Bundesanstalt für Materialforschung und -prüfung

International Symposium (NDT-CE 2003)

Non-Destructive Testing in Civil Engineering 2003
Start > Contributions >Posters > Acoustical: Print

NON-DESTRUCTIVE EVALUATION OF SLENDER STEEL MEMBERS USING ELASTIC GUIDED WAVES

Laurent Laguerre, Laboratoire Central des Ponts et Chaussées, Division Reconnaissance et Mécanique des Sols,
Section Reconnaissance et Géophysique, Route de Bouaye, BP 4129, 44341 Bouguenais, France, Laurent.Laguerre@lcpc.fr

ABSTRACT

A magnetostrictive- transducer-based device used for the low-frequency ultrasonic longitudinal wave reflectometry (<500kHz) is designed for the non-destructive evaluation of steel bars and cables. The major factors influencing the system performances for the production of guided wave were studied from laboratory-based experiments. Detection and localization of artificial defects in a seven-wire strand using pulse compression signal was performed. The experiments conducted for a prestressing strand under longitudinal tensile loading showed the modifications of the propagation conditions with loading. A time frequency analysis of the dataset showed a direct relation between a missing frequency band in the propagated signal and the mechanical tension.

INTRODUCTION

Magnetostrictive-based pulse-echo device generates and detects mechanical waves using the direct and inverse magnetostrictive effect [1,2] which states that ferromagnetic material has the property to mechanically strain under a magnetic change and to change its magnetization under a mechanical stress.

In our case, this non-contact device is tested on steel members of civil engineering structures from laboratory-based experiments for the flaws detection. The steel material acts both as a part of the transducer and medium of propagation. A local superimposition of a static polarizing magnetic field to a transient magnetic field allows to enhance the excitation or detection of mechanical waves into the specimen under testing. Both magnetic fields are produced by a static polarizing coil and a small dynamic coil at the emission and at the reception. The two coils are coaxial and encircle the specimen under evaluation, the magnetic core. Our experimental configuration, imposes axial and axisymmetric static magnetic fields to produce a local longitudinal mechanical deformation.

The proposed method is an alternative to conventional ones as for example eddy current and magnetic flux leakage methods.

MAGNETOSTRICTIVE DEVICE CALIBRATION

Experimental set-up
A DC power supply is connected to each polarization coil to vary the polarizing field at the emission and detection independently. The transmitter is driven by a controlled-current amplifier piloted with an arbitrary waveform voltage generator. The detector is connected to a voltage amplifier. The time signal is visualized on a digital oscilloscope. The time voltage excitation is a low radio-frequency burst, typically lower than 500kHz in our experiments. In the following, it consists of a sinusoidal carrier (centre frequency fc) enclosed in a gaussian window such that the impulse bandwidth is Df = fc at -20 dB. To improve the signal-to-noise-ratio, the time detected signal was time-averaged over N samples (500 < N < 1000) resulting from N emission/detection cycles. Then, data are collected and processed using a PC computer.

The specimens used in this experiment are elongated cylindrical bars and strands of various homogeneous steel materials.

Transduction nonlinearities
In our experiments, we first considered the material influence on the mechanical generation because

  1. the material to be tested is an intrinsic part of the transducer and
  2. the control of the mechanical wave generation is critical in the use of guided waves.

Hence, we monitored the material magnetostrictive emission versus polarizing magnetic field strength, dynamic magnetic field amplitude and frequency content. The magnetostrictive emission indicator was determined from the characteristics of the first mechanical wave-packet arriving at the detector. The specimen used was a steel cylindrical bar.

Figure 1a shows the variation of magnetostrictive emission rms voltage, with increasing amplitude of the dynamic drive current intensity (from 4.5 to 41.2 A) at different polarizing current intensities from 0 to 9 A for the steel cylindrical bar specimen. The excitation is a low frequency transient centred around 8.5 kHz. Figure 1a shows that the magnetostrictive emission rms voltage gives rise to a non-linear behaviour versus the polarizing current which roughly decreases for higher polarizing currents. The magnetostrictive emission rms voltage increases with the amplitude of the dynamic intensity. The polarity inversion in the time signals for polarizing current below and above 1A (not shown here) and the relatively high magnetostrictive emission at lower polarizing intensities are in quantitative agreement with common steel magnetostriction curve versus polarizing field. This shows the dynamic magnetostriction is the main driving force of the generation of mechanical waves at polarizing levels used here [3].

Fig 1: (a) influence of the polarizing and dynamic current on the magnetostrictive emission (b) amplitude Fourier spectrum of the generated elastic wave versus the polarizing intensity for a low and high dynamic peak current values, at fc=8.5kHz for a steel cylindrical bar.

Figure1b shows the spectrum of the mechanical generated waveform for two low and high driving current values for the range of polarizing intensities inspected above. Globally, the presence of harmonics to the fundamental frequency fc in the detected signal for the low polarizing intensities are a drawback for the well-controlled generation of guided waves (REF). Indeed, they tend to generate higher modes of propagation which alter the information carried by the fundamental mode.

These harmonics tend to disappear for higher polarizing current intensities (> 2.5 A). It shows that a satisfactory level of linearization of the transfer function between electrical and mechanical quantities can be achieved between 6 and 8 A (i.e. for polarizing magnetic field values between 18 and 24 kAm-1) over the possible range of driving currents, even if the linearization degradation increases with the driving current.

In Figure 2, we observe the influence of the source harmonics on the time-history waveforms at fc= 85kHz for a cylindrical bar (L = 6m long, 15.5 in diameter) at
fc = 85kHz. At 9A, in a linear regime, the mechanical wave is dispersive because the fundamental mode is dispersive for this frequency content [4]. With no polarization, the presence of nonlinearities at the emission strongly distorts the signal which comprised now higher modes [4].

Fig 2: influence of magnetostrictive emission nonlinearities on the generated elastic wave at fc=85kHz for a steel cylindrical bar. The transmitter and detector are at L/4 and L/2 respectively, in order to well separate the different contributions (direct and endreflected waves).

On the basis of above study, defect detection was performed on a seven-wire prestressing strand.

DEFECT DETECTION

A seven wire-strand of 15.7 mm in diameter consists of a straight center wire of 5.3 mm diameter surrounded by 6 helical wires (figure 3).

Fig 3: seven-wire strand geometry.

The strand used is 6 m long. Two defects were made on two distinct helical wires. One surrounding wire was broken at 680mm from one strand end and 150 mm of one surrounding wire was cut-off from the same strand end.

To overcome the limitation of strand attenuation to a long range inspection, coded signals can be an alternative to conventional burst signals. Indeed, the use of a chirp signal, for example, is known to improve both signal-to-noise-ratio and time resolution [5]. Furthermore, this signal is well-suited to our magnetostrictive device which is first a peak-current limited device and second works as a relative wide-band transducer rather than a resonant one [6].

In the following, we used a linear-frequency modulated chirp. There are several advantages to use a such type of excitation. First, the long duration signal allows to give more energy to the system, second the frequency modulation increases the frequency bandwidth, third a coded information is transmitted to the system.

The excitation signal is of the form LFC(t) = W(t) Sin (2p f(t) t) with a linear-frequency with a linear frequency modulation dependence f(t) such as with 0 = £ t £ t

t is the chirp duration and fmin, fmax are the lower and upper bounds of the chirp bandwidth b. W(t) is a given gating function.

Figure 4a shows the time waveform of the chirp excitation signal.

In order, to take full benefit of this technique, it is necessary to consider the cross-correlation between the transmitted coded signal and the detected signal instead of the detected signal. In that case, it is common to define a compression ratio equal to the bandwidth duration product bt which represents the gain of the technique. We use different conditions in our experiments, we chose to work at constant excitation duration and to vary the signal bandwidth. The duration is 2 milliseconds. The bandwidth started from a very low frequency frequency fixed at a fmin = 200 Hz value to a varying maximum frequency fmax. The frequency fmax varied from 50 to 140 kHz. This leads to bt ratios from 100 to 280.

Fig 4: (a) chirp time signal, (b) cross-correlograms of the damaged seven-wire strand, for different product duration-bandwidth chirp excitations (the higher frequency is from 50 to 140 kHz, the chirp duration is 2 ms).

Figure 4b represents for the damaged strand the cross-correlation between the measured time driving signal and the detected time voltage versus the time lag for the above mentioned fmax.

It shows the improvement in the defects detection (D1 and D2) and spatial resolution especially for the defect closer to the end of the strand using chirp excitation with increasing higher frequencies.

Quantitative estimation of the defects locations can be made by retrieving the propagation velocity in the strand, from the cross-correlated resulting signal, by determining the travel time of the pulse over a known distance. Typically, we use the time-delay separating 2 echoes from the strand ends, each echo-time position being determined from the maximum of correlation. The estimated strand velocity is Vstrand = 5153.2ms-1. The same procedure is repeated for the determination of the defects from the nearest strand end knowing Vstrand = 5175.7 18.1 ms-1.Hence, the respective predicted locations of D1 and D2 defects from the nearest strand end are d1 = 655 ± 26.7 mm and d 2 = 14.6 ± 2.1 mm . These results are to compare to actual locations of 680 mm and 15 mm, respectively. This is an encouraging result particularly for the detection of the nearest strand-end defect D2, considering the large transmitter-to-detector travel distances after the interaction with D1 and D2 defects of 6.14 m and 7.2 m, respectively. The conventional tone-burst excitation could not separate the D2 defect echo from the strand end echo.

SEVEN-WIRE STRAND UNDER TENSILE LOADING

Then, we performed measurements on a seven wire strand of ten meters long under different longitudinal tensile loading varying from 2% (5.8kN) to 60% (141kN) of the breaking load. The strand was loaded gradually from approximately no load to 60%, by step of 10%. We used the magnetostrictive device in a linear regime of transduction. A serie of measurement consisted of recording the time history at a given load and excitation frequency content (fc was varied from 8 to 112 kHz). The magnetostrictive transmitter and detector are located between the anchorages. A such useful configuration is allowed by the contactless ability of the magnetostrictive device (contrary to a piezoelectric system), and simply avoids the interaction of the wave with the loading system. Figure 5 shows the time histories for three fc 24, 80, 112 kHz and 2, 10 and 60% of the breaking load. The five traces of each serie of measurement correspond to increasing receiver locations with the transmitter at a fixed position. Hence, the first arrival of each trace is the direct wave between the transmitter and detector which has never interacted with clamped ends. The very first arrival whose time arrival is independent of the transmitter-to-receiver distance is simultaneous to the energizing of the device and relative to the electromagnetic coupling between the two transducer coils.

Fig 5: time traces for a seven-wire strand under different longitudinal tensile loading conditions and for fc =24, 80 and 112 KHz).

For the lower frequency, the first arrival is similar to a cylindrical bar of same diameter especially for the higher loading conditions. We can observe the varying influence of the clamped ends with the load. Especially, at higher loads, the time separation between a near- clamped-end-detector signal and the associated reflected wave is possible.

At the intermediate frequency (80kHz), we can observe the signal weakness for the lower load which is immediately increased for a low load increase. The effect of the dispersion of the wave as it propagates is clearly observed for the different transmitter-to-detector distances.

At the higher frequency, there again a strong signal which suffers spatial filtering as it propagates

The time-frequency analysis (Figure 6) performed for the higher fc and different loading conditions recordings highlights the above mentioned observations. On a first hand, it shows that a frequency band really electrically excited almost disappears from the propagated signal,and on the other hand, the average frequency of this frequency band increases with the applied tensile loading. This phenomenon is not yet understood to our knowledge but can reasonably be attributed to the structure of strand where conditions of contact between adjacent wires are dependent upon the loading strengths. Concerning acoustoelastic effect relying the propagation velocity to the stress level of the material, as a potential cause to these observations, the literature [7,8] shows this effect is very weak, so from our operating experimental conditions we deduced we can not measured it (we did not observe any velocity variations under the different loads).

Fig 6: time-frequency representation of the magnetostrictive detected signal for the seven wire strand under different tensile loading at fc=112kHz.

CONCLUDING REMARKS

An experimental magnetostrictive device has been designed for the detection of defects in long steel cylindrical bars and strand using low-frequency ultrasonic longitudinal guided waves. Experimental results have demonstrated the necessity to control the magnetostrictive emission especially in terms of polarizing current intensity in order to linearize the magnetostrictive transduction and maximize the mechanical energy inserted in the objects to be tested. The linearization step is crucial in the case of the use guided wave because it allows the control of the frequency content of the mechanical pulse excitation which in turn conditions the propagation of the pulse into the object. The application of the magnetostrictive device to the detection of artificial defects was carried out for a seven-wire strand. An alternative approach to the conventional tone-burst excitation, i.e., the frequency modulated excitation, for the improvement of both signal to noise ratio and time resolution and because the magnetostrictive device is a non-resonant transducer. Hence, minimizing the dispersive character of propagated guided waves, near-end defects (positioned at 15 cm from the end) were detected and located over non negligible travel distances (typically around 10 meters) which can be considered as a satisfactory result for civil engineering applications.

Finally, measurements made on a seven-wire prestressing strand under tensile loading have shown the influence of the contact conditions between wires on the wave propagation at moderate frequencies. Through a time-frequency analysis of the dataset, a relation of causality can be found between the modification of the propagation conditions and the mechanical tension level.

REFERENCES

  1. R. M. Bozorth, "Ferromagnetism", Van Nostrand Company Inc., New York, (1951).
  2. H. Kwun and A. E. Holt, "Feasability of under-lagging corrosion detection in steel pipe using the magnetostrictive sensor technique", NDT&E International, 28, (4), (1995), 211-214.
  3. R. B. Thompson, "Electromagnetic noncontact transducers", Ultrasonics symposium, Phoenix, (1977), 385-392.
  4. L. Laguerre , J.-C. Aime and M. Brissaud, "Magnetostrictive pulse-echo device for non destructive evaluation of cylindrical steel materials using longitudinal guided waves", Ultrasonics, Vol. 39, no 7, pp 503-514, April 2002.
  5. M. Skolnik, "Radar Handbook", Mc Graw-Hill, New-York, (1970).
  6. M. Brissaud, J.-C Aime. and L. Laguerre, "Analytical modelling of magnetostrictive transducers used for non-destructive evaluation", 8th Int. Cong. on Sound and Vibration, Honk-Kong, (2001), July.
  7. G. Washer, "The acoustoelastic effect in prestressing tendons", PhD Thesis manuscript, 163p, May 2001.
  8. H. L. Chen and K. Wissawapaisal, "Application of Wigner-Ville transform to evaluate tensile forces in seven-wire prestressing strands", J. Eng. Mech., Vo. 128, no11, pp1206- 1214, November 2002.
STARTPublisher: DGfZPPrograming: NDT.net