1、oZI-t_Z!iNASA TN D-960TECHNICAL NOTED-960EFFECTS OF CHANGING STRESS AMPLITUDE ON THERATE OF FATIGUE-CRACK PROPAGATIONIN TWO ALUMINUM ALLOYSBy C. Michael Hudson and Herbert F. HardrathLangley Research CenterLangley Field, Va.NATIONAL AERONAUTICS AND SPACE ADMINISTRATIONWASHINGTON September 1961Provid
2、ed by IHSNot for ResaleNo reproduction or networking permitted without license from IHS-,-,-Provided by IHSNot for ResaleNo reproduction or networking permitted without license from IHS-,-,-1QNATIONAL AERONAUTICS AND SPACE ADMINISTRATIONTECHNICAL NOTE D-960EFFECTS OF CHANGING STRESS AMPLITUDE ON THE
3、RATE OF FATIGUE-CRACK PROPAGATIONIN TWOALUMINUMALLOYSBy C. Michael Hudson and Herbert F. HardrathSDMMARYA series of fatigue tests with specimens subjected to constant-amplitude and two-step axial loads were conducted on 12-inch-wide sheetspecimens of 2024-T3 and 7075-T6 aluminum alloy to study the e
4、ffects ofa change in stress level on fatigue-crack propagation. Comparison ofthe results of the tests in which the specimens were tested at first ahigh and then a low stress level with those of the constant-stress-amplitude tests indicated that crack propagation was generally delayedafter the transi
5、tion to the lower stress level. In the tests in whichthe specimens were tested at first a low and then a high stress level_crack propagation continued at the expected rate after the change instress levels.INTRODUCTIONThe evolution of the fail-safe design philosophy in aircraft con-struction has pres
6、ented designers with a number of new design consid-erations. One of the most important of these considerations is theprediction of fatigue-crack propagation rates. A number of investi-gators have developed empirical expressions for predicting crack propa-gation rates by using the results of constant
7、-stress-amplitude fatiguetests. This work has been extended to include tests in which fatiguecracks were propagated at first one stress level and then another_ asa first step toward the study of effects of the variable-amplitudeloading to which aircraft are subjected. In separate investigations_Jenn
8、ey and Christensen (ref. i) and Schijve (ref. 2) found that highload cycles succeeded by lower ones produced delays in fatigue-crackpropagation. The present investigation was conducted to provide amore quantitative evaluation of the delay in fatigue-crack propagationin 2024-T3 and 7075-T6 aluminum-a
9、lloy specimens when these specimensare tested at two stress levels. These tests are referred to hereinProvided by IHSNot for ResaleNo reproduction or networking permitted without license from IHS-,-,-as two-step tests. The delay in crack propa_;ation was measuredby com-paring the results of the two-
10、step tests witL the results of companionconstant-amplitude tests.SYMBOLSNNcN2ScSI$2number of cycles from crack initiationnumber of cycles required to propagate crack to a given lengthat stress level in constant-amplitude testsnumber of cycles required to propagate crack to a given lengthat second st
11、ress level in two-step testsstress in constant-amplitude tests, ksiinitial stress in two-step tests, k_ifinal stress in two-step tests, ksiSPECIMEN PREPARATIO_IThe materials for these tests were takel_ from the special stocksof 2024-T3 and 7075-T6 aluminum alloys described in reference 3 andretained
12、 at the Langley Research Center for _atigue testing. Thetensile properties of the materials tested are given in table I. Thespecimen configuration used is shown in figure i. Sheet specimens12 inches wide, 35 inches long, and with a ncminal thickness of0.090 inch were used in this investigation. A 1/
13、16-inch-diameter holewas drilled at the center of each specimen a1_d a 1/32-inch-deep notchwas cut into each side of the hole with a thread impregnated with finevalve-grinding compound. The thread was dragon across the edge to becut with a reciprocating motion. A very gentle cutting process isinvolv
14、ed in making notches in this manner; ccnsequently the residualstresses resulting from cutting are believed to be small. The radiiof the notches were within 6 percent of 0.0C5 inch. The theoreticalstress-concentration factor for this configuration was computed to be7.9 by the method outlined in refer
15、ence 4.The surface area through which the crac_ was expected to propagatewas polished with No. 600 alundum powder to _acilitate observation ofthe crack. Fine lines were scribed on the ssecimen with a razor bladeto define intervals along the crack path. Nc stress concentration wasProvided by IHSNot f
16、or ResaleNo reproduction or networking permitted without license from IHS-,-,-expected as a result of these scribe lines as they were parallel to thedirection of loading._S_NG _C_SThree types of axial-load testing machines were employed in thisinvestigation. Fatigue machines operating on the subreso
17、nanceprinciple(ref. 5) were employed for tests in which the applied load did notexceed i0 kips. The loading rate for these machines was i_800 cpm. Thecycles counter read in thousands of cycles. A lO0,O00-pound-capacityhydraulic fatigue machine (ref. 4) was employed for tests in which theapplied load
18、 did not exceed 20 kips. This machine applied loads at therate of 1,200 cpm, and its counter read in hundreds of cycles. A120,000-pound-capacity hydraulic jack (ref. 6) was employedwhen theload was to exceed 20 kips. The jack applied load at the rate of 20 to50 cpmdepending upon the magnitude of the
19、 load. The cycles counterread in cycles.TESTPROCEDUREBoth constant-amplitude and two-step axial-load fatigue tests wereconducted. In the two-step tests the cracks were initiated and propa-gated to a desired length at one stress level and then propagated tofailure at another. Tests in which the high
20、stress cycles were appliedinitially will be referred to hereinafter as high-low two-step tests,and tests in which the low stress was applied initially will be referredto as low-high two-step tests. In the constant-amplitude tests, thefatigue cracks were initiated and propagated to failure at one str
21、esslevel.All the specimenswere clampedbetween lubricated guides similarto those described in reference 7 in order to prevent buckling shouldthe specimenbe accidentally loaded in compression and to prevent out-of-plane vibrations during testing. A minimumtensile stress of i ksiwas maintained in all t
22、ests.Loads were monitored continuously by measuring the output of astrain-gage bridge attached to a weigh bar through which the load wastransmitted to the specimen. The maximumerror in loading was i per-cent of the applied load.In all tests crack growth was observed through 30-power microscopes.In t
23、he two faster testing machines a stroboscopic light was employed sothat crack growth could be followed without interrupting the tests. AllProvided by IHSNot for ResaleNo reproduction or networking permitted without license from IHS-,-,-crack lengths were measuredfrom the center of the specimens. The
24、numberof cycles required to propagate the crack to each scribed linewas recorded so that the rate of crack propsgation could be determined.In a numberof the low-high two-step tests it was desirable to usean initial stress level of 6 ksi. Since th_s stress was so close to thefatigue limit for the spe
25、cimenconfiguratior, it was decided to initiatethe cracks at i0 ksi. The cracks were then propagated to the desiredlength at 6 ksi at which point the stress icvel was raised and the crackwas propagated.RESULTSAND DISCUSSIONCrack-propagation test results are summarized in tables II, llI,and IV for low
26、-high two-step tests, high-low two-step tests, andconstant-amplitude tests, respectively. The quantity “number of cycles“given in these tables and in the figures is the mean of the numbers ofcycles required to produce cracks of equal length on both sides of thespecimens.The results of tests conducte
27、d at constlnt-amplitude stress Scwere used as a reference, and all the two-step test data were comparedwith these results to determine the effects of the initial loading SIon subsequent propagation at a second stress level S2. This compari-son was made by plotting on the same figure the variation of
28、 cracklength with number of cycles for both the constant-amplitude tests andthe second portion of the two-step tests. T_e starting point for bothcurves was the crack length at the time of the change in stress levelsin the two-step tests. The difference between the two curves is ameasure of the effec
29、t of previous loading hi;tory.Figure 2 shows the plots of the variation of crack length withnumber of cycles under load for the low-high test series. Inspectionof the figure indicates that crack propagati)n at the second stresslevel was not generally affected by previous loading history in eithermat
30、erial.Figure 3 shows the same type of plots f_)r the low stress portionof the high-low test series. Comparison of _he curves for the con-stant and two-step tests indicates that crac propagation at thesecond stress level was significantly delayec_ in both materials as aresult of previous loading hist
31、ory. Similar results were obtained byJenney and Christensen (ref. i) and Schijve Jref. 2) in their multi-step tests. The delay in crack propagation _hich resulted from previousProvided by IHSNot for ResaleNo reproduction or networking permitted without license from IHS-,-,-i340loading in the high-lo
32、w tests is plotted against the second stress infigure 4. Examination of this figure reveals that for a given secondstress the higher the initial stress the greater the delay in crackpropagation. The probable cause of this delay in crack growth is theexistence of residual compressive stresses at the
33、tip of the crack atthe time of the change in stress level. It is believed that thesestresses were present as a result of the large amount of plastic defor-mation which occurred at the tip of the crack during propagation at thehigh stress level.It was also of interest to determine whether previous lo
34、adinghistory affected the rate of crack propagation once crack growth hadagain started at the second stress level. This determination was madeby a comparison of the number of cycles required to propagate the cracksequal increments in the high-low two-step tests and in the constant-amplitude tests wi
35、th Sc = S2. The interval over which this comparisonwas made began when the crack had propagated 0.i inch past the cracklength at which the stress levels were changed, and the interval extendedto the point at which the specimens failed. This comparison is shownin figure 5. The reference line shown on
36、 the figure is the locus ofpoints along which the test points would lie if the rates of propagationin the constant-amplitude and the high-low two-step tests were the same.The generally close proximity of the test results to the reference lineindicates that there is little difference between the rate
37、s of propa-gation in the constant-amplitude and two-step tests.In the high-low test series_ the lowest stress at which fatiguecracks would propagate in 107 cycles was 16 ksi. This stress wasconsiderably higher than the lO-ksi stress at which fatigue crackswere initiated and propagated to failure in
38、constant-amplitude tests.Thus_ it appears that the fatigue limit has increased. This resultindicates that specimens subjected to variable-amplitude loadings maynot be damaged by some stress cycles with magnitudes above the normalfatigue limit of the specimens.These results help to explain why the li
39、near cumulative-damagerule frequently produces erroneous estimations of the fatigue life oftest specimens. This rule assumes that damage accumulates at a rateequal to the percentage of life used at a given stress level. Thus,this rule cannot predict the observed delay in crack propagation andthe res
40、ultant increase in fatigue life.In several instances a great deal of bifurcation was observed atthe tip of the cracks following the transition to the second stresslevel. In the 2024-T3 specimen tested first at 30 ksi and then 16 ksi,for example, the crack which finally propagated to failure at 16 ks
41、istarted behind the tip of the crack produced by the 30-ksi loading(fig. 6). Figure 7 shows the crack tip after the change in stressProvided by IHSNot for ResaleNo reproduction or networking permitted without license from IHS-,-,-6levels for the 7075-T6 aluminum-alloy specimen tested first at 40 ksi
42、and then 16 ksi. It appears that residual s_resses producedby highstresses at the crack tip were sufficient to render other portions ofthe specimen more vulnerable to fatigue-crac_ growth at subsequentlower stresses.CONCLUSIONSComparisons of the rates of crack propa_ation in constant- andvariable-am
43、plitude fatigue tests support the following general con-clusions:i. When the initial stress level was higher than the second,crack propagation at the second stress level was delayed. It was alsoobserved that for a given second stress leve the higher the initialstress the greater the delay in propaga
44、tion. The probable cause ofthis delay was the presence of residual compressive stresses at thetip of the crack which result from plastic d_:formation near the tip ofthe crack during propagation at the initial _tress level.2. Once crack propagation had commenced in the second step of thehigh-low test
45、s, the propagation rate quickly approached that ofconstant-amplitude specimens tested at the s_me stress level and con-taining cracks of equal length.3. Cracks propagated at the normal rate during the second stresslevel in the specimens tested first at a low and then a high stresslevel.4. The fatigu
46、e limit of specimens tested at first a high and thena low stress level was increased following tYe application of the ini-tial loading.5. The results of these tests help explain why the linearcumulative-damage rule is often in error. T_is rule assumes thatdamage accumulates at a rate equal to the pe
47、centage of life used ata given stress level, and thus cannot predict the observed delay inpropagation and the resultant increase in fatigue life.Li340Langley Research Center,National Aeronautics and Space Administration_Langley Field, Va., July Ii, 1961.Provided by IHSNot for ResaleNo reproduction o
48、r networking permitted without license from IHS-,-,-7REFERENCESi. Anon.: Discussion in New York by W. W. Jenney and R. Christensen(Santa Monica, Calif.). Proc. Int. Conf. on Fatigue of Metals(London and New York), Inst. Mech. Eng. and A.S.M.E., 1956,pp. 859-861.2. Schijve, J.: Fatigue Crack Propagat
49、ion in Light Alloy Sheet Materialand Structures. Rep. MP.195, Nationaal Luchtvaartlaboratorium(Amsterdam), Aug. 1960.3. Grover, H. J., Bishop, S. M., and Jackson, L. R.: Fatigue Strengthsof Aircraft Materials. Axial-Load Fatigue Tests on Unnotched SheetSpecimens of 24S-T3 and 75S-T6 Aluminum Alloys and of SAE 4130 Steel.NACA TN 2324, 1951.4. McEvily, Arthur
