EDDY CURRENT TESTING MANUAL ON EDDY CURRENT METHOD

AECL-7523

ATOMIC ENERGY OF CANADA UMITED

^ ^ 3 E & ^ W

L'ENERGIE ATOMIQUE DU CANADA LIMITEE

EDDY CURRENT TESTING MANUAL ON EDDY CURRENT METHOD Essais par courant de Foucault Manuel des rne'thodes d'essai par courant de Foucault Volume 1

V.S. CECCO, _£* point C is at a higher potential than point A. 2 4 This implies that when Z-y increases (i.e., coil moving across a defect) with Z2. Z3 & Z4 constant, the bridge voltage unbalance increases,and the opposite happens when Zq increases* It is this bridge unbalance characteristic that results in a plus-minus or 'figure 8' signal as the differential probe moves across a localized defect. This signal occurs independent of whether the two coils are wound in opposition or in addition. Z

4.2.2 Typical Bridge Circuit in Eddy Current Instruments Figure 4.3 illustrates a typical AC bridge used in eddy current instruments. It is similar to the bridge in Figure 4.2 except for two additional arms. In this bridge the probe coils are placed in parallel with variable resistors. The balancing, or matching of voltage vector phase and amplitude, is achieved by varying these resistors until a null is achieved. Potentiometer R2 balances the reactive component of the coils to make the phase angle of each coil circuit equal. Potentiometer R^ balances the resultant voltage with an equal voltage amplitude to null the instantaneous voltage between R^ and R£>

TEST COIL

REFERENCE COIL

Fig. 4 . 3 : Typical Bridge Circuit Used in Eddy Current Instruments

-37The Inductive voltage drop across each coll Is equalized by controlling the current passing through the colls. This Is done by varying potentiometer R2« However, when the test coll Inductance differs significantly from reference coil inductance, potentiometer R2 will have to be rotated to one extremity. This means less must be close to the capacitive reactance, X c . In most crack detectors this is in the range of 20 to 100 ohms. Crack detectors that operate at or close to resonance do not have selectable test frequencies. Crack detectors for non-ferromagnetic, high electrical resistivity materials such as Type 304 stainless steel typically operate between 1 and 3 MHz; those for low resistivity materials (aluminum alloys, brasses) operate at lower frequency, normally in the 10 to 100 kHz range. Some crack detectors for high resistivity materials can also be used to inspect ferromagnetic materials, such as carbon steel, for surface defects. Normally a different probe is required; however, coil impedance and test frequency change very little.

-43-

PROBE WITH L I F T - 0 F F = 0 . 1 mm

PRO8E WITH L I F T - O F F = 0 mm

SAMPLE WlTH

0.8

0 . 9

1.0

1.1

OSCILLATOR FREQUENCY.

DEFECT

1.2 _L f

g. 4.7; Meter Output with Varying Oscillator Frequency Crack detectors have a meter output and three basic controls: balance, lift-off, and sensitivity. BALANCING control is performed by adjusting the potentiometer on the adjacent bridge arm, until bridge output is zero (or close to zero). GAIN control (sensitivity) adjustment occurs at the bridge output. The signal is then rectified and displayed on a METER. Because the signal is filtered, in addition to the mechanical inertia of the pointer, the frequency response of a meter is very low (less than 10 Hz). LIFT-OFF CONTROL adjusts the test frequency (by less than 25%) to operate slightly off resonance. In crack detectors the test frequency is chosen to minimize the effect of probe wobble (lift-off), not to change the skin depth or phase lag. The set-up to compensate for probe wobble can be described with the help of Figure 4.7. Frequency is adjusted by trial-and-error to obtain the same output signal on the meter with the probe touching the sample and at some specified lift-off (normally 0.1 mm). At this frequency a deap surface defect will give a different reading on the meter, as shown in Figure 4.7. However,the meter output is a complex function of signal phase and amplitude, and cannot be used to reliably measure depth of real defects. Nor can they be used to distinguish between real and false indications such as ferromagnetic inclusions.

-444.4.3 Material Sorting and Conductivity Instruments Material sorting,or conductivity instruments,have a precalibrated meter output and have a unique way of compensating for lift-off.. Instruments for sorting of high resistivity materials (Type 304 stainless steel) use a fixed, high test frequency, normally between 200 and 500 kHz,and those for low resistivity materials (aluminum alloys),a low test frequency, between 20 and 100 kHz. They incorporate AC bridges and normally have two coils (one as reference). Coll impedance is in the range of 20 to 100 ohms. They either have bridge balancing or a zeroing control, to keep the signal on scale. GAIN CONTROL or sensitivity adjustment occurs at the bridge output. The signal is then rectified and displayed on a METER. LIFT-OFF compensation is normally pre-aet. Figure 4.8 explains how the probe-wobble (lift-off) Bignal is eliminated. The bridge is purposely unbalanced (by pre-set internal adjustment)* such that the unbalance point, P, is at the centre of curvature of the lift-off impedance locus, AB. The instrument meter reads a voltage proportional to the distance, FB 1 or PA 1 , from the chosen unbalance point to the impedance curves. The amplitude of this voltage remains constant with probe wobble but changes significantly for wall thickness (and resistivity) variations. In fact any signal that traces an impedance locus different from lift-off will change meter output.

PRESET UNBALANCE

Fig. 4.8: Unbalanced Bridge Method Showing Selection of Operating Point

*This is achieved by subtracting a signal equal to OP from the signal OA.

-45With this type of instrument only the magnitude of the impedance change is measured. This instrument is effective for conductivity and wall thickness measurement (and deep defects) and is simple to operate. It has only two basic controls: balance and sensitivity. 4.5

SEND-RECEIVE EDDY CURRENT SYSTEMS The "send-receive" eddy current method eliminates the temperature drift sensed by general purpose instruments. The flow of eddy currents is monitored by observing the effect of their associated electromagnetic fields on the voltage induced in an independent coil(s), Figure 4.9. The excitation or primary coil is driven with a sinusoidal current with constant peak-to-peak amplitude to obtain a constant magnetomotive force, (2.3)

sin

RECEIVE CO ILS

TEST ARTICLE

Fig. 4.9:

Send-Receive Circuit

-46Thia makes the excitation primary coil resistance. connected to a high input induced voltage V9 is not

magnetic flux $ independent of The secondary or receive coil(s) is impedance amplifier, hence the affected by receive coil resistance.

d* ** N

CO8

(lit

(2.5)

The wire resistance of both the excitation and receive coils can change, because of temperature, without affecting the output signals; temperature drift has thus been eliminated* Temperature independence makes this method useful for measuring resistivity, wall thickness and spacing between components. It has no significant advantage over the impedance method for defect detection, except in the through-wall transmission system discussed in Section 5.4. 4.5.1 Hall-Effect Detector Most send-receive circuits consist of one excitation (or driver) coil and one or more receive (or pick-up) coils. However, the induced magnetic flux 4>s can be measured with a Hall-effect detector rather than by monitoring the induced voltage V B across a pick-up coil, see Figures 2.2b and 2.2c.

Fig. 4.10;

Hall Detector Circuit

-47The induced voltage in a pick-up coil is proportional to the time rate of change of the magnetic flux and therefore is proportional to the test frequency, V

Pick-u P "

f

The Hall detector instead responds to the instantaneous magnitude of the magnetic flux, 0. This means the output voltage is independent of test frequency, making it useful for low frequency inspection (especially if the detector has to be small). The Hall detector works as follows: When direct current is passed through a Hall element, voltage (electric potential) is produced, perpendicular to current flow, see Figure 4.10. This voltage is proportional to the component of magnetic flux perpendicular to the element and the element surface area. This voltage is NOT from a change in element resistance. Hall elements as small as 1 mm square are commercially available. 4.5.2 Send-Receive Coils and Lift-Off Compensation General purpose "send-receive" instruments are similar to "impedance" instruments, as described in Section 4.4.1. The main difference is the method of balancing because of the different coil configuration. Most send-receive circuits consist of one excitation coil and two receive coils positioned symmetrically inside or outside the excitation coil. They can either be differential where both coils sense the test specimen or absolute where only one coil senses the test specimen, as shown in Figure 4.9. Although coil impedance is not important in send-receive instruments, the induced voltage is a function of number of windings and test frequency. Therefore their inductive reactance tend to be similar to coils used in impedance instruments. The sensing coils are wound in opposition so the excitation field induces no net voltage in the receive coils when they both sense the same material. In the presence of a defect, the voltage changes as each coil moves over it. Figure 4.9 illustrates a surface reflection type probe where both excitation and pick-up coils are on the same side of the test sample. However, the excitation coil and pick-up coils can be placed on opposite sides of the sample; this method is referred to as through-wall transmission. The two methods are compared in Section 5.4. The output signals in most send-receive instruments are the quadrature components of the secondary voltage. However, in some special purpose instruments, one output signal is proportional to amplitude and the other to phase of the secondary voltage (relative to primary voltage). They

-48compensate for LIFT-OFF as follows: if coil-to-sample spacing varies there is a large change in amplitude of the secondary voltage but little change in phase. The phase shift between the secondary and primary sinusoidal voltages Is measured at a voltage level V o slightly larger than zero, Figure 4.11. At this voltage the sinusoidal voltages have the same phase shift for aero lift-off as for maximum (perhaps 0.1 mm) lift-off. The voltage discriminator in these phase-shift measuring eddy current instruments trigger on the V o voltage pointfand therefore»the output signal for lift-off between 0 and 0.1 mm is minimized. Measurement of resistivity, wall thickness or deep defects can be made without lift-off noise. V(t) PROBE SIGNAL,

FIR.

4.6

4.11;

LIFT-OFF=O

PROBE SIGNAL,

L I FT - O F F = 0 . i mm

PROBE SIGNAL,

DEFECT

IN

TEST

ARTICLE

Secondary Voltage Waveform for Various Test Conditions

MDLTIFREQUENCY EQUIPMENTThe eddy current NDT method is sensitive to many test parameters, making It very versatile. However, one is usually only interested in a single parameter such as defects. Insignificant parameters such as changes in electrical or magnetic properties, the presence of dents or support plates in tube inspection and lift-off in surface probe inspection can mask defect signals. The multifrequency eddy current method was developed to eliminate the effect of undesirable parameters.

i

1

-49-

The response to various anomalies changes with test frequency. This allows a means of discriminating against unimportant changes. In multifrequency instruments, two or more frequencies are used simultaneously (through the same coil(s)). Coil current consists of two or more superimposed frequencies, i.e., the coil(s) is excited with more than one test frequency simultaneously. A three-frequency multifrequency instrument acts the same way as three separate (single-frequency) eddy current instruments. Band-pass filters separate the signals at each frequency. The discrimination or elimination process is accomplished by combining the output signals (DC signals) from individual frequencies in a manner similar to simultaneous solution of multiple equations. The elimination of extraneous signals is achieved by matching the signal at two test frequencies and subtracting. This process is continued for other unwanted signals using other test frequencies until the final output only consists of only the defect signal. A discussion of inspection results with multi-frequency is covered in Section 8.4. Multifrequency instruments have the same controls and functions as general purpose "impedance" type instruments, described in Section 4.4.1, with the addition of mixing modules. These modules are used to combine or subtract the output signals from each combination of frequencies. 4.7

PULSED EDDY CURRENT EQUIPMENT Faraday's Law states that eddy currents are induced in a conductor by a varying magnetic field. This magnetic field can be generated by passing sinusoidally varying current through a coil. However, the current can be of other waveforms such as a train of pulses. This method works only on the send-receive principle where the flow of eddy currents is monitored by observing the effect of their associated electromagnetic fields on the induced voltage of the receive coil(s). The voltage pulse is analyzed by observing its amplitude with time, Figure 4.12. To compensate for LIFT-OFF, the voltage is sampled at a preset time, tj. When the waveform is triggered (measured) at time tj_, the voltage for zero lift-off and maximum lift-off is the same, whereas the voltage waveform in the presence of a defect is different. This method is quite similar to the send-receive method described in Section 4.5.3. Therefore, by measuring the voltage at the appropriate crossing point, lift-off effect can be drastically decreased.

-50-

V(t)

DEFECT IN TEST ARTICLE LIFT -OFF = 0.1 mi*

I Fig. 4.12: Voltage Across a Pulsed Eddy Current Pick-Up Coll as a Function of Time The pulsed eddy current method offers another advantage. The pulsed driving current produces an inherently wideband frequency spectrum, permitting extraction of more selective information than can be determined from the test specimen by a single frequency method. Unfortunately, there is at present no commercially available instrument that operates on this principle. 4.8

SPECIAL TECHNIQUES Two old methods used to measure large coil impedance variations (greater than 5%) are the ELLIPSE and SLIT methods. These methods analyse the AC signal directly on an oscilloscope (without converting it to DC). They were mainly used for material sorting. They are obsolete methods and a detailed description is not warranted in this manual; a full description is contained in Reference 5. Another technique, MODULATION ANALYSIS, is also described in Reference 5. It works on the same principle as "frequency spectrum analysis" where a discrete frequency component of a waveform can be analysed without interference from lower or higher frequency noise. The inspection must be performed at constant speed (in fact it only works if there is relative motion between coil and sample). It is used in productionline testing at speeds up to 2 m/s or higher. It is a very specialized and complicated method and a detailed description is not warranted in this manual.

-51-

4.9

RECORDING EQUIPMENT During inspection, eddy current instruments and recording equipment are typically connected as in Figure 4.13. The eddy current signal is monitored on a storage CRT (cathode ray tube) and recorded on X-Y and two-channel recorders. Recording on an FM tape recorder for subsequent playback is also common. The important characteristic of these recording instruments is FREQUENCY RESPONSE, or speed response, which limits inspection speed. Section 4.4.1 indicated general eddy current instruments have a frequency response of 100 t;o 300 Hz, limiting the inspection speed to 0.25 m/s. To be compatible, recording instruments must have the same or higher frequency response.

X-Y STORAGE MONITOR

O

EDDY CURRENT INSTRUMENT

PROBE

X? ?Y

o

X

i

X-Y RECORDER

Fig. 4.13; Equipment

6

O

X ,Y 2-CHANNEL CHART RECORDER

Block Diagram of Eddy Current

FM TAPE RECORDER

Monitoring

-52-

X-Y Recorders Signal analysis for signal discrimination and defect depth estimation is normally done on X-Y signal patterns. The CRT storage monitors have a frequency response of at least 1 kHz and therefore do not restrict maximum inspection speed. However, to obtain a permanent visual record of the signal, it must be recorded on X-Y recorders. The fastest recorders have a speed of response of 8 Hz for small signals. This drastically limits inspection speed if used on-line. It is therefore only used in the laboratory or to record playback from tape recorders (this is done by recording at the highest tape speed and playing back at the lowest, a factor of 8:1 for most tape recorder). One solution to on-line recording of X-Y signals is to photograph the CRT display; however, this is not practical for recording many signals. Another solution is to use storage monitors with hard copy (paper output) capability. These exist commercially but require custom-made control units. They have a frequency response of 1 kHz or higher. Strip Chart Recorders Recording X and Y signal components against time is useful in locating defects and determining their length. Common two channel ink-pen strip chart recorders have a speed response of approximately 100 H z . At maximum inspection speed (0.25 m/s) the recorded signal will decrease in amplitude and be slightly distorted. Ink-ejection strip chart recorders have a speed response of 1 kHz. These recorders are not readily available in North America and use a lot of paper. Ultraviolet light recorders have a speed response higher than 1 k H z , but require special paper. These recorders are rarely used in eddy current testing.

p Ji

FM Tape Recorders Tape recorders allow storage of eddy current signals (on magnetic tape) for subsequent retrieval. They have a frequency response proportional to recording speed. The lowest recording speed is 24 mm/sec (15/16 ips) giving a frequency response of 300 H z , and the fastest, 380 mm/s (15 i p s ) , will respond to 4.8 kHz.

• r

f ; • \

-53-

4.9.1 Frequency Response Eddy current instruments an^ recording instrumentation have limited frequency response. This means they require finite time to respond to an input signal. Frequency response, sometimes called speed of response, is defined as the frequency at which the output signal falls to 0.707 (-3 dB) of the maximum input signal. A test coil with an effective sensing width w passing over a localized defect at a speed s will sense the point defect for a duration of w/s seconds. This signal is approximately equal to one wavelength with a frequency f - s/w where 3 is speed in mm/s

hertz

(4.6)

and w is width in mm.

For example, at a probe speed of 0.5 m/s and probe sensing width of 2 mm, f » 250 hertz. If the instrumentation has a frequency response of 250 hertz, the output signal is reduced to 0.707 the input signal and the X-Y signal is distorted. If the instrumentation frequency response is 500 hertz, the output signal decreases only slightly. For this example, the eddy current instrument should have a frequency response equal to or greater than 500 hertz to obtain undistorted signals. Or inversely, if the instrument frequency response is only 250 hertz, the maximum inspection speed should be reduced to 0.25 m/s 4.10

SUMMARY Basic eddy current equipment consists of an alternating current source (oscillator), voltmeter and probe. When the probe is brought close to a conductor or moved past a defect, the voltage across the coil changes and this is read off the voltmeter. The oscillator sets the test frequency and the probe governs coupling and sensitivity to defects. For effective purchase or use of an eddy current instrument, the following information is needed: (a) type of instrument: impedance, send-receive, crack detector, etc. (b) type of outputs: single (meter) or quadrature (X-Y) component outputs (c) test frequency (d) type of lift-off compensation. Most eddy current instruments use an AC bridge for balancing but use various methods for lift-off compensation. Send-receive instrument should be used for accurate absolute measurements in the presence of temperature fluctuations. Multifrequency instruments can be used to simplify defect signals in the presence of extraneous signals.

-54Eddy current instruments and recording equipment have a finite frequency response limiting the Inspection speed to normally 0.25 m/s. Moat Instruments tolerate probe impedance between 10 and 200 ohms. Crack detectors operate close to coil-cable resonance. The resonant test frequency Is given by

(4.4a)

f r - l/2ir/LC

where L is coll Inductance in henries and C is cable capacitance in farads. The lift-off signal is minimized by adjusting the frequency (slightly off resonance) until zero and a small probe lift-off gives zero output signal. High test frequencies are normally used to inspect for shallow defects in high resistivity or ferromagnetic materials. Low test frequencies are used for detecting deep defects or inspecting good conductors. Crack detectors have a meter output, and cannot be used to reliably measure defect depth. 4.11

WORKED EXAMPLES

4.11.1 Impedance at Resonance PROBLEM:

In a parallel L-C circuit, inductance is 80 ? 10-6 henries, capacitance is 5 x 10~ 9 farads and resistance is negligible. Calculate (a) resonant frequency, (b) inductive reactance and (c) capacitlve reactance.

SOLUTION: (4.4a)

(a) f 252 kHz 2TT

(b)

6

9

^(80 x 10~ ) (5 x 10~ ) - 2irfL

Inductive Reactance,

2TT X 2 5 2 x 10 3 x 80 x 10" 6

(c)

Capacitive Reactance, X

C

"

1 2ir x 252 x 1 0 3 x 5 x 10" 9

(3.4b)

- 126.5 ohms

l/2irfC » 1 2 6 . 5 ohms

(3.5)

-55CHAPTER 5 - TESTING WITH SURFACE PROBES 5.1

INTRODUCTION The goal of this chapter Is to present a practical approach to eddy current Inspections using surface probes. The emphasis is on test variables such as test frequency, probe size and type; these are normally the only variables an inspector has at his control. These selections are usually determined by skin depth considerations, defect size, and probe size. Impedance graphs and the Characteristic Parameter are included because the/ are tools that an inspector should not be without. A thorough understanding of impedance graphs is essential to manipulate test conditions to minimize and/or to •cope with undesirable test variables. Erroneous conclusions are often made by persons who do not have a working knowledge of impedance graphs. The scope of the approach to an eddy current inspection can be very broad; a successful outcome usually depends on the proper approach.. When planning an inspection the first questions that must be answered before proceeding are; For what type of defects is the inspection being conducted? If the expected defects are cracks, how big are they? Do they have directional properties? What is the minimum acceptable defect size? Does the material have ferromagnetic properties? Other variables will, of course, influence the test but these questions must be answered in order to select an appropriate probe size and test frequency.

5.2

SURFACE PROBES The eddy current probe plays two important roles: it induces eddy currents, and senses the distortion of their flow caused by defects. Sensitivity to defects and other variables in the test article can be affected by probe design. This is achieved by controlling direction of eddy current flow, by controlling the coil's magnetic field, and by selecting an appropriate coil size. The effects of undesirable material variations and/or variations in probe to test article coupling (lift-off) can often be decreased by using multiple coils. A surface probe, as the name implies, is used for inspecting surfaces, flat or contoured, for defects or material properties. Defects can be either surface or subsurface. (Surface defects are those that bre-\k through, or originate at the surface - typically cracks, voids, or inclusions: a subsurface defect does not break the surface and is therefore not visible). Other names used for variations of surface probe designs are pancake probe, flat probe, spring probe or coil, spinning probe, and pencil probe.

-56-

5.2.1 Probe Types Simple Probes Surface probe designs can vary from a simple, single coil attached to lead wires, to complex arrangements, as shown in Figure 5.1. Most eddy current instruments require two

ZIRCONIUM ALLOV

Fig. 5.1:

Surface Probes

slmilai coils to satisfy their AC bridge network as discussed in Chapter 4. If only one coil senses the test material, it is an absolute probe; if both coils sense the test material, it is a differential probe. The simple probe in Figure 5.1(a) is therefore undesirable because a second coil or electrical device with similar impedance will be necessary for bridge nulling. An exception would be in the use of Crack Detectors; these instruments operate with an internal balancing circuit (see Section 4.2.3). A better arrangement is shown in the pencil probe of Figure 5.1(b). This probe incorporates a second coil (reference) mounted far enough from the test article that it will not be influenced by it. The tvo colls have the same impedance when the probe is balanced In air, but will change relative to each other when the test coil is coupled to a sample. However, the degree of coupling is usually small because of the inherent small size of pencil probes so the coils still match well enough for most instruments over a reasonable frequency range. The probe shown has ferrite cores; ferrite is used for three reasons:

-571. 2. 3.

higher inductance from a given coil size, small surface area in contact with the material, the coil can be further from the contact surface providing greater wear protection.

A further improvement in reference coil arrangement is shown in Figure 5.1(c); it is attached to a disc whose properties are similar to the test material. With this arrangement the relative impedance of the two coils will not be affected by test frequency. The probe shown in Figure 5.1(d) is a spring loaded type designed to minimize lift-off. The shoe provides a broad area for squarely positioning the probe on a flat surface, while the spring maintains probe contact at constant force. Figure 5.1(e) shows a probe used for inspecting large diameter tubing. The probe can be rotated and/or moved axially. The design shown incorporates a replaceable wear cap. Other Probe Designs A multi-coil array as shown in Figure 5.2(a) is useful for inspecting tubes. This type of probe could detect defects SURFACE COILS TEST TUBE

TORROIDAL REFERENCE COIL PROBE CENTERING DISCS

TEST COILS

U) (a) DIFFERENTIAL SURFACE PROBE MUUI SURFACE -COIL PROBE

. FERROMAGNETIC CORE •COILS

FIELD

SENSING CDII

(b)

(H)

GAP PROBE

LIFT-OFF COMPENSATING PROBE

Fig. 5.2:

Special Surface Probes

-58-

that would not be detected by a conventional circumferential coil (discussed in Section 7.5). A gap probe, Figure 5.2(b), uses ferromagnetic material to shape the magnetic field. The field is confined by the core causing eddy currents to flow in circular loops perpendicular to the flux lines. A differential configuration is shown in Figure 5.2(c); the two coils are placed side-by-side. Both coils have high sensitivity to localized variations but tend to cancel out the effect of lift-off, gradual material variations, or ambient temperature changes. A lift-off compensating probe is shown in Figure 5.2(d); this probe combines the signals from two coils to effectively rotate the defect signal relative to the lift-off signal. Therefore, even on "rough" surfaces, shallow defects can be detected.

SEND . COIL \ (DRIVER COIL! \

I

TEST ARTICLE RECEIVER COIL

( a)

PICK-UP COILS (WOUND OPPOSING EACH OTHER)

ELECTRICAL CONNECTIONS (c)

Fig.

5.3:

Send-Receive Probes

-59-

Send-Recelve Probes Figure 5.3(a) shows a through-transmission probe arrangement. Current flowing in the SEND coil produces a magnetic field, part of which is transmitted through the test article. The field is detected by the RECEIVER coil, inducing a voltage. There will be no signal variation from the receiver coil when a defect-free test article is moved anywhere between the two coils as long as the coil-to-coil spacing remains constant. Figure 5.3(b) shows a reflection-type probe arrangement. The probe consists of a large send coil which generates a field, and two small receiver coils wound In opposite directions, as mirror images to one another, as shown in Figure 5.3(c). With the probe in air, net output is zero. However, If one end is placed near a test article, the field differs at the two ends, and a net voltage appears across the two coils. 5.2.2 Directional Properties Eddy currents are closed loops of induced current circulating in a plane perpendicular to the direction of magnetic flux. Their normal direction of travel is parallel to the coil winding and parallel to the surface. See Figure 5.4. Pancake type surface probes are therefore insensitive to poor bonding of coatings and flaws parallel to the surface of a sample.

SURFACE CRACK

EDDY CURRENTS LAMINAR CRACK TEST PLATE

EDDY CURRENT FLOWS PARALLEL TO COIL WINDINGS - POOR SENSITIVITY TO LAMINATIONS SURFACE CRACK IN PLATE

ZERO SENSITIVITY AT CENTRE OF COIL

Fig. 5.4:

LOW SENSITIVITY PARALLEL TO WINDINGS

MAXIMUM SENSITIVITY ACROSS WINDINGS

Directional Properties of a Surface Probe

-60-

When testing for flaws such as cracks, It Is essential that the eddy current flow be at a large angle (preferably perpendicular) to the crack to obtain maximum response. If eddy current flow is parallel to the defect there will be little or no disruption of currents and hence no coil impedance change* When testing for flaws parallel to the surface, such as laminations, a horseshoe shaped probe (a gap probe with a very large gap) has reasonable sensitivity. 5.2.2.1 Sensitivity at Centre of a Coil Probe impedance changes with coil diameter, as will be discussed further in Section 5.5. A simplified derivation of this diameter effect is derived below, for the case of no skin depth attenuation or phase lag and long coils. From Faraday's Law,

* . - • • »

The magnetic flux density, B, is approximately constant across a coil's diameter, hence (j> » BA - (B)(irr2) where r is radial distance from centre of probe; therefore,

or a r

N(turns)

Ac

-61-

Reslstance to flow of current Is proportional to flow path length and resistivity and inversely proportional to crosssectional area, A c , R

2urp s unit depth x unit width

or

R

s

CC

Y

s - V /Z

Since

by Ohm's Law

S

and

Z

-V.

1 + (U)L)2 - R n ,

at low test frequency

and no skin depth effect, therefore, s or

I

s ince

*s

s

R~ "

~

« r = - I

from Lens's Law, it follows s

that

Therefore, eddy current flow and its associated magnetic flux are proportional to radial distance from the centre of a coil. Hence no current flows in the centre (r » 0) and there is no sensitivity to defects at the centre of a coil. 5.2.3 Probe Inductance The factor governing coupling and induced voltage in test material is the magnetic flux surrounding the coil. The total magnetic flux („) is proportional to probe inductance (L) and current (I), i.e., 4>poc LI. In most eddy current instruments excitation current is kept Reasonably constant (in the milliampere range) but probe inductance could vary by a factor of one thousand. The most important aspect of inductance is that probe impedance, which is a function of inductance, must be compatible with the instrument and signal cable, and 6

Arctan -~ K

2 TfL when f is in hertz, L in henries and R is where X coil wire resistance in ohms.

- 62 -

TABLE 5.1

o - 1.6 mm

SURFACE COIL IMPEDANCE

» 3.2 mm

D

- 6.3 mm

•12.7 mm

D >= 25.4 .mm 0

L • 0.27 yH R - 0.2 ft

L - 0.54 yH R - 0.1ft

L - l.i yH L - 2,,lyH R - 0.05 ft R - 0.02ft

40 AWG (0 .080 mm)

34 AWG

28 AWG

(0 .16 mm)

(0 .32 mm)

22 AWG (0 .64 mm)

16 AWG (1 . 3 mm)

L - 3,0

L - 4.3 yH

R - 0.01 ft

N - 21

L >• 1.5 R - 24 R >• 0.06

43 AWG (0 .056 mm)

37 AWG (0 .11 mm)

31 AWG (0 .23 mm)

25 AWG (0 .45 mm)

19 AWG (0 .91 mm)

L >• 5 . 8 R >• 4

L - 12 R - 2

L - 23 R - 1

L - 47 R - 0.5

L >• 94 R >« 0.3

46 AWG (0 040 mm)

40 AWG (0 .080 mm)

34 AWG (0 .16 mm)

28 AWG (0 .32 mm)

22 AWG (0 .6 4 mm)

L •• 11 R - 9

L - 23 R - 3

L « 45 R - 2

L - 90 R - 0.9

L »• 180 R >• 0.5

48 AWG (0. 031 mm)

41 AWG (0 .071 mm)

36 AWG (0 .13 mm)

29 AWG (0 .29 mm)

23 AWG (0 57 mm)

L •' 24 R -• 1 7

L - 49 R - 8

L - 97 R - 4

L - 195 R - 2

R ••

49 AWG (0. 028 mm)

43 AWG (0 .056 mm)

37 AWG (0 .11 mm)

31 AWG .23 mm)

25 AWG (0. 45 mm)

N - 50

N - 98

N « 136

L »•

390 1

N » 200

i

PA

.

•f = Dj =0.2 D o I

-63-

The self-inductance of a long coil (solenoid) can be calculated from the equation

L Q -= 4ir x 1 0 " 1 0 y r N2A/Jt

where

LQ Hr A &

is is is is

(5.1a)

self-inductance in henries relative permeability of core (normally -1.0) coil's planar surface area, millimetres2 coil length, millimetres.

This formula is a good approximation for coils of length/diameter ratio greater than 10. For a short coil, end effects will reduce inductance because of lower flux at coil ends. The N 2 term remains since N enters in N $_ (total number of flux linkages) and again since _ itself is proportional to N. The following approximate equation can be used to calculate inductance of short coils: L Ox,

- 4m

r N2Un ~

where r is mean coil radius •

and

- 2) 1 0 " 1 0

(5.1b)

JAI

D +D ;——

, mm

K • 0.112 (2£ + D Q + D ) , mm

Most eddy current instruments will operate over a fairly broad range of probe impedance (and probe inductance) without substantial reduction in signal-to-noise ratio and signal amplitude. An instrument input impedance of 100 ohms is typical, although any impedance between 20 and 200 ohms is generally acceptable, unless test frequency is too close to probe-cable resonance; see Section 5.9. Exact probe inductance calculations are therefore not essential. To facilitate impedance calculations, Table S.I has been prepared. This table lists coil inductance and resistance (with probe away from test material) for various outside diameters and number of coil turns, keeping both the inside diameter and coil length equal to 0.2 times the outside diameter. Wire diameter is chosen to fill available coil cross-sectional space. Using this table and the knowledge that Inductance, L « N2D2

(5.2)

where N is number of turns of wire and I) is average coil diameter, one can usually make a reasonable estimate of wire size and number of turns required to achieve a particular inductance.

NORMALIZED DEFECT SIGNAL AMPLITUDE

NORMALIZED DEFECT SIGNAL AMPLITUDE Vx./Vx=l G9

CD

o

7

ii

en

^ ;

/ r

/

/

o l-t (D HI CO

/

ro CJI (D

3

CO —I

CO

o CO Ui

7 (•

DEFE 2 mm

CO

y k

n>

/

CJI

; \

cD rn pn

—|NX VC v

i"

ro

A)

OS>

II

CJI

I

"o

c

OS?

II

CJI

2 C

\A 7/A

o rt er

:i

11

h

'

R) o

en

LIF T-OFF

^—|

O

-65-

5.3

PARAMETERS AFFECTING SENSITIVITY TO DEFECTS During eddy current inspection one must be aware of the limitations of the technique and should take maximum advantage of its potential. Although sensitivity to deep surface defects is excellent, sensitivity to deep sub-surface defects is very poor. A subsurface defect only 5 mm from the surface is considered very deep for eddy current test purposes• There are two factors that contribute to this limitation. The skin depth effect causes eddy currents to attenuate with depth depending on the aaterial properties and test frequency. This effect is normally minor and can be controlled (within limits) by reducing test frequency. The predominant effect (rarely mentioned) is the decrease in magnetic flux, and consequently eddy current density, with depth because of the small diameter of most practical probes. One can increase penetration by increasing probe diameter, but as a consequence sensitivity to short defects decreases. One could optimize sensitivity if defect length is known; however,the maximum depth of detectability is still very small. Unlike ultrasonic inspection where a defect is detected many transducer diameters away, eddy current testing is limited to detecting defects at a depth of less than one probe diameter. It is this effect of probe diameter that limits most volumetric eddy current inspection to materials less than 5 mm thick. In following subsections, limitations are discussed and empirical examples presented.

5.3.1 Sensitivity with Lift-Off and Defect Depth There is a decrease in sensitivity to defects as a coil is moved away from the surface. This is caused by the decrease in magnetic flux density with distance resulting from finite probe diameter. Figure 5.5(a) shows the extent of this decrease for three probes of different diameters. Note,for example, the sensitivity of the smallest probe (5 mm diameter) decreases a factor of four when moved about 1 mm from the surface. This loss of sensitivity with distance will also apply to defects in a solid, in addition there will be a decrease due to skin depth attenuation. Figure 5.5(b) illustrates the decrease in signal amplitude with subsurface defect depth without skin depth attenuation (solid lines) and with skin depth attenuation (dashed lines). With large skin depths (low test frequency) the decrease in subsurface defect sensitivity with depth is similar to the decrease in sensitivity with distance for surface defects shown in Figure 5.5(a). This implies magnetic flux density decreases with distance from the coil in air as in a solid (without skin depth attenuation).

-66At a typical test frequency, where one skin depth equals defect depth ( Change in phase of secondary voltage as probe is moved over a r'efect. This is approximately the phase measured by some send-receive eddy current instruments without X-Y outputs. 5. 0 , Phase between the voltage signals obtained from LIFT-OFF and a crack or void. It is related to PHASE LAG '3 • explained below. (0^ 1 B about double the phase lag.) 0 3 is used to estimate defect depth during E T . 6. g, PHASE LAG (not shown in Figure 5.16) of eddy currents below the surface relative to those at the surface. It was derived in the eddy current density equation Chapter 2, i.e. g • x/6for semi-infinite p l a t e s , where x is the distance below the surface and B is in radians. 7. 0,, Many eddy current instruments have a PHASE knob by which the entire impedance voltage plane display can be rotated. It is coataon practice to rotate the display to meke LIFT-OFF horizontal. (On an eddy current instrument display, absolute orientation of inductive and resistive axes may be unknown). 8.0_, Phase between inductive voltage and current in a circuit; 0 - 90° •

• K £ _• ™ • g _ I H B gj m

• II m If I; I:

I! I;

5.8

SELECTION OF TEST FREQUENCY

5.8.1 Inspecting for Defects The first question that must be answered before proceeding with a n inspection is: For what type of defects is the inspection being done? If the defects are cracks: What is the smallest defect that must be detected? Are the cracks surface or subsurface? A r e they likely to be laminar cracks or normal to the test surface? A single general inspection procedure to verify the absence of any and all types of defects often has little merit. Inspections often require t w o or more teit fraquencias and/or different probes to accurately identify dafacta. Taat frequency can ba aalactad without knowledge of the characteristic parameter, P c , or tha oparatlnf point on tha impedance graph. It ahould ba chosen for good discrimination batwaan dafacta and othar varlablna. Tha moat troublasoma variabla la LIFT-OFF variations, a o aaparatlon of dafacts from lift-off ia tha foraaoat consideration.

I! a?

f I r|| p r• '

-79-

Only the skin depth equation has to be used, 6 -

nun

(2.13a)

A test frequency where 6 is about equal to the expected defect depth provides good phase separation between lift-off and defect signals. Figure 5.17 illustrates the display on

COIL

LIFT-OFF

SURFACE CRACK SUBSURFACE VOID (A) SUBSURFACE VOID (B)

\ 1X1

4&W' \

SURFACE CRACK

SUBSURFACE VOID (A) SUBSURFACE VOID (B)

(a)

INCREASING LIFT-OFF X -Y DEFECT SIGNALS

(b)

Fig. 5 . 1 7 ; T y p i c a l Response Signals for Two Types of Defects

an eddy current instrument monitor as a probe passes over surface and subsurface d e f e c t s . Test frequency is such that $ equals depth of deepest defect, and instrument controls are selected such that a signal from lift-off is h o r i z o n t a l . Note the difference in signal amplitude and angle relative to lift-off of subsurface voids A and B . This results from skin depth attenuation and phase lag. I f , during inspection, a signal indicating a defect is observed, test frequency may be altered to verify whether the signal represents a real defect or the effect of another v a r i a b l e . This discussion is expanded in the next chapter under Signal A n a l y s i s .

-80-

5.8.2 Measuring Resistivity Resistivity can be measured at small localised areas or by sampling a larger volume of a test article to determine bulk resistivity. The volume of material interrogated depends on probe size and test frequency. For bulk measurements a large probe would be used and a low frequency to maximize penetration. The skin depth equation is again used to estimate depth of penetration at the test frequency. Electrical resistivity measurement is a comparative technique; reference samples of known resistivity must be used for calibration. Variables that affect the accuracy of resistivity measurement are lift-off, temperature, and changes in the flow of eddy currents in test articles not related to electrical resistivity (such as cracks, thickness and surface geometry). For best discrimination between resistivity and other variables the operating point on an impedance graph should be considered. Figure 5.12 illustrated the effect of test frequency on normalized probe impedance. At the top of the graph the angle, between lift-off variations and the resistivity curve, is small. Moving down the curve the angle, separating the two variables, increases towards the knee with no appreciable change beyond that. However,small lift-off variations, at the bottom of the curve, produce a large impedance change. The best operating point is somewhere between the two extremes, near the knee of the impedance curve.

INCREASING

•

REFERENCE SAM H.E PEDANCE FOIIi

RESISTIVITY

p - 5S / « " em

MPEDANCE POINT OF UHKNniM ^

|

-_ —

INC (EASING ' LI FT-OFF

s

REFERENCE " SAMPLE MONITOR DISPLAY EODV CURRENT INSTRUMENT MONITOR DISPLAY

RESISTANCE M K M K E MMM • KIISTJWn fFFICT

Fig. 5.18;

Resistivity Measurement and the Impedance Graph

-81-

Figure 5.18 shows the method of manipulating test conditions to best deal with lift-off. Figure 5.18(a) shows the resistivity impedance curve with a frequency and probe selected to operate near the knee. Figure 5.18(b) is an enlarged section of the curve rotated so lift-off signals are approximately horizontal. This is the view on an eddy current instrument monitor. Next consider temperature effects. First, test article resistivity will be a function of temperature so test sample and standards should be at uniform temperature. A greater potential error is in probe wire resistance, R-QQ . The coil wire resistance is a part of the probe impedance circuit, so variations in temperature which affect coil resistance will appear as an impedance change. For greatest accuracy, the inductive reactance, Xj.» should be large compared to coil Wire resistance; X^/R.. > 50 is desirable. Obviously this condition is not easily satisfied at low test frequencies where inductive reactance is low. One solution is to use a large diameter probe cupped in ferrite. The large diameter and ferrite cup will both increase X L / R Q ^ Another solution is to use a Send-Receive instrument. Such an instrument has a high input impedance, sensing only voltage changes in the receive coil. Coil wire resistance is insignificantly small in comparison to instrument impedance and therefore has no effect. Consider next the effect of changes in eddy current path not related to electrical resistivity. If the test is supposed to be a measurement of electrical resistivity, thickness should not influence the signal. The skin depth equation must again be used. Test article thickness should be equal to or greater than three skin depths, t > 3 { ,

t > 3 x 50jf ,

r- 22500

t

is thickness, p is resistivity centimetres, and f is frequency.

f

where

, mm

>—-

Hz

in microhm-

Other sources of signals are edge effects and surface geometry. When the test article's edge is within the probe's magnetic field, an increase in resistance to eddy current flow will be detected. Edge effect can be reduced by probe design, such as a ferrite cupped probe, or by increasing test frequency.

n

-82-

If the surface of the test article Is contoured, the magnetic flux coupling will differ from that of a flat surface and a correction factor may be required. Cracks or voids are usually less of a problem. The signal from a crack will be very localized whereas resistivity variations are usually more gradual. The best procedure to determine ii a localised signal is from a change in resistivity is to rescan with a smaller probe at higher and lower frequency (at least three times and one third the test frequency). The angle between the signals from lift-off and resistivity should vary only slightly whereaB the angle between lift-oft and defect signals will Increase with frequency. An example of resistivity variations in a zirconium alloy, due to a change In oxygen concentration, is shown in Figure 5.19.

ii

TEST ARTICLE WIDTH

f! f! fi

X,VOLTS (a) X-Y DISPLAY OF COIL IMPEOANCE FROM CHANGE IN ELECTRICAL RESISTIVITY

n

(b) MODIFIED C-SCAN DISPLAYING Y-COMPONENT OF COIL IMPEOANCE VECTOR FROM A CHANGE IN ELECTRICAL RESISTIVITY Fig. 5.19: Eddy Current Signals from a Change In Electrical Resistivity on the Surtace of a Zr-Nb Test Article." Test Frequency • 300 kHz.

I I I I I I I I I I

-835.8.3 Measuring Thickness Test frequency should be chosen so 'lift-off and 'change in thickness1 signals are separated by a 50° phase angle, see Figure 5.2O(a). This frequency can be calculated using the skin depth equacion. A 'reasonable approximation for thin sections is when obtained when (5.6)

0.8

t/6 which converts to f where

6 t P f Vr

1.6

p/t5

(5.7a)

kHz

is skin depth, mm is test article thickness, mm is electrical resistivity, microhm-centimetres is frequency, kHz is relative permeability (y • 1 for nonferromagnetic material).

In testing thick material, this equation can similarly ba used to choose a test frequency to separate lift-off and subsurface defect signals by 90°. Fquation 5.7(a) can be used by replacing t with x, f - 1.6 p/x2

(5.7)

kHz

where x is depth of subsurface defect.

1

s I I I I I I I

INCREASING RESISTIVITY _|

rmcKNESs \

LIFT-OFF

BA ANC M NT FOf KOI IM1 -TH CKNI SS-

J i HERE»SIN G HICK IESS

(b) EDDY CURRENT INSTRUMENT MONITOR DISPLAY

RESISTANCE (=> IUPEDSNCE GRAPH - RESISTIVITY AND THICKNESS EFFECT

Fig. 5.20:

Thickness Measurement and the Impedance Graph

-84-

Conventlonal thickness measurement Is to display the lift-off signal horizontal (along the X-axis) and use the vertical signal (along the Y axis) to measure thickness, see Figure 5.20(b). It the signal on the instrument monitor is set to move from right to left as the probe is moved away from the test article, a vertical movement up or down denotes decreasing and increasing thickness respectively. 5.8.4 Measuring Thickness of a Non-Conducting Layer on a Conductor An insulating layer will not conduct eddy currents so measurement ox its thickness is essentially a lift-off measurement (provided it is non-ferromagnetic), i.e. the distance between the coll and test article. At high test frequency a small variation in lift-off produces a large change in probe impedance as shown in the impedance graph of Figure 5.9. To minimize the signal from variations in the base material, the test should therefore be done at the highest practical frequency* The maximum frequency would be limited by probe-to-instrument impedance matching, cable resonance problems and cable noise. The measurement is a comparative technique so standard reference thicknesses must be used for calibration. 5.8.5 Measuring Thickness of a Conducting Layer on a Conductor Measurement or the thickness of a conducting layer on a conducting test article can be done provided there is a • ditterence in electrical resistivity (Ap) between the two. The measurement is essentially the same as the thickness measurement described in Section 5.8.3. There is one important difference; variables in the base plate, in addition to the variables in the layer, will affect the signal. Figure 5.21(a) shows a computer simulation of a layer thickness measurement. The model shows the magnitude and direction of variables when attempting to measure a layer (clad l ) , nominally 0.75 mm thick, with resistivity p •= 3 yfi.cm on a base (clad 2) with resistivity 5 yfl.cm. The plot is part of a normalized impedance graph. In addition to material property variables, the parameter of space (gap) between the layers is shown as well as the effect of an increase in test coil temperature. At 10 kHz, t/S Is 0.8 and, as predicted, the angle separating signals from

-85-

E00» CURRENT 1HPEDINCE PLANE

1

I

F

I

I

I

I

I

I

.1

!I I

RESISTKITr I c 3 i 2 0 t u f l - .« HIR CUf

\

t («)

I

-ULJ»

0 TO .37 m

RESISTIVITY 3 -- 5 ! 101 nCl-

RJNGE OF VARIABLES SHOIN IK COKPUTOR PLOTS (D)

,170

7

.0S2O

l'"/1 .0940 ,0H0

I

I

I

.OMO .MOO .0020 .0140

I

l

l

.OHD

.0110

I .0)00

IIOWJLIZED RESISTANCE. _ ! ! L

Fig. 5.21; Computer Simulation of a Multi-Layer Sample

lift-off and layer (clad 1) thickness is about 9 0 ° . Unfortunately, so are the signals from test coil temperature, gap, and resistivity of the base (clad 3 ) . Some of these parameters can be discriminated against at higher and/or lower test frequencies. 5.9

PROBE-CABLE

RESONANCE

Probe-cable resonance must be considered when operating at high test frequencies and/or using long signal cables, e.g.. frequencies greater than 100 kHz and cables longer than 30 m. Most general purpose eddy current instruments cannot operate at or close to resc-ince. Probe-cable resonance can be modelled as shown in Figure 4.5. In simple terms, resonance occurs when inductive reactance of the coil equals capacitlve reactance of the cable, i.e. when ti)L - 1/oiC where

to

is angular frequency, in radians/secpnd, L is coil inductance in henries and C is total cable capacitance In farads.

-86-

Transforming this equation and substituting resonance occurs when frequency is f r - l/2ir i/LC

U)=27rf

shows (4.6a)

This approach is sufficiently accurate for most practical applications. A more rigorous approach to resonance is presented in Section 4.3. Resonance is apparent when a probe and cable combination, which balances at a low frequency, will not balance as frequency is increased. At the approach of resonance, the balance lines on the eddy current storage monitor will not converge to a null. The two balancing (X and R) controls will produce nearly parallel lines rather than the normal perpendicular traces, on the storage monitor. A number of steps can be taken to avoid resonance: 1. 2. 3.

Operate at a test frequency below resonance, such that f is less than 0.8f r . Select a probe with lower inductance. (Since f r is proportional to 1/ /U7 inductance must be decreased by a factor of four to double resonant frequency). Reduce cable length or use a cable with lower capacitance per unit length (such as multi-coax cables). This will raise the resonance frequency since capacitance is proportional to cable length and f r is proportional

to 1/ /cT 4.

Operate at a test frequency above resonance, such that f is greater than 1.2f r . However, above resonance the sensitivity of all eddy current instruments decreases rapidly with increasing frequency because capacitive reactance (X c »l/ 0)C) decreases, and current short circuits across the cable, rather than passing through the coil.

5.10 SUMMARY Test probes induce eddy currents and also sense the distortion of their flow caused by defects. Surface probes contain a coil mounted with its axis perpendicular to the test specimen. Because it induces eddy currents to flow in a circular path it can be used to sense all defects independent of orientation, as long as they have a component perpendicular to the surface. It cannot be used to detect laminar defects. For good sensitivity to short defects, a small probe should be used; probe diameter should be approximately equal or less than the expected defect length. Sensitivity to short subsurface defects decreases drastically with depth; even a 'thin' 5 mm sample is considered very thick for eddy currant testing.

-87-

The analysis of eddy current signals is the most important ana unfortunately the most difficult task in a successful inspection. A thorough understanding of impedance graphs is essential to manipulate test conditions to minimize undesirable test variables. The characteristic parameter for surface probes is used to locate the operating point on the impedance diagram. It is given by P - 7.9 x 10"4 i2 f/p c

(5.5)

where r is mean radius, mm; f is test frequency, Hz; and p is electrical resistivity, microhm-centimeters. The criterion for defect detection with impedance plane instruments is phase discrimination between lift-off noise and defect signals. Test frequency is chosen such that 'lift-off and 'change in wall thickness* signals are separated by a 90° phase angle. This can be derived from the following equation: f - 1.6 p/t2 , kHz

(5.7)

where t is sample thickness, mm. If inspection is performed at high test frequencies and/or with long cables, it is desirable to operate below probe-cable resonance frequency. This is normally achieved by using a probe of sufficiently low inductance. To optimize test results, the inspector has control over probe size and test frequency. In choosing probe diameter the following must be considered: (a) (b) (c) (d) (e) (f) (g) (h)

operating point on impedance diagram probe inductance and resistance sensing area sensitivity to defect length sensitivity to defect depth sensitivity to litt-off sensitivity changes across coil diameter (zero at centre) sensitivity changes with ferrite core or cup.

Choice of test frequency depends on: (a) (b) (c) (d) (e)

depth of penetration phase lag operating point on impedance diagram inductive reactance probe-cable resonance.

-88-

5.11

WORKED EXAMPLES

5.11.1 Effective Probe Diameter PROBLEM:

Determine sensing diameter of a 5 mm probe when (a) testing 316 stainless steel (p - 72 microhmcentimetres) at 2 MHz,

and (b)

testing brass ( p -6.2 microhm-cm) at 10 kHz.

SOLUTION: (a) 6-5

(2.13a)

72 50

+h

x

- 0.30 mm

106

D -. « D + 46 - 5.Q + 1.2 - 6.2 mm eft c

1.25 mm

(b)

D ,- - D + 46 - 5.0 + 5.0 - 10 mm err c

5.11.2 Characteristic Parameter PROBLEM:

If an available probe had coil dimensions of 10 mm outer diameter and 4 mm inner diameter, determine the best frequency for resistivity measurements of a zirconium alloy (P " 50 microhm-cm).

SOLUTION:

The best frequency for resistivity measurements is when the operating point is at the knee location on the impedance diagram. This occurs when the characteristic parameter P c «10. Using equation 5.5,

-4 /lO.O + 4.0 \

P c - 7.9 x 10 therefore,

f

f/50 - 10

50 kHz.

(This calculation places no emphasis on skin depth effect, which may be an overriding consideration).

-89-

CHAPTER 6 6.1

-

SURFACE PROBE SIGNAL ANALYSIS

INTRODUCTION Manufacturing and preventive maintenance inspection of "flat" components with surface probes is one of the oldest and most important applications of eddy current testing. Manufacturing inspection of small steel components for defects and hardness is almost exclusively performed by eddy current methods. For safety reasons and preventive maintenance (savings on replacement costs and downtime) inspection of aircraft components for cracks and heat treatment effects has been performed since commercial aircraft first went into service. Eddy current testing is one of the most effective NDT methods for the above applications because it doesn't need couplants, it is fast, and 100% volumetric inspection is often possible. . This chapter describes how to maximize signal-to-noise by proper choice of teat frequency and minimizing "lift-off" noise. Emphasis is given to signal analysis and how to recognize and discriminate between defect signals and false indications. An attempt is made throughout this chapter to illustrate discussion with real or simulated eddy current signals.

6.2

EDDY CURRENT SIGNAL CHARACTERISTICS

6.2.1 Defect Signal Amplitude A defect, which disrupts eddy current flow, changes test coil impedance as f.he coil is scanned past a defect. This condition is shown pictorially in Figure 6.1 which portrays eddy currents induced by a surface probe in a defective plate. Eddy currents flow in closed loops as illustrated in Figure 6.1(a). When a defect interferes with the normal path, current is forced to flow around or under it or is interrupted completely. The increased distance of the distorted path increases the resistance to current just as a long length of wire has more resistance than a short length. Eddy currents always take the path of least resistance; if a defect is very deep but short, current will flow around the ends; conversely, if a defect is very long (compared to the coil diameter) but shallow, the current will flow underneath. In summary, defect length and depth (and width to some degree) increase resistance to eddy current flow and this, in turn, changes coil impedance. (The effect of defect size on flow resistance in tube testing is derived in Section 8.2.1) .

-90-

COIL BOUNDARIES EBDY CURRENTS

SURFACE COIL WINDINGS , TEST PLATE

TEST PLATE EDDY CURRENT DISTORTION AT CRACK

CRACK (b) EDDY CURRENTS TAKE THE PATH OF LEAST RESISTANCE UNDER OR AROUND A DEFECT

(9) EDDV CURRENTS FLOW IN CLOSED PATHS. A DEFECT INTERFERES WITH THE NORMAL PATH.

Fig. 6.1;

Eddy Currents In a Defective Plate

In terms of the equivalent coll circuit of a resistor In parallel with an Inductor and Its associated semi-circular Impedance diagram (Section 3.5), a defect moves the operating point up the impedance diagram. Increasing resistance in a test article changes both probe inductance and resistance. In the preceding discussion the defect was considered to disrupt the surface currents closest to the coil. Consider the difference between surface and subsurface defects. When a surface probe is placed over a deep crack of infinite length, the surface currents must pass underneath the defect if they are.to form a closed loop, see Figure 6.2(a). This is not the case with subsurface defects as shown in Figure 6.2 (b). Although the void in this picture is not as far from the surface as the bottom of the crack, the void may not be detected. Eddy currents concentrate near the surface of a conductor,and therefore, tests are more sensitive to surface defects than internal defects. The skin depth equation helps in the understanding of this phenomenon. In Chapter 2 it was shown that current density decreased with distance from the surface in the following proportions: - 63% of the current flows in a layer equivalent in thickness to one skin depth, 5 , - 87Z flows in a layer equivalent to two skin depths, 2 5 , - 95% flows in a layer equivalent to three skin depths, 36 .

-91-

SURFACE COIL

IX'

TEST PLATE

CRACK

(a) EDDY CURRENT FLOW UNDER A CRACK

(b) EDDY CURRENT FLOW AROUNO A SUBSURFACE VOID

Fig. 6.2: Eddy Current Flow in the Presence of (a) Surface and (b) Subsurface Defect Since only 5% of the current flows at depths greater than the 3 6 , there is no practical way to detect a subsurface defect at this distance from the surface. But in the case of a long surface defect 3 8 or greater in equivalent depth, most of the current is flowing under the defect. Surface cracks will be detected and depth can be estimated even if eddy current penetration is a small fraction of the defect depth. Once eddy currents are generated in a metal surface, they will follow the contour of a crack because a potential is set-up about the crack. 6.2.2 Defect Signal Phase From the above description one cannot predict a defect signal in detail, only its relative amplitude and direction on the impedance diagram. A more complete explanation requires inclusion of phase lag. Consider the cross section of a surface probe as shown in Figure 6.3(a). This pictorial view shows the distribution of magnetic field magnitude and phase around a coil as derived by Dodd(2). The solid lines are contours of constant magnetic field strength; the dashed lines represent constant phase. Since the magnetic field and induced eddy currents have approximately the same phase, the dashed lines will also represent the phase (g) of the eddy currents. Amplitude drops off exponentially with distance and eddy current flow increasingly lags in phase ([relative to eddy currents adjacent to the coil) both with depth and with axial distance from the coil. Skin depth effect occurs in both radial and axial directions. Figure 6.3(a) permits an approximate derivation of eddy current signals for the shallow surface, subsurface and deep surface defects illustrated. One needs to establish a

-92-

COHSUNT »«PUTUOE

OEEP DEFECTSHULLOI DEFECT SUBSURFUCE DEFECT-

DEFECT POSITION

Fig. 6.3;

(a)

Derivation of Eddy Current Signal Appearance for Three Types of Defects"

-93-

reference phase direction as starting point; the LIFT-OFF direction Is convenient and can be defined as the signal resulting from Increasing the space between the coil and test article, starting from the point when the space is minimum. The signal or effect of defects can be imagined as the absence of eddy currents which were flowing in the area before the defect existed at this location. As the defects approach the coil from positions 0 to 5 in Figure 6.3(a), the signal on the eddy current storage monitor moves from point 0 to 5, tracing the curves illustrated in Figure 6.3(b). This procedure is reasonably straight forward for shallow surface and subsurface defects since they are localized and only intersect one phase and amplitude contour at any given position. For the deep defect one has to divide the defect into sections and determine weighted average values for amplitude and phase at each position. The shallow surface defect in Figure 6.3(b) has a large component in the lift-off direction; primarily its approach signal makes it distinguishable from lift-off. As defect depth increases, signals rotate clockwise due to increasing phase angle. The angle indicated in Figure 6.3(b) is not the value calculated from the phase lag equation, 3 = x/6

(2.14)

where (3 is phase lag (radians), x is distance of defect below the surface (mm) and 6 is skin depth (mm). The angle between lift-off and defect signals is about 2 g . Although probably not strictly true, one can imagine defect phase angle as the sum of a lag from the coil to the defect and the same lag back to the coil. The foregoing discussion assumes that the defect is a total barrier to the flow of current. Although this assumption is valid for most cracks or discontinuities, some cracks are partial conductors. Fatigue cracks, formed when the test article is under a tensile stress, can become tightly closed when stress is released. The result is that some fraction of eddy currents could be conducted across the crack interface and the magnitude of the coil impedance change due to the defect will be less. The phase lag argument is still valid; a deep crack will still be distinguishable from a shallow crack by the shape of the eddy current signal, but the sensitivity to such a crack will be reduced because of smaller amplitude, 6.3

EFFECT OF MATERIAL VARIATIONS AND DEFECTS IN A FINITE THICKNESS For each test, one must decide on the test frequency to use and on the phase setting. The conventional way of setting

-94phase on an eddy current Instrument is to display the "lift-off" signal horizontally (on the X-axis) with the impedance point moving from right-to-left as the probe is raised. All material variables will then display an eddy current signal at an angle clockwise to the lift-off signal.

7 iran

LIFT-OFF

1.5 mm 2.0 mm

- -At LIFT-OFF FREQUENCY = 10 kHz

LIFT-OFF FREQUENCY = 50 kHz

LIFT-OFF FREQUENCY - 200 kHz

Fig. 6.4: Probe Response to Various Test Parameters at Three Frequencies Discrimination between defects and other variables is accomplished through pattern recognition and varying test frequency. Figure 6.4 displays the change in coil impedance loci for various parameters at different test frequencies. The electrical resistivity (Ap) signal angle, relative to lift-off, increases only slightly as frequency is increased, whereas a change in plate thickness ( At) signal angle continually increases with frequency. The angle, between the signal from lift-off and plate thickness change, equals about twice the phase lag across the plate thickness. The signal from a change in magnetic permeability (Ay) of the plate is approximately 90° to the lift-off signal at low frequency and decreases only slightly with Increasing frequency. Figure 6.5(a) illustrates a computer simulation of coil response to various test parameters. The simulation is based on the same probe and test sample used in the previous figure. Comparison of these two figures reveals computer simulation gives very realistic results.

-95-

1.0 LlfT-OFF ,2nn

10 kHz

\

1.5 mi.

x

\

5 0 KHZ \

0.25

-\>

0.9

0.J5

A/. = t25* A t - -25» A M = .25*

\\

p. - 1.0

m\

0.2

o.i

(a) v

-..

•-

V

IIFT-OFF I 2 Ml

I1«.f'

1 . 5 mm

\ 0.25 mm *>3(

to kHz

O.I

0.7

*J— 2 r»

\ = 72 f > a • cm

0.6

\

\

• = •25% =1.0

0.25 m 0.5

\ I SC kHz

0.4

0.25 MH '*»/

J 0.1

1 0.2

J 0.3

L. 0.4

0.5

(b) Flg» 6.5:

Computer Simulation of Probe Response to Various Test Parameters

M

-96-

Note at 50 kHz the increase in magnetic permeability signal (Ay) is to the right of the electrical resistivity signal for the 7 mm probe. For the 25 mm probe at 50 kHz it is to the left of the Ap signal. As the operating point moves down the impedance curve with increasing probe diameter, a resistivity signal rotates CW relative to a permeability signal. Note also that the permeability signal is not perfectly parallel to the inductive reactance axis. This is due to the skin depth and phase lag changing with permeability, rotating the signal CW. During general inspection for all parameters in a thin plate test frequency is normally chosen such that 'lift-off and 'change in plate thickness1 signals are separated by 90° on the impedance plane. This frequency is empirically derived by setting ratio between plate thickness and skin depth equal to approximately 0.8, t/5

=

0.8

Substituting in equation 2.13 f - 1.6 p/t 2

L C

is angular frequency, radians/second Is coil inductance,henries is total cable capacitance, farads

Transposing this equation and substituting shows resonance occurs when frequency is

to - 2irf

]

'

This approach is sufficiently accurate for most practical applications. A more rigorous approach to resonance is presented in Section 4.3. Resonance is apparent when a probe and cable combination, which balances at a low frequency, will not balance as frequency is increased. At the approach of resonance, the balance lines on the eddy current storage monitor will not converge to a null. The two balancing (X and R) controls will produce nearly parallel lines, rather than the normal perpendicular traces, on the storage monitor. A number of steps can be taken to avoid resonance: 1. 2. 3.

4.

Operate at a test frequency below resonance, such that f te st is less than 0.8 f r . Select a probe with_lower inductance. (Since f r is proportional to 1//L , inductance must be decreased a factor of four to double the resonant frequency). Reduce cable length or use a cable with lower capacitance per unit length (such as multi-coax cables). This will raise the resonance frequency since capacitance is proportional to cable length and f is proportional to r 1/^C , Operate at a test frequency above resonance, such that f test l s greater than 1.2 f r . However, above resonance the sensitivity of all eddy current instruments decreases rapidly with increasing frequency because capacj.tive reactance (Xc. • 1 / W C ) decreases, and current short circuits across the cable rather than passing through the coil.

I '

H i i

A "1.0. A few slightly magnetic materials can be heated above their Curie temperature to make them nonmagnetic. Monel. 400 heated to between 50° and 70°C has been tested in this manner. Most materials have too high a Curie temperature to be tested by this approach. The only other way to decrease U^ to unity is by magnetic saturation. This topic is treated in a subsequent section. 9.4

TESTING MAGNETIC

MATERIALS

9.4.1 Simplified Impedance Diagrams A qualitative understanding of the effect of permeability on coil impedance can also be obtained by the equivalent circuit and its associated semicircular impedance diagram treatment of Section 3.5. Coil inductance is a function of magnetic flux through it; flux increases in the presence of a magnetic material. For a cylinder surrounded by an encircling coil, coil inductance is proportional to both the cylinder's permeability and its cross-sectional area, L p

2 " yr o

where L_ is primary coil (probe) inductance, U r * V ^ is the cylinder's incremental permeability and D Q its diameter. An increase in permeability or diameter will increase coil inductance. By a similar treatment to that presented in Chapter 3, one can generate the impedance diagrams of Figure 9.11. Figure 9.11(a) is obtained by plotting the encircling coil Impedance normalized to the inductive reactance in sir. It illustrates the effect of permeability and cylinder diameter. As permeability or cylinder diameter increases (with constant coil diameter) coil impedance increases drastically. (This explains the good response to ferromagnetic inclusions and deposits discussed in Sections 6.5.1 and 8.3.1). There is no phase separation and hence no discrimination between variations in permeability and cylinder diameter. However, there is about 90° phase separation and hence excellent discrimination between variations in permeability and resistivity.

-175-

uiL

1.0

1.0

0.5 RL/