Ultrasonics 37 (2000) 589594 www.elsevier.

nl/ locate/ultras

Design of a self-calibrating simulated acoustic emission source

M.J. Evans a, J.R. Webster b, P. Cawley a, *
a Department of Mechanical Engineering, Imperial College, Exhibition Road, London SW7 2BX, UK b Strategic Research Centre, Rolls Royce plc, PO Box 31, Derby DE2 8BJ, UK Received 1 March 1999; received in revised form 1 September 1999

Abstract The use of conical piezoelectric transducers as point acoustic sources has been investigated. It has been shown that transducers based on a design originally developed at the National Institute for Standards and Technology in the USA can be used as point transmitters over the frequency range of interest in acoustic emission measurements (100 kHz to around 1 MHz). They should, therefore, be suitable for use in experiments to calibrate structures so that acoustic emission source strengths can be determined. It has also been shown that measurements of the response of the transmitting transducer backing can be used to assess the coupling eciency, and hence to remove concerns about inconsistent coupling aecting the calibration measurements. The results indicate that the variation of the backing response with coupling is due to a shift in the resonance frequencies of the transducer with the mechanical load impedance. If other transducers can be shown to behave in a similar fashion this eect could be used to measure coupling in standard acoustic emission and ultrasonic transducers. 2000 Elsevier Science B.V. All rights reserved.
Keywords: Acoustic emission; Conical transducers; Point acoustic sources; Self-calibration

1. Introduction Acoustic emission techniques were rst used commercially in the late 1960s for the testing of pressurised systems for the chemical and aerospace industries and have since become widely used. Applications for which acoustic emission is routinely used include quality assurance of new vessels [1], regular short-term eld evaluations to assess vessel degradation, and continuous long-term monitoring [2]. Hydrostatic proof tests are normally carried out on new pressure vessels to 150% of the working pressure to ensure system integrity, the acoustic emission during the test being monitored. Acoustic emission has also been successfully used for the non-destructive testing of steel and pre-stressed concrete structures such as bridges, dams and skyscrapers [3,4]. Transducers are often mounted permanently to important points on the structure, such as around joints, and can be continuously monitored. The number and placement of transducers depend on the specic structure, but single transducers can be used if one particular joint in a structure is under examination.
* Corresponding author. Tel.: +44-171-594-7069; fax: +44-171-584-1560. E-mail address: p.cawley@ic.ac.uk (P. Cawley)

The experimental layout can aect some or all of the measured waveform parameters in an unpredictable fashion; for example, the relative locations of the source (defect) and receiver can alter the maximum amplitude of the resulting waveform. Consequently, measurements taken on identical structures may not be directly comparable if the source location is dierent in each case. Therefore, current methods only give a qualitative indication of the change of state of the component rather than a quantitative indication of the absolute level of damage. However, if the structure could be calibrated, by taking measurements at the selected receiver positions with a source of known intensity placed at dierent possible damage locations, it would be possible to infer the energy released by acoustic emission events detected subsequently. If the eld generated in the structure is diuse [57], the same signal amplitude is received irrespective of the source position. Also, if the eld is diuse, the received signal amplitude is dependent only on the energy input by the source, not on, for example, the direction in which the force is applied, since in a diuse eld the ratios of the dierent modes present in the structure tend to a constant, irrespective of their initial ratios. Therefore, in this case, only one calibration measurement is needed; however, in general, it would be necessary to calibrate with the simulated source at

0041-624X/00/$ - see front matter 2000 Elsevier Science B.V. All rights reserved. PII: S0 0 4 1 -6 2 4 X ( 9 9 ) 0 0 11 0 - 9

590

M.J. Evans et al. / Ultrasonics 37 (2000) 589594

all likely damage sites, and to ensure that the chosen source is representative of the force generated by the real acoustic emission event. In order to do the calibration, a simple, reproducible source is required. The most commonly used articial sources are a pencil break [8,9] or a glass capillary fracture [912]. A micro-hammer has been proposed [13], and it would also be possible to use laser excitation [14]. The fracture sources have a bandwidth up to about 1 MHz and are fairly reproducible (10% shotshot amplitude variation can readily be achieved [15]). Their major disadvantage is that they would be awkward and time consuming to use for routine, industrial calibration. The micro-hammer is more convenient to use, but in order to obtain signicant energy at frequencies above 500 kHz, a very small impactor must be used; this limits the energy in the pulse and can lead to signalnoise problems. Laser excitation is reproducible and has a wide bandwidth, but the high-power lasers required are prohibitively expensive for many applications. It should be noted that although these sources may give reproducible results on a given structure, the results obtained are a function of material, surface nish, etc. Piezoelectric transducers are used as both transmitters and receivers in most ultrasonic testing applications. They are simple, robust and can be made with wide bandwidths, so giving the possibility of tailoring the bandwidth of the excitation applied to the test structure by varying the excitation signal applied to the transducer. A conical piezoelectric transducer was developed at the National Bureau of Standards (now the National Institute for Standards and Technology) in the USA for use as a wideband acoustic emission receiver [16,17]. This typically has a small (diameter 1 mm or less) contact tip and so can be regarded as a point receiver at typical acoustic emission frequencies (<1 MHz). A device of this type could also be used as a source. However, the drawback of piezoelectric devices is that their output is very sensitive to the coupling between the transducer and the structure [7,18]. This means that the coupling must be very carefully controlled if reliable calibration measurements are to be obtained. An alternative to ensuring consistent coupling would be to devise a method of monitoring the coupling, so producing a self-calibrating transducer. This paper investigates the use of conical transducers as point acoustic sources and discusses adding a measurement element in the transducer to facilitate correction for coupling variations.

Fig. 1. Schematic diagram of self-calibrating conical transducer.

2. Performance as transmitter and receiver The basic conical transducer construction employed was similar to that used in the NIST transducer [16,17]. The construction is shown schematically in Fig. 1, but the original version did not include the Pinducer, which was introduced later to form a self-calibrating unit (see

below). The conical tip was machined from a 10 mm diameter, 3 mm thick PZT disc and had a tip diameter of 1 mm. A 50 mm thick brass shim was bonded to the tip with conducting epoxy and served as a wear plate and electrical contact; this shim was earthed to the transducer case. The wider end of the cone was bonded to tungsten-impregnated epoxy backing and the other electrical contact was made at this interface. It was found that a 14 mm thick cylinder of tungsten-loaded epoxy gave sucient damping, but it was convenient to extend the backing by a further 14 mm of pure epoxy. The backing assembly was mounted in an aluminium case. PTFE guides were used to allow the backing to move freely in and out, while keeping it oriented normal to the test-piece surface. The case was mounted on three feet to keep it clear of the surface while maintaining the correct orientation. During operation the contact tip was pressed onto the test surface with a force of 10 N, which was controlled by a spring. The compression was either maintained by pressing the case down manually or, more commonly, by applying weights to the top of the case. The transducer has been calibrated as a receiver by comparing its response with that of a laser interferometer. A typical calibration curve is shown in Fig. 2, from which it can be seen that the conical transducer is wideband and can be used satisfactorily over the frequency range of interest in acoustic emission. Further details of the calibration measurements are given in Ref. [15]. Initial tests of the performance of the conical transducer as a transmitter were carried out on a solid cylindrical aluminium block 200 mm in diameter and 144 mm long. Two similar transducers were placed 100 mm apart on a diameter of the top surface and

M.J. Evans et al. / Ultrasonics 37 (2000) 589594

591

placing the transducer, so it is likely that results on rougher surfaces, or where access is more dicult, would be less reproducible. Therefore, if the transducer is to be used as part of an acoustic emission calibration procedure, it would be desirable to incorporate some form of calibration for coupling variations.

3. Modied transducer
Fig. 2. Typical reception sensitivity of conical transducer measured in calibration experiment with laser interferometer.

3.1. Construction The idea of the self-calibrating transducer is that the amplitude of the waves generated in the transducer backing is related to the coupling of the transducer onto the test piece. Therefore, measurement of the amplitude of the waves in the backing may give the required information about the degree of coupling. The construction of the self-calibrating transducer is shown in Fig. 1 and was essentially the same as that used previously, except that a commercial broad-band transducer ( ValpeyFisher VP-1093 Pinducer) was permanently bonded into the tungsten-epoxy backing, its tip being placed 15 mm from the back face of the conical element. The Pinducer itself consists of a 2 mm diameter piezoelectric disc element mounted on the end of a 33 mm long brass stem. The small element dimensions give the Pinducer a wide operating range, which is quoted by the manufacturer to be from DC up to 10 MHz (40 dB down points). The brass backing, although well acoustically matched to the PZT element, has very low attenuation, so the typical response of a Pinducer shows some ringing within the brass backing. In the current application it is the rst arrival waveform measured by the Pinducer that is of interest and, therefore, the ringing is of little consequence. The self-calibrating transducer unit, therefore, has two independent piezoelectric elements, the intention being to use the conical element as an ultrasonic point source and the Pinducer as a passive receiver. Driving the conical element with an electrical signal generates ultrasonic waves in both the surface to which it is coupled and the backing. The corresponding signal received by the Pinducer will be termed the backing signal. This signal is rather complex, consisting of several arrivals corresponding to multiple reections within the backing and ringing within the Pinducer, as shown in Fig. 4. The back face of the conical element, the backing material and the Pinducer are permanently bonded together, making the path length and coupling between the piezoelectric elements constant. Variation in the amplitude or frequency content of the backing signal (for a xed excitation waveform and environmental conditions) can only be caused by modications in the response of the conical element due to changes in the boundary conditions present at the front face.

equidistant from the axis of the cylinder. One transducer acted as a receiver while the other was excited with a single cycle, 0.5 MHz tone burst generated by a LeCroy 9101 arbitrary function generator and amplied to 200 V peakpeak by an ENI power amplier. Fig. 3(a) shows the received time domain traces in three experiments, the transmitter being removed and replaced at nominally the same position between the experiments while the receiver remained xed. A clear signal corresponding to the Rayleigh wave arrival is produced and the spectra of the signals are shown in Fig. 3(b). This test indicates that the conical transducer performs well both as a transmitter and as receiver. However, although the three signals of Fig. 3 are very similar, their peak amplitudes span a range of around 10% from the mean. This is almost certainly due to coupling variations. The surface of the cylinder was smooth and care was taken in

Fig. 3. Signals received in three transmission experiments on an aluminium block: (a) time domain; (b) frequency domain including spectrum of excitation signal.

592

M.J. Evans et al. / Ultrasonics 37 (2000) 589594

Fig. 4. Typical signal measured by Pinducer in backing. Excitation was a ve-cycle, 140 kHz, Hanning windowed tone burst.

must remain constant and the electrical input signal to the source be unchanged. The peakpeak amplitude of the reference signal will vary from zero (corresponding to no coupling) to a maximum V (corresponding to max 100% coupling). The value of V depends on the input max voltage to the source, the material properties of the surface, the type of wave being transmitted, the distance between the source and the receiver and the sensitivity of the receiveramplier system. The value of V for max a given arrangement of transducers can be estimated by adjusting the orientation of the transducer and moving it around the surface a little until a maximum value is obtained. The percentage coupling eciency is then given by V g =100 R %, (1) c V max where V is the amplitude of the signal from the remote R receiver transducer in a given test. Initial experiments were carried out on a variety of materials using an excitation frequency of 140 kHz. Large stainless steel, brass and aluminium blocks were used to simulate half-spaces and a 5 mm thick aluminium plate was used to assess the eect of material thickness. The results obtained are shown in Fig. 6 where the amplitude of the backing signal is plotted against the measured coupling eciency, calculated using Eq. (1), and the lines are linear least squares ts to the experimental data. The backing signal amplitude was measured according to the denition shown in Fig. 4. A good t is obtained in each case, which demonstrates a linear relationship between backing voltage and coupling eciency. This linear relationship is most evident at coupling eciencies above 35%; below

3.2. Results The self-calibrating transducer was tested using the setup shown in Fig. 5. A signal generator was used to excite the conical element with a ve-cycle, Hanning windowed tone burst with an amplitude of 200 V peak peak. The backing signal, measured by the Pinducer, was amplied using a 40 dB preamplier and sampled at 10 MHz using a digital storage oscilloscope. An additional conical transducer was permanently bonded to the structure with cyanoacrylate adhesive, and used as a reference transducer to measure the signal received at a remote location on the surface. The signal from the reference transducer (subsequently referred to as the reference signal ) was also amplied using a 40 dB preamplier and sampled simultaneously by the oscilloscope. The amplitude of the rst arrival of the reference signal was used to indicate the coupling eciency g ; c for this to be valid the distance from source to receiver

Fig. 5. Setup used in tests of self-calibrating transducer.

M.J. Evans et al. / Ultrasonics 37 (2000) 589594

593

Fig. 6. Backing amplitude as a function of coupling eciency for selfcalibrating transducer on stainless steel, brass and aluminium blocks, and 5 mm thick aluminium plate. Input was a ve-cycle, 140 kHz Hanning windowed tone burst.

this value the backing amplitude tends rapidly towards the 0% coupled case. Coupling eciencies of less than 35% were dicult to achieve, requiring deliberate tilting and lifting of the transducer. It is concluded that these coupling eciencies are unlikely to occur accidentally, making this loss of correlation unimportant. In all cases the backing signal drops with coupling eciency, with a minimum occurring at 100% coupling, which will be termed V %. Clearly, the backing ampli100 tude at no coupling, V %, is the same for all cases, but 0 the value of V % is material dependent. The minimum 100 V % was measured for steel (with aluminium having 100 the highest and brass in between); this indicates that the acoustic impedance of the material is the controlling factor, with the higher impedance materials having lower V %. The results obtained for the aluminium 100 half-space and the 5 mm thick aluminium plate are similar, indicating that, with plate thicknesses of 5 mm and above, the material impedance is the controlling factor. An additional experiment was carried out on the steel block with an excitation frequency of 850 kHz, and the results are shown in Fig. 7. The eect of increasing the

coupling eciency of the transducer now has the opposite eect on the backing signal, although the magnitude of this change in backing amplitude is much smaller than that previously measured at 140 kHz (note the dierent scales on Figs. 6 and 7). The reversal in the relationship between backing voltage and coupling eciency has also been measured for aluminium and brass, but the change in backing amplitude was so small that it was dicult to measure using the current experimental setup. It is thought that the variation of backing amplitude with coupling is controlled by the eect of coupling on the resonance frequencies of the transducer. Fig. 2 indicates that the transducer has a (damped) resonance at around 0.5 MHz. This resonance frequency and its variation with coupling may be predicted using the model of the conical transducer developed by Greenspan [17] who extended the well-known Mason model [19] to account for the conical piezoelectric element. This model predicts that the resonance frequencies increase as the impedance of the surface to which the transducer is coupled increases. Increasing the coupling eciency eectively increases the load impedance, so the amplitude of vibration of the transducer backing as a function of frequency will change as shown schematically in Fig. 8. The excitation signal used to drive the transducer in the experiments of Fig. 6 was a ve-cycle tone burst whose energy was concentrated in region (a) of Fig. 8, representing the below resonance case, whereas the energy was concentrated in region (b) in the above resonance case of Fig. 7. The eect of a slight increase in the resonance frequency is to decrease the response of the transducer within region (a) and to increase the response in region (b). This explains the reduction in backing amplitude at a given coupling eciency seen in Fig. 6 as the impedance of the load material increases. It also explains the reversal in the eect of coupling eciency on backing voltage when the operating frequency is above the resonance. The eects shown schematically in Fig. 8 were demonstrated in an experiment on the steel block. The excitation signal used was a single-cycle tone burst at 450 kHz,

Fig. 7. Backing amplitude as a function of coupling eciency for selfcalibrating transducer on stainless steel block. Input was a ve-cycle, 850 kHz, Hanning windowed tone burst.

Fig. 8. Schematic diagram of the eect of coupling on the backing amplitude.

594

M.J. Evans et al. / Ultrasonics 37 (2000) 589594

the band of interest. If other transducers can be shown to behave in a similar fashion this eect could be used to measure coupling in standard acoustic emission and ultrasonic transducers.

Acknowledgement This work was funded by an EPSRC CASE studentship for Dr Evans in conjunction with Rolls Royce plc.
Fig. 9. Spectra of backing signals obtained at 0, 50 and 100% coupling on a stainless steel block. Input was a single cycle, 450 kHz tone burst.

References
[1] N.O. Cross, Acoustic Emission Testing of Pressure Vessels for Petroleum Reneries and Chemical Plants, ASTM, Philadelphia, 1982. [2] P. Tscheliesnig, H. Theiretzbacher, Leakage test by acoustic emission testing (AET ) on at bottomed tanks, Proceedings of the Second International Conference on Acoustic Emission, Lake Tahoe, 1985, pp. 2427. [3] S.E. Dunne, Discussion of the application of acoustic emission methods for eld monitoring of corrosion induced structural deterioration of reinforced and pre-stressed concrete structures, Florida Atlantic University, Boca Raton, 1985. [4] S. Lovass, Acoustic emission of oshore structures: attenuation, noise, crack monitoring, J. Acoust. Emission 26 (8) (1985) A161A164. [5] D.M. Egle, A stochastic model for transient acoustic emission signals, J. Acoust. Soc. Am. 65 (5) (1979) 11981203. [6 ] R.L. Weaver, Diuse waves for materials NDE, AcoustoUltrasonics, Plenum, New York, 1988. [7] M.J. Evans, B. Pavlakovic, P. Cawley, Measurement and prediction of diuse elds in structures, J. Acoust. Soc. Am. 106 (6) (1999), in press. [8] N.N. Hsu, Acoustic emission simulator, US Patent no. 4 018 084, 1976. [9] J.E. Michaels, T.E. Michaels, W. Sachse, Applications of deconvolution to acoustic emission signal analysis, Mater. Eval. 39 (1981) 10321036. [10] F.R. Breckenridge, Acoustic emission transducer calibration by means of the seismic surface pulse, J. Acoust. Emission 1 (2) (1982) 8794. [11] J.E. Michaels, T.E. Michaels, W. Sachse, Applications of deconvolution to acoustic emission signal analysis, Mater. Eval. 39 (11) (1981) 10321036. [12] Y.H. Kim, H.C. Kim, Source function determination of glass capillary breaks, J. Physics D: Appl. Phys. 26 (1993) 253258. [13] D.G. Chetwynd, W. Sachse, Design of micro-hammers for ultrasonic source applications, Ultrasonics 29 (1991) 6875. [14] D.A. Hutchins, Ultrasonic generation by pulsed lasers, in: W.P. Mason, R.N. Thurston ( Eds.), Physical Acoustics, Vol. 18, Academic Press, London, UK, 1989. [15] M.J. Evans, The use of diuse eld measurements for acoustic emission, PhD Thesis, Imperial College, University of London, 1997. [16 ] T.M. Proctor, An improved piezoelectric acoustic emission transducer, J. Acoust. Soc. Am. 71 (5) (1982) 11631168. [17] M. Greenspan, The NBS conical transducer: analysis, J. Acoust. Soc. Am. 81 (1) (1987) 173183. [18] M.G. Silk, Ultrasonic Transducers for Non-Destructive Testing, Adam Hilger, Bristol, 1984. [19] W.P. Mason, R.N. Thurston, Physical Acoustics, Principles and Methods, Academic Press, London, 1972.

having energy over a broad bandwidth from 50 kHz to above 1 MHz. The backing signal was recorded and the rst arrival pulse of Fig. 4 was gated to remove the multiple reections in the backing and the ringing of the Pinducer. The resulting time trace was Fourier transformed and the resulting spectra are given in Fig. 9 for coupling eciencies of 0, 50 and 100%. The small resonance frequency shift caused by the coupling changes can be seen, together with a decrease in response with coupling eciency below 0.5 MHz and an increase above 0.5 MHz. The variation of the sensitivity of the backing response to changes in coupling eciency with frequency from a negative correlation with frequency to a positive correlation through zero means that the proposed calibration procedure will not work well with all excitation signals. However, if necessary, the coupling eciency could be measured using a narrow-band input at a frequency at which the backing response is sensitive to coupling. Having determined the coupling, the required experiment could then be carried out without moving the transducer.

4. Conclusions It has been shown that conical piezoelectric transducers can be used as point transmitters over the frequency range of interest in acoustic emission measurements. They should, therefore, be suitable for use in experiments to calibrate structures so that acoustic emission source strengths can be determined. It has also been shown that measurements of the response of the transmitting transducer backing can be used to assess the coupling eciency, and hence to remove concerns about inconsistent coupling aecting the calibration measurements. The results indicate that the variation of the backing response with coupling is due to a shift in the resonance frequencies of the transducer with the mechanical load impedance. Further work is needed to develop the idea into a fully practical technique at all frequencies across