Using acoustic emission techniques to aid physical testing of leather

Introduction Due to the heterogenous, fibrous nature of leather, a great deal of information can be gained through the monitoring of acoustic emissions during testing. This research may provide fundamental theories as to the mechanisms occurring in leather under stress. When a force is applied to any material, resulting movements within the structure form sound waves. These sound waves can be monitored as acoustic emissions. The definition of an acoustic emission is the transient elastic wave generated by the rapid release of energy within a material¹. Another way to describe this is that when a load is applied to a specimen, energy is stored as strain energy. When an inherent critical point is reached in the sample, there is a sudden redistribution of the energy. At this point, some of the strain energy is converted into acoustic energy². Various properties of sound waves can be measured, the most common being the amplitude and frequency. The amplitude is a measure of loudness, the frequency is the number of cycles per second (hertz) and is perceived as the pitch of the sound. Generally, the frequency range that is studied is the ultrasound region. In some cases, the audible range is evaluated. Acoustic testing is an area that has been incorporated into many industries. Typically, it is used for the non-destructive testing of materials such as pressure vessels. Some other applications are mentioned below. However, it should be noted that a large number of studies have been performed on acoustic emission data³ because it is such a useful non-destructive testing method and so the list is not exhaustive. Results obtained have been correlated with properties such as the applied stress and the stress intensity at a crack tip. Another application is the evaluation of fundamental deformation and failure behaviour of geological materials⁴. In this case, both the acoustic activity and the frequency spectra were evaluated. It should be noted that there are difficulties in interpreting frequency spectra due to a number of issues. These include the inherent resonance frequencies in the test materials and loading system, the frequency dependent attenuation characteristic of the test material and the need to ensure that the test equipment has a flat/linear response in the region of interest. It is worth discussing the issue of attenuation of sound waves1, as it has implications for the work reported here. As a wave travels through any material, the peak amplitude diminishes. It is harder to detect sources at a greater distance from the sensor (a useful analogy being that it is more difficult to hear someone speak when you are some distance from them). Some causes of this phenomenon are the geometric spreading of the wave front, absorption or damping of the propagating medium and the 'leaking' of the wave form into adjacent media. Acoustic emission studies have been applied to areas where it is difficult to observe the sample under test. An example of this is the testing of a fibre-reinforced composite material⁵ where acoustic techniques enabled an assessment of the breakage of fibres while within the matrix. Also of importance is that the onset of failure can be detected before actual catastrophic sample failure occurs⁶. This allows action to be taken to prevent the destruction of the test sample. Work has been carried out (mostly in the USA) to evaluate the potential for acoustic emission studies in leather. The work^7,8,9,10 investigated the emissions within the ultrasound range. Various parameters were investigated, such as the effect of mechanical softening⁷, and the possibility for non destructive tensile testing⁹. Some information was gained concerning the mechanisms of leather failure and deformation under strain⁸, and the effects of applying polymers to soften leather¹⁰. While published work has mostly concerned the evaluation of the ultrasound region of acoustic emissions, there are some drawbacks to this approach. As the frequency of sound increases, its attenuation increases also, ie higher frequency sound does not travel as far as low frequency sound. In the ultrasound region, the sound waves do not propagate through air, and a coupling material is required. This is the reason behind using gels when carrying out medical ultrasound examinations. However, the presence of any gel or grease to improve propagation of sound between leather and the sensor could interfere with the results. Some work has been published linking the audible acoustic properties of materials with subjective properties. Of particular interest is work carried out by the food industry to evaluate the crispness of foods^11,12. The main disadvantage of the audible range is the effect of background noise. This is an area that is addressed within this work. Experimental methods - apparatus The system that has been evaluated is shown in Figure 1. A materials tester (in this case a MT-LQ materials tester) was used to carry out the physical testing of the sample. A commercially available tie clip microphone was used to record the sound emitted and a computer with a sound card and software processes the data (the software used was a package called 'Cool Edit' which was designed for the music recording industry). Within this work the audible range was evaluated between 200 and 22,000Hz. Production of an acoustic chamber One possible problem with carrying out the analysis within the audible range of frequencies, is the presence of background noise. Therefore, in order to obtain audio information from product measurements, it is necessary to be able to achieve a suitable ratio between the amplitude of the wanted signal and any unwanted background noise. This is particularly pertinent when determining the acoustic emissions from products which are being physically deformed, as the process of producing physical deformation involves the use of mechanical instruments that by their very nature emit sounds themselves. By using a directional microphone mounted very close to the sample, it is possible to get a useful gain on the wanted signals, and a beneficial attenuation of the unwanted machine noise. Further to this, it should be possible to make a cabinet or similar structure that provides a degree of acoustic isolation between the test sample and the environment. Several microphones were chosen. To evaluate the response of the microphones, an audio signal was produced. This contained elements of all frequencies in the band of interest in equal amplitudes, with more than one frequency being present at any given time (white noise was generated which is a random and uniform mix of noise at all frequencies in the audio band). The recorded signal was then subjected to a fourier transformation to extract the relative amplitudes of individual frequencies of interest. This approach is achievable using mainly proprietary equipment. Several types of microphone were assessed for their suitability to this application. A large studio microphone, the Audio Technica AT30M, (a small diaphragm back-electret condenser based design) was compared with an Altai tie clip microphone intended for conference usage. Both microphones incorporate a transducer that works on the principle of measuring the variance in capacitance caused by sound waves displacing one of the capacitors' plates. There was a slight difference between the two, but overall the tie clip microphone performed so well for its size that its advantages far outweighed any slight limitation in performance. Due to the problems associated with the manufacture of an acoustic cabinet, other options for noise reduction were considered that were tailored to the specific test being evaluated. This, in conjunction with carrying out testing in a relatively quiet environment, is sufficient for routine testing purposes. It is envisaged that the acoustic chamber would only be required if a quiet environment is not available. Acoustic emission to aid lastometer tests The official lastometer test for the leather industry is detailed in method SLP9 (IUP/9; BS 3144: method 8) 'measurement of distension and strength of grain by the ball burst test'. The method is designed to evaluate the suitability of a leather for the lasting process during shoe manufacture. During the test, a disk of leather is clamped around its circumference, and a probe pushed perpendicular to the flesh side until cracking of the grain surface is observed. Testing can then continue if required until the probe punctures the leather. Measurements are taken of the distension required to cause grain crack and ball burst. Stosic13 described the use of a materials tester to automate the technique and this will be used during this research. Stosic also mentions that it is possible to determine the point of grain crack via deflections in the force/displacement curves obtained during testing. While this may be the case in a large majority of samples, the deflection can be quite subtle and difficult to distinguish from other events on the curve. It is, therefore, necessary to observe the test and determine the point of grain crack visually. This is time consuming and may result in inaccuracies due to human reaction speeds. The project, therefore, investigated the use of acoustic emission techniques to evaluate the incidence of grain crack in leather. Optimisation of the testing conditions As discussed previously, the presence of background noise needs to be considered when carrying out acoustic emission studies in the audible range. This is especially true when trying to evaluate a subtle event such as the grain crack. The use of an acoustic chamber is one approach to avoiding unwanted background noise. However, it is not practical to build a system to remove noise in the low frequency range. An alternative approach was the construction of an array of low cost microphones mounted in a physical layout designed to increase the rejection of background noise. The design of the lastometer test jig is such that it is possible for microphones to be mounted very close to the sample surface. The microphones were mounted in an annular ring around the lastometer jig so that the microphones all pointed to the surface of the sample and were equally distant from the centre of the jig (see Figure 2 and 3). Sound emanating from the centre of the jig will have a wave front that reaches all the microphones at the same time. If all the output signals are summed, the sounds from the centre point should converge, giving a large output signal. The annular ring was produced from a stiff foam material and four microphones were fitted into holes in the ring all pointing towards the centre of the unit. This was then fitted to the outside of a Lastometer jig. Testing of the new four-microphone array showed an improvement in noise rejection performance of 1.5 times over that achieved from a single microphone. While this is below the theoretical maximum possible, it should be noted that a constant wave sound source (which was used for evaluation purposes) is not random. The answer predicted by the statistics of random events should not be expected to hold for non-random wave forms. When background noise occurs, it is unlikely to emanate from the centre of the transducer array. The wave fronts will therefore reach different microphones at various times. The summation of all the transducer noise outputs will not be accumulative. The output from some transducers will cancel the output from others, such that the overall gain given to these signals is less than that experienced by signals emanating from the centre of the array. From these experiments, it is clear that the array of microphones offers a useful improvement in noise rejection over a single microphone. Evaluation of the lastometer test Lastometer testing was carried out on leather samples while recording the acoustic data. During testing, it was possible to note the time at which the grain was observed to crack by pressing the space bar on the computer keyboard. This results in a marker on the force/displacement curve. In order to allow the acoustic data to be correlated to the data from the materials tester, the acoustic data was collected from two microphones using a stereo technique. This allowed one microphone to record the sound of the key press, while the other detected any acoustic data from the leather sample. (Preliminary trials ensured that the microphone monitoring the leather did not detect the keyboard noise.) A transient drop in the force usually accompanies grain crack. This is seen as a deflection in the force displacement curve. Upon studying the acoustic data, it was seen, as reported by Stosic13, that there is a definite deflection in the force/time curve at the point of grain crack. In some cases this was not readily visible on the force graph and so visual inspection is still currently required during routine testing. During these experiments, however, the acoustic event was found to occur at the point of deflection on the force/time curve. In some cases, the visual detection of grain crack occurs slightly after the acoustic detection but this is indicative of the reaction times of the operator. By using the acoustic method to detect grain crack, it was observed that in many of the samples that do not initially appear to exhibit a drop in the force at grain crack, there was a subtle deflection. Several leathers were selected to be tested according to the lastometer test. This was to determine whether the technique was sensitive enough for routine testing. Testing of commercial leather samples illustrated that it is possible to correlate the observed grain crack with the acoustic data and this is seen in Figure 4. During the testing, there were just two leathers that did not produce an acoustic event at grain crack. However, these leathers were not designed for use in shoe uppers and so may not typically be tested using this method. This is further proof of the success of the acoustic lastometer test and its ability to reliably detect grain crack. Evaluation of the aesthetic properties of leather The subjective properties of leather are currently assessed manually and described through the use of a unique vocabulary such as soft, tinny, round or loose. Currently, there are no methods to quantify these subjective parameters. Research has been carried out to develop a series of test methods, the results from which can be combined to allow leather handle to be quantified. Full details of this work will be published separately. However, additional work has been carried out to evaluate the potential for acoustic emission techniques to be used to aid the evaluation of leather handle. Experimental methods The apparatus used during this section of the research was similar to that used for the lastometer test. The major difference was the use of a single tie clip microphone rather than the array. By using a directional microphone mounted very close to the sample, it is possible to get a useful gain on the wanted signals, and a beneficial attenuation of the unwanted machine noise. This, in combination with carrying out the analysis in a quiet area, was more than adequate for these experiments. Evaluation of tearing tests using acoustic emission To minimise the chance of interference from background noise, most of the experimental work was done in a quiet area with the cooling fan from the materials tester disconnected during analysis. Even with optimised acoustic conditions etc, there is always likely to be some background noise. The computer software can compensate for this by subtracting a pre-recorded waveform of background noise from the waveform of interest and this was carried out on all the samples analysed. The waveform is comprised of a series of acoustic events or sound bursts. These have a characteristic shape. It should be noted that initially, the shape was typical of a 'ringing' effect, and was an artefact of the test jig vibrating after each pulse from the leather. Damping of the test jig jaws removed this problem and Figure 5 illustrates the results after this. These pulses are typical of damped harmonic motion seen, for example, in a musical string being plucked. It was considered initially that the breakage of individual fibres or fibre bundles was being monitored using this technique. To investigate the technique, preliminary samples were evaluated that had been prepared using different post tanning conditions. This was to determine the sensitivity of the technique to different leather types and determine the potential for any further analysis. It is clear that there are more acoustic events occurring in the sample treated with a higher offer of mimosa. It is also possible that the amplitude distribution may be different, however it is not possible to quantify this at the moment. Comparison of these samples with a leather treated with a waterproofing fatliquor illustrates the extremes of acoustic behaviour that can be obtained. This is almost certainly indicative of variations in the failure mechanisms of the leather types. The waterproof leather shows few events, however the fibres must be breaking for failure to occur. It is therefore possible that the breakage of bonds between fibres rather than the breakage of the fibres themselves is being observed. This is in agreement with the conclusions of ERRC4 in their study of mechanical softening of leather. It is quite possible that the data obtained in the audible region is related to the subjective parameters of leather (such as softness). This is indicated by the extreme differences observed in the acoustic emissions from the post tanned samples. During the previous experiments detailed above, qualitative observations on many acoustic emission waveforms were carried out. One easily identifiable common factor was the way that many of the sounds were constructed from individual bursts or pulses of sound. These featured an initial sound event followed by an exponential decay. The events varied greatly in amplitude but occurred with a greater regularity during times of greater material deformation. A software programme was developed by Stable Micro Systems Ltd in Visual Basic to allow the identification and tagging of these acoustic pulses (referred to henceforth as 'events'). The programme applied a user specified amplitude threshold to the acoustic emissions waveform to allow the detection and time stamping of individual events. The programme also applied a 'dead time' after identified events during which any further acoustic information was ignored. This was to allow the exponential decay of the event to fall to a level below the detection threshold before continuing with the detection process. Hence multiple triggers from single events could be avoided, provided the threshold and dead time were set to suitable levels. Output from the programme was in the form of an ASCII text file consisting of a list of times that corresponded to detected events. This file was then imported to analysis software in order that the event density against time could be calculated. To evaluate the technique, the tear test was investigated further. Leather samples were prepared using identical tanning conditions. The only variation between the samples was the offer of fatliquor. Observation of the waveforms indicated that there were differences in the number of events occurring. This was, therefore, quantified using the event counter software. Figure 9 illustrates the results obtained. It is clear that the level of fatliquor applied significantly influences the acoustic emissions observed. At offers in excess of 5%, the pattern observed is as expected. Increasing the offer of fatliquor results in increased lubrication of the leather fibres. As the lubrication is increased, there is a decrease in the number of acoustic events recorded. This indicates that the pulses are due to the bonds between the fibres being broken, rather than recording the breakage of the fibres themselves. At between 0-5% offer, there is an increase in acoustic events as the offer of fatliquor is increased which is unexpected. At 0% offer of fatliquor, there is no lubrication of the fibres and, therefore, failure is most likely due to fibre breakage. At low fatliquor offers, there will be increased sound from the fibres being able to move past each other as they are lubricated slightly (consider the analogy of a bow on a violin string). Acoustic emission to evaluate handle The results illustrate that the data obtained from the acoustic emission of leather tear tests may be related to the subjective properties of leather. Within this research, the subjective evaluation of leather handle was based on characteristics such as grain firmness (looseness), stiffness (softness) and fullness (emptiness). Twenty-five commercial leather samples were assessed for handle by a panel of eleven experts and the subjective parameters listed previously were assessed on a five-point scale. The results from these were then correlated with the data obtained from analysing the acoustic emission data obtained from tear tests carried out on these samples. In order to carry out the statistical analysis, it was necessary to determine whether the data obtained fitted a normal distribution. Using the Saphiro-Wilks test, the variables were tested for normality and transformations through LOG were made to reach normality where required. The following variables were considered in this analysis: * Log (counts per second across tear propagation) * The average RMS power across tear propagation Maximum amplitude across tear propagation The RMS power is a parameter that is calculated through the software package and is a measure of the perceived 'loudness' of the noise or effective energy of the signal. It was found that several significant correlations occur between the acoustic data and the subjective parameters. Log (counts per second across tear propagation) was strongly related with stiffness, fullness and emptiness. RMS power was strongly correlated to softness and emptiness. While the acoustic emission data shows good correlation with the subjective parameters of leather, other work has resulted in the development of a mathematical model to allow these parameters to be quantified. It is, therefore, of interest to compare the results from the two techniques and combine them to determine if improvements can be made in the prediction of handle. Table 1 illustrates the correlation coefficients obtained from the acoustic emission data, the mathematical model, and a combination of these two methods. From Table 1, it can be seen that there is a significant improvement in the correlation of subjective assessment with objective measurment when the acoustic data is included in the mathematical model. The acoustic emission data alone can be used as an indication of properties such as the stiffness/ softness and fullness of the leather. Conclusion Through the work detailed in this paper, two applications for acoustic emission techniques have been developed for leather testing. It has been possible to demonstrate that there is a relationship with the aesthetic properties of leather. The previously subjective parameters of grain firmness/looseness, stiffness/ softness and fullness show a relationship with quantifiable parameters through the use of acoustic emission techniques. A working system has been developed that has been used to evaluate data from tear tests. The system is not limited to this alone and allows acoustic data to be collected from a number of physical tests. Evaluation of the lastometer test using acoustic emission has provided some extremely promising results. Using the technique, it is possible to detect grain crack during testing without resorting to visual observation. This is a more accurate test, which is easier to carry out than visual inspection. It is possible to fully automate the analysis on a materials tester, removing the need to use hand driven apparatus (with inherent inaccuracies). The problem of background noise was investigated and it was concluded that an acoustic chamber would remove some of the higher frequency noises. However, it was not able to remove low frequency noise. The use of a multi transducer array resulted in an increase in the signal to noise ratio observed and this combined with carrying out testing in a quiet area is perfectly adequate for the collection of acoustic data. Due to the success of this research, a patent has been applied for (application number GB0016063.0). It is planned that the apparatus will be developed by Stable Micro Systems Ltd into a commercial product. Acknowledgements The authors wish to acknowledge the financial support of the European Commission through the Standards Measurement and Testing (SMT) programme (Contract number SMT4-CT98-5508). Also recognised is the valuable input from the partners in this project: Pittards plc, Industrias Peleteras SA, Teneria Moderna Franco Espanola SAL, and Maximo Mor SA. A version of this paper was presented at the XXVI IULTCS Congress held in Cape Town, South Africa, March 7-10, 2001. It was awarded the second prize in the scientific category during the Congress.