History of Testing

Key Principles

• Failure in the composite structure results from the accumulation of microscopic damage in a local region of the composite.

• There is a high payoff for the nondestructive testing application if accurate acoustic emission source location can be determined in the composite vessel.

• This task is complicated by the following: significant geometric signal attenuation and significant signal dispersion (both of which occur in all materials); material-based signal attenuation (which is especially pronounced in composites); and variations in velocities in different directions of signal propagation.

• Many acoustic emission systems determine signal arrival times using fixed threshold techniques; because of the aforementioned complications, such AE systems measure arrival times for signals using various portions of the acoustic emission signal which travel at different velocities. Many "triangulation" type computations that are used to determine the AE source location based on its arrival time make the erroneous assumption that all AE signals travel at a single velocity; hence such computations often result in incorrect locations of AE sources.

An alternative location technique uses the concept of "the first sensor hit by an AE event" (also called a "first hits" concept) to identify a more generalized region around each sensor (vs. a specific location on the test specimen) from which the acoustic emission signal likely originated. Thus, depending on certain conditions, one can determine which one of the several sensor regions on the test specimen has more concentrated AE activity. We considered the case of AE monitoring of aerospace-type composite pressure vessels that have relatively uniform stress levels. For this case, the technique of using first hits to identify localized regions of significant concentrated damage works best when there is a relatively high density of acoustic emission sensors (i.e., a close spacing between adjacent sensors) on the composite vessel. In such cases, we have shown that the best correlations between residual strength and proof-test acoustic emission occur when only the acoustic emission generated from the strength-limiting part of the pressure vessel is used in the correlation.

Our results with impact-damaged graphite/epoxy vessels have also shown that using the localized acoustic emission signals from the unloading portion of a proof cycle is superior to using those from other portions of a proof cycle. We reached this conclusion since this technique results in a more exact correlation with the actual residual strength. To evaluate the acoustic emission recorded during unloading, we have defined a so-called "Shelby ratio" to provide a correlation that could be used in a nondestructive testing application. We found this ratio provides better correlations than the Felicity ratio for the impact-damaged vessels we studied. Additionally, we found that the acoustic emission generated during a hold portion of a proof cycle provides the poorest correlation with the residual strength of impact damaged graphite/epoxy pressure vessels. This latter result is probably related to the reduced sensitivity to stress rupture of graphite/epoxy as contrasted to glass/epoxy composites in which AE data from holds at pressure were more helpful.

The titles, abstracts, references and in some cases additional information from our relevant papers in this subject area are listed below. Multiple other appropriate references can by found by searching through the complete bibliography.

Six aerospace-type, filament-wound graphite/epoxy pressure vessels were studied. Four of these cylindrical vessels had each received a single, controlled impact; four had received thermal exposures including moderate heat and/or cryogenic cold. Acoustic emission (AE) was monitored during a proof test sequence (after the impact/thermal exposure), including the final depressurization from proof pressure (i.e., the second unload cycle). A single set of wideband AE sensors was used to simultaneously record parameter-based AE and waveform-based AE on two independent AE instrumentation systems. Some slight to moderately large differences in AE activity for impacted vs. non-impacted vessels were noted for AE from Felicity ramps and holds at pressure. However, dramatic differences were apparent in quantities of AE from the final depressurization cycle; impacted vessels experienced unload AE activity which was a minimum of an order of magnitude greater in quantity than the unload AE activity for non-impacted vessels. This difference provided a distinctive means of identifying those vessels with impact damage. A newly defined "Shelby ratio" was introduced as a means of quantitatively assessing the unload AE; correlations were obtained between vessels' residual strength and the unload "Shelby ratios." Refinements of the parameter-based data analysis approach were made possible by waveform information. Additionally, examination of the waveform data of the unload cycle revealed a significant quantity of "friction" type waveforms having a repeated lower frequency character; similar "friction" waveforms were able to be artificially generated by rubbing a pencil lead across a composite vessel surface.

1. Unload AE data displayed a minimum order of magnitude distinction in the quantity of AE data generated for vessels having no impact damage vs. those with impact damage. This fact was true for both the total quantity of AE events and some waveform-based subsets of the total data even though the vessels had various levels of thermal exposure (i.e., dwells at moderate heat and/or cryogenic cold). In contrast, AE data from Felicity ramps or holds only displayed about a minimum two-to-fourfold distinction between impacted and nonimpacted vessels.
2. The newly defined "Shelby ratios" provide a methodology for assessing unload AE data in a fashion somewhat analogous to the Felicity ratio which has often been used for assessing AE data from increasing loading ramps. Approximately linear correlations were obtained for various individual and average "Shelby ratios" vs. residual vessel strength, with countup data (vs. countdown) resulting in the best correlations.
3. A standardized methodology for computing an average "Shelby ratio" was presented. This averaging approach eliminates subjectiveness in its determination of an "average Shelby ratio" value, and it is "robust" because it accommodates data sets containing widely differing amounts of AE activity. The same averaging approach could be applied to create an "average Felicity ratio" that is "robust" and non-subjective.
4. A significant fraction of the unload AE data displays a characteristic "friction" waveform having a low frequency (e.g., 50-130 kHz) and low amplitude nature.
5. If it is desired to take the greatest advantage of the "friction" unload AE data and to increase sensor spacing for an unload cycle, then it will necessitate the use of wideband sensors with high sensitivity in the low frequency ranges (vs. sensors without such sensitivity) and/or sensors with a higher signal-to-noise ratio than were used in this research.

This study examined the effects of exterior thermal insulation foam and previous thermal exposure upon recorded acoustic emission (AE) signals from the pressurization of graphite/epoxy vessels. The specific conditions imparted (in various combinations) to ten vessels before proof testing were as follows:

• controlled impact damage
• exposure to moderate heat and/or cryogenic cold
• and, low-density thermal insulation foam bonded to the exterior surface.

1. Adhesively bonding a thermal foam insulation layer to a composite vessel exterior surface generated an abundance of additional AE during the 1st loading after bonding vs. vessels without foam.
2. Foam insulation caused significant attenuation of AE signals. This resulted in a significant reduction in the quantity of recorded AE vs. generated AE data for vessels with a bonded foam layer.
3. The two competing effects of additional AE generation and AE signal attenuation caused by a bonded foam layer can have various net effects on the cumulative sum of AE data recorded from foam-covered vessels depending upon whether the vessel had impact damage and whether data from the 1st or 2nd ramp were considered. The competing effects are both deleterious with regard to the ease of distinguishing between vessels with and without structurally significant impact damage.
4. Thermal cycling (cryogenic cold and moderate heat) caused a significant increase in the quantity and earliness of generated and recorded AE from graphite/epoxy pressure vessels during the 1st pressurization after thermal cycling.
5. The additional AE caused by thermal cycling may be beneficial when using AE data to assess the structural integrity of real composite structures, especially when the maximum loading (with AE monitoring) after thermal cycling may be significantly smaller than the original (maximum) proof load.
6. The additional AE generated (due to previous thermal cycling or the presence of a bonded foam/adhesive layer) and recorded during the next loading was similar in character to the AE generated during the formation of a composite's "characteristic damage state" in two main aspects: a) the recorded AE was from uniformly distributed sources; b) the quantity of AE due to the previous thermal exposure or foam/adhesive was significant during the 1st loading, and was substantially reduced (perhaps depleted) during the 2nd loading. For example, the quantity of AE events recorded (up to 57.2 MPa) in the 2nd ramp of a non-impacted vessel with thermal cycling was about 6% of the AE quantity recorded (up to 57.2 MPa) in the 1st ramp; similarly for a non-impacted foamed vessel, the events recorded in the 2nd ramp were about 4% of the total for the 1st ramp.

Numerous acoustic emission waveforms were recorded using wideband, non-resonant sensors during the initial proof-pressurization ramp of a graphite/epoxy pressure vessel following a single controlled impact. The waveforms exhibited a variety of characters and frequency spectra. Some of the observed differences are presented, and potential source distinctions are discussed.

1. Recorded waveforms for various AE events from proof testing of a graphite/epoxy pressure vessel exhibited a wide variety of waveform characters and frequency spectra.
2. In contrast, waveforms for various AE events (recorded at a fixed distance from the AE source) from four-point bend testing of an isotropic chopped fiberglass/epoxy composite exhibited remarkably uniform waveform character and frequency spectra for all waveforms recorded.
3. Signal-propagation distance (from the AE source to the sensor) significantly affects waveform character and frequency spectra for any single event recorded from the graphite/epoxy pressure vessel
   1. The most dramatic changes occur within the immediate vicinity of the AE source.
   2. At larger distances from the AE source, differences are typically much less dramatic, especially for the frequency spectra. More specifically, those events that have significantly different characters when observed very near their AE sources may appear to have quite similar characters when examined farther away from their AE sources.
4. Some potential reasons for the differences in graphite/epoxy vs. glass/epoxy, respectively, are as follows:
   1. Layered vs. uniform construction.
   2. Biaxial vs. unidirectional normal stresses.
   3. Anisotropic vs. isotropic material.

A sixteen channel acoustic emission (AE) system was used to evaluate thirty-two graphite/epoxy pressure vessels (260 mm outside diameter) during a proof-pressure test to approximately 70% of the nominal vessel burst strength. Later the vessels were pressurized to failure. AE data (85-325 kHz) were gathered during the tests by non-resonant, small aperture sensors. AE data were examined (for a subset of eleven vessels which experienced consistent testing conditions) in terms of the cumulative AE activity occurring on each individual channel during a secondary pressure ramp and a second hold at fixed pressure. For each vessel, the probable vessel failure location was compared to the geographic region of the "most active" sensor for three points of interest in the proof test: early in the second ramp (i.e., near pressures at which Felicity ratios were determined), near the end of the second ramp (i.e., at the previous maximum pressure), and for the second hold at pressure.

Results showed that for lower strength vessels there was generally a good match between probable failure location and the specific geographic sensor regions which had the greatest cumulative "first hit" AE activity at Felicity-ratio-type pressures. No similar match was found for higher strength vessels. No consistent correlation was found between the probable burst location and most active regions of AE, activity occurring at higher pressures in the second ramp or during the second hold, regardless of burst strength. It was also discovered that efforts to assess the significance of the AE data are substantially influenced by the effects of geometric attenuation.

1. Because of geometric attenuation due to spreading, most AE hits recorded from a well-designed graphite/epoxy pressure vessel originate within the immediate, localized vicinity of the sensor recording the hit. This localized vicinity is substantially smaller than the "first hit" geographic region surrounding the sensor.
2. The "first hit" geographic sensor regions used in the analysis presented here are substantially larger in physical size than any localized, failure-controlling zone which may produce damage-based AE. Thus, it is difficult to correlate the "most active" geographic sensor region with the eventual failure location unless one or both of the following are true:
   1. the previous maximum pressure prior to the FR ramp was high enough relative to the eventual failure pressure; and/or,
   2. the localized zone of the vessel controlling the burst strength is located sufficiently near (or directly below) the sensor within the "most active" geographic region.
3. Use of a waveform-based AE system with high speed processing capability should improve the ability to detect localized zones of damage concentration, and to correlate the AE from that localized zone with both burst pressure and eventual failure location.
4. Successful correlations of the Felicity ratio (FR) and burst pressure for those vessels which were not proofed to a significant fraction of the ultimate burst strength prior to the FR ramp are probably due to the FR being related to the overall quality of the composite material as fabricated, rather than being related to localized zones of damage.

Thirty-two graphite/epoxy pressure vessels (260 mm outside diameter) were evaluated using a sixteen-channel acoustic emission (AE) system. The vessels were proof-pressurized to approximately 70% of the nominal vessel burst strength. Later the vessels were pressurized to failure. Acoustic emission data (85-325 kHz) were gathered during the tests by non-resonant, small aperture sensors. The proof pressure profiles were structured to allow calculation of Felicity ratios (from subsequent pressure ramp cycles) and weighted event rate data (from holds at fixed pressure). Correlations between these AE data parameters and vessel failure pressures are presented for a subset of eleven vessels which experienced consistent testing conditions. An area location scheme based on hit arrival time analysis was used to identify the first sensor hit for each event.

1. The correlations of burst pressure to Felicity Ratio (FR) were frequently excellent for the graphite/epoxy vessels studied. This is similar to previous work on Kevlar®/epoxy vessels.
2. The best correlations of burst pressure to FR were linear for both "low burst" vessels (62-70 Mpa burst) and "high burst" vessels (74-81 Mpa bust), although the correlation slopes differed substantially for the two populations of vessels.
3. Using a high sensor density allowed meaningful analysis schemes which selectively examined data from "first hits" and a "most active channel." Since attenuation of AE signals within the graphite/epoxy composite was substantial (with typically only 2.6 to 5.3 of the 16 AE sensors being hit by significant amplitude events), the high sensor density also maximized the opportunity to record information from a greater quantity of events.
4. Substantial reductions in data scatter were achieved for correlations of burst pressure to FR when selectively analyzing data solely from "first hits". Additional improvements in the correlation of burst pressure to FR were achieved for "low burst" vessels when further reducing the data analyzed to that from a "most active channel."
5. The specific degree of correlation of burst pressure to FR was highly dependent upon the definition used for "AE activity" when determining the FR value. Also, the specific choice of quantitative value for "significant" when determining "significant AE activity" in defining the FR greatly affected the resulting degree of correlation of burst pressure with FR.
6. FR definitions based on sums of AE activity gave the most "robust" correlaions of Bust pressure to FR. FR definitions based on rates of AE activity produced correlations which were also good, but which were slightly less stable than correlations from FR definitions based on sums of AE activity. FR definitions based on AE-activity-per-hit produced correlations with an erratic degree of fit.
7. The correlations of burst pressure to two rate moments (specifically the Event Rate Moment and Energy Rate moment) calculated from AE data gathered during holds at fixed pressure were generally poor for the graphite/epoxy vessels studied here. This is in contrast with previous work using an event rate moment for Kevlar®/epoxy vessels.

A series of pairs of acoustic emission (AE) waveforms taken from either a closely spaced pair (64 mm) or a widely spaced pair (410 mm) of AE sensors mounted on a spherical graphite/epoxy pressure vessel of 260 mm outside diameter were examined. The waveforms from the small (6.3 mm dia.) aperture non-resonant sensors bandpassed from 85 to 325 kHz were recorded simultaneously on a 12-bit waveform recorder. Changes due to wave propagation in the typical AE hit characteristics (peakamplitude, duration, energy, spectrum, etc.) were correlated with distances from source to sensors and physical principles of wave propagation.

Based upon the results, implications for AE monitoring and future AE studies of large fiber/polymer composite structures are presented. Also, AE source location in conjunction with a model for localized failure of a large aerospace type fiber/polymer composite structure was considered. Large changes in typical AE hit parameters were observed, particularly in the first 60 mm of propagation. These results bring into question typical approaches that have been developed for source identification and assessment of source significance in small test specimens. Large changes in arrival times of the same event at separate sensors were observed as a function of the system threshold setting. Possible approaches to overcome arrival time deficiencies as well as suggestions for future studies are presented.

1. Standard AE parameters such as peak amplitude, rise-time, energy and duration used for characterization of AE hits in graphite/epoxy composite plate or shell-like structures experience large changes in the transition from the near-field to the far-field. Also, the spectral content of an AE signal changes significantly for near-field versus far-field measurements.
2. Even over rather short distances (e.g., 60 mm) of propagation, large changes in hit peak amplitude and energy occur. Thus, it does not make sense to characterize the significance of an event or the type of source by using standard AE parameters or spectral content unless the effect of the propagation distance from the source to the sensor is taken into account.
3. Differences in the far-field AE waveform characteristics arising from different composite A.E source mechanisms require fundamental study with large composite specimens.
4. Large source location inaccuracies in a graphite/epoxy plate or shell-like structure are primarily due to the use of fixed thresholds to determine hit arrival times, rather than being due to the inherent nature of the AE signals.

Small, filament-wound, Kevlar/epoxy, biaxial test specimens were subjected to various levels of impact damage at the Oak Ridge Y-12 Plant. The specimens were pressurized in a proof test cycle to 58% of their nominal, undamaged strength and then pressurized to failure. Acoustic emission data were gathered by multiple sensors during a 10 minute hold at peak proof pressure. Post-test filtering of the data was performed to study composite behavior in the damaged region and other areas. The rate and total amount of AE produced depends on the duration of the static load and degree of damage. The concept of the event rate moment is introduced as a method of quantifying a structure's total AE behavior when under static load. Average event rate, total long duration events, and event rate moments provided various degrees of correlation between AE and residual strength.

The following conclusions are valid for the specimens and test conditions outlined in this paper:

1. First hits from the damaged region are necessary for strength correlations.
2. Undamaged and damaged specimens produce AE when subjected to static loads. The emission produced by undamaged and lightly damaged specimens diminishes with time, which implies that a more stable stress state is being approached.
3. A damaged specimen always produces more AE than does an undamaged one.
4. Heavily damaged specimens with reduced residual strengths produce much more emission and exhibit emission rates which fluctuate in an unstable manner.
5. The normalized cumulative event rate moment can be used to quantify the static load event rate behavior and differentiate between stable and unstable rate behavior. Distinctions between behaviors increase with elapsed hold time.
6. Average event rates can also be used to differentiate between stable and unstable behavior when emission is known to come from damaged material. Distinctions are not as good when emission is detected at sites remote from the damage.
7. AE event duration distributions during the initial portion of a load hold are independent of damage. The number of long duration events (>1000 µs) increases with elapsed hold time for heavily damaged specimens and decreases for less damaged specimens.
8. Long duration events can be used to differentiate between stable and unstable behavior and to predict residual strength when unstable behavior occurs.

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