We have experimentally studied the potential advantages and payoff of wideband waveform-based acoustic emission by monitoring fatigue cracks in metal plates. The work was done in two main stages.
Stage I
The first stage involved an extensive examination of extraneous acoustic emission generated during fatigue testing of metal specimens. This broadly based study included variations in specimen geometry as well as characterization of the differences in extraneous acoustic emission signal waveforms as recorded from multiple resonant sensors and wideband sensors. Two plate thicknesses (typically 6 mm and 25 mm) were used along with two different sizes of the lateral dimensions of the specimens. The larger specimens had sufficiently large lateral dimensions such that signal reflections from the edges of the specimens did not superimpose on the direct arrivals of the signals at the sensors (as can commonly occur for smaller specimens). As a bit of background information, some types of AE equipment define the arrival time of an AE signal as the time when a fixed threshold amplitude value is first exceeded. Two key results were noted from these first stage experiments.- The first key result of our experiments showed that arrival times as defined by the time when a fixed threshold is first exceeded are not always a reliable indicator of where a signal originated. Thus, extraneous AE signals (such as those arising from the test apparatus grips) can be confused with the more important crack-based signals and vice-versa if one depends upon the use of fixed thresholds to determine signal arrival times.
This first result is of considerable significance when one considers the fact that often more than 95% of all the acoustic emission signals recorded in a fatigue experiment are of extraneous origin, and thus are not of the same significance as crack-generated signals.
- The second key result from the first stage of study concerned the contrast between the waveforms of extraneous acoustic emission signals as recorded from wideband sensors compared to those recorded from resonant sensors. The signals from wideband sensors had distinctive characteristics for different acoustic emission events while the signals from resonant sensors all were remarkably similar to each other for the same set of extraneous acoustic emission events. In other words, the wideband sensors were sensitive to differences in the AE source characteristics as evidenced by the different modes excited in the plate specimens. Thus, various extraneous acoustic emission source types could be distinguished from each other when viewing the waveforms as recorded with wideband sensors; however, the same extraneous AE signals as recorded with resonant sensors resulted in waveforms which all looked about the same
Stage II
The second stage of the study examined acoustic emission generated by fatigue cracking in a unique large specimen. This specimen was a metal plate with large lateral dimensions such that signal reflections from the specimen edges did not superimpose on the direct signal arrivals at the sensors. The specimen was also of sufficient lateral dimensions such that the sensors could be located in the far field (i.e., at least a distance of 10 to 15 times the plate thickness away from the acoustic emission sources). In a typical (i.e., smaller) laboratory specimen, the AE signals are reflected at the specimen edges at a short distance from the AE sensors. Thus, the reflected signals (and their consequent mode conversions) often are superimposed on the direct (i.e., unreflected) signal when it arrives at the sensor. Such a superposition of signals can create an artificial increase in the signal amplitude as detected at the sensor for small specimens (as compared to the case of a typical "real" specimen used in field testing in which the specimen dimensions are usually large enough to prevent any reflected signals from superimposing on the direct signals at the sensor). As expected, various characteristics of the crack-based signals were observed for the signals recorded by wideband sensors whereas such differences were not apparent in the signals recorded from resonant sensors. These larger samples also provided an opportunity to study variations in the signal radiation pattern from a single acoustic emission source location (i.e., to study the differences in the AE signals as detected at various positions in the plane of the plate in different directions away from the AE source).In addition, we studied the significant signal amplitude reduction due to wave propagation for the crack-generated signals; this effect can be observed by placing sensors at the crack tips where the signals are generated and a few inches away. This amplitude reduction results most significantly from the geometric spreading of the signal, and to a lesser extent from the frequency-based dispersion of the signals as they propagate. Further, this fall-off in signal amplitude has important implications when one tries to apply laboratory results from small specimens to field cases with large specimen dimensions. With a large specimen (such as often the case in the field application) the AE signals may not be detectable above the inherent electronic noise, since there are few (if any) edge-reflected signals superimposing onto the original signal when it is detected at the sensor. This situation could invalidate the relevance of the small specimen laboratory study for intended field applications.
Additionally, the high sensitivity over a broad frequency bandwidth of wideband sensors results in signal distinctions, which directly relate to source characterization/identification. For example, the effects of the depths of the AE sources within the plate thickness have clearly been observed in the anti-symmetrical aspects of the experimentally recorded wideband signals. Also, the presence of friction/fretting of crack surfaces has been observed in the form of repeated acoustic emission waveforms which were identical (except for small amplitude changes) on successive load cycles. Two papers have been published on this subject and a third is in progress. Also we expect in the future to "marry" the results of our finite element modeling to the fatigue crack results from these large plate studies.
The titles, abstracts, references and where possible additional information from our papers in this subject area are listed below.
| 1."Detectability of Slow Crack Growth in Bridge Steels by Acoustic Emission," M. A. Hamstad and J. D. McColskey, Materials Evaluation, Vol. 57, No. 11, 1999, pp.1165-1174. |
The detectability of slow growth of cracks in bridge steels has been studied by use of acoustic emission testing technology. Fatigue crack growth rates of nominally 1 x 10-4 mm/cycle (4 x 10-6 in./cycle) and 1 x 10-3 mm/cycle (4 x 10-5 in./cycle) were monitored with an eight channel, 12-bit, waveform-based, acoustic emission measurement system. Eight acoustic emission sensors were arranged on plates nominally 25 mm (1 in.) thick (A588, A572, A36, A7, A514 steels and 2024 T351 aluminum alloy). Four of the sensors were typical resonant, commercial acoustic emission sensors, and four sensors were wideband, high sensitivity sensors developed at NIST. The resonant sensors were band passed from 100 to 300 kHz and the wideband sensors from 50 kHz to 1.5 MHz. Each recorded acoustic emission event produced eight simultaneous waveforms. A representative series of event waveforms from each test was analyzed. Crack events were sorted from extraneous acoustic emission events visually, using the relative arrival sequence of sets of whole waveforms. The extraneous events normally originated in the specimen grip region. Typically, at the lower crack growth rate, several thousand cycles were required to obtain one valid crack acoustic emission signal. The results from A514 steel and 2024 aluminum were exceptions. The A514 steel had 20 crack acoustic emission events over eight consecutive cycles and none over the other 12 461 cycles examined. The 2024 aluminum had a crack event approximately every 12 cycles. At the higher crack growth rate, approximately one or two cycles were required for a valid event from A514 steel and 2024 aluminum. The other steels required 18 to 130 cycles for a valid event. The behavior of crack generated acoustic emission will be discussed in terms of rise times of acoustic emission sources as related to material microstructure. In addition, indications of classes of crack events as observed with the wideband sensors will be discussed. Supporting experiments and results from plate specimens with large lateral dimensions (533 x 457 mm [21 x 18 in.]) are presented. These studies on steel 6.4 mm (0.25 in.) thick demonstrate the significance of attenuation of the signals from a sensor located at the crack tip to a sensor approximately 102 mm (4 in.) away. These supporting experiments also demonstrate the changes in signal amplitudes due to acoustic emission source radiation pattern effects as well as signal reflections.
- Many fatigue cycles were required to generate one detectable acoustic emission crack event from a crack in the tested bridge steel parent materials at low crack growth rates for the current R ratio.
- An increase of about an order of magnitude in the growth rate of cracks substantially reduces the number of cycles needed to detect a crack event.
- Since bridge steels are generally ductile (except for possible scattered brittle inclusion particles) with high values of fracture toughness, the acoustic emission source rise times during crack extension are not expected to be short in most alloys. Hence, the amplitude of the generated acoustic emission will be small.
- Cracks in bridge steel parent materials were more readily detected by a sensor positioned at the crack tip. The attenuation due to geometrical spreading and wave dispersion is minimized for a sensor located at this position.
- When geometric features which cause early reflections of acoustic emission signals are not present, detection of crack growth acoustic emission depends on proper placement of acoustic emission sensors as dictated by the radiation pattern of the acoustic emission source.
| 2. "Wideband and Narrowband Acoustic Emission Waveforms From Extraneous Sources During Fatigue of Steel Samples, "M. A. Hamstad and J. D. McColskey, Journal of Acoustic Emission, Vol. 15 No. 1-4, 1997, pp. 1-18. |
An extensive study of acoustic emission (AE) generated from extraneous sources during fatigue loading of steel samples has been performed. AE waveforms were emphasized instead of traditional AE parameters. Typically, four resonant sensors and four wideband high-sensitivity sensors were used to gather eight waveforms for each AE event. The fatigue tests were conducted in a servo-hydraulic fatigue test machine, using hydraulic grips. Specimen variables included two different thicknesses and different lateral dimensions of the plate-like samples. In addition to grip-based extraneous AE sources, artificial extraneous sources were also examined. Acoustic emission signals from both solid- and air transmitted artificial sources were studied. The extraneous signals were contrasted to an AE event from a fatigue crack. A motivation for this research was the fact that applications of AE to monitor steel bridges encounter extensive extraneous AE.
The waveform database established in this work indicates that recognition of extraneous AE is not always easy. This is due to the variety of source types, the effects of thickness changes, and dispersion effects during wave propagation. Some extraneous signal types could be easily distinguished from crack signals in the wideband AE data. Other extraneous signals are not as easily distinguished, even with wideband sensors. Using either resonant or wideband sensors, waveform-based AE analysis offers advantages to fixed-threshold approach for spatial discrimination of extraneous AE.
- Extraneous waveforms from wideband sensors show significant differences in both shape and frequency content for different types of sources, whereas relatively small differences in shape and frequency content are seen in resonant-sensor waveforms from the same sources.
- Techniques, which rely on using rise-time or frequency-spectrum discrimination to distinguish real crack-induced AE sources from extraneous sources, may be ineffective for resonant sensors.
- A crack-induced AE source and an extraneous source can sometimes result in large shape and frequency-content differences in the wideband-sensor waveforms from the two sources, whereas in the narrowband resonant-sensor waveforms, these differences are typically relatively small.
- Great care must be used in specimen design and spacing of sensors for spatial or guard-sensor discrimination of grip-based AE to be effective, when using threshold-based arrival times. Visual arrival-time determination or equivalent signal processing of recorded waveforms from resonant sensors can be used much more effectively for spatial discrimination of extraneous grip sources.
- Approaches to distinguish extraneous AE from crack-induced AE with wideband sensors will need to account for sample-thickness variations and their effects on wave propagation.
- Extraneous AE occurs in the peak-load region during fatigue cycling. Hence, load-gating is useful but not 100% effective in eliminating extraneous sources.
These web pages are maintained by Dr. Hamstad. They are protected by Copyright © 2000. Last update: December, 2003.
For information on web site development, contact ACC at dpackard@acentauri.org