·Home ·Table of Contents ·Civil Engineering

Generation and Detection of Elastic Guided Waves with Magnetoelastic Device for the Nondestructive Evaluation of Steel Cables and Bars Laurent LAGUERRE Laboratoire Central des Ponts et Chaussées, BP 4129 - 44341 Bouguenais Cedex, France Jean-Christian AIME, Michel BRISSAUD Laboratoire de Génie Electrique et Ferroélectricité, Institut National des Sciences Appliquées de Lyon, 20 Avenue Albert Einstein, 69621 Villeurbanne Cedex - France Contact

Abstract

The aim of this study is to present the potentialities of a system based on the magnetoelastic effect to evaluate ferromagnetic structures through the generation and detection of mechanical waves in these structures. First, experimental conditions for a well-adapted use of the magnetoelastic emitter are presented. Then, the noncontact magnetoelastic device is shown to be a good candidate for the experimental observation of the characteristics of mechanical wave propagation in ferromagnetic structures. Finally, detection of simple defects is discussed according to the operating conditions achieved before.
Keywords: magnetoelastic effect, defect detection, nondestructive testing, mechanical guided waves.

Introduction

There is an increasing need for health monitoring of civil engineering structures. Steel members constitute an important part of these structures and the evaluation of their integrity is essential. Various methods of nondestructive evaluation (NDE) already exist and the most commonly used are radiography, acoustic monitoring, eddy current and magnetic flux leakage methods. Each method has its own operating condition, validity range and potentialities. Nevertheless these potentialities can be strongly affected by the structure configuration (low accessibility for the application of a particular NDE method, for example). Hence, synergism of NDE techniques could be a powerful tool to overcome this restriction.
In this way, the method presented here may appear as a complementary NDE method. It allows the generation and detection of mechanical wave in ferromagnetic structures using the direct magnetoelastic (or magnetostriction phenomenon) and inverse magnetoelastic effect. The method is based on the ability of a ferromagnetic material to deform mechanically under the application of a magnetic field (direct effect) or to change its magnetization under a mechanical stress.
Both passive and active magnetoelastic systems have already been designed as acoustic or ultrasonic detector [1-2] and as ultrasonic pulse-echo. In the last case, the application is rather dedicated to the detection of flaws in tubes and pipes [3-4].
The major advantage of the method is the absence of coupling between the sensor and the device under testing. So, sensors can be easily moved along the structure to allow, in conjunction with the pulse-echo technique, the inspection of large structures.
In the present experiment-based study, conditions for a good production of mechanical waves using the magnetoelastic effect are first shown. Then, propagation characteristics are briefly examined for a cylindrical bar and a seven-wire strand. Finally, detection of defects for these two configurations is presented.

Experimental setup

Two types of experiments were conducted:
• the first experiment was designed to point out the non-linear behaviour of the magnetoelastic conversion versus the polarizing field. This allows one to specify the polarizing field conditions for an optimized magnetoelastic generation relative to the material under investigation. In our case, the emitter is a magnetoelastic device, the detector is a piezoelectric sensor mounted on the bar end.
• the second experiment was designed to examine the conditions of mechanical waves propagation in ferromagnetic structures (bar, cable) using the magnetoelastic devices and then to test the ability of the system to detect flaws.
In both cases, ferromagnetic structures are axisymmetric structures.

The magnetoelastic devices consist of two encircling coils. One for the emission (driving coil) and one for the detection (sensing coil). These two small coils are placed inside magnetizing (or polarizing) coils, respectively. The axis of symmetry of the coils are parallel to the length of the specimen. A such configuration makes the device predominantly sensitive to the generation and detection of longitudinal mechanical waves [5].
The magnetizing coil is fed with a static current source. The current intensity is controlled to provide magnetic fields up to 60 kAm^-1. The driving coil is powered by a current amplifier driven by an arbitrary function voltage generator. The nominal performances of the current amplifier are 1kVA and 50Hz -100 kHz bandwidth. The detection coil is connected to a digital oscilloscope via a wideband voltage amplifier and a digital filter. The whole system is piloted by a personal computer using a Labview National Instrument software.

Nonlinear behavior of the magnetoelastic emitter as a function of the polarizing field strength.

Low polarizing fields
The ferromagnetic material is a steel bar of 6 m long and 15.5 mm in diameter. The excitation signal is a 3-cycle sinusoidal burst of 9 kHz center-frequency. The intensity level in the driving coil is measured using a wide-band current monitor and fixed at 5 A peak. The magnetoelastic device - piezoelectric sensor distance is about 1.5 m.
The experiment is conducted as the following sequence: the bar is first demagnetized, the polarizing field strength is selected, and the pulse is launched. This sequence is repeated for increasing polarizing fields (from no polarizing field up to 25 kAm^-1). The H values are measured, using a gaussmeter, at the center of the polarizing coil with no bar in the coil.

According to the propagation theory of longitudinal waves in cylindrical axisymmetric structures, we can consider the bar as a nondispersive and monomodal medium at this frequency. Therefore, the characteristics of the mechanical pulse emitted are not affected by the propagation distance down the bar between the emitter and detector (# 1.5m).

The results are summarized in figures 1 (a) and (b). Figure 1(a) represents the time evolution of the driving coil intensity and the detected signal for different low levels of polarizing field, respectively. It illustrates the influence of the polarizing field strength on the mechnical wave amplitude generated using magnetoelastic effect. The following H values are 0, 1, 2, 2.6, 5.2, 21 kAm^-1 for H0, H1, H2, H3, H4 and H5, respectively.

(a)	(b)
Fig 1: (a) influence of the polarizing field on the mechanical wave generated with the magnetoelastic device, (b) representation of the same signals in the Fourier domain (normalized amplitude spectrum).

Qualitatively, we observe 3 major phenomena in the variation of the detected signal with the polarizing field:

• a modification of waveform. For the H0 condition, the double frequency effect relative to the quadratic behaviour of the magnetostrictive effect is illustrated for nonpolarized material excited at a single frequency. For the H5 condition, the time detected signal approaches the driving coil intensity behavior. The magnetostrictive effect is quasi-linearized
• a variation of the amplitude,
• a modification of signal polarity. We can observe a polarity change between 2 and 2.5 kAm^-1.
This is a typical illustration of the nonlinear behaviour of the magnetoelastic effect versus polarizing field.
The nonlinear behaviour of the magnetoelastic device can be summarized in the figure 1(b). This figure represents the normalized amplitude spectrum of the Fourier transform of the time excitation and detected signal, respectively, for the static magnetic conditions mentionned above. We can observe, with no polarizing field, the dominant frequency is the double frequency 2f of the excitation frequency f. As the polarizing field strength increases, we note the appearance and extinction of the second harmonic 2f and the subharmonic f/2 in the mechanical wave generated, to finally reach a quasi-linear behavior at H5.
This experiment was conducted with a driving coil intensity equal to 5A peak. Some results not presented here, would have shown, a very similar behavior for H5 at higher driving intensity (up to 20A peak).

Medium polarizing fields
For this experiment, we use a more powerful system to produce higher amperage in the polarizing coil. The same sample was used. Figure 2 is the variation of the mechanical energy of the detected wave versus the polarizing field (from no polarizing field to 60 kAm^-1). We remark the magnetoelastic coupling decreases with increasing polarizing field. The maximun range is found to be between around 30 kAm^-1. Magnetoelastic generators produce mechanical waves for a limited range of polarizing field. These main properties have already pointed out by [6]

Fig 2: evolution of the detected normalized mechanical intensity versus the polarizing field for the steel cylindrical bar

This simple procedure allows one to select the experimentally matched configuration of the magnetoelastic transducer to enhance the production of mechanical waves in a given material sample.

Propagation characteristics in a cylindrical steel bar and a seven-wire strand

In order to know the potentialities of the magnetoelastic device in the detection of defects, it is important to evaluate first, its possibilities as a system of generation and detection of mechanical waves. To this end, we first use a cylindrical steel bar as propagation medium. Mechanical wave propagation characteristics are well-known in a such axisymmetric structure [7]. This constitutes a validation of our system.

According to the magnetoelastic detector, [8] showed it was possible to detect mechanical waves, using at least a 60 db voltage amplifier and residual magnetization. We use this configuration in our study.
The magnetoelastic detector (or inductive sensor) is not only sensitive to the magnetoelastic effect but to electromagnetic noise as well. Assuming a random noise, the detected magnetoelastic signal is the result of a systematic N-sample time-averaging (with ), to improve the signal-to-noise ratio. Moreover, each sample is passband filtered between 700Hz and 1MHz.
The excitation pulse is a gaussian windowed sinusoidal burst. The pulse center frequency f_c can be varied. The pulse normalized frequency content Df / f_c is unity at -20dB.

Cylindrical steel bar
The sample used is the same as in the above study. The emitter and detector are positioned at 3 m and 1.5 m from the bar end, respectively Figure 3 is an example of the waveforms of the received time signal using the magnetoelastic device for a free cylindrical bar for the three following fc: 8.4, 67, 150 kHz, respectively). The amplitude waveforms are normalized with respect to the peak amplitude of the first arrival (as 1). Recording duration is 5 ms. So, assuming a propagation velocity of 5100 ms^-1, the propagation distance is about 25 m.

Fig 3: normalized amplitude waveforms of the magnetoelastic detected signal for 3 operating center frequency of excitation fc = 8.5, 67 and 150 kHz, respectively, for a cylindrical steel bar 15.5 mm in diameter and 6 m long..

For f_c = 8.4 kHz, we can observe the time waveform ot the received signal composed of the first wave packet (denoted 1, figure 1) relative to the first detected mechanical pulse and the successive reflexions from each end. The emitter produces simultaneously a mechanical wave in opposite directions. So, according to the sensor positions, two successive echoes have the same polarity (denoted 2 and 3), which is the opposite polarity to the two following ones (denoted 3 and 4).
The signal suffers from no attenuation and no dispersion occurs.
For f_c = 67 kHz, we can observe the time signal composed of a first pulse different from the following ones (denoted 0). This first pulse is the electromagnetic coupling between the emitter and detector. It is simultaneous to the excitation. Its amplitude increases with increasing frequency. The second wave packet is relative to the first detected mechanical wave (denoted 1).
Here, the dispersive nature of longitudinal waves is evident. We can observe, the spreading of the pulse with time. The front of each received wave packet is composed of lower frequencies and the tail of higher frequencies. This is in agreement with the theory [5], which predicts a velocity of propagation frequency-dependent and decreasing as frequency increases for the longitudinal fundamental mode L01.

For f_c = 150 kHz, the signal is dispersive and multimodal. A time-frequency representation can help one to a more insight of the bar behavior at this frequency content. A classical short time Fourier transform using a Hanning window is applied to the time signal (Figure 4). Each wave packet arrival can clearly be seen. The first one is the electromagnetic signal. The second wave packet is the first arrival of the mechanical wave. Two distinct time -frequency behaviors can be isolated:

Fig 4: time-frequency diagram of the received time waveform, for 150 kHz excitation pulse, for the steel cylindrical bar of 6 m long, 15 mm in diameter

• under 200 kHz, the lower frequencies propagates faster than the higher frequencies,
• above 200 kHz, the lower frequencies propagates slower than the higher frequencies.

A comparison between the experimental frequency-dependent group velocity and the theoritecal one allows one to retrieved the following two modes of propagation, i.e. the fundamental longitudinal mode L(0,1) and the second longitudinal mode L(0,2). A third mode is present around 180 kHz and cannot be attributed to a longitudinal mode according to its frequency content. It should be a flexural mode (as the F(2,1) mode ) coupled to the first fundamental longitudinal mode as discussed in [7].

The time-frequency representation allows one to quantitatively estimate the dispersive and multimodal nature of the structure under testing.

Seven-wire strand
A serie of similar experiments was conducted on a prestressing seven-wire strand of 6 m long. The strand that is 15.7 mm in diameter consists of a 5.3 mm center wire and 6 surrounding helical wires. Time waveforms of magnetoelastic detected signals are displayed in figure 5 for 3 different f_c (8.4, 50 and 125 kHz).

Fig 5: normalized amplitude waveforms of the magnetoelastic detected signal for 3 operating center frequency of excitation fc = 8.5, 50 and 125 kHz, respectively, for a seven-wire strand of 6 m long and 15.7 mm in diameter..

In the low-frequency range, same remarks as for the cylindrical bar can be made. No dispersion and attenuation occur. In the medium-frequency range (i.e. f_c = 50 kHz), the strand signal exhibits a higher amplitude attenuation with the propagation distance than the bar signal at a close frequency. Finally, time signals of the bar and strand are quite different at the higher frequency. The attenuation is particularly high in the strand sample.
Moreover, according to the strand, the amplitude attenuation is stronger at 125 kHz than 50 kHz. The major difference between these two strand signals lies in the time behaviour of the wave packets following the first arrival (as 1). The second arrival (as 2) of the 150 kHz-waveform undergoes a stronger distorsion than the one at 50 kHz.

A complementary study was conducted for two receiving coil positions (1.5 and 2.5 m, respectively) and a fixed emitter position, for f_c = 110 kHz. The time waveforms of the received signals and the associated time-frequency representation are illustrated in fig 6 and fig 7.

Fig 6: normalized time waveforms of the rceived, fc = 110 kHz, for two receiving emission-detection distance: (a) 1.5 m and (b) 2.5 m
Fig 7a:	Fig 7b:
Fig 7: time-frequency representation of the above waveforms (a) and (b)

From figure 7, we observe the frequency components around 100 kHz are rapidly attenuated. These frequencies are missing (figure 6, image a), i.e. beyond a propagation distance of 1.5 m. As the emission-detection distance increases, we can observe an attenuation of higher frequencies from this 100 kHz missing-frequency (figure 6, image b).
The 110-150 kHz band has really disappeared beyond a propagation distance of 2.5 m. The progressive extinction of the higher frequencies is such that no frequency content still exists after the first end reflexion (i.e 5.5 m) . Concurrently, attenuation of lower frequencies is weaker, following an opposite scenario (highest relative frequencies are first attenuated). Beyond a propagation distance of 25 m, the wave packet is composed of frequencies lower than 50 kHz. We remark the lower frequency content is not very dispersive.
The strand consists of a bundle of wires acting as a complicated frequency filter for the propagation of mechanical waves. The attenuation of higher frequencies is strong, and only lower frequencies still compose the pulse beyond a propagation distance of a few meters.
Further studies will have to be conducted for different type of strands with different contact conditions between adjacent wires.

Default detection

In this part, we try to appreciate the sensitivity of the method to a defect close to an end of the samples. It gives one an estimation of the spatial resolution of the pulse-echo system.

Cylindrical bar:
The sample used was a cylindrical bar 2.92 m long and 14 mm in diameter. One simultated defect was made at 150mm from the end of the sample. The defect is a small hole of 5 mm in diameter in a transverse direction. The emitter and detector are at a 1.13 and 2 m from the beginning of the sample, respectively. The magnetic conditions and pulse excitation are the same as those used for the 15 mm bar above. The center frequency is around 84 kHz. The detected time waveform of 1 ms duration is in figure 8. First reflection from the defect (as D1) can clearly be observed. We can recognize the above mentionned (1), (2), and (3) wavepackets. Defect and end echoes can be easily time resolved. Moreover, according to the emission coil, detection coil and defect position along the sample, multiple reflexions between the defect and the near end of the bar can also be observed (as D2, D3 and D4 respectively).

Fig 8: time waveform of the received signal for a 2.92 m long cylindrical bar and 14 mm in diameter bar with a defect at 150 mm from the bar end.

Seven-wire strand
One defect was simulated: 150 mm from one of thee six surrounding wires was cut off from the end of the strand. The same strand as above was used. The emitter and detector are positioned at about 1.80 and 3.9 m from the beginning of the strand. The center frequency is around 50 kHz.The time waveform is in figure 9. We can recognize the echo ends as (1), (2), (3) and (4) respectively. According to the sensors positions, the defect echo must be time-positioned before the second wave packet which is the reflexion of the first wave packet at the end of the strand. The defect echo can be identified (as D) just before the front of the second wave packet and can be seen before the fourth wave packet as well. We must note the amplitude of the second wave packet is weaker than the third wave packet amplitude. The second wave packet (2) is the transmitted amplitude of the initial pulse. beyond the defect. In contrast, the third wave packet (3) amplitude is the reflexion from the beginning of the strand of initial pulse. This pulse does not encounter the defect along its propagation path, so it cannot be attenuated as (2) was.

Fig 9: time waveform of the received signal for a 6 m long seven-wire strand and 15.7 mm in diameter with 150mm of one surrounding wire cut off from the strand end

It is worth noting a defect far from the end would be clearly seen. Figure 11 represents one defect at 680 mm from the near end of the strand. One surrounding wire is cut at 680 mm from the near end and the remaining two parts are distant from 2 mm.

Fig 10: time waveform of the received signal for a 6 m long seven-wire strand and 15.7 mm in diameter with one surrounding wire cut at 680 mm from the strand end

Conclusion

This preliminary study has experimentally shown the potentialities of a magnetoelastic device to assess the evaluation of the integrity of some steel members commonly used in civil engineering. Operating as a low-frequency ultrasonic pulse-echo system, the magnetoelastic device can be used
1. to retrieve the characteristics of mechanical wave propagation of various samples, and
2. knowing these characteristics, to optimize the generated wave in terms of frequency content for a better detection of defects.
Further studies have to be done on different type of strands (geometry and daimeter) and signal processing technique as well.

References

1. L. C Lynnworth and J. E. Bradshaw, "Magnetostriction transducer for acoustic emission, impulse vibration, and noise analysis, Materials Research and Standards, Vol. 11, No3, March 1971, pp 33-35.
2. R.Hecker and R. K. Schulz, "Messung elasticher Wellen in Stahl - und gesteinssäben mit Hilfe des magnetoelastischen Effektes", Mat.-wiss. u. Werkstofftech, 28, 1997, pp 571-574
3. W. Böttger, H. Schneider and W. Weingarten, "Prototype emat system for tube inspection with ultrasonic guided waves", Nuclear Engineering and Design, 102, 1987, pp 369-376.
4. H. Kwun and A. E. Holt, "Feasability of under-lagging corrosion detection in steel pipe using the magnetostrictive sensor technique", NDT&E International, Vol. 28, No4, 1995, pp 211-214.
5. V. A. Komarov and P.S. Kononov, "Direct and inverse electromagnetic-acoustic conversion in ferromagnetic rods", Soviet Journal of Nondestructive Testing, Vol. 14, n°5, May 1978, pp398-403.
6. R. B. Thompson, "Electromagnetic noncontact transducers", Ultrasonics symposium, Phoenix, 1977, pp385-392.
7. T. R. Meeker and A. H. Meitzler, "Guided wave propagation in elongated cylinders and plates", in Physical Acoustics, Vol. 1, 1964, W. P. Mason, Ed., Academic, New York
8. C. M. Teller and H. Kwun, "Monitoring structural degradation of concrete beams using magnetostrictive sensors", Proceedings Int. Conf. Nondetructive Testing of Concrete in the Infrastructure, Dearborn, Michigan, June 1993, pp 559-566.

© AIPnD , created by NDT.net

|Home|    |Top|