·Table of Contents
·Methods and Instrumentation
Threshold Independent AE TestingHartmut Vallen, Jens Forker,
Vallen-Systeme GmbH, Icking (Munich) Germany,
Tel. +49-8178-9674-400, Fax -444, Email: firstname.lastname@example.org
Jürgen von Stebut
Ecole des Mines, Lab. de Science et Genie des Surfaces, Nancy, France,
Tel. +33-383-584245, Email: email@example.com
The acquisition threshold is the most important setting for the measurement of short acoustic emission (AE) burst signals. Only signals that have crossed a pre-defined threshold level are recorded. Even more tricky: some of the AE-feature results depend on the threshold setting. Profound expertise is required to set the threshold correctly. For many potential industrial applications the availability of such expertise can't be expected.
The paper describes a new approach to measure AE bursts by means of a new signal conditioner. The signal conditioner provides the amplitude of even the shortest AE bursts and information about the (continuous) background signal as well - without the need for any parameter setting. It converts the short high frequency AE-signal into a DC signal that can be measured by a simple measurement equipment. The signal conditioner allows for recording the amplitude of AE bursts from noise level up to pencil break, in a single measurement range, accurate to 1 dBAE. This signal conditioner turns a simple analogue data acquisition system that might be already in use to measure process parameters like temperature, force, pressure etc,., into an efficient AE monitoring system.
The AE module is rugged and can be easily integrated into industrial environments where it can reliably detect burst signal sources, such as crack formation and growth, delamination, fiber breakage, partial discharge, particle impact, and more, and it can also be used to detect continuous signals from, for instance, leakages, flow of liquid or powder, continuous friction and wear, drilling, bearing, cutting or other like machining processes.
The signal conditioning principle is described and, ans an example, the use for crack detection during the scratch tests is described.
An AE application works with either burst or continuous AE signals.
AE burst signals are distinct signals where the beginning and end can be recognised. Common examples include crack growth, particle impact, partial discharge, or other source mechanisms of short duration.
Continuous AE are endless although the amplitude and frequency content can change over time. Examples are flow noise from leaks and rubbing noise among others. A rapid sequence of overlapping burst signals can also appear as a continuous signal, when the individual signals can no longer be separated.
The measurement type is chosen according to whether burst or continuous signal types are being investigated.
For continuous signals, one measures usually the averaged amplitude. A signal conditioner amplifies, rectifies, and averages the high frequency AE-signal over an integration time and by a specific method. The RMS-value is often used which is the square root of the averaged signal square.
From burst signals one usually extracts a variety of features. These features include arrival time, peak amplitude, rise time, signal duration, counts and energy. The acquisition threshold is the most important criterion for discriminating AE burst signals.
Signals whose peak amplitude is not large enough to cross the threshold are not detected. Setting the threshold too high will prevent potentially important signals from being recorded. Setting the threshold too low will cause the background noise to cross the threshold and will result in a great deal of unwanted data to be recorded.
If the background level rises during the test, the signals may remain above the threshold for long times, so that the AE system "sees" only a few especially long hits. In this case, despite a large number of burst signals occurring, only a small amount of data will be produced. Possibly a lot of important information is lost.
Next to the severe influence of threshold on general detection of the signals is the dependence between the threshold setting and the values obtained for many measured features. An example is the value measured for the duration of the AE signals. The duration is defined as the time period between the first and last threshold crossing of a signal. A signal which decays slowly will result in a much longer duration for low thresholds than for higher thresholds.
The optimum threshold setting depends specifically on the amplitude of the AE signals of interest and on the actual noise signals. Further, the signal amplitude depends on many factors, such as sensor sensitivity, the sensor coupling quality, the source energy, the signal attenuation related to distance between source and sensor, changes in wave propagation conditions, and so forth. With presently available AE systems, the optimum threshold setting is of primary importance and fundamental for the success of every application that makes use of burst emission analysis. One needs experience to recognise if acquired data is plausible and then conclude a correct threshold setting.
There are numerous possible applications, where threshold independent detection of short AE burst-signals in a wide range of amplitudes would be helpful. These measurements should be both foolproof and require no special AE knowledge.
As an example the scratch tester application is presented.
|Fig 1: Scratch testing principle and set-up for research of AE-behaviour|
Scratch testers are used to determine failure limits of surface coatings. Consider the following: a test fixture set-up draws the tip of a diamond indenter with continuously increasing force and constant speed across a test surface. The steadily increasing contact load causes tensile stress behind the indenter tip (trailing edge) and compressive stress ahead of the cutting tip (leading edge).
At some point during scratching, a crack is initiated. The position of the first crack within the scratch path is determined visually using a microscope. From the crack position within the scratch track one can determine the corresponding critical contact load
With modern coatings, the first cracks are so small that they are difficult to perceive even under the microscope. They may even close-up within a few milliseconds and thus become optically "undetectable". This is a task just waiting for AE as a "crack damage warning - validation" indicator. However, this requires an AE system that is easy to use and cost effective.
In the course of an EC funded project (SMT4- CT97-2150) a better scratch test procedure is currently under development. Here, an AE module and a coupled video system are added to a conventional scratch tester. The AE module is used to determine the cracking and trigger the video system. These video images (potentially taken on-line, in the stress applied state) will then be the material proof of AE signals caused by cracking.
For this application, it was necessary to study the relation between AE and different variables - such as different diamond tips, coatings and substrates. These measurements were made with the multipurpose Vallen AMSY4 instrument in a set-up as shown in Figure 1. Results were published by von Stebut et al. 
The goal was to find means to differentiate between crack emissions and noise, such as friction, mechanical vibrations etc. For this purpose, a great deal of burst signals were studied, including investigation of waveforms and their frequency content.
For this application, the sensor has to be attached to the shaft of the scratch tester for practical reasons. In this position, the geometry of the shaft influences the wave propagation very strongly. It was not possible to find a waveform or frequency analysis criterion suited to differentiate crack signals from friction and mechanical vibration artefacts that could be obtained with a realistic and cost effective instrument.
The best criterion found for separating cracking and unwanted noise is the peak amplitude. It is easy to understand that the absolute peak amplitude is strongly dependent on the brittleness of the test sample, the shape and size of the diamond tip as well as the quality of the AE sensor coupling on the shaft, along with other factors.
Several requirements were established from the preliminary study and knowledge of how scratch testers are used. These requirements and the corresponding solutions are described hereafter.
4.1 First Requirement
The AE module should be as foolproof as possible. This means that as many conditions which cause errors or mistakes, such as bad settings, should be eliminated. One can not expect that a user will perform multiple tests on each new coating only for the reason of finding the optimum threshold. Despite these constraints, determining the instant of crack formation remains the goal which must be reached in practice. This must even be true when the amplitude dependent AE conditions change, as can be expected with every new test material. It must also hold true when the diamond tip sharpness and the quality of the coupling changes.
The AE signal is converted into a DC-signal which is proportional to the logarithm of the peak amplitude detected. The scaling factor is 40mV/dBAE. A noise signal at the input of 20dBAE, or 10µV, produces an ASCO-P output of 800mV (40mV/dB * 20dBAE).
A pencil lead break which generates a pulse of 100dBAE produces 4V (40mV/dB * 100dBAE).
After each increase in amplitude, the peak signal remains for 50 milliseconds at this last peak. This time period is called the peak stretching time (PST), because the peak output voltage is maintained constant for this period of time. This allows a slower, more cost efficient data acquisition system to be used, such as one for measuring the force. If the amplitude increases still further during the 50ms PST, the output increases immediately to this new value and the PST begins again. After the expiration of the PST, the signal quickly falls to the value of the input amplitude, thereby allowing further hits with lower amplitudes to be easily recognised.
Figure 2 shows the ASCO-output signal (upper curve) caused by a simulated AE-Signal (lower curve). To make this process clearer, the peak-stretching time was lowered from 50ms to approximately 0.080ms in this example.
|Fig 2: ASCO-Output reply to a short transient AE-signal|
The arrows 1 and 2 on the left edge show the zero point of both signals. The ASCO output. begins at the far left at about 0.8 V, before the AE-Signal begins. This corresponds to the peak value of the electrical noise which is around 20 dBAE (40m V/dBAE * 20 dBAE = 800mV).
Tha ASCO-output follows the rising AE-Signal without delay. The peak amplitude of the signal reaches approximately 99 dBAE (90m V) which causes a corresponding increase in the output signal to 3960 mV (99 dBAE * 40 mV/dBAE = 3960 mV) and remains at this value for the duration of the PST (here reduced to 0.08 ms, default is 50ms). When the PST is over, the output signal falls immediately to the current value of the input signal. This allows short duration signals to be separated even when they are following quickly after each other.
To further exploit these signals, a U-t-recorder or a simple data acquisition system can be used. The sampling rate should be shorter than the PST; with a 50ms PST a 25-50 Hz sampling rate is recommended. This combination will guarantee that short bursts can be measured with an arrival time resolution of 20-40ms. If one acquires an additional test parameter, such as load or displacement, then one can read the activation point of the first crack in a simple load-time diagram. An example is shown in Figure 2. The steadily increasing line curve shows the contact load of the diamond normal to the surface (right axis). Every bar represents the peak amplitude of at least one AE-signal. The dB scale (left axis) is obtained by the factor 40mV/dB.
The high amplitude signal at t = 17 seconds was correlated unambiguously with a crack under the microscope. All activity before this peak was essentially low amplitude friction or other noise. The 60dB signal at 10s could not be correlated to visible cracks either. The first crack is found to occur at 27 Newtons from the correlation with the load curve shown here.
Materials behave differently. Most, however, follow a similar pattern: a clear increase of amplitude at a specific load. The actual value of the peak amplitude is not important, as long as crack signals rise above the amplitude range of the friction noise.
4.2 Second Requirement
When crack growth occurs, a digital switch shall be activated. In the case of the scratch tester this switch triggers a video system.
To satisfy this requirement, an absolute threshold setting is used. This is defined as the level which is expected for peak amplitudes of cracking. Because the measurements are completely independent of the threshold, the first test with a crack is already sufficient to show the optimum threshold for a reliable separation of useful signals from the noise. In the example shown in Figure 3, one would set the threshold at about 75-80 dB.
|Fig 3: ASCO-Output and Load vs. Time|
The ASCO-P does not have a direct threshold setting, instead an alarm threshold is specified by an analogue input signal. A PC-plug-in-board used for data acquisition can normally also supply this analogue setting.
The ASCO-P compares the logarithmic AE-Signal with the supplied alarm threshold signal. When the AE signal exceeds the alarm threshold, a 50ms long impulse is produced. With the scratch tester application, this alarm signal is used to trigger the video system to record crack initiation.
Triggering the (106 pixel) video system by the AE module reduces the amount of video data considerably. Images are stored only when there is a high probability of crack activity. The recorded video verifies the indication of the AE-module. New knowledge of not yet understood artefacts then helps to improve the understanding of the AE-module indications. In other applications, this output can have other functions, for example to turn off a test or as an impulse for an event counter.
Presently available commercial instrumentation often requires a threshold setting that determines whether data is acquired and the values of the recorded features. On the contrary, the ASCO-P measurements are not influenced by a threshold. Merely the digital switch function is dependent on a threshold.
Figure 4 shows the ASCO-P block diagram with essential functional groups of the ASCO-P. The filter module can be chosen appropriate for specific applications.
|Fig 4: ASCO-P Block Diagram|
The frequency range which is best for crack detection is 100-300 kHz. Even higher frequency ranges are useful to get farther away from noise. For detection of leakage the frequency range of 20-100 kHz is recommended.
The ASCO-P has the additional outputs OutputFilter and OutputASL which are useful for maximising potential applicability. The OutputFilter signal is the amplified and filtered high frequency AE Signal. This can be used with a transient recorder, for example.
The logarithmic AE-Signal is fed through a low pass filter and presented as Output ASL signal. On this output the influence of short burst signals is insignificant and it shows the essential features of the continuous background noise. It is especially useful in monitoring continuous AE-signals
Figure 5 shows the actual device, cover removed. The module fastened by 2 screws is the filter module that can be easily replaced.
|Fig 5: ASCO-P with removed cover|
The ASCO-P is an analogue front-end, easy to use, and very cost effective. It is perfectly suited for industrial applications as well as for research activities. But where more detailed information about the AE signals is required, e.g. multi-parameter analysis, sophisticated statistics, frequency analysis, location calculation, clustering, etc., a high-end system like the AMSY4 is needed. The possibilities of the ASCO-P can be easily extended by changing the PST when industrial applications require this. The analogue output can then be combined with other instruments and the ACSO-P then becomes an AE front end providing filtered AE signal output that can be acquired and analysed with an external transient recorder, for example.
The AMSY4 AE system has been extended to include a threshold independent mode that corresponds to the function of the ASCO-P. Thereby new applications of the ASCO-P can be developed and documented with both the convenience and greater possibilities of the AMSY4. When the initial work is completed, the ASCO-P is then ready to be put right into service.
The ASCO-P is also the basis for a new battery-powered hand held instrument LSM1: The Leak Signal Monitor. The peak amplitude of the AE Signals can be read in dBAE on a digital display. With this device, the AE operator can test a structure for background noise levels before testing. Either the background level (ASL) or peak amplitude level can be chosen for display. Headphones can be used with an adjustable threshold so that even the smallest variations can be tracked.
Differences in signal intensity can be used to find the direction from where noise signals arrive.
The LMS1 helps determining the origin of specific noise sources before a time-consuming multi-channel test set-up. The field test engineer, and even the field test client, can use the LSM1 to find out if a structure is sufficiently "quiet" for AE testing. The device pays for itself, because it allows for scheduling only quiet tanks for AE field test and avoids costs that unnecessarily arise, when a AE test crew comes and cannot perform the test because the structure is too noisy.
The ASCO-P allows for the measurement of short peak amplitude signals with large dynamic range in a simple, easy to use instrument. It makes the realisation of new, simpler and cost effective applications for the AE method possible.
The ASCO-P is an ideal accessory for every mechanical loading device, for example any universal testing machine. This is especially true when the machine is already equipped with an existing data acquisition system for measuring load and/or displacement. One can then add the ASCO-P output to an unused channel of the data acquisition recorder. One then possesses a simple, manageable, yet very powerful AE System that shows the first onset of damage as it occurs.
This work has been accomplished in the framework of the European programme "Standards, Measurements and Testing" as part of an on-going EC funded project (SMT4- CT97-2150).
|© AIPnD , created by NDT.net|||Home| |Top||