59015 Ultrasonic Detection and Measurement of Fatigue

Cracks in Notched Specimens

A reflection technique is employed to detect and measure fatigue cracks, nondestructively during test, in circumferentially notched cylindrical specimens subjected to reversed axial-fatigue loading -

by S. J. Klima and J. C. Freche

ABSTRACT--~An ultt'asonic technique was developed and used to observe the formation and growth of fatigue cracks in notched cylindrical specimens subjected to reversed axial cyclic loading. Fatigue curves of Iife-to­initial detectable cracks as well as life-to-fracture were obtained for an aluminum-. a titanium- and a cobalt­base alloy, as well as a maraging steel. Depth of initially detectable cracks ranged between approximately 0.0005 and 0.004 in. ,0.013 and 0.10 mm). Curves were also obtained I'elating ultrasonic system output voltage to crack depth IIp to 0.030 in. (0.76 mm) fol' three materials. These curves were llsed to demonstrate the capability of the device for monitol'ing crack growth.

Introduction

Fatigue involves the processes of crack initiation and propagation prior to fracture. Any method that can be used to detect small fatigue cracks non­destructively during the course of a fatigue test would be extremely useful as a research test tool. If the method could also be used to monitor fatigue­crack growth, its usefulness would be even greater.

Methods to detect fatigue cracks are presently available. Generally speaking. each has difficulties associated with its use which are more or less severe, depending on the intended application. For example, commonly used inspection methods such as pene­trating-liquid, magnetic-particle and radiographic techniques, when applied to fatigue specimens, all require interruption of the fatigue test. There are additional limitations becau~e the penetrating-liquid and magnetic-particle techniques can only be used to detect cracks at or near the surface, and X-ray techniques p08e problems of safety and interpreta­tion. Optical microscopy is probably the most posi­tive means of measuring the size of fatigue cracks, but this method requires highly polished surfaces and generally involves termination of the test and sectioning of the specimen before examination. A recent preliminary investigation \ utilized microwave methods to detect machined grooves approximately O.OOOl-in. (0.0025 mm) deep and O.S-in. (20.3 mm) long in the surface of an untested specimen. The investigators predict that further development of

s. J. J{lima is Research l'-Ietallurgist and J. C. Frcchc is Chief, Fatigue & Alluys Uescarch Branch, NASA-Lewis Research Center, Clet'eland, Ohio. Paper u'as presented at 1968 SESA Fall ~"Ueeting held in San Francisco, Calif. 011 Octuber ~s ·~Vot'ember 1.

the technique should permit detection of even smaller cracks. Applicability of the technique to specimens while they are undergoing a fatigue test was not demonstrated, however.

In order to detect fatigue cracks accurately, regardless of material, and without interruption of the fatigue test, we originally applied an ultrasonic method similar to that used for nondestructive­inspection purposes for the early detection of fatigue cracks in center-notched sheet specirnens." In that initial investigation, the reflection technique was used and fatigue cracks ranging in length from 0.0005 to 0.005 in. rO.013 to 0.13 mm) were detected while the tests were in progress. Another investi­gator" used ultrasonic surface waves to detect surface flaws in bending-fatigue specimens, but crade sizes were not determined. Still others I have applied ultrasonic inspection techniques to detect cracks 0.003 to 0.004-in. 10.08 to 0.10 mm) in length in thin (0.039 in.) (1.0 mm) center-notched steel-sheet specimens that were tested in axial fatigue.

Our earlier investigation" was limited to axial tension-fatigue tests of sheet specimens and the transducer design was therefore tailored for applica­tion to sheet specimens. In the present investigation, the transducer design was altered to permit applica­tion of the ultrasonic method to cylindrical (notched) specimens subje~ted to reversed axial-fatigue load­ing. The reflection technique was employed to detect fatigue cracks during test in 2014-T6 aluminum, 5Al ~ 2.58n ~ titanium, a 300 grade mal' aging steel, and a cobalt-base alloy, L-605. Fatigue curves were obtained showing cycles-to-initial detectable crack, and life-to-fracture. Some specimens were removed from test, sectioned and examined optically to relate crack length to system output voltage. The device was calibrated in terms of output voltage as a function of crack depth for three materials over a range of stresses.

Principles of Crack Detection by Ultrasonic Reflection Technique

The principles of ultrasonic-wave propagation are described in detail in Refs. 5 to 7. The following is a brief review of the theory involved in applying the

Experimental Mechanics I 193

reflection technique to the detection of fatigue cracks. The technique is similar to the use of radar in the detection of distant objects. Acoustic energy, in the form of high-frequency waves, is transmitted from a transducer into the test specimen. After transmission of each pulse of energy, the transducer acts as a receiver for energy reflected from any dis­continuity in the specimen. The metal-air interface of a fatigue crack constitutes such a discontinuity. The low density of air and the relatively low velocity of ultrasonic waves in air result in an acoustic mis­match that causes the reflection of incident ultra­sonic waves. The amount of energy reflected from a crack is directly related to crack area, the intensity of the incident ultrasonic wave and the orientation of the crack. In this report, we will discuss crack size in terms of the average depth of penetration around the entire circumference of the notched cylindrical specimen.

System Design and Operation

A block diagram of the ultrasonic crack-detection system is shown in Fig. 1. A commercial ultrasonic flaw detector was used in this investigation. This unit contained a pulse generator used to drive the piezoelectric crystal in the transducer. It also con­tained amplifiers and a cathode-ray tube that dis­played the reflected energy pattern, as well as time­gating and integrator circuitry. Filter circuitry and an oscillograph were added. The oscillograph was used to obtain a permanent record of the signal reflected from the notch and the propagating crack in the specimen.

Specimen-Transducer Design and Assembly

Figure 2 shows a sketch of the notched-cylindrical fatigue specimen used in this investigation and Fig. 3 shows the specin:en-transducer-loading rod assembly. Longitudinal ultrasonic waves were transmitted to the specimen by a transducer mounted on the speci­men end. Access to the end of the specirren when installed in the fatigue machine was achieved through an opening in the upper loading rod (Fig. 3). The transducer was clamped in position by a yoke which screwed onto the end of the specimen and was large enough to accommodate a H-in. (13 mm)-diam, %-in. (10 mm)-thick transducer.

The specimen was machined so that a O.4-in. (10 mm )-diam section extended beyond the threaded portions (Fig. 2). Since the transducer was clamped against this button which had a diameter less than the root diameter of the threads, the amount of stray ultrasonic waves which would ordinarily reach the tranl:'ducer from reflective surfaces on the threads was reduced. No fillet was provided between the test section and the threads so that reflected waves from the specimen shoulder could return directly to the transducer and be displayed on the cathode-ray tube as a distinct spike. It had previously been observed that a generous fillet in this region caused the re­flected waves from the fillet to travel a longer, more

194 I May 1969

indirect route, so that they arrived at the transducer at about the same time as the reflected signal fro111 the specimen notch. This caused a superposition of the signals which is undesirable since it tends to mask the reflected signal from the specimen notch. For the same reason, it was necessary to position the specimen notch so that its distance from the trans­ducer was not a multiple of the distance between anv other strongly reflective surface and the transduce;'. Thus, secondary reflections caused by waves bounc­ing back and forth between the end of the specimen adjacent to the transducer and discontinuities such as the shoulder were not superimposed on the signal from the notch.

It was necessary to provide a coupling medium between the transducer and the fatigue specimen to eliminate air from the interface and allow efficient transmission of ultrasonic energy. Lubriplate was used, but a wide choice of petroleum or silicone lubricants would probably be satisfactory. To insure a good fit between the transducer and the specimen, the specimen ends were polished with 600-grit abra­sive paper to obtain a Slr.ooth flat surface.

Transducer Characteristics

A transducer frequency of 7.0 111Hz was cho:'.:n for this investigation. Theoretically, higher frequency transducers, because of their associated short wave­length, are sensitive to smaller flaws than are low­frequency (long-wavelength) transducers. However, energy losses by attenuation are greater with short wavelengths than with long wavelengths. It is desirable, therefore, to seek a compromi~'e and choose as high a frequency as possible without excessive attenuation losses. Generally, this mu~t be done b~' experiment.

Determination of the optimum frequency for a particular application can be hampered by apparent differences in energy levels produced by different transducers. In this investigation, for example, pairs of 7.0-MHz transducers were purchased from the same manufacturer on three different occasions (a pair at a time). Each pair provided a different level of reflected energy from the fatigue-specimen notch. The pair which was found to be the n:ost Eenf'itivc in obtaining reflections from the specimen notch wa:, used for our tests. A pair of lO.O-MHz transducers purchased from the same supplier had approxin:ately the same sensitivity as the least sensitiw pair of 7.0-MHz transducers. This points up a difficulty likely to be encountered by investigators in selecting the proper transducer for a particular experiment.

Operation of System Electronics

Ultrasonic pulses were transmitted at the rate of 500/sec with a pulse time of about one )1sec. Since a typical velocity for longitudinal waves in the speci­mens used in this investigation was about 0.2 in. (5 mm),',usec, sufficient time was provided between pulses for all reflected signals to retuhl to the drive crystal. These reflected pulses were reconverted to

--------------1 I 1

t 1 ,--+~ I

1

I 1

I

Fig. 1-Crack-detection system

513 -IS'lf thl€.'dljS

i l: I: _ ~- .

, ' -I ,II 'I I I" I' : ' [ .Ii

..-- -- ~'Xk nut

~. ______ rransducer clamplr; ::

-. --~ SO~('lmen end

_..--- Th readed loading r :':

J \: .~;I-~i2-4S1~b ~-,n~-l[======-=:::::]J==~-.---L-fl-~"J20 ;,,;;, i Centerless ends

~-~a:i~s,t).':LG -, '0.2::: - -

·4,00 in,Ii02 mml----

Fig. 3-Fatigue specimen, transducer a -: loading·rod assembly

Fig 2-Notched cylindrical fatigue specimen. Theoretical stress concentration, 3.65

electrical signals by the transducer, amplified and displayed on the cathode-ray tube.

The commercial ultrasonic equipment included a time gate and integrator circuitry (Fig. 11. The gate allowed only the reflected signals occurring within a preselected time interval after each transmitted pulse to pass through to the integrator circuitry. Because the distance traveled by an ultrasonic pulse is proportional to time, the time gate may be inter­preted as a "propagation-distance" gate. The output of the amplifier was gated for the specimen position at which fatigue cracking was expected to occur. Extraneous reflections from the specimen-transducer interface and from within the specimen were blocked by this gate.

The integrator circuitry provided a d-c voltage proportional to the signal that passed through the time gate. After filtering to remove minor fluctua­tions in the integrator output, the resulting d-c voltage was recorded on an oscillograph. Changes in the recorded voltages were proportional to changes in the amount of ultrasonic energy received.

Material and Fatigue-test Procedure

Specimen lHaterials

Specimens (Fig. 2) were machined from %-in. (19 mm) bar stock of four alloys, 2014-T6 aluminum, 5Al(aluminuml - 2.5Sn(tin) - titanium, a maraging steel, and a cobalt-base alloy, L-605. The nominal composition of the alloys is given in Table 1. The condition of these materials and their tensile strengths as obtained with the specimens used in this investigation are listed in Table 2. All test speci­mens contained a circumferential notch having a theoretical stress concentration, K/, of 3.65.' Stress

TASLE I-NOMINAL COMPOSITION OF ALLOYS TESE:: . ..

Element -Alloy

2014-T6 5Al·2.5Sn-Ti Maraging L-GO: (ELI) steel Composition, wt. %

AI Sal. 5.2 0.10 Ca 0.05 C 0.08 max. 0.03 ma:<. 0.10 Cr 0.10 max. 20.0 Co 9.00 Sal. Cu 4.4 Fe 1.00 max, 0.15 max. Sal. 3.00 ma / Mg 0.5 Mn 0.8 0.10 max. 1.5 Mo 4.80 Ni 18.50 10.0 Si 0.8 0.10 max. 1.00 mal Sn 2.5 Ti 0.15 max. Sal. 0.60 W 15.0 Zr 0.02 Zn 0.25 max, S 0.003 S 0.010 max. 0.030 rna ( p 0.010 max. 0.040rr~/_

0 0.12 max. H 0.0125 max. N 0.05 max.

values were calculated on the basis of the minimw::::. cross-sectional area bounded by the notch root. Th", maraging steel was tested, both in the aged and iL the as-received condition, This was done to det,,::-­mine the difference in response of the ultrasonic' method to a material with the same composition bu,: with different mechanical properties.

Fatigue Tests

Specimens were subjected to reversed loadin?

Experimental lVlechanics I 195

TABLE 2-TENSILE STRENGTH AND CONDITION OF ALLOYS TESTED

Tensile strength, notched

specimens Material Condition Ib/in.' MN/m'

2014-T6 T6 (artificially aged) 97 400 672 5Al-2.5Sn-Ti Annealed (as-received). 213 400 1471

(ELI) Rockwell hardness, Rc, 30-33

Maraging Sol ution annealed 1500° F 413 700 2852 steel for % hr, air cooled; aged

9000 F for 3 hr, air cooled Maraging Solution annealed 15000 F 253 500 1748

steel for % hr, air cooled (as-received)

L-605 Solution treated (as- 166 500 1148 received)

applied in a sinusoidal pattern in a hydraulically actuated axial-fatigue machine. The frequency of load application varied between 0.3 and 1.2 Hz depending on expected life-to-fracture. All tests were conducted in air at ambient temperatures.

Detection 0/ I nitiaZ Crachs by Ultrasonic JI ethod

Depth of the initially detectable cracks by the ultrasonic reflection technique was less than 0.005 in. (0.13 mm) for all of the materials considered. The transducer was positioned at one end of the specimen and longitudinal waves traveled along the entire specimen length. The reflections of these waves from the notch were observed as a spike on the cathode­ray tube. The same reflected signal, after passing through the time gate, was also recorded on an oscillograph (Fig. 1). The amplifier suppression (d-c bias) was adjusted to reduce the notch signal to an output level such that an oscillograph recording was made near the middle of the chart at the start of the test, leaving room to record either a decreasing or an increasing signal.

Figure 4 schematically illustrates how the recorded voltage output was affected by applied load, as well as crack formation and crack growth, during the early part of a fatigue test for our test specimen. The direction of increasing reflected signal received by the transducer is toward the top of the figure. The broken line EF (Fig. 4d) represents the level of out­put voltage recorded whenever the peak compressive load was applied to the specimen. The solid line ABCD represents the level of output voltage when­ever peak tensile load was applied. Note that the latter first decreases and then increases with increas­ing mmlber of applied load cycles. Since the output voltage at peak tensile load is of primary interest for the detection of cracks this curve will be discussed in detail.

A sketch of a longitudinal section in the region of the specimen notch at no load prior to beginning the fatigue test is shown in Fig. 4a. The angle between the notch surface nearest the transducer and a plane perpendicular to the direction of the ultrasonic waves is designated avo The ultrasonic waves reflected from the notch, before load is applied, provide an output-

HlR I i\1a.v 1969

voltage level which is about midway between points A and E on the two output-voltage curves. vVhenever the peak compressive load is applied, the notch angle is reduced to ex,. (Fig. 4bl. This causes the notch surface from which the inciclent waves are reflected to be more nearly perpendicular to the direction of the incident ultrasonic waves and results in increased refleded energy over the no-load condition. The voltage level at peak compressive load remains con­stant even when cracks first appear because the crack surfaces are so tightly squeezed together as to permit transmission of ultrasonic waves across the interface. When the tensile load is applied, the notch angle is increased to some value greater than a o and the output voltage level is correspondingly reduced from the no-load condition. As long as no cracks form, the voltage output at peak tensile load remains constant as shown in the region A-B. When tiny cracks (too small to reflect a significant amount of ultrasonic energy) are formed in the notch root, the notch angle can increm:e progressively with increas­ing crack depth. This, in turn, probably causes the level of the reflected ultrasonic energy to be progres­sively reduced (region from B to C). The initial decrease in output voltage, point B, can therefore be used as an indication of initial cracking in the speci­men. It is of interest that the voltage level used as an indication of initial cracking in this inve"tigation was considerably lower than that of point B, but higher than point C. Thus, by arbitrarily choosing a smaller voltage decrease than we did, it is probable that smaller cracks could be detected with this apparatus. It should be emphasized that the only requirement to establish the presence of a crack is that the decrease in output voltage from point B exceeds the minor fluctuations inherent in the instrumentation.

Continued crack growth affects the recorded voltage output as follows: the net output voltage is influenced by two opposing factors, (1) a del'1'easing amount of reflected energy from the notch, and (2) an increasing amount of reflected energy from the crack itself. Thus when point C is reached, the crack surface area is large enough to reflect sufficient ultrasonic energy to overcome the effects of the increasing angle at (Fig. 4l'1. The recorded output voltage beyond point C increases (in a nonlinear fashion) until the specimen fractures.

Nleasurement 0/ Large Cracks b.y Ultrasollic Method

The ultrasonic device was also used to measure cracks between 0.005 and 0.030-in. (0.13 and 0.76 mm) deep. To accomplish this, only one change was made in the system; an attenuation unit was inserted between the transducer and the reflected signal input to the amplifier (Fig. l'i. The attenuation was variable from zero to 42 decibels and permitted a ratio to be obtained between any two levels of reflected ultrasonic energy from the fatigue crack. The manner of obtaining this ratio can be illustrated by again making use of the schematic outP~lt-voltage trace of Fig. 4. In any test, the value of lowest output

1

1 t IJ\ Unl::d(f,r: ~pf·r.lr,pn. Ultra­

\onl(. ,',d','f'\ r~lh;r:\t{j trom

n(Jt~r jnl , .

(0) Cracht.'o \1J!'(ll11f'l1 in (or.)- 1.:1 ',; ·f-e }ff'CII~!?n In leflo:.ion. Lf'~""

Fig. 4-Schematic illustration of recorded voltage output and effects of cyclic loading and crack formation. ac < a" < at pression. :\10re ultraso!ll( I'.avh . "~;'_'nl( ',',C'.f:S reflf'f:tfd {rum notcll.

reflpcted fro[1l nDtch. little or ,.~JS rf't:f'ctfd by crad. none reflectprl by crac~. 0

Recorded voltage output ~· .. hen ~pecirnen at peak compressivp. load-7

~ E ____ ~ _____ --------~---- ---------F

o c ~

AppllHI IOllll cyclt,S

L Recor.:.:-: ',r;ltauf: output ',','hen speci--:-- at ~}f'-a~ t~nsile load

1(11 Level 01 vallely? outplJl at limr or ppa~, loac ~:. <~ti~r.

I point C I was arbitrarily designated as the reference for all subsequent output-voltage readings and was given unit designation. By inserting sufficient attenu­ation, the level of any subsequent recorded voltage output could be brought back to unity i i.e., the level recorded at point C). The number of decibels required to achieve this was noted. By reference to H table provided by the n;anufacturer of the attenu­ator, the ratio of output voltage to the reference voltage was determined. By stopping the fatigue test, breaking the specimen and optically measuring the size of the fatigue crack corresponding to this ratio, a relation"hip between crack depth or crack area and normalized voltage output was established. "When repeated for several different crack depths and corresponding voltage-output levels, a calibration C'urve for a given material was obtained. SuC'h cali­bration C'urves may be used in subsequent tests of the particular material to measure crack size without interruption of the fatigue test.

Optical Crack 111 ewwrement lor Calibl'CltiO/l Purposes

Upon first detection of a crack, some specimens were removed from the fatigue machine and sec­tioned longitudinally for microscopic examination. The sections were polished and etched to better define the crack. The image of the area containing the crack was projected on a metallograph screen at a magnification of 500, and the crack image was measured to the nearest 0.010 in. (to the nearest 0.00002 in. or 0.0005 mm of actual crack depth). Four crack-depth measurements were made on each specimen at points 90 deg apart and the average was used.

To measure deeper cracks (up to 0.030 in. or 0.16 mm) specimens were first broken in tension and the depth of the fatigue-cracked portion was measured at a magnification of 10 using a graduated eyepiece. The reported crack depths represent the average of 12 readings evenly spaced around the circumference of each specimen. If no crack existed at a particular measuring point, a crack length of zero was assigned to that reading so that all 12 points would be in­cluded in the calculated average depth. This was rarely necessary since, in most cases, the crack front progressed uniformly around the specimen circum­ference.

Results and Discussion

Fatigue Data

Fatigue curves showing cycles-to-initial detectable cracks and cycles-to-fracture are plotted in Fig. 5. The shaded circles represent those tests \\"hich were stopped upon initial indication of a L'l'ack. The specimens were then sectioned so that the crack depths could be measured as previously dl~slTibed.

The range of crack depths observed in these speci­mens for each material is indicated on the figure. These crack depths do not necessarily represent the smallest cracks that could be detected with this device. Instead, these crack depths correspond to an arbitrary voltage-output change \\"hich we selected early in the investigation as being indicative of the presence of a measurable crack :-:ee section on "Detection of Initial Cracks by Cltrasonic Nlethod"). As our experience with system operation and data readout increased, it became e\':dent that. smaller cracks could have been detected earlier in specimen life. Other tests were also run to obtain life-to-fracture. These are represented by the open triangles. The open circles represent the number of cycles-to-the-first detectable crack, as observed by the ultrasonic device, in the specimens that were run to fracture. For the particular specimen configura­tion used, cracks were detected within approximately 10 to 40 percent of life-to-fracture depending on material and cyclic stress. The dotted curves (Fig. 5c) represent life to various crack depths as observed by the ultrasonic method. The latter curns will be discussed in t.he next section of the paper.

Relation Between Output Voltage and Crack Depth

Tests were run with three materials using the ultrasonic method as a means of nondestl'uctively measuring crack growth. A series of tests was run for each material at a relatively low stress level to obtain significantly different crack sizes. The fatigue tests were stopped, the specimens broken in tension, and the cracks measured optically. These crack depths were related to output voltage observed at. the time each test was terminated. A gel'ond series of specimens run at a high stress level was removed from test when similar output voltages were ob­served. The details of the experiment.al techniques

Experimental JIechanics I 197

Fig. 5-Fatigue curves showing cycles to first detectable cracks and cycles to fracture for cylindrical notched specimens subjected to reversed loading. Theoretical stress con­centration.3.65

198 I May 1969

E Z

" ~.

~

, ~

414

2f6 ~-

IJ~, -

(Jl-

~27

6"

55z

414 .~

~.

27f) ~

1J, j

llf!:'r-

I 41,'~

:; ... ~

I bU,llJ)

411

.0

:~r

60

! t

60~

I.'.' ','J.(),!2Sq'.

'.,1)(,:,',,"

41j5~ I' i r· >.\11

Ihl 5 AI-2. ') 511-Ti ;,.110"

I 1..1 iliLJ-.- _L-~llL~. 1t! ~ w1

I , .. II-~

I It'. -,1),0012111,

l :::~'\. \

I , 1,1,1,1

::ll ,~ ;::[;L""J 'd"'~:" '7

6S9

552

414

2ib

135

100

so

60

40

il'_ ' .. :)klll.I02 wnl

C' •

2<l,O--"----'-LLLl~-"----L.l-LU-D:LLLL_LJ 1 101 lt1~ 103 IVl

Cj'clt',

I,:

employed have been described in a previous section. The results of these tests are shown in Fig. 6 for 2014-T6 aluminum, 5Al -- 2.5Sn - titanium and the mm'aging steel. Crack depth is plotted as a function of normalized voltage output. All curves begin with an output voltage of 1 for a crack depth of 0.003 to 0.005 in. (0.076 to 0.13 mm). This was the approxi­mate depth of crack when the reference voltage output was defined (see "Test Procedure" section). All other output voltages were measured in multiples of the reference value. The curves are terminated at slightly under 0.030 in. (0.76 mm) because specimens tested at the highest stress level fractured when the crack depth reached approximately this value. For each material, the curves obtained at a high stress and at a low stress are virtually identi<:al I maximum difference 0.002 in. or 0.05 mm I, indicating that voltage output for a given crack size is not affected by cyclic stress level. The figure indicates that although the output voltage is apparently unaffected by stress level, it is different for different materials. The ultrasonic crack-measuring apparatus, there­fore, should be calibrated for each material before attempting to use the device for monitoring crack growth.

The results of applying the ultrasonic technique in monitoring fatigue-crack growth are shown in Fig. 5c. Each dotted curve represents a particular crack depth. The data through which these curves are drawn were obtained by applying the ultrasonic device in conjunction with the calibration curves for the maraging steel. Figure 5c serves to illustrate how the ultrasonic technique may be used to nondestruc­tively measure increasing crack size as a function of cycles during a fatigue test. Such data can, in turn, be used to obtain fatigue-crack propagation rates. Thus, from a single test, it becomes possible with this technique to determine crack-propagation rates as well as life-to-fracture.

}.1 etallographic Observations

Photomicrographs of longitudinal sections of speci­mens tested to initial detection of cracks for each material are presented in Fig. 7. Cracks emanate from the root of the notch and progress toward the center of the specimen and are generally trans­granular. In most specimens, there were several crad:s which began at different points along the notch-root radius as shown in Fig. 7. This was most commonly observed in the 5Al- 2.5Sn - titanium alloy, the maraging steel and L-605. Cracks were also observed to emanate from the notch root at many points around the circumference of a given specimen. As the test progressed, these small cracks grew, some of them joining together to form the crack that eventually caused fracture.

When the applied cyclic stresses corresponded to those on the inclined portion of the fatigue curve (above the fatigue limit of the notched specimens), the main crack extended all the way around the specimen circumference and was generally uniform in depth until very rapid crack growth occurred just

:g '" > w u c 2 w

~ '0 '" ill

:::-

c

,-

~

":f G > :::J co.

0

6

2014-T6 Aluminum

4

2

a

6

5Al- 2.5 Sn-Ti

4

a

6

Maraging steel

4 (heat treated)

0 C rack depth, in.

I I a .25 .50

C rack depth, mm

0 0

C c:

I\\aximum str'ess, ::psi (:±MN/m2)

35 000 (241) 25 000 (172)

',"iiximum stress, =psi 1:b\\N/m2)

100 000 16891 60 000 (414)

I\\aximum stress, :':psi (:±iv\N/m2)

o 125 000 (862) o 75 000 (5171

.75 1.0

Fig. 6-Effect of crack depth on normalized output voltage of crack-detection device

prior to fracture. This is illustrated in Fig. 8, which shows fracture surfaces of maraging-steel specimens tested to various stages of fatigue damage. To obtain this sequential series of macrographs, the specimens were fatigue tested until cracks of a pre­determined size were indicated by the ultrasonic detection device. They were then removed from test and broken in tension. The fatigue-cracked portion

Experimental lVlechanics I 199

\ r

I"J)

"~\ 2014·TG

b SAI·2.5 S,,·Ti

,.'

i, v

, ,

Fig, 7-Micrographs of tested fatigue specimens of each material showing first detectable cracks emanating from notch root. Longitudinal sections

of the fracturL' surface is clearly distinguishable in the figurl'. It i" the lightly shaded area at the periph­er~' of the 0.;) in. 17,6 I11ml-diam te"t section. The clark. rough rcogion in the cconter of the specimen Tcopre>'conts the' tensile fracture. Thco fatigue crack is not a" apparent in specimens that were not te3ted beyond tilt' point of initial crad, detection (Fig. 8a and b,. However, microscopic examination showed that the nack extended around the entire circum­ference and that the depth varied between 0.001 and 0.00:3 in, \(),O~5 and 0.076 mml in both specimens. Reiatiw to the average crack depth of about 0.002 in. (0.050 mm). this represents a substantial varia­tion. HllwL'vL'r. for larger cracks (Fig. 8e to 8h), the crack de'pt h was relatively uniform around the entire circumference. Note that the normalized voltage output for each specimen is the same for a given size l'l'adi:, even though the number of applied c~'de>, and maximum cyclic stress le\'el applied were different. indiL'ating that the l'l'ack-detection system

200 I M(lY 1%9

I,e) Maraging steel; heat treated

rei L·605

is sensitive to crack area and independent of how the crack was formed.

It has been shown in Fig. 8 that the main fatigue crack generally progressed inward uniformly a'round the entire circumference of the notched specimens

Ma¥imum cyclic stress, ± 75 0Cf.I ~i t5111\\Nlm2)

,1' ';'iero3')e crack (lepth, 0.0013 inci11O,l)46 mm); aDph(-(llo~ ~·Icles. 1103; non'1alized output 'JoltJ'(Je, '.1.0,

-, :".",r,i',!! cr"ck depth, U,0105 incn ~O,267 f'W.l: applie-d 103£' _ .. CI£'$, 2,~3'~: normalizro output ',l)lt.i']i? 1.4,

1';1 4.\da.;e crack depth, O.OlSO inch (0.J57 mm); applied load c).:les. 34Q2; norma!i!eu cutput \1)ltiNi?, 1.~.

lSI :\vaa~(' crack dpplh, 0.024<1 inch 10.6.::"0 n'm); appllfdload c~cles, 4177; n0fJ'1aI12~ 0utP'.Jt \o1t;),J~, 1.6.

11' A','".:raj!:- uack depth, 0.0018 inr;h iO.fJ46 mm); applied load (\'cI€s, 141: normalized output voltage, '1.0.

Idl A,\f'~;-9"l (.(,10, deplh. O.Olu3 inch 10.262 mini: applied load ':F1"s, £1S5: norrn2lizr;d o'~t~ut '.,'Jltd~E, 1.4.

In Avprage ((a.:~ depth, 0,0187 inch m.475mm); appHed 1000 c!cl€'~, i50: normJllzed output voltage, 1.9.

Ihl AI€rag, crao:k deplh, 0.01Q7 inch 10.678 mml; appllw load cycles, 93&; normallzed output vo!t;!~e. 3,75.

Fig, 8-Fracture surfaces of maraging-steel specimens showing crack depths cor· responding to normalized out­put voltage of crack-detection system and applied cycles. Unetched

Experimental .jYlechanics I 201

,", Fatigue test stopped after 31,450 cycles

bl Fatigue tested to fracture 37,500 cyciesl

Fig. 9-Fracture surfaces of maraging·steel specimens te3ted near fatigue limit. Maximum cyclic stress, ±50,OOO psi (345 MN/m")

used in this investigation when the applied cyclic stresses were high. This was not the case when the maximum cyclic stress level was low, i.e., in the region near the fatigue limit. This may be seen in Fig. 9 which shows the fracture surfaces of maraging­steel specimens tested at ±50,000 psi (345 MN/m2). One of the tests (Fig, 9a I was stopped after operation for approximately 75 percent of the expected life-to­fracture. After removal from test, it was broken in tension. A second test \Fig. 9b) was run to fracture. It is apparent from the macrographs of the fracture surfaces that fatigue cracking did not occur around the entire specimen circumference. Instead, the

202 I klay 1969

fatigue crack, which is evidenced by the smooth light surfaces, progressed primarily from one side of the specimen (see arrows).

Summary of Results

An ultraEonic method was developed to detect and measure fatigue cracks, nondestructively during test, in circumferentially notched cylindrical specimens (K{ = 3.65) subjected to reversed axial-fatigue loading. The reft~ction technique was employed. Materials used were 2014-T6 aluminum, 5AI--2.5Sn­titanium, a maraging steel and a cobalt-base alloy, L-605. Fatigue curves were obtained showing cycles­to-initial detectable crack as well as cycles-to-frac­ture. The following major results were obtained.

1. The depth of initially detectable cracks, which generally extended around the specin:en circumfer­ence was approximately 0.0005 to O.OO.J in. (0.013 and 0.10 mm) for the materials used. By introducing additional refinements in operating techniques, smaller cracks can be detected with this method.

2. For the specimen geometry investigated, initi­ally detectable cracks were formed within approxi­mately 10 to 40 percent of life-to-fracture depending on material and cyclic stress.

3. Crack depths up to approximately 0.030 in. (0.76 mm) were reproducibly measured with the ultrasonic method. This suggests that this technique can provide a convenient means of monitoring fatigue-crack growth in notched specimens during test.

4. The output voltage vs. crack-depth curves were virtually the same regardless of cyclic stress level for a given material. However, these calibration curves were different for the different materials.

5. For all materials inv~stigated, the fatigue cracks were generally transgranular.

6. At stresses above the region of the fatigue limit (inclined portion of fatigue curve) the fatigue macl'O­crack front advanced uniformly around the specimen circumference until a point of rapidly accelerated crack growth was reached. At lower stresses, near the fatigue limit of the notched specimens, the crack that caused fracture usually propagated from one side.

References 1. Feinstein, L., and Hrub.", R. J., "Surface Crack Detection by..111eTo"

wave j\letlzods," presented at the Sixth Symposium on lVondesfruction Evaluation of Aerospace and lVeapons Systems CompOllents and J.llaterials, San Antonio, Tex. (April 17-19, 1967).

2. Klima, S. J., Leseo, D. J., and Freche, J. C., "A.pplication of Ultra­sonics to Detection of Fatigue Cracks," EXPERfMENTAL MECHANICS, 6, (3),154-161 (1966).

3. Rasmussen, J. G., "Prediction of Fatigue Failure Using Ultrasonic SUlfaee lVal'es," Jill. Soc. for Nondestructive Testing, 20, (2), 103-110 (1962).

4. Rublev, Ya. A., and DanilDu, S. Yu., "Ultrasonic Detection of Fatigue Cracks During Repeated Static Tests," Industrial Labol'atOl:v, 29, 1306-130.9 (April 1964).

6. Hinsley, Jo Fo, Non-Destructit'c Testing, J.lacDonald and El'ans Ltd., London (1959).

6. Carlin, B., Ultrasonics, ~lcGraw-Hill Booh Co., Inc., Neu: Y01'h, 2nd ed. (1960).

70 Banks, B., Oldfield, Go Eo, and Ruuxling, H., Ultrasonic Flaw Detection in AIdals, Iliffe Books Ltd., London (J.962).

80 Anon., "J.Ylanualon Fatigue Testing," ASTlY[ STP 1.\'"0.91, American Society for Testing and Materials, Philadelphia, 33 (19-19\.

90 J.lfanson, So S., "Inler/ace Between Fatigue, CI't:!ep, and Fractu.re," Internat. Jnl. Fmc. AIech., 2, (1),327-36.1 (1966).