scale with the horizontal axis).
L.S. Rosado et al. / NDT&E International 51 (2012) 85–93
3
x 10-3
3 2
1
1 0 -7
-3.5
0 X [mm]
4.5
3.5
4
0 2 1.5
¯
2.5
Phase { U I¯ } [°]
¯
Amplitude { U I¯ } [Ω]
90
3
-90
1 0.5 0
-8
-6
-4
-2
0
2
4
6
8
-180
X [ mm ]
Defect
Sweep Direction Fig. 9. Simulated amplitude and phase responses for the adjacent notches of the defect number 5 (note that the probe representation is not in scale with the horizontal axis).
2
x 10-3
1.5
Frequency
1 0.5 0 -7
7
-3.5
0 X [mm]
3.5
7
-3.5
0 X [mm]
3.5
7
100
Phase { U I¯ } [ ° ]
0
¯
0
¯ U
I¯
} [ °]
180
3.5
100
Phase {
x 10
¯
4 2
The response to the defect number 5 was also simulated and registered in Fig. 9 where the position 0 was defined when the probe is centered between the two notches. As the adjacent notches have different depths, they form a perturbation that is no longer symmetrical in the sweep direction leading to a nonsymmetrical response and to output amplitudes different than 0 when the probe is centered with the defect. To evaluate the input current frequency influence on the probe response signal, a set of 16 logarithmic spaced frequencies between 10 kHz and 10 MHz was evaluated. Using the same sweep parameters and the input amplitude current of 1 A, a sweep over the defect number two was simulated for each of the 16 frequencies. The output amplitude and phase responses for the tested frequencies between 100 kHz and 1 MHz are shown in the top of Fig. 10 where the highest amplitude curves correspond to the highest tested frequencies. In the output phase responses, the frequency variation seems to have small influence as the several curves almost overlap.
Amplitude { U I¯ } [Ω]
Amplitude {
¯ U
I¯
} [Ω]
field asymmetry and maximum output voltage amplitude. After this, the output amplitude decreases until the defect is perfectly centered with the probe leading to an output amplitude close to 0. The sweep continues with the second sensing coil moving over the defect while the output signal amplitude describes a symmetrical evolution with respect to the horizontal axis. The defect also affects the output phase response mainly by causing a 1801 phase shift resulting from the change of the sensing coil with highest induced voltage amplitude. Using the same probe model and parameters for the input current and the positioning sweep, the remaining four defects described in Fig. 6 were analyzed and the results are shown in Fig. 8. The main differences in the response signals to the four defects can be found in the amplitude response evolution which substantially increases with the depth of the defects. Even though the selected operating frequency and the simulated aluminum conductivity of 34.5% IACS lead to a standard penetration depth of 84.4 mm which is much smaller than the defects depths, there is a significant difference on the probe amplitude for the different defects. Around the defects with higher depths, the induced eddy currents are subject to a greater modification causing higher asymmetry on the magnetic field disposition which in turn results in an increased amplitude response. A different situation occurs in the output phase evolution where the defect depth has reduced influence. Another noteworthy characteristic observed in Fig. 8 is the amplitude response shape modification for the several defects. In the defects with lower depths (defect numbers 1 and 2), the maximum values of the amplitude response are registered when the driver trace is near the defect (such as in Fig. 7). For the two defects with higher depths (defect numbers 2 and 4), the maximum amplitude values occur when each of the sensitive coils is centered with the defect. In these cases, the amplitude response contribution due to the eddy currents reduction beneath the driver trace (when it is near the defect) is almost not affected by its depth. However, the same is not true when the defect disturbs the eddy currents beneath the sensitive coils. In this situation, the defect depths prevent eddy currents from flowing under the defect causing its concentration near the defect edge. This increased eddy currents modification leads to increased magnetic field asymmetries and consequently to higher output amplitudes.
89
-100 -200 -7
-3.5
0 X [mm]
3.5
7
Fig. 8. Simulated amplitude and phase responses for the defect numbers 1–4.
-100 -200 -7
Fig. 10. Simulated amplitude and phase responses to the defect number 2 for the several tested input current frequencies between 100 kHz and 1 MHz.
L.S. Rosado et al. / NDT&E International 51 (2012) 85–93
10-3
10-3
-4
-4
10
10-5
10-5
ΔX
10
1
10-6 101
10-6 102
103
104
f [kHz]
10-4
10-4
10-5
10-5
¯
10-3
{| U I¯ |} [Ω]
10-3
ΔX
10-2
1
ΔX
¯
Max {| U I¯ |} [Ω]
Fig. 11. Maximum and normalized trapezoidal integration of the simulated amplitude responses with the several input current frequencies.
10-2
10-6 -2 10
10-1 I [A]
1.4
0.7
1.2
0.6
1
0.5
0.8
0.4
0.6
0.3
0.4
0.2
0.2
0.1
0
0
0.1
0.2
0.3
0.4 0.5 0.6 Lift – Off [mm]
0.7
0.8
0.9
1
¯
0.8
ΔX
0.9
1.6
1 ΔX
¯
1.8
{| U I¯ |} [Ω]
x 10-3 1
x 10-3 2
0
Fig. 13. Maximum and normalized trapezoidal integration of the simulated amplitude responses in the several lift-off conditions.
the frequency to 1 MHz, a sweep as specified for the previous results was simulated for each of the lift-off conditions. The results for a linearly spaced lift-off between 0.1 and 1 mm are shown in Fig. 13. In addition to signal intensity reduction clearly shown in the figure, in the increased lift-off conditions a significant modification on the amplitude response evolution was also observed.
4. Experimental evaluation
¯
10-2
{| U I¯ |} [Ω]
10-2
ΔX
¯
Max {| U I¯ |} [Ω]
The simulated amplitude responses for the 16 different frequencies were processed in order to quantify their intensity. The maximum value and the normalized trapezoidal integration of the amplitudes on each response were computed and presented in Fig. 11. Due to the sensing coils operation principle, the output voltage will grow proportionally with the derivative of the magnetic flux across them. As the input current frequency appears as a multiplicative constant in the magnetic flux derivative, the signal intensity is expected to grow approximately linearly with the input current frequency. The input current intensity influence on the response signal was also tested in a set of 11 logarithmic spaced current amplitudes defined in the interval 10 mA to 1 A. Using the defect number 2 and setting the input current frequency to 1 MHz a sweep was simulated for each of the current amplitudes. In Fig. 12, the maximum detected amplitude and the trapezoidal integration of the amplitudes values of each sweep are shown. By increasing the current intensity, the same happens with the magnetic flux on the sensing coils leading to increased induced voltages. In the two plots, the relation between the intensity indicators and the input current is almost linear. To test the lift-off influence, the model was modified by changing the distance between the probe plane and the aluminum block surface. Setting the input current to 1 A amplitude and
Max {| U I¯ |} [Ω]
90
10-6 100
Fig. 12. Maximum and normalized trapezoidal integration of the simulated amplitude responses with the several input current amplitudes.
The experimental validation of the probe operation and its FEM model is described in this section. As seen in the Section 3, the probe sensitivity has a strong dependence on the driver trace current parameters, increasing with both its amplitude and frequency. To properly operate the probe, a dedicated eddy currents testing instrument was used whose detailed description was presented in [21]. An important feature of this new system is the possibility of generating a driver trace current with amplitude as high as 1 A in a wide range of frequencies. The measurements on this section were done by digitalizing both the I in and U out waveforms, performing some simple filtering operations and analyzing its fast Fourier Transform coefficients to compute the amplitude and phase of the dominant component taking into account spectral leakage effects. The probe was manufactured in a printed circuit board substrate with a copper layer of 35 mm thickness using a photoresist and etching process. The produced probe prototype is shown in Fig. 14. A shielding plane was added to reduce the influence of external undesired electromagnetic sources. This plane is etched in the probe substrate using the same material and process of the probe elements. A circular cutout was selected so a constant gap between the sensitive coils external winding and the plane is achieved. The FEM tool was used to investigate the cutout radius so the shielding plane starts when the magnetic field absolute value is reduced to about 10% from its maximum value on the probe center. The single defects described in Section 3 were reproduced on an aluminum part using electro-discharge machining and characterized in a sweep with similar parameters to the simulated ones. The driver trace current is set to 1 A amplitude and frequency 1 MHz. The comparison between the simulated and measured responses for the defects 1–4 is shown in Fig. 15. For all the defects, the simulated and measured responses are in good agreement. Notice that the phase differences registered for the highest absolute values of relative positioning between the probe
L.S. Rosado et al. / NDT&E International 51 (2012) 85–93
and the defect are not significant since they correspond to small output amplitude responses. The same procedure was done using the reproduced adjacent notches defects (defect number 5) and its response is shown in Fig. 16 with a good agreement between the simulations and measurements. The small differences in the amplitude response may result from the weak accuracy on the reproduction of the defects width and also the difficult to set the distance between the probe and the metallic part to the simulated 200 mm lift-off. As done in the Section 3, defect number 2 was repeatedly tested using a set of frequencies linearly spaced between 100 kHz and 1 MHz and the results are shown in Fig. 17. The higher value amplitudes correspond to higher frequencies thus validating the behavior found in the simulation. The influence of the input current amplitude on the probe response was also experimentally validated. The same set of logarithmic spaced input current amplitudes with the frequency set to 1 MHz was applied. In Fig. 18 both the simulated maximum amplitude and the trapezoidal integration are compared with the measured values for each of the input currents. In all the previous tests, the driver trace was placed parallel to the defect orientation. This orientation leads to highest symmetry modifications when the probe is placed over the defect and consequently to its maximum sensitivity. Inversely, if the driver trace is placed perpendicularly to the defect orientation, the
PCB Substrate
Sensitive Coils
Driver Trace
Shielding Plane
−3.5
0 3.5 X [mm]
0 −7
7
−3.5
0 3.5 X [mm]
100 0 −100
−3.5
0 3.5 X [mm]
7
U¯out } [Ω] I¯
1
−3.5
0 3.5 X [mm]
0 −100
−3.5
0 3.5 X [mm]
7
1
0 −7
−3.5
0 3.5 X [mm]
7
−3.5
0 3.5 X [mm]
7
200
100 0 −100 −200 −7
x 10−3
2
7
200
100
−200 −7
3
2
0 −7
7
200 U¯ Phase{ ¯out } [°] I
U¯ Phase{ ¯out } [°] I
200
−200 −7
1
U¯ Phase{ ¯out } [°] I
0 −7
2
Defect 4
x 10−3
U¯ Phase{ ¯out } [°] I
1
3 U¯out } [Ω] I¯
2
Defect 3
Amplitude {
U¯out } [Ω] I¯
3
Amplitude {
Amplitude {
U¯out } [Ω] I¯
3
Defect 2 x 10−3
Amplitude {
Defect 1
symmetry is kept and the probe output voltage remains close to 0. The influence of the driver trace placement with respect to the driver trace orientation was also evaluated. This was done by rotating the probe in 7.51 steps and performing a sweep over the defect number 2 for each of the orientations. The results of each sweep were processed and the maximum registered amplitude and the integral of the output voltage amplitude computed as described in Section 5 are shown in the polar plot of Fig. 19. Note that all the values were normalized so that the maximum measured quantity among the several steps (which corresponds to an angle equal to 0) has a value of one. As explained before, when the driver trace is perpendicular to the defect orientation, in both 901 and 901, the maximum amplitude and integrals are both near 0. The quantities start increasing while the angle is successively modified from 901 or
Fig. 16. Comparison between the simulated and the measured probe responses for the adjacent notches defects (defect number 5).
Fig. 14. Prototype probe processed in a printed circuit board substrate.
x 10−3
91
−3.5
0 3.5 X [mm]
7
100 0 −100 −200 −7
Simulated Fig. 15. Comparison between the simulated and the measured probe responses for the defect numbers 1–4.
Measured
92
L.S. Rosado et al. / NDT&E International 51 (2012) 85–93
10−2
10−2
Simulated Measured
Z
10−4
10−4
10−5
10−5
10−6 1 10
10−6
ΔX
102
103
Fig. 20. Notch defect with ramp depth profile dimensions.
104
f [kHz] Fig. 17. Maximum and trapezoidal integration of the measured and simulated amplitude responses with the several input current frequencies.
10−4
10−4
10−5
10−5
10−6 10−2
10−6
¯
10−3
ΔX
10−3
{| U I¯ |}[Ω]
10−2
Simulated Measured
registered amplitude while the amplitudes integral remains almost unchanged leading to the directivity differences presented in Fig. 19. Another defect was processed in an aluminum part using electro discharge machining. This notch defect has a ramp depth profile from 34 mm to 0 mm as shown in Fig. 20. In this defect, the probe is placed with the driver trace parallel to the ramp and multiple sweeps are performed in the perpendicular direction. The resolution on the perpendicular direction was 100 mm while the step between each sweep was 1 mm. The driver trace current amplitude is 1 A and the frequency is 1 MHz. The measured amplitude and phase responses are shown in the two dimensional intensity plots of Fig. 21 where the influence of the increasing defect depth is clearly shown. Note that the phase estimates for the highest absolute values in the X-axis are compromised by the small output voltage amplitudes.
1 ΔX
¯
Max {| U I¯ |} [Ω]
10−2
10−1 I [A]
5. Conclusions
100
Fig. 18. Maximum and trapezoidal integration of the measured and simulated amplitude responses with the several input current amplitudes.
¯
Max {| U I¯ |} 1 X
0° 1 −30°
X {|
¯ U
I¯
|}
30° 0.8
Defect
0.6 −60°
60° 0.4
0.2
−90°
Y
1 ΔX
¯
Max {| U I¯ |} [Ω]
10
¯
10−3
{| U I¯ |} [Ω]
X −3
Sweep Direction
90°
Fig. 19. Polar plots representing the maximum and the trapezoidal integration of the measured amplitude responses for the several angles between the driver trace and the defect. 01 correspond to a parallel orientation between the driver trace and the defect.
901 to 01, when both achieve its maximum values. The change of this angle affects the output voltage amplitudes evolution both qualitatively and quantitatively. For instance, in the tested defect, the high amplitudes registered between about 1 mm away from the defect start vanishing while the angle approaches 901 or 901. This qualitative change has a strong effect on the maximum
In this work, a new eddy current probe intended for NDT of Friction Stir Welding joints was presented, modeled and experimentally evaluated. The main characteristic of the proposed sensor is an increased sensitivity to defects following a specific orientation. This property is usually located at root and top zones of Friction Stir Welding beads where the conventional probes experience difficulties from several origins on the defects detection. In the proposed probe, eddy currents are generated by the flow of an alternate current on a planar trace which is placed between two symmetrical sensing coils. These sensing coils share 1 terminal so in the 2 remaining output terminals it is possible to measure the sum of their induced voltages. In a symmetry condition, these induced voltages are equal in amplitude and in phase opposition so its sum is 0. The presence of defects or other conductivity variations modifies the magnetic field disposition and consequently the induced voltage equilibrium is also changed leading to an output voltage different from 0. The defect detection is made by characterizing the complex ratio between the output voltage on the sensing coils terminals and the input current flowing in the driver trace. A finite element modeling tool was used to study the probe operation principle and to evaluate the influence of the defect on its response. The small differences on the probe response to the several defects appear in the output voltage amplitudes while the output voltage phase is almost similar among them. Later, the influence of the input current waveform parameters on the probe response was studied by performing tests with several input current amplitudes and frequencies. It was shown that either input current amplitude or frequency increases result in an increase of the output voltage amplitudes. The probe and the set of defects were reproduced and used to experimentally validate the simulation results and the proposed design. A good agreement between the simulated and measured
L.S. Rosado et al. / NDT&E International 51 (2012) 85–93
93
Fig. 21. Measured notch defect with ramp depth profile response. Intensity graphs for amplitude and phase responses.
probe responses was registered for all the tested defects demonstrating the accuracy of the simulation model. Finally, some tests to evaluate the influence of the angle between the symmetry axis and the defect orientation were made showing the directivity pattern in the probe sensitivity.
Acknowledgments The authors would like to acknowledge Fundac- a~ o para a Ciˆencia e Tecnologia (FCT) for its financial support via the Ph.D. scholarship FCT- SFRH/BD/65860/2009. References [1] Raj B, Jayakumar T, Thavasimuthu M. Practical Non-destructive testing.Woodhead Publishing; 2002. [2] Fu F, Bowler J. Transient eddy-current driver pickup probe response due to a conductive plate. IEEE Trans Magn 2006;42:2029–37. [3] Sabariego RV, Dular P. A perturbation approach for the modeling of eddy current nondestructive testing problems with differential probes. IEEE Trans Magn 2007;43:1289–92. [4] Panaitov G, Krause HJ, Zhang Y. Pulsed eddy current transient technique with HTS SQUID magnetometer for non-destructive evaluation. Physica C 2002;372: 278–81. [5] Sophian A, Tian GY, Taylor D, Rudlin J. Design of a pulsed eddy current sensor for detection of defects in aircraft lap-joints. Sensors Actuators A: Phys 2002;101:92–8. [6] He Y, Luo F, Pan M, Weng F, Hu X, Gao J, et al. Pulsed eddy current technique for defect detection in aircraft riveted structures. NDT&E Int 2010;43:176–81. [7] Dogaru T, Smith ST. Giant magnetoresistance-based eddy-current sensor. IEEE Trans Magn 2001;37:3831–8.
[8] Smith CH, Schneider RW, Dogaru T, Smith ST. GMR magnetic sensor arrays for NDE eddy-current testing. Rev Prog QNDE 2003;22:419–26. [9] Yang G, Tamburrino A, Udpa L, Udpa SS, Zeng Z, Deng Y, et al. Pulsed eddycurrent based giant magnetoresistive system for the inspection of aircraft structures. IEEE Trans Magn 2010;46:910–7. [10] Wincheski B, Yu F, Simpon J, Williams P, Rackow K. Development of SDT sensor based eddy current probe for detection of deep fatigue cracks in multi-layer structure. NDT&E Int 2010;43:718–25. [11] Ditchburn RJ, Burke SK, Posada M. Eddy-current nondestructive inspection with thin spiral coils: long cracks in steel. J Nondestr Eval 2003;22:63–77. [12] Ditchburn RJ, Burke SK. Planar rectangular spiral coils in eddy-current nondestructive inspection. NDT&E Int 2005;38:690–700. [13] Gilles-Pascaud C, Vacher F, Decitre JM Eddy current array probe for complex geometries, In: Proceedings of the review of progress in ICNDE 2006. [14] Ravat C, Woytasik M, Joubert PY, Bihan Y, Marchand C, Dufour-Gergam E et al., Study for the design of eddy current microsensor arrays for non destructive testing applications. In: Proceedings of the transducers—Solidstate sensors, actuators and microsystems conference; 2007. [15] Sadler D, Ahn C. On-chip eddy current sensor for proximity sensing and crack detection. Sensors Actuators A 2001;91:340–5. [16] Yamada S, Katou M, Iwahara M, Dawson FP. Eddy current testing probe composed of planar coils. IEEE Trans Magn 1995;31:3185–7. [17] Goldfine N, Clark D, Eckhardt H, Meandering Winding Test Circuit, US Patent 5,793,206; 1998. [18] Schlicker V, Walrath D, Weiss K, Goldfine N. MWM eddy current sensors for monitoring of crack initiation and growth during fatigue tests and in service. Int J Fatigue 2001;23:477–85. [19] Rosado L, Santos T, Piedade M, Ramos PM, Vilac- a P. Advanced technique for non-destructive testing of friction stir welding of metals. Measurement 2010;43(8):1021–30. [20] TEAM Benchmark 8—Non-destructive testing example: coil above a crack, CST AG; April 2009. [21] Rosado LS, Piedade M, Ramos PM, Santos TG, Vilac- a P. A reconfigurable digital signal processing system for eddy currents non-destructive testing. In: Proceedings of the IEEE I2MTC 2010—international instrumentation and measurement technology conference; 2010. p. 1284–9.
