1) INTRODUCTION

Piezoelectric ceramics are widely used in various applications like low and high power ultrasonic generation and reception, ultrasonic cleaners, SODAR, SONAR, medical diagnostics, printer heads, gas lighters, micro positioners, actuators, annunciators and memory devices in thin film form etc.

The most commonly used piezoelectric ceramics are Lead Zirconate Titanate (PZT) with several additives giving rise to PZT-A to H varieties, Lead Lanthanum Zirconate Titanate (PLZT) and Barium Titanate (BaTiO₃). These ceramics before being put to use in different applications, have to be poled to orient the domains in a given direction. This poling is done by the application of an electric field greater than the coercive field, across the specimen, which is kept at a suitably high temperature. After some time the sample is allowed to cool to room temperature with the field appplied, to 'freeze' the positioning of domains. Poling is said to be complete, when all possible domains align themselves in the direction of the field applied and the specimen yields high values of piezoelectric parameters. In order that the poling is complete, the parameters like the field to be applied, the duration for which it is applied and the temperature at which it has to be poled, have to be optimised. It has been found that excessive poling parameters bring about micro cracking in the specimen and this in turn deteriorates the performance parameters of the piezoelectric ceramic which is reflected in the decrease of its piezoelectric parameters like d₃₃ - charge coefficient, k_p - electro mechanical coupling coefficient, Q_M - mechanical quality factor and P_r - remanent polarisation (Cheon et al, 1991, Nejezchleb etal, 1988, Chung and Kim, 1987, Freiman, 1986). Thus the useful life of operation of the device reduces. Similarly, in high power applications, the piezoelectric sample will be subjected to large ac electric fields resulting in repeated polarisation reversals. Due to this continuous switching, alternating mechanical stresses are induced piezoelectrically and result in the initiation and propagation of micro crack. Thus the initiation and propagation of micro cracks during poling and during electric fatigue in piezoelectric ceramics both in bulk and thin film form, is a matter of serious concern to the manufacturer as well as the user of the above ceramics (Hill et al, 1996, Lynch et al, 1995, Jiang et al, 1994). The optical method of in-situ monitoring of domain alignment is not only labourious and time consuming but also not a practically viable tool for industry. Thus, there exists a need to monitor in-situ, either during poling or under actual usage, the micro cracking in piezoelectric ceramics by means of a simple technique. Also, there exists a need to check the quality (ie the existance of micro / macro cracks which might grow under usage) of a poled piezoelectric ceramic at the user end. The authors have addressed this industrially useful problem of discriminating the domain alignment and micro cracking mechanisms based on pattern recognition of AE signal waveforms.

2) EXPERIMENTAL

The main aim of the study was to monitor micro-cracking during poling of piezoelectric ceramics using acoustic emission. Since it is known that there can be several other sources of AE apart from micro cracking, it was proposed to obtain AE signals from all the possible different sources by using varied experimental conditions on poled and unpoled specimens to build a knowledge base. Studies were made on PZT (both soft and hard varieties) circular disk samples of 20 mm dia and 1 mm thick as purchased from M/S Central Electronics, Sahibabad, India and M/S Concord Electroceramic Industries, New Delhi, India. Studies were also made on in-house prepared BaTiO₃ samples of specific porosity.

The basic AE experimental set up is shown in fig. 1. The acoustic signals obtained from the sample (S) under study are led by a commercial aluminum cylindrical (48 cm x 9 mm) wave guide (WG) to a wide band transducer (100 kHz to 2 MHz) SE2MEGP [Dunegan Engineering Consultants Inc. (DECI), USA ] - 'Td' and the resultant electrical signals are amplified (80 dB), sampled and digitally stored and also processed on-line to give ring down count - RDC and event count - EVC. The digitally stored signals are analysed off-line. The sample is kept immersed in silicone oil kept in a brass cup (B). The temperature of the sample can be varied by keeping the entire setup in a temperature controlled oven ( ± 2^o C ). The extent of poling or depoling is quantified by a measurement of piezoelectric parameters d₃₃, Kp, Qm, Ps, and Ec at various stages of experimentation. Further,microphotographs of the body of the ceramic (fractographs) are used to visualise the damage (i.e., extent of micro or macro cracking, pore discharges etc.,) within the sample.

Fig 1: Block Diagram of the Acoustic Emission Experiment Setup.

3) SIGNAL ANALYSIS

As mentioned in the beginning , it was reasoned, that there could be several sources of AE during the experimental studies on these ceramics and variation of RDC and EVC alone can not identify and discriminate between them. However it was expected that the AE waveform could reflect the source characteristics to a certain extent. This was fortified from the fact that under similar experimental conditions different AE signal waveforms were observed. Thus waveform analysis was undertaken.

The authors used only 180 micro second length (being the return travel time of AE signal in the waveguide used) of the AE signal and modelled it using auto regression (AR) of the 12^th order. AR coefficients were obtained using Yule- walker's method.

Selection of the order of AR model has been made based on the following criterion and this verification was made on every AE signal used in the present work:


1. The variance ie the root mean square of the error signal must be minimum
2. The power spectral density of the AR modeled signal must approximately equal that obtained from a Discrete Fourier Transform of the actual AE signal.
3. The residual signal (difference between between the AE signal and the AR model) must consist of random noise and hence has a fourier spectrum which is white noise-like i.e. the frequency spectrum should be flat and not contain any significant peaks. This can also be checked by taking the autocorrelation of the residual signal which should give a single peak at a lag=0 (lag 1 means, one sampling interval).

The 12 AR coefficients representing the signals were used in a supervised pattern recognition as described later in the next section. All the signal analysis was done under MATLAB environment.

4) RESULTS

Initially, to form a knowledge base or "Learning set" of AE signals, specific experiments were performed in such a way that the signals could have resulted predominantly from only one source mechanism. The type of the experiments performed and the concerned source mechanisms responsible for AE signals obtained are listed below.

a) Propagating Macro crack (category - M)
The AE signals which were captured just before the physical breakdown of the sample during the application of a large dc field, 3 to 4 times E_c, would have resulted from a propagating macro crack. A representative time domain signal (top trace) and the microphotograph (bottom) of such an experimented sample is shown in fig. 2. The photograph Clearly indicates the presence of a propagated macro crack (circled).

b) Macro crack (category - C)
AE signals obtained during the application of large dc (3 to 4 times E_c ) fields- but much before physical breakdown of poled / unpoled sample are likely to represent the above source. Obviously several trials were made to achieve this. Fig.3 shows macro cracking in a PZT-A sample (bottom photograph), which was subjected to 40 kV/cm dc voltage at 100^° C for about 10 minutes and the corresponding representative time domain AE signal (top trace). Note the difference in magnification scales used in figs. 2 & 3.

Fig 2: Time domain AE signal of PROPAGATED MACROCRACK- M category and microphotography of a propogated macrocrack in PZT-A (X-100)

Fig 3: Time domain AE signal of MACRO CRACk-C category(top) and microphotograph of a macrocrack in PZT-A (X-500)

c) Pore discharge (category - P)
AE signals captured during the application of medium dc fields (2 to 3 times E_c ) on highly porous specimen (10 to 20% porosity) would arise due to partial electrical discharges in the pores of the ceramic. PZT specimens having about 10% porosity were chosen from among a large number of samples obtained from the manufacturer (this was indeed difficult as most of the specimens obtained had porosity < 5%). However, BaTiO₃ specimens with a porosity of 13% were easily prepared in one batch by increasing the amount of binder used for mixing the powders to make green compacts. A representative Fig. 4 (bottom photograph) shows the damage caused around a pore by the electrical discharges in the cavity for a PZT-A specimen of 10% porosity with 30 kV/cm field application at 100^° C. Top trace of Fig.4 shows the corresponding time domain AE signal.

Fig 4: Time domain AE signal of pore DISCHARGE-P category (top) and microphotography (x-500) of 10% porous PZT-A sample showing the destruction around the pore cavity.

d) The following two categories were found to give similar type of AE signals and are described below.
Debonding of electrode (category - B)
AE signals obtained from samples which were coated manually by silver paste (but not cured at high temperature) and subjected to medium dc fields (2 to 3 times E_c) at 100^°C, with sample immersed in oil, fell under this category. Fig. 5(a) shows a typical AE signal obtained and Fig.5(b) shows the surface of a PZT- A sample subjected to such a study. The sample surface became black (which otherwise is yellowish in color) due to the electrical discharges taking place in between the electrode and the top surface of the ceramic. Fig. 5(c) shows the peeled out electrode from the above sample which came out as a thin sheet of silver after experimentation.

Fig 5: Time domain AE signal of DEBONDING- B CATEGORY (a), photography of the blackned surface of the PZT-A sample (b) and the debonded silver electrode (c) which peeled off easily after expt.

Grain pull-out (category - B)
This was specifically observed only when a PZT-A sample was cleaned in an ultrasonic cleaner in order to degrease the surface. However after cleaning, the etched out surface showed regions where grains were missing (fig. 5d). The same sample was recoated with silver electrodes cured at 650 ^°C for one hour to preclude the possibility of debonding of the elctrodes and later subjected to 25 kV/cm poling field in oil for about 5 minutes. This gave rise to several AE signals. The microphotograph of this worked out sample (fig. 5e) showed wide regions from where the grains have been dislodged. The AE signal for this category was similar to the one showed in fig.5(a).

Fig 5a: Microphotographs of PZT-A sample showing GRAIN PULL OUT after ultrasonic cleaning but before expt. (d) and showing extensive pullout after expt (e) the AE signal for this phenomenon fell under B-category (Sa)

e) Domain alignment / fall backs (category - D)
AE signals captured during the initial stages of application of low poling field on unpoled samples at RT or 100^°C, or around phase transition temperature for poled samples - during temperature rising cycle or the signals obtained during the cooling phase of poling when the poling field has been switched off wantonly, would fall under this category. The last category of signals are caused by the already aligned domains falling back to their original position. A typical time domain AE signal is shown in the top trace of fig.6. However, the microphotograph (500 magnification) of the PZT-A sample which has undergone poling study for a limited time of about 2 minutes looked exactly the same as that of an unexperimented sample (bottom photograph of fig.6). This should be so, as the AE signals could have come from micro deformations in grains with no permanent damage to the microstructure.

Fig 6: Time domain AE signal of DOMAIN- D category (top) and Microphotograph (x-500) of the PZT-A sample which gave the above signal.

f) Surface to surface electrical discharges (category - S)
AE signals obtained during DC field application on unpoled samples which are not immersed in oil would give rise to electrical discharges across the surface of ceramic around it's periphery. Similarly DC field application on unpoled samples (immersed in silicone oil) with weak peripheries can also give rise to the above source. Fig.7-top trace, shows a typical AE signal belonging to this category and the surface (bottom photograph) of a Titanium electroded BaTiO₃ sample which was subjected to 25 kV/cm electric field at RT. The whitish spot seen on the left is the weak peripheral part which got damaged due to surface to surface electrical discharges.

Fig 7: Time domain AE signal of SURFACE DISCHARGE- 5 Category(top) And photograph of a titanium electrode BatiO3 sample showing a white spot where the electrode got burnt out due to surface discharge.

g) Initiation of microcracking (category - I)
AE signals obtained during poling study from samples (immersed in silicone oil) subjected to medium electric fields of about 30 kV/cm. It is observed usually in such an experiment there will be a sudden spurt of AE activity as soon as the field is applied. This activity which is due to domain alignment is followed by a lull and then once again the AE activity starts. The signals obtained in the second round of activity are usually due to initiation of micro cracking. A typical AE signal belonging to this category is shown in fig.(8) and a micro Photograph of the body of the ceramic at 500 X looked like that of an unexperimented sample. However, micro cracks could have got un-noticed because of the low magnification used. In the literature (Jiang, 1992 and Cheon et al.,1991) micro cracks have been noticed at a magnification of 1000 and above using Scanning Electron Microscope.

Fig 8: Time domain AE signal of INITIATION OF MICROCRACKING - I category in PZT-A sample.

The authors have segregated several signals in each category which arose under specified experimental conditions as mentioned in (a) to (g) above, and have prepared a Pattern Recognition (PR) chart for these signals. A good clustering of signal points was obtained when AR parameter-1 (on x-axis) was drawn against AR parameter-2 (on y-axis). This PR chart is shown in fig.9 and is referred as Master Learning Set (MLS) in future. In this chart, each AE signal is represented by a serial number. Seven clusters were noticed for the 7 categories of AE source mechanisms. This chart was used for future identification of AE signals from specimen which were subjected to different experimental conditions.

It may not be out of place to mention that, under each of the above categories several samples (minimum of TEN at least) were tested and only representative signals - out of the many which were captured and analysed, are shown in the MLS. This is done to avoid crowding of signal points in a given category leading to loss of clarity.

Fig 9: PR chart of the Master Learning Set (MLS). The elliptical boundaries from the necessary template for MLS.

After the MLS had been prepared, each one of the AE signal obtained during any test and under any experimental condition is classified as follows. The signal is first AR modeled & checked whether 12^th. order is sufficient as per the mentioned criterion and then the signal is plotted as a point on MLS by taking it's AR parameter-1 and parameter-2 values. If the signal falls into any of the categories of MLS, the signal is accordingly labeled as belonging to that category. Alternatively, for a given specimen and under a given set of experimental conditions, all the possible signals captured are analysed i.e., AR modeled and the necessary AR parameters obtained. A PR chart of AR parameter-1 Vs AR parameter-2 for all the signals, with the same scales as that of MLS is plotted. Then the external elliptical boundaries of the different categories in MLS are drawn on the present PR chart using a MLS Template as shown in fig. 9. The signals falling with in an elliptical boundary belong to that category or class. This was verified on several samples vis-à-vis their fractographs.

5) CONCLUSIONS

It is shown in this study that it is possible discriminate and differentiate between several source mechanisms of AE that occurr in PZT ceramics using pattern recognition of AE signal waveforms. The authors made studies on PZT ceramics during poling, phase transition and ac fatigue. These results which formed a part of first author's doctoral thesis work (Aparna, 1999), are being communicated elsewhere.

With suitable modifications one can make the studies, on-line, in order to monitor, in-situ the poling process. Thus AE technique can be used as a monitoring tool for optimal poling of the PZT ceramic samples with minimized micro cracking. The technique can probably also be used to test the quality of a PZT sample as regards the presence of micro cracks since any growing micro-crack would emanate AE signal which could be identified.

6) ACKNOWLEDGEMENTS

The authors would wish to greatfully acknowledge the help rendered by Dr. Harlod Dunegan of DECI, USA, in gifting the broad-band transducer used in this work and Prof.LE.Cross of The Pennsylvania State University for sending a copy of his student's thesis (Jiang, 1992). The first author wishes to thank UGC , New Delhi, for the award of a fellowship.

7) REFERENCES

1. Aparna,D., "Pattern recognition of acoustic emission signals from piezoelectric ceramics" Doctoral Thesis submitted to Osmania University, Hyderabad, India 1999.
2. Cheon Chae-Il, Kim,S. and Kim,H, Ferroelectrics, 115, 35-42, (1991).
3. Chung,H.T. and Kim,H.G., Ferroelectrics, 76, 327-333 (1987).
4. Freiman,S.W., Proc.Sixth IEEE Intl.Symp. on Applications of ferroelectrics, Leigh Univ., Bethelhem, 367 (1986).
5. Hill, M.D.,White G.S., Hwang C.S. and Lloyd,I.K., J.Am.Ceram.Soc., 79, 1915 (1996).
6. Jiang,Q.Y., Ph.D.Thesis : "Electrically induced fatigue effects and reliability in piezoelectric and electrostrictive ceramics", The Pennsylvania State University, December (1992).
7. Jiang,Q.Y., Subbarao,E.C.,and Cross,L.E., Ferroelectrics, 154, 119-124 (1994).
8. Lynch,C.S., Yang,W., Collier,L., Suo,Z. and McMeeking,R.M., Ferroelectrics, 166, 11-30 (1995).
9. Nejezchleb,K., Kroupa,F., Boudys,M. and Zelenka,J., Ferroelectrics, 81, 339-342 (1988).

© NDT.net - info@ndt.net

|Top|