The Measurement and Significance of Energy in Acoustic.emission Testing

Purpose of investigation is to present an approach to measuring the energy sensed by an acoustic-emission transducer and to present experimental results that compare energy measurements with counting and RMS measurements in several different types of tests

by D.O. Harris and R.L. Bell

ABSTRACT--A technique for measuring the energy sensed at t . * = an acoustic-emission transducer is presented that utilizes a squaring circuit and digital integrator. Theoretical relationships between energy and other more conventional acoustic-emisw u = parameters, such as counts and RMS voltage, are derived for U = certain idealized cases. Experimental results from the following 0 = types of tests are presented : (1) unflawed tensile ('continuous'

U r = emission); (2)precracked stress-corrosion cracking; (3)pre- cracked fracture toughness; and (4)fatigue-crack growth. V ( t ) = Energy, counts, RMS-voltage, energy/event and counts/event Vo = measurements are included. In the case of unflawed tensile Vo. = specimens, energy techniques appeared somewhat superior to counts. In all other cases, a direct relationship between counts Vo. = and energy was obtained. Energy measurements tended to give a larger weight to higher amplitude events. Other than I/".= this, energy measurements appeared to have no advantage over counts. The theoretical relationship predicted between energy/event and count/event agreed quite welt with experi- V. = mental observations, Overall, the test results presented indicate that energy techniques provide no significant advantage over II,. = counting threshold crossings in cases in which crack extension 1/",. = in metals is the primary source of acoustic emission, e =

T :

Notation

time for acoustic-emission signal to digital integrator in energy processor to ring down to trigger level of processor energy for a single event total acoustic-emission energy energy rate/s, W-s/s acoustic-emission energy/load cycle volts amplitude of continuous-emission signal initial amplified and filtered voltage from single event to counter initial amplified and filtered voltage from single event to energy processor amplified and filtered acoustic-emission signal to counter amplified and filtered acoustic-emission signal to energy processor threshold voltage in counter threshold voltage in energy processor tensile strain, percent decay time of acoustic-emission signal [see eq (3)]

V n a , =

n =

N = N =

E l =

P = S =

t = tn ~

C = a constant [see eq (3)1 Introduction f = frequency of oscillations in an acoustic-emission

signal gain of amplifier chain to counter gain of amplifier chain to energy processor counts for a single event total acoustic-emission counts acoustic-emission count rate acoustic-emission counts/load cycle tensile load value of RMS-voltage signal above background noise time time for acoustic-emission signal to counter to ring down below trigger level of counter

D.O. Harris and R.L. Bell were associated with Dunegan/Endevco at time that paper was prepared; they are now associated with Science Appli- cations, Inc., Palo Alto, CA 94304 and Celesco Industries, Canoga Park, CA 91305, respectively.

Paper was presented at 1975 SESA Spring Meeting held in Chicago, IL on May 11-16.

Original manuscript submitted : November 16, 1974. Revised version received, May 20, 1977.

Acoustic emission is the term applied to the very low- level stress waves impulsively produced within a material subjected to stress. A review of this relatively new technology is provided in Ref. 1. Several techniques for measuring acoustic-emission activity have been used, such as counting the number of threshold crossings, RMS- voltage measurements, event counting, energy and ampli- tude measurements. Each of these techniques has certain advantages and disadvantages. Historically, threshold- counting techniques have probably been the most widely used, and counts have been successfully correlated with many physical activities within various materials (for example, see various papers in Ref. 1, and Refs. 2-8). Threshold-counting techniques are also the simplest to accomplish, are applicable to either continuous emission or emission events widely separated in time, and provide some measure of the amplitude of the event. However, the number of counts from a given event is only indirectly related to the physical characteristics of the event as seen at the sensor location. A more direct relation to such physical characteristics as event amplitude and duration is perhaps desirable, in that it may provide acoustic-emission

Exper imenta l Mechan ics I 347

data having the following desirable features : a. The data may be more readily relatable to the processes

occurring within the material. b. The data may not be so sensitive to the many factors

that affect acoustic-emission data. 9 This could eliminate the need to 'recalibrate' each time a new test situation arises.

c. The data may provide a clearer warning of impending failure. This would be a desirable feature in averting failure during proof testing.

Energy is a parameter that has been suggested that could provide the desirable features listed above. Previous investigators (for example, see Ref. 10)have presented data labeled energy release without providing information on how the energy was measured. Beattie and Jaramillo ~' have presented a circuit for energy measurement that appears to provide plausible measurements, and provided limited experimental data to support its utility.

The purpose of this investigation is to present an approach to measuring the energy sensed by an acoustic- emission transducer, and to present experimental results that compare energy measurements with counting and RMS measurements in several different types of tests. Conclusions regarding the merits of energy measurements are drawn based on these actual test results.

Description of Energy Technique The energy in an electrical signal is proportional to the

square of the voltage. Hence, the first step in a realistic energy-measuring unit as envisioned for acoustic-emission testing is to square the signal. It is then necessary to measure the area under the resulting curve. These steps are shown schematically in Fig. 1, where the signal from a single event is shown at various stages of processing. On an oscilloscope, the filtered and amplified signal appears as shown in the top trace of Fig. 1. ',8 A filter has been added between the squared signal and the integrator, and a digital integrator is used in order to minimize the drift that is commonly observed in analog equipment. This technique for energy measurement has several additional desirable characteristics such as the ability to multiply two signals from a given sensor but with separate signal conditioning (amplifiers and filters). This improves the signal-to-noise ratio, since the product of two random electronic-noise signals will be nearly zero, whereas the signals themselves will be in phase, and will therefore be enhanced. This could prove to be of great benefit in areas in which the emission signals themselves are of very low amplitude. Event counting is also simpler to accomplish when working with the squared signal, because the single event now produces a single pulse (bottom trace of Fig. 1), rather than several oscillations (top trace of Fig. 1). These two features of providing signal enhancement and simplifying event counting could make it _possible to count events in 'continuous' emission from tensile specimens--a measurement very difficult to perform using other techniques.

Relation of Energy to Other Acoustic-emission Parameters

The acoustic emission as measured by the proposed energy technique can be related to other more conventional parameters by making simplifying assumptions regarding the nature of the conditioned signal. In the case of 'continuous' emission, if it is assumed that the signal is of constant amplitude, V., and frequency, f , it can be

represented as*

V ( t ) = Vo sin 2 r f t (1)

Ignoring the effect of the filter between the squaring circuit and integrator, the energy rate, ~r, is simply

~l = I ~o II2(t) d t = Vo 2 I o~sin~2 l r f t d t o: Vo ~ o: (RMS voltage) 2

(2)

In the case of burst-type emission, the assumption that the signal from a single event is a damped sinusoid appears reasonable, and has been used in the past. '.7A2 Considering the case of using r gains for measure- ments of counts and energy, the following expressions for the voltage to the counter (V.) and energy processor (V.) are applicable :

V. = Vo, e - " ' s i n 27rf t = G . C e - " ' s i n 2~rf t (3)

V, = Vo, e- ' / ' s in 2 w f t = G . C e - " ' s i n 27rf t (4)

Assuming that the ringdown time is large compared to the period of oscillation, the following expression is obtained:

n = f t . * = f 7 In ( C G . / V , . ) (5)

The signal to the energy processor, after being squared, is

* The notation used is summarized at the beginning of thispaper.

AND FILTER ]

I l AMPLIFIER I

AND FILTER

SQUARING CIBCUIT

DIGITAL INTE .GRATO R I

I DIGITAL/ANALOG

COtiVERTER

1 DC SIGNAL

V v v V'

Fig. 1--Schematic representation of components in energy-measuring device

348 I September 1977

I1". 2 = ( C G . ) 2 e -2,/, sina21rft

1 ( C G D ~ e _ 2 . , (1 - cos 4 7 r f t ) (6)

Assuming that the effect of the filter between the squaring circuit and integrator is to eliminate the cos 47rft term in eq (6), the voltage signal to the integrator is

1 V . 2 = ~-- (CG.)2e -2"" (7)

The energy for the event can then be calculated

G.2C 2 u = "")o'*V." ( t ) dt -

2 G . 2 C 2 r

- _ _ q _ _ _ (] - e -2 , . , . )

Solving the relationship

g o u 2 ,

v,2 = - - 2 - e - - : -

- - j ? ' e -2''" d t

(8)

(9)

for t.* leads to

t.* = - ~ I n [ ~ - ( )21 (1o)

Substituting t-his into eq (8), observing that Vo. = C G . , and using the expression for C obtained from eq (5), results in the following relationship between u and n

rV,. 2 ( 3 . V,.G. 2 u - ( (3.-)2 e2.m [1 - 2 ( .~----~-~ ) e- 2./ y.] (11)

4 v , . t . r . n

This equation is simplified to the following in the case

V..G. 2 e-2.u. where ( ~ ) is much less than one.

r V " 2 u ~ ( )2 e2.,p (12)

The above relationships should not be considered to be exact, or even very accurate, since the model employing damped sine waves has been found to be inaccurate in some cases2 Experimental results presented in later sections of this report show the usefulness of these relationships,

E x p e r i m e n t a l R e s u l t s and D i s c u s s i o n

Turning now to experimental comparisons of acoustic- emission data as taken by various techniques, a counter trigger level of 1 volt peak was used in all cases. A 10- KHz low-pass filter was used in the energy module, and various trigger levels for energy measurements were utilized. The integrator integrated only when the level of the squared and filtered voltage exceeded the trigger level, and the area under the curve all the way down to zero volts was measured. The transducers, gain settings, bandpasses and trigger levels in the energy module are indicated in the corresponding figures.

Tensife Tests

The emission from an initially flaw-free tensile speci- men of 7075-T6 aluminum was selected as an example of 'continuous' emission. Count rate, RMS voltage and energy rate were simultaneously monitored. The RMS voltage has been shown by Hamstad and Mukherjee '3 to be superior to counting techniques in this instance, because the RMS voltage was linearly related to the strain rate, whereas no simple relation between count rate and strain rate was observed. Typical tests results obtained at a crosshead velocity of 2.54 mm/min (0.10 in. /min) are presented in Fig. 2. The RMS level above background

i

Cr

/ D 1 4 0 , 9 6 d b . 0.1 - O . 3 M H z

500

Z -- 400

- - 300 qlJ

u c

~ r4_~_ - ~oo

i ~176 E o

�9 = o.o, .,o " NIII

I 0

2 4 b 8 1O

tensile s | ra in , percent

11o 2

lO

r 0

c o u

.Z

Fig. 2--Acoust ic emission and stress as a function of strain for 7075-T6 aluminum tensile specimen tested at a cross- head velocity of 2.54 mm/min (0,10 in./min)

Experimental Mechanics i 349

1 0 5

l o 4

10 3

e I#1

C

0 U

. 10 2

-Z

10

8

1

5

S L O P E = 4 . 3 3

§

+ §

+ + 4

4-§ 4 4 4 0 § § 4" §

4* +

't " t

/ D 1 4 0 , 9 6 d b

0 . 1 - 0 . 3 M H z

10 5 0 100 5 0 0

S, millivolts rms

Fig. 3--Acoust ic-emission count rate vs. RMS signal above noise for tensile tests on 7075-16 aluminum conducted at various crosshead velocities

noise (denoted as S) is shown in this figure. The results show the typical behavior of the emission activity beginning with the onset of plastic deformation, passing through a peak at a few-percent strain, and decreasing at larger strains. All three acoustic-emission parameters peak at the same strain. The tests were stopped at 10-percent strain, and the gage length of the specimens was about 5 cm (2 in.) in all cases.

Tests were performed at various crosshead velocities in the range of .5 to 25 mm/min (0.020 in. /min to 1.00 in./min). In all cases, the peaks occurred at the same strain level for all three acoustic-emission parameters.

Data points were taken at various strain levels in the tests performed at various rates. A fair amount of scatter was observed, and the linear relation observed by Hamstad and Mukherjee '3 was not apparent. Attempts were made at changing the cross-head rate during the test to eliminate specimen-to-specimen variations in the effect of strain rate on acoustic emission. However, considerable overshooting of the load when the rate was increased (and 'undershooting' when decreased) with corresponding overshooting (and undershooting) of the acoustic emission complicated analysis of the test results.

16 2 -L

U 0 bq

t-~l ~

0

- 11~3

�9 ~ 8

~4

| i i

D140 , 9 6 d b , 0.1 " 0 . 3 M H z

g br 0

o o

Slope ~ 2 .84

u

o

oo

,t.

7075-T6 AI 4" Nodular I ron % Nodularlty

0 Q

25

5 0 a

7S o

100 o

5 10 50 100 5 0 0

S, millivolls rms

Fig. 4--Acoust ic-emission energy rate vs. RMS signal above noise for tensile tests on 7075-T6 aluminum conducted at various crosshead velocities, and 3.9-percent carbon nodular cast iron with varying nodularity conducted at a crosshead velocity of 1.3 mm/min

A plot of the count rate vs. the RMS signal is presented in Fig. 3. A straight line with a slope of 4.33 provides a fit to the data, except at extreme values of S. The energy- rate data as a function of the RMS signal are presented in Fig. 4. A straight line with a slope of 2.84 provides a fit to the data, except at higher values of S. Hence, the second-power relation predicted by eq (2) does not hold, which is undoubtedly due to the truly discontinuous nature of the emission. The data of Figs. 3 and 4 show that the count rate varies much more strongly with RMS voltage than the energy rate. This suggests that energy rate is somewhat superior to count rate in this case.

The results of tensile tests performed on cast iron containing 3.9-percent carbon with varying degrees of nodularity are also shown in Fig. 4. The cast-iron tests were performed at a crosshead velocity of 1.3 mm/min (0.05 in./min) on specimens having the same geometry as the 7075-T62aluminum specimens. The cast-iron data . show a 1.70 power relationship between the energy rate and RMS-signal level. Hence, it is apparent that a general relationship between energy, counts and RMS voltage does not exist, and that the relation between these parameters depends on the material.

Stress-corrosion Cracking and Rising-load Fracture Toughness

Tests were performed on 6.4-mm (�88 compact tension specimens" of 4340 steel quenched and tempered to a Rockwell C hardness of 52 and 7075-T6 aluminum. Fatigue precracks were introduced into all specimens. The

350 I September 1977

100 I I I I I

I:

o t~

o v l

"o c o 3 o

,.1=

=s

80

60

40

S t r e s s C o r r o s i o n (S140) ( 9 2 0 1 )

( S t e e l ) s p e c 1 v �9 s p e c 2 G A

U N F r a c t u r e T o u g h n e s s

. . ~ S t e e l .

.1: m s p o c 1 a I �9

N C) s p e c 2 r �9 �9 ~

I A l u m i n u m �9 d spoc 1 o . ~ 6

s p e r 2 (> �9 �9 o

, , 2 0 ~ 4'

�9 & v &n

All& �9 V �9 6"

::,', ~176 201- D

�9 �9 a v

/ o 0 _ j I I

0 5 10 15 20

v

v v v

9201 , 4 6 d b , V t ~ : 4 x l O - 3 V 2

1 0 K M Z l o w p a s s

I 25 30 X 10"3

U j V 2 - s e c

Fig. 5--Total counts as a function of total energy for stress-corrosion cracking of 4340 steel in salt water, and fracture toughness tests on 4340 steel and 7075-T6 aluminum

stress-corrosion specimens (all of which were steel) w e r e loaded to values larger than used in the precracking, and were subjected to salt-water--which produced stress- corrosion cracking. ~' Two sensors, Dunegan/Endevco Models S140 and 9201, were used. Figure 5 presents plots of the count data as a function of energy, for both the stress corrosion and rising-load fracture tests. It is seen that, in each case, a linear relation between counts and energy was obtained, although the slope of the N-U relation varied from test to test. The reason for this variance is not known. The observance of a linear relationship between counts and energy shows that energy techniques have no advantage in this case.

RMS measurements of the emission were not made in this case, since the highly transient signals are not well suited for analysis by this technique.

Another series of risingdoad fracture tests was per- formed using the materials mentioned above. In this series, the instrumentation was set up to print out the counts and energy per event for each event, so that the relationship between n and u could be obtained directly. Figure 6 presents a semi-log plot of u versus n for the aluminum. The gain and threshold settings used are such that eq (12) is applicable. Figure 6 shows that eq (12) fits the experimental data well, except at large values of u. The energy processor has a limited dynamic range (26 dB), so that it would underestimate u when u is large. This is also the reason for the plateau in u.

The data of Fig. 6 show that a direct relation exists between n and u, and that this relation is quantitatively predicted by the damped sine model�9 The direct relation between n and u indicates that energy measurements have no advantage over counts, except that energy does weigh the higher-amplitude events more heavily. However, the

~2 1 I I : I I

103

16 4

i .

2 2 l|nT Gu/ e2n//h"

5 -4 Ix10 sec 50 K H z)

§

$140 , 0.1- 0.3MH, 7075 -T6 Aluminum precracked ~ rislng-load

80db

n , c o u n t s / e v e n t

Fig. 6--Energy per event as a function of counts per event for acoustic emission from precracked 7075-T6 aluminum during rising load

Experimental Mechanics t 351

'~176 s,.., sp . ..... . - / Z T - - pulJer on side

�9 b<~

+ o e / 300

Spark of center ~ + ~r C �9 ,~

~P o., -o.~, , . ,~,; ~,,o o

o ~ . o �9

~&// �9 * o

P oN + o ~ e k e

,oo /..

40 50 60 70 80 90

g a i n , d b

Fig. 7--Counts per event vs. amplifier gain for repeatable events using various simutated acoustic-emission sources. 4340-steel compact tension specimen. Pulser mounted on specimen face unless otherwise noted

limited dynamic range of the energy processor is a disadvantage.

The parameter f r can be adjusted to provide an optimum fit between the theoretical and experimental n-u relation. If the acoustic emission is white noise, f will be determined primarily by the response of the sensor, r is related to the ringdown time of the processed sensor signal, which will be a function of the sensor damping, specimen geometry and specimen damping. The value of f r gives the slope of the n-u relation on a semi-log plot, with a value of 35 being obtained for the aluminum, as shown in Fig. 6. The best fit for a similar plot of the steel data provided a value of f r of 80, but the scatter in the steel data was larger than for the aluminum. The value of r controls the position of the n-u relation, with a value of 1.4 x 10-' s resulting for the aluminum (which gives f = 250 KHz, which is within the bandpass used). r for the steel data was not calculated, because of the poorly defined n-u relation for this material.

Values of f r can be obtained by means other than the slope of the n-u relation. One way is to measure count/event at two different gain settings for a given event. Using eq (5),

CG1 CG2 G, n , - n, = f r [ l n ~ - ln---~-.] = f r l n

G,

03)

Hence, if n for a given event at two different gain settings is known, f r can be calculated. Such measurements were made from a steel specimen during rising load, and a value o f f r of 190 was obtained.

Another technique for evaluating f r is to use a simulated acoustic-emission source, and determine how n varies with gain. Equation (5) predicts a linear relation between n and gain expressed in terms of dB (since gain in db = 20 log G), with f r being the slope of the line. Such measurements were made using an acoustic-emission sensor (Dunegan/Endevco S140 or $750) or an ultrasonic transducer (SIL5.0) as a sender. The sender was coupled to the specimen and driven by short-duration voltage pulses, which caused short-duration stress waves of

repeatable amplitude to be introduced into the specimen. Of the three senders, the SIL5.0 ultrasonic transducer has the flattest frequency response, and should therefore be the best simulation of white noise. A spark discharge was also used as a repeatable simulated acoustic-emission source," and measurements of n vs. gain were again made. Figure 7 presents the n-gain results for the steel, as measured by the various simulated emission sources. The good degree of linearity between n and gain is shown. The values of f r determined by the various techniques for both the steel and aluminum are summarized in Table 1. Figure 9 and Table 1 show that the values of f r vary widely depending on the technique used for its evaluation. Hence, it is not possible at this time to determine a definitive value o f f r for a given sensor and specimen.

Measurements of u were made in conjunction with the n-gain tests using the various simulated-emission sources. The results showed that the exponential relation between n and u predicted by eq (12) holds, but that the value of f r is dependent on the source of the simulated emission. This is consistent with the results presented in Fig. 7 and Table 1.

Hence, it can be concluded that the damped sine-wave model provides accurate quantitative relations between counts and energy, when f r is considered to be an adjustable parameter. Values of f r in the n-u relation cannot be reliably determined by 'calibrations' involving simulated-emission sources, or even by measuring n at different gain settings for a real-emission event.

Fatigue

Fatigue tests were performed on 25.4-mm (1-in.)-wide, 0.8-ram (1/32-in.)-thick single-edge-notched specimens of 7075-T6 aluminum using procedures outlined in Refs. 4 and 17. Briefly, a sharply notched specimen was cycled between fixed loads while continuously monitoring for

TABLE 1--VALUES OF fr AS DETERMINED BY VARIOUS TECHNIQUES

Equation Technique Used Aluminum Steel

$140 sensor used as pulsed 5 sender mounted on speci- men face

$140 sensor used as pulsed 5 sender mounted on speci- men edge

$750 sensor used as pulsed 5 sender mounted on speci- men face

SIL5.0 ultrasonic transducer 5 used as pulsed sender mounted on specimen face

Spark used as repeatable 5 source

Acoustic-emission counts per 13 event from cracking of 4340 steel under rising-load at two different gain settings

n-u relation from acoustic 12 emission from cracking during rising-load tests

31 64

106

31 64

19 26

114 120

- - 1 9 0

34 80

352 I September 1977

acoustic emission. Count and energy measurements were made. The results presented in Refs. 4 and 17 showed that the counts per cycle (N ' ) passed through a peak, even though the crack-growth rate increased monotonically. This peak is believed to be associated with a plane-strain to plane-stress transition that occurs as the crack extends. Figure 8 presents a plot of the energy per cycle ( U ' ) as a function of the counts per cycle (N') . Open data points are for information before the peak in N ' , and solid ones are for data after the peak. A unique relation between N ' and U ' is shown in Fig. 8. In fact, a linear relation between N ' and U ' would be a good fit to the data. Hence, energy techniques have no advantage over counts in this case.

C o n c l u s i o n s

The following conclusions can be drawn from the results presented in this paper :

1. The ability to obtain enhanced signal-to-noise ratios by energy techniques may be advantageous in certain instances where the emission amplitudes are very low,

2. Energy rate appeared superior to count rate in the case of 'continuous' emission because energy rate did not vary as strongly with RMS voltage, which, in turn, is closely related to the important test parameter of strain rate.

3. A linear relation between counts and energy was obtained during stress-corrosion-cracking tests and rising- load fracture-toughness tests.

4. The linear relation observed between counts and energy in the cracking tests indicates that energy techniques have no advantage over simple counting techniques in this case.

5. A unique relation between counts/event and energy/event was observed during rising-load fracture- toughness tests.

6. The counts/event and energy/event data showed that energy techniques tend to weigh high-amplitude events more heavily than count techniques. This could result in energy measurements providing clearer warning of impending failure. However, the limited dynamic range of a practical energy-measuring instrument would tend to nullify this potential advantage.

7. The damped sine-wave model provided a quantitative relation between counts/event and energy/event, if f r is considered as an adjustable parameter.

8. The several different techniques for evaluating f r provided widely varying values. Hence, it is not possible to reliably evaluate f r by 'calibrations' involving simulated emission sources.

9. A linear relation between counts/cycle and energy/cycle was observed in the case of continuous monitoring of fatigue-crack growth. Hence, energy techniques have no advantage over counting techniques in this case.

Although the results presented here do not indicate any marked advantage of energy over counts, this does not eliminate the possibility that energy techniques may be advantageous in certain instances.

References

1. Acoustic Emiss'ion, ASTM Special Tech. Pub. No. 505, ASTM, Philadelphia, PA (1972).

2. Dunegan, H.L. and Tetelman, A.S., "'Nondestructive Characteriza- tion of Hydrogen Embrittlement Cracking by Acoustic Emission Techniques, "" Engrg. Fract. Mech., 2, 387-402 (Jun. 1971).

3. Harris, D.O., Dunegan, H.L. and Tetelman, A.S., "'Prediction of Fatigue Lifetime by Combined Fracture Mechanics and Acoustic

10 - 2 i i i

S140, 0 . 1 - 0 , 3 H H Z

P m o z , Pmin , N N �9

~0-3 & 2200 450 Spec 1 I~

D 2660 1060 ~per 2 2 6 6 0 1 0 6 0 r162 3

.- I

I0 -4 _,' : . "

>. _o

o > o ."

> ~>-= / L , ~ v n

10 -? []

oo

[]

8 6 d h

10-8 -3 ,;i I I 1 I I

10 10 -2 10 1 I0 10

N~ c o u n t s / c y c l e

Fig. 8--Acoustic-emission energy per cycle vs. counts per cycle for continuous monitoring of fatigue-crack growth in 7075-T6 aluminum

Emission," Proc. Air Force Conf. Fatigue and Fracture of Aircraft Structures and Materials, AFFDL Rep. AFFDL TR 70-144, 459-471 (1970).

4. Harris, D.O. and Dunegan, H.L., "Continuous Monitoring of Fatigue Crack Growth by Acoustic-emission Techniques, '" EXPERIMENTAL MECHANICS, 14 (2), 71-81 (Feb. 1974).

5. Dunegan, H.L. and Harris, D.O., "Acoustic Emission Techniques, "" Exper. Tech. Fract. Mech., Kobayashi, A.S., ed., SESA, Westport, CT, Ch. 3, 38-75 (1973).

6. Magnani, N.J., "Acoustic Emission and Stress-corrosion Cracking of U-4�89 wt. O/o Nb,"EXPERIMENTAL MECHANICS, 13 (12)(Dec. 1973).

7".Harris, D.O., Tetelman, A.S. and Darwish, F.A.L, "Detection of Fiber Cracking by Acoustic Emission, "" Acoustic Emission, ASTM Special Tech. Pub. No. 505, ASTM, Philadelphia, PA, 238-249 (1972).

8. Brindley, B.J., Holt, J. and Palmer, L G., "'Acoustic Emission--3, The Use of Ringdown Counting, "' Non-Destructive Testing, 6, 299-306 (Dec. 1973).

9. Dunegan, H.L. and Green, A.T., "'Factors Effecting Acoustic Emission Response from Materials, "" Acoustic Emission, ASTM Special Tech. Pub. No. 505, ASTM, Philadelphia, PA, 100-113 (1972).

10. Hutton, P.H., "'Acoustic Emission Applied Outside the Laboratory, "" Acoustic Emission, ibid., 114-128.

11. Beattie, A.G. and Jaramillo, R.A., "'The Measurement of Energy in Acoustic Emission, "" Sandia Laboratories, Albuquerque, NM (undated),

12. Tetelman, A.S., "'Acoustic Emission and Fracture Mechanics Testing of Metals and Composites, "" Materials Department, UCLA, No. UCLA-ENG-7249, presented U.S.-Japan Joint Syrup. Acoustic Emission, Tokyo, Japan (JuL 1972).

13. Hamstad, M.A. and Mukherjee, A.K., "'The Dependence of Acoustic Emission on Strain Rate in 7075-7"6 Aluminum, "" EXPERIMENTAL MECHANICS, 14 (1), 33-41 (Jan. 1974).

14. Wessel, E,T., "'State of the Art of the WOL Specimen for Klc Fracture Toughness Testing, "' Engrg. Fract. Mech., 1 (1), 77-103 (Jun. 1968).

15. Johnson, H.H. and Paris, P.C., "'Subcritical Flaw Growth, "" ibid., 3-46.

16. Bell, R.L., "'Acoustic Emission Transducer Calibration--Transient Pulse Method, "" Dunegan/Endevco Tech. Rep. No. DE-73-3, Dunegan/- Endevco, San Juan Capistrano, CA (Feb. 1973).

17. Harris, D, 0., "'The Effect 'bf Gain and Frequency Bandpass on Acoustic Emission Observed from Growing Fatigue Cracks, "" Dunegan/- Endevco Technical Report No. DE-74-4, San Juan Capistrano, CA (Jan. 1974). Presented at Syrup. Schallemission Anwendung bei der Untersuchung, Prufung und Uherwaehung metallischer Werkstoffe, sponsored by Deutsche Gesellschaft fur Metallkunde, Munich, Germany (Apr. 1974).

Experimental Mechanics [ 353