Cumulative Damage in Fatigue
By MILTON A. MINER, SANTA MONICA, CALIF
‘The phenomenon of cumulative damage under repeated
loads was assumed (o be related to the net work absorbed
by a specimen. The number of loading cycles applied
expressed as a percentage of the number to failure at a
slven stress level would be the proportion of useful life
expended. When the total damage, ar defined by this
concept, reached 100 per cent, the fatigue specimen should
fail, Experimental verification of this concept for an
aluminum alloy, using different types of specimens, ¥
fous stress ratios, and vatious combinations of loading
cycles is presented, These data are also analyzed to pro-
‘ide information on different stress ratios when an S-N
curve for any one ratio ia known. Results of a sample
analysie based on experiments are given. It is con:
luded that a simple and conservative analysis is possible
‘using the concept of cumulative fatigue damage.
Inmaoouertox
A YROGRESS with methods of fatigue anslysis has been slow
1B. F. Zanger* presente! a method fo estimating the elects of
Toading cycles of differnt stross amplitudes in 1037, ‘The
‘propagation of a erack is related to a damage ourve, obtained 23
in-work at Battlle Institsto,or tothe start of @erac, Langer
also suggested certain experimenta to verify the proposed ealea
lation. “A discursiont of the work of Thum and Fedor, Bellen
rath, Nisen;-Snd Kaul indicated crends of thought. wat
eablshing a cemulative damage approach to fatigue failure
"Thum and Bouts alzo discussed somo of the agpects of this
problem.
Consideration of these various approaches led to an atlerapt
to simplify the various phases of the damago problem ia fatigue
Sinoe aircraft structures ars mainly made of alumina alloys,
this material was chosen forthe teats although allay stecls are
stil important, Also, the appearance of a crak in an sirplace
structural member has roquired repair or replacement befere
further operation was posible; thus a crack can be considered
‘asa failure in an aircraft structural component. 'To account for
the effect of atest eyes above sme arbitrary lower lit (forpAUGNECO win MIEHIOUs OF IBUgUe BMAIYSIS DAS Ooo SOW
because of lack of basi information on material behsvior
and, moreimportant, because no method of handling any
‘but the simplest problem has been available. Any solution r=
quizee several toole before other than simple comparative
‘sualysie based on tests can be made. (Bven in compare-
‘tivo tects the experiment ray easly be oversimplified to the ex-
tent that significant effects are complataly lost.)
"These tools must provide the folowing
1 nformstion on the loading conditions to be encountered
Coumber of eye, frequency of occurrence of gusts, et.).
‘2. Taformatioa regarding the number of oyeles to Tulare of tho
smatesial or part at various etre, jc, SAN carves,
‘3. Means for evalusting the cumulative effots of eyeles of
stress at various lresslovels,
‘4 A method of relating various loading ayes (1.0, eyshs of
treo with difrontratioa of mininsum to maize stress)
{tom 1 of this list iso gonoral problem of a magnitude bayond
the scope of the laboratory experimenter, the designer, or the
sivess gnalyet. ‘They must rely on service information where
availble or otherwise make conservative louding estimates.
‘This problem is important and mast be undertaken on a larger
seale than is posible inthe laboratory.
‘tem 2 ia obtainable experimentally in most eaams, and data
‘can be augmented with instrumentation of pars in service and
‘carefully prepared servis records
‘Tho principal purpose of this paper deals with item 8, # means
for evaluating the eamalative effects of cycles of stress at various
sees lovels, Item 4 is supplemental to item 3 and is also dealt
‘with since it is very necessary in most practical problems.
5 Stone Teat Kaginer, Dousan Alvaalt Compaty, In,
convenience say the stress at N’ = 16) a simple concept of damm
sige was frst considered. ‘This concept held that the damage
could be expresed in terms of the number of eyeles applied
divided by tho number to produce failure at a given stress lval
‘When the summation of these “increments of damage’ at several
stress Ievels bocame unity, failure occurred,
Asswurrions avo Leattasions oP Conse
‘The complexity of loadings enoountered in wireraft makes any
approach toa fatigue analysis npproximate at best. Therefore,
plifying assumptions must be made so that a start is possible,
‘To this end the following assumptions sre made:
1 The lading cycois sinusoidal.
2 The total amount of work that can be absorbed produces
failure (under the farther assumption that no work-hardening
‘ceurs).
3 The rlationship between various stress ratios (ratio of
‘minimum stress to maximum) i approximately as shown in
Figs 1a) and 10)
4 The Inception of « crack, when observed, is considera to
conetiuate fale,
‘These assumptions for thesuka of experimental brevity arwalza
subject to the following limitations:
41. Only aluminum alloys azo to bo stadod
2, Only maximum stresses abovo the steest at which fetigue
failire oscars st 10 eyoles azo to be coasidered.
+ “Patigu Paiiur From Stress Cycles of Varying Amplitude” by
B. Rr Lnoger, Jovusas or Avruise Mrcnaiey Tras ASAE
val. 89, 7p. AO.
ia ei a chs Mlb Mika Tic iat MsiPresented at mesting of the Avistion Division, Lor Angeles,
(a Sone 871915 f Tan Anca Socio oF Macuanont
DDiswarion of this paper should be addramed to the Ssorotary,
ASIN, 20 Wort hh Street, New York, Ne ¥., and wil bo ae-
‘apted ute Oct. 10, 1048.” Dissuasion vossived ‘afer the elosing
Site wll bo vataran
"Nom: Statomonts ard opinions adeaneod in papers are 40 be
‘understood as inividnal exprossions of their ashore snd ot those
ot the Sonat
+ “Prevention of the Failure of Metals Under Repeated Stress,”
Battle Merril Tnvutute Stall, John Wiley & Sota, Toe,
‘York, N.Y. 1911, pp. 92-06 and pp. 188-208.
‘Tia, 205,
* Diosiution of Langor's paper * by Thum snd W.Buuts, Joon
vat. op Apri Mueanios, Tras. ASME, val 60, 1938, p-
aia,
Buen of Strong of Materila" by S. Timoshenko and G.
FE: MacCulouoh D. Van Notrard Ca fats Now Yor N.Y 399,
ps
Ants
sin JOURNAL OF APPHED MECHANICS
a
Ne
rhe al
‘This exyreiniaetly demnonatratntl 10 lili love exnerieiasaiteaes ne eae eatin
Fra, 1 f@) Moptrime Gooostax Draciese
———
Fie, 1 () Pion ov Starse Viesus H Shown Aoneewaire Bee
‘twas Stanton Line Paow Fe. 1a) xo Tier Reauces
Nowavcusrues Rerenaey ro Pies. 2) ao 2(0)
= cithor maximum or minima stress
S, = ultimate static tonsil strength
Si; = maximum stres in a londing epele (Le, the stress of
‘eater magnitude rgardies of whether it is tension
‘or compression)
1, = number of stress eyeles to fail
at stress, 8,
22 CAYEUSION Is AeINONSETALED 10 hold Hue Experimenta
‘within the nite ofthe asuptions.
should he noted that che significant festuncof the emmudative
darnagr concept lias in the hypathees that failare occurs wit
22
In Fig. 8 are shown the typos of specimens eaployed in the
cxpenments. IL showkrbe notod that all specimens were made
from Alelad 245-7 sheet wnat
Fia.2 (6) Tenican SLY Conve Tusommustive Nowexetaromsim = nuunber of eyes applied at,
WY = work done to flare
my = work done at my evoles
‘Soin = minimum stross in given losding eyele (may be aty
‘value loss than 8, either tension or compression)
1W = ratio of minimum stees to maximum stress in lending
DeveLormmsr oF Dassen Coscarr
Voaetestioss
ap Exramncesrst
LEW represents the not work absorbed at failure, then
mn
wo
Piette Bat a1
and cubstivuting values of Equation [1]
1)
‘The uipmont wi oper show ast was etl the
‘ani tat shows in Pig, and inthe ne agama
Posie dyna led conel was mattaned trough th a
elcrcimatnce sin gag ene i san wih the te
spesinen and inating ‘mf Dunfont pe 308 Since
Citas lessor, Fig Sap 80).
‘hbase 2 ares archon in Fig, 7, Sissratio valu
, (AEE of 4080, 0.20; nt 020 carn
experimentally are preseted, Fg Siva mproduton of a po
Gon of Fi 7 wth reste pote for tree tea eto
irom Table. Fig Si pled Uitte grabaly te eae
ingeycles cpap an the arco of tel nae tho
‘snes downge coon
"Thomo of ptlng te cuultive eyes ia Fg, 9, and
10 is teed om the proportion, = At any ratio of Za per
®
contage of wscfal life has been consumed, This proportion is
ted to deturmino tho starting point of ne at @ diferent stres.
‘Thus specimen $19 was run at 89,800 pal for 135,000 eyeles.
‘ny _ 135,000
~ 29,000 eles, 2 oy
lower stres.of 29,000 psi, this was eruivalent to 0.55. Ne
'= 0.55 X 830,000 = 181,000 ey les whieh ithe point from which
vm sturis in Fig, 8. Now since ny = 141,000 eycles at failure,
me _ 141,000 A
200 = 0.45, and the value of 2 = 0488. ‘This same
opp 7 0A andthe valu of = = 0.98
Since = 0055, sind at the:LL
MINER-CUMULATIVE DAMAGE 1 FATIGUE sein
Fe. ay Puree Fyoue Spores
er neta Sal mt Na aera
Fis. 8 @) Dawene-Freen-Sousw Famave Soro
ie
Fro, 6 (@) Lave Discnas oF Tyetoxt Famous Macmtsi
ten | fe OPS]a
Fis. (0). Laws Disawas oF Loxo-Conraou Cincorr
Sy
5
5
i
Rio. 7 S:¥ Conves a7 Vamous & Vaexs vow 28. Aver
procedure is used to plot the other points serving st exemple
‘Table gis the reslts ofa number of experiments all of whicFro. # Axaicboine Farays Macme Srets-Gaae Rove
Nor Sows)
REAR
Fue. cress Tascs ov Seauss‘Tante Wave Fou Faout Osc1-10-
‘cura Reconons
Acts
JOURNAL OF APPLIED MECHANTSS
‘wore run ut {wo oF moro iross lovele at a streseretio of +0.20.
“Th ava tt vale or is 098, Te maxim an wi
mum values ave 1.45 and 0.61, If only the last oye of lout
fs plotted for these tle (assuming that only this final sts
cycling cased the fare), she average value of 2 is 0.87, thus
x
Indicating that the damage from the other loading eycles
be ignored
Fig. @ is representative of a more complex type specimen {in
this ease a dimpled fash serow Jolt) in which tho Su¥’ eurve lise
hon established and tivo Usts have been run at different str
levels,
{eshould be pointal out at this point that the author beliew=
‘that “‘stresemisers” or strostconoentrations are best handled in
SEPTEMBER, 1045,YANLE 1 MESULTS OF EXPRRIMENTS RON AT TWO OR MORR STW
STEVE
1008 x sce
at ° % ia ocr 10)
co ara RB 15 om WIS
i 8, 2 i
Ba
= Ge ho og & oo
RR 8B RB iendytetee
i, gee oe penn
ges os oa i knlgtem
i. Gen as ag if vest
2 2500 1330 3000, Odd Faltwe °
ogee —_
Se RE TT pty
co ge a wows
aa te Rak Teter
SHA (pam 18 a 3 7 Yee decrommed for
ee a ee
<< Gel ow oe
18 HE # RESiER
Gos Ne OSB ico
SEA i) 35000 2 HE ae * retcermedte
(3) 27500 320 420, 0.66 aa ee
“Ue oe og ae
o 2 8 gests
Nore: Stree rato (HRA BIER) oo. feral specimens
Specimen type Ais snad ln Fp Sa), and type Bin Fe, 30
Average eal of 2 = 098i Table
ee8
1 Saas
Fea. § Parmar, Reronvcvion ov Tio. 7, Grvmma 4 Gaapmoun
Parsewranon oF Comvuartvn Dawaen Concare ton Twice
‘Skene
{atigue by experimentally establishing S-N curves on specimens oF
parts which contain the actual concentrationé wherever pombe,
“Applying factors or eorzoations dow not seem feasible atthe
recent tase,
‘Mumioo o? Retna Vantovs Sranse Razioa
OEE SRE
Max. TENSION LOAD ON JOINT-I485)
Fie. § Loan-Creue Cosve rom "Yavin: FuomSeaew Daur
Jobe in 218-7 kes
characteristics of amateria or part ean be comparvd oven though
‘he information available hss bean obtained for varying ratios of
nninimum to maximum stress, Experiments with Alelad 248-7
‘sluminam-afloy specimens do not result in a straight line such ae
‘that in Fig. (a); however, the variation i not great. Tn Fig
1(), the lowerlimit curve, bounding the shaded afea, has heen
‘obtained from the modified Goodman diagram in Fig. 1a). This
‘useful in presenti a orase-plot of stress (ceprossed ara porsent-
tage of the ultima! tensile strength) against stress ratio R, for aified Goodman diner, Fig Ce}, the fatigue given mimber of cys, ‘The tppor eure of te shaded ea
MINER—CUMULATIVE DAMAGE IN FATIGUE. Aa
1X WITCH PERCENTAGE OF DAMAGE WAS I31Posco)
Ho 'Snb bab ChiaNceD UNTIL Pateone oucuMMED
“= gimp "gfe
S84 0.20 Bae my at 9.08 “ie Hope iporeeeed for co
SOA =o. 1) 9 48 oe 7° pes
BEG 8 8 OS
eB! easanis hal tered
essA 0.00 ys ta.HE BER OE
sw. 38 QE Ae
ar ae
so. 83 GER Be
son 8B BOB
to. 88 GER ORs
toe 8B GHB Be
so. 82 BER
‘Soom: Avecage von for = 1.05 Table 2
represents the maximum deviation of test values of various exe
perimeters including some unpublished data on Alclad 24ST
‘material.
"Ths i is apparent that the lower eurve is in fair agreement
sith experiments and gives conservative stress values. Also the
expense and simplicity of handling of pertienlar problems using
the straight line are greatly improved, since the large numbers of
tects necessary io establish several SN curves oan thus be
‘voided. For the purposes of this paper, S-V eurves at various
FR values have been experimentally eatablished, Fig. 7, 60 that
sianged: Sota Inrenced
3 ie
2 Rugg tect
a satay
8 Eafaye td cent
& aiey
ug Agee nos
& Sefagr at ert
“3 Saige nr
48 Baits fond
& Sefer trate
“
2 shaaaeds toad incr
fughsgrs toad nore
the snecimen (oeltig Uae of worchaneing) would
‘peat at et lace obo dprovedy cover previo
‘Meminum al However, data by lohan aad Oberg! when
reat vo anaverage rate of 2 f 1.08. ‘Thine done by
simply shifting the basic S-V curve slightly (entirely within the
rein of mea Cot vals.‘ sgl tt rn presed by
Benycanot be choked without seo Wh orepndng
5. eure; sve, ia ly hase 3% = 200 bod 00‘he accuracy of ti damage concep! can be ascertainod.
“Table 2 gives results of experiments in which a pereentage of
damage was imposed upon a specimen at one stress ratio then
jn meet easee tho strest ratio and load were changed and the
specimen tested until faire oecurred, In several ences, three
diferent tress ratios were employed both in the positive and
negative range (.e, both tensen-tension and tonion-compression
Toads). As in the case of tests rum ata constant stress ratio, the
‘value of 2 is taken directly from the S-N curves tho diforence
there being that S-N curves are required for each stress ratio.
ig, 10 pricents several ofthe resultefeom Table 2in graphieat
form. It is to bo noted that the average value of 2 is 1.05 or
there experiments, The maximum and minimom values aro 149
and 080,
Dascussion ov Resvas
Although only twenty-two tests are prevented in the resus of|
‘Tobles 1 and 2, othor date are available which further tend to
confirm the ress herein presented. ‘The concept that damage
in fatigue ie cumulative and is proportional to the work done on
the lowest stress for the materiel, a value of not ovor 1.50 would
be reasonsble if the mean strass value were used.
"The practice of overstreaing specimens for eome number of
cycles then running to failure (or “Tunaing out”) at the endorance
limit is usally used to prove the presence or laek of damage >
‘This procedure obscures the reults quantitatively in torms of the
cumulative damage concept because of the wide seattor of test
‘Points in the gion of tho enduranco limit, For this reszon no
‘aeourate pieturo enn be drawn a to the validity of the concept
for materials other than Alelad 266-7 aluminum alloy because of
lack of test data Some investigaiors have had individual vst
specimeas run beyond th S-V earvo during the “overstresing”
ran and sill not fail when “wun out” at the enduranoo limit
‘The normal scator in fatiguo-tsting, howevor, acount for this
behavior in opeasional specimens.
‘Tho accuracy of test reculls is pariculaly important in at-
lami o eta damage information, tal astra
ob
""Bheundon, Proordlge of ke ALTA, Oy G. W. Sickle, vol
48, 1985, p. 703 °A164
; frifine owe |)
5
i
Fic. 10. Parral Repgoouctox ov Fi. 7, Suowina Guat
Fuluve Pores soe ‘Temes Soscnacys Wen Skea, Sos
‘eemo (A), ano Nona op Creti Aiu Vans
JOURNAL OF APPLIED MECHANICS
SEPTEMBER, 1935,
[Nov rom au 8-3" env for the asain in question (rae hs
assumed foe thisexsnupe)
Ny = 1,000,000 excles 10 (suze at 834000 pal and # = + 0.30
Xy = "215,000 eyelet fun st 83.000 poi and He =f 020
Now
$4 B= 28 = porion of ale conned por minute
of operation, oF
2 8 os
Too x toe * aS sap 01 IP
and since falyep oestfe st 2
1, tho pressure unit ean be
pected to Fail ia
1
oa
340 br
00
‘continuous operation under these acoslersted conditions,
OF course for an actual case, there are several problems which
ariso to maks tho dotannination of fatigue life anything but
simple, such as the example tha is llustrated. ‘Tho deternins-
tion of the working stress and munabor of eyces of opsration is
usually a difeult mattor involving precise instrumentation and
caroful experimertal work or a vary earaul evaluation of serview
‘experience and‘ e other factors that can be determined. ‘Thon,in loading (which are not so likely in rotating-bow snsehines,
‘but are common in axial-loading types) will result i large varia
tions in the number of eyeles to fallure. Considering tho usual
‘srge seater, the results cbtainod inthe tast of Tables | and 2 re
felt to be both accurate and signiflcant. The average value for
2 hb 1018 forthe two cio The asin a
‘minimum values are 1.45 and 0.61, respectively. Taspestion of
‘an S.N curve on which ate plotied actual test points will show
that the scatter of thesp maximum and minimum valuoe is not
‘excosive™
Sum Avan”
‘The problem i o find the expected life in hours of a bydeaulice
pressure unit.
“Assume thatthe pressure veszel is aade of aluminum alloy and
‘operates from 100 psi to 3000 psi during part ofits operafion and
‘rom 600 to $000 pel during the remainder. Thasyatom oparates
‘ab 20 eycles por min under aocelerated text conditions and 80 per
‘eat of these operating eycas aro of the formes type. At 3000 psi
pressure, the mariraum tensile tras s known to bo 35,000 pa in
‘the erica region
Now
1300)
By = = $0.90
and
Als
‘ao, S-¥’ curves are roquired for dhe particular part in question.
Honovor, if tho extieal rogion of tho part is mown to have
negligible stress concentration, SV curves of the basic material
fam bousod. The modified Goodman diagram in Fig.
curves in Fig. 1(), provide a ueeful means of ech
sites ration at N= 10% but, of cour, test curves are ti
preferred for greatar accuracy. It should be nated that when
‘perimental curves are to be established, many aetusl eases are
‘ost eoouoaoslly handled with an S-¥ ourve of constant mean
stress (Le, with the stras ratio varying) rathor then x number of
‘SAN eurves for constant values of
Aa octal problem involving slumiowm hydraaio-preseire
accumulators has beon zuccesstully solved by the procedures
outlined. The aceumulators were made of forged 148-T slumi-
rum alloy for which basi meatorial SV eurvss wero available,
‘Test runs on the accumulators indicated that a relatively small
numberof eyeles ceourting wa lower stress rato but atthe same
‘maximum stress might be eaasing premature failures. Analysis
of tho various possibilities for fatigue failure in th diferent tpe
‘systems in which the accumulators were used indicated thst ene
syslam was entirely satisfactory while another was not. Actual
service experience proved the analysis to bo correct in both eats
A long history of satisfactory secvico existed in ono ease while
in the other, several fires had occtared ia a relatively sort
Coxerverona
“From the evidenco available at prseat, itis concluded that the
concept of eumulative damage:
1. Holds true for Alclad 248-P aluminun alloy aad probably
for tho other high-strength aluminum alloys.
2. Provides ¢ simple, practical (and eonservative) mene for
mnshapion fatsanlproltenn,‘8620 eyclea per main under anceerated text conditions and 60 per
‘oat ef theso operating cycles aro of tho former type. At 3000 psi
pressure, the maxim tensile stras fa Known fe 35,000 ps in
‘tho eitical rogion.
$0.50
and
coo
pn eo
Alwo .
Tu ste = 20 eyes per min
or effectively,
ay = 0.6 % 20 = 12 cyeles por min
m= 0.4 X20 = Boyde per min
TRet 8, Pa, 188, p. 2
‘makinum stress mignt be causing premature lauures. Anslysis
of the various possibilities for fatigue failure in the diferent type
systems in which the acramulatoes wore usod indicated that one
system was entirely satisfactory while another was not. Actual
‘service experience proved the analysis to bo corect in both eases
A long history of satisfactory secvies existed in one ease whilo
in the other, soveral failures bad occurred in a rolatively short
time.
Coxcuvsions
“From the evideuco availablo ak present, its oneluded that the
‘concapt of sumulative damage:
1 Holds true for Alclad 248-P aluminum alloy and probably
for tho other high-etrength aluminum alloys.
2 Provides a simple, practical (and eonservative) means for
analyzing fatigue problens
8 Should be experimentally investigatod for stels and other
‘materials to dotermine ies rango of usfulness.
Acxxowuinours?
‘The author wishes to acknowledge the part that Mr. RH,
(Christensen and Mr. P.T. Mahaffey played in the dovelopment
ofthis paper.