STP 338-1963
STP 338-1963
STP 338-1963
OF AIRCRAFT STRUCTURES:
LOW-CYCLE, FULL-SCALE,
AND HELICOPTERS
Presented at the
FOURTH PACIFIC AREA NATIONAL MEETING
Published by the
AMERICAN SOCIETY FOR TESTING AND MATERIALS
1916 Race St., Philadelphia 3, Pa.
BY AMERICAN SOCIETY FOR TESTING AND MATERIALS 1963
Introduction—H. F. Hardrath 1
Low-Cycle Fatigue
Low-Cycle Axial Fatigue Behavior of Mild Steel—J. T. P. Yao and W. H. Munse. . 5
The Effect of Mean Stress on Fatigue Strength of Plain and Notched Stainless Steel
Sheet in the Range from 10 to 10' Cjxles—W. J. Bell and P. P. Benham 25
Low-Cycle Fatigue of Characteristics of Ultrahigh-Strength Steels—C, M. Carman,
D. F. Armiento, and H. Markus 47
Low-Cycle Fatigue of Ti-6A1-4V at - 4 2 3 F—R. R. Hilsen, C. S. Yen, and B. V.
Whiteson 62
Low-Cycle Fatigue Properties of Complex V\'elded Joints of High Strength 301,
304L, 310, and AM-355 Stainless Steel Sheet Materials at Cyrogenic Tem-
peratures—J. L. Christian, A. Hurlich, and J. F. Watson 76
Effect of Stress State on High-Temperature Low-Cycle Fatigue—C. R. Kennedy. . 92
INTRODUCTION
BY H. F. HARDRATH1
B Y J. T. P. Y A O 1 AND W. H. MUNSE2
SYNOPSIS
NOTATIONS
Ac, Cross-sectional area at the test section qc\, Plastic true precompressive strain, per
of the specimen after precompression, cent
sq in. qf , Plastic true strain at fracture in
Af , Cross-sectional area at the test section simple tension, per cent
of the specimen at fracture, sq in. <7max , Cyclic maximum plastic true strain,
A0, Original cross-sectional area at the per cent
test section of the specimen, sq in. 9min , Cyclic minimum plastic true strain,
Ar, Cross-sectional area at the test section per cent
of the specimen, re-machined after qm , Plastic true mean strain, per cent
precompression, sq in. qa , Plastic tensile true strain at fracture,
C, A constant per cent
Dc, Diameter at the test section of the R, Absolute-strain ratio, cyclic minimum
specimen after precompression, in. plastic true strain to cyclic maximum
df , Diameter at the test section of the plastic true strain, qm-,n/qm&x
specimen at fracture, in. r, Relative-strain ratio; cyclic com-
d0 , Original diameter at the test section of pressive change in plastic true strain
the specimen, in. to cyclic tensile change in plastic true
dr, Diameter at the test section of the strain, Aqc/Aqt
specimen, re-machined after precom- S, Engineering stress, psi
pression, in. Aqe, Cyclic compressive change in plastic
i, Number of applications of tensile load true strain, per cent
m, Empirical parameter obtained from Aqt , Cyclic tensile change in plastic true
slope of log Aet versus log N diagram strain, per cent
«, Number of applications of tensile load Aqa , Cyclic tensile change in plastic true
prior to fracture strain at w = 1, per cent
N, Number of cycles to failure
Aec, Cyclic compressive change in plastic
q, Plastic true strain, per cent
engineering strain, per cent
1
Assistant Professor of Civil Engineering. Aej, Cyclic tensile change in plastic engi-
University of N e w Mexico, Albuquerque, N . neering strain, per cent
Mex.
2
Professor of Civil Engineering, U n i v e r s i t y At a , Cyclic tensile change in plastic engi-
of Illinois, TJrbana, 111. neering strain at n = 1, per cent
During the past two decades, an in- On the basis of a literature review, a
creasing amount of information on the general hypothesis was developed to de-
low-cycle fatigue behavior of metals has scribe the cumulative effect of various
been published (1,2).^ Nevertheless, these types of plastic strain cycles on the low-
studies have provided data of only lim- cycle fatigue behavior of metals. Special
ited scope. tests on three mild steels, ABS-C* as-
Since in low-cycle fatigue tests the rolled, ABS-C normahzed, and a rimmed
applied loads are generally high enough steel, were then carried out to verify the
to cause plastic deformation and a cor- hypothesis. In addition, hmited correla-
responding hysteresis in the stress-strain tions were made with published test data
relationship, limits of either load or de- from other types of low-cycle fatigue
formation are usually maintained con- tests on 2024 aluminum alloy to indicate
stant in any particular test. Low-cycle the possibilities of extending the applica-
fatigue tests are therefore further iden- tion of this hypothesis to metals other
tified as either constant-load or con- than mild steel.
stant-deformation tests.
LOW-CYCLE FATIGUE HYPOTHESIS
In general the results of constant-load
low-cycle fatigue tests are presented in In 1912, Kommers (3) concluded from
the form of conventional S-N curves, a series of cyclic bending tests that the
where S and N are respectively the nom- magnitude of the cyclic deflection was
inal stress or stress range and the cor- an important factor in low-cycle fatigue
responding life of the specimens. Al- studies. Orowan (4) suggested that the
though the shape of a typical S-N curve following expression be used for cyclic
can be qualitatively described, it is diffi- strain tests:
cult to make a precise analysis for this
iV(Ae,) = C. (1)
type of test. On the other hand, results
of constant-deformation low-cycle fa- This relationship has served as a basis
tigue tests have consistently shown a for most of the hypotheses that have
linear relationship between the tensile since been developed.
change in plastic deformation and the On the basis of the results from re-
number of cycles to failure on a log-log versed-strain tests, Manson (5), as well
basis. as Gross and Stout (6), empirically mod-
Empirical relationships have been de- ified Eq 1 to the following form:
veloped to describe the effect of fully
iV"(A€,) = C. .(2)
reversed cychc strain on the low-cycle
fatigue life of metals. However, these Then Coffin and his associates (7-Q)
hypotheses do not portray adequately found that straight lines with a slope of
the data from all types of low-cycle fa- approximately —0.50 (t» = ^) best fit
tigue tests. their extensive data. Moreover, to cor-
The objectives of this investigation relate the results of cyclic strain tests
were to study in a general manner the and those of simple tension tests. Coffin
cumulative effect of the changes in plas- presented the following equation:
tic deformation on the low-cycle fatigue
behavior of axially loaded steel speci- IV"HA€,) = - . .(3)
mens and to develop a more general 2
low-cycle fatigue hypothesis. Many constant-deformation tests con-
ducted on various metals have shown
' T h e boldface n u m b e r s in parentheses refer
to t h e Ust of references a p p e n d e d t o t h i s p a p e r . ' American B u r e a u of Shipping designation.
YAO AND MUNSE ON LOW-CYCLE AXIAL FATIGUE OF MILD STEEL
Material"
Mang- Phos- Chro- Alu-
Carbon anese phorus Sulfur silicon Copper mium Niciiel minum
C N a n d CA steels 0.24 0.69 0.022 0.030 0.20 0.22 0.08 0.15 0.034
E-steel 0.21 0.34 0.019 0.030 0.01 0.18 0.12 0.19 0.003
" CN, ABS-C normalized steel; CA, ABS-C as-rolled steel; and E, rimmed steel.
instantaneous diameter at the critical true strain of zero. After the compression
section (Fig. 1). loading, the test section in the precom-
pressed state was enlarged and had a
Description of Tests: new diameter, dc, and a corresponding
In Fig. 1 two types of C-2 specimens plastic true strain of qd . At this stage,
are shown. The conventional threaded some of the specimens were re-machined
specimens were used first, but when to approximately the original size and
specimens with large precompressive shape. These specimens with a new
Virgin State
^
Q.
^r-'^k'-^'^i
r Re-mochriing
At Frocture
S
55
Q. iSl,
Q- q.
"tl^
A = 2—. n ^
'"A. d. "f^
's.
%-'"A^'^'"d^Sr'"^'^'"d;
strains failed in the threaded section, the diameter, dr, are assumed to possess a
specimen with flat ends was adopted to plastic prestrain of ^^i • The specimens,
provide more bearing areas at the ends. either in as-compressed or in re-machined
In testing these latter specimens, the condition, were then loaded in tension to
tensile forces were transmitted through fracture. The specimen diameter at the
pin connections and compressive forces fractured section, d/, was used for the
were applied on the flat ends. computation of the plastic true strain at
The strain calculation procedure for fracture, ga , and the tensile change in
"one-cycle" tests is illustrated in Fig. 2. plastic true strain, Agn , with the re-
In the virgin state, the test section had lationships shown in Fig. 2.
an original diameter of da and a plastic All one-cycle tests were conducted on
10 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
C-2 type specimens. To prevent buckling The subsequent tension tests were
of the test section at extremely high conducted in the same testing machine
compressive loads, a special guide and using the same fixtures. After the
assembly was used (Fig. 3). The whole specimen failed, the diameter at the
assembly was placed in the testing fractured section was again measured
machine, and then a dial-type diameter with the optical diameter-measuring
gage was manipulated through the device.
key-hole of the "sleeve" to measure the Cyclic strain tests were conducted at
minimum diameter of the specimen. constant relative-strain ratios of —J,
n = 3 ( Three-cycle Test)
n = 4 (Four-cycle Test)
n = I (One-cyde Test)
n = I (One-cycli
Test)
2V 3 4 5 i
i.i.
n = 5 ( Five-cycle' Test )
r =-l J i- = - l
n= I (One-cycle Test] n = 4 (Four-cycle Test)
H Failure
FIG. 5.—One-Cycle and Cyclic Strain Tests with Various Relative-Strain Ratios.
specimen to measure the change in its plotted in Fig. 7 for each of the steels
diameter. The electrical output of the tested. From these figures, the tensile
gage and that of the load dynamometer change in plastic true strain at w = 1
of the fatigue machine were recorded on may be obtained for any relative-strain
an X-Y recorder. This record was used
ratio. For example, in Fig. 7(c) the
for control purposes in the conduct of
the tests. Typical stress-diameter dia- values of A^^ are 66 per cent and 75 per
grams selected from the record of test cent for relative-strain ratios of — 1 and
C-2-CN522 are shown in Fig. 6. _ i , respectively.
12 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
100 XIO'
Specimen No. C-2 -CN 5 2 2
r » - l , Aq, • 5 % , r = 130 / /I
" /I r\ ^ f ( ( / / [1
40
~l_ o o o o
i£
1
'1-
II - II II
II
rocvj
II II
- ^
t
-20 _ _ „ _ — -—
-40 —
, -
/ , 1
1
-80
- / / uzy
/ / /
/
/
j
/ y / y
-100
- Plastic Chonge n Specimen Diameter, i^d= 0.012-in. 1/ ^
J90
FIG. 6.—Typical Stress Versus Diameter Diagrams.
1
(a) c
"o
100 ^1 lOOK
\- •••tf-- "?! i
--k
80 :^
. 60
- 6 0 ^ 8 0
Plastic True P r e c o m p r e s s i v e Strain < \\< per.cent
100
100
Cyclic Strain Test: Aqt versus log n plots but with various
The hypothesis presented here for slopes. To verify this observation 35
constant relative-strain ratios suggests specimens were tested in cyclic strain
that linear relationships exist for log tests at constant values of r. The results
•| ip "iOC Two
Number of Applications of Tensile Load Prior to Fracture, n
100
\
N (b)
hN
^r =-!
\„
r •• -
10
^
v \ .
I V
V r =- 1 ^^
10 100 1000
Number of Applications of Tensile Load Prior to Fracture, n
100
^,
"<x (c)
--r = - |
^
1
r • — •
5"
10
\ "^ "^
o
O^
o r =- 1
10 100 1000
Number of Applications of Tensile Load Prior to Fracture, n
100
10
of these tests are plotted in Fig. 8 for row, three CA-steel specimens are shown
the three steels tested. These data may after being tested at relative-strain
be further combined by dividing all ratios of -0.25, -0.50, and -0.75.
Aqi values by the corresponding values These gave lives of 10, 14, and 17 cycles
of Aqa . Figure 9 is a diagram of this respectively. All six of these specimens
normalized cyclic tensile change in exhibited cup-and-cone type of fractures.
plastic true strain plotted against « on a In the bottom row of Fig. 10 are shown
log-log scale for all three steels tested. specimens of the three steels tested with
All of these figures indicate that straight r = — 1. These specimens gave lives of
lines with slopes varying with r ratios more than 260 cycles and failed with
2.0
1.8
TL=l-0.86r- /
Ttr 1.6 V
/
/
1.4
x'
7^
/
1.2
/
^'
/
/
/
1.0 /
-0.2 -a4 -0.6 ^T
Relative-Strain Ratio, r
FIG. 11.—Variations of l/m with Respect to the Relative-Strain Ratio, r.
fit the test points quite well and that propagating fatigue type cracks. There
there does not seem to be any effect of was evidence of numerous surface cracks
material on the slopes of these relation- on the specimens, thereby demonstrating
ships. that these specimens were close to failure
A group of the fractured specimens at a number of locations.
is shown in Fig. 10. In the top row,
three one-cycle test specimens are pre- Analysis of Test Results:
sented, one for each material. Vertical Evans (22) obtained a constant true
cracks often appeared on the surfaces of strain at fracture in his repeated tension
the E-steel specimens when large com- tests, regardless of the number of cycles
pression loads were employed, but only applied prior to fracture. In the program
on the E-steel specimens. In the second reported here, it was observed that for a
18 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
group of C-2 type CN-steel specimens C-2 type CN-steel specimens were tested
subjected to various amounts of repeated in reversed-load low-cycle fatigue tests.
tension, regardless of the number of The plastic true strain history for each
cycles of tensile load before fracture, the of the specimens is shown in Fig, 12. By
final value of plastic true strain at frac- evaluating each strain-cycle of the tests
ture was more or less a constant for low- and summing, the quantity
cycle tests. Therefore, it would seem
reasonable to assume that for tests with
repeated tension only (r = 0), the cyclic i~i [\Agn/ J.
M i n i !"i I T'l
4
-I
=1
si
til
VXi
^^//*
.K,OT///^**
Note- Solid Points Show Tests With
Initial Load Applied in Tension
11 i I i j . ' 11 ' I J - ' ' ' ' ' J - ' '' ' ' L'' *-''
li 18 ^T ^Sg ^SS^ ^ 46
Number of Applications of Tensile Load, i
tensile change in plastic true strain is for these tests was found to vary from
linearly cumulative, that is i/m = 1. 0,94 to 1.08, which was close to the
The slopes of the four lines in Fig. 9 value 1.0 presented in the hypothesis.
may be described in the form of Eq 6
and give values of l/w equal to 1.22, CORRELATIONS WITH OTHER DATA
1.43, 1.65, and 1,86, respectively, for Data in the literature are generally
r values of —J, — j , —f, and —1. When reported for low-cycle strain tests con-
these corresponding values of 1/m and r ducted by cycling the specimen between
are plotted (Fig. 11), the following rela-
a constant maximum plastic strain
tionship is obtained; (jmax or €max) and a constant minimum
plastic strain {qmm or tmin)- In most
1 - 0.86f. .(8) cases, the tests were started with a
tensile load to produce the upper or
With the empirical relationships for maximum strain limit which was fol-
both l/m and Aga with respect to r, the lowed by fully reversed strains. Some
hypothesis is now complete. tests were carried out with constant
To verify the general hypothesis six absolute-strain ratios (R = constant)
YAO AND MUNSE ON LOW-CYCLE AXIAL FATIGUE OF MILD STEEL 19
+^«-4sr
9m a
1.0.. (9(a))
For the subsequent cycles (i > 1), 34
20 SYMPOSIUM ON FATIGUE or AIRCRAFT STRUCTURES
a.
E
E
-100
s EqiKd
55
-10
10 100 1000
Number of Applications of Tensile Lood Prior to Fracture, n
10
•Eq IO(e
^ o
1.0
^ 0
^^\^o
o Test< i by Dubiic(l3g
3
(Test points ore plotted on tti( basis
of the plastic er gineering St ain meaiured at 1/2)
m
0.1 SAEIO 30 Steel
10 100 KDOO 10,000
Nurrber of Appllcotlons of Tensile Load Prior to Fracture, n
FIG. 16.- -Correlation of Hypothesis with Tests Conducted on the Basis of Engineering Strains.
22 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
For q„ + 13.5,
y max to be approximately equal, we
38 then obtain.
5iTiax 13.5 1.S6
+ («-!) 19
= 1.0 -(10(6))
1)
I tmaX I
1.0... (10(e))
44
terms of ^min and n, along with the cor- 2. For constant relative-strain ratios,
responding test data from the same r, a linear relationship exists between
reference. Excellent correlations are log Aqt and log n.
again obtained. 3. There does not seem to be any
Dubuc (13) tested SAE 1030 steel effect of material on the slope of the
specimens with a gage length of 1 in. relationship between log Aqt and log n.
These tests, with «„ = 0, were con- 4. A linear relationship was found
ducted by controlling the total engineer- also to exist between the quantities
ing strain range. T o apply the present 1/w and r, which for mild steel may be
hypothesis, Atn was assumed to be expressed as 1/w = 1 — 0.86r.
constant and equal to the elongation at 5. Plastic strain-histories obtained
static tensile fracture, 44 per cent in this from nine reversed-load cyclic tests
case. From Eq 10, assuming Smax and were analyzed in terms of the general
YAO AND MUNSE ON LOW-CYCLE AXIAL FATIGUE OF MILD STEEL 23
hypothesis. It was found that the condi- under sponsorship of the Ship Structure
tions specified by the hypothesis were Committee, National Academy of Sci-
satisfied. ences, through the Bureau of Ships,
6. Correlations of the general hypothe- U. S. Navy. However, the opinions
sis with various published test data indi- expressed in this paper are those of the
cate that the hypothesis may be appli- authors and do not necessarily represent
cable to cyclic strain tests on metals those of the Ship Structure Committee
other than mild steel. However, addi- or its member agencies. The investiga-
tional confirmations of this correlation tion is a part of the structural research
would be desirable. program of the Department of Civil
7. Under low-cycle fatigue conditions, Engineering, of which N. M. Newmark
any tensile change in plastic strain is is the Head.
cumulative, and the manner in which The authors wish to express their
this accumulation takes place is de- sincere appreciation to R. J. Mosborg,
pendent upon the amount of compressive V. J. McDonald, and S. T. Rolfe for
plastic strain in each cycle. their helpful suggestions in many phases
of this research. Special acknowledgment
A cknowledgment: is due D. F. Lange, W. F. Wilsky, and
The work described in this paper was others in the laboratory shop for their
conducted in the Structural Research excellent workmanship in making speci-
Laboratory of the Department of Civil mens and maintaining the test equipment
Engineering at the University of Illinois, in this program.
REFERENCES
(1) P. P. Benham, "Fatigue of Metals Caused (8) L. F. Coffin, Jr., "The StabiUty of Metals
by a Relatively Few Cycles of High Load Under Cyclic Plastic Strain," Journal of
or Strain Amplitude," Metallurgical Re- Basic Engineering, Series D, Vol. 82, No.
views, Vol. 3, No. 11 (1958). 3, p. 671, Sept., 1960.
(2) J. T. P. Yao and W. H. Munse, "Low- (9) J. F. Tavernelli and L. F. Coffin, Jr., "A
Cycle Fatigue of Metals—Literature Compilation and Interpretation of Cyclic
Review," Welding Journal, Research Sup- Strain Fatigue Tests on Metals," Transac-
plement, Vol. 41, p. 182s, April, 1962. tions, Am. Soc. Metals, Vol. 51, p. 438
(3) T. B. Kommers, "Repeated Stress Test- (1959).
ing," Vlth Congress, International Assn. (10) P. P. Benham and Hugh Ford, "Low
for Testing Mats., New York, N. Y. Endurance Fatigue of a Mild Steel and an
(1912). Aluminum .Alloy," Journal Mechanical
(4) E. Orowan, "Stress Concentrations in Engineering Science, Vol. 3, No. 2 (1961).
Steel Under Cyclic Loading," Welding (11) R. D'Amato, "A Study of the Strain
Journal, Research Supplement, Vol. 31, Hardening and Cumulative Damage
Behavior of 2024-T4 Aluminum .\lloy in
p. 273 (1952).
the Low-Cycle Fatigue Range," WADD
(5) S. S. Manson, "Behavior of Materials
TR 60-175, April, 1960.
Under Conditions of Thermal Stress,"
(12) D. A. Douglas and R. W. Swindeman, "The
A'^C.4 TN 2933 (1953).
Failure of Structure Metals Subjected to
(6) J. H. Gross and R. D. Stout, "Plastic Strain CycHng Conditions," Am. Soc.
Fatigue Properties of High-Strength Mechanical Engrs., Paper 58-A-198 (1958).
Pressure-Vessel Steels," Welding Journal, (13) J. Dubuc, "Plastic Fatigue Under Cyclic
Vol. 34, p. 161s (1955). Stress and Cyclic Strain with a Study of
(7) L. F. Coffin, Jr. and J. F. Tavernelh, "The the Bauschinger Effect," Ph.D. Thesis,
Cyclic Straining and Fatigue of Metals," Ecole Polytechnique, Universite de Mon-
Transactions, Metallurgical Soc, Am. Inst. treal, Montreal, Canada, Jan., 1961.
Mining, Metallurgical, and Petroleum (14) A. Johansson, "Fatigue of Steels at
Engrs., Vol. 215, p. 794, Oct., 1959. Constant Strain Amphtude and Elevated
24 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
T H E E F F E C T OF M E A N STRESS ON F A T I G U E S T R E N G T H OF P L A I N
AND N O T C H E D STAINLESS STEEL S H E E T I N T H E
R A N G E F R O M 10 TO 107 CYCLES
B Y W. J. B E L L 1 AND P . P . BENHAM 2
SYNOPSIS
endurance is used to express cyclic life- predicting notched behavior under mean
times leading to both types of failure. stress, it was hoped that the results
would indicate various trends of be-
Stress concentration and mean cyclic
havior at low and high endurances.
stress are two important aspects of
metal fatigue that have been studied
TEST EQUIPMENT
extensively in the past, both separately
and in relation to each other. This paper Tests in the range from 5 X 10' to
is concerned with the eSects of local 10' cycles were conducted on a 6-ton
and general plastic deformation at a Haigh axial-load fatigue machine. This
stress concentration in relation to the operates on the principle of electro-
mean stress imposed under axial-load magnetic excitation of a resonant spring
cycling conditions over a wide range of and mass system at a frequency of
endurance. Yielding can be caused by 3000 cpm. The required full load can be
various combinations of mean and set up on the test specimen in about 15
alternating stress, the extreme cases sec.
being a large alternating range of stress Tests in the range from 10 to 10' cycles
superimposed on a zero or low value of were carried out on a 6-ton Schenck
mean stress, or a high mean stress, which
in itself causes yielding, plus some alter- TABLE I.—CHEMICAL COMPOSITION.
nating component. Both generally may Element Per Cent
result in failure in a relatively few Carbon 0.10
cycles, but so far more attention has Silicon 0.68
Manganese 0.75
been paid to the fatigue-limit region. Sulfur 0.019
Several investigators (1-3)' have studied Phosphorus 0.018
high mean stresses with stress concen- Chromium 18.25
Nickel 9.76
tration for long endurances, and a Titanium 0.70
theoretical method of predicting a
notched 5a-S„ diagram to allow for axial-load fatigue machine incorporating
yielding under high mean stress has a low-frequency mechanical drive of
been proposed (4). approximately 10 cpm. This machine is
described in more detail elsewhere (6).
The information so far available at low
endurances which deals with notched Some of the tests on the Schenck
material is generally only related to machine were instrumented to obtain
one or two mean stresses {R = — 1 or 0) stress-strain records (hysteresis loops)
(S-7). automatically during the life to fracture.
The machine is fitted with a dynamome-
lUg (8) has studied the effect of three
ter loop, the deformation of which is
mean stresses on notched aluminum and
proportional to load. One variable-
steel alloys over a range of endurance
inductance probe was mounted across
from 2 to 10' cycles. The present program
the dynamometer to record the load,
has studied stress ratios from R =
while a second was attached by exten-
— 1.0 to +0.91 for endurances from 10
someter clamps to a gage length on the
to 10' cycles for stainless steel sheet,
specimen. The signals from the probes
plain and notched with several circular
were fed into amplifiers which gave a
holes. Although these are insufficient
visual reading of displacement, and then
data to propound a general law for
to the X and Y plates of an oscilloscope.
Permanent records were obtained by
' The boldface numbers in parentheses refer
to the hst of references appended to this paper. photography from the latter.
BELL AND BENHAM ON ErPECT OF MEAN STRESS ON FATIGUE STRENGTH 27
Two stages of signal amplification tained and the number of holes reduced
were available to cover strain ranges of to two (Fig. i{d)). A photoelastic stress
up to 2 and 20 per cent respectively. analysis was conducted for both types
A simple relay and timer circuit was of notched specimen. The stress concen-
arranged to obtain automatically a tration was the same at each of the
suitable number of records during a holes in both the triple and double
test. group, giving Kt = 2.44. Fatigue tests
conducted under the same stress condi-
MATERIAL AND SPECIMENS tions for each type of notched specimen
Thin sheet material was chosen to give gave the same average endurance.
approximate plane stress conditions, For the fatigue tests in which the
TABLE I I , — M E C H A N I C A L PROPERTIES,
specimen was subjected to some com- (9). Oil-impregnated filter paper at low
pressive loading during the cycle, guide loads, and graphite grease at high loads
plates were used to prevent the test between the specimen and plates kept
70 /
>-»'
f
60
50
/ '^
k
40 III 1
30
\ • static Curve
X Dynamic Curve
20
10
Cycles, N
FIG. 3.—S-N Curves for Various Stress Ratios, Unnotched Specimens.
moved quite easily along the specimen and short-term (up to 30 min) room-
when under load. temperature creep tests were conducted
Unnotched and notched specimens tests. On the Schenck machine the upper
were also pulled to fracture on the limit of endurance, 10^ cycles, was
Schenck machine at the same strain rate selected because of the length of time
as was used for the low-cycle fatigue required for a test at 10 cpm. The above
tests, and the load-extension record limits provided a sufiBxient overlap of
obtained from the instrumentation. endurance to compare the effect of the
The "dynamic" stress-strain curve two frequencies and machines. The
(strain rate about 5 in. per in. per min) results of the tests on unnotched speci-
for an unnotched specimen is also shown mens on both machines are plotted as
in Fig. 2 up to 2 per cent strain. The curves of maximum stress against cycles
tensile strength was raised from 68.5 to to failure in Fig. 3.
70 tons per sq in. at the higher strain In some of the tests at high loads and
10^ 10*
Cycles, N
rate. Notched specimens gave an even short endurance, both at high and low
higher nominal tensile strength, 74 tons frequency, the specimens exhibited a
per sq in. continuous cyclical extension (ratchet-
ting) until failure finally occurred by
FATIGUE T E S T S tensile rupture rather than fatigue
The fatigue program was designed to cracking. In other cases, a small fatigue
cover a wide range of mean and cyclic crack occurred at one edge or internally
stress combinations. The tests were con- in the specimen, and a small visible zone
ducted at constant values of stress ratio of plastic deformation at the tip of the
of: R = - 1 . 0 , - 0 . 4 6 , +0.075, + 0 . 3 3 , crack soon spread into one or more yield
+ 0 . 5 , +0.725, and + 0 . 9 1 . bands at 45 deg to the axis of the speci-
The lower limit of endurance, 5 X 10^ men. Some of these fractures are illus-
cycles, on the Haigh machine was defined trated in Fig. 4.
by the time required to set up a test, or The results of the notched fatigue
excessive creep in the high-mean stress tests are plotted as Snax. against N'
BELL AND BENHAM ON EFFECT OF MEAN STRESS ON FATIGUE STRENGTH 31
curves for the various stress ratios and could occur on initial setting-up of a
both frequencies in Fig. 5. In general high mean stress, or on introduction of
the same hmitations on endurance the alternating component, took the
existed as above for the unnotched speci- form of small dimples on either side of
mens. each hole. At higher stresses the dimples
Fatigue cracks developed on both developed into yield bands looped be-
sides of two or all of the holes, with tween the holes and from outer holes to
little preference being shown between the edges of the specimen at 45 deg. In
the outside or the central holes. In all certain circumstances plastic dimpling
but the very low stress conditions, could be obtained at the notches without
intense localized plastic deformation any subsequent fatigue failure, while
preceded the crack tip; however, for under other conditions cracks would
very low stresses no permanent deforma- develop and propagate without any
tion could be observed, and the position prior macroscopic plastic deformation.
of the crack was difficult to locate. The Some examples of yield band and crack
localized plastic deformation, which formation are shown in Fig. 6.
32 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
20 30 40 50 60 70
Mean Stress {Sm),tons per sq in.
UTS
E
<
10 20 30 40 50 60 70 80
Mean Stress (Sm), tons per sq in.
(b) Notched specimens.
FIG. 7.—Continued.
latter is not zero, there is much specula- predict notched values on the basis of
tion and still no certain answer as to the unnotched data.
effect of stress concentration on the The results shown in Figs. 3 and 5
actual local values of mean and alternat- are reasonably comprehensive and pro-
ing stress. Without this latter knowledge vide a basis for detailed graphical
34 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
analysis of the various quantities. This Since there are generally more data on
has been done elsewhere (10), but due to fatigue at various mean stresses for
space limitations only some of the unnotched than notched specimens, it is
interesting features can be discussed very useful to have a method for pre-
here. dicting the effect of mean stress for
From the S-N curves (Figs. 3 and 5) notched specimens from unnotched data.
for unnotched and notched specimens, The simplest approach, when consider-
the Sa-Sm diagrams in Figs. 7(a) and (J) ing, say, a modified Goodman diagram,
have been derived for various values of is to divide the ordinate and abscissa by
endurance. Considering first the diagram the theoretical stress concentration
for unnotched specimens, the curves factor, Kt, so that
are bounded by the lines Sa/2 = 0,
5™ = 0, and 5„/2 + 5™ = 5„ = 70, and
their shape is very much dependent on '• • •'• (• - 1 )
the relative positions of the 5-A' curves.
For this reason, the rather sharp kink becomes
in the Sa-S^ curves for long endurances
at approximately R = 0 might have
been attributed to a misplaced S-N
-l('-f^)
curve; however, a study of Fig. 3 indicates The main weaknesses in this approach
that a fairly unreasonable shift of the are: (1) The strength reduction in
curves would be required to smooth out fatigue seldom achieves the value of Kt
the Sa-Sn curves. This has been noticed and is, therefore, given a separate symbol
on unnotched specimens by other in- Kf (to be defined later); and (2) the
vestigators (3), but it is certainly not the effect of local and then general yielding
general rule. A further point of note is around the notch influences the effective
that because the S-N curves intersect values of Sa and 5„ and hence the
with horizontal line S'max = 70 at various alternating-mean stress diagram. The
R values and endurances, the 5o-5„ general assumption that when local
curves are either asymptotic at low R, or yielding occurs at the stress raiser the
intersect at high R with the line Sa/2 + maximum stress is relieved to some
5„ = 70. extent, but that the range of stress is un-
Empirical expressions devised by affected implies that the local mean
Peterson (11), Stuessi (12), and Burdon stress will be reduced and a greater
(13) have each been compared with the range of stress can be maintained for a
experimental results, the last of those particular nominal mean stress.
giving the best correlation. For this Since the nominal static, tensile
particular material, difficulties arise strength of a notched specimen is gen-
because of the sharp inversion of Sa erally equal to or slightly higher than
versus S^ nt R = 0. A relationship of for an unnotched specimen, a better
the form approach to obtaining the notched
Sa-Sm diagram would be to join the
(1 - R)
-[t]"
where C and n are constants dependent
points Sa = So/Kt or So/kf to Sm =
Su notched by a straight line.
Reference to Fig. 1(b) for the present
on N, gives a better fit to the results, but tests shows that the above method of
has the disadvantage of requiring more predicting the notched relationship
initial experimental data for evaluation. would be in considerable error. Gunn
BELL AND BKNHAM ON EFFECT OF MEAN STRESS ON FATIGUE STRENGTH 35
(4), who has studied this problem, sug- These latter parts of the diagram,
gests that three distinct stages in a described in detail elsewhere (4,14), can
notched Sa-S^ diagram are: (1) the be constructed either by using a plastic-
condition that the local maximum stress stress concentration factor, Kp , or the
is still elastic, for which the first part of stress-strain curve for the material and a
E
<
20 30 40 50 60
Mean Stress ( S ^ ) , tons per sq in.
Su Plain^
Su Notched-
(a) iV > 10° (Fatigue endurance limit).
FIG. 8.—Comparison of Experimental and Empirical S„-Sm Curves.
the notched Sa-S^ diagram can be strain concentration factor equal to the
drawn up to the point of intersection elastic stress concentration factor. The
with the line joining Sa = Sy/Kt to Sm = three stages mentioned above are self
Sy/Kt or when local yielding com- evident in Fig. 7(6) at endurances down
mences; (2) occurrence of local yielding, to 10* cycles; while for unnotched speci-
with the body of the material remaining mens the inversion occurs at constant
elastic, and (3) general yielding in the stress ratio {R = 0), for the notched
material. curves the change in curvature occurs
36 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
at almost a constant mean stress of dieted curve and the notched Sa-Sm
approximately 25 tons per sq in. At curve for high, medium, and low en-
endurances below W cycles, where durances is shown in Fig. 8(a), (b), and
maximum stresses are high and more (c), respectively. In each case the theo-
general yielding is taking place, the retical curve is overly conservative.
10 20 30 40 50 60
Sa-Sm curves are fairly smooth and principally owing to the use of theoretical
parabolic in form. elastic and plastic stress concentration
Gunn's method of analysis avoids the factors rather than a fatigue strength
need of any initial notched fatigue data, reduction factor. The definition of the
because predictions are made directly latter term is well established in the
from unnotched results. That approach case of fully reversed stress cycling,
was applied to the present results, and that is, constant zero mean stress;
the comparison between Gunn's pre- however, if various mean stresses also
BELL AND BENHAM ON EFFECT OF MEAN STRESS ON FATIGUE STRENGTH 37
10 20 30 40 50 60 70 80
Mean Stress (Sm), tons per sq in.
(c) .V = 5 X 10^ cycles (Schenck machine data).
FIG. 8—Concluded.
Cyclic Lifetime
FIG. 9.—Strength Reduction Factor (kf^) Versus Cyclic Lifetime.
factor for 10^ <N < and Sa-Sm curves are derived and plotted on
S„, = 0, Figs. 8(a), (i), and (c). For each en-
S — tensile strength of un- durance the agreement with the experi-
notched specimen, and mental points is quite reasonable.
^ up
tensile strength of notched Until there is more detailed informa-
specimen. tion on the local stress conditions, Sa and
•^ un ^^
cycles yielding is no longer S„t, and the way in which these may
Below
localized,10^the mean stress is less effective, alter under elastic and plastic conditions,
and the nominal maximum stress it is not possible to predict accurately
governs k/^ • It is found that a linear notched fatigue data from unnotched
variation of k/„ with 5max fits the results fatigue tests. Thus, a notched fatigue
fairly well and may be expressed as: curve for at least zero mean stress, from
which other mean stress conditions
f^fm — W'Jmax ~\~ ^
might be reasonably derived, is neces-
where: sary.
B E L L AND BENHAM ON E F F E C T OF M E A N STRESS ON FATIGUE STRENGTH 39
specimens are rather remarkable in that Fig. 11 for various tests on unnotched
at higher endurances the material has a specimens a.t R = —0.46. To observe
greater fatigue strength than at high the shape of the latter part of the curves
frequency. This is completely opposite right through to fracture, the above
to what is usually postulated, and no results plus a few for other stress ratios
conclusive reasoning can be put forward have been plotted on a linear cycle scale
at this stage. in Fig. 12. To some extent these curves
Cyclic Lifefime
FIG. 11.—Cyclic Strain Versus Lifetime at Constant Stress Ratio (R = —0.46).
Y
^ L-!-/Vo. 061 Smax SB R-0.46
X-^No. 064 S„ax 68 R-0.46
[ 1 j
l ^ / V o 066 Smax 68 R-0.33
/
/
\
No. 050
64 R-OA 4
y
7^
/
^
, /
/
i 8
^
Jo. or.? Sma
4 R-l.0.60
y / ^
No. 0 6rs„ ax
66 R -0.33
1 /
/
/
/
/
r
I
A )
^
y
--'
&
.s
100 200 300 400 500 600 700 800 900 1000 MOO 1200 1300 1400 1500 1600 1700 1800
Cyclic Lifetime
FIG. 12.—Cyclic Creep Data Plotted to a Linear Lifetime Scale, Plain Specimens.
2.6
\ R-0.46
2.4
Smax rZOi c
R+0.075
2.2
72.0
\ JR + 0 33
S 2.0
Smax 68.0
R-0.46
•^ 1.8
Smox7l.5
o
C
/ R+0.075 M
^ 1.6 8/nox^
'y^J
o> R+O.C775
c
o
ID 1.4
! 1.2 X ^0.0 A
0)
R + 0.525
o
^8ffj ax 60.0
1.0 SmaxSO-O
> - 0.46 R+0075 ~
as • ^
0.6
Cyclic Lifetime
FIG. 13,—Cyclic Elongation, Notched Specimens.
specimens were small and relatively results obtained for very high maximum
localized; however, extensometer meas- stresses and various stress ratios are
urements were made on a Ij-in. gage shown in Fig. 13. Fracture elongations
length across the notched section, and for all the tests lie between 1 and 2.6
42 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
3°
10" 2 5 IQI 2 5 IQZ 2 5 |o3 2 5 ,34 2 5 .QS
Cyclic Lifetime
FIG. 14.—Variation of Total and Repeated Hysteresis Loop Width During Lifetime.
Cyclic Lifetime
per cent, and for a particular stress the solid curve is the total plastic strain
ratio elongation increases with the maxi- range, tj,i (Fig. 10). The difference be-
mum stress. tween the two curves (Fig. 14) represents
From the records of hysteresis loops, the cyclic creep strain per cycle.
the variation of plastic strain range Each curve shows an initial period of
during a number of the tests on plain reducing strain range which is greater as
specimens was determined. It has been stress ratio changes from negative to
10 20 30 •40
' 4 0 * f5;0 ' ' 7 0 80 100 150
Stress Range, tons per sq in.
FIG. 16.—Cyclic Plastic Strain Range Versus Total Stress Range, Unnotched Specimens.
1.6x10°
1.5 J
• R = -l.00
1.4 » R=-0.60 /
• R=-0.46
].3 0 R = -0.I0 /
» R =+0.075
a R = +0.33 /
1.2 " R-+0.48
c = Creep Faiure
1.0
0.9
•
0.8
w 1
0.7 i o
a
• / (Very Sm ill Loops )
06 Y^ • /
0.5
c 1
0.4 T 9
/
/
•
0.3
A
0.2
0.1 v ^ B
(Very Small Loops)
0.0
2 5 IQS 2 5 |o4
Cyclic L i f e T i m e , f r a c t u r e
FIG. 17.—Total Cyclic Energy to Fracture Versus Lifetime, Repeated Loops (Summation of
Repeated Loop Areas for Each Cycle to Fracture).
mens failing by cyclic creep, in which measuring the area of very small loops,
case tpr has been used. Over the range it appears that the results might be repre-
studied, mean stress appears to have sented by one smooth curve, as shown,
virtually no influence on plastic-strain of increasing energy with increasing en-
range, and the line in Fig. 16 is repre- durance, which again appears to be
sented by the equation: independent of stress ratio.
SUMMARY
It is probably mere coincidence that The effect of local yielding on notched
the constant 141 is almost twice the specimens was to reduce the effective
tensile strength of the material, which mean stress and hence increase the al-
would give for this steel: lowable stress amplitude for a particular
" > = < ? , 0.127 endurance, as illustrated in the Sa-Sm
B E L L AND BENHAM ON E F F E C T OF M E A N STRESS ON FATIGUE STRENGTH 45
REFERENCES
(1) J. 0 . Smith, "The Effect of Range of (8) W. Illg, "Fatigue Tests on Notched and
Stress on the Fatigue Strength of Metals," Unnotched Sheet Specimens of 2024-T3
University of Illinois Bulletin, Vol. XXXIX, and 7075-T6 Aluminum Alloys and of
No. 26 (1942). SAE 4130 Steel with Special Considera-
(2) B. J. Lazan and A. A. Blatherwick, tion of the Life Range from 2 to 10,000
"Strength Properties of Rolled Aluminum Cycles," NACA TN 3866 (1956).
Alloys Under Various Combinations of (9) W. C. Brueggeman and M. Mayer, "Guides
Alternating and Mean Axial Fatigue for Preventing Buckling in Axial F'atigue
Stresses," Proceedings, Am. Soc. Testing Tests on Thin Sheet Metal Specimens,"
Mats., Vol. 53, p. 856 (1953). NACA TN 931 (1941).
(3) A. R. Woodward, K. G. Gunn, and G.
(10) W. J. Bell, "The Effect of Stress Concen-
Forrest, "The Effect of Mean Stress on the
trations and Local Plastic Deformation
Fatigue of Aluminum Alloys," Inter-
on the Fatigue Properties of Stainless
national Conference on Fatigue, Inst.
Steel Sheet," Ph.D. Thesis, London
Mechanical Engrs., p. 158, Sept., 1956.
University (1961).
(4) K. Gunn, "Effect of Yielding on the
Fatigue Properties of Test Pieces Contain- (11) R. E. Peterson, "Review of F'atigue of
ing Stress Concentrations," Aeronautical Metals Field," Applied Mechanical Review,
Quarterly, Vol. VI, Nov., 1955. Vol. 5, No. 1 (1952).
(5) P. P. Benham, "Axial Load and Stran- (12) F. Stuessi, "Theory and Test Result on the
Cycling Fatigue of Copper at Low Ein Fatigue of Metals," Proceedings, Am. Soc.
durance," Journal Inst, of Metals, Vol. 89, Civil Engrs., Journal Structures Div., Vol.
No. 9, p. 328 (1961). 85, No. ST8, p. 65, Oct., 1959.
(6) P. P. Benham and H. Ford, "Low En- (13) W. H. Burton, "A Practical Method of
durance Fatigue of a Mild Steel and an Fatigue Stress Analysis," Aircraft Engi-
Aluminum Alloy," Journal of Mechanical neering, Vol. XXVII, p. 299, Sept., 1955.
Engineering Science, Vol. 3, No. 2, p. 119 (14) H. F. Hardrath and L. Ohman, "A Study
(1961). of Elastic and Plastic Stress Concentra-
(7) M. H. Weisman and M. H. Kaplan, "The tion Factors Due to Notches and Fillets in
Fatigue Strength of Steel Through the Flat Plates," NACA Report 1117 {and TN
Range from 5 to 30,000 Cycles of Stress," 2566) (1951).
Proceedings, Am. Soc. Testing Mats., Vol. (15) G. P. Tilly and P. P. Benham, "Load
50, p. 649 (1950). Cycling in the Low Endurance Range in
46 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
Relation to Brittle Fracture of Mild Steel," Stainless Steel and Heat Treated 403 Stain-
Journal, Iron and Steel Inst., Vol. 200, less Steel," iV/lC4 TN 3017 (1953).
No. 3, p. 216 (1962). (17) C. T. Mackenzie and P. P. Benham,
(16) H. F. Hardrath, C. B. Landers, and E. C. "Low Cycle Fatigue of an Aluminum-
Utley, "Axial I^oad Fatigue Tests on Copper Alloy B. S. L65," Journal, Royal
Notched and Unnotched Sheet Specimens Aeronautical Soc, Vol. 66, No. 614, p. 129
of 615-T6 Aluminum Alloy, Annealed 347 (1962).
STP338-EB/Sep. 1963
SYNOPSIS
successfully for other high-strength ma- mens were then subjected to reversed
terials problems, would be appropriate bending in a Krouse plate fatigue ma-
to obtain meaningful behavior data chine to initiate a fatigue crack at the
which would lend itself to analysis. ends of the notch to act as crack starters.
The basic mathematical model of These fatigue cracks are indicated on the
fracture mechanics has been carefully fracture surfaces in Fig. 1.
reviewed by Irwin (2). In this review he Both a flat fracture and a 45-deg
has defined the force conjugate to speed shear lip were present in the fracture of
of crack extension as the rate of dissipa- the 0.160-in. thick specimens, as shown
tion of strain energy with crack exten- in Fig. 1. Ink staining indicated that the
sion, g. The study of the rate of slow flat portion of the crack tunneled into
crack extension, as previously advocated the specimen a greater distance than the
(3), as a function of the g value or crack surface crack. To obtain the relation be-
tip strain field was followed in this in- tween the internal flat crack length and
vestigation. the surface crack length, a series of
specimens for each thickness and direc-
MATERIAL tion were ink stained or heat tinted, or
The material used for this investiga- both, at various stages of crack exten-
tion was D6Ac steel supplied by the sion. The results for the 0.160-in. thick
specimens are shown in Fig. 2. The ad-
TABLE I.—CHEMICAL COMPOSITION justed crack length, arbitrarily computed
OF D6Ac STEEL. by dividing the area of the actual crack
Carbon, per cent 0.46 by the thickness of the specimen rather
Manganese, per cent 0.79
Phosphorus, per cent 0.005 than using the surface crack length,
Sulfur, per cent 0.007 was used to compute values of g.
Sihcon, per cent 0.18
Nickel, per cent 0.51 The results obtained for the 0.075-in.
Chromium, per cent 0.98 thick specimens (Figs. 3 and 4) indicate
Molybdenum 0.95
Vanadium, per cent 0.07
that the fractures of these specimens
showed full shear.
Ladish Co. Plate samples 0.160 and The initial crack length (notch plus
0.075 in. thick were obtained by flatten- fatigue crack) of the specimen was
ing and grinding roll ring forgings. The measured with a traveling micrometer
chemical composition of the material is microscope, and the value of load re-
given in Table I. quired to produce a given value of crack
Samples of this steel were austenitized tip driving force, g, was computed by
at 1550 F and cooled in an air blast. The solving Eq 1 (4) for a:
specimens were tempered for 4 hr at a--W
650 F to develop 240,000 psi yield Q =
t)
h +
-t- ^
tan I - (1)
E \W 2Wa
strength at 0.20 per cent offset.
The specimen was loaded in uniaxial
EXPERIMENTAL PROCEDURE
tension to the calculated value, held for 3
Centrally notched fracture test speci- min, and then released. The surface
mens 3 by 12 in. from both the longi- crack length, measured with the travel-
tudinal and the transverse directions, ing micrometer microscope, was used to
were used for this investigation. After determine the effective crack length
heat treatment, a central slot 0.500-in. from Figs. 2 or 4.
long was cut in the specimen by elec- The value of the effective crack length
trical-discharge machining. The speci- was used in Eq 1 to compute the value
CARMAN ET AL ON FATIGUE or ULTRAHIGH-STRENGTH STEELS 49
FiG. 1.—Study of Crack Shape for D6Ac Steel, 0.160 In. Thick; Yield Strength, 240,000 psi.
Arrows indicate extent of heat-tinted crack.
50 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
X A
ff»
o
J3
z
0>
O
LJ fl
01
u
1-
(J
to fa
I -5
y
^ \ ^ N.
,^ < (fl (/J
0) T3
C
>
1-
H
2 H
°g CI J^
tin
\K - H
„ F-l
H Q 1^
1 UJ
o A
H
1 z
Q. o
^ 3 \
l\K
H <
• 1
H
6
D a
Xx-^ H
73
o
lil
1-
" ^ </)
< I^
X T3 tfi
:
zo if M u5
u J
-1
a.
. ^\ rt r m
?| \ \
S: S
t- <
!! M
rt tn M
^
3"
1^
J3
b
1 ~;
<^^ S
Q,
. !/l
3
n
2 '^.J 5
[n.Si H
Ni'MisNan Novtio nvNuaiNi J3
CAKMAN ET AL OK FATIGUE OF ULTRAHIGH-STRENGTH STEELS 51
mens there is a fairly rapid crack growth this phenomenon is associated with the
for the first three to four cycles, after shear lip development. There is an
which the rate of crack growth becomes equilibrium value of shear lip for each
relatively constant. I t is believed that value of 9 3.t which the specimen is
at'-*im
O
A
d
o
\ X
\ "*
X N\
\\ ^ 1-
v> o
3
^ a< I: ti5
\\r\ V ^
IN \
•v\i "
1 "
-u
ri
rt
\ <
x\ t«
^hn
te^,'
tn
I
r\i . ~ G S
J 1 \
a.
1
°fe o \ \ E
TS t^
^t- 23 a. \ \
1-
1 \ \//
\ l\ -a
A
oc
c
• 1
\ h-l OS
\ ^u
OS
•n
B
•-<
U
'Nl' H10N3n >OVtl3 nVNb31NI
c
H a
o aj
^ \ V X
a
s H -*J
•n
c
pp
\ X
\ \ •** O
H LJ
3
1^^
~>
a
<
i H
o
w
z
Q. ^ O
0. -»
J3
\I' m 5!
9 M o
h-
o
. M
\ N
H
a
. 1
Si
\
tn
J3
-Nl H19N3-I yiytva -iwa^iM
CARMAN ET AL ON FATIGUE OF ULTRAHIGH-STRENGTH STEELS 53
cycled. During the first few cycles of length-number of cycles curve in the
loading the crack growth is fairly high linear part, r^s is the plastic zone size
until this equilibrium value of shear lip and is equal to E^/irays^, and C is a
is attained. This effect is illustrated by numerical constant. In analyzing fatigue
both the curves and small macrographs data on 7075-T6 aluminum, Irwin found
in Figs. 5 and 6. that a value for C of 0.02 fit the data
Irwin (5) has suggested a steady-state points fairly well.
NUMBER OF CYCLES,N
model to describe crack growth in fa- Corten (3) has modified the expression
tigue, in which the coordinates are fixed to account for the zero intercept as
and the material flows past the co- follows:
ordinates at a fixed rate. Mathematically
this may be expressed as:
da
TN = ^'-
(2)
da
— = 0.02
dN
EQ
^
Tro-ys* ar
where £ic is the plane strain fracture
.(3)
where da/dN is the rate of crack propa- toughness. This expression was used to
gation defined as the slope of the crack analyze the data in this investigation.
54 SYMPOSIUM ON F A T I G U E O F AiECRArx STRUCTURES
Table III summarizes the data ob- mens T5, TB, L l l , L7, L4, and L3. The
tained for the 0.160-in. thick specimens. experimental values of da/dN are con-
In Fig. 7(a) are plotted the values of siderably larger than predicted from Eq
1.24
1.20
G = I25
1.16
1.12
i.oa
1.04
1.00
G =IIS
0.96
0.92
o.ae
0.84
G= 100
0.76
NUMBER OF CYCLES, N
da/dN versus g for specimens T22, T24, 3 using the same value of the constant
L5, L2, L6, and T21, the results of which C. Scatter will be observed in these data,
are in agreement with the values pre- but this is not considered unusual in
dicted from Eq 3. In Fig. 7(6) are plotted fatigue phenomena.
the values of da/dN versus g for speci- In supposedly identical specimens of
CARMAN ET AL ON FATIGUE OF ULTKAHIGH-STRENGTH STEELS 55
the same material, 50 per cent of the crack approaches the surface with a
data agree with calculated values from characteristic lead of the tip to shear
the steady-state model, and the re- lag, and the crack front shape is estab-
mainder show growth rates in excess of lished (Fig. 8(a)).
those predicted. The fracture toughness Upon unloading, the crack is forced
developed in these specimens does not closed by the surrounding elastic field,
vary as inferred by the per cent shear but the y displacements cannot be
measured after cycling at g = 125. zeroed because the crack is propped
Therefore, this effect is not the con- open by the crack tip plastic zone. How-
trolling factor (see Table I I I ) . Further ever, the plastic zone is not about the
examination of this table shows that the axis perpendicular to the sheet surface
fracture contours of the specimens that but along the tunneled front profile.
agree with predictions from Eq 3 show Collapse of the load will leave a residual
TABLE III. -SUMMARY OF SLOW CRACK GROWTH DATA FOR 0,160-IN. THICK
SPECIMENS OF D6Ac STEEL.
^ FOR A EQUAL TO ;
SHEAR,
FRACTURE
PER CENT
PROFILE
100 115 125 150 175 185 200 215 225 240
aligned 45-deg obhque shear lips except stress pattern sufficient to accomplish
for specimen L5; the fracture contour some compressive plastic strain toward
of the specimens that disagree show the tail as shown in Fig. 8{b).
elements of the cup-cone fracture, except The compressive plastic deformation
for specimens L7 and L4. These speci- occurs to an increasing degree toward
mens show 45-deg oblique shear lips the tail of the crack as the elastic open-
which are oriented in opposite directions ing, Av, contracts through greater y
on each side of the specimen. displacement toward this region. This
The following tentative phenomeno- is the irreversible part of the cycle, and
logical explanation has been proposed in the case of high elastic constraint
(6). I t will be assumed that a more severe the effect is intensified.
restraint is placed on the plastic strain Upon reloading, the yield stress pat-
zone at the crack tip by the cup-cone tern changes; the Bauschinger effect
type fracture. prevents the tail of the crack tip from
As the load is applied, the crack ex- supporting its previous unitensile share
tends in the center, shear occurs as the of the load. 9 is intensified at the crack
56 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
z m • •+
UJ <D
z
o
tu lO CD = K ^ •<>
a. 1 1 1 1 I 1
v> 1- H -1 -1 -J -1
1-1
11
0 \
\ a .5
\ 03 >
1
1
ill
-i
<
>
• «i
•
^^'—41"
a • u
u
t-
<
_i
O
_J
JO
<
o
o
b
z
Ui • 0 > 0 x <]
J3
UJ eg M <" C^ U> N
(0 H H J J J h-
a; O
X d
rt J5
1 \. X
o x < \ a
o a o
o
OJ (T u
UJ d
o
1 NyX \
i5 CO Ml
UJ \ 3 a
3 z o
3
O ® 2 - -7^,
UJ
t-
< TS O
O
-J
anoAD dSd'Ni'
CARMAN ET AL ON FATIGUE o r ULTRAHIGH-STRENGTH STEELS 57
1 X
•H- 6^ FOR SPECIMENS NOT
MTI6UE CYCLED
e
9
y-
STRESS FIELD
^
! V^^^<r„ l<^
STRAIN FIELD
100 200 9 00
FINAL A IN-LB PER SO IN. VALUE IN FATIGUE
G-600
2 4 fi 8 10 12 14 16 IS 2<
NUMBER OF CYCLES, N
FIG. 10.—Slow Crack Extension of Longitudinal Specimen L-4, 0.075 In. Thick.
Note extent of fatigue crack and fracture formation (approx X12).
G • 600
y
/
NUMBER OF CYCLES, N
FIG. IF—Slow Crack Extension of Transverse Specimen T-3, 0.075 In. Thick.
Note extent of fatigue crack and shear fracture formation (approx X12).
CARMAN ET AL ON FATIGUE OF ULTRAHIGH-STRENGTH STEELS 59
The other specimen shows a rapid than for the cup-cone fracture; thus it
and fairly constant rate of slow crack would be anticipated that the 45-deg
growth (Fig. 11) and a fracture of 45 deg oblique fracture profile would show the
oblique shear. These specimens could be faster slow crack-growth rates.
cycled at higher values of g than the
SUMMARY
heavy 0.160-in. thick specimens. Table
IV summarizes the data obtained for A correlation exists between the steady
the 0.075-in. specimens. state model as proposed by Irwin and
In Fig. 12(a) are plotted the values of the results of this investigation. How-
da/dN as a function of 9 for 0.075-in. ever, two specific types of behavior are
thick specimens L2, L4, T7, T4, and evident in the data which are a function
T5, which exhibited elements of the of specimen thickness, and they can be
cup-cone fracture. These results are in explained to some degree by the con-
qualitative agreement with the predicted sideration of two factors: (1) the elastic
values of Eq 3. In Fig. 12(6) are plotted constraint present in the specimen, and
the values of da/dN versus g for speci- (2) the ease with which vertical shear
mens T6, L I , T l , and T3 which showed displacement can be attained.
45-deg oblique fracture and much greater In the thick-plate behavior, which is
rates of slow crack extension. These exemplified by the 0.160-in. specimens,
effects are opposite to those observed the plate thickness is substantially
for the heavier specimens. The 0.160-in. larger than the local plastic zone size
thick specimens with the cup-cone type at the advancing tip of the crack. The
of fracture showed the greatest rate of elastic constraint, therefore, dominates
crack growth, whereas the 0.075-in. thick the rate of crack propagation. Because
specimens with oblique shear showed of the lower elastic constraint present
the largest crack growth rate. These in a 45 deg obhque shear fracture, the
results, however, are congruent. For the rate of crack propagation at a constant
thin specimens which fail in full shear, driving force is less than that for the
the vertical y displacements are easier cup-cone condition.
to attain for the 45 deg oblique shear In the thin-plate behavior, the shear
60 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
1
1
1 hn
ffl
l n
o, >
^"^"^ a
\ fl
Li
Q
1 Cl fi,
1 J2
1
H
U \ u
<>
_J
1 U
; 3
1
• X + <1 Q o o
U
\ u £3
5 •a
|I- _l H t- O lO
_l
•1 0
o 3
-c h
!
X + <] a
1
^ cj • m
! -
1
J
< •
1 - 1- I \ t §
-+ ' r-V
\
1-
— —A— •i
1
1 \ o PL,
t
1 '
\
1 I
;
)f
-
^ S
CO C J
i
i
I 2
fa
c ) c 10
5
< < :>
313Jk3 H3<j Nl '
CARMAN ET AL ON FATIGUE OF ULTRAHIGH-STRENGTH STEELS 61
REFERENCES
(1) G. R. Irwin, Encyclopedia of Physics, Vol. (4) "Fracture Testing of High-Strength Sheet
VI, Julius Springer, Heidelberg (1958). Materials: A Report of a Special ASTM
(2) G. R. Irwin, "Relation of Crack Toughness Committee," ASTM BULLETIN, N O . 243,
Measurements to Practical Applications," Jan., 1960, p. 29; No. 244, Feb., 1960, p. 18.
Welding Journal Research Supplement, (5) G. R. Irwin, Lecture Series Presented at the
November, 1962. Department of Theoretical and Applied
(3) H. T. Corten, "Slow Crack Extension in Mechanics, University of Illinois, December,
High-Strength Steels," Proceedings, Seventh 1961.
Sagamore Ordnance Materials Research (6) J. M. Krafft, private communications.
Conference, Aug. 16 to 19, 1960, Vol. IV, (7) A. Nadai, Plasticity, McGraw Hill Book
p. 117. Co., New York, N. Y. (1931).
STP338-EB/Sep. 1963
LOW-CYCLE F A T I G U E OF Ti-6A1-4V A T - 4 2 3 F
B Y R. R. H I L S E N , 1 C. S. Y E N , 1 AND B. V. WHITESON 1
SYNOPSIS
3%" - 3Vi"
1.5"
r.
r
0.6"
lb) SECTION A-A, AFTER MACHINING, BEFORE WELDING,
TWO SYMMETRICAL PCS.
~1
0.750- D ,, ,-
0.000- — ( I H-'
0.005- ^^
of the 3-in. diameter bar stock to make and base metal specimens were annealed
it a tube. Then two 3|-in. long tubes for 1 hr at 1300 F followed by air cooling.
were welded together yielding six "tube" A typical fatigue specimen is shown
specimens per "tube stock" (Fig. 1). in Fig. 2. Welded specimens were ma-
This method of welding together tubes chined with the weld located at the
approximates the welding practice used center of the bar where the diameter
on actual pressure vessels. All welded was a minimum (highest stress section).
(a)
280 000
270 000-
A STATIC TENSION
O R = 0.02
o R = 0.2
240 000-
oi
290 000.
(b)
280 000
•%
I 260 000
s
X
<
^ 250 000-
t. STATIC TENSION FATIGUE STRENGTH FOR GIVEN LIFE . psi
O R = 0.02 10 CYCLES 100 CYCLES 1000 CYCLES
O R = 0.2 271,000 270 500 270 0 0 0
(c)
280 000
270 000
oi
102 103
CYCLES TO FAILURE
Static tension tests were conducted on a wave form. Specimens were immersed in
standard universal testing machine. In Uquid hydrogen within a double-wall
the fatigue tests, repeated tension-ten- stainless steel cryostat during testing
sion load was applied by a 15,000-lb at - 4 2 3 F.
280 0 0 0 -
u! 270 0 0 0 - ^O^-^
250 000
10 IC IflJ
CYCLES TO FAILURE
FIG. 5.—Effect of Cycling Speed on Fatigue Strength of Ti-6A1-4V (Heat M9779) Pressure
Weldments at - 4 2 3 F.
290
270-
250
^
"2 180-
PRESSURE WELDMENTS ROOM
X
< 160
-1 TEMPERATURE R ^ 0.03
CYCLES TO FAILURE
E F F E C T OF LOAD RATIO
Base M e t a L G8549 . . 0.02 12 -423 263 500 262 500 260 500
Base M e t a l , G 8 5 4 9 . . 0.2 12 -423 271 500 269 500 267 500
P e r cent C h a n g e " . 3.0 2.7 2.7
Base M e t a L M 9 7 7 9 . . 0.02, 0.2 12 -423 273 000 272 500 271 500
Weldment, M 9 7 7 9 . . . 0.02, 0.2 12 -423 271 000 270 500 270 000
Weldment, M9779.. . 0.03 12 Room tem- 165 000 163 000 161 500
perature
QUALITY OF WELD
and Sinclair (S) in steel at room tempera- - 4 2 3 F (Table IV). This cannot be
ture. A possible explanation for this be- attributed to the statistical scattering
havior is suggested by Sines (6) in which effect, since sometimes the whole group
the energy to activate the plastic de- of fatigue specimens exhibited strength
formation or "cycHc creep" is obtained higher than static. This interesting phe-
from microacoustic waves produced from nomenon prompted further investiga-
dislocation interactions during cycling. tions using the third heat of Ti-6A1-4V.
HiLSEN ET AL ON LOW-CYCLE FATIGUE OE Ti-6A1-4V AT - 4 2 3 F 69
When the fatigue machine was started, machines, static tension tests of six
for usually about three to eight cycles, identical specimens were conducted on
the maximum loads on the specimen both machines. The strain rate on the
were slightly below the designated maxi- fatigue machine was the same as that in
mum. It seems that the material is normal fatigue testing. The average
strengthened rather than damaged dur- strength of three specimens on each
ing these few cycles so that it can with- machine was the same, 264,000 psi
stand a stress higher than the static. (Table V).
To be certain that the effect was not As further evidence of the strengthen-
due to strain rate or any other differences ing eilect, six specimens were cycled on
between the universal (1200 lb per min) the universal and fatigue testing ma-
and fatigue (about 3,600,000 lb per min) chines while the load was increased
70 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
CYCLES TO FAILURE
COMPLHE FRAaURE -
CENTER NUCLEATED (a)
K-
SPECIMEN 5-4
SPECIMEN 5-3
o-p^^j^tisai^
(b)
< 10- COMPIETE FRACTURE
< (SURFACE NUCLEATEI^)
Z
Z 5 SPECIMEN B-7
O
(cj
COMPLETE FAILURE -
25-
^ 60 CYCLES
10-
T A B L E I I I . — S T A T I C T E N S I L E S T R E N G T H , Ti-6A1^V (ANNEALED).
[ Room temperature
Room temperature
138 800
138 700
149
149
300
100
9.5
11.5
38.9
41.8
Weldment, M9779 \
-423 257 800 2.5 22.2
-423 257 100 3.0 21.2
I
" Values listed on t e s t report for h e a t of m e t a l .
' F r a c t u r e originated at n o t c h caused b y extensometer grip.
B.\8E M E T A L , M9779
WELDMENT, M9779
u l t i m a t e tensile s t r e n g t h — fatigue s t r e n g t h
" P e r cent change
u l t i m a t e tensile s t r e n g t h
'' Average tensile s t r e n g t h from static tension t e s t s .
' P a r t of high fatigue s t r e n g t h is p r o b a b l y due to strain r a t e since t i t a n i u m is s t r a i n - r a t e sensitive
a t room t e m p e r a t u r e .
HiLSEN ET AL ON LOW-CYCLE FATIGUE OF Ti-6A1-4V AT —423 F ?3
TABLE v . — E F F E C T OF LOAD-RATE
A 2000 258 000 271 000
ON U L T I M A T E T E N S I L E STRENGTH
B 2000 260 000 271 000
O F Ti-6A1-4V A T - 4 2 3 F , H E A T D-1700.
commercial aluminum alloys and steels. suited from decreasing the test tempera-
An explanation is illustrated in Fig. 12. ture from ambient to —423 F.
At the point where the original material 5. Cycling below ultimate tensile
would have failed, sufficient strengthen- strength tended to raise the fatigue
ing has occurred to change the S-N curve strength above the static tension
to the upper one. strength.
6. A transition in the fracture appear-
CONCLUSIONS
ance occurred between 700 and 2000
At —423 F, for any fatigue life of cycles, where the fracture changes from
annealed Ti-6A1-4V from 10 to 1000 a cup-and-cone to a more typical surface-
cycles: initiated fatigue fracture.
STRENGTHENED MATERIAL
N' N
REFERENCES
(1) J. T. P. Yao and W. H. Munse, Welding Sons, Inc., New York, N. Y,, p. 450
Research Supplement, April, 1962, p. 182-s. (1959).
(2) W. A. Wood, Fracture, Technology Press, (9) A. J. McEvily, Jr. and E. S. Machlin,
Massachusetts Inst. Technology, John Transactions, Am. Inst. Mining, Metal-
Wiley & Sons, Inc., New York, N. Y., p. lurgical, and Petroleum Engrs., Vol. 221,
412 (1959). p. 1086, Oct. 1961.
(3) N. E. Frost, Nature, Vol. 192, p. 446, Nov. (10) D. M. Fegredo and G. B. Greenough,
4, 1961. Journal, Inst, of Metals, Vol. 87, p. 1
(4) P. P. Benham and H. Ford, Journal, (1958-1959).
Mechanical and Engineering See, Vol. 3, (11) W. J. Wadsworth, Dislocations and Me-
p. 119 (1961). chanical Properties of Crystals, John Wiley
(5) JoDean Morrow and G. M. Sinclair, T & & Sons, Inc., New York, N. Y., p. 477
AM Report No. 345, Dept. of Theoretical (1957).
and Applied Mechanics, University of Illi- (12) H. A. IJpsitt and D. Y. Wang, Transac-
nois, Urbana, 111. (1957). tions, Am. Inst. Mining, Metallurgical, and
(6) G. Sines, Discussion of the paper by F. H. Petroleum Engrs., Vol. 221, p. 918, Oct.,
Vitrovec, "Combined Stress Effect at 1961.
Elevated Temperatures on Super Alloys," (13) J. C. Levy and S. L. Kanitkar, Journal,
presented at SAE Winter Meeting, Detroit, Iron and Steel Inst,, Vol. 197, p. 296, April,
Mich., 1961. 1961.
(7) T. Broom, J. H. Molineaux, and V. N. (14) H. F. Hardrath, B. C. Landers, and E. C.
Whittaker, Journal, Inst, of Metals, Vol. Utley, Jr., NACA TN 3017, Oct., 1953.
84, p. 356 (1955-1956). (15) H. F. Hardrath and W. Illg, NACA TN
(8) A. J. McEvily, Jr. and E. S. Machlin, 3132, Jan., 1954.
Fracture, Technology Press, Massachu- (16) W. Illg and H. F. Hardrath, NACA RM
setts Inst. Technology, John Wiley & L55D15a (1955).
STP338-EB/Sep. 1963
LOW-CYCLE FATIGUE P R O P E R T I E S OF C O M P L E X W E L D E D J O I N T S
OF H I G H - S T R E N G T H 301, 304L, 310, AND AM-355
STAINLESS STEEL S H E E T MATERIALS AT
CRYOGENIC T E M P E R A T U R E S *
SYNOPSIS
and are of thin gages (0.010 to 0.040 in. which gives nearly 100 per cent joint
thick). The conditions imposed by strength efficiency is used.
environment and structural design are The use of stainless steels for cryogenic
cryogenic temperatures, high stresses, tankage therefore created the need for
and the use of complex welded joints. fatigue data on thin-gage, high-strength
Cryogenic temperatures result from the sheet materials over the temperature
use of liquid oxygen ( —297F) as an range from 78 to —423 F. Because of
oxidizer in the Atlas and Centaur and the finite operating life of missiles and
liquid hydrogen (—423 F) as a fuel in most space vehicles, the amount of
the Centaur. To minimize the weight fatigue or number of repeated loadings
Temper.. 60 per cent" 60 per cent" 50 per cent" 75 per cent" CRT*
Gage, in. 0.025 0.010 0.012 0.020 0.032
Supplier. Washington Washington Rodney Washington Wallingford
Steel Steel Metals Steel Steel
Heat No 49061 57644 33251 43631 38174
Coil No 7450 11976 44942
Rockwell Superficial Hard-
ness, 15N Scale 83.9 81. 76.8 79.3 86.6
Martensite, per cent 76 62 0 0 95
Chemical Composition,
weight per cent:
Carbon 0.07 0.09 0.023 0.060 0.14
Chromium 17.28 17.38 18.04 24.62 15.60
Copper 0.23
Iron balance balance balance balance balance
Manganese 0.66 1.04 1.54 1.60 0.72
Molybdenum 0.32 2.71
Nitrogen 0.031 0.11
Nickel 6.70 7.32 10.39 19.66 4.38
Phosphorus 0.022 0.026 0.030 0.018
Sulfur 0.015 0.011 0.011 0.018
Silicon 0.63 0.66 0.58 0.29
•» Cold rolled.
* Cold rolled and tempered.
of flight vehicles, maximum operating to which they are subjected is quite low;
stresses are about 90 per cent of the however, the operating stresses are quite
0.2 per cent yield strength. Also use is high. Therefore the data of primary
being made of low-temperature design interest are very-high-stress, low-cycle
allowables when appHcable. Because fatigue. It has been found in past studies
pressurization of the tanks is required (4) that the complex welded joints
for stabilization and flight operation, generally have a much shorter fatigue
it is necessary to fabricate the tanks by life than does the parent metal; therefore
leak-tight fusion or resistance roll seam the fatigue studies described here were
welds. The joint efficiencies of these on welded joints typical of those used in
welds may be quite low (30 to 80 per missiles.
cent). Therefore, to take advantage of In addition to obtaining the fatigue
the materials' high strengths, a doubler data, one of the primary purposes of this
78 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
tension and shear tests were conducted cold rolling. This transformation is
on individual resistance spot welds at advantageous for obtaining high yield
the same temperatures. The data from and tensile strengths but is also responsi-
these evaluation tests were analyzed to ble for the tendency toward decreased
determine which test or tests best pre- toughness at extreme cryogenic tempera-
dicted the fatigue life of the materials tures (—423 F). As may be seen in
investigated. Another objective of this Table I, the amounts of martensite
program was to provide useful engineer- present for the two heats of 60 per cent
ing data on the mechanical properties cold-rolled type 301 are 62 and 76 per
of several alloys at room and cryogenic cent. Nearly 100 per cent martensitic
temperatures. Therefore five or more transformation occurs in the 301 alloy
replicate tests were made and many of during tension testing at cryogenic
the data were statistically reduced. A temperatures. Type 304L stainless steel
thorough description of the test program contains a sufficient amount of nickel to
and results may be found in reference (S). be stable during room-temperature roll-
CHRISTIAN ET AL ON LOW-CYCLE FATIGUE or WELDED JOINTS 79
ing; however, a considerable amount of Neuber's concept (7) are given in Table
martensite forms during tension testing II.
at cryogenic temperatures (—320F). The cross-tension and tension-shear
The 20 per cent of nickel present in specimens used for testing individual
type 310 stainless steel makes it fully resistance spot welds conform to Specifi-
stable during cold rolling or tension cation MIL-W-6868A. The fatigue speci-
testing at cryogenic temperatures. Cold- mens had a test section 16 in. long by 4
rolled and tempered AM-355 stainless in. wide containing a centrally located
steel is nearly 100 per cent martensitic. complex welded joint. Longitudinal
The test specimens used in this in- joint No. 1 was composed of a linear
vestigation (Figs. 1 to 4) include: stand- butt fusion weld (no filler metal) which
ard pin-loaded flat tension specimens; was strengthened to nearly 100 per cent
notched tension specimens having three joint efficiency by means of a doubler
different notch geometries; cross-tension sheet attached to the specimen by four
and tension-shear specimens of individual rows of resistance spot welds on each
resistance spot welds; and several con- side of the fusion weld. Longitudinal
figurations of large fatigue specimens joint No. 2 was similar to No. 1 except
(38 in. long) containing complex welded that only 2 rows of spot welds were used
joints. The notched tension specimens' to attach the doubler sheet. Transverse
dimensions and stress concentration joint No. 1 was composed of an over-
factors, Ki, as determined by the lapping joint made by resistance roll
w air, by Peterson's equation (6), and by seam welding with one row of resistance
80 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
V'V<.'
spot welds on each side of the seam weld. Figs. 5 and 6 show typical tension and
Transverse joint No. 2 was the same as fatigue testing cryostats.
longitudinal joint No. 1 except that the
material was tested transverse to the EXPERIMENTAL RESULTS
direction of rolling. A thorough descrip-
Table III gives the tensile and notched
tion of the test specimens may be found
tensile properties of types 301, 304L,
in reference (5).
310, and AM-355 stainless steels as a
EXPERIMENTAL PROCEDURE function of direction and temperature.
The values reported for 0.2 per cent
The low-temperature tension tests
yield strength, tensile strength, and
were performed in cryostats mounted on
elongation are averages of five replicate
standard universal testing machines.
tests at each condition of testing. The
TABLE XL—DIMENSIONS AND STRESS notched: unnotched tensile strength ra-
CONCENTRATION FACTORS FOR tios were determined by dividing the
NOTCHED TENSION SPECIMENS.
average notched tensile strength by the
Mild Medium Severe
tensile strength. The notched tensile
Notch Notch Notch strengths were also determined by five
replicate tests. Table IV gives the
Total length, in 9 9 8 mechanical properties of linear fusion
Total width, in 1.25 1.25 2.0
Width in reduced sec- welds and resistance spot welds. The
tion, in 0.4 0.4 1.0 weld schedules are given in reference (5).
Width between notches,
2a, in.. 0.2 0.2 0.7
The values reported for the linear fusion
Notch root radius, r, in.. 0 . 0 1 0 . 0 0 2 5 0.001 welds are averages of five tests and are
K, W^r) 3.2 6.3 18.7 averages of twenty tests for the resistance
Ki (Peterson)" 3,8 7.2 21.0
Kt (Neuber)'' 3.9 7.5 spot welds. The elongations reported in
Table IV are total elongations as
" Reference (6). measured over a 2-in. gage length.
' Reference (7).
The results of axial fatigue tests on
Temperatures were maintained by im- complex welded joints are given in Table
mersion in dry ice and alcohol (—100 F), V and Figs. 7 through 9. Two different
liquid nitrogen (—320F), and liquid joint configurations were tested in the
hydrogen ( — 423 F). Details of the longitudinal direction and two in the
apparatus and experimental procedure transverse direction. The static strengths
for performing the tension tests are are results of one or two tests for each
given in reference (8). The axial fatigue condition. The values reported for
tests were conducted on a series of test number of cycles to failure at each
beds equipped with hydraulic rams, temperature and for number of cycles to
cryostats, automatic cycling apparatus, leak at room temperature are generally
and counters. The rate of loading was 6 averages of five tests. The number of
cpm. Leaks were detected by standard cycles to leak at — 320 and — 423 F were
dye penetrant tests. A full description generally determined on only one speci-
of the fatigue test apparatus and testing men and at 50-cycle intervals. Cracks
procedure is given in reference (5), while initiated at the edge of the weld nuggets
82 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
on the outer rows of the spot welds to —423 F. The joint efficiencies con-
where eventual specimen failure also tinuously decreased with decreasing
usually occurred. Figures 7 to 9 show temperature from 78 to —423 F (85 to
some of the fatigue data in the form of 99 per cent at 78 F, 68 to 80 per cent at
S-N curves. - 3 2 0 F, and 47 to 63 per cent at - 4 2 3 F
from 78 to — 423 F, and the joint effi- temperature for types 301 (heat 49061)
ciencies were above 90 per cent for all and 310 stainless steels and 85 per cent
conditions tested. Static strengths of the of typical yield strength for types 304L
complex welded joints of AM-355 stain- and AM-355 stainless steels except for
less steel continuously decreased with those cases in which the static tensile
decrease in temperature with resulting strength of the joint was less than the
joint efficiencies of 84 to 100 per cent at yield strength of the base metal. There-
78 F, 32 to 42 per cent at - 3 2 0 F, and fore, the stress levels for type 301 (heat
27 to 29 per cent at - 4 2 3 F. 49061) at - 4 2 3 F were from 41 to 63 per
stress level than 304L or 310 steels; thus strengths and fatigue hfe of types 304L
they are more desirable for room-tem- and 310 are superior to those of type 301
perature application in which weight is (heat 49061). AM-355 has low joint
a critical design factor. efficiencies and poor fatigue properties
Notched: Unnotched
Yield Tensile Strength Ratio
Tempera- Direc Strength Ultimate Elonga- Obtained from Notched
Stain ess Steel Alloy ture, deg tion" (0.2% Strength, tion, Specimens with Kt
Fahr Offset), psi per cent
psi
The joint fatigue properties at — 320 F at —320 F and would not be considered
are given in Fig. 8. Types 301, 304L, and acceptable for structural applications
310 stainless steel have about the same at this temperature. Figure 9 shows that
fatigue properties for the longitudinal types 304L and 310 stainless possess
joint. For the transverse joint the joint excellent joint fatigue properties at
CHRISTIAN ET AL ON LOW-CYCLE FATIGUE OF WELDED JOINTS 85
— 423 F, whereas the joint strengths and amount of martensite present in this ma-
fatigue life of 301 (heat 49061) decreased terial. Heat 49061 had 76 per cent mar-
considerably from —320 to —423 F. tensite in the as-received material as
There is, however, an appreciable varia- compared to 62 per cent for heat 57644.
Number Number
Temp- Joint Direc- Static of Cycles of Cycles
Stainless Steel Alloy erature, Confi g- tion" Strengtli, psi
Stress Range, psi to First to
deg Fahr uration Leak Failure
f
78 L 222 000 0 to 150 000 473 862
L 0 to 170 000 376 544
L 0 to 190 000 300 420
L 203 000 0 to 170 000 370 590
T 203 000 0 to 132 000 357 405
T 316 366
1 T
0 to 150 000
347
1 -320 L 259 0 0 0
0
0
to
to
170
189
000
000 700 1029
L 0 to 214 000 300 406
L 0 to 239 000 74
301 (Heat 49061) L 244 000 0 to 214 000 147
T 220 000 0 to 165 000 778
T 0 to 189 000 266
T 0 to 214 000 14
-423 L 209 000 0 to 157 000 73
L 0 to 178 000 48 53
L 0 to 194 000 4
L 212 000 0 to 178 000 17
T 164 000 0 to 123 000 79
T 0 to 139 000 22
> T 0 to 156 000 3
86
CHRISTIAN ET AL ON LOW-CYCLE FATIGUE or WELDED JOINTS 87
TABLE ^.—Continued.
Number Number
Temp- Joint Direc- Static of Cycles of Cycles
Stainless Steel Alloy erature, Config- tion" Strength, psi Stress Range, psi to First to
deg Fahr uration Leak Failure
300 000
280 000 01 ( H eat 4 9
260 000 V I D 304
O 3 10
240 000
220 000
A
\ ;
M 35
200 000 N, ^
180 OOO'L , • ^ 1?.
160 000
140 000
X^ L^
V -~>. ^
120 000
-V
100 000
80 000
(a)
60 000
300 000
280 000
260 000 [
240 000 \,
220 000
\
200 000 1
180 000' ^.
160 000
^ i^.'
140 000 ' ^ •^
»T>~-^
120 000
100 000
80 000
(b)
60 000
0 200 400 600 800 1000 1200 1400 1600
CYCLES TO FAILURE
(a) Longitudinal J o i n t N o . 1.
(b) T r a n s v e r s e J o i n t No. 1.
F I G . I.—S-N Curve—Stainless Steels a t 78 F .
if any, martensitic transformation occurs the reason for its poor fatigue properties
in these alloys due to the stabihzing at cryogenic temperatures.
influence of large nickel contents. Also, The number of cycles required to
the presence of nearly 100 per cent produce the first through crack, or leak,
martensite in AM-355 is believed to be was determined by dye penetrants to
88 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
havior. Table V shows that those alloys 310, and AM-355 stainless steels; for
which possessed the most favorable type 304L the reverse was true. Although
fatigue Hfe also withstood a large number transverse joint No. 1 had a different
of cycles from first leak to failure. configuration than the longitudinal
The fatigue life of longitudinal joint joints, the same configuration was used
No. 2 was nearly the same as for longi- for transverse joint No. 2 (type 310 only)
tudinal joint No. 1 at 78 F. At cryogenic as for longitudinal joint No. 1, and a
temperatures the fatigue properties were fewer number of cycles to failure were
only slightly superior for joint No. 1 than still obtained for the transverse direction.
for joint No. 2, indicating that the joint Based upon the results obtained from
efficiency and fatigue properties are the fatigue testing of complex welded
CHRISTIAN ET AL ON LOW-CYCLE FATIGUE OF WELDED JOINTS 89
joints, the following alloys may be rec- joint efficiency, and elongation of fusion
ommended for structural applications: welds; and tension and shear tests of in-
types 301, 304L, 310, and AM-355 at dividual resistance spot welds.
78 F; types 301, 304L, and 310 at The results of notched and unnotched
- 3 2 0 F; and types 304L and 310 at tension tests are given in Table III. The
— 423 F. (Type 301 may be acceptable at unnotched data were analyzed to deter-
— 423 F but, due to the variations of mine possible correlations with the fa-
300 000
280 OOO'1 01 (H<
"°"
260 000 V^ \ n 30 4
0 3 10
240 000
220 000
1
K M 35.
200 000
L
I BO 000
160 000 1
140 000
I 20 000
100 000
80 000 V • ( a ) -
60 000
300 000
280 000 I
260 OOo' Q^.
240 000 ^
220 000 ^ r^ ^v;
^
200 000
ISO 000
160 000'
140 000 I
120 000 \,
100 000 \
80 000
60 000
..^ •(b)-
40 000
200 400 600 800 1000 1200 1400 (600
CYCLES TO FAILURE
properties of the alloy from one heat to tigue life of the complex welded joints.
another, it requires the evaluation of The decrease in tensile strength of AM-
each heat of material prior to use.) 355 from —320 to —423 F indicates a
Several types of evaluation tests were possible embrittlement of this alloy at
conducted to see which test or tests best — 423 F. Changes in per cent elongation
predicted the fatigue life of the alloys correctly predicted a decrease in tough-
investigated at cryogenic temperatures. ness of 301 (heat 49061) at - 4 2 3 F and
The types of tests used included three AM-355 at - 4 2 3 F (also - 3 2 0 F for
different notched tension tests, having the transverse direction); however the
stress concentration factors of 3.2, 6.3, decrease in elongation of 304L at — 423 F
and 19; the determination of strength, did not correlate with fatigue Hfe nor did
90 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
REFERENCES
(1) R. J. Favor et al, "Investigation of Fatigue ASTM STF No. 302, Am. Soc. Testing
Behavior of Certain Alloys in the Tempera- Mats., p. 129, March, 1962.
ture Range Room Temperature to —423 F," (5) J. L. Christian, "Physical and Mechanical
WADD TR 61-132, June, 1961. Properties of Pressure Vessel Materials for
(2) J. W. Spretnak, M. G. Fontana, and H. F. .Application in a Cryogenic Environment,"
Brooks, "Notched and Unnotched Tensile ASD-TDR-62~258 March, 1962.
and Fatigue Properties of Ten Engineering (6) R, E. Peterson, Stress Concentration Design
Alloys at 25 C and — 196C," Transactions, Factors, Appendix A, John WUey and Sons,
Am. Soc. Metals, Vol. 43, p. 547 (1951). Inc., New York, N. Y., pp. 136-138 (1953).
(3) M. G. Fontana, "Investigation of Mechani- (7) Heinz Neuber, Theory of Notch Stresses,
cal Properties and Physical Metallurgy of English Translation, J. W. Edwards, Ann
Aircraft Alloys at Very Low Temperatures, Arbor, Mich. (1946).
Part II: Strength Properties and Hardness," (8) J. F. Watson and J. L, Christian, "Cryostat
USAF Air Material Command Technical and Accessories for Tension Testing at
Report No. 5662, October 18, 1948. — 423F," Materials Research & Standards,
(4) J. F. Watson et al, "Correlation of Notched: Vol. 1, No. 2, p. 87, Feb., 1961.
Unnotched Tensile Ratios with Tensile (9) .\. Hurlich, "Mechanical Model Testing,"
Fatigue Properties of Complex Welded Proceedings, Seventh Sagamore Ordnance
Joints in High Strength 300 Series Stainless Materials Research Conference on Mechani-
Steels at Cryogenic Temperatures," Sym- cal and Metallurgical Behavior of Sheet Ma-
posium on Evaluation of Metallic Materials terials, Aug. 16-19, 1960, published by
in Design tor Low Temperature Service, Syracuse University Research Inst.
STP338-EB/Sep. 1963
E F F E C T OF STRESS STATE ON H I G H - T E M P E R A T U R E
LOW-CYCLE FATIGUE
B Y C. R. KENNEDY1
SYNOPSIS
peratures. The experimental work con- pressure that is thermally cycled in the
sisted of two test series on Inconel tubes axial direction. The stress states applied
at 1500 F. must, therefore, produce an axial strain
cycle about zero strain with the internal
PROCEDURE
pressure held constant. This was done
The material used is the same as that with the stress states shown in Table I.
used in previous investigations of creep All tests were performed with the tan-
under complex stresses (8) and in uni- gential stress equal to 4000 psi for the
axial low-cycle fatigue (3). Specimens complete cycle. The half-cycle square-
were machined from J-in. schedule-40 wave stress states were calculated to
seamless Inconel pipe as shown in Fig. produce principal strains equal and op-
posite in the axial direction; however,
-0.06010.00025
, 50" the radial and tangential strains were
RAD
VB
not equal and opposite. This type of
.^CZT" testing is comparable to standard fatigue
i • • • ' 1
Test
Axial stress. Tangential Radial Stress, Axial Stress, Tangential Radial Stress,
Stress, OQ Stress, (r0 "R
The plastic strain per cycle did in- average plastic strain ranges, considering
crease from the first half cycle, remained all cycles to failure. The effective plastic
constant for over 75 per cent of the cycles strain per cycle, e, was calculated from
to failure, and increased again before the axial strain using the relation-
fracture. The values reported are the ships in Eqs 2:
Alternating
Shear Stress, Shear Strain
Effective Stress Effective Strain Time to Failure Number of
per Cycle, A7,
Test
or( ffi
psi'
2
g-A
ffi —- <Ti\
/'
or (e. - esf,
per cent
per Cycle, ^, per Cycle,
A€, per cent
If , hr Cycles to
Failure, N/
4x10^
-i UNIAXIAL FATIGUE
• TORSIONAL FATIGUE
o UNIAXIAL TUBE CREEP
DATA
* INTERNAL PRESSURE
PLUS A X I A L CYCLE ,
10 10 10 10 10
EFFECTIVE STRAIN RATE , t , per cent per hr
2x10"
! 1 1
i
!
10" 1
—t[-f 1 _i-1
8 ^.' 1 .r'*!
I 1 1
^H
6 t-p-rr^ o
<
i .°§ 11
•s.
)
1 0
1
! 1 S.j|!
4 AXIAL UNIAXIAL TES rs ; rii
o TANG ENTIAL UNIAXIA L TESTS ' i Ii 1 . X-
o :c MP LEX STRESS-RUPTURE TESTS ^ATIC OF 00 TO •2) :
1 1
1
1
10"
0
10 10 .(
10 10 10 10"
RUPTURE TIMEx-|^,hr
2x10^
10 2 5 10 10
RUPTURE TIME x - j - , s e c
Hull and Rimmer (11) have reported developed to satisfy the requirements of
that void formation under the same shear considering both the shear and tensile
stress is a simple function of the maxi- stresses:
mum tensile stress. From this study and
the other evidence cited, it was apparent
that a shear stress and its accompanying
deformation are responsible for the
(0'^) .(5)
where:
KKNNEDY ON EFFECT OF STRESS STATE ON LOW-CYCLE FATIGUE 99
4x10'
o UNIAXIAL FATIGUE
• TORSIONAL FATIGUE
V STATIC TORSION TEST " ^
A PRESSURIZED TUBE AXIALLY
CYCLED FAILURE NORMAL TO -.'-r-
AXIAL STRESS
* PRESSURIZED TUBE AXIALLY CYCLED '
FAILURE NORMAL TO TANGENTIAL STRESS
3x10"^
0.1 10.0 100 1000 10,000
static creep and two failed as a result of data from the complex stress tests using
a cyclic axial stress. Results of tests 1 a parameter based on effective strain
and 2 are plotted in Fig. 7 with the other range per cycle. These data demonstrate
fatigue tests; however, since failure was that both the effect of frequency and
a result of a static load, they should be stress state must be considered before a
compared with the data in Fig. 5. If this direct comparison can be made.
is done, these two points fall in the Although it has been demonstrated
middle of the scatter. Tests 3 and 4, how- that the cyclic creep test results do not
ever, failed due to the axial tensile stress, follow the static creep results implicitly,
and, as shown in Fig. 7, the rupture static creep-rupture relationships can be
lO'-
<
a:
? ID"
ISl
< 10"'
Q.
11
UNIAXIAL FATIGUE,
TORSION FATIGUE
PRESSURIZED TUBES AXLALLY
to" CYCLED, FAILURE NORMAL TO -la ^t
AXIAL STRESS ^ "3:
PRESSURIZED TUBES AXlALLY CYCLED, ^
3
FAILURE NORMAL TO TANGENTIAL STRESS
10
10" 10' 10^ 10-=' IC 10=
NUMBER OF CYCLES TO FAILURE
Figure 4 indicates that the fatigue should be equal to aj^. This is not the
data do not follow the static creep data case, since aj^ = 1.45 and n from Fig. 8
exactly; however, the lack of agreement is equal to 0.762. Regardless of the agree-
does not appear to be associated with ment of the cyclic creep with the static
frequency or stress state. The cyclic creep constants, a and /3 are almost equal
creep tests can then be represented by: to one another and the values of a./^ and
(a — )3)/|8 will not vary significantly.
(15) Thus, the effects of stress state and fre-
'""'^ = ( l ) quency can be generalized.
where: The solution for the second case, that
of a static stress causing failure, is ob-
Acj, = effective plastic strain range, and tained by combining Eqs 15, 16, and 17
Ic = time per cycle. in a similar manner to obtain:
Equation 5 is solved for the effective
a \«lf
stress to obtain, iV/""* At
>-©1-fe) .(21)
'-(ra)"'
The time, frequency, and number of
.(16) Thus, to compare the results from tests
1 and 2, in which fracture resulted from
the static tangential stress, to uniaxial
cycles are related depending upon the cyclic data, both Eqs 20 and 21 must be
type of stress cycle. For those cases used. A comparison is accomphshed by
where the specimen fails as a result of a considering the effect of differing fre-
tensile stress acting on the specimen for quencies and stress states on tests with
half of the cycle, the relation is: the same plastic strain range. The com-
parison for tests 1 and 2 with the uni-
.(17) axial fatigue data is
2v '•
where 17 = frequency.
.(22)
For those cases where the failure is a ^iii \nu/ \5/ii
result of a static tensile stress acting on
the specimen, the relation is: (Subscript I for uniaxial data and sub-
script II for tests 1 and 2.)
N,
= I,. .(18) The value of the exponent for Inconel
at 1500 F was determined to be
Considering the case of cyclic stresses,
Eqs 16 and 17 are combined to obtain: = 0,31. .(23)
'-(r,rC^)"' (19)
Since the exponent is less than 1, the
number of cycles to failure should not be
This equation can now be substituted seriously affected by small changes in
into Eq 15: frequency. The factors obtained from Eq
22 for tests 1 and 2 are:
(20)
The same type of comparison is made Inconel at 1500 F to furnish failure test
for tests 3 and 4 to yield the factors data under complex dynamic stresses at
high temperatures. The results obtained
\VinJ Vo'/iil
from these series have been correlated
with the results of previous static and
Test 3. 4.4 X 1.11 = 4.94 dynamic stress tests. This investigation
Test 4. 3.7 X 1.13 = 4.18 has furnished a general method of solv-
(Subscript III for tests 3 and 4.) ing complex stress creep problems in-
The comparisons can now be made, volving both static and dynamic stress
and the results are shown in Fig. 8 in states. The general solution is based upon
which the factors determined above are the developed relationship.
multiplied by the number of cycles to
failure. As expected, the data do not
correlate exactly owing to the discrepan- '-(?)'(•,)•
cies between the fatigue and static data. and upon Minor's damage accumulation
This method of adjusting data with hypothesis. The damage accumulation,
respect to stress state and frequency however, was shown to be independently
does, however, demonstrate the magni- additive in each principal stress axis.
tude of these effects on life expectancy. The conclusions of this investigation
The torsional results can be compared are:
in a similar manner using Eq. 20 to ob- 1. An expression has been developed
tain: for creep rupture under complex stress
states.
^ / i = A^/iv (r)^ .(24) 2. A method for generating true stress
rupture data from constant load data is
(Subscript IV for torsion test data.) given.
or 3. The effect of cyclic stresses upon
strain rate is demonstrated.
Nn = Nnv (0.577) (25) 4. Results of the two test series
demonstrate the independence of damage
The broken line in Fig. 8 is the strain accumulation in each principal stress
fatigue data expected from testing under axis.
cyclic torsion. As seen in this figure, the 5. A method for determining the
data all fall slightly above this line on magnitude of the frequency effect on the
the conservative side. The relative agree- number of cycles to failure is given when
ment adds further support to the applica- the plastic strain range is used as a cri-
bihty of this method to determine the terion for fracture.
magnitude of the effect of stress state 6. A method is given for determining
upon the strain cycle behavior. the magnitude of the stress state effect
SUMMARY AND CONCLUSIONS
on the cycles to failure when the plastic
strain range is used as a criterion for
Two test series were performed on fracture.
REFERENCES
(1) L. F. Coffin, Transactions, Am. Soc. Me- as a Criterion for High-Temperature De-
chanical Engrs., Vol. 76, p. 931 (1954). sign," in High-Temperature Materials,
(2) C. R. Kennedy, "Plastic Strain Absorption Their Strength Potentials and Limitations,
104 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
the range of stress that is greater in the did in our laboratory, which Mr, Bell
unnotched than in the notched specimen cites in his paper, would suggest that the
data. Clearly, the ratio factor is thus stress concentration factor reduces as a
reduced in accordance with the curves. result of plastic flow. If this is truly the
Although I have no strong opinion to case, then Mr. Symonds' comment is a
express as to the cause of this phenome- valid one. You would expect this factor
non, I seem to remember that others to level off or reach its maximum value
have recorded failure data of this type. when you are definitely in the elastic
I think perhaps Mr. Grover produced range. I should like to suggest to those
fatigue strength reduction curves of the who are finding drop-offs of the nature
same form as shown in this paper. under discussion here, that if they were
M R . H . J. GROVER^ (panelist).—Some to assume, as Mr. BeU did in his paper,
years ago we conducted some experi- that a maximum value does exist and
ments on a box-beam type of structure. persists over some range of lifetimes or,
In connection with these, we also tested preferably, range of elastic stresses, then
a number of coupons of the material it does not take a major refairing of the
from which the beam was made, of which S-N curve to straighten out the curva-
some were notched. I n general, we found ture. To do this one might have to adjust
in the notched coupons, and more the S-N curves by 1000 psi or so. Quite
particularly in some of the riveted ones, frequently our data are no more reliable
there was redundancy and the possibility than that. I do not know how many
of load sharing and redistribution, but tests Mr. Bell made in this particular
notably in the box beam, the same kind case nor the amount of refairing that
of a curve for the fatigue strength re- might be required. However, I think that
duction factor as Mr. Bell has indicated. this technique might work in many
I am not quite sure why this occurs, b u t cases.
I should like to add that we found some- M R . GROVER.—In the box-beam tests
thing equivalent to this in a structure, mentioned previously and in some other
so that it may have practical implica- experiments I have seen, I do not be-
tions in structures at long lifetimes. lieve refairing of the S-N curve is a valid
M R . LEONARD M O R D F I N . ^ — M y paper explanation of the variation of Kf ; there-
on fatigue tests of box beams does not fore at the risk of disagreeing with Mr.
include a figure of Kf versus cycles to Hardrath, I do not think this technique
failure, but we also found that the fatigue would work for Mr. Bell's data.
strength reduction factor started to fall MR. C . R . SMITH.*—One of the most
oii a t high cycles. confusing things in fatigue is the effective
M R . H . F . HARDRATH.'—Discussion of stress concentration. Perhaps a more
how the stress concentration factor reliable understanding could be had by
varies with life might be putting the visualizing the concentration in terms of
shoe on the wrong foot. I tend to look strain rather than stress. We would un-
at stress concentrations as being stress- doubtedly find that the strain at the
dependent. Some of the work that we notch was nearly the theoretical concen-
tration times the nominal strain, even
^ Research Supervisor, Battelle Memorial though localized plastic deformation
Inst., Columbus, Ohio.
* Aeronautical Research Engineer, Engineer- occurred during the first cycle.
ing Mechanics Section, National Bureau of M R . HARDRATH.—I agree with most of
Standards, Washington, D. C.
' Fatigue Branch, National Aeronautics and ' Fatigue Laboratory, General Dynamics/
Space Administration, Langley Station, Va. Convair, San Diego, Calif.
PANEL DISCUSSION ON LOW-CYCLE FATIGUE 109
what Mr. Smith has said, but I think the lower stresses and long endurances at
point under discussion here has to do low mean stresses. Under the conditions
with what is happening in the elastic of long endurance and zero mean stress,
range, in which we would not find we did not experience any looping type
measurable plastic deformations. When of plastic deformation with the notched
we go to higher stresses and shorter lives, specimens; however, these remarks all
this effect comes into play and can be refer to macroscopic deformation. It was
rather significant. The question was all "as observed" by microscope and
asked primarily with respect to com- naked eye, no close measurements being
pletely reversed loading. For this type made, and it might well be that an inter-
of loading the plastic actions are prob- action between neighboring holes does
ably symmetrical. Mr. Bell could choose occur long before there is any noticeable
to talk about a strain concentration fac- deformation.
tor or a stress concentration factor, but MR. W . H . MUNSE^ (panelist).—
he would find changes in value in the There are several questions I should like
plastic range either way. to ask Mr. Bell. The first concerns the
MR. GROVEE.—In the box-beam ex- Fig. 7(b) we have been discussing. I
periments we found a more certain maxi- wonder if Mr. Bell has any comments as
mum in the Kf-vevsus-N curve when to what might happen to some of his
there was some redundancy in our speci- information if it were to be extended into
men. Mr. Bell had some specimens with the negative mean stress range. There
double notches, and I wonder if he has has not been much work done in this
any comment concerning possible dual direction; however, I know that there has
modes of cracking for such specimens. been a little thought given to this sub-
MR. BELL.—All notched specimens ject. Some of the information I have seen,
used in these tests had at least two holes; indicates that these fatigue relationships
the Haigh machine data, which are of begin to change very markedly in the
particular interest, are for three-hole negative mean stress range, particularly
specimens. Figure 6 of my paper shows when the members contain stress con-
some types of plastic deformations which centrations such as exist in the specimens
occurred in the notched specimens. At shown in Fig. 7(b). The second question
high stresses these deformations tend to concerns the type of information that
encompass the entire cross-section of the was obtained in these tests, particularly
specimen. Three typical examples of at the lower number of cycles. I wondered
plastic deformation are shown in this whether there were any strain data ob-
figure. The smallest or least noticeable tained by which this information might
type of deformation appeared as small be examined in terms of strain rather
dimples at the sides of each hole and than stress. At the lower number of
was, in this respect, highly localized, as cycles we have often seen that strain
might be expected. With increasing provides a much better correlation,
stressing conditions, there was a tend- whereas with stress we get rather flat
ency to develop a more extensive type curves, such as those in Fig. 5 of Mr.
of deformation pattern that developed Bell's paper.
in a looping action between the holes MR. BELL.—I believe in your first
and extended to the edges of the speci- question that you were looking for a
men as two bands at 45 deg to the speci- theoretical approach to what is happen-
men major axis. It is not necessarily true
' Professor of Civil Engineering, University
that this can be tied in exactly with the of Illinois, Urbana, 111.
no SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
directed to those who have discussed the which is small compared to the plate
two different types of fracture obtained thickness will result in flat fracture,
at short hfetimes. Taking one particular whereas a shear-type fracture will occur
example, it appears from Fig. 10 of Mr. when the plastic zone size is large com-
Whiteson's paper that he experienced pared to the plate thickness. The plastic
greater reduction in area with the ductile zone size is also a function of the stress;
type of fracture. We noticed, as was therefore high stresses result in a large
shown by the cyclic creep curves as we plastic zone together with faster crack
call them, that a considerable increase in propagation and shear-type failure. On
the total extension of the gage length of the other hand, the lower stress levels
the material was found with the ductile result in the opposite type of material
or cyclic creep type of fracture. Do the behavior.
observations of other experimenters in- It is felt that this approach should be
dicate that the more ductile type of given greater consideration when at-
fracture is always associated with an tempting to analyze fatigue data. Paris
increase in the elongation of the gage and Anderson of Boeing have used a
length? fracture mechanics approach in the
MR. WHITESON.—No strain measure- analysis of fatigue studies of aluminum
ments were made during the low-cycle panels, with good results. Their results
fatigue test; also, the fatigue specimens show the more familiar flat fracture and
were a standard type of specimen con- were run at lower values of the crack tip
figuration as shown in Fig. 2 of the paper. driving force.
It should be noted that there is no realis- MR. H . B . DUNTHORNE.I^—I would
tic gage length that could provide reason- like to ask a very general question that
able measurement of uniform elongation. would combine the talents of the entire
The tapered shape of the specimen forces panel as well as a little prophetic work.
all failures to take place in a very lo- As we move into the supersonic area in
calized area; thus, any deformation is aircraft, we shall experience a lot of
quite localized. thermal cycling, a very wide range of
MR. C. M. CARMAN" (panelist).—The temperature, and designs operating very
appearance of plastic deformation, re- close to the yield strength; therefore, the
duction of area, or shear-type fracture cycles to failure are bound to be low.
observed during high-stress, short-hfe Because of the thermal cycling nature,
fatigue tests as compared to a mixed the frequency will be relatively low. We
mode or flat fracture observed for the will be in temperature ranges in which
fatigue tests conducted at lower stress it has been found that a thermally in-
levels appears to be consistent if one duced strain aging metallurgical in-
uses the fracture mechanics viewpoint. stability type of phenomenon can occur,
In the fracture-mechanics, steady- and this has sometimes led to unusual
state model described in the paper brittleness. When all these factors are
"Low-Cycle Fatigue Characteristics of combined, would the panel care to offer
Ultrahigh-Strength Steels," the rate of an opinion, based upon their many
crack propagation is expressed as a func- researches, as to when we could expect
tion of the plastic zone size. The size of any untoward effects, any cumulative
the plastic zone also determines the frac- results, showing a sudden or sharp
ture appearance. A plastic zone size loss in the otherwise expected properties?
'^ Metallurgist, Research and Development " Metallurgy Department, Curtiss-Wright
Group, Frankford Arsenal, Philadelphia, Pa. Corp., Woodbridge, N. J.
112 SYMPOSIUM ON FATIGUE or AIRCRAFT STRUCTURES
MR. SYMONDS.—In our normal low- location, or was the fact that it occurred
stress, high-cycle, fatigue-type failures, right in the center significant?
we have a mechanism of failure for MR. WHITESON.—Analysis of the failed
which there is still considerable discus- specimens using electron fractographic
sion with respect to exactly how the techniques (see the accompanying Fig.
fatigue fracture starts. However, the 1) indicates that no fatigue-crack pro-
failure is usually associated with a surface gression occurred when fracture was
type of discontinuity, an inclusion, or internally nucleated. The fractographic
FIG. 1.—Typical Electron Fractograph Taken from Center of Internally Nucleated Fatigue Frac-
ture Surface (X 18,490).
some other type of stress raiser. But analysis showed no difference in fracture
with respect to this other phenomenon surface appearance between a static
of the fatigue crack starting right in the tensile failure and the internally nucle-
center of a specimen that is loaded to a ated fatigue failure. Surface-nucleated
fairly high stress level and that fails at a fatigue failures showed the normal
low number of cycles, would anyone arrest lines associated with high-cycle
care to comment on what type of fatigue type of fatigue failure, indicating that
crack growth mechanism was occurring stable fatigue crack propagation took
in the center of this particular specimen? place before catastrophic failure of the
Would this be considered a random specimen (see the accompanying Fig. 2).
phenomenon that could occur at any MR. SMITH.—The fact was mentioned
P A N E L DISCUSSION ON LOW-CYCLE FATIGUE 113
that the necking type of failure origi- years ago and getting a normal S-N
nated in the center. Was that estabHshed curve at room temperature. At 450 F,
in fact, or could it be under the surface, however, the tests produced a near-
but not necessarily at the center? If it straight-line curve with essentially no
were not, then it is possible that we slope out to 10^ cycles. The failures in
might have sources for nucleation al- this instance had an appearance identical
most anywhere inside the material. to that of creep fracture, with consider-
M R . WHITESON.—Electron fracto- able necking down and no evidence of
FIG. 2.—Typical Electron Fraclograph Taken from Initiation Point of Surface Nucleated Fatigue
Fracture Surface (X25,900).
Note the fatigue arrest markings.
seen that there is a flat portion in the able to reduce this creeping mechanism.
center, and in the following figures pro- The resulting fracture appeared as a
gressive fatigue-type cracks are shown. standard fatigue-type of fracture, and
Both types were found in the tests. the lifetime was extended to that which
M R . T . J. DOLAN1«—This question is would be anticipated by extrapolation of
one that should be related to the existing the short, steeply sloping portion of the
stress state. I t is well known that when a standard fatigue curve. There seems to
specimen begins to neck locally, the be more evidence in favor of the hy-
stress state changes rather drastically. pothesis of having two separate failure
You are really building a circumferential mechanisms, both of which can progress
notch into the specimen as it begins to simultaneously under certam stress con-
neck down. The interior then is subjected ditions. An extrapolation from this
to a rather complex three-dimensional might be that you simply have two
state of stress which varies with the criteria for failure and it is a matter of
degree of reduction caused by the neck- which one reaches its critical condition
ing down. Under cycHc loading at the first.
higher stresses, the necking may cause M R . HARDRATH.—Is it reasonable to
fracture by a high degree of three-di- expect to see crack progression at these
mensional stress in the center, initiating extremely high stresses? I t seems to me
the interior fractures, whereas at slightly that, in the fatigue failure mechanism we
lower stresses, necking is restricted and are discussing, some phenomenon not
you are repeating a surface stress which completely understood makes the first
becomes the criterion for failure. I t seems couple of atoms let go of each other to
quite logical that one can expect either initiate a crack. This crack then grows in
of these failure types, depending on the some manner for some time, and the net
amount of necking which occurs. This effect is to reduce the residual static
would mean, of course, that the type of strength of the specimen. Working this
failure would vary considerably from one specimen through rather heavy plastic
material to another depending on the deformations depletes a large amount of
degree of ductility exhibited by the the ductihty in many cases. Conse-
material. There has been evidence even quently, the sensitivity to whatever
in ordinary static tests that ductile and crack is present is very much increased.
semiductile materials frequently fracture Since we are working very close to the
from the interior rather than from the ultimate tensile strength of the material,
surface after they have necked down and a very large crack should not be needed to
caused this complex three-dimensional reduce the static strength to the stress
interior state of stress. level being applied to the specimen.
M R . B E L L . — I n some of his previous Thus, it seems to me that we should not
work, Mr. Benham found that under expect to see much crack progression at
certain conditions in which he obtained these high stress levels; and perhaps the
a cyclic creep-type failure during which over-all mechanism is not necessarily
he was observing cyclic extension and different from that which applies at
cyclic strain, by slightly decreasing one lower stress levels.
of the minimum stresses by only about 1 MR. SYMONDS.—If we had a tiny
per cent of the minimum stress, he was crack starting and propagating rapidly,
we might, at random, expect our crack to
" Professor and Head, Department of The- start occasionally at the surface of these
oretical and Applied Mechanics, University' of
Ilhnois, Urbana, 111. high-stress low-cycle specimens. We
P A N E L DISCUSSION ON LOW-CYCLE FATIGUE 115
would find a certain scatter with re- take exception to the remarks of Mr.
spect to where this crack did first start. Schijve that, in general, creep fractures
However, it does seem to be the opinion are intercrystaUine and fatigue fractures
of the investigators that the crack had a are transcrystalline. T h a t assumption
tendency to start at the center of the cost the Bell System a lot of money
specimen, or at least that the total before it was demonstrated that this is
fracture surface tended to look as if it not necessarily true. When cable sheath
were a static specimen. Mr. Whiteson was shipped across the country a lot of
comments very appropriately that we fractures developed which were trans-
might be describing a situation in which crystalline, whereas normally fatigue
the reduction in area, due to necking cracks in cable sheath are intercrystal-
down, was such that a static fracture line. It was therefore assumed that these
occurred after a few hundred cycles. failures during shipping were not fatigue
M R . SCHIJVE.—We have tested un- cracks. However, upon study it was
notched aluminum alloy sheet specimens found that when fatigue was due to low-
at endurances as low as about 5000 cycle straining, such as occurs in cable
cycles. We fitted a set of two microscopes sheath due to daily and seasonal tem-
to the fatigue machine and were able to perature changes, the failures are inter-
discern readily the amount of crack crystaUine—that is, they resemble creep
propagation starting at the surface. Our failures in appearance—but high-cycle
present definition of whether a failure is failures induced by flat wheels on freight
a creep failure or a fatigue failure is cars are transcrystalhne. Studies on
that a fatigue failure generally is sup- lead sheath and aluminum sheath tested
posed to be transcrystalline and a creep at J cpm for lives ranging from 200 to
failure, in general, is supposed to be 40,000 cycles (the latter represents 100
intercrystaUine. Have Mr. Bell and his yr life) indicate that almost invariably
staff made any observations of this these materials fail in fatigue with an
type, especially in the low-cycle fatigue intercrystaUine type of failure. In such
range? Did he observe any crack propa- tests the materials are strained in the
gation at all and, if so, could he say plastic range. The transition and the type
whether this was transcrystalhne or of failure appear to be dependent upon
intercrystaUine? whether the test is made in the plastic
or the elastic range. If fatigue failures
MR. BELL.—We did not make any
occur in the elastic range the type of
microscopic examinations of the speci-
failure is predominantly transcrystalline,
men. However, for those specimens which
but if the failure occurs in areas that are
did finally fail in a ductile manner and
strained plastically the type of failure
produce a fracture with a ductile ap-
seems to be predominantly intercrystal-
pearance, there was no observed propa-
line. This has been observed in high-
gation of the crack. We observed these
purity aluminum (99.99+ per cent
only with a magnifying glass, but we
aluminum) down to the commercial
were able to look continuously and quite
grade ( 9 9 + per cent aluminum) and in
closely at the slow-cycle tests at 10
many of the leads and lead alloys, such
cycles, 15 cycles, and up to 100 cycles.
as pure lead, chemical lead, lead-silver,
There appeared to be a continued in-
lead-calcium, and lead-antimony. This
crease in the strain and a final fracture
which gave all the appearance, at the
last cycle, of a tension test. ^' Supervisor, Mechanical Testing Laboratory,
Bell Telephone Laboratories, Inc., New York,
M R . G . R . G O H N . " — I would like to N. Y.
116 SYMPOSIUM ON FATIGUE or AIRCRAFT STRUCTURES
has been observed over and over again loading and things of this nature. Some
so that I believe the wrong conclusion of the fatigue tests were run at much
would be drawn if attempts were made shorter durations. One run at 0.18 hr
to distinguish between fatigue and creep resulted in a stress of about 29,000 psi;
failures on the basis of whether the failure this is much higher than the actual
is transcrystalline or intercrystalline. ultimate tensile strength normally de-
MR. SCHIJVE.—I think Mr. Gohn is termined for the material.
quite right in saying that for the more MR. GROVER.—In the past several
pure materials the intercrystalline failure years there have appeared many ad-
is much easier to initiate than the trans- vantages in considering the stages of
crystaUine failure. I also agree that in- crack initiation and of crack propagation
creasing the stress amplitude may favor for fairly long-cycle fatigue, although the
the intercrystalline failure. Increasing defining line between them is not clear.
the temperatures also favors the inter- To what extent is this sort of thinking
crystalline failure. Whether failure is a useful in low-cycle fatigue? To what ex-
creep failure or a fatigue failure might be tent is it desirable to study carefully the
a question of definition. The definition I crack propagation stage? Does anyone
gave before could not be generalized, have any ideas about the cycle limits
but it seems to work in our experience where this procedure would begin to be
for those materials that have a limited less helpful?
amount of ductility, which usually MR. SMITH.—It does not appear worth
applies in materials used in aircraft while to do much study at the crack
structures. propagation rate at low-cycle fatigue
MR. SMITH.—We are actually talking except as it influences the propagation
about creep. This may be a little beside rate at the low level or long-cycle fatigue
the point, but I have a question on Mr. strength. For example, a very high load
Kennedy's paper. It seems that all his would introduce a residual compressive
data start at about f hr for rupture time. stress at the end of a crack which would
Were there any shorter times? certainly retard the crack growth at
MR. C . R . KENNEDYIS (panelist) .—The lower loads. To study this problem,
definition of an ultimate strength at this quantitative measurements of the re-
particular temperature is dependent on sidual stress produced as a result of
the stress-strain test itself. This particu- these high loads would be required. As
lar specimen was loaded to about 95 per yet, there is no way of measuring this,
cent of the ultimate tensile strength. The but there is some talk about getting
normal tension test is run at about a strain gages that measure stress instead
strain rate of 100 per cent per hr with a of strain. This might be a good place for
resulting close representation of an one.
ideally plastic material. It starts elas- MR. SCHIJVE.—Our study of crack
tically, begins yielding, and the strain propagation under low-cycle fatigue was
hardening coefficient is very small. It to see whether a fatigue mechanism
flows out at the stress that will produce that was valid for low-stress-amplitude
the strain rate that is introduced by the fatigue could also be applicable to high-
tension test, and creep tests were not stress-amplitude fatigue. We were in-
run faster than this. To do so would terested in this problem in view of the
involve some question as to the time of well-known cumulative damage problem;
^* Metals and Ceramics Division, Oak Ridge
and restricting our studies to an alumi-
National Laboratory, Oak Ridge, Tenn. num alloy, we observed that there was
PANEL DISCUSSION ON LOW-CYCLE FATIGUE 117
probably not very much difference in range of crack propagation, even where
fatigue behavior. It is generally supposed the constant stress condition exists, we
that the fatigue phenomenon at high find a linear rate of crack propagation.
endurances and low amphtudes is quite This type of information may be of con-
a localized phenomenon and not a bulk siderable help when we reach the design
phenomenon. However, at high stress of larger structures.
amplitudes, speaking in terms as used Another part of this program was
in the paper by Munse and Yao, start- concerned with the strain patterns in
ing from plastic strains, you are definitely the members under the different loading
considering bulk properties. Still, we did conditions. Photostress and strain gage
observe crack propagation at those high studies were made to check these condi-
stress amplitudes, but I would not like tions and to show how these fields change
to suggest that at these high stress as the cracks propagate. These strain
amplitudes fatigue is a localized phe- patterns verify the assumed condition of
nomenon. On the contrary, we did ob- constant stress. A constant stress is ob-
serve quite a lot of fatigue cracks all tained if the compressive load is based
over the specimen. I do not think this on the original net section and if the
will answer the question raised by Mr.
Grover. At any rate, this is our evidence.
MR. MUNSE.—In recent studies at the
University of Illinois on the question of Bending
Moment
crack propagation, consideration was
given to what type of loading cycle
might be used in studies of this problem. Cycles, N
In a number of cases in which we used a FIG. 3.—Bending Moment Required to Bend
constant-load cycle in our tests, the the Specimen.
stresses changed remarkably as the
cracks propagated. This had a marked tensile load is based on the changing
effect on the type of behavior that we section. However, this matter needs a
observed. It is rather difficult to apply great deal more study before it can be
this type of information to an actual tied directly into the behavior of an
structure that may be much larger than actual structure.
the test member. To overcome the MR. GOHN.—Is there a significant
problem of increasing stress we con- difference between crack initiation and
ducted tests in terms of a constant-net- crack propagation? Crack propagation
section stress and a constant-stress load- in thin sheets was studied in terms of
ing. In the latter case the applied load loss of bending moment using a device
was adjusted on the basis of the changing that strains a specimen through a con-
net section in tension and the original stant amplitude and simultaneously
net section in compression, thus giving a indicates the bending moment required
constant stress range during the test. to bend the specimen. If the bending
We found quite a difference in the crack moment is plotted against the number of
propagation rates under these different cycles (this varied from one to something
loading conditions. less than 10,000 cycles), three types of
Looking at the crack propagation curves can be obtained as shown in the
data, we found a small region that accompanying Fig. 3. Depending upon
initially has a changing rate of crack the initial strain, cases were observed
propagation. Then, over a rather long (see Fig. 3) in which (1) there is no
118 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
damage at all (curve / ) ; (2) there was tion rather than two as we have been
some improvement and then the bend- discussing.
ing moment fell ofi (curve 2); and (3) MR. WHITESON.—I concur with Mr.
there was apparently some cracking Symonds. In the high-stress, low-cycle
which started immediately and propa- fatigue evaluation, the maximum fatigue
gated very rapidly as indicated by the stresses were generally within 98 per cent
loss in bending moment (curve J). of tensile ultimate stress or higher. A loss
For the same material, it is possible to of only 2 per cent of the cross-sectional
go from one of these representative area would increase the net section stress
types of failure to the other merely by to a level where catastrophic failure
changing the speed at which the test is could occur. At this point, it then be-
made. These phenomena were observed comes a problem of the material's notch
at testing speeds ranging from a 4-min sensitivity, or its ability to withstand
cycle to 75 cpm.^" the presence of a small surface flaw,
MR. SYMONDS.—It might be very without showing a decrease in load-
possible that we really should be defining carrying ability. This would vary con-
three different regions of fatigue crack siderably with the material and the
propagation. The first region would be testing environment. Fracture toughness
at the extremely low cycles to failure in considerations might be a good model for
which would be found a certain amount determining critical flaw size for fast
of necking down in the specimen. The fracture under high-stress, low-cycle
propagation that Mr. Whiteson de- fatigue once a fatigue crack has initiated.
scribed would be of this type. We would I would like to ask Mr. Schijve if he
get necking down and a fatigue failure saw any reduction of area on those
that would appear to be a static failure. specimens he tested that had surface
The second region would occur if we nucleated fatigue cracks?
tested at a stress level slightly below the MR. SCHIJVE.—There was almost no
previous level, such that we would not reduction of area.
have any appreciable necking in our MR. WHITESON.—This would also con-
specimen. In this case, we would find firm our observations that once a surface
that a typical surface propagation of the crack has nucleated, failure is accom-
fatigue crack would start. We would panied by very little reduction of area.
then find a crack propagation mechanism MR. R . F . BRODRICK.^"—Our observa-
such as Mr. Hardrath described. Once tions were similar for Inconel specimens
the fatigue crack started, the net reduc- that were subjected to completely re-
tion in cross-sectional area due to the versed strain of an amplitude resulting
fatigue crack would appreciably affect in failure in about 500 cycles. There were
the strength of the specimen and we a multitude of cracks, and some of those
would then get extremely rapid crack adjacent to the ultimate failure were
propagation to failure. Finally, the third quite large. Although no accurate meas-
region would be the typical low-stress, urements were taken, some of these
high-cycle fatigue crack propagation cracks covered about 30 per cent of the
mechanism in which it would take a long area. Furthermore, for those specimens
time for the crack to grow. Thus, there with relatively long, straight, slender sec-
might really be three areas of investiga- tions, the area of maximum reversed
strain seemed to move along the speci-
^' "Discussion on Bending Fatigue Strength,"
Proceedings, Am. See. Testing Mats., Vol. 56. ^° Senior Project Engineer, Lessels and As-
p. 1034 (1956). sociates, Inc., Boston, Mass.
P A N E L DISCUSSION ON L O W - C Y C L E FATIGUE 119
men as testing progressed. This area possible to observe them through the
started a t one location and moved to a number cycles to failure since they are
new location as the number of cycles inside the furnace. However, they are of
increased, probably due to some strain relatively uniform appearance with re-
hardening of the part that was subject to spect to the strain. The failure is pri-
plastic straining. This result may also be marily an intergranular failure, and it is
experienced in the work presented in Mr. not exactly the same as you obtain under
Kennedy's paper; I wonder if it is not static creep conditions. Under fatigue or
possible that in a complex stress state low-cycle fatigue conditions it is often
you would experience some strain hard- observed that the specimen has only one
ening in one direction and even transfer crack which has gone through the wall.
the failure from one stress direction to Under creep conditions the specimen has
another. In such highly strained states, many intergranular cracks all over the
the strains in the two directions can be specimen; in this respect, it is possible to
drastically different. I n cylinders under get nucleation of the first crack which
pressure, for example, where you may then begins propagating throughout the
get 5 per cent plastic strain in a matter specimens before the others can begin to
of tens of milliseconds, you m a y also
nucleate. With respect to strain harden-
get as much as a 50:1 difference in strain
ing under complex stresses, a number of
in the two directions. We have accom-
experiments have been performed. An
pHshed such cycling under shock load-
example is tension testing at low tem-
ing, and high differences in the strain in
peratures under complex stresses in which
the two directions have been the result.
the shear stress - shear strain relation-
MR. KENNEDY.—Are you talking ships are relatively uniform except in
about the effect of stress state on work those cases where anisotropic effects
hardening? exist. This implies that the strain harden-
M R . BRODRICK.—I first remarked that ing under complex stress states would be
the area of maximum strains can ap- essentially the same, regardless of the
parently migrate, at least in a straight stress state. I t can always be expected
cyhndrical specimen subject to axial that one particular unit or small section
loading. Therefore, I wonder if there in a specimen will yield just on a statis-
could be, perhaps, a change in direction tical basis, and in turn, it will probably
of the worst condition under biaxial strain harden. If it doesn't strain harden,
loading. or if there is an absence of strain hard-
MR. KENNEDY.—Are you talking ening, localized yielding results in neck-
about high-velocity loadings? ing. This is observed to a large extent
M R . BRODRICK.^—The first set I men- in materials which have been strain hard-
tioned was cycled at low speeds with ened or cold worked considerably. When
fully reversed strain. In the second case, these materials are tension tested, their
the only occurrence has been in high strain hardening ability is reduced to such
velocity in one-shot loading of high an extent that the first yielding pro-
velocity. I was wondering if there would
duces necking and failure. This is a
be some interaction at lower speeds under
large reduction of area rather than a
a biaxial stress type of test. These re-
brittle failure. You may be seeing a
marks are based upon room-temperature
condition of essentially localized strain-
testing.
ing, localized yielding on a statistical
M R . K E N N E D Y . — F o r the particular
basis, and then strain hardening to the
specimens in this report, it was not
120 SYMPOSIUM ON FATIGUE OF AIKCRAFT STRUCTURES
point where yielding begins in some I have thought about this condition, but
other area. I have not come to any conclusions on
MR. DOLAN.—Would Mr. Kennedy how to handle principal axes that are
clarify conclusion 4 of his paper? This rotating. This is essentially what you
conclusion states that results of the two have—the shear stress from the torsion
test series demonstrate the independence builds up, and the principal axes are
of damage accumulation on each princi- rotating. Again, this is a high-tempera-
pal stress axis. This needs seme clarifica- ture phenomenon and this holds true
tion because we have been running only when Eq 5 in my paper, for creep
experiments at Illinois on specimens of rupture, is appHcable.
tubing roughly 5 in. in diameter by 6 ft MR. DOLAN.—The tests we have been
long suspended at the top and loaded at making are at room temperature on
the bottom by a fixed dead weight copper and aluminum tubing, and the
which incurs a nominal axial stress be- large amounts of plastic strain are rather
low the elastic strength of the tubing. surprising. Similar results were obtained
The tubing is then subjected to alter- by Coffin in some tests which used a
nating torsion (plastically twisted) under cylindrical specimen with a small torque
either controlled strain cycles or con- applied to the specimen which was then
trolled load cycles. Surprisingly enough, plastically cycled in the axial direction.
the axial load with low nominal stresses In this case, his principal stress directions
is sufiicient to extend the tube by very were again at odd angles, producing large
large amounts; extensions can be meas- twists for small steady stress.
ured with a yardstick. Unlocking the MR. KENNEDY.—CoflSn's test condi-
plastic deformation by the torsional tions were very poorly defined. I do not
action causes large extensional strains believe he thoroughly analyzed this
for very low tensile stresses. In this case, particular test. The shear stresses or the
the addition of effects from an axial effective stress on these particular oc-
load independently of those from the casions can be quite large; if they are
torsional load is not a legitimate summa- compared in this respect, surprising
tion; in fact, the deformations are so amounts of strains are found.
large that there is no simple plasticity MR. MoRDFiN.—Some years ago Voor-
theory that can explain them. This is a hees, at the University of Michigan,
case in which the principal stress axes are handled a problem in which the principal
at 45 deg to each other, rather than at stress axes were changing with time due
90 deg to each other as I understand they to deformation of the specimen; how-
were in Mr. Kennedy's tests. Is the ever, this was more of a creep test than a
statement that these are independent in fatigue test. He took an average initial
the accumulation on each principal stress octahedral shear stress and obtained
axis meant to imply some special cases, good agreement with test results for a
or is this type of phenomenon explainable notched specimen. This was tried at the
on this basis? National Bureau of Standards with
I have in mind the case in which the simple riveted joints, and, using this
axial load is a constant, steady load. The idea of an average stress, we obtained
torsional load builds up from zero to results that were satisfactory for en-
maximum and has been completely re- gineering purposes.
versed, on a sinusoidal loading basis. MR. C. P . BAUM.2'—A question Mr.
MR. KENNEDY.—Do not the principal 2' Aerospace Engineer, Bureau of Weapons,
stress axes vary with that stress cycle? U. S. Navy Department, Washington, D. C.
PANEL DISCUSSION ON LOW-CYCLE FATIGUE 121
Grover asked earlier deserves a little ticularly in the case of very low-cycle
more discussion in view of the emphasis fatigue. Work on the factor of cumulative
on fatigue in the Department of Defense. plastic strain has been going on at the
The Air Force has put a lot of its money University of Illinois under Mr. Dolan.
and effort into this subject. Since 1959, Also, Mr. Bell reported some data on
fatigue work has been accelerated ap- this subject. The second factor involved
preciably, and answers to Mr. Grover's is that the modified shear strain is
question wovild be helpful in giving some probably higher at the surface of a
guidance; this group could provide such uniaxially loaded specimen. The third
guidance in this particular case. Per- factor is the introduction of residual
haps Mr. Grover may wish to restate his stresses, and the fourth one is the initia-
question. tion and propagation of cracks. Finally,
MR. GROVER.—What is being done the fifth effect is the critical static crack
with regard to crack propagation versus length, or the unstable crack length.
crack initiation; and of what use is this This last factor is associated with the
work from the point of view of practical reduction of the tensile strength. The
application? I shall propose a couple of final fracture may differ depending upon
thoughts I pretend to believe; however, which mechanism overrides the others.
I do not want to say outright that I In the case of very low cycles, the yield
believe them. In the intermediate range, point at the surface, I believe, is some-
at least, there seems sometimes to be what less than in the interior. After a
much more reproducibility in the rate of certain number of cycles, compressive
crack propagation than in the number stresses develop at the surface of a plain
of cycles required to initiate the crack. specimen, and tensile stresses in the
For this purpose, let me define a crack interior. Then under certain conditions
as something 0.001 in. long. If you the cracks first form internally because
measure the growth from 0.001 in. to 0.1 of this high internal tensile stress. When
in. you may get, for a given loading these cracks reach a critical length, an
condition, quite a lot of reproducibility "instantaneous" propagation occurs, the
from one specimen to another in the specimen fails in the center by a cohesive
rate of growth. However, if you measure failure, and the typical shear lip failure
the number of cycles at which you first is exhibited at the edges of the specimen.
see this small crack, you will get a lot At a somewhat higher number of cycles,
of scatter. Does this mean that our we do not get enough plastic strain
statistical scatter is wholly explained by accumulation at the surface to cause an
the very early stages of cracking? Does internal rupture. As a result, you have
it mean that we must apply statistics to the so-called local sore spots which
crack initiation and can forget about initiate fatigue cracks, and there are a
them for crack growth? Are there any multiplicity of these cracks on the surface.
comments from others along any such In this case I think the fatigue cracks
line as this? join together to cause the final failure.
MR. F . B . STULEN.22^1 think there
Thus, there are two mechanisms of
are a number of mechanisms that are failure, an internal failure that Professor
competing in low-cycle fatigue, and I Dolan and Mr. Hardrath were talking
should like to list several of the factors about, and a failure that starts from a
that may enter into this problem, par- multiplicity of surface cracks that Mr.
Schijve discussed, and which occurs
^* Chief Scientist, Curtiss Division, Curtiss- probably at a somewhat higher number of
Wright Corp., Caldwell, N. J.
122 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
cycles. Thus, the reduction of area de- crack propagation versus crack initiation,
pends on the type of failure. In the case one immediately gets into a complex
of a notch, Mr. Smith brought some of field because of difficulties in understand-
these points out. A unidirectional plastic ing scale effects and actual terminology.
strain near and at the edge of the hole is If we regard cracks as something visible
developed, and this plastic strain results to the naked eye, we are talking about
in a mean compressive stress which one set of phenomena. If we look on the
inhibits crack initiation propagation. If microscopic scale within crystals, we are
this occurs for R = —1, for example, talking about a different observational
there is a bias strain, that is, you might level and the things we see are different.
say there is more yielding in the tensile If we go a step further and look at the
strain resulting in a mean compressive atomic arrangement within the crystals,
stress. If this condition continues long again we are looking at another type of
enough it may explain the anomaly that phenomenon; thus it is hard to say what
came up earlier where the Kf factor, a crack is and when it does initiate.
when plotted against .V, actually seemed The present discussion represents a
to decrease a bit. This may be a result mixture of phenomena also, in that we
of the accumulation of a mean compres- are looking at fracture resulting from
sive stress at the edges of the discon- very few cycles of loading. These are in
tinuity. To introduce an externally ap- the realm of sudden brittle fractures or
plied mean compressive stress in the sudden ductile fractures—an area that
specimen would result in an accumula- has been under constant study by the
tion of plastic strain in such a way as to A S T M Committee on Fracture Testing
give rise to a residual tensile stress at the of High-Strength Steels. This committee
notch, so that there is actually a reduc- is looking at the characteristics of crack
tion in the fatigue strength compared propagation in slow or rapid fashion
with what one would expect for the using built-in simulated cracks and
smooth (unnotched) condition. I do not notches, and it is finding the field ex-
know whether this clarifies the subject tremely complex with just a single static
but I think there may be several con- application of loading, or with two or
cepts that may explain the various three static applications of loading. This
phenomena that we have seen. is not a cycle-dependent phenomenon as
MR. SMITH.—In my paper, short-cycle much as it is involved with the material
fatigue data are used to predict the life ductility characteristics, the build-up of
of full-scale structures for any desired shear lips on the running crack, and the
life or any load history. In so doing, it toughness characteristics of the material
introduces something that is quite in resisting the propagation of a flaw once
foreign to most of our thinking. I should formed. Again, let me emphasize that the
like to compliment Mr. Bell on his fracture is essentially that for a single
presentation of data in this manner. application of gradually increasing load.
This is the very type of data that I re- When we develop low-cycle fatigue, we
quire to make the system work. Actu- are introducing or adding to this the
ally, it uses plastic flow data of the type progressive type of cumulative damage
shown by Mr. Bell and makes use of the which is cycle dependent. The phenom-
fact that when a specimen failed at a ena observed are different at each of the
certain number of cycles, we know observational levels, and we are mixing
definitely that yielding had taken place. in low-cycle fatigue a combination of
MR. DOLAN.—When one discusses both the static fracture and the pro-
PANEL DISCUSSION ON LOW-CYCLE FATIGUE 123
gressive cracking. The sudden fracture is some movies showing the propagation of
evidence of the flaw size reaching a fatigue cracks. I do not know whether
critical level and causing the toughness these were high-cycle or low-cycle tests
characteristics of the material to be over- but the movies show this jumping phe-
come. To reach that flaw size in a low- nomenon. If the surface is observed,
cycle fatigue test, we are also generating there is no crack in a given area and then
cycle-dependent flaws which may be suddenly a crack in an adjacent section
quite random in character and rather jumps over a rather extensive area. This
complex. movie is available for those who want to
If a specimen is repeatedly cycled borrow and study it. More details can
with fairly uniform stress over a large be obtained from Mr. William Hols-
zone, the result is not one fatigue crack; houser of the Civil Aeronautics Board.
at the various levels of observation there MR. W . L . HOLSHOUSER.-'—The movie
are hundreds of fatigue cracks rather dealt with high-cycle fatigue. It was
randomly distributed. The rate of growth made in connection with the study of the
of each flaw on a microscale is very origin of fatigue cracks around a small
erratic; it does not grow with uniform hole. The movie shows that slip lines
modes of growth. Even if you start a appear in certain areas around the hole
single crack in a sheet by putting a sharp and that the cracks originate in the slip
notch and loading it cyclically, careful lines and progress in a rather discon-
observations indicate it does not grow tinuous fashion.
uniformly but by erratic jumps. One MR. SMITH.—Mr. Dolan made a
usually puts a smooth curve through statement that the final failure is not
these points. However, it is quite fre- necessarily at the crack that came into
quently observed that the curve is not view first. This is especially true in
of a uniform growth character per cycle. large-scale structures. With increasing
Engineers like to look at models of numbers of cycles the first crack just
things in an idealized fashion and think stops growing and another crack appears
in terms of linear relationships. Here, elsewhere; maybe even a third crack will
however, we are dealing with phenomena culminate in the final failure. This might
that are very complex and nonlinear; if have been visualized as being typical of
we talk about propagation of a crack we smooth specimen data, but it is also true
are looking at a zone around the tip of in full-scale structures.
the crack and in advance of the crack MR. STULEN.—About 1957 we ran
that has suffered some damage from some fatigue tests on smooth rotating-
plastic deformation. This plastic zone in beam specimens and carefully photo-
the advance of the tip of the crack, de- graphed the whole surface of each speci-
pending upon the rate of growth, may men at frequent intervals. At a given
actually contain microscopic flaws or stress level, if the crack length was
voids caused by the progressive damage plotted against the number of cycles, we
under the previous cycling. Therefore, found exactly what Mr. Dolan described,
the whole field is extremely complex, that is, the crack was initiated at an
particularly in the low-cycle range in inclusion. At the beginning of the cychng,
which the cycle-dependent flaw propaga- the crack would hesitate for a long period
gation mixes with the sudden running and then it would appear to jump. It
fracture under a static load. seemed to grow rather rapidly for a few
M R . GOHN.—Mr. J. A. Bennett of the ^' Metallurgist, Civil Aeronautics Board,
National Bureau of Standards has made Washington, D. C.
124 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
ten thousandths of an inch and then there is probably some other value for
hesitate for a long time. The hesitation the exponent.
periods would continue during growth, The other phenomenon was that the
but they would become shorter as the high variability in fatigue was caused by
number of cycles increased. Because the initiation period. Practically 70 or
measurements were taken every 5000 80 per cent of the variance is caused by
cycles, the hesitation periods later in the the initiation period which, in turn,
process would not be apparent and the depends upon the variation in the sizes
growth appeared continuous. If we took of the flaws. If you were to drill a small
another specimen that happened to have hole in a smooth specimen, say of the
a somewhat larger inclusion, of the order order of 0.005 in. (larger than an inclu-
of 0.002 in. compared with the previous sion), no doubt you would get the same
inclusion of 0.001 in., the initial crack type of crack propagation relationship
growth would be more rapid, and the that you have in the smooth specimens,
curve of growth versus cycles would be except the crack probably starts earlier.
above and to the left of the previous You can rationalize the old French dam-
curve. The crack growth on a number of age line on this basis and show that there
specimens tested at the same cyclic is a specific crack length that can exist
stress was measured; it was found that in a material at a given cyclic stress
the crack curves above about 0.003 in. level without propagation. If a step test
could be shifted laterally so that they is run on a specimen with a crack in it,
were more or less coincident. If the it will have very large life up to a certain
starting point of the propagation curve critical alternating stress value, depend-
is taken at No cycles and the correspond- ing on the crack length. If this threshold
ing crack length k , the relation obtained value or fatigue limit of the crack length
is: is plotted against the stress level on log-
log paper, a linear relationship is ob-
log
(0= ks'iN - iV„) tained. The fatigue limit of a crack in a
given material is inversely proportional
to its length raised to a given power.
Frost and Dugdale in England pub-
lished this same formula and experimen- Therefore a certain maximum length of
tally determined the exponent, a, to be crack or defect in a given material will
3. By using some plastic calculations at not affect the fatigue hmit of the ma-
the tip of the crack and making some terial. This observation indicates that
rather simplifying assumptions, the ex- the French damage line is a convenient
ponent a should either be equal to 3 method of dividing the crack initiation
based on the concept of depletion of stage and crack propagation stage by
ductility at failure or should be equal to the specific crack length that does not
3 plus the strain-hardening index, as- affect the fatigue strength. To introduce
suming the total specific damping is small holes artificially is to start out
constant at failure. We do not know with a larger critical dynamic crack,
quite what the correct value for this thus lowering the original fatigue limit.
exponent is because the strain-hardening Very clean steel raises the threshold
index is usually about 0.1 and it is value and the fatigue hmit. Steel with
difficult to distinguish between the small notches that are larger than the
theories. However, this relationship was inherent defects produce a lower fatigue
developed for the case of zero stress limit. These threshold values may not be
gradient. When there is a stress gradient, horizontal; they may actually dip a bit.
P A N E L DISCUSSION ON LOW-CYCLE FATIGUE 125
depending upon whether it is ferrous or claimed the fatigue crack growth period
nonferrous and depending upon whether was about 2 to 5 per cent of hfe, while
there is strain aging or not. a discussion claimed that the cracking
MR. GROVER.—How far up in high phase represented 90 to 95 per cent of
strain and how low down in number of the hfe. Actually both were correct in
cycles do you think this kind of model their statements; what was not made
usefully apphes? clear at the time was that one was re-
M R . STULEN.—This model gets con- ferring to the behavior on plain (high
fused when you go up to a higher stress apphed stresses) specimens while the
level because there is a multiplicity of other was observing crack formation in
cracks forming in a smooth specimen as notched (low apphed stresses) specimens.
mentioned by previous speakers. A In defense of Mr. Dolan's comments
large crack can start from a large defect concerning erratic and spurting growth
first, b u t if the many cracks starting of cracks, I will agree because I have
from the small defects happen to join observed this type of behavior in many
together quicker than the initially formed materials. I also believe that scatter in
crack, they will cause the failure. Single- the crack growth phase is low, so here
nucleus failures will occur in most prac- again we may be intermixing our dis-
tical devices because holes, notches, or cussions about two separate things. T h e
stress raisers are present in everything spurting and erratic growth in the micro-
that is built, and the crack will start as a phase is continuing in an orderly fashion,
single-nucleus crack. Even in plain speci- if it is possible to describe it in this way,
mens, as Mr. Dolan has pointed out, at and these curves can be translated one
the low values of stress, a very large upon the other as Mr. Grover has
number of cracks will be found all over mentioned.
the specimen, b u t the one that gets Did I get the correct impression from
started first is usually the one that Mr. Kennedy's paper that the compres-
dictates the failure. However, this is not sion half of the cycle in the combined
true at the high stress levels, in which creep and fatigue program at low-cycles
many cracks start; it is not necessarily had no effect in damaging the material?
the first one that starts, but rather those MR. KENNEDY.—That is what was
that happen to line up and join that said, b u t it is not certain about the low-
dictate the failure. stress tests that were performed. The
CHAIRMAN RATTINGER.—Would Mr. high-stress tests and the torsional tests
Christensen, who is actively engaged in did not indicate this.
similar work, like to comment on this MR. CHRISTENSEN.—Well, I would
topic? have challenged the statement for long-
M R . R . H . CHRISTENSEN.2^—The dis- time tests because data exist showing
cussion on the sporadic nature of crack that it does. If fatigue and creep are
growth versus a continuous behavior conjointly going on, the compression
reminds me of a similar discussion when has a large effect within the fatigue
Mr. Wilkov presented his paper on component when alternating from ten-
Mechanism of Fatigue Failure^* at the sion to compression, whether the load
Fatigue Session in Atlantic City during is a sine wave or a square wave. If the
the 1961 Annual Meeting. Mr. Wilkov
^*M. A. Wilkov, "New Observations Relat-
^* Staff Assistant, Strength Engineering De- ing to the Mechanism of Fatigue Failure,"
partment, Space System Division, Douglas Proceedings. Am. Soc. Testing Mats., Vol. 62,
Aircraft Co., Santa Monica, Calif. p. 540 (1962).
126 SYMPOSIUM ON FATIGXIE OF AIRCRAFT STRUCTURES
effects of material degradation and the plex joint. For the type of test conducted,
mechanism of creep cracking are occur- the notched {Kt = 6.3) data gave the
ring simultaneously with fatigue, then I best correlation. For the very shallow
think these processes going on simul- notch {Kt = 3.0), the stress concentra-
taneously and acting one upon the other tion was not sufficiently sharp to give
certainly would be affected somewhat by any indication of embrittlement for any
the compression cycle. Was your wave a of the materials tested.
square wave? MR. CHRISTENSEN.—While this has
MR. KENNEDY.—All of the experi- appeared to be true for some materials,
ments used square waves, and, in this the only caution I should like to make is
respect, it simplified analysis consider- that there are materials for which the
ably. fatigue crack is the much severer notch,
MR. C H R I S T E N S E N . — I was going to and at liquid nitrogen or hydrogen tem-
challenge your taking half of the time as peratures the curve actually flips over
damaging time on this; b u t I was think- and decreases in strength while the
ing of a sine wave load shape. strength may increase with notches of
I have a question for Mr. Christian. Kt of 18.
He tested a considerable number of M R . CHRISTIAN.—In essence, then,
notched specimens with Kt of 3, 19, for a sharper (fatigue) notch, the notched
and so on, at different testing tempera- tensile value may decrease rather than
tures. Then he ran many fatigue tests. increase with reduction in testing tem-
Does he believe that if he had made a perature. This seemed to be indicated
tension test on the fatigue-cracked panels from even the three notch configurations
he would have obtained strength reduc- which were tested. There was a decrease
tions similar to the sharp-notch data? in the notched tensile strength with
MR. J. L. CHRISTIAN^^ {panelist).— reduction in temperature in the follow-
The use of a static test for correlating ing cases:
with or predicting a fatigue failure has
been discussed. In the very high-stress Ki = 3.2 for type 301 stainless steel (heat
low-cycle fatigue of a complex joint, such 49061), transverse direction, at
as was tested in the subject investiga- - 423F,
tion, it appears that a static notched K, = 6.3 for AM-355, longitudinal and
transverse directions, at — 320 and
test is capable of indicating whether the
- 423F, and
material is capable of withstanding a K, = 19 for type 304 stainless steel, long-
large number of cycles in axial fatigue. itudinal direction, at — 423F, and
The data indicated that the Kt of 6.3 types 301 and 310 stainless steel,
more closely correlated or more closely transverse direction, at — 423F.
indicated fatigue life than did the Kt of
19. I believe the reason for this was due
to the presence of resistance spot welds CHAIRMAN R A T T I N G E R . — I should like
in the complex joints in the fatigue to thank the panel for participating; I
specimens. H a d there been a fatigue think that these thanks are also shared
crack present rather than the notch due on the part of the audience. Finally, Mr.
to the resistance spot weld, it is possible Grover has a few concluding remarks.
that the sharper notch data would better M R . GROVER.—I should hke to express
predict the fatigue behavior of the com- our appreciation to those who arranged
this symposium, to Mr. Hardrath who
^^ General Dynamics/Astronautics, San Di-
ego, Calif. has done a lot of work for it, and to
PANEL DISCUSSION ON LOW-CYCLE FATIGUE 127
our chairman this morning, Mr. Rattin- 13 of Mr. Whiteson's paper show two
ger. different forms of fracture that I have
[CHAIRMAN'S NOTE.—The following also found.2' It is suggested that the
additional remarks were supplied by Mr. term fatigue should be reserved for the
Benham, who was unable to attend the classical type of cyclical failure, the
meeting. They are included for a num- crack area which shows no gross plastic
ber of reasons, not the least of which is deformation. This behavior is obtained
their thought-provoking nature.] at low endurances particularly under
MR. P . P . BENHAM^' (by letter).—Mr. restricted strain cycling conditions.
Bell has dealt quite fully with points Alternatively, under load cycling condi-
raised on our paper; however, a few tions, because strain is "free," an incre-
further remarks might be made regarding mental, unidirectional strain can occur
the apparent drop in strength reduction that can overrule the fatigue slip and
factor, Kf, in the high-cycle range. cracking mechanism, and eventually
Based on theoretical and static experi- lead to fully ductile failure indistinguish-
mental elastic and plastic stress concen- able micrographically from static mono-
tration factors, one would expect fatigue tonic fracture (cup and cone in tension).
strength reduction to be greatest at high I have also found instances of small
cycles and least at low cycles. However, surface-fatigue-type cracks occurring in
in the particular multiple notched speci- the neck of a ductile failure, particularly
mens tested here, there are three fairly in the range of endurance in which there
distinct stages of deformation across the is the change from the ductile to the
notched section: "elastic," elastoplastic fatigue mechanism.
(local yielding), and fully plastic, related Mr. Munse's tests are quite fascinat-
to high, medium, and low endurances, ing in conception, and they provide some
respectively. In the second of the three
valuable information on residual duc-
phases there is some interaction, redistri-
tility. However, with the exception of the
bution, and residual stress between the
R = —1 series which may be regarded
notches. Since there is no simple relation-
as reversed strain cycling after com-
ship established between Kt and Kf ,
pressive prestrain, the other R ratios
there is no conclusive evidence why there
should not be a peak in some cases in are actually biased cyclical tension tests
Kf-N curves. It must be admitted, resulting in necking and ductile cup-and-
however, that the shape of those curves cone fractures. The R = —I tests can
can be effected by the position of the be regarded as true low-cycle fatigue
S-N curves as drawn, as was pointed out failures. It is interesting to note that the
by Mr. Hardrath. relationships derived for predicting
strain and endurance agree well with
The remaining comments are mainly Messrs. Pian and D'Amato's results,
concerned with the nature of fracture in which were essentially prestrain plus
the low-endurance range, a topic dealt reversed strain cycling, and also Mr.
with by a number of the speakers. A
Munse's R = —1 results, but do not
number of people would define fatigue
represent the results very well (see Figs.
as a failure caused by any form of cyclical
application of stress and strain. Yet II to 13) for the other R ratios which
Fig. 14 of Mr. Munse's paper and Fig. lead to tensile type failure.
Figures 14 to 16 in Mr. Whiteson's
^' Lecturer in Applied Mechanics, Depart- paper parallel the behavior found in
ment of Mechanical Engineering, Imperial Col-
lege, London, England. copper and mild steel at room tempera-
128 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
^yj.g28,29 Figure 14 again emphasizes fractures would have been true fatigue
that under mean load cycling at low type. It may be that some materials just
endurances two different mechanisms do not exhibit the ratcheting-ductile
can operate in some materials. In the failure mechanism; in fact under various
present tests, fatigue failures were not stress conditions there is a genuine
obtained earlier than about 1000 cycles. minimum number of cycles required to
One aspect of Mr. Whiteson's paper, initiate and propagate a true fatigue
not fully understood by me, concerns fracture.
the so-called strengthening effect. If Finally, in relation to Mr. Kennedy's
these were constant-load tension cycling paper, I have also found good correlation
tests, why did the specimen not break between low-endurance reversed torsion
on the first half cycle? In other words, and reversed axial plastic-strain cycling,
how can one achieve a fatigue strength on pure copper, on the basis of the Von
greater than the static strength other Mises equivalent stress - equivalent strain
than by coaxing, which is not constant criterion.
load cycling, over a number of cycles? MR. WHITESON.—Mr. Benham is cor-
The point at issue is whether the tests rect in assuming that tensile failure
referred to by Hardrath et al., Illg, and would have taken place if we had reached
also the writer^'' ^° on aluminum alloys a stress level corresponding to ultimate
could have been made to exhibit the tensile strength of the material during the
ductile type of failure, by coaxing, at first half cycle of loading. However, we
shorter lives than the minimum fatigue
pointed out in the paper that we did not
life quoted, or alternatively whether the
achieve a constant load cycling type of
2' P. P. Benham, Journal, Inst. Metals, Vol. test, in that the first few cycles of load
89, p. 328 (1961). application were slightly below the
^' P. P. Benham and H. Ford, Journal, Inst.
Mechanical Engineering Science, Vol. 3, No. 2, designated stress level. We believe that
p. 119 (1961). these first few load cycles were responsi-
™ C. T. Mackenzie and P. P. Benham, Jour- ble for the strengthening effect, due to
nal, Royal Aeronautical Soc, Vol. 66, No. 614,
p. 128 (1962). some coaxing phenomenon.
Helicopter Fatigue
STP338-EB/Sep. 1963
E M P I R I C A L ANALYSIS OF F A T I G U E S T R E N G T H OF P I N - L O A D E D
LUG JOINTS
B Y A. A. MITTENBERGS 1
SYNOPSIS
d-pin installed
with 0.0005-in.
per in. interference
pin diameter,
lug and lever width,
clevis lug thickness,
lever thickness,
throat depth,
fillet radius,
locations of failure origins (Oc , in clevis lugs; OL , in lever),
steady tension load,
alternating bending force, and
distance from centerline of the pin to the point of application of force B.
FIG. 1.—Specimen Configuration.
A rational engineering-type approach The clevises and levers were made from
to the problem, therefore, was taken to aircraft-quality forged SAE 4340 steel
establish a correlation between the bars and from extruded 2024-T4 alumi-
nominal stresses at failure and the test num alloy bars. The steel parts were
results observed in order to enable pre- quenched and tempered to an ultimate
diction of alternating stresses at failure. tensile strength of 145,000 psi. The ulti-
This paper describes the approach and mate tensile strength of aluminum alloy
presents the results of an empirical was 82,000 psi. All pins for both steel
analysis of fatigue strength of pin- and aluminum specimens were made from
loaded lug joints investigated in the aircraft-quahty SAE 8620 steel. They
experimental program (I). were carburized, heat treated to a case
M i T T E N B E R G S ON F A T I G U E STRENGTH OF P i N - L o A D E D L u G J o i N T S 133
STEEI. SPECIMENS
.4-1.... 0.625 2.00 0 . 3 7 5 1.50 1.25 0.19 0.250 0.313 0.188 0.750
A-2.... 0.625 2.00 0 . 5 0 0 1.50 1.25 0.19 0.333 0.313 0.250 0.750
A-3.. .. 0.625 2.00 0 . 7 5 0 1.50 1.25 0.19 0.500 0.313 0.375 0.750
B-0.... 1.00 2.00 0 . 2 5 0 1.50 1.25 0.19 0.167 0.500 0.125 0.750
B-l. . . 1.00 2.00 0 . 3 7 5 1.50 1.25 0.19 0.250 0.500 0.188 0.750
B-2.. .. 1.00 2.00 0.500 1.50 1.25 0.19 0.333 0.500 0.250 0.750
B-3.... 1.00 2.00 0.750 1.50 1.25 0.19 0.500 0.500 0.375 0.750
B-4.... 1.00 2.00 1.000 1.50 1.25 0.19 0.667 0.500 0.500 0.750
C-1.... 1.3125 2.00 0.375 1.50 1.25 0.19 0.250 0.657 0.188 0.750
C-2.. . . 1.3125 2.00 0.500 1.50 1.25 0.19 0.333 0.657 0.250 0.750
C-2.5. . 1.3125 2.00 0 . 6 2 5 1.50 1.25 0.19 0.417 0.657 0.313 0.750
C-3.. .. 1.3125 2.00 0 . 7 5 0 1.50 1.25 0.19 0.500 0.657 0.375 0.750
D-2.... 0.625 1.25 0 . 5 0 0 1.50 0.781 0.12 0.333 0.500 0 . 4 0 0 1.200
E-2.. . . 1.3125 2.625 0 . 5 0 0 1.50 1.641 0 . 2 5 0.333 0.500 0.190 0.572
F-1. . . . 1.00 2.00 0 . 3 7 5 1.125 1.25 0.19 0.333 0.500 0.188 0.564
G-3... . 1.00 2.00 0 . 6 2 5 1.875 1.25 0.19 0.333 0.500 0.313 0.937
L-2. . . . 0.75 1.50 0 . 5 0 0 1.50 0.940 0.14 0.333 0.500 0 . 3 3 3 1.000
M-2. . . 0.50 1.00 0 . 5 0 0 1.50 0.625 0.094 0.333 0.500 0 . 5 0 0 1.500
N-2.... 0.8125 2.00 0 . 5 0 0 1.50 1.25 0.19 0.333 0.406 0.250 0.750
R-1... . 0.75 1.50 0.281 0.844 0.940 0.14 0.333 0.500 0.188 0.564
ALUMINUM SPECIMEN.?'
AL-A-2. 0.625 2.00 0.500 1.50 1.25 0.19 0.333 0.313 0.250 0.750
ALB 0 1.00 2.00 0.250 1.50 1.25 0.19 0.167 0.500 0.125 0.750
AL-B-1 1.00 2.00 0.375 1.50 1.25 0.19 0.250 0.500 0.188 0.750
AL-B-2 1.00 2.00 0.500 1.50 1.25 0.19 0.333 0.500 0.250 0.750
AL-B-3 1.00 2.00 0.750 1.50 1.25 0.19 0.500 0.500 0.375 0.750
AL-C-2 1.3125 2.00 0.500 1.50 1.25 0.19 0.333 0.657 0.250 0.750
AL-D-2 0.625 1.25 0.500 1.50 0.781 0.12 0.333 0.500 0.400 1.200
AL-L-2 0.75 1.50 0.500 1.50 0.940 0.14 0.333 0.500 0.333 1.000
Applied bending moment versus life- points are assumed to be the critical
time (Af-.V) curves were developed for locations.
each specimen series up to about The nominal stresses at the critical
15 X 10^ cycles, using three to eleven locations, Oc and OL , are defined as
specimens per test condition. For each follows:
M-N curve the steady-tension load was
kept constant. The alternating bending Sc = ~ .(1)
moments applied at the centerline of the
pins were varied from specimen to
specimen by adjusting the alternating SL = .(2)
bending force (-B, Fig. 1) applied at the
distance / from the pin axis. This
Sbc (3)
distance was 16 in. for the steel and 21 c
11.125 in. for the aluminum specimens.
Mh
(4)
NOMINAL STRESSES AT
CRITICAL LOCATIONS
S,c = (5)
The fatigue failures occurred either
in the lever or in the clevis lugs, de-
pending primarily upon the relative SsL = .(6)
thicknesses of the clevis lugs, h , and
the lever, t^. Thus, the ratio h/li is the where
principal criterion which determines the Sc = Nominal steady tensile stress in
failing component. Generally, at ti/l2 < the net cross-section of the clevis
0.333 clevis failures were predominant; lugs through the centerline of the
at h/k > 0.333 all failures occurred in pin hole, psi,
the levers; and at a ratio of h/h = 0.333, SL = Nominal steady tensile stress in
failures occurred in both components. the net cross-section of the lever
The failures, originating on the inside through the centerline of the pin
of the pin hole for both levers and clevis hole, psi,
lugs (0, Fig. 1), were located roughly in Sbc = Nominal alternating bending
a plane through the pin axis perpen- (tensile and compressive) stress
dicular to the longitudinal direction of at the inside top and inside
the joint with the origins located near bottom surfaces of the clevis lugs
the top or bottom surface of the lever, in the net cross-section through
OL , or near the inner top or inner the centerline of the pin hole, psi,
bottom surfaces of the clevis lugs, Oc • SbL = Nominal alternating bending
Failures at the latter locations were stress (tensile and compressive)
caused by the effects of pin bending and at the top and bottom surfaces of
lug deformations. According to an ele- the lever in the net cross-section
mentary analysis that neglects these through the centerline of the pin
effects, the failures should occur at the hole, psi,
outer fibers of the clevis lugs, since the Ssc = Nominal alternating shear stress
bending stresses are nominally the in the net cross-section of the
highest at the outside surfaces. Although clevis lugs through the centerline
the exact locations of the failure origins of the pin hole, psi,
varied somewhat from the points desig- SsL = Nominal alternating shear stress
nated by 0, OL , and Oc in Fig. 1, these in the net cross-section of the
MiTTENBERGS ON F A T I G U E STRENGTH OF PiN-LOADED L u G JOINTS 135
STEEL SPECIMENS
ALUMINUM SPECIMENS
In the calculations for Sbc (Eq 3) and the stresses from pin-interference fits.
for / e , the clearances between the lever Although the tangential stresses at the
and the clevis lugs are neglected since pin-hole surfaces from pin interference
they are small in comparison with ti and of 0.0005 in. per in. of pin diameter are
since the error is insignificant in view of significant (on the order of 10,000 psi for
the normal scatter in fatigue data. steel and 4000 psi for aluminum joints
The fatigue data obtained in the ex- used in the experimental program), their
perimental program for the various test- effects on the fatigue strength of these
specimen series at 15 X 10^ cycles are joints are negligible as compared with
summarized in Table II. The lifetime of joints having closely (metal-to-metal)
15 X 10^ cycles was selected arbitrarily fitted pins. It can be shown (2,3) that,
to provide a common basis for comparing under conditions used in the experi-
80,000
(b) Aluminum (a) Steel
s SA S SA
10,000 18,000 7500 64,200
60,000
15,000 17,500 10,000 64,000
<
</5 20,000 17,000 11,250 63,800
30,000 15.000 15,000 63,000
,-t 40,000 — 18,000 62,600
to 62,000
22,500
30,000 60,000
45,000 55,000
fc 20,000
and analyzing the results. The alter- mental program, these interference fits
nating bending moments, M, were taken decrease slightly the peak steady (mean)
from the developed M-N curves. The stresses, but they have no effect on the
applied nominal tensile stresses and the amplitudes of peak alternating stresses.
nominal alternating bending stresses at Therefore, the analysis described in this
failure were computed from Eqs 1 to 4. paper is applicable to both pin-loaded
The nominal alternating shear stresses joints with slight pin-interference fits
are small (generally less than 10 per cent and those having a close fit between the
for the failing member) in comparison pin and pin hole.
with the alternating bending stresses;
therefore, they are not included in Table BASIC CONSIDERATIONS
II, and they are excluded from further Generally, the fatigue strength of pin-
consideration. loaded lug joints may be influenced by a
Excluded also from this analysis are large number of factors, many of which
MiTTENBERGS ON FATIGXIE STRENGTH OF PiN-LOADED LUG JOINTS 137
tions and all other geometric effects localized yielding suspected in steel
influencing the fatigue strength of the specimens with thin pins. Therefore, a
joints. plasticity factor of Pc = 1.0 (no plastic
In the experimental program (1), the deformation) is assumed for all clevises.
joints were evaluated as a unit. Here, The selection of the plasticity factor,
the clevis and lever must be considered PL , for the levers is based on a com-
separately because there are two separate parison between the nominal alternating
load-stress relationships: one for the stresses at failure for clevises, Sbc, and
clevis lugs (Eqs 1 and 3) and another levers, SbL , for specimen series that had
for the levers (Eqs 2 and 4). However, failures in both elements (see Table I I ) .
the fatigue behaviors of these two Except for the steel specimens A-1 and
elements are interdependent. Therefore, B-1, all these series had a tjti ratio of
the interaction between the two elements 0.333 and a relationship SbL = 3.6 Sbc
and the pin, and the influences of one (defined by the /1//2 ratio). The effects
element on the fatigue behavior of the of the various factors in the clevises and
other element, must be considered. An levers should be on the same order of
empirical treatment of the various magnitude, except for the plastic
factors included in Eq 9 is given below. deformations and the steady stress
levels. The clevises had 50 per cent
Plasticity Factor, P: higher nominal steady stresses in all
As shown in Table II, the nominal these specimens (with ti/li = 0.333).
alternating bending stresses at failure, Thus, the plasticity factor for the levers
SbL , at the outer surfaces of the levers of these series should be somewhat lower
were in all cases considerably higher than the value 3.6. For simplicity, a
than the nominal alternating bending plasticity factor of PL = 3.0 is assumed
stresses at failure, Sbc, at the inside for all levers. Such an assumption may
surfaces of the clevis lugs. In those be in error for levers that do not fail but
specimens that failed in clevis lugs, the should not affect the correlation or pre-
nominal steady tensile stresses in the diction of alternating stresses at failure
clevises were higher than those in the in levers that fail.
levers. This difference in nominal steady
stresses alone, however, cannot explain Steady-Load Factor, L:
the large differences in the nominal The effect of steady-load or mean-
alternating stresses, even with the effects stress levels should increase with the
of the pin bending considered. mean stress, S. From test results
The failure origins in the levers were obtained on the same steel specimen
generally located some distance from the configurations (series) at two steady-
outer surfaces toward the inside of the stress levels and on other test series at
levers. This indicated that plastic defor- various steady-stress levels, it was found
mations had taken place at the outer by trial that the following simple linear
surfaces, alleviating the peak stresses function affords a rather good correlation
from stress concentrations and work among all experimental results:
hardening the material in the yielded 35
regions between the origin and the outer + 1. (10)
10*
surface. Hence, the levers had sustained
the higher nominal alternating stresses. This relationship is applied to both
No evidence of plastic deformation was clevises and levers in both steel and
found in the clevis lugs except for some aluminum joints. I t should be mentioned,
140 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
however, that no test data were available surface finish and aluminum is usually
at two steady-stress levels for aluminum somewhat less notch-sensitive than steel
specimens. Therefore, the validity of Eq at the strength level used in the experi-
10 for aluminum joints cannot be verified,
except that it appears to fit approxi-
mately the test results.
The selected form of the L-function
fulfills the boundary condition in that L (a) \ w •-Load
becomes 1.0 (no steady-stress effect)
when the steady stress, S, is zero.
Combined Factor, FNQ:
The fretting factor, F, and the notch-
sensitivity factor, ..V, may both change
with the joint configurations and other
factors. The surface-quality factor, Q,
\ (b)
represents the clevis-lug flexibility and, original axis of the clevis-lug holes, the
thus, the capability of the lugs to adjust stresses induced at the critical locations
themselves to the pin deflections. At of clevis lugs will be about proportional
S T E E L J O I N T S , FXQ = 2.0
B-1 2.46
15 000 I 1,20 1 5 940" 7 500
1.38
20 700
30 000 4 310° 15 000 17 000
15 000 5 050" 10 000 17 200"
B-2. . . . 2,46 22 500 1.85 1 4 290" 15 000 1,85
15 100"
B-3 2.46 15 000 3.39 3 710 15 000 2.77 12 460°
low h/w ratios, the lugs will flex more to the lug thickness, h . Thus, both above
readily, and therefore, the peak stresses effects cause higher stress concentrations
at the critical locations will be alleviated. at higher h/w values.
Also, at a given misalignment between The ratio t^/w included in Eq 13 repre-
the axis of the deflected pin and the sents the pin fixity provided by the
144 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
levers. At high Ujiio ratios the lever will thus, the stress-concentration factor for
restrict the pin deflections more ef- the clevises depends also on the geo-
fectively than at low fe/w ratios, and metric configuration of the levers.
12,000
fc 2000
d/w
FIG. 5.—Nominal Bending Stress as Function of d/w Ratio for Steel and Aluminum Clevises at
15 X 106 Cycles.
0.300
0.4
t/t^
FIG. 6.—Nominal Bending Stress as Function of /1//2 or /i/w Ratio for Steel and Aluminum
Clevises at 15 X 10" Cycles.
MiTTENBERGS ON F A T I G U E STRENGTH OF PiN-LOADED LUG JOINTS 145
F--I^^^^E-2
5000 4000
R-l^^^
FIG. 7.—Nominal Bending Stress as Func- FIG. 8.—Nominal Bending Stress as Func-
tion of t\/w or h/w Ratio for Steel Clevises at tion of Joint Dimensions for Steel Clevises at
15 X 10" Cycles. 15 X 10" Cycles.
term d/d^, the variation in pin-bending There are close similarities among the
effect for similar aluminum joints of four equations (Eqs 13 to 16) derived
various sizes should be lower than for empirically, and the differences among
similar steel joints, and it may be them can be readily explained. It also
neutralized by the usual "size effect" should be pointed out that the stress-
found in fatigue applications. No test concentration factor, Kp, generally
data, however, were available to verify increases with a decreasing d/w ratio,
this conclusion. as does the theoretical stress-concentra-
The stress-concentration factors, Kp , tion factor, Ki. Thus, joints with low
for steel and aluminum levers are, d/w ratios have a high over-all stress-
respectively: concentration factor, K.
(1 - d/w)
K^t-si = 7.4 {h/w) , ALTERNATING STRESSES AT FAILURE
{d/w) {d/d,)
30,000
:S 5000
<
FIG. 9.—Nominal Bending Stress as Function of d/w Ratio for Steel and Aluminum Levers at
15 X 10« Cycles.
0.525
t,/t2
FIG. 10.—Nominal Bending Stress as Function of h/h or h/w Ratio for Steel and Aluminum
Levers at 15 X 10« Cycles.
MiTTENBERGS ON F A T I G U E STRENGTH OF PiN-LoADED LUG JoiNTS 147
the stress-range diagrams used in the metric configurations, thus reflecting the
above computations are listed in Fig. 2. relatively stiffer behavior of the steel
The steady-load factors, L, were com- pins in aluminum joints. As compared to
puted from Eq 10; the theoretical stress- the theoretical stress-concentration fac-
concentration factors, Kt, were taken tor, Kt, the values of Kp may be higher
from Fig. 3; and those of the geometric than Ki (mostly for low d/w ratios) or
effects, Kp, were computed from Eqs 13 lower (for high d/w ratios). Since the
through 16. Table III lists the values of geometric factor, Kp, expresses mainly
both factors, Kt and Kp, and gives the the effects of pin bending, it can be
nominal steady-stress levels, Sc and stated that the contribution of pin
t|/W
0.267 0.400 0.533
tg/w
FIG. 11.—Nominal Bending Stress as Function of li/w or tt/w Ratio for Steel and Aluminum
Levers at 15 X 10« Cycles.
SL , and the empirical values for factors bending toward increasing the stress
P and FNQ. concentrations in pin-loaded lug joints
As can be seen from the table, the is quantitatively significant and, de-
computed geometric factors, Kp, vary pending on the joint geometry, may be
over a wide range (0.48 to 11.94) de- higher or lower than the effects of the
pending on the joint configurations. theoretical stress-concentration factor,
The values for levers, KpL , are of the Kt.
same order of magnitude as those for A comparison between the computed
clevises, Kpc , and are lower than the alternating stresses at failure (Table III)
latter for some configurations and higher and those obtained in the experimental
for others. The computed Kp values for program (Table II) shows a general
aluminum joints are generally lower than agreement for the failing element (a few
those for steel joints of the same geo- exceptions are discussed below). For the
148 SYMPOSIUM ON FATIGUK OF AIRCRAFT STRUCTURES
element which did not fail in the experi- considerably lower than the computed
mental program (or where the predomi- value (Figs. 9 and 10). All joints of the
nant failure location was in the other C-series had large-diameter pins (high
element), the computed stresses are dlw ratio), and therefore the pin bending
higher than those hsted in Table I I in did not cause as high stress concentra-
all cases. In this instance, the computed tions as in the specimens with thinner
values indicate the stress level at which pins. For that reason, there was obvi-
this element would fail if the failure did ously less plastic deformation and a
not occur in the other element. smaller effect of plasticity in joints C-2
The computed nominal alternating- than that predicted by a plasticity factor,
bending stresses at failure for ranges of P — 3.0, used in computations for all
variables are shown graphically by con- levers. The latter conclusion implies that
tinuous curves in Figs. 5 to 11 as func- for large-diameter pins a smaller plas-
tions of the various variables (for the ticity factor should, perhaps, be used.
failing element only). Also shown in the For example, a plasticity factor of
figures are the data concerning the ma- P = 2.2 would make the computed SbL
terials, steady-stress levels, geometric value for C-2 joints agree with the experi-
ratios, and one of the joint dimensions. mental value. Because of the limited
For comparison, the nominal alternating experimental data available, however,
bending stresses at failure, determined no attempt was made to introduce a
experimentally (hsted in Table II), are variable plasticity factor, P.
shown in these figures as single points With the exception of the discrepancies
designated by the symbols of the various discussed above and some minor dis-
test series. crepancies that may be attributed to the
As can be seen from Figs. 5 to 8, the scatter, the correlation between the
agreement between the computed and coniputed and experimentally deter-
the experimental alternating-stress val- mined alternating stresses for the levers
ues for the clevises is rather good. Some was satisfactory, as illustrated in Figs.
of the discrepancies could, perhaps, be 9 to 11.
attributed to the scatter in the experi-
SUMMARY
mental data. In Fig. 6, only one test
point per curve (per steady-stress level) The preceding analysis and discussion
was available, except for steel clevises at indicate validity of an empirical approach
15,000 psi. for failure prediction in pin-loaded lug
For levers (Figs. 9 through 11), the joints. It was shown that the nominal
agreement is not as good as for the alternating stresses at failure can be ex-
clevises because of (1) effects of plastic pressed in terms of geometric parameters
deformations occurring in the levers and and other variables. A rather good corre-
(2) changes in part interaction in the lation was established between computed
series C-2.5 and C-3 joints (1,3). The and experimental stress values for the
alternating-stress distribution pattern in clevis lugs. For levers the agreement was
the series C-2.5 and C-3 specimens was not so good, mainly because of plastic
different from that in other .specimens, deformations occurring in the levers and
causing lower stress concentrations from because of the change in part interaction.
pin bending, and therefore these two lest The observations of the experimental
points (Figs. 9 and 10) do not fall in the program, supported by the analysis
category of all other specimens. results, indicated that pin bending
For series C-2 joints, the experimental contributes significantly to the stress
alternating-stress value at failure was concentrations and, thus, toward lower-
MiTTENBERGS ON F A T I G U E STRENGTH OF PiN-LOADED L u G JOINTS 149
ing the fatigue strength of pin-loaded lug various conditions. It is also believed
joints, particularly of those with rela- that the basic approach of this analysis
tively thin pins. The effects of fretting may be applicable to structural members
were significant in lowering the fatigue (other than pin-loaded lug joints) in
strength of both clevises and levers. In general.
levers, however, the effect of fretting was At the present, this method can be
more than offset by the effect of plastic used for correlating test data after they
deformations cccurring at the outer are available. However, if enough similar
surfaces. analyses were conducted and the various
It should be realized that this analysis factors involved were more firmly estab-
was based on a somewhat limited lished, a procedure may evolve whereby
amount of experimental data (1). Because these factors could be estimated on some
of insufficient data, some assumptions realistic basis without conducting a large
had to be made and some uncertainties number of tests. This could lead to design
had to be resolved on the basis of procedures that do not require prior
judgment. Therefore, it is considered testing, except perhaps for some spot-
that the usefulness of this analysis lies check or verification tests. Furthermore,
primarily in the approach used and in such an approach may eventually lead
the method developed, rather than in to the formulation of an analytical
the numerical values derived for the method for estimating fatigue behavior
empirical constants, which may change of structures.
for different conditions. A cknowledgment:
In this analysis only two variable
This research was supported by the
numerical factors (the combined factor,
Air Force Systems Command, United
FNQ, and the numerical values included
States Air Force, under Contract No.
in Eqs 13 through 16) were used to
AF 33(616)-6751, Department of the
establish the correlation. Even then, it
Army Project/Task No. 9R38-01-017-22.
was observed that the method used
It was monitored by the V/STOL
permits slight simultaneous changes in Propulsion Branch, Directorate of De-
these factors without greatly affecting fense and Transport Systems Engineer-
the correlation. If more variable factors ing, Aeronautical Systems Division,
were employed, the correlation could be Wright-Patterson Air Force Base, Ohio,
inaproved further. Thus, the niethod is with Hollis A. Cochran acting as project
quite elastic, and therefore it is believed monitor. The original research work was
that the approach and the method described in reference (3). Since then, the
presented in this paper are potentially author has reworked a part of the
useful for analyses of the fatigue strength analysis, particularly concerning the
of pin-loaded lug joints in general, under aluminum joints.
REFERENCES
(1) A. A. Mittenbergs and I.. G. Beall, Jr., (3) A. A. Mittenbergs, L. G. Beall, and H. J.
"Fatigue Strength of Pin-Loaded Lug Grover, "Fatigue Investigation on Pin-
Joints," Proceedings Am. Soc. Testing Joined Lug Connections," ASD Technical
Mats., Vol, 62, p. 710 (1962).
Report 61-163, United States Air Force
(2) S. J. Ligenza, "On Cyclic Stress Reduction
(1961).
Within Pin-Loaded Lugs Resulting from
Optimum Interference Fits," Experimental (4) R. E. Peterson, Stress Concentration Design
Mechanics, Soc. Experimental Stress Analy- Factors, John VV'ile)' and Sons, Inc., New
sis, Vol. 3, No. 1, p. 21, Jan., 1963. York, N. Y., 1953, p. 99.
STP338-EB/Sep. 1963
B Y CARL 0 . ALBRECHT 1
SYNOPSIS
Many full-scale component fatigue limits are developed on the basis of less
than five tests. Use of statistical concepts on such a limited number of speci-
mens would result in extremely low and unrealistic allowable stresses for any
reasonable probability and confidence level.
Fatigue data from element tests and full-scale component tests, where
many specimens have been tested, can be used to evaluate a coefficient of
variation that is representative of the material and type of component being
considered. By use of this information, and with the application of basic sta-
tistical techniques, a design S-N-P (probability) curve can be determined for
components where only a limited number of tests are available, for any proba-
bility and confidence level desired. The methods employed for the evaluation
of available data are discussed.
Typical values of coefficient of variation are tabulated for components of
typical steel and aluminum alloys such as 4130 and 4340 alloy steel, 150,000
to 160,000 psi ultimate tensile strength, and 2024-T4, 2014-T6, and 7075-T6
aluminum alloys, which contain stress concentrations or areas of fretting
corrosion or both.
Suitable probability and confidence levels are selected for components
where a single noninspectable failure would result in a catastrophic loss.
Determination of a design fatigue limit can be obtained directly by using
the method described in this paper or by applying a factor of safety to the
average test fatigue limit that simplifies the determination of an appropriate
design S-N-P curve for the designer and structures engineer. A table is pre-
sented showing the factor of safety required to ensure a minimum desired
probability and confidence level when only a limited number of test specimens
is available.
150
Rotating beam:
Smooth 0.40 X 102 0.30 (27)
2024-T4 aluminum. Notched 0.65 X 102 0..35 (27)
Tension-tension:
Smooth, ij = 0 0.40 X 102 0.36 (27)
Rotating beam
Smooth 0.48 X 102 0.30 (27)
2014-T6 aluminum. Notched 0.74 X 102 0.34 (27)
Tension-tension:
Smooth, fl = 0 0.25 X 102 0.30 (27)
Rotating beam:
Smooth 0.76 X 102 0.40 (27)
7075-T6 aluminum. Notched 0.51 X 102 0.30 (27)
Tension-tension;
Smooth, R = 0 1.02 X 102 0.44 (27)
AZ80-A and ZK60-A mag Tension-tension:
Smooth, R = 0 0.23 X 102 0.35 (33)
Rotating Beam:
Smooth, 140 000 psi UTS 3.50 X 10' 1.00 '3)
4340 steel. Smooth, 190 000 psi UTS 2.10 X 10" 1.00 (3)
Notched, 140 000 psi UTS 0.92 X 10* 0.89 (3)
Notched, 190 000 psi UTS 0.47 X 10* 0.85 (3)
Tension-tension:
4130 steel. Smooth, 140 000 psi UTS 3.50 X lO-" 1.00 (25)
Notched, 140 000 psi UTS 0.92 X 10" 0.89 (25)
Tension-tension:
18-8 stainless steel 14 hard Smooth, R = 0 0.13 X 10I* 0 74 (25)
The values given by the above equa- evaluated for various materials (3,25,27,
tion when plotted on semilog paper show 28) and are shown for reference in Table I.
satisfactory correlation for failure To evaluate the constants a and 7,
1.00
Reference (3) Fig. 15
V
0.50 \
1
s . S-S
• \. "' S
1 Slope, T = 1.0
For i4 ^ = 0 1
0.20
a: 0.1 N S
= 0,1(3.5 x i o ' i s
Jiw 0.10 = 3.5x 0<S
1
0.05 \
\ .
0.02
i t \
0.0\
2 5
N, Cycles
loox lo-
go
80
.- 70
'60
(A
= 68 0 0 0 * 3.5X10 ( 6 8 . 0 0 0 )
t/>
4U '- N'-°° -. ^
<n
50
40
30
20
10
10" 5 10° 5 10^ 5 lO' 2 5 10°
0 N, Cycles
stresses not too far above the fatigue a mean S-^ curve is drawn through the
limit. test data using the method of least
The material constants a. and 7 were squares. A plot is then made on log-log
154 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
paper using (5 — S)/S as an ordinate log paper, it varies from tiie mean S-N
and N as the abscissa. The points when curve shape by no more than 5 per cent,
plotted in this manner lie in nearly a An example of such a plot for smooth,
60
Reference (3) Fig, 15.
4 3 4 0 Steel, 140,000 psi
\
50 Ultimate Tensile Strength
Smooth Rotating Beam,R=-l
Normal Curve
J. (a)
W
Log-Normal Curve
li.
o 30
u I
/A
I
\ \
10 viZ
>-^^.
56 58 60 62 64 66 68 70 72 74 76 78x10^
0
Alternating Stress at i d ' Cycles
40
i
L o g - N o r m a l Curve
20
%
ja
E
''\
10
li \
2 4 6 8 10 12 14 16 18 20xl0'
Alternating Stress at I o ' Cycles
straight line. Some curvature does exist, unnotched 4340 steel with a stress ratio
but when the curve is straightened out of i? = — 1 obtained by using data from
and replotted as an S-N curve on semi- reference (3) is shown in Fig. 1. From
ALBRECHT ON STATISTICAL EVALUATION or TEST SPECIMENS 155
T A B L E I I . — C O E F F I C I E N T O F V A R I A T I O N , V. F R O M LARGE-SCALE
FATIGUE TESTS.
NOTES:
1. All of t h e following d a t a are based on testing ranging from a stress ratio R = — 1 t o a stress
ratio R = 0. In applications where stress ratios are appreciably greater t h a n 0, test d a t a should be
e v a l u a t e d t o determine effect of m e a n stress on V.
2. N o a t t e m p t h a s been m a d e t o e v a l u a t e t h e effect of stress c o n c e n t r a t i o n on t h e notched
coefficient of variation.
3. Failures occurred in areas of fretting.
4340 (steel) 150 000 3.9 144 3.1 139 (3), (25)
2024-T4 ( a l u m i n u m ) . 57 000 9.6 94 14.5 82 (27)
2014-T6 ( a l u m i n u m ) . 60 000 8.5 170 15.2 107 (27)
7075-T6 ( a l u m i n u m ) . 79 000 13.8 131 16.0 112 (27)
AZ80 (magnesium) . . 53 000 6.1 34 7.9 26 (28)
ZK60 (magnesium) . . 48 000 7.1 26 7.1 21 (28)
D-Design
F-Failure
L-Lower Limit
U-Upper Limit
Stress
FIG. 4.—Probable Variation of the Mean.
w h e r e (from E q 10)
TABLE III—PROBABILITY VERSUS a
RELATIONSHIP (KEFEREN'CES (5) AMD
(32)). "Ln = log
a Probability a Probability
then
0.0 . . 0.500 3.00 . . . . , . 0.99865
l.OO. . 0.84134 3.09 . . . . . . 0.99900
1.28. . 0.90000 3.29.... . . 0.99950 'og -z^ = log
1.64 . , 0.95000 3.72 . . 0.99990 •JO
2.00. . 0.97725 3.89-. .. . , 0.99995
2.05. . 0.98000 4 . 0 0 . .. . . . 0.99997 or
2.33. . 0.99000 4.2(5... . . . 0.99999
2.57 . . 0.99500 4.75 . . 0.999999
^ = ( ^ \ ^ - 1
So V - V) "' SL ~ / 1 Y
SL and a five per cent probability that
SF will fall above Su, then there is a 90 By subsitution of SL from Eq 9
per cent probabiUty that SF will fall
between these limits. ID 1
.(12)
SF / 1 \ {o+a'/V»)
Analytical Evaluation of Design Fatigue
Limit:
With the establishment of SL , the The next requirement imposed on the
next item under consideration is the derivation is the selection of an ap-
determination of the design fatigue propriate survival rate. Since the scatter
limit, SD • that normally exists in full-scale tests,
imposing a requirement of greater than
For a normal distribution:
0.9999 survival with a normal distribu-
SD — SL — aff„ where o-„ = VSL tion would result in an allowable design
stress that can be unrealistically low,
then and in some cases reduce the allowable
SD = SL (1 - aV)
to zero.
A literature review indicates that
By substitution of SL from Eq 8, distributions other than normal have
ALBRECHT ON STATISTICAL EVALUATION OF T E S T SPECIMENS 159
Factor of SD
Safety
2.50
0.50
0.40
\iA N.
S
333 030
X
500 0.20 ^ \ Is PFR CFN T <: IIRVIV Al
SF
One Test Specimen
1000 J 0.10 /—
90 P« r Cen1 Conf de ice Le vel
0.01 Ol 5 10 30 50 70 90 95 99 9 9 9 99.99
Per Cent Survival
Factor of SD
Safety ^-
SF
1.00 -^ 1.00
1.1 1 - 0.90 1
1.25 - 0.80 "^ —-^^^.OF.
0.70
\ J
143 -
167 - 0.60 ^^^r^ '^^
2 00 - 0.50 L ^ ^ *
^ v . ' ' -
2.50 - 0.40
(b)
r^
333 - 0.30
4.00 - 0.25
5.00 - 0.20
SD
667 - 0.15 — V i PER CFN1 SURVIV/W
SF
OneTe St Sp«cirnen
Coefficie ntof V ar otion ,\1
90 P er CenI Conf de nee Level
1000 -• 0.10 I 1
0.01 - 0.1 - I 5 10 30 50 70 90 95 99 99.9 99.9?
Per Cent Survival
Factor of
Safety
SF
99,99 99.9 9 9 95 90 70 50 30 10 5 Ql 001
1.00 1,00
III 0.90
^ ^
1.25 O80
^vt; 3D/„
1.43 070 ^ N ^
0,60
\/-
1,67 H X
2,00 050
Co)
^1
2.50 040 % '
3 33 - 0.30
So
5,00 - 0.20 — V s PER CENT SIJRVIV l\l
SF
Four Test Specimens
10,00 0.10
90 F'erCent Confidence Level
1 1 1
"0,01 0.1 I 5 10 30 50 70 90 95 99 999 9 9 9 9
Per Cent Survival
Fbctor of So
Safety
5,00 - 0,20
So ,
6,67 - 0.15 — V i PER CENT SU RVIVA
bF
F()ur Tes t'Spe cim ens
Coe ftiden of V irio tion,V
90 F>er Cent Confi dence Lev(!l
1000 0.10
0 01 0.1 5 10 30 50 70 90 95 99 99.9 99.99
Per Cent Survival
r = 0.05 V = 0.10
Number of Specimens
Mean Stress, Design Stress, Factor of Mean Stress, Design Stress, Factor of
SD Safety SD Safety
V = 0.15 V = 0.20
Number of Specimens
Mean Stress, Design Stress, Factor of Mean Stress, Design Stress, Factor of
SL SD Safety SL SD Safety
V = 0.05 V = 0.10
Number of Specimens
Mean Stress, Design Stress, Factor of Mean Stress, Design Stress, Factor of
SL SD Safety SL SD Safety
V = 0.15 V = 0.20
Nun^bet of Specimens
Mean Stress, Design Stress, Factor of Mean Stress, Design Stress, Factor of
SL SD Safety SL SD Safety
normal distribution and for a single test tion of the flight load spectrum be used.
specimen is 5.3. This may seem severe, For example, only maximum measured
but it is statistically correct for the loads resulting from helicopter operation
probability of survival required. For a at the most critical center of gravity,
log-normal distribution, however, a rotor rpm for each forward speed and
factor of safety of only 3.31 is required, maneuver are to be used in the calcula-
thus justifying the importance of the tion of fatigue life. This approach
proper distribution. usually results in reasonable safe life
Another item of significance that can predictions.
be deduced from Table IV is the effect With the aid of references (35) and
of the number of tests conducted. For a (36), an evaluation of mission profiles
given known coefScient of variation, and statistical evaluation of flight
very little difference exists between the stresses can be made. Calculation of
factor of safety for 1 specimen or 20 component life using the method of this
specimens. The most significant change paper results in somewhat lower lives
exists between one and two specimens. than are determined by using CAM 6
This would not be true if the standard (34). More important than actual life
deviation were estimated from a limited magnitudes, however, is the better under-
number of tests, in which case the standing and greater reliability in pre-
number of specimens would have a large dicting safe life afforded by the proposed
effect on the factor of safety. method.
Another factor of safety approach is
proposed in reference (37). A factor of STRUCTURAL COMPONENT L I F E
safety approach is also presently a re- DETERMINATION
quirement of the FAA Civil Aeronautics The life can be determined by first
Manual 6, Appendix A (34): obtaining the appropriate value of SD
"The minimum number of test specimens by dividing the average fatigue limit of
required is dependent on the oscillatory test the few tests conducted by the factor of
level, in the following manner: safety required. The allowable S-N-P
"(a) A minimum number of 4 test speci- curve is then plotted. A modified form
mens if the oscillatory level is chosen at 1.1 of Eq 6 is used for this purpose:
times the critical oscillatory stress level.
"(b) A minimum of 3 test specimens if the
oscillatory level is chosen at 1.25 times the S = So + Ny
critical oscillatory stress level.
"(c) A minimum of 2 test specimens if the a and y are selected from Table I.
oscillatory level is chosen at 1.5 times the Since a and 7 are based primarily on
critical oscillatory stress level. tests conducted with a stress ratio R =
"(d) One specimen if the oscillatory level
0 or i? = — 1 , the design S-N-P curve
is chosen at twice the critical oscillatory
stress level." should be corrected by use of an ap-
propriate Goodman diagram for the
For the above requirement, the lowest effects of steady stresses or loads. Cumu-
failure level is used as a design fatigue lative damage theory can then be used
limit instead of a mean fatigue limit in conjunction with measured data to
reduced for given confidence and survival determine the life.
rates, as proposed in this paper. In order to be completely fair in the
Civil Aeronautics Manual 6 (CAM 6) life calculation, measured load and stress
(34) requires that a conservative defini- data should be evaluated statistically. A
164 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
REFERENCES
(1) H. Grover, S. Gordon, and L. Jackson, (12) H. N. Cummings, "Some Quantitative
"Fatigue of Metals and Structures," .Aspects of Fatigue of Materials," WADD
NA VAER 00-25-534. TR 60-42, July, 1960.
(2) W. Weibull, "A Statistical Representation (13) W. Weibull, "Scatter of Fatigue Life and
of Fatigue Failures in Solids," Transactions, Fatigue Strength in Aircraft Structural
Royal Inst. Technology, Stockholm, Materials and Parts," Fatigue in Aircraft
Sweden, 1949. Structures, Academic Press, Inc., New
(3) H. Cummings, F. Stulen, and W. Schulte, York, N . Y. (1956).
"Investigation of Materials Fatigue Prob- (14) Symposium on Statistical Aspects of
lems Applicable to Propeller Design," Fatigue, ASTM STP No. 121, Am. Soc.
WADC TR 54-531, May, 1955 Testing Mats. (1951).
(4) E. H. Schuette, "The Significance of Test (15) Symposium on Fatigue with Emphasis on
Results from Small Groups of Specimens," Statistical Approach, ASTM STP No. 137,
Proceedings, Am. Soc. Testing Mats., Vol. Am. Soc. Testing Mats. (1952).
57, p. 1304 (1957). (16) T. T. Oberg and E. J. Ward, "Fatigue of
(5) G. J. Schott, "The Statistical Significance Alloy Steels at High Stress Levels," WADC
of a Few Fatigue Results," National TR 53-256, Oct., 1953.
Aeronautical Establishment Laboratory Re- (17) G. E. Dieter and R. F. Mehl, "Investigation
port LR-58, May 26, 1953. of the Statistical Nature of the Fatigue of
(6) E. Epremian and R. F. Mehl, "Investiga- Metals," Af/IC^ TN 3019, Sept., 1953.
tion of Statistical Nature of Fatigue (18) J. K. Childs and M. M. Lemcoe, "Fatigue
Properties," NACA TN-2719, June, 1952. Investigation on High Strength Steel,"
(7) "A Tentative Guide for Fatigue Testing WADC TR 56-205, July, 1957.
and Statistical Analysis of Fatigue Data," (19) E. H. Schuette, "Fatigue Properties of
ASTM STP No. 91-A (1958). Magnesium Alloy Forgings," WADD TR
(8) H. E. LeSueur, "The Certification of Civil 60-854, D e c , 1960.
Transport Rotor Craft with Particular (20) H. J. Grover, S. M. Bishop, and L. R.
Reference to Multi-Engines," Journal, Jackson, "Fatigue Strengths of Aircraft
Royal Aeronautical Soc, Aug., 1961. Materials, Axial-Load Fatigue Tests on
(9) A. M. Freudenthal, "Fatigue Sensitivity Unnotched Sheet Specimens of 24S-T3
and Reliability of Mechanical Systems, and 75S-T6 Aluminum Alloys and of 4130
Especially Aircraft Structures," WADD Steel," NACA TN 2324, March, 1951.
TR 61-53, July, 1961. (21) H. J. Grover, W. S. Hyler, and L. R.
(10) J. S. Lariviere, "Method of Calculation to Jackson, "Fatigue Strengths of Aircraft
Determine Helicopter Blade Life," Transla- Materials, Axial-Load Fatigue Tests on
tion by J. Koch, Journal, Am. Helicopter Notched Sheet Specimens of 24S-T3 and
Soc, April, 1961. 75S-T6 Aluminum Alloys with Stress
(11) H. N . Cummings, "Quahtative Aspects of Concentration Factor of 1.5," NACA TN-
Fatigue of Materials," WADC TR 59-230, 2639, Feb., 1952.
Sept., 1959. (22) H. J. Grover, S. M. Bishop, and L. R.
166 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
SYNOPSIS
This paper presents four aspects of the helicopter fatigue problem as viewed
by the Federal Aviation Agency. They relate to: (1) development of the
fatigue requirements for civil helicopters; (2) application of these fatigue
requirements; (.?) history of failures known to have been attributed to or
influenced by fatigue; and (4) future prospects for new and improved means
to cope with fatigue evaluation problems.
Under Section 601(a) (l)2 of the Federal and the current civil certificating au-
Aviation Agency Act of 1958, "The thority, the Federal Aviation Agency.
Administrator is empowered and it shall The standards and their various means
be his duty to provide safety of flight of of interpretation have evolved over the
civil aircraft in air commerce by prescrib- years primarily through the combined
ing and revising from time to time, such efforts of many interested parties in the
minimum standards covering the design, aviation community, both within and
materials, workmanship, construction outside Government, and to the best of
and performance of a i r c r a f t . . . as may our knowledge are reasonable, realistic,
be required in the interest of safety." minimum safety standards.
Section 603(a) of the same act places
DEVELOPMENT OF FATIGUE
the responsibility on the Administrator
REQUIREMENTS
to conduct such findings as are necessary
to establish that the aircraft is of proper The early design criteria for helicopters
design and meets the minimum stand- were based on the experience gained by
ards, rules, and regulations prescribed by the autogiro makers. Cierva in his text-
the Administrator. book, The Theory of Stresses in Autogiro
The airworthiness standards for heli- Rotor Blades, cited two empirical factors,
copters, presented in Parts 6 and 7 of the concerning fatigue, to be used in design.
Civil Air Regulations (2,3), contain One was related to material fatigue
specific provisions on structural fatigue. characteristics and the other was a
These fatigue criteria have been the correction for stress concentrations.
subject of extensive interpretative ma- The first comprehensive set of air-
terial issued by the predecessor agency, worthiness standards for rotorcraft was
the Civil Aeronautics Administration, put forth in Civil Aeronautics Adminis-
tration Engineering Report No. 32,
1 "Proposed Rotorcraft Airworthiness,"
Assistant Chief and Aerospace Engineer,
respectively. Airframe Branch, Engineering and dated Feb. 16, 1943. This document
Manufacturing Div., Flight Standards Service, covered the basic safety parameters
Federal Aviation Agency, Washington, D. O.
2 such as speeds, loading conditions, load
The boldface numbers in parentheses refer
to the list of references appended to this paper. factors, proof of strength, autorotation
167
Copyright^ 1963 by ASTM International www.astm.org
168 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
capability, and fatigue. The fatigue stresses do not exceed safe values for con-
requirement was as follows: tinuous operation."
"At the critical helicopter airspeed and This regulation did not imply that
rotor rpm and under hovering conditions the safe life or even infinite lives had to be
hub, blade attachments, blade controls established, though the manufacturer
and mechanisms essential to operation as an could do so on the basis of fatigue tests
autogiro which are under cyclic flexing or or by rational application of material
alternating stresses, shall be adequately fatigue properties modified to account
investigated for fatigue strength against the for stress concentrations.
endurance limit of the materials used."
As additional experience was gained it
The evaluation was based on materials became apparent that measured stresses
data only; no measurment of stresses or in some instances were above the
fatigue testing was required. The criteria endurance limit of the material and the
in this report served as the basis for evaluation methods using material cor-
issuance of type certificates for the rection factors were found to be inade-
world's first civil helicopters—the Bell quate. It was evident that fatigue testing
Model 47 and the Sikorsky Model S-51. to determine safe lives was a vital part
On April 3, 1946, a certificated heli- of fatigue evaluation. This provision,
copter with less than 100 hr total time introduced into Section 6.250(a), Part 6,
crashed. The cause was fatigue failure of of the Civil Air Regulations on Jan 15,
the rotor hub. This raised questions 1951, reads as follows:
about the validity of the analytical
"The hubs, blades, blade attachments,
procedure used, particularly with regard
and blade controls which are subject to
to the stress levels present and the alternating stresses shall be designed to
strength characteristics of the fabricated withstand repeated loading conditions. The
parts. The Civil Aeronautics Administra- stresses of critical parts shall be determined
tion subsequently urged that flight strain in flight in all attitudes appropriate to the
measurements be made and that the type of rotorcraft throughout the ranges of
results be used in testing full-scale com- limitations prescribed in Section 6.204. The
ponents. service life of such parts shall be established
When the first Civil Air Regulations, by the applicant on the basis of fatigue tests
or by other methods found acceptable to the
Part 6, were adopted in May 26, 1946,
Administrator."
there was general agreement within the
helicopter industry and the certificating During the early 1950's, the rapid de-
agency on the need for strain gage velopment of the helicopter and the use
measurements on all parts of the rotor of the larger rotorcraft in transport-
system subject to possible fatigue condi- type operations resulted in a new Part 7
tions in flight. The specific requirement to the Civil Air Regulations effective
read: Aug. 1, 1956 (3), based on Civil Air
Regulations, Part 6 (CAR 6). Since the
"The hub(s), blades, blade attachments fatigue problems of both categories.
which are under cyclic flexing or alternating Transport and Normal, are quite similar
stresses, shall be substantiated to demon-
in nature, the requirements for main and
strate the airworthiness of these parts under
repeated loading conditions associated with auxiliary rotor assemblies read sub-
normal operation. The vibration stresses of stantially alike in both documents.
critical metal parts shall be determined in Very recently, CAR 6 was amended to
flight and it shall be demonstrated that these provide for fatigue evaluation of the
DOUGHERTY AND SPICER ON FATIGUE SUBSTANTIATION 169
fuselage and rotor pylon structure. The regulation, aimed at determining the
changes contained in Amendments 6-5 magnitude of steady and oscillatory
to Section 6.251(e) of CAR 6, effective stresses associated with normal opera-
May 3, 1962, brought the provisions of tion. Prior to conducting the flight strain
CAR 6 and CAR 7 substantially into
agreement on the matter of structural TABLE I.—FLIGHT MANEUVEKS—
fatigue. PER CENT OCCURRENCE.
Ground Conditions:
APPLICATION OF FATIGUE Rapid increase of rpm on ground to
REQUIREMENTS
quickly engage clutch 0.5
Taxiing with full cyclic control 0.5
With the adoption of the provisions Jump takeoff 0.5
Hovering:
contained in the January 15, 1951, Steady hovering 0.5
version of CAR 6, it was apparent that Lateral reversal 1.0
additional guidance material would be Longitudinal reversal 1.5
Rudder reversal 1.0
necessary to establish a minimum sub- Forward Flight Power On:
stantiation program leading toward Level flight, 20 per cent VNE 5.0
certification and to provide for reason- Level flight, 40 per cent VNE 10.0
Level flight, 60 per cent VNE 18.0
able uniformity in the administration of Level flight, 80 per cent VNE 18.0
this regulation. This was further neces- Maximum level flight (but not greater
sitated by the high cost of testing and the than VNE) 10.0
VNE 3.0
length of time required to perform tests. 111 per cent VNE 0.5
Hence, Appendix A to CAR 6, "Methods Right turns 3.0
Left turns 3.0
of Rotor Service Life Determination," Climb (max continuous power) 4.0
effective May 1, 1951, was developed. Cyclic and collective pull-ups from
This Appendix A has served for years as level flight 0.5
Change to autorotation from power-on
the mainstay for showing compliance flight 0.5
with CAR, Parts 6.221 and 6.250(a). Partial power descent (including condi-
tion of zero flow through rotor) 2.0
Although it would be desirable to stand- Landing approach 3.0
ardize on a method of resolving fatigue Lateral reversals at VH 0.5
problems, it is recognized that variables Longitudinal reversals at VH 0.5
Rudder reversals at VH 0.5
such as new design features, different Climb (takeoff power) 2.0
fabrication methods, and use of new Autorotation-Power Off:
materials have required variations from Steady forward flight 2.5
Right turns 1.0
the recommendations in Appendix A. Left turns 1.0
These variations in all respects have at Lateral reversals 0.5
least provided an equivalent level of Longitudinal reversals 0.5
Rudder reversals 0.5
safety. Cyclic and collective pull-ups 2.0
Landings (including flares) 2.5
In any rational method of determining
fatigue life of a structure, three basic 100.0
factors must be known: (1) the flight
maneuvers and associated stresses to survey, critical stress areas must be
be expected in normal operation, (2) the determined to insure proper distribution
frequency of occurrence of specific load- of the strain gages. Here engineering
ings, and (3) the fatigue strength char- judgment is important and use is made
acteristics of the structure. of such tools as brittle coatings, photo-
Stress levels must be found by care- elastic methods, analysis, and experience
fully controlled, instrumented flight with related projects. When performing
strain measurements, as required by the the survey, factors such as rotor rpm.
170 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
speed, altitude, weight, and center of for example, partial power operation for
gravity must be considered to ensure multiengine helicopters and the effect
complete coverage of critical conditions. of gust loadings.
Table I, taken from Appendix A of the In Appendix A of the Civil Air Regula-
Civil Air Regulations, sets forth a list of tions four simplified approaches are out-
occurrences consisting of some 33 maneu- lined to determine the fatigue strength
vers during which strain measurements characteristics. A number of empirical
are to be made. The objective is to collect factors used in these methods were
information on stresses that the heli- accepted by those concerned when Ap-
copter will be subjected to during opera- pendix A was adopted as being reason-
tion. Each maneuver is ascribed a able. The service experience to date has
percentage of time; therefore integration confirmed the validity of these factors.
of the stress envelope and the frequency Method 1.—This curtails the fatigue
Y i e l d Stress
Steady Stress
of occurrence can serve as the basis for a test program by establishing infinite
fatigue test program. The maneuvers are life for the parts in question at several
grouped under the following headings: measured stress levels. The approach
ground, hovering, forward flight, and hinges on construction of a modified
autorotation, and include conditions Goodman diagram as shown in Fig. 1,
such as clutch engagement, taxiing, level using polished specimen data for the
flight, turns, pull-ups, control reversals, material under consideration to establish
power recovery, and landing. This ap- the endurance boundary. This boundary
proach is not always applicable to the is reduced to account for the appropriate
problem; our exposures to some industry theoretical stress concentration factor
segments reveal instances where this for the component area involved, and the
spectrum has to be deliberately altered failure boundary is then reduced by a
in favor of other equally acceptable and factor of 3. If all of the operating stresses
often more sophisticated loading sched- fall below this boundary, infinite life can
ules. It is emphasized that this table is be established and no further testing is
a guide and must be tempered to suit the required.
anticipated operation of the helicopter, Method 2.—This calls for establishing
DOUGHERTY AXD SPICEE ON FATIGUE SUBSTANTIATION 171
infinite life through fatigue testing of In this method, the number of speci-
actual parts to show that the most mens tested to determine the endurance
critical stress measured in flight is less limit must be increased as the margin
T A B L E I I . — H E L I C O P T E R A C C I D E N T S , 1950-1961.
1950 42 9 3 0 0 3 2 0 1 0 0
1951 29 6 4 3 2 2 1 0 2 1 0
1952 26 4 i 0 0 1 1 0 0 0 0
1953 110 17 0 1 0 0 1 1 0 0 0 0
1954 137 36 5 2 0 0 2 1 0 1 0 0
1955 165 39 2 3 0 0 3 2 0 0 0 1
1956 256 53 12 3 1 0 3 3 0 0 0 0
1957 302 56 7 10 4 1 9 8 0 0 0 2
1958 370 83 9 12 5 1 11 9 1 2 0 0
1959 459 89 8 7 2 0 7 5 1 0 0 1
1960 543 75 11 12 6 1 11 7 0 1 1 3
1961 651 124 13 13 3 1 12 7 1 0 0 5
Total.... 670 86 71 24 6 65 47 3 7 2 12
T A B L E I I I . — S U M M A R Y OF H E L I C O P T E R A I R W O R T H I N E S S D I R E C T I V E S , "
1950 T O J U N E 1962.
N O T E . — T h i r t y - n i n e per cent of t o t a l Airworthines.s Directives concern airframe s t r u c t u r a l fatigue
discrepancies. Power p l a n t , cowling, drive system, and engine m o u n t discrepancies n o t included in
s t r u c t u r a l category.
Total "A" Models 'B" Models 'C" Models ' D " Models ' E " Models *F" Models
Structural
Total
Fatigue
Year Airwor- Struc- Struc- Struc- Struc- Struc- Struc-
thiness Total tural Total tural Total tural Total tural Total tural tural
Directives Fa- Fa- Fa- Fa- Fa- Fa-
tigue tigue tigue tigue tigue tigue
1950. 3 2
1951. 6 2
1952. 13 3
1953. 4 3
1954. 5 2
1955. 7 4
1956. 13 7
1957. 9 2
1958. 10 3
1959. 8 4
1960. 15 11
1961. 13 0
1962. 5 1
Total 111 44 50 15 19 13 19 10 13
than the endurance limit, 10' cycles for between the endurance limit and the
ferrous materials or 5 X 10'' cycles for critical flight stress decreases. To account
nonferrous materials. for statistical variation in allowable
172 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
cyclic stress from one specimen to an- ing stresses, is required to establish S-N
other, if one specimen is tested, a factor curves for each critical steady stress
of 2.0 is applied to the cyclic stress level. level.
This factor can be decreased to 1.5, 1.25,
or 1.1, respectively, if 2, 3, or 4 specimens In either method 3 or 4, the initial
are tested. service life is established at 75 per cent of
Method 3.—A finite life is established the calculated life, or 2500 hr, whichever
for a component based on tests in cyclical is less. If the calculated life exceeds 3333
K
LANDING GEAR ||4%
Z
1-
LANDING GEAR 1% NOTE: BASED ON 44 DIRECTIVES
ISSUED BETWEEN 19508 1962
z
z
Ul
o TAIL ROTOR BLADE a HUB • l<)%
s
o MAIN ROTOR BLADE a HUB ^ ^ ^ ^ 2 5 %
1-
CONTROL SYSTEM
J 1 1 1 1
PER CENT
units at a series of stress levels and a hr (2500/0.75), the initial life can be
number of cycles to simulate a random extended in increments up to 75 per cent
load sequence. Thus, the spectrum is a of the calculated life, provided the eval-
sequential loading representative of a uation of service experience on the life-
block of time such as 100 hr, repeatedly limited parts warrants such action.
applied. The life established is based on
SERVICE EXPERIENCE
the actual number of blocks completed
in the test program. The Federal Aviation Agency also has
Method 4.—Finite life is established by the responsibility to assure continued
use of an S-N curve and a cumulative airworthiness of aircraft already certifi-
damage analysis. A minimum of four test cated and in service. In the past decade
specimens, covering the range of alternat- we have witnessed a many-fold increase
DOUGHERTY AND SPICER ON FATIGUE SUBSTANTIATION 173
review of detail design to insure removal tory of the rotor-craft or the damage
of undesirable features of design. A vast effect of loads encountered in service in
reservoir of useful guidance information an orderly and systematic fashion. The
exists in the justification bases used for former should prove useful in establish-
issuance of the airworthiness directives ing a more precise loading spectrum. The
mentioned earlier and the accident latter would have as its primary objec-
records cited. Hopefully, this can be tive failure detection before progression
made available to designers in a useful beyond safe limits, and, with appropriate
form with the objective of reducing fatal service experience, could prove to be an
fatigue failures to an extremely low rate. adjunct to the operator responsible for
7. Encouragement by the FAA of the the conduct of a helicopter maintenance
installation of fatigue metering equip- program, including life-limited structural
ment capable of recording the load his- parts.
REFERENCES
(1) Civil Air Regulations, Part 4b: Airplane ing Conditions Experienced by Three
Airworthiness; Transport Category. Military and Mountain-based Commercial
(2) Civil Air Regulations, Part 6: Rotorcraft Helicopters," NASA TN D-432.
Airworthiness; Normal Category. (7) George Sines and J. L. Waisman, Metal
(3) Civil Air Regulations, Part 7: Rotorcraft Fatigue, McGraw-Hill Book Co., Inc.,
Airworthiness; Transport Category. New York, N. Y. (1959).
(4) "Strength of Metal Aircraft Elements," (8) "An Engineering Evaluation of Methods for
MILHDBK-5, March 1959.
the Prediction of Fatigue Life in Airframe
(5) Andrew B. Connor and Leroy H. Ludi, ".A
Summary of Operating Conditions Experi- Structures," ASD-TR-61-434, March 1962.
enced by Two Helicopters in a Commercial (9) Bo Lundberg, "Fatigue Life of Airplane
and Military Operation," NASA TN D 251. Structures," Report 60, The Aeronautical
(6) Andrew B. Connor, "A Summary of Operat- Research Inst, of Sweden (1955).
Design and Evaluation of
Full-Scale Structures
STP338-EB/Sep. 1963
Progress has been made in recent years For example, as part of certification
in studying the effects of aerodynamic requirements, realistic cyclic loads have
heating, which is a relatively new factor been applied to the B-58 bomber at
that must be accounted for in designing rates up to 14 cpm, thus requiring
and testing aircraft structures. A great the operation of as many as 64 hydraulic
deal of effort has been spent to develop rams simultaneously.
analytical methods that include the Since aircraft are designed to fly
effects of elevated temperature. Most continuously in elevated-temperature
such methods, applicable to full-scale environments, repeated-load testing ob-
structures, are approximations at best viously must produce the effects of
and make no accounting for strength accumulated exposure. Such effects in-
degradation due to transient conditions clude: (1) degradation of material
that vary with time. Therefore testing properties; (2) thermal stress, including
that must include time-dependent cyclic thermal stress fatigue; and (3) creep and
loads as well as static loads for maximum stress relaxation and their effects on
conditions continues to be a primary static and fatigue strengths of the struc-
tool of the structures engineer in evaluat- ture.
ing structures designed for sustained A test program for a full-scale alumi-
supersonic flight. num sandwich panel to produce such
Methods and techniques for static effects has been developed and conducted
loading, including simulation of aero- at General Dynamics/Fort Worth on
dynamic heating of full-scale structures, panels similar in construction to those
have been concerned with facilities and that have been proved in supersonic
equipment for heating the structure to a flight on the B-58 bomber. The purpose
given design temperature and then of this test was (1) to show the capability
applying static design loads. Until now, of bonded aluminum sandwich construc-
fatigue test programs of full-scale struc- tion for use in continuous Mach-2.4
ture have not been required to include flight, and (2) to establish the feasibility
aerodynamic heating effects. Attention of testing full-scale panels simulating
has been directed to developing loading cyclic aerodynamic heating in addition
devices which will provide high loading to cyclic loads.
rates for full-scale structures in a realistic A typical flight profile was established,
manner at room temperature. This has and parameters describing an aircraft
been accomplished reasonably well in that could fly such a profile were deter-
several full-scale fatigue test programs. mined. A delta-wing aircraft was as-
sumed with four engines mounted in pods
1
Senior Structures Engineer and Senior Test beneath the wing. Wing panels were
Engineer, respectively, General Dynamics/Fort
Worth, Fort Worth, Tex. assumed to act as fuel tank walls. An
analytical study was made to estimate thermal gradients, and flight loads on the
the effects of a sustained supersonic test specimen, all of which are outlined
flight environment on the wing panel in the following paragraphs.
t = 200 MIN
M = 2.4
W = 65 000 LB
a 40
t = 2.08 MIN
D 30 M=0.9 t = 206 MIN
W= 185 000 LB M= 1.0
W= 63 200 LB
20
<
10 t =0.85 NUN t = 217 Mm
M-.0.75 M=0.45
W= 190 000 LB W= 60 000 LB
50 100 150 200
TIME , MIN
FIG. 1.—Mach-2.4 Flight Profile.
OUTER V, UPPER
^3 2 SKIN WING SURFACE
OUTER LOWER
SI^N T4 Tj T ; WING .SURFACE
EDGE ^ ' I.I'l.l'iMia
IQl
10°
10
10-2
10-3
0 0.4 0.8 1.2 1.6. 0.4 0 . 8 1.2
' Incremental Gust Positive Incremental
Factor (An) Load Factor (An)
altitude than transport usage. A typical through the supersonic cruise region.
profile that created the structural During the descent phase, the heated
problems that might be expected in surfaces experience rapid cooling, and
Mach-2.4 cruising operations was se- the aircraft once again enters the altitude
lected for study (Fig. 1). regions which produce significant gust
loads.
The profile can be divided essentially
into three phases, chmb-out, cruise, and M A C H - 2 . 4 THERMAL ENVIRONMENT
descent, each of which imposes different Aerodynamic heating is a function
requirements on the structure. During primarily of Mach number and ambient
182 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
described in reference (1)^ was assumed to tion at Mach 1.0. The gust and maneuver
be applicable. For convenience, the stress spectra are shown in Fig. 4.
corresponding to gust loads was ex- The thermal stresses in the wing
pressed in terms of the calculated \-g panels, caused by transient temperatures
stress for a 190,000-lb, Mach-1.0 condi- (Fig. 2), are primarily the result of
+ 15 000
+ 10 COO
+ 5 000
- 5 000
INNER SKIN
- 10 000
- 15 000
- 20 000
OUTER SKIN
CO - 25 000 UPPER SURFACE
u
« - 30 000
co
< 4- 15 000
s + 10 000
K
U
K
H + 5 000
- 5 000
INNER SKIN
- 10 000
- 15 000
- 20 000
OUTER SKIN
- 25 000 LOWER SURFACE
- 30 000
0 20 40 60 80 100 120 140 160 180 200 220
TIME, MIN
tion. The maneuver stress spectrum was differential temperatures between the
based on the appropriate frequency of outer skin, the internal grid (slug), and
occurrence data given in reference (1). It the inner skin. For the purposes of
is also expressed in terms of the l-g condi- calculating thermal stresses, sandwich
" The boldface numbers in parentheses refer
construction can be considered as an
to the list of references appended to this paper. assembly of elements (Fig. 5). The slugs
184 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
and spar-cap areas are essentially sembly and the effective temperature,
columns stabilized by the substructure and in terms of over-all panel strain and
and the sandwich panel proper. Stress temperature. The inner and outer skins
equations for the elements can be written of the sandwich panels are treated sepa-
in terms of the total strain of the as- rately as plates. Stress equations could
Slug and Skins, 2024- T86- Skin Thickness 0,038 and 0.098 in.
in
&< 60 000
Room Temperature
ri 40 000 - -_ 325 F
Exposure at 325 F , h r
-23-1/2"-
^ ^
\ ~-\ 1^0.75"
SECTION A A
also be written in terms of over-all panel conditions were based upon box-beam
strain and temperature. Strains in the tests and resulted in a more severe stress
skins could then be obtained by finite- condition than would develop in an
difference solutions of the plate equations actual aircraft wing. Calculated thermal
of equilibrium and compatibility. The stresses are shown in Fig. 6 for both
surfaces at the selected wing location.
TEST SETUP Tensile stresses are shown positive and
compressive stresses negative.
T E S T LO.^D B L O C K I
±3 450
±5 940
±3 530
-1-4 620
+7 700
T E S T LO.\D BLOCK II
1 RT ±3 450 4
2 RT ±5 940 2
3 RT ±8 470 45
4 RT ±11 000 12
5 RT ±13 530 3
6 . . RT ±15 379 2
7 ET ±3 530 5
8 ET -f4 620 4
9 ET ±7 060 14
10 ET +7 700 30
11 ET + 10 780 125
12 .... ET + 13 860 38
13 ET + 16 940 13
14 ET + 20 020 4
15 ET + 23 100 2
N O T E S — 1 . R T = room t e m p e r a t u r e , E T =
elevated t e m p e r a t u r e .
2. For test load block I, E T =
F I G . 9.—Test L o a d - T e m p e r a t u r e - T i m e Profiles. 250 F in outside skin.
T e s t Load Block I represents gust a n d ma- 3. For test load block I I , E T =
neuver loads t h a t occur each flight. I t is applied 250 F in outside skin for one half
once each simulated flight. load cycles. E T = 300 F in out-
T e s t L o a d Block I I represents gust and ma- side skin for t h e other half load
neuver loads t h a t occur less frequently. I t is ap- cycles.
plied once each 400th simulated flight.
and fatigue. The ability of the aluminum The maximum permanent strain due
sandwich panels to resist static loads as to creep was predicted to be less than
a function of elevated-temperature ex- 0.1 per cent after 10,000 hr of Mach-2.4
posure is discussed in reference (2). The operations. Creep analyses were based on
compressive strength of the panel con- data obtained from Alcoa Aluminum
figuration chosen for analysis and testing Research Laboratory extrapolated to
(Fig. 7) is shown as a function of a 325 F meet the conditions of the problem.
exposure temperature. Relaxation of thermal stresses due to
. TEST PANEL
u I I I n I j'l I I I J ^ I r 1 } / n n 11 f 1
COLD WATER - 40 F ©
- IRON MAIDEN
TEST BOX
APPLIED
y QUARTZ LAMPS / LOAD
I / I 111 / 11 I / I : 11 > I I ! 11 11 / . I y
•>-t
©
"ir
HOT AIR ©
©
FIG. 10.—Test Cycle of Temperature Profile.
Fatigue is most significant in the lower such creep was predicted to occur early
or tension side of the wing. The fatigue during service for the upper wing panels.
analyses were based upon an approach Subsequent flights should then experi-
which accounted for fatigue strength ence essentially an unchanging pattern.
degradation due to elevated-temperature Such relaxation is expected to be less
exposure and existing fatigue strength than 5000 psi in the outer skin of the
data for aluminum sandwich construc- upper surface panels. No relaxation is
tion. The results of these analyses expected to occur in the lower panel due
indicated that service according to the to the combination of lower stresses and
chosen profile would be trouble-free in temperatures.
excess of 15,000 hr. The Miner-Palmgren
TEST OBJECTIVES
cumulative damage rule was used in
estimating the fatigue effects of the The analyses indicated that the upper
cyclic loads and cyclic thermal stresses. surface would experience: (1) strength
BuNTiN AND LOVE ON ALUMINUM SANDWICH PANEL TEST 187
elevated temperature, (2) creep and The test specimen (Fig. 8) consists of
stress relaxation as a result of elevated two aluminum sandwich panels spliced
temperature, applied loads, and thermal together in exactly the same manner
loads, and (3) some fatigue as a result of that production panels would be joined.
cyclic service loads and cyclic thermal The panels were made by using produc-
loads. The lower surface case, on the tion tooling and methods with two
other hand, appeared to be one primarily 2024-T86 aluminum alloy skins bonded
of fatigue, both at room and elevated to fiberglass honeycomb core and to
temperature. Less strength degradation, 2024-T86 grids or slugs. The slugs
no creep or stress relaxation, and less frame the honeycomb core and provide
severe effects from the cyclic thermal means of sphcing the panels and attach-
loads were predicted. ing them to understructure—in this case
The analysis showed that the highest the test jig. The outer and inner skins
combined stress develops in the upper were clad on one side only and had a
f «^-
1 eA
1 ii'i " * 1
( _ _ _ J , . L ^ , — ^ - ^
( ' yi.r- - ^,
1 ®^
1 !'[[ * J
'• (?r - -
® STRAIN GAGES ON OUTER SKIN
surface panel. In bonded aluminum nominal thickness of 0.038 and 0.098 in.
sandwich panels, the greatest structural respectively. The unclad sides of the
concern is compression and skin buckling skins were bonded to the core and slug.
predictions. Therefore, the upper surface The honeycomb core, made from a glass-
conditions were selected as initial test reinforced plastic using a high-tempera-
conditions. As previously stated, the ture-resistant resin, has a i^-in. cell
objectives of the test were to determine size and weighs 7 lb per cu ft. Greater
if bonded aluminum sandwich construc- detail concerning bonded aluminum
tion could withstand the elevated tem- sandwich construction is given in refer-
peratures and loads associated with the ence (4).
Mach-2.4 profile for a reasonable period
TEST PROCEDURE
of service. To withstand cyclic thermal
loads successfully would mean that no The load-temperature-time profiles
failure would occur, that excessive applied to the test panels are shown in
permanent deformation would not de- Fig. 9. The load profiles, defined in terms
velop, and that the required static of nominal stress at the minimum cross-
strength would remain throughout the section in the specimen, are further
planned service life. described in Table I. The temperature
188 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
profile and test load block I simulate the condition is being simulated, hot air is
environment and flight loads correspond- blown into the test box to accelerate the
ing to one mission. The test time cycle build-up of temperature on the inside
corresponding to one mission was face of the test panel.
shortened to 60 min. The effect of cyclic 4. The quartz lamps and hot air are
thermal loads was of prime concern in turned off and cold air is blown across
this test. By shortening the test cycle the outer surface of the test specimen,
time, a greater number of such thermal thus simulating deceleration.
loads could be achieved within the
planned test span. Test load block II These events are repeated for each
was substituted for I every 400th applica- simulated flight. The required loads are
tion in order to account for the higher applied according to the previously de-
loads that occur periodically. scribed schedules. Each event required
The sequence of events required to in producing the temperature profile
produce the temperature profile in the and the applied loads is automatically
test specimen (Fig. 10) is as follows. controlled and sequenced by program-
1. Cold water is passed into the "iron ming equipment.
maiden" test box and cold air is blown Instrumentation consisted of both
over the test panel until the starting- thermocouples and strain gages (Fig. 11).
point temperature is reached. The thermocouples were used both for
2. Quartz lamps apply heat to the control and as a means of recording
outer surface of the specimen, simulating temperatures in both the skins and
aerodynamic heating. The cold water is slugs throughout the test. Strain gages
gradually released, simulating the con- were used to monitor the applied loads
sumption of fuel in an aircraft wing. in the panel. Paperback gages were used
3. At the point at which a dry-tank and replaced as required. No attempt
BuNTiN AND LOVE ON ALUMINUM SANDWICH P A N E L T E S T 189
ALUMINUM
TEST PANELS
TRACK
ROLLER BOLTS
-^ AIR,
WATER INLET
WATER
OUTLET REINFORCED
SECTION A A BACKING
PLATE
LOAD
PROGRAMMER
MA.STER CONSOLE
was made to install strain gages that lengthwise in the plane of the panel
would remain intact throughout the independently from the "iron maiden"
test. box. As a result, practically all applied
load is carried by the test specimen and
DESCRIPTION OF T E S T SETUP AND very httle by the supporting box. The
CONTROL EQUIPMENT rollers follow fixed ground tracks inside
the box. Slotted holes are machined in
A universal shear frame (Fig. 12)
the box at each bolt location. SiHcone
reacts the applied loads and serves as a
rubber strips, capable of withstanding
base on which to mount equipment; the
high temperatures, were inserted between
loading arm is pivoted on a roller-
the faying surfaces of the panel and the
support to minimize vertical components
box. These strips, applied along each
of applied loads. The load application
boltline, form a seal and reduce the fric-
and reaction links are located on the
tional load as panel deformation occurs.
vertical centerline of the pivoted loading
arm. The panel ends are centered on the The activities of each piece of test
arm and base and encased in half- equipment, shown in Fig. 14, are co-
round bronze bars. A dry Molykote ordinated and sequenced properly by a
powder hand-rubbed into the bronze master console. External heaters con-
bars reduces edge fixity. The compres- sisting of radiant quartz tubes, spaced
sion loads are therefore evenly distrib- approximately 2 in. apart and backed by
uted over the ends and enter the speci- a flat aluminum reflector, are hung 3 in.
men parallel to the plane of the panel. above the test specimen. The heater
rectangle overlaps the pane! 3 in. on all
The magnitude of the load is moni-
sides. Power to the heaters is provided
tored with three load sensors as shown
in Fig. 12. The upper sensor measures by an Ignitron unit controlled by an
the steady-state load which remains arbitrary function generator. The genera-
relatively constant throughout the test. tor produces its demand signal by com-
The sensor in series with the hydraulic paring a thermocouple signal from the
ram not only measures the compressive middle of the test specimen to a pre-
load in its link, but also feeds a signal selected heating curve mounted inside
to the control equipment which is the generator. The system has proved to
plotted on an electronic recorder. This be very reliable in satisfying the auto-
plot is a visual display available for mated temperature cycling requirements.
every load cycle, whereas the lower Temperature recorders monitor con-
sensor, beneath the panel, is used for an tinuously three thermocouples on the
occasional check for correct load applica- panel; they also provide a means of
tion. incorporating safety circuits to shut
The test panel is mounted on a steel, down the test in the event of overheating.
air-tight box, commonly referred to as an Hot air, blown inside the "iron
"iron maiden" (see Fig. 13). I t is bolted maiden" to accelerate temperature
to the box every 2^ in. along each edge buildup on the inner face of the test
and down the longitudinal and trans- panel, is provided by a 12-kw calrod-
verse centerlines. The attaching bolts heater. Control of the air is accomp-
are threaded on one end and attached to hshed with by-pass valves that direct
2-in. steel rollers on the opposite end. the heated air to either the test box or
These rollers permit the panel to deform the atmosphere. The by-pass valves are
BuNTiN AND LOVE ON ALUMINUM SANDWICH P A N E L T E S T 191
controlled by the master console box tion is required in applying test load
through motor-driven cam relays. The block I I previously described.
air temperature control is accomplished
TEST RESULTS AND DISCUSSION
with a preset thermostat built into the
heater. As of July 1, 1962, 2750 test cycles
The cold air and cold water are sup- simulating 2750 missions (approximately
plied by a 5-ton air cooler which is also 10,000 flight hr) have been apphed to
controlled with by-pass valves operated the test specimen. Panel deformation
by cam relays in the master console. has been negligible, and no failure has
The air is dried in existing dehumidifiers occurred. The panel was removed from
before entering the heat exchangers in the test setup after 1500 test cycles and
carefully inspected for failure and as to
the cooler; this prevents ice from forming
flatness. After 2000 test cycles, new
in the coils in the cooling chamber.
strain gages were installed and load
Sufficient antifreeze material is added distributions were checked. For a given
to the cooling water to prevent freezing applied load, the distribution across the
in the cooler's water chamber. The exit panel was found to be essentially the
temperature of both the air and the water same as in the beginning of the test.
is approximately 40 F during the demand
period. SUMMARY
The load programmer controls the
hydraulic ram to apply the spectrum The test has demonstrated the capa-
bility of bonded aluminum sandwich
loading. This programmer supplies step-
construction to withstand cychc thermal
input signals to amplifiers controlling a
loads, cyclic applied loads, and elevated
set of servo valves. The rate of loading is
temperatures corresponding to long
determined by a function generator built
service hfe in a Mach-2.4 environment.
into the electrical circuit. The amplitude
The test apparatus and equipment have
of each load, as well as the time the load proven to be more reliable than antici-
remains on the specimen, is preset in the pated. The test procedures are quite
programmer. Once the programmer general and can be extended to apply for
receives the "go" signal from the master flight conditions beyond Mach 2.4.
console, it completes the desired load Time is perhaps the biggest factor in
spectrum without further external signal. running tests of this sort. Minimum ex-
The safety circuit to prevent overloading pense demands a system that requires
and the apphcation of an excessive minimum manpower support, and there-
number of loads is incorporated in the fore one that is automatically controlled
load recorder. for continuous operation. A large expense
The equipment that has been described problem occurs in dissipating the heat
has functioned as a completely auto- energy as required. Refrigeration equip-
mated system requiring minimum super- ment and materials required quickly
vision with good reliability. All of the become significant economically. In
equipment can be operated manually conclusion, it is believed that structural
to apply arbitrary loads, load-times, integrity testing of full-scale structure to
heating requirements, or any combina- include cyclic heating in addition to
tion of these parameters. Manual opera- cyclic loads is both desirable and feasible.
192 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
REFERENCES
(1) "Airplane Strength and Rigidity Reliability (3) Aluminum Research Laboratories, unpub-
Requirements, Repeated Loads, and Fa- lished creep data, Oct., 1959.
tigue," MIL-A-8866 {ASG), May 18, 1960. (4) L. M. Smith and C. W. Rogers, "Bonded
(2) W. D. Buntin, "Some Materials and Struc- Bomber—B-58," SAE Paper Presented at
tures Problems Associated with Designing
the National Aeronautics and Space Engi-
B-58 Type Structure for Supersonic Cruise
neering and Manufacturing Meeting, Los
Capability," Paper Presented at the Golden
Gate Metals Conference, Am. Soc. Metals Angeles, Calif., Oct., 1961.
Feb., 1962.
STP338-EB/Sep. 1963
E S T I M A T I O N OF FATIGUE P E R F O R M A N C E OF A I R C R A F T
STRUCTURES
BY J. SCHIJVE 1
SYNOPSIS
AIRCRAFT A
AIRCRAFT F
Isec
The leading question here is: The struc- illustrated by some test results recently
ture and its loads being given, what is its obtained at the National Aero- and
fatigue performance? Nowadays the Astronautical Research Institute (NLR),
fatigue performance does not simply Amsterdam, and by similar data derived
mean the fatigue life until failure or the from the literature. Some results of the
fatigue life until visible cracking; it also NLR were obtained under contracts
includes crack propagation data and the with the Netherlands Aircraft Develop-
residual strength in the cracked condi- ment Board (NIV) and with the Aero-
tion. Crack propagation is discussed to nautical Systems Division, United States
some extent in this paper, but the Air Force, European Office, Brussels.
strength of damaged structures will not Permission for publication is acknowl-
be considered. In fact, the latter problem edged here.
ScHijvE ON ESTIMATION OF FATIGUE PERFORMANCE 195
BENDING MOMENT
tg LEVEL, 2 0 0 KNOTS
£ MOMENT OF TOUCH DOWN -TAXIING LOADS
-*-TIME
V LEVEL AT STAND STILL
OF THE AIRCRAFT
-H
FIG. 2.—Record of the Bending Moment at the Root of the Wing During a Rough Landing
of Aircraft F.
equivalent if they yield the same fatigue giving a complete description of a service
lives, the same crack rates, and if they loading further complicate the empirical
indicate the same components of a struc- comparison with cyclic loading.
ture to be fatigue-critical. Tests to study Although cyclic loads can be defined
these points have just been started at the very easily they still can be applied in
NLR and will be discussed later. A numerous ways in a fatigue test. In the
quantitative impression on the equiva- conventional fatigue test the mean load
lence can be obtained at this moment by as well as the load amplitude are con-
comparative testing only. stant. Some equivalence with service
It is not so easy to give a complete loading has been claimed (14), but it is
definition of a random load-time history. thought that program-fatigue loading as
Usually the distribution function of the shown in Fig. 3, as proposed by Gassner
peak loads is presented. Although this (11), is more likely to be equivalent to
gives some information on what has oc- random loading. The loading in a pro-
1 PERIOD
FIG. 3.—Schematic Load Sequence in a Program-Fatigue Test with Increasing and Decreasing
Load Amphtudes in Each Period.
600
NUMBER OF EXCEEDINGS
FIG. 4,—Comparison of the Load Statistics of a Gust Record Obtained by Various Counting
Methods.
loads have been developed. All these range-mean count method gave results
methods but one involve counting of that are essentially different from the re-
certain occurrences: (1) peak loads, or sults of the other methods. It was shown
maxima and minima of the load-time that these range methods did not give
history, and (2) load variations, or the realistic information. Counting according
differences between maxima and minima. to the Fatiguemeter gave almost the
A further differentiation stems from two same results as the mean-crossing peak
questions: (1) Should all peak loads or count method (VGH-recorder), while the
load variations be counted or may a part range-pair count method (strain range
of them be neglected? (2) In which way counter) gave counts that were about 10
are counts being made? A survey of to 20 per cent lower than both previous
counting methods was recently given by methods, except at low load amplitudes
the author (31). Methods discussed were: (for an example, see Fig. 4). Reference (31)
(1) peak count method, (2) mean-cross- presents a more elaborate discussion as
198 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
well as a full description of all counting cording to the range-pair count method.
methods. The main conclusion reached Power spectral analysis has advantages
there is that probably the range-pair for studying the dynamic behavior of air-
count method must be preferred from craft, but its possibilities for the evalua-
the fatigue point of view and that it is tion of fatigue properties seem still to be
fortunate that both best-known methods limited at this moment.
for load statistics, by the Fatiguemeter
and the VGH-recorder, offer almost the ACCUMULATION OF FATIGUE DAMAGE
same results. This conclusion obtained UNDER VARIABLE-AMPLITUDE
for gust loadings will probably also hold LOADING
for maneuver loads. An elaboration of A discussion of the accumulation of
this point is worthwhile. fatigue damage requires a definition of
The fortunate coincidence of the re- fatigue damage, although a completely
sults of three important counting meth- satisfactory definition cannot be given
ods, as illustrated in Fig. 4, implies that at this moment. On the basis of micro-
there is little to choose between them for scopical evidence it has become generally
making life estimates. The usefulness of accepted that microcracks are initiated
these counting methods can be shown very early in the fatigue life. Shanley (34)
only by comparative testing, as by com- assumed that the fatigue damage was
paring the fatigue results of random-load fully determined by the crack length;
tests with those obtained by program- and similar ideas have been put forward
fatigue testing or calculations. For the by Corten and Dolan (6) and Valluri (36).
program-fatigue test the program of loads In reference (27) an analysis of fatigue
is based on such data as those shown in damage and its accumulation has been
Fig. 4. For life calculations the same presented. Some conclusions were:
data, relevant S-N curves, and a cumula- 1. The fatigue damage cannot be ac-
tive damage rule are required. Such curately indicated by a single damage
comparative studies give an empirical parameter, for instance by the crack
answer to the question whether a reduc- length. More parameters are necessary
tion of a random load-time history to a to give a complete description of the
distribution function of load increments fatigue damage. Two additional parame-
can be useful for practical purposes. ters are the residual stress around the tip
One method of analyzing a random of the crack and the strain hardening in
trace, other than a counting method, is the same region.
by using a mathematical evaluation. It is 2. The Palmgren-Miner rule
assumed that the loading behaves as a
random Gaussian phenomenon (3) which
is fully described by a power spectrum ^ N
function. From this function the distri- is not compatible with more than one in-
bution function of all peak values can be dependent damage parameter. So its
derived. Some practical difficulties in validity must be discounted for physical
applying the theory are described in ref- reasons.
erence (31). Further, the theory does not In reference (26) data on variable-
yet allow the calculation of either the amplitude loading of light alloy speci-
distribution function of peak values mens available in 1955 were summarized.
counted according to the mean-crossing A conclusion of the review was that un-
peak count method, or the Fatiguemeter, notched rotating-beam specimens showed
or the distribution function of ranges ac- a cumulative-damage behavior that was
ScHijvE ON ESTIMATION OF FATIGUE PERFORMANCE 199
1 X — 1 7075 plane
bending
random n o i s e Sm = 0 1.2 to 8 F r a l i c h (9)
2024 1.3 to 1 .9
H ^ 7075
axial maneuver spectrum Smin = 0
1.1 to 1 .9
W a l l g r e n and
Svensson (41)
c o n s t a n t S^ and
Riveted joint 2024 axial ground-to-alr- Sn, > 0 --0.5 B a r r o i s (1)
cycle
gust spectrum
2024 . axial gus t s p e c t r u m and S^> 0 G a s s n e r and
1 : 1 ground-to-air- Horstmann (15)
cycle
Commando wing 2024 wing random g n s t spectrum S^ > 0 3.5 Whaley (37)
bending
" The minimum and maximum values of averages of test series are presented.
200 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
STRESS
TEST RELATION WITH OTHER TEST
SERIES SERIES
LOAD HISTORY
- S _ = 20 percent Su
SIMILAR TO SERIES 2 4 ,
ONE HIGH NEGATIVE AND
POSITIVE LOAD AT THE
END OF EACH PERIOD.
FIG. 5.—Effect of High Peak Loads on the Fatigue Life of a 2024-T3 Riveted Lap Joint Under
Program-Fatigue Loading.
reputation as being unsafe for life pre- testing variables in accordance with
dictions. service conditions. It is felt that neither
However, tests on axially loaded, the academic approach nor the practical
notched specimens with a positive mean approach should be neglected. At the
stress very often yield conservative ( > 1) NLR the main effort was directed to-
wards the practical side. However, an at-
J2 -^ values. The latter type of testing is tempt was made to analyze theoretically
A the test results and to obtain a quahta-
usually a more realistic representation of tive understanding of the accumulation
the circumstances that cause fatigue of fatigue damage.
troubles in service. Apparently there is a
In reference (29) the results of program-
large difference between the academic ap-
proach and the practical approach to the fatigue tests on riveted lap joints of
problem of cumulative damage. A test 2024-T3 and 7075-T6 were presented.
ScHijvE ON ESTIMATION OF FATIGUE PERFORMANCE 201
Data from all available similar investiga- will induce compressive residual stresses
tions were compiled and summarized that are favorable for preventing crack
(Table I). The conclusion to be drawn growth; a negative peak load will induce
from the table is that on the average tensile residual stresses that will acceler-
ate crack growth. In applying two peak
values of ^ — obtained for notched light loads of opposite sign the latter appar-
alloy specimens loaded by a gust or a ently has the dominating effect. Similar
maneuver spectrum at a positive mean tests with positive peak loads only or
stress are larger than unity. Therefore, negative peak loads only confirmed the
it is incorrect to say that the Palmgren- above reasoning.
Miner rule is an unconservative design Most of the deviations from the
rule for aircraft structures. On the other Palmgren-Miner rule could be explained
hand, Table I also shows that the rule is qualitatively (see reference (29)) by the
not very accurate. Before labeling the effect of residual stresses on the growth
rule as inaccurate, what is meant by ac- of the fatigue crack, irrespective of
curate has to be defined. In reference whether this crack was an invisible
(29) it was said that a life prediction which microcrack or a macrocrack. Compres-
does not err more than by a factor of 2 sive residual stresses, which are favorable,
are induced by tensile loads; therefore a
(0.5 < J^ ^ < 2) may be regarded as tensile mean load allows a more easy
a good estimate from the technical point creation of favorable residual stresses.
of view. Therefore, the Palmgren-Miner
This is one reason why J^ ~ values are
rule may be considered as a useful tool
for design purposes, and this aspect will generally over 1 for program-fatigue
be further discussed later in this paper. tests on notched light alloy specimens
Deviations from the Palmgren-Miner with a positive mean stress. Another
rule were studied systematically (29); it reason is that a residual stress field can
was concluded that in addition to the be formed only if there is a nonhomo-
amount of cracking a second parameter geneous stress distribution (stress concen-
of great importance for the accumulation tration). Such a distribution is present
of fatigue damage was the residual stress. in notched specimens. Both arguments
Negative (compressive) residual stresses
are compatible with X! T7 values below
decelerate the growth of fatigue cracks,
and positive (tensile) residual stresses unity frequently found for unnotched
have an accelerating effect. This effect is specimens loaded at zero mean stress.
illustrated in Fig. 5 (29). A riveted lap The well-known detrimental effect of
joint was loaded in a program-fatigue ground-to-air cycles can be interpreted
test with four or five different load ampli- in two ways: (1) These loads periodically
tudes in increasing order representing a destroy any favorable residual stress
gust spectrum. The result of adding pe- field present; and (2) they may even
riodically a positive peak load followed form an unfavorable residual stress field.
by a negative peak load (test series 27) It has been argued that the Palmgren-
was a considerable reduction of the fa- Miner rule need not necessarily lead to
tigue life (compare with test series 21). unconservative life estimates for a gust
Adding the same peak loads in the re- load spectrum combined with ground-to-
verse order results in a large increase of air cycles (29).
the fatigue life (compare test series 28 Recently tests to study the macrocrack
with test series 24). A positive peak load propagation in sheet specimens under
202 SYMPOSIUM ON FATIGUE or AIRCRAFT STRUCTURES
i CENTRAL NOTCH
^- 2
,\^'^!^~\ CRACK
NTIP RADIUS
o
©' 0.4
o
SCRIBE LINE MARKINGS:
SPACING,1mm
t
160 DIMENSIONS IN mm
1 in.= 25 4 m m )
FIG. 6.—Sheet Specimen Used tor Crack-Propagation Studies with Variable-AmpHtude Loading,
suits, presented in references (32) and higher load amplitude will increase the
(33), were in good agreement with those amount of crack extension at the lower
of Hudson and Hardrath (18). The results amplitudes by making available a larger
fully confirm the importance of residual number of nuclei than the lower ampli-
stresses around the tip of the crack. A tudes could have created themselves.
dramatic illustration is given in Fig. 7, Consequently the interaction of high
showing the effect of three intermittent load amplitudes on low amplitudes would
positive peak loads on crack propagation always be unfavorable. This model may
in a constant-amphtude test. After the be applicable to unnotched rotating
application of the peak loads the crack beam specimens, but it is not considered
almost refuses to grow for a considerable to be an appropriate representation of
time, because it has to break through a the damage accumulation in notched
barrier of a residual stress field. Crack specimens loaded at a positive mean
propagation in the program-fatigue tests stress. For the latter case a favorable
with three, four, and i]ve different stress interaction is certainly possible due to
amplitudes in increasing and decreasing the effect of residual stresses. Freuden-
order was about two to four times slower thal (10) relates the interaction to the
than predicted by the Palmgren-Miner density of slip bands. The density is in-
rule; thus the rule seems to be conserva- creased by higher stress amplitudes; this
tive for macrocrack propagation, al- has an unfavorable effect on the damage
though not accurate. rate at lower stress amplitudes. This con-
Among all published cumulative-dam- cept shows some similarity with the
age theories only the simple Palmgren- theory of Corten and Dolan. A theory re-
Miner rule has found wide application. cently published by Valluri (36), based on
However, this is not only due to its physical arguments, is aimed at the ex-
simplicity; the main argument is that planation of a large number of fatigue
other theories have not yet succeeded in aspects. One argument is that of the
gaining more confidence with respect to variable-amplitude loading problem. The
their reliability. Some comments on the theory includes neither interaction ef-
most important theories will be made. fects of different stress levels in program-
It was shown before that the residual fatigue tests nor the effect of residual
stress is an important second damage stresses. Analytical evaluations of the
parameter in addition to the micro- theories of Corten and Dolan, Freuden-
crack length. Since the residual stress thal, and Valluri will not be presented
will be changed by all load cycles of a here. The formulas contain some con-
sufficiently large magnitude, there will be stants which have to be derived from
an interaction between load cycles of dif- program-fatigue tests or constant-ampli-
ferent magnitudes. Such an interaction is tude tests, unlike the Palmgren-Miner
excluded by the Palmgren-Miner rule rule. The formulas therefore allow a bet-
and also by the theory of Shanley (34). ter adaptation to test results at the price
Theories of Corten and Dolan (6) and of somewhat more complication. Their
Freudenthal (10) involve an interaction, validity still has to be shown by tests for
although not on the basis of residual all those circumstances for which they
stress. According to Corten and Dolan have to be applied. The theories have
the interaction is a consequence of the
been checked for unnotched rotating
number of fatigue crack nuclei, which is
beam and wire specimens and did yield
larger for higher stress amplitudes. The
useful results for that purpose. Their
nuclei are supposed to be able to grow at
applicability for program-fatigue tests on
all amplitudes. This means that the
204 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
GROUND-TO-AIR CYCLE
GROUND LOAD
1 SIMULATED FLIGHT
-•-TIME
So
\ 1.S-N CURVE ° DAMAGE DISTRIBUTION
X ^ ^ LEVEL OF MAXIMUM
\H \N ^ / ^ . ^ ^ ^ DAMAGE Som
f-
_^
NmJX
— ^ N (FATIGUE LIFE) —^^ / N ("DAMAGE \
dSo/ \, INTENSITY/
—»-H (NUMBER OF EXCEEDINGS)
TOTAL DAMAGE . T^ ^ /n)<ISa
FIG. 9.—Damage Distribution Curve and Reduction of a Load Spectrum to One Load Level
some danger of misjudging the fatigue quences of the second and the third
properties of several components. (2) If method do not outweigh the inferior rep-
components are tested separately there resentation of the load-time history. In
will be some uncertainty about the exact fact, the only essential difficulty involved
simulation of the load transmission into in the first method is to obtain a repre-
the component. (3) It is more realistic sentative load-time history. However
and informative to obtain crack propaga- this problem will not be discussed in de-
tion data from a full-scale test than from tail here. It is thought that with the
component testing. present knowledge of loads on aircraft
An exact simulation of service condi- and aircraft response it should be possi-
tions is either impossible or impractica- ble to ascertain a good record. This result
ble. It is not feasible to simulate such could be improved by recording strains
factors as the occurrence of rest periods during prototype flying. Such strain
and the rate of loading which, fortunately measurements are also advisable to learn
do not have a large effect on the fatigue whether the loads in flight correspond
life (29). With respect to the crack propa- with their calculated values.
gation rate, a small systematic effect of A simplified flight-simulation accord-
the load frequency has been found (30) and ing to Fig. 8 can no longer be recom-
should be kept in mind if the loading rate mended, since, at the present time, the
in service and in the test are largely dif- simplification is not excused by the in-
ferent. Not very much is known about creased experimental effort for a com-
the effect of the environmental condi- plete simulation of the service loads. For
tions. Most fatigue failures start at rivet program-fatigue tests a good deal of ap-
holes or bolt holes. The environment in plication is still foreseen because: (1) such
such holes probably is not very much dif- tests can be executed on components at a
ferent in the laboratory and in service. much larger speed with commerciaUy
However, once cracks are growing and available fatigue machines; and (2) for
are accessible to the open air one may rough life estimates and for comparative
expect some effect of corrosion or water testing in design studies the method is
vapor. This is a problem that should be considered acceptable.
studied. The NLR has recently started In view of the latter argument it is de-
some comparative testing on this topic. sirable to improve the knowledge of the
It was pointed out previously that a similarity between service loading and
service load-time history will be a mix- program-fatigue loading. Studies of this
ture of several types of loadings varying similarity might well extend the applica-
from random loads to systematic loads. bility of program-fatigue testing. Some
With the present-day experimental facil- test programs on this problem are re-
ities for large structures, an exact viewed in the following paragraph. The
simulation of such a load-time history load sequence in a program-fatigue test
can be faithfully accomplished. For a can be arranged in many different ways.
large full-scale structure for which a sin- Some thoughts on this topic are pre-
gle test has to give the most accurate sented later in this paper.
information for the operation of the air-
craft the following three alternative test Comparison of Random-Load Tests and
methods can be considered: (1) simula- Program-Fatigue Tests:
tion of the service load-time history, (2) The number of investigations to study
program-fatigue test (Fig. 3), and (3) the similarity between random loads and
simplified flight-simulation test (Fig. 8). program-fatigue loading is rather limited.
It can be said that the simpler load se- The first study was made some ten years
208 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
ago in Australia (23) on Mustang wings. An extensive test program was con-
Random tests and program-fatigue tests ducted by Lockheed (20) on 7075-T6
showed the same weak spots. The com- sheet specimens with a central elhptical
parative results for the fatigue lives were notch. Kt values of 4.0 and 7.0 were
not so conclusive. The random load and used. Load-time histories studied were
the program-fatigue load did not fully based on gust loads (constant mean load
satisfy the conditions considered here. = l-g load), maneuver loads (constant
In the random tests each gust load re- minimum load = l-g load), taxiing loads
turned immediately to the level of the (random vibrations superimposed on
mean load (l-g level), which is a restric- ground load), ground-to-air cycles (cy-
tion on the randomness. In the program- cles between l-g load and ground load),
fatigue tests only three different values and combinations of these loads. The
of the load amplitude were used; this is fatigue machine, designed by Lock-
considered to be too small for a correct heed, was fed by a magnetic tape con-
respresentation of a gust spectrum. taining the load sequence as a magnetic
A second and more recent study was signal. The programmed sequence was
conducted by Kowalewski (19). Small derived from the random trace by em-
light alloy specimens were loaded in ploying the mean-crossing-peak-count
plane bending. The random load was ob- method. Load cycles started with the
tained as random noise of a limited fre- negative part followed by the positive
quency band (about 10 to 35 cps). part. Load cycles in one period were ap-
Kowalewski studied four types of ran- plied in a sequence of increasing ampli-
dom loading with different power spectra. tudes only. The number of variables
He derived the program loading from the studied in these investigations is too large
random load sequence in two ways. In to allow a quantitative summary of the
the first program the distribution of peak results here. The most salient conclusions
loads was the same as for the random were:
load. In the second program the distribu- 1. The ratio of the fatigue life under
tion of ranges (the differences between random loading and the fatigue life un-
successive maxima and minima, P.+i — der program loading was found to vary
Pi) was the same as for the random load. from values noticeably lower than unity
For the first type of program loading the to values noticeably higher than unity,
life was about half the life under random depending on the type of loading being
loading for all four types of random load- considered. The best agreement was ob-
ing. For the second type of program tained for a severe gust spectrum.
loading no systematic correlation was 2. The effect of the ground-to-air cycle
found. In reference (31) it was stated that was much larger than expected on .he
a representation of a random load by its basis of the Palmgren-Miner rule.
ranges cannot be expected to be a good 3. A representation of the ground-to-
representation. Briefly, the reason is that air cycles in a program-fatigue test by
a large load variation from Pmin to Pmax , batches of load cycles between the mean
which is interrupted by a small load load in flight and the mean load of the
variation, is split up in three small load taxiing loads on the ground underesti-
ranges, thus ignoring the large variation mates the severity of the ground-to-air
from Pmin to Pmax • Kowalewskl's results cycle.
have a Umited applicability, since the
specimen was only mildly notched The first conclusion indicates that a
(Kt = 1.77) and the loading was plane unique correlation between random load
bending with Sm = 0. and program load was not found. The
ScHijvE ON ESTIMATION OF FATIGUE PERFORMANCE 209
conclusion was arrived at for one method (3) to compare the crack propagation
of reducing the random loads to cyclic data, and (4) to study the damaging ef-
loads and for a certain sequence of the fect of the ground-to-air cycle.
cyclic loads in the program-fatigue test. The random loading was derived from
Therefore it cannot be generalized to a strain-gage record of the bending
other reduction methods and other types moment at the root of the wing of a large
of program-fatigue tests. aircraft flying in turbulent air. The
Recently the NLR started a program principle of the hydraulic testing machine
on full-scale structures using tension- with an electronic servo-apparatus,
skin specimens from a predmodification designed by the NLR (42), is shown in
type of the Fokker F-27 wing. They Fig. 10. The random load sequence to be
consist of the stiffened skin between front applied to the specimen is punched in a
and rear spar of the center section of the tape as a series of code numbers (binary
LOAD RECORDER
FIG. 10.—Principle of the NLR Random-Load Machine CARLA (Code-Actuated Random Load
Apparatus).
wing. The material is 7075-T6 (material digit numbers) representing all succes-
of the post-modification type is 2024-T3). sive maxima and minima. The oil supply
The length of the specimen is 8.3 m (27 of the pump to the hydraulic jack is
ft). About 50 per cent of the tension ma- monitored by the electronic apparatus
terial is in the skin; spar caps are rela- which receives its input data from the
tively light. Some skin splices, cutouts, tape reader and from the load cell via a
reinforcements, and other discontinuities feedback line. The tape reader, the load
are incorporated. The specimen is con- cell, the recorder, and the pump were
sidered to be representative of a modern commercially available. The electronic
aircraft structure. Three types of tests apparatus with the code name CARLA
planned are: (1) random-load tests, (2) (Code Actuated Random Load Appa-
program-fatigue tests, and (3) random- ratus) was designed and built by the
load tests with ground-to-air cycles. The NLR. Punched tapes for the random-
aim of the comparative tests is (1) to see load tests are made from the strain rec-
whether the same components are fatigue ord by a semiautomatic data reading
critical, (2) to compare the fatigue lives. apparatus and an electronic computer.
210 SYMPOSIUM ON FATIGUE or AIRCRAFT STRUCTURES
These tapes could also be used for a method has to be preferred from a fatigue
statistical evaluation of the loadings on point of view. Fortunately the results of
the record in accordance with several the mean-crossing-peak-count method
counting methods (31), the evaluation are almost the same. The Fatiguemeter
being performed by the electronic com- and the VGH-recorder give results that
puter. The latter also generates the tapes are almost or exactly in accordance with
for the program loading. A testing ma- the latter counting method. Therefore it
chine operating according to more or less is advised for practical reasons to adopt,
similar principles was developed in the mean-crossing-peak-count method.
Switzerland (4). Further studies on this topic are desir-
Until now only two tests have been able.
completed; therefore it is not yet useful Whether a load cycle should start
to present quantitative data. The speci- with its maximum or its minimum is im-
men is continuously inspected for cracks portant only for the more infrequent
without stopping the testing machine; loads with large amplitudes. The effect,
cracks are far more easily found if the discussed previously, was clearly shown
test is running due to "breathing" of the in Fig. 5, although this figure may prob-
cracks. Very minute cracks have been ably give a somewhat exaggerated im-
found. Quantitative results of the tests pression. The conclusion to be deduced
will be published in due course. from this figure and the earlier discussion
In summary, the knowledge of the on the effect of residual stresses is that a
similarity between random loads and sequence starting with the maximum,
program-fatigue loading is still far from followed by the minimum, will give con-
complete. servative results, and the reversed se-
quence may give unconservative results.
Planning of a Program-Fatigue Test: Although the conservatism of the results
In reducing a service load-lime history is perhaps not a very rational argument,
to a load sequence for a program-fatigue it is recommended to apply load cycles
test there are many questions for which starting with the maximum followed by
no exact answer exists, although most the minimum.
of them have already been touched upon Figure 3 shows a schematic example of
in this paper. Some comments will be a single load sequence in a program-fa-
made on the following questions: tigue test. Other sequences being studied
1. How should the irregular load-time (29) are: increasing load amplitudes only,
history be reduced to a population of decreasing load amplitudes only, and a
complete load cycles? randomized sequence of load amplitudes.
2. Each load cycle comprises a maxi- It was said earlier that damage accum-
mum and a minimum. Which of these lation was largely governed by the
two should be applied iirst? amount of microcracking and the resid-
3. In which sequence will the load ual-stress field around the tip of the
cycles with different amplitudes be ap- cracks. During a random loading there
plied? will be a certain random variation of the
residual stress, which can be considered
The reduction of the service load- as a random variable with an average
time history to cyclic loads has been value depending on the mean value of the
discussed previously. The conclusion random load. It is now suggested that a
reached was that there is neither a ra- load sequence in a program-fatigue test
tional nor an empirical answer for this should be adopted for which it can be
problem. Probably the range-pair count expected that the average residual stress
ScHijVE ON ESTIMATION or FATIGUE PERFORMANCE 211
has about the same value as under the Finally, some comments will be made
corresponding random load. Although on the application of ground-to-air cycles
this criterion cannot be evaluated quan- in a program-fatigue test with a gust-
titatively it is thought that a sequence spectrum loading. In view of the damag-
with increasing amplitudes only or with ing effect of the ground-to-air cycles they
decreasing amphtudes only will satisfy cannot be omitted. The best solution
the criterion in an inferior way as com- would be to insert one ground-to-air
pared with the sequence indicated in Fig. cycle at regular periods between the gust
3. A randomized sequence may give some load cycles. This would involve a certain
further improvement, but it is thought type of flight simulation. Experimentally
that this complication of the test is prob- this is not a simple solution because the
ably of fictitious importance. Therefore number of ground-to-air cycles in one
it is advised to perform program-fatigue period would be large. A simpler applica-
tests with a sequence as proposed origi- tion is possible if the ground-to-air cycle
nally by Gassner (11) (Fig. 3). is considered as being a negative gust
Other variables to be selected for plan- load. The gust spectrum will become
ning a program-fatigue test are: (1) the noticeably asymmetric then, but this is
lowest amplitude to be included in the not necessarily an objection for the ap-
program, (2) the highest amplitude to be plication in a program-fatigue test as
included in the program, (3) the number indicated by Payne (23). It implies that
of load steps (five in Fig. 3), and (4) the the minimum load of the ground-to-air
size of or the number of cycles in one cycle is combined with the positive gust
period. Comments on the lowest ampli- load which is anticipated on the average
tude to be included in the program have once per flight. A similar suggestion was
been made previously. With respect to made by Barrels (2). It is then possible
the highest amplitude to be included, a to apply ground-to-air cycles in batches
^alue which is exceeded about 10 times of more than one cj^cle. As has been said
in the anticipated fatigue life was some- before, a combination of the minimum
what arbitrarily recommended in refer- load of the ground-to-air cycle with the
ence (28). However, it is felt that cycles mean load of the gust loading (1-g level)
with larger amplitudes are too rare to is considered to be an underestimation of
contribute noticeably to the fatigue the real damage of the ground-to-air
damage. Moreover, if they occur in serv- cycle.
ice their effect will probably be favorable; The conclusion is that no ideal method
but since it is not certain that each air- can be given for planning a program-
craft will meet these rarely occurring fatigue test; therefore no guarantee of a
loads their omission in the program-fa- high precision of the data can be obtained
tigue test seems to be justified. with such tests for the fatigue perform-
The number of load steps should be as ance under service loading. However, the
large as possible. A minimum of six or incorporation of present experience and
eight steps is probably acceptable. The understanding of the fatigue damage ac-
size of the period should not be too large; cumulation has been attempted.
a maximum of one-tenth of the antici-
CONCLUSIONS
pated life is probably still acceptable,
although a smaller size is preferred. Both
1. The values of X! ^ '"^^^ generally
these recommendations cannot be justi-
fied on a rational basis and are to some larger than 1 for notched light alloy
extent made intuitively, based on the specimens tested under program-fatigue
experience discussed in reference (29), loading with a positive mean stress.
212 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
2. The Palmgren-Miner rule gives con- liable for judging whether a certain type
servative estimates for the rate of macro- of service load will contribute substan-
crack propagation in sheet material un- tially to the damage induced by other
der variable-amplitude loading. types of loadings.
3. For notched Ught alloy specimens, 6. Program-fatigue testing should be
the accumulation of fatigue damage is strongly preferred over constant-ampli-
considerably affected by residual stresses. tude testing in comparative design
The residual stress at the tip of the studies of components. Recommenda-
microcrack is considered to be a sec- tions are made for a careful planning of
ond damage parameter of great impor- the load sequence in a program-fatigue
tance; the first parameter, the amount of test.
cracking. 7. Full-scale testing is strongly recom-
4. The Palmgren-Miner rule can be mended to assess the fatigue performance
considered a useful tool for obtain- of an aircraft structure. The loading to
ing rough life estimates in the early de- be applied should preferably be a simula-
sign stage of an aircraft structure. The tion of the anticipated load-time history
major problem is not the inaccuracy of in service.
this rule but the availability of relevant 8. The simulation of random loads by
S-N data of sufficient accuracy. cyclic loads is not yet fully understood
5. The Palmgren-Miner rule is unre- and should be studied further.
REFERENCES
(1) W. Barrois, "Physical Interpretation of craft," RAE-Technical Note, Structures
Metal Fatigue," Minutes of the 5th 235, March, 1958.
Conference, of the International Com- (8) W. A. P. Fisher, "Programme-Fatigue Tests
mittee on Aeronautical Fatigue, Brussels, on Notched Bars to a Gust Load Spec-
October, 1957. trum," RAE-Technical Note, Structures
(2) W. Barrois, "The Philosophy of Fatigue 236, March, 1958.
Tests on Large Dimension Aircraft (9) R. W. Fralich, "Experimental Investigation
Structures," Proceedings, Symposium on of Effects of Random Loading on the
FuU-Scale Fatigue Testing of Aircraft Fatigue Life of Notched Cantilever-
Structures, Amsterdam, June, 1959, p. Beam Specimens of 7075-T6 Aluminum
239. Alloy," NASA Memorandum 4-12-59 L,
(3) J. S. Bendat, Principles and Applications June, 1959.
oj Random Noise Theory, John Wiley and (10) A. M. Freudenthal, "Physical and Sta-
Sons, New York, N. Y., 1958. tistical Aspects of Cumulative Damage,"
(4) J. Branger, "A Review of Work in Switzer- Colloquium on Fatigue, Stockholm, May
land on Aircraft Fatigue During the 1955, edited by W. Weibull and F. K. G.
Period March 1959-April 1961," EUge- Odquist, Springer Verlag, BerUn, p. 53.
nossisches Flugzeugwerk Emmen, Report (11) E. Gassner, "Festigkeits-Versuche mit
S-129, May, 1961 (restricted). wiederholter Beanspruchung im Flug-
(5) R. A. Carl and T. J. Wegeng, "Investiga- zeugbau," Luftwissen, Vol. 6, p. 61,
tions Concerning the Fatigue of Aircraft 1939.
Structures," Proceedings, Am. Soc. Test- (12) E. Gassner, "Auswirkung betriebsahnlicher
ing Mats., Vol. 54, 1954, p. 903. Belastungsfolgen auf die Festigkeit von
(6) H. T. Corten and T. J. Dolan, "Cumula- Ilugzeugbauteilen," Berichl der DVL, Cf
tive Fatigue Damage," International 407/5, August 1941.
Conference on Fatigue of Metals, Lon- (13) E. Gassner, "Some Remarks About Wing
don, Inst. Mechanical Engrs., p. 235, Load Spectra, Program and Flight
September, 1956. Tests," Minutes of the 4th Conference
(7) W. A. P. Fisher, "Some Fatigue Tests on of the International Committee on
Notched Specimens with Programme Aeronautical Fatigue, Zurich, May
Loading for a 'Ground-Attack' Air- 1956.
S c H i j v E ON ESTIMATION OF FATIGUE PERFORMANCE 213
(14) E. Gassner and W. Schiitz, "The Signifi- (25) H. Press and J. W. Tukey, "Power Spectral
cance of Constant-Load-Amplitude Tests Methods of Analysis and Their Applica-
for the Fatigue Evaluation of Aircraft tion to Problems in Airplane Dynamics,"
Structures," Proceedings, Symposium on AGARD Flight Test Manual, Vol. 4, Part
Fatigue Testing of Aircraft Structures, IV C, 1957.
Amsterdam, June, 1959, p. 14. (26) J. Schijve and F. A. Jacobs, "Research on
(15) E. Gassner and K. F. Horstmann, "Einfluss Cumulative Damage in Fatigue of Riveted
des Start-Lande Lastwechsels auf die Aluminum .\lloy Joints," Nat. Aeronautical
Lebensdauer der boenbeanspruchten F'liigel Research Irs!., Amsterdam, Report M.
von VerlNe>.f flugzeugen," Advances in 1999, ]anuav , 1956.
Aeronautical Sciences, Vol. 4, Pergamon (27) J. Schijve, "Critical Analysis of the
Press Ltd., London, 1961, p. 763. Fatigue Damage Concept and Some
(16) H. F. Hardrath, E. C. Utley, and D. E. Consequences for Fatigue Testing of
Guthrie, "Rotating-Beam Fatigue Tests of Aircraft Structures," Minutes, 4th Con-
Notched and Unnotched 7075-T6 Alumi- ference of the International Committee
num Alloy Specimens Under Stresses of on Aeronautical Fatigue, Zurich, May,
Constant and Varying Amplitudes," NASA 1956, Appendix 2.
TN D-210, December, 1959. (28) J. Schijve, "The Endurance Under Program-
(17) A. K. Head and F. H. Hooke, "Random Fatigue Testint-," Proceedings, Symposium
Noise Fatigue Testing," International on I'atigue Testing of Aircraft Structures,
Conference on Fatigue of Metals, Inst, of Amsterdam, June, 1956, p. 41.
Mech. Engrs., London, 1956, p. 301.
(29) J. Schijve and F. A. Jacobs, "Program-
(18) C. M. Hudson and H. F. Hardrath, Fatigue Tests on Notched Light Alloy
"Effects of Changing Stress Amplitude on Specimens of 2024 and 7075 Material,"
the Rate of F'atigue-Crack Propagation in Nat. Aeronautical Research Inst., Amster-
Two Aluminum Alloys," NASA TN D-960, dam, Technical Report M. 2070 (1960).
September, 1961.
(30) J. Schijve, "Fatigue Crack Propagation in
(19) J. Kowalewski, "On the Relation Between
Light Alloy Sheet Material and Struc-
Fatigue Lives Under Random Loading
tures," Advances in Aeronautical Sciences,
and Under Corresponding Program Load-
Vol. 3, Pergamon Press Ltd., London,
ing," Proceedings, Symposium on F'atigue
1961, p. 387.
Testing of Aircraft Structures, .Amsterdam,
June, 1956, p. 60. (31) J. Schijve, "The Analysis of Random
Load-Time Histories with Relation to
(20) A. J. McCulloch, M. A. Melcon, W. J.
I'atigue Tests and Life Calculations," 2nd
Crichlow, H. W. Foster, and J. Rebman,
ICAF-AGARD Symposium, Paris, 1961,
"Investigation of the Representation of
NLR-Report MP. 201, May, 1961.
Aircraft Service Loadings in F'atigue
Tests," ASD Technical Report 61^435, (32) J. Schijve and D. Broek, "Crack Propaga-
September, 1961. tion Under Variable-Amplitude Loading,"
(21) E. C. Naumann, H. F. Hardrath, and D. E. Aircraft Engineering, Vol. 34, 1962, p. 314,
Guthrie, "Axial-Load Fatigue Tests of and NLR-Report MP. 208, Amsterdam,
2024-T3 and 7075-T6 Aluminum Alloy D e c , 1961.
Sheet Specimens Under Constant- and (33) J. Schijve, D. Broek, and P. de Rijk,
Variable-Amplitude Loads," NASA TN "Crack Propagation Under Variable-
D-212, December, 1959. Amplitude Loading," NLR-Report M.
(22) A. O. Payne, "Random and Programmed 2094, Jan., 1962.
Load Sequence Fatigue Tests on 24ST (34) F. R. Shanley, "A Theory of Fatigue
Aluminum Alloy Wings," ARL Report Based on Unbending During Reversed
SM.244, Dept. of Supply, Melbourne, Slip " The Rand Corp. Report No. P-350,
September, 1956. Nov., 1952, Supplement, May, 1953.
(23) A. O. Payne, "Determination of the Fatigue (35) C. R. Smith, "Fatigue-Service Ijfe Pre-
Resistance of Aircraft Wings by Full- diction Based on Tests at Constant Stress
Scale Testing," Proceedings, Symposium Levels," Proceedings, Soc. Experimental
on F'atigue Testing of Aircraft Structures, Stress Analysis, Vol. 16, No. 1, p. 9 1958.
Amsterdam, June, 1956, p. 76. (36) S. R. Valluri, "A Unified Engineering
(24) F. J. Plantema and J. Schijve (Editors), Theory of High Stress Level Fatigue,"
Proceedings, Symposium on Full Scale Aerospace Engineering, Vol. 20, 1961, p. 10.
Fatigue Testing of Aircraft Structures, (37) R. E. Whaley, "Fatigue Investigation of
Amsterdam, June, 1959, Pergamon Press Full-Scale Transport-Airplane Wings,"
Ltd., London, 1961. and "Variable Amplitude Tests with a
214 SYMPOSIUM ON FATIGUE or AIRCRAFT STRUCTURES
DISCUSSION
MR. K . HORSTMANN (presented in Smax/Sm = 2. In thls case, the effect of
written form)}—These comments on the the GTAC is large, since the minimum
paper by Mr. Schijve will show some stress due to gust loads is zero; but the
results of the tests carried out in the GTAC will induce stresses in the com-
Laboratorium fiir Betriebsfestigkeit, pressive range (see Fig. 11).^ The GTAC
-10
-12
Darmstadt, Germany. They are con- may then reduce the favorable effects of
cerned with the effect of the ground-to- residual stresses built up at the maximum
air cycle (GTAC) on the fatigue life of
gust-loaded aircraft structures. The - Assumptions in Fig. 11.
•Sm.ix = 20 kg per sq mm = constant (induced
effect of the GTAC depends on the design by a gust load of 40 ft per sec);
characteristics of the aircraft, as will be Sm = stress at 1 — 9 level;
shown in the following: Gust frequenej' distribution = linear relation on
double-log scale from H = 1 at a gust veloc-ity
A normal transport-aircraft structure of 40 ft i)er sec to H = 10° at a gust velocity
may be designed for a maximum stress, of 2 ft per sec; H is the number of exceedings
of the corresponding gust velocity.
Smax (due to gust loads), which is twice Number of GTAC's, n = 150 (see reference
the stress S„i at the 1-g load level, or (15) of the paper);
Minimum stress of GTAC = —4 kg per sq mm;
^ Laboratorium fUr Betriebsfestigkeit, Darm- and
stadt, Germany. Maximum stress of GTAC = Sm .
DISCUSSION ON ESTIMATION OF FATIGUE PERFORMANCE 215
positive gust loads. Results of recent tests smaller transport aircraft are based on
(see reference (7) of the paper) have the ratio Smax/S„ = 3.5 and even more.
shown a reduction of the fatigue life to Finally the cumulative damage rate
40 per cent of the original fatigue life
without GTAC's (see Fig. 12). Other is better approximated by 2.^ — = 1 at
results of current tests with a gust-load higher values of Smcx/S,,,. Further tests
spectrum, for which Smax/S„ = 1.8, have to be carried out to confirm the
show an even larger effect, the fatigue curve in Fig. 12.
life being reduced to 20 per cent of the MR. J. ScHijVE {author).—Mr. Horst-
fatigue life without GTAC's. mann's discussion is a welcome addition
For the hypothetical case to the paper. His results confirm the
the gust loads do not cause any damage author's statement that the damagitig
and the GTAC theoretically reduces the effect of the GTAC is caused by the
elimination of favorable residual stresses.
The point which he made that the
damaging effect of the GTAC will
depend on the severity of the GTAC, as
compared with the maximum negative
gust loads, is well taken. This is illus-
0 10 2 0 30 4 0 50 6 0 70 8 0 9 0 100 trated in Mr. Horstmann's second figure
Percentage of Fatigue Life Without GTAC (Fig. 12). Although the trend of this
FIG. 12.—Effect of the Ground-to-Air Cycle
figure is qualitatively correct, it should
(GTAC) on the Life of Gust-Loaded Wings be pointed out that the curve will
of Aircraft Structures for Different Ratios of depend, among other variables, on the
magnitude of the load caused by the
life to 0 per cent, as indicated in Fig. 12. GTAC. This load will vary from one type
For large values of Smax/S,,,, the GTAC of aircraft to another. But it will also
will lose its significance, and it? effect vary within a wing structure, even as a
will become negligible at approximately percentage of the 1-g load level. The
Smax/Sm = 5 (sce Figs. U and 12). This author agrees that a further study of the
can be explained since large negative damaging effect of the GTAC is worth-
gust loads will enter the compressive while. We are now performing some
range and will also eliminate the favor- research on this topic.
able residual stresses. Tests carried out It is doubted, however, whether a
by Payne (reference (23) of the paper), curve such as shown in Fig. 12 could have
at Smax/Sm = 2.3, confirm the conclusion, a general applicability, since too many
as can be seen in the curve of Fig. 12, variables are involved. To make a
which shows the trend of fatigue life quantitative estimate of the effect of
reduction by the GTAC. the GTAC the designer will have to give
These conclusions may be important due consideration to his particular
from the point of view of the designer, circumstances. To improve his under-
for many questions on the effect of standing of the problem he needs
GTAC on the life of aircraft structures empirical evidence. Therefore, Mr. Horst-
are still unanswered. Newer designs of mann's data are most welcome.
STP338-EB/Sep. 1963
B Y M. S. ROSENFELD 1
SYNOPSIS
Twenty-six unused fighter aircraft were made available for use in an investi-
gation to determine the fatigue characteristics of a typical fighter airplane.
The primary purpose of this investigation was to obtain fatigue data on full-
scale structures for use in designing new aircraft. The aircraft were disassem-
bled into major components: wing, horizontal tail, vertical tail, nose landing
gear, and main landing gear. Constant-amplitude and variable-amplitude tests
were performed on the various components to determine: (1) the degree of
scatter, (2) the effect of cumulative damage and the validity of various cumu-
lative damage hypotheses for full-scale structures, (3) the effect of randomiza-
tion of the load sequence within a spectrum block, (4) the effect of superim-
posing landing loads on a flight load spectrum, (5) the effect of prestressing,
and (6) the effect of varying the mean and alternating loads. A discussion of the
test results and their significance for future design use is included.
216
Copyright^ 1963 by ASTM International www.astm.org
ROSENFELD ON N A V Y RESEARCH ON STRUCTURAL F A T I G U E 217
the critical design condition for each Wing specimens were used for con-
structure. Figure 1 shows the horizontal stant- and variable-amplitude tests. T h e
tail test setup, which is typical of the minimum load for both types of tests
wing and vertical tail test setups b u t less was established at 1 g (13.33 per cent
complicated. This type of setup is limit load) which is the normal, un-
typical for static and fatigue tests of air- accelerated flight load for this airplane.
craft structures. The maximum loads for the constant-
All tests were load-amplitude con- amplitude tests ranged from 40 to 120
trolled to simulate actual operating per cent limit load; 100 per cent limit
conditions. Load cycling was automati- load corresponds to 7.5 g for this air-
cally controlled by means of micro- plane. The static strength of the wing
switches installed on the dynamometers. as determined from the manufacturer's
Actuation of a switch controlled the static tests is 183 per cent limit load;
Operation of the hydraulic by-pass valve this value was used to convert the load
permitting the load to build up or fall limits from a limit load basis to an actual
off as required. The basic load control static strength basis. The maximum
and hydraulic systems were previously load-test spectra simulating 20-hr blocks
described (2). of flight loads are given in Table I ;
The load frequency was dependent spectra I and V I I were used for the
upon the magnitude of the structural wing tests. The fixed-sequence tests for
deflection for the load applied. For the spectra I and V I I were performed by
wing, the frequency varied from 3 cpm applying the lowest load in the block for
at limit load to 10 cpm at 40 per cent the required number of cycles and then
limit load. The horizontal tail load successively applying the next higher
frequency varied from 3 cpm at 150 per load levels for the required number of
cent limit load to 45 cpm at 40 per cent cycles. At the completion of each block,
limit load. Similar frequency ranges were the lowest load was then applied to start
used for the vertical tail tests. the succeeding block. The random-se-
220 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
quence tests for spectrum I were per- sec. The application of 20 landing-load
formed by using the same simulated cycles per block was arbitrarily assumed
20-hr block size as for the fixed-sequence to represent one landing per hr of flight.
tests, but the maximum loads within Similar constant-amplitude and fixed-
T A B L E II.- - R E S U L T S O F C O N S T A N T - A M P L I T U D E T E S T S O F H O R I Z O N T A L
TAIL SURFACE.
33 249 nc
150 ( 8 5 . 2 ) . 34 158 nc
Median 204 1.58
31 553
125 ( 7 1 . 0 ) . 32 313
Median 433 1.77
1 2 252 6 0.27
2 1 382 4 0.29
9 1 371 9 0.66
100 ( 5 6 . 9 ) . 10 352 9 0.67
11 370 7 1.97
12 431 33 2.31
Median 1 376 0.66 1.66 090
5 4 077 5 0.12
6 3 316 170 5.12
80 ( 4 5 . 5 ) . 19 4 180 160 3.83
20 3 811 132 3.46
Median 3 944 146 3.64 1.26 045
" I n per cent of design limit load. N u m b e r s in parentheses are m a x i m u m loads in per cent of static
strength.
' Visible crack.
comparable to that from the wing tests. to 100 per cent limit load; the vertical
The maximum loads for the constant- tail static strength as indicated by the
amplitude tests ranged from 40 to 150 manufacturer's static tests is approxi-
per cent limit load. For the variable- mately 176 per cent limit load. Tests
amplitude tests, wing load spectra I to VI were performed for i? = 0, — 5, and —1
of Table I were used for convenience to with the direction of loading for i? = 0
permit direct comparison with the wing assumed as positive. The minimum loads
data based on the assumption that the were zero for R = Q, —50 per cent of the
tail surfaces can be considered as small maximum load for R = — | , and — 100
wings. This eliminated the complexity of per cent of the maximum (complete rever-
-
160
^^«sr^>^ B>>^ ^ ^
Wing^
.^V«rtical Tail
R=0
20
_
-
1 1 1 1 1 111 1 1 - _.. 1 II mil 1 1 1 1 1 III 1 1 1 1 1 III
coping with the controversial problem of sal) for i? = — 1. The first two tests of the
what constitutes an acceptable horizontal vertical tail were performed with the
tail spectrum and permitted direct com- vertical tail installed on the fuselage to
parison of the fatigue characteristics of obtain data on the load distribution in
two different structures. The loads were the attachment fittings, since the attach-
applied in a low-high-low sequence that ment to the fuselage is statically indeter-
approximated a sinusoidal sequence minate. These data were used to con-
insofar as possible within the capabilities struct a support jig for the remainder of
of the loading system. the tests; the spring constant of the
Only the constant-amplitude tests of support jig closely matched the spring
the vertical tail were performed. For all constant of the fuselage in the vicinity
tests, the maximum load ranged from 40 of the attachment fittings.
222 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
^,e,. -i>»;-^
T A B L E III,—RESULTS OF C O N S T A N T - A M P L I T U D E T E S T S OF W I N G .
16L 619 A
120 ( 6 6 . 7 ) , 17L 900 A
Median 760 45
3R 409 B
4R 146 A
100 (55.6) , 5R 345 A
6R 534 A
Median 246 57 090
9L 656 B
80 ( 4 4 , 5 ) , lOL 826 A
Median 241 25
IIL 24 288 A
60 (33,3), 12L 29 558 C
Median 26 923 22
° I n per cent ot design limit load. N u m b e r s in parentheses are m a x i m u m loads in per cent of
static s t r e n g t h .
' B o t h left (L) and right (R) h a n d sides of each specimen were tested simultaneously. L or R
n e x t t o specimen n u m b e r indicates which side failed.
•^ A = M a i n spar c a r r y - t h r o u g h fitting at wing station 33,25,
B = M a i n spar c a r r y - t h r o u g h fitting a t wing station 25,5,
C = M a i n spar carry-through fitting at wing statioti 3.0, and
D = M a i n spar lower flange at wing s t a t i o n 117,0.
223
224 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
and the life to failure is plotted in Fig. 2. Constant-amplitude tests of the wing
The failures, which occurred at the were also performed. The maximum
fuselage attachment, were almost iden- loads ranging from 40 to 120 per cent
tical (Fig. 3). Attempts were made to design limit load correspond to 22.2 and
detect initial cracking, but the crack 66.7 per cent of the wing static strength.
could not be found before it extended aft Results are summarized in Table I I I
of the leading edge skin as shown in Fig. and plotted in Fig. 2. The failures ob-
3, upper left. The cracking was usually tained were not all identical as indicated
accompanied by a loud noise during the in Table I I I ; all failures occurred in the
cycle immediately preceding crack detec- wing-fuselage attachment except for the
tion. Data on the life remaining after two at wing station 117.0. Ten of the
crack detection are included in Table I I ; twelve failures occurred in the main spat
this life is plotted against maximum load carry-through fitting; seven of these ten
in Fig. 4. Table I I also includes data failures occurred at the same location
indicating the scatter in lifetime for (Fig. 5). Crack detection prior to failure
identical specimens tested under identi- was considerably more difficult during
cal conditions. Scatter is represented as the wing tests than during the horizontal
follows: in all cases the ratio of maximunT tail tests, and the data obtained were not
to minimum life is shown; and where consistent. About halfway through this
sufficient data are available, the standard test program, an eddy-current crack de-
deviation of the log-life was computed. tector was used to provide a consistent
ROSENFELD ON N A V Y RESEARCH ON STRUCTURAL FATIGUE 225
Specimen Life to
R Maximum Load" Number Failure, Rear Spar Skin
cycles (N)
Cycles Per cent N Cycles Per cent N
" In per cent of design limit load. Numbers in parentheses are maximum loads in per cent static
strength.
* Visible crack.
crack detection method. The area in- detector was used, cracking was indicated
spected was the inside of the two holes prior to 50 per cent of the specimen life;
through which the most frequent failures however, the crack was not visible for a
occurred. Whenever the eddy-current considerable period of time after this.
226 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
The vertical tail specimens were the fuselage attachment fitting (Fig. 7).
divided into three groups: R = 0, —5, Prior to failure, initial cracks were de-
and — 1 . Two specimens of each group tected during most of the tests in the
were tested for each jnaximum load, tension flange of the rear spar at the
which ranged between 40 and 100 per upper end of the attachment fitting and
cent design limit load and corresponded in the skin in the same vicinity. The life
to 22.7 to 56.9 per cent of the actual remaining after cracking is included in
static strength. Test results are sum- Table IV, and these data ior R = 0 are
marized in Table IV^ and plotted in Fig. plotted in Fig. 4 for comparison with the
6; the life lor R = 0 is also shown in Fig. wing and horizontal tail data.
2 for comparison with the horizontal tail
Variable-Amplitude Tests:
and wing data. In Fig. 6, scatter bands
are indicated for the R = Q and R = —1 An extensive investigation of the
results. It was impossible to draw a horizontal tail was performed for the
logical band ior R = —^ because of the basic and five variations of the basic
great amount of scatter. This scatter was fixed-sequence spectrum. Eighteen speci-
caused by diifiiculty in maintaining the mens were tested, three for each spec-
proper spring constant in the support jig trum, and the results are summarized in
for the 2? = — I group of tests. All failures Table V. All the failures were similar to
were similar and occurred at the upper the failures obtained during the con-
end of the attachment of the rear spar to stant-amplitude tests. Only visual de-
R O S E N F E L D ON N A V Y R E S E A R C H ON STRUCTURAL FATIGUE 227
tection of cracks was attempted. Al- to minimum lives for each group and as
though all the constant-amplitude tests the standard deviation of the log-life.
of the tail for maximum loads up to 100 The test results for spectra I, II, and
per cent limit load exhibited visible III seem to overlap sufficiently that
cracking prior to failure, only eleven of there appears to be no significant eiiFect
„. f [ 21
22
210
174
113
93
209
770
nc
539
0
0.57
8.02
6.64
1.718
1.423
Lo-Hi-Lo.. 1 26 125 67 309 539 0.80 4.77 1.021
Average 170 91 429 359 0.46 1.68 0.114 6.48 1.387
i [
[ f 27 107 107 560 27 229 25.3 3.92 0.867
28 75" 74 536" 27 561 37.8 2.75 0.601
Lo-Hi-Lo.. 1
"•- 35 109 109 572 nc 0 3.99 0.883
[ I Average 97 97 223 18 260 21.0 1.47 0.095 3.55 0.784
[ 40
41
85
95
45
50
459
883
nc
1
0.0
0.00
3.06
3.43
1.016
1.136
Lo-Hi-Lo I
42 66 35 256 537 1.52 2.38 0.788
- \ [ Average 82 43 849 179 0.51 1.44 0.081 2.96 0.980
If
29 124 31 680 nc 0.0 3.97 0.825
6 5 . 3 per cent
30 144 37 020 nc 0.0 4.61 0.964
v.,J Prestress <
36 157 40 160 10 090 25.1 5.02 1.046
Lo-Hi-Lo .
Average 142 36 287 3 363 8.37 1,27 0.052 4.53 0.945
i
t 37 215 55 019 15 954 29.0 6.89 1.273
7 3 . 8 per cent
38 253 64 833 nc 0.0 8.10 1.500
VI...I Prestress <
39 267 68 430 3 428 5.0 8.55 1.584
Lo-Hi-Lo .
Average 245 62 761 6 461 11.3 1.24 0.049 7.85 1.452
" See T a b l e I .
'' For spectra I, V, a n d V I ; 70, 100 85, 55 per cent limit load
F o r spectra I I : 40, 70, 100, 85, 55 per cent limit load
F o r s p e c t r u m I I I : 30, 70, 100, 85, 55 per cent limit load
F o r s p e c t r u m I V : 40, 70, 85, 55 per cent limit load
' Twent.y-hour p r o g r a m blocks.
'' Visible crack.
the eighteen specimens tested for flight- of including loads below 31.3 per cent of
loads spectra cracked prior to failure. the static strength. The tests for spec-
For four of these eleven speciinens, the trum IV, which is identical to spectrum
life remaining after crack detection was II except for the omission of the three
less than 1 per cent of the total specimen cycles per block at 100 per cent limit
life. Here again, the scatter in life is load, showed a significant reduction in
presented both as the ratio of maximum life resulting from the omission of these
228 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
three cycles per block. This is contrary to used. Two specimens were tested for
the effect predicted by the Miner-Palm- spectrum VII, which is the positive load
gren cumulative damage hypothesis. The portion of the fighter spectrum specified
tests for spectra V and VI indicated that in the latest design specification, MIL-
prestressing can have a beneficial effect, A-8866, to determine experimentally the
provided the prestressing is sufficiently relative severity of the new and old
high. flight spectra and to obtain an approxi-
Life to Failure
Specimen Type A^max a (log A'Se
Spectrum'* Sequence'' Number" of
Cycles Failure* A min A) A|7
Blocks''
(A')
The results of the wing tests for mate correlation factor between the two
variable-amplitude loading are sum- spectra for a given structure. The test
marized in Table VI. The first six speci- results indicate that spectrum VII is
mens were tested for the same spectrum; almost twice as severe as spectrum I.
three specimens were tested for a fixed The last two wing specimens were used
sequence of loading and the other three for an investigation of the effects of
specimens were tested for a random ground - air - ground cycling. A 20-hr
sequence of loads within each block. The flight load spectrum, spectrum I, was
results show no significant difference applied. At the completion of each block
between random and fixed sequence of flight loads, 20 cycles of a constant-
loading for the 20-hr program block amplitude landing load, approximately
ROSENFELD ON N A V Y RESEARCH ON STRUCTURAL F A T I G U E 229
80 per cent of the design limit landing bending moment for the flight load
load, were applied. This test sequence condition and of the same sign. Here,
was continued to failure. The results again, the data on cracking prior to
indicated no significant effect of super- failure were inconsistent. The eddy-
position of the landing loads on the current crack detection method indi-
flight load spectrum. This was to be cated the presence of microscopic crack
expected as a result of the landing gear prior to 50 per cent of the life. The
location and the magnitude of the wing failures again were not all in the same
bending moments in the critical region. location; however, six of the ten failures
The wing bending moment for the land- occurred in the same location as the
ing condition was equivalent to approxi- majority of the constant-amplitude test
mately 50 per cent of the limit load wing failures.
230 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
0.35
0,30
© Limiting Line
® A u s t r d l i a n Wing T e s t s ) Reference ( 4 )
0.25
N L L Tests of Joints
„ 0.20
0.15
0.10
O.OS
lO"* 10=
Median Life >o Failure , Cycles
FIG. 8.—Standard Deviation of Log N for Constant and Variable-Amplitude Fatigue Tests.
group tested. These ratios range from and the regression line is superposed on
1.06 to a maximum of 3.26; the mean the three regression lines shown by
ratio is 1.52 and the standard deviation Freudenthal (4). The regression hne for
is 0.52. These ratios are based on a the Aeronautical Structures Laboratory
limited number of observations for each (ASL) tests, which is based on both
group; one group consisted of only two constant- and variable-amplitude test
specimens and only two groups con- data, falls between the line for the
sisted of as many as six specimens. The Australian wing tests and the regression
results of the spectrum load tests fall line for the riveted joint tests. It is
within the same range of maximum to believed that the line for the Aus-
600
500
400
300
^ 200
100 —t
70 75 80 85
Prestress in Per Cent of Stotic Strength
minimum life ratio as the constant- tralian wing tests might represent a
amplitude test data and are included in fair estimate of the standard devi-
the determination of the mean and ation for new design purposes. Quality
standard deviation for this ratio. control can seriously affect the mag-
A better estimate of the scatter can be nitude of the standard deviation for a
obtained from the standard deviation of given population of specimens; therefore,
the log-life, which was estiniated only for if the quahty control is exceptionally
groups containing three or more speci- good, the standard deviation will ap-
mens. These values are also included in proach the line for the riveted joint
Tables I I to VI. The values of the laboratory specimens. If quality control
standard deviation are plotted against is poor, then the standard deviation will
the median life for each group in Fig. 8, probably approach the limiting line.
232 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
One of the test points falls considerably to cycling for the load spectrum. The
above the limiting line. The results from results (Fig. 9) show that the lower value
the constant-amplitude test of a single of the prestress had no significant bene-
specimen (horizontal tail specimen 13) ficial effect whereas the high prestress
far exceeded the lives of the other five approximately doubled the life. For
specimens tested in that group. If the comparison purposes the curves showing
data from specimen 13 are omitted, then the beneficial effects of prestressing ob-
the standard deviation could be reduced tained by Payne et al (5) are also shown.
• Wins
O Horizontal Toil
A Wing- Reference (I)
I I I I I 11
I0-' 10^ 10° 10
Life Estimated By Linear Rule , Cycles
I'lG. 10.—Comparison of Actual Life and Life Estimated by the Linear Rule.
from a value of 0.21 to 0.13 for the group All the curves indicate a beneficial effect
of specimens tested at 40 per cent limit resulting from prestressing, but the
load. magnitude of this beneficial effect ap-
pears to be as much a function of the
Effects of Prestressing: loading subsequent to prestressing as it
Six horizontal tail specimens were is of the magnitude of the prestress itself.
tested for a fixed-sequence load spectrum From the curves it is seen that pre-
(spectrum I) after application of a single stressing between 71 and 86 per cent of
prestress. Three specimens were pre- the static strength will be required to
stressed to 65.3 per cent of the static double the life of the structure, and each
strength and three were prestressed to curve is applicable to a single loading
73.8 per cent of the static strength prior condition only. For the multiplicity of
ROSENFELD ON N A V Y RESEARCH ON STRUCTURAL F A T I G U E 233
loading conditions required for the de- were analyzed by this method and the
sign of an airframe, the airframe should summation of the damage ratios are
be prestressed for a number of different shown. These ratios range from 2.38 to
loading conditions to obtain over-all 8.20 based on actual failure of the struc-
beneficial effects. This is economically tures tested. Figure 10 shows the actual
impractical. Furthermore, prestressing life to failure plotted against the life ex-
that would introduce a residual stress pectancy estimated by this method. A line
with a resultant beneficial effect in one was drawn through the lowest test points
part for one loading condition could to represent the minimum "safe" life for
introduce a residual stress in another full-scale structures for i? > 0. Many
part that would have a deleterious effect. investigators have suggested that the
Therefore, prestressing should not be summation of the damage ratios should
used as a means of increasing fatigue life be a value X instead of unity; however,
of a complicated structure. this plot indicates that the summation
of the damage ratios may be a linearly
Cumulative Damage: increasing function of the log-life rather
In recent years a considerable number than a constant. Freudenthal (6) shows
of hypotheses have been proposed for similar plots for rotating-beam specimens
estimating the fatigue life of aircraft tested for randomly applied variable-
structural components subjected to re- amplitude loads; however, these data
peated loads of varying amplitudes. One plot below the line representing the life
of the primary objectives of the ASL expectancy determined by the linear
fatigue investigations has been to rule. Comparison of the data presented
evaluate the applicability of these cumu- here and by Freudenthal indicates that
lative damage hypotheses to full-scale the summation of the damage ratios
structures. All the proposed hypotheses might possibly be a function of the
except four were eliminated from con- minimum to maximum load ratio and of
sideration because: (1) the physical the total life. Furthermore, from an
mechanism was not clearly defined so analysis of the data in Table V, it can be
that the relation did not contain con- shown that this hypothesis does not
cepts readily identifiable with usable always indicate the direction of life
design concepts, (2) an excessive amount expectancy change when the loading
of experimental data are required for spectrum is altered. Thus, the Miner-
practical engineering application, or (3) Palmgren hypothesis will not produce
mathematical calculations are cumber- valid life estimates for the high-stress,
some. Three of these hypotheses are low-life end of the S-N curve.
discussed here; the fourth, which is based
upon residual stresses introduced by the Tangent-Intercept Method:
high loads in the spectrum, is discussed The tangent-intercept method pro-
in a paper by C. R. Smith (U). posed by Schleicher (7) and Bland and
Sandorf fl3) defines the life for variable-
Miner-Palmgren Method: amplitude loading conditions as the
Although its validity has often been number of cycles under which the cumu-
questioned, the Miner-Palmgren or lative load-frequency curve for a given
linear rule method for estimating the time interval can be translated to the
effects of cumulative damage is still point of tangency with the constant-
widely used because of its simplicity. amplitude load-life curve. This method
The data presented in Tables V and VI predicts life expectancy with a smaller
234 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
error than the linear rule; however, the at the highest stress. Si, in the
results are still inadequate. One draw- stress spectrum
back of this method is its relative in-
sensitivity to changes in the load spec- .(2)
trum. The cumulative load-frequency - - ( ! ) '
curves for spectra I and I I and the mean where:
load-life curve for the horizontal tail Si = maximum stress applied by the
specimens are tangent in the range be- stress spectrum.
tween 40 and 70 per cent limit load. Si = stress at any level in the spectrum
Variations of the loads above and below below 5 i , and
this range, as in spectra I I I and IV, had
d = experimentally determined con-
no effect on the region of tangency or on
stant.
the life prediction even though the actual
The proportion of cycles applied at any
test results demonstrated significant
stress. Si, is given by the relation
variations in life.
.(3)
Corlen-Dolan Method: N'
One of the most recent cumulative
where:
damage hypotheses was proposed bj'
fti = number of cycles applied at Si
Corten and Dolan (8) and Liu and Corten
during one program block, and
(9). This method takes stress history into
N = l^tii = total number of cycles per
account in terms of the number of
program block.
damage nuclei initiated by the highest
The basic equation for the estimated
applied stress and in the propagation of
fatigue life, Ng, now becomes
damage by all subsequent stress cycles.
For these investigations the method re- Ni
N„ = (4)
quired a number of simplifications.
In the Corten-Dolan hypothesis for
estimating the fatigue life of a part
s (^) (sr
below the proportional Hmit, where Pi is
subjected to cyclic loading for any load
spectrum, the fatigue damage was the highest load in the spectrum and Pi
visualized as a joining or accumulation of is proportional to 5 , . To distinguish
cracks and was treated in terms of the between experimental and calculated
number of damage nuclei formed and the values for Ng, the subscripts e and t are
rate of crack propagation. The basic used; thus the experimental life is ^„„
equation for the estimated life, .V„, is: and the calculated (theoretical) hfe is
various assumed values of d using the tested at P i , nor does the estimated life,
values of »,•, N, Pi, and Pi from the -Y„, , reflect the distribution that would
horizontal tail data for spectrum I. be obtained if a large sample were tested
2. The average value of .Yi/.Y„^ was for the load spectrum. This introduces an
determined using the average values of error in the determination of the life
the lives (number of cycles) from Table V ratios since the predicted lives used for
and the mean value of -Yi at 100 per this determination were based on the
cent limit load (1500 cycles) from Fig. 2. mean life, Xi, and the test lives are at
3. Using this value of Xi/,Y„, , the unknown points within the distribution
value of d was determined from the for the load spectrum.
iVi/,Y.„, versus d curve for .A'i/.Y„^ = To estimate the possible magnitude of
.Vi/.V,,, . The value oi d = 8.25 for this error, the life ratios for four speci-
complete failure was determined by this mens were recalculated. The specimens
method. selected (Nos. 21, 22, 27, and 28 from
For analysis of the data, only the Table V) have life ratios more than one
ratio of experimental to theoretical life
was desired—not an estimate of the TABLE VII.—MAGNITUDE OF ERROR
theoretical life. This ratio is obtained IN LIFE RATIO OF HORIZONTAL TAIL
from reference (4) and becomes: SPECIMENS.
Life Ratios
(5) Specimen
Based on Based on 95 Per
Mean Life cent Coniidence
Limits
where:
No. 21 1.718 1.408
(6) No. 22 1.423 1.167
No. 27 0.867 1.084
No, 28 0.601 0.751
Using the value oi d = 8.25, the ratio
of experimental to predicted life was
determined for each wing and horizontal Standard deviation above or below the
tail specimen tested using a load spec- average value of 1.036. For specimens 21
trum, and the results are included in and 22, the life ratios were recalculated
Tables V and VI. The extreme values of with the arbitrary assumption that the
this ratio are 0.601 and 1.718, the average load-life curve for these specimens corre-
value is 1.036, and the median is 0.992. sponds to the upper limit of the 95 per
The proximity of the median and average cent confidence interval for the mean,
values indicates that the data are sym- and the value for Y'l was obtained from
metrically distributed. Assuming a nor- this curve (10). The life ratios for speci-
mal distribution of these life ratios, the iTiens 27 and 28 were recalculated simi-
estimated standard deviation for the data larly, except that the lower limit of the
in Tables V and VI is 0.272. confidence interval was used. The origi-
Equations 4 and 5 are both functions nal and recalculated life ratios for these
of the mean life, .Yi, corresponding to four specimens are shown in Table VII.
the maximum load. Pi, of the spectrum. The recalculated values more nearly
The arithmetic value of iYi does not take approach unity than do the original
into account the life distribution that values, and the difference between the
would be obtained for a large sample values is a measure of the possible error
236 SYMPOSIUM ON FATIGUE or AIRCRAFT STRUCTURES
REFERENCES
(1) R. A. Carl and T. J. Wegeng, "Investi- nautical Systems Division, U. S. Air Force,
gations Concerning the Fatigue of Aircraft WADD Technical Report 61-53 (1961).
Structures," Proceedings, Am. Soc. Testing (5) A. 0 . Payne, D. G. Ford, et al, "Fatigue
Mats., Vol. 54, p. 903 (1954). Characteristics of a Riveted 24ST Alumi-
(2) J. Rudnick and R. Friedman, "Loading num Alloy Wing, Part V: Discussion of
Systems for Conducting FuIl-Scale Aircraft Results and Conclusions," Aeronautical
Fatigue Tests," U. S. Naval Air Material Research Laboratories, Report SM 268,
Center, Report No. NAMATCEN-ASL- Melbourne (1959).
1028 (1960). (6) A. M. Freudenthal and R. A. Heller, "On
(3) W. W. Johnstone and A. O. Payne, "Deter- Stress Interaction in Fatigue and a Cumu-
mination of the Fatigue Characteristics of lative Damage Rule, Part I: 2024 Alumi-
Mustang P-51D Wings," Aeronautical num and SAE 4340 Steel Alloys," Wright
Research Consultative Committee, Report Air Development Center, U. S. Air F"orce,
No. 30, Melbourne, June, 1953. WADD Technical Report 58-69, Part I
(4) A. M. Freudenthal, "Fatigue Sensitivity (1958).
and Reliability of Mechanical Systems, (7) R. L. Schleicher, "Practical Aspects of
Especially Aircraft Structures," Aero- Fatigue in Aircraft Structures," Fatigue
238 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
DISCUSSION
M R . J. ScHijVE.'—The information on P.N„
program-fatigue tests is steadily increas- E-^T^Ci)'}-
Ni
• (b)
N^ ZP,
fer= 1, .(a)
Miner rule is applied, employing, how-
ever, a modified Wohler curve, namely,
xYv instead of Ni.
or ^ International Conference on Fatigue of
^ National Aero- and Astronautical Research ^Metals, Inst. Mechanical Engrs., London.
Inst., Amsterdam, Holland. Sept., 1956, p. 262.
DISCUSSION ON NAVY RESEARCH ON STRUCTURAL FATIGUE 239
It may be remarked that the physical wing tests with the bolt-hole probes sup-
model of the Corten-Dolan theory does plied with the ED-500. Once the location
not allow a favorable interaction, that is, of the crack origin was known, all subse-
CO J < 1, because the number of damage quent specimens were examined by this
nuclei m-i at the maximum load P] will be device. Location of the failures is shown
larger than the number nii created by on Fig. 5 of the paper. The two counter-
the lower level Pi. In accordance with sunk-head bolts at the failure location
reference (8) of the author's paper one were removed periodically and the in-
may write sides of the holes were inspected using the
bolt-hole probe. Meter readings obtained
were plotted against cycles, and curves
> 1, since nii > Wj
similar to crack propagation curves were
obtained. Cracks became visible at ap-
It is encouraging to see that the author proximately 50 per cent of the meter full-
could show that the analytical relation of scale deflection.
the Corten-Dolan theory can be made MR. R. L . BENEDICTO.^—My question
compatible with favorable interaction ef- refers to the effect of the lowest load level
fects. This proves the relation to be more in the spectrum on the lives of the hori-
flexible than the Palmgren-Miner rule, zontal tail specimens and the beam speci-
which is obviously due to the additional mens. Table V of Mr. Rosenfeld's paper
parameter, d. shows an increase in life from spectrum
The similarity of the analytical rela- I to II and a decrease from spectra I or
tions of Freudenthal, and Corten and II to III. Mr. Mordfin's paper^ shows a
Dolan was previously indicated by me.' large decrease in life from spectrum I to
MR. M . S. ROSENFELD (author) .—The II. The spectra in both papers are the
following are answers to Mr. Schijve's same, defined in terms of per cent limit
direct questions: load. Would the author care to comment
1. All specimens subjected to variable- on this apparent discrepancy?
amplitude loads failed at the highest load MR. ROSENFELD.—The data from the
in the spectrum. horizontal-tail tests for spectra I, II, and
2. As discussed in the paper, all fail- III show a considerable overlap and fall
ures originated at bolt holes. within the scatter normally found in full-
3. Visual examination of the fracture scale tests of structures. Consequently,
surfaces immediately after failure re- it was concluded that loads below 55 per
vealed typical clamshell marks surround- cent limit load (approximately 31 per
ing the crack nucleation point. This cent of the actual static strength) had
nucleus always occurred at the juncture a negligible effect on the fatigue damage
of the outermost surface of the part and for this spectrum.
hole surface. Evidences of fretting prior This statement appears to conflict with
to failure were detected as discussed in the results of box-beam tests presented
the paper. by Mordfin and Halsey. However, no
4. The eddy-current detectors used conflict actually exists. The apparent
were the Magnatest ED-500 and an ex- conflict is caused by the difference in the
perimental modification of this unit that definition of limit load. For the hori-
was limited to use with aluminum alloys. zontal-tail tests, 100 per cent limit load
Both devices were used only during the
' International Conference on Fatigue of •' Senior Structures Engineer, Republic
Metals, Inst. Mechanical Engrs., London, Aviation Corp., Farmingdale, L. I., N. Y.
Sept., 1956, p. 755. ' S e e p . 251.
240 SYMPOSIUM ON FATIGUK o r AIRCRAFT STRUCTURES
is the design limit load for the structure Mordfin and Halsey concluded that
and corresponds to 57 per cent of the truncation of the spectrum by elimina-
actual static strength. For the box-beam tion of loads above 40 per cent limit load
tests, 100 per cent limit load was ar- (approximately 27 per cent of the actual
bitrarily estabhshed as f of the actual static strength) produced unconservative
static strength. If the spectra are ex- results. This is in approximate agreement
pressed in terms of the static strength, it with the horizontal-tail results from
is evident that the spectra for the hori- which it was concluded that loads below
zontal tail and for the box beam are not 31 per cent of the actual static strength
the same. produced negligible damage.
STP338-EB/Sep. 1963
SYNOPSIS
This paper presents a method for predicting the fatigue life of a full-scale
structure for any combination of loading provided that a single datum point
representing the life of the structure for 104 or fewer cycles is known. The only
additional information required is stress-strain and S-N data for smooth
axially loaded coupons. An illustrative example of the use of this method is the
prediction of lives for tail structures yet to be tested at the Navy Aeronautical
Structures Laboratory, Philadelphia, Pa.
Any method for predicting fatigue life Thus, in trying to translate informa-
of a structure for spectrum-type loading tion from one loading spectrum to an-
should consider the position of the de- other, the designer does not know which
signer. In the early stages of design he is parts of the spectra are applicable, and
fortunate if he has some idea of load mag- therefore would be in a better position if
nitude, let along appropriate stresses and he had a single datum point that could
stress concentrations. Usually the de- be used for any load spectrum.
signer must estimate concentrations from This paper presents a method whereby
configurations that are never quite like the results of a constant-amplitude test
the one chosen in design. on a structure can be used to predict the
Accordingly it is common practice to life of that structure for any load spec-
build up a structural component in early trum. This method for predicting the life
stages of design and subject it to fatigue of structures is dependent upon the fol-
or static tests or both. Also, it is common lowing qualifications: (1) that the datum
for the loads used in final design to be point represents a life of less than 104
substantially different from those used in cycles, and (2) data for smooth axially
preliminary tests. If the structure had loaded coupons of the material are avail-
been subjected to fatigue tests involving able.
spectrum loading, the problem of then
applying the data to the succeeding de- MATERIAL PROPERTIES
sign becomes virtually impossible. This For the purpose of this discussion, ma-
stems from the fact that the highest load terial properties will be those exhibited
in the spectrum changes the life of the by the simple stress-strain curve and
structure for all subsequent lower loads, families of S-N curves for smoothly axi-
and the designer does not know the rela- ally loaded coupons. Material properties
tive amounts of damage contributed by illustrated here will be of 7075-T6 alu-
each load level in the spectrum. minum alloy.
1
General Dynamics/Convair, Fatigue Lab-
Figure 1 is a stress-strain curve for
oratory, San Diego, Calif. 7075-T6 aluminum alloy. Note that the
241
Copyright^ 1963 by ASTM International www.astm.org
242 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
slope of the curve above the yield most pertinent in this theory are pre-
strength is approximately 1000 psi per 1 sented below:
per cent strain. 1. Fatigue failure of a structure will al-
Four S-N curves are shown in Fig. 2 for ways occur at a stress concentration. This
four values of stress ratio, R. Note that assumption should be almost axiomatic;
the yield strength for this material equals in fact, if airplane structures were de-
the stress at about 10^ cycles for J? = 0 signed without concentrations there
8 3 . 0 ULTIMATE
ao
70 <kp^
(0
•a. 60
in 7075 -T6 ALUMIN JM AL LOY SMEET
en
UJ 50
().09'THICK
ir
1-
(0
40
30
20
1 2 3 4 5 6 7 8
S T R A I N , P E R CENT
80 1 II
R»0
70
60
-o.5r
50
- L- V
40
30
20
10
1
10^ lO- 10" 10" 10
CYCLES
FIG. 2.— 5-A" Curves for 7075-T6 Sheet.
and that at 10' cycles the stress for i? = 0 would be no need for preparing this
is approximately 7000 psi higher than for paper.
10^ cycles. According to the stress-strain 2. Strain at a concentration will be pro-
curve shown in Fig. 1, this would amount portional to load, even though localized
to about 7 per cent elongation. yielding occurs. This assumption is based
on the premise that locally yielded ma-
THEORY terial is adjacent to unyielded material
Any theory will of necessity employ a and is therefore constrained to deform at
certain number of assumptions; the four the same rate as the unyielded material.
SMITH ON PREDICTING LIFE or FULL-SCALE STRUCTURES 243
This has some experimental verification residual compressive stress at the concen-
within nominal limits. tration after load is removed, provided that
Accordingly, repeated loading (R = 0) the adjacent material is still unyielded.
that causes failure of a structure at 10^ This can be more easily visualized with
cycles is accompanied by a localized the aid of a photoelastic model. Figure 3,
strain cycle of less than 1 per cent, as for example, shows a piece of polyethyl-
shown in Figs. 1 and 2. In order to realize ene plastic film loaded in tension. The
= 0.003" INTERFERENCE
UJ
u
LU
2 3 4 5
SPECTRUM
i, R = 0, descending spectrum.
2, R = 0.5, descending spectrum.
3, fi = 0, ascending spectrum.
^, B = 0, ascending spectrum (40,000 psi load deleted).
5, iJ = 0, ascending spectrum (40,000 and 34,000 psi loads deleted).
FIG. S.—Results of Spectrum Tests on 7075-T6 Lugs, (Lugs Were Loaded with |-In.-Diameter
Tapered Bolts.).
NOTE.—Data for this figure were acquired under Air Force Contract AF-33(616)-8228, "Research
on Linear Strain Theory of Fatigue Life Prediction."
2 per cent strain, which according to the dark fringe area is plastically deformed
stress-strain curve would increase the and will be forced into compression (or
stress 2(X)0 psi, a load twice the magni- buckling) after load is removed. Being
tude of that causing failure at W cycles extremely thin, it chose the latter
would be required. A load of that magni- (Fig._4).
tude would usually cause the structure to With the exception of thin shear panels
fail and is beyond the scope of the present having lightening holes, buckling in a
paper. fabricated structure cannot occur. There-
3. Localized yielding will result in a fore, the locally yielded material will
SMITH ON PREDICTING LIFE OF FULL-SCALE STRUCTURES 245
have to go into compression, since there The previous assumptions supply the
is no other place to go. tools for estimating the behavior of a
4. Fatigue damage will be accumulated full-scale structure provided that a
at a linear rate with reference to stresses^ datum point for i? = 0 and a fatigue life
causing failure at constant amplitude. This of less than 10^ cycles were given. This is
means that the damage rate would be best shown by the tests discussed in the
linear if estimated in terms of life after a following paragraphs.
single application of the highest load to
be used in the loading spectrum. FULL-SCALE STRUCTURES
Figure 5 shows the estimated versus
actual lives of lug-type specimens sub- Constant A mplitude Tests:
jected to spectrum-type loading. Speci- In the first of a number of tests cur-
mens were made of 0.1-in.-thick 7075-T6 rently being conducted at the Aero-
sheet. Dimensions afforded a theoretical nautical Structures Laboratory, Naval
stress concentration factor of 3.6 in terms Air Material Center, Philadelphia, Pa.,
of average stress through the net area. one aircraft vertical stabilizer failed after
Loading pins were f-in.-diam taper bolts 1474 cycles of limit loading at i? = 0.^
Since this fin was tested at a load ratio of
TABLE I.—CYCLE LOADING SPECTRUM
U S E D I N T E S T S O N 7075-T6 L U G S .
zero (minimum to maximum load = 0),
it is an especially suitable example for the
Maximum Stress
(Nominal Net), psi Cycles per Sequence purposes of this paper. While the method
described here lends itself to any positive
40 000 3 load ratio, the description for a load ratio
34 000 17 of zero is easier to follow. Also, fatigue
28 000 65 life estimates made in this particular test
22 000 172
16 000 283 are predictions in the true sense of the
word.
in order to simulate swelling of rivets in a In the S-N curves in Fig. 2, note that
riveted structure. Except for the two ior R = 0 and a lifetime of 1474 cycles
cases with interference' (shown in Fig. 5), the stress for a smooth specimen would
the structure lives estimated from the have to be about 80,000 psi, which is sub-
S-N curves, developed after a single stantially above the yield strength for
40,000-psi preload, were very close to the material and could not be obtained
experimental values. Bars indicate av- without elongating the material 5 per
erages of four or more tests in each case. cent or more. The material used for de-
Table I indicates the test spectrum used veloping the S-N curves had a yield
in Fig. 5. strength (0.2 per cent offset) of 74,000
psi.
^ T h i s differs from t h e Miner (1) relationship
only in t h e respect t h a t a c t u a l stress a t t h e con- Since the yielded material is con-
c e n t r a t i o n is considered instead of n o m i n a l strained and is deformed along with the
stress. T h u s , t h e influence of compressive
residual stresses resulting from t h e highest load unyielded material adjacent to the con-
in t h e load s p e c t r u m is considered in life evalua- centration, this theory states that the
tion. Other work b y t h e a u t h o r (2) h a s indicated stress at the concentration could not ap-
t h a t this is a reasonable assumption. L a t e r work
b y C o r t e n and D o l a n (3) introduces these effects preciably exceed the yield strength for
as exponential functions which can be obtained the material without yielding the entire
empirically.
cross-section. Also, according to the tan-
' T h e t a p e r e d bolt was d r a w n u p sufficiently
t o increase t h e hole diameter 0.003 in. Provision
was made t o prevent t h e specimen from binding •* As of J u l y 1962, only one specimen had
in t h e test fixture. failed.
246 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
gent modulus for the material, 1 per cent strength' or 75,000 psi. This would indi-
strain would amount to less than 1000 cate that the actual stress cycle at the
psi in the plastically deformed material concentration was that conforming to a
at the concentration. stress ratio of —0.6, or a range from
How can it be known whether the en- + 75,000 to -45,000 psi. The total stress
tire cross-section was not yielded in the range of 120,000 psi can now be prorated
test that is used as an example? Again to estimate structure lives at other stress
referring to the S-N curves in Fig. 2, note levels.* To eliminate cumbersome data,
that the life for cycling at the yield the S-X curves in Fig, 2 are replotted
strength for i? = - 1 (±74,000 psi) is 500 (Fig. 6) in terms of effective yield
cycles. This represents strain cycling of strength of the material. This has the
from +0.8 to —0.8 per cent, or a total advantage of presenting numbers that
strain of 1.6 per cent. This amount of are easier to deal with, as well as making
110
100 II
III
u 80
- f (=0
1
Ki ;=o.
60 R = - 1' \
- o
5- R=-2
t
UJ
u
40 11 • |
10 10 10 10 10
CYCLES TO FAILURE
P"iG. 6.—Dimensionless S-N Curves for 7075-T6 Sheet.
160 = 128 per cent. Since 128 per cent per sequence are tabulated in Table II.
is still above the yield for the material, From this, n/N per sequence = 0.0089,
the actual cycle will be from +100 per from which total predicted life =
cent to — 28 per cent, and the predicted 1/0.0089 = 112 sequences.
lifetime will be found at the 100 per cent The remaining three loading conditions
load level fori? = -0.28, or 4800 cycles. presented in reference (5) are:
For 60 per cent limit loading, the stress
range will be 0.6 X 160, or 96 per cent. Loading condition 2 = Same as 1, except
Being below the yield strength, structure for addition of 283 cycles of 40 per cent
life can now be determined directly from loading.
the R = 0 curves at 96 per cent, or 13,000 Loading condition 3 = Same as 2, except
cycles. Similarly, the structure life for 40 for replacing 40 per cent loading with
per cent loading is found to be 60,000 750 cycles of 30 per cent loading.
cycles. Loading condition 4 = Same as 2, except
for deletion of 100 per cent loads.
Spectrum Tests:
For predicting life under spectrum Under conditions 2 and 3, the predicted
loading, the problem will be different in life of the structure in the test example
TABLE II.—CALCULATED LIVES A N D FRACTIONAL DAMAGE P E R SEQUENCE.
in that the residual stress resulting from would also be 112 sequences, since the
the 100 per cent load (60 per cent of damage at lower levels was considered
yield) will be retained throughout the neghgible* in view of the beneficial 60 per
test. Assuming that the test example will cent residual compressive stress intro-
be loaded in the same manner as that duced at the 100 per cent limit load. In
used for horizontal stabilizers (5),' the the fourth condition, where the 100 per
load spectrum for the first loading condi- cent loading was omitted, structure
tion would contain the following cycles fatigue lives at other levels would be re-
per sequence: duced. Also, the 40 per cent loading,
which otherwise does not contribute' to
Load, per cent of limit Cycles per Sequence the fatigue damage, contributes 0.0001
damage per sequence. Total damage per
100 3
85 17 * D a m a g e was considered negUgible a t lower
70 65 load levels only in view of beneficial residual
55 172 stresses resulting from high-level loading. I t was
assumed t h a t neither s u b s e q u e n t l o a d i n g nor
relaxation due to t i m e would erase very m u c h
Using the method described above, of t h e residual stress (60 p e r cent of yield).
calculated lives and fractional damage R e l a x a t i o n a t room t e m p e r a t u r e should be con-
sidered where t h e residual stress is greater t h a n
' T h e boldface n u m b e r s in parentheses refer t w o t h i r d s t h e yield s t r e n g t h , or where load
t o t h e list of references a p p e n d e d t o t h i s p a p e r . reversals a c t u a l l y reverse t h e stress (6).
248 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
sequence would be 0.0162 for a predicted In the examples discussed thus far,
life of 62 sequences. no consideration was given to the fact
Estimates were also made on the hori- that the mode of failure for low-level
zontal-tail surfaces (5). Ratios of test to loading may differ from that for high
predicted lives were 0.98, 1.35, 0.85, and loading. It is known that small amounts
1.19 for the four conditions, respectively. of preloading on a riveted joint may
Average for all conditions was 1.09. lessen its life, while high preloading im-
proves it (7), thus accounting for differ-
DISCUSSION OF RESULTS ent modes of failure. However, the mode
In the test selected as an example for of failure will be reflected by whatever
demonstrating this method of predicting influence the highest load in the spectrum
structure life, only one test point was has on the structure.'
available. This does not leave much Accordingly, where the predicted life
choice in a statistical analysis, and it is based on spectrum-type loading, and
10 10 10
CYCLES TO FAILURE
will be a problem that most likely con- the highest load in the spectrum is real-
fronts the designer. What should he do? istic with respect to probability of early
Assuming a standard deviation of log occurrence in life, the mode of failure
life equal to 0.15, the structure which should be the same for spectrum-type
was tested could have failed after only loading as for loading at the highest
700 cycles of 100 per cent loading instead load in the spectrum. Two or more
of 1474 cycles, or it might have failed modes of failure, however, could occur
after 3000 cycles. These structure lives at one load level, as it is reasonable to
would represent a spread of approxi- assume that more than one position in
mately plus or minus two standard devi- an entire structure could be suffering
ations from the original datum point. the same stress. This would not influence
Using the procedure described pre- the predicted life, since for equal stress-
viously, the expected structure life under ing the prediction is in terms of first
the first loading condition (5) would be crack wherever it may occur. For the
70 sequences for the low-life structure structure chosen, so little time elapsed
and 220 sequences for the other. The between first crack and final failure that
original calculation of 112 sequences failing hfe would be equally applicable.
does not represent a median for the two ' Except where low loading is accompanied
estimates. by local bending stress. See footnote 6.
SMITH ON PREDICTING LIFE OF FULL-SCALE STRUCTURES 249
Also, in the examples cited here, high action factor would further reduce the
loading increased fatigue life. This would predicted life.
generally be the case for airplanes having
high maneuver-load factors. In the case SUMMARY
of transport aircraft, maximum loading A method of estimating fatigue life of
may not exceed 1.8 g for a substantial structures for spectrum loading has been
portion of their service life. In the latter presented which requires only: (1) basic
case, a loss in fatigue life would be ex- material properties, and (2) a fatigue
perienced. This is shown by the S-N life for the structure under single-level
curves on riveted lap joints (Fig. 7), loading that causes failure in less than
where a fatigue-strength loss of more ICH cycles. Neither nominal stresses nor
than one third accompanied small concentration factors need be known.
amounts of prestressing (8). Being independent of the load spec-
One explanation for these losses would trum, this method lends itself to use with
be that the tightly driven rivets, which any load spectrum, as long as the load
heretofore prevented stress from return- producing failure at less than 10^ cycles
ing to zero, were loosened to permit full- is representative of one of the loads in
stress cycling. Larger amounts of pre- the spectrum, and the load ratio, R, is
stress introduced residual compressive zero or more. This is a special asset where
stresses which more than offset the effect structural tests are made in early stages
of rivet loosening (8). of design before firm loads are estab-
For spectrum-type testing, the S«/iV lished, and new life estimates are neces-
based on the lower curve was 1.28. Based sitated by later load information.
on the normally accepted middle curve,
the Zn/N would have been 0.23. This A cknowledgmenls:
may be the reason for the service life of The author wishes to express his ap-
some transport airplanes falling short, preciation to M. S. Rosenfeld of the
even though the fatigue life as indicated Naval Air Material Center and Charles
by conventional 5-iV curves would seem A. Davies of the Aeronautical Research
adequate. Laboratories, Office of Aerospace Re-
Although an additional effect of stress search, for permission to use data from
interaction (9) may also influence final Government contracts.
fatigue life, it would seem small in com- Credit is also due G. Garner Green,
parison with changes in residual stress chief structures engineer of General
and stiffness factors. This is particularly Dynamics/Convair, for valuable sug-
true in the case of lugs having interfer- gestions, to Glenn D. Lindeneau, research
ence fit loading pins, where the predicted test engineer, and other members of the
fatigue life fell far short of that shown Structures and Fatigue Laboratories for
in spectrum tests (Fig. 5). A stress inter- obtaining data used here.
REFERENCES
(1) M. A. Miner, "Cumulative Damage in Fa- Conference on Fatigue of Metals, Inst.
tigue," Journal of Applied Mechanics, Vol. Mechanical Engrs. (London) (1956).
12, pp. A159-A169 (1945). (4) C. R. Smith, "A System of Estimating
(2) C. R. Smith, "Prediction of Fatigue Failures Cumulative Fatigue Damage by Using the
in Aluminum Alloy Structures," Proceedings, Miner Rule Corrected for Residual Stress,
Soc. Experimental Stress Analysis, Vol. XII, Preprint No. 353B from 1961 SAE National
No. 2, pp. 21-28 (1955). Aeronautic Meeting, New York, N. Y.,
(3) H. T. Corten and T. J. Dolan, "Cumulative April, 1961.
Fatigue Damage, Proceedings, International (5) R. P. Swartz and M. S. Rosenfeld, "Variable
250 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
Amplitude Fatigue Characteristics of a Slab of Aircraft Structures, ASTM STP No. 203,
Horizontal Tail for a Typical Fighter Air- Am. Soc. Testing Mats., pp. 10-25 (1956).
plane," U.S. Naval Air Material Center Re- (8) C. R. Smith, "Influence of Residual Stresses
port No. NAMATCHEN'ASL-1023, Part and Stiffness Factors on Fatigue Strength of
II, Sept. 18, 1961. Metals and Structures," Proceedings, Third
(6) J. Schijve and F. A. Jacobs, "Program-Fa- Annual Engineering Symposium, Brigham
tigue Tests on Notched Light Alloy Speci- Young University, pp. 321-366, April 21,
mens of 2024 and 7075 Material," National 1962.
Luchtvaarlaboratorium, NLL-TR M 2070, (9) A. M. Freudenthal and R. A. Heller, "On
pp. 1-41 (1960). Stress Interaction in Fatigue and Cumula-
(7) C. R. Smith and G. D. Lindeneau, "Riveted- tive Damage Rule," Journal of Aerospace
Joints Fatigue Strength," Papers on Fatigue Sciences, Vol. 26, No. 7, July 1959.
STP338-EB/Sep. 1963
P R O G R A M M E D M A N E U V E R - S P E C T R U M F A T I G U E TESTS OF
A I R C R A F T BEAM S P E C I M E N S
SYNOPSIS
Aircraft structures are subjected to theories, in general, has been less than
cyclic loads of various magnitudes and adequate. As a result, full-scale fatigue
frequencies due to maneuvers, gusts, testing of aircraft structures under
landing and take-off, and so on. The multiple load levels is generally con-
estimation of fatigue life under these sidered a necessary step in the estima-
conditions is one of the more difficult tion of the fatigue life of aircraft under
problems of aircraft design. Many service conditions.
theories of cumulative fatigue damage Thousands of dynamic measurements
have been proposed over the years to have been made on aircraft in flight to
provide a rational basis for making these assess the actual loads imposed on the
estimates. T h e reliability of these structure so that realistic load-time
histories could be established for labora-
1
Aeronautical Research Engineer and Physi- tory simultation. T h e cost of full-scale
fatl ue testi
t^T^bJ^T^JS^- § "g I * * hi .s h '. and > n f d -
is
ington, D. c. less to say, the exact duplication of these
251
Copyright^ 1963 by ASTM International www.astm.org
252 SYMPOSIUM ON FATIGUE or AIRCRAFT STRUCTURES
histories in the laboratory would be too specimen provides some measure of the
time-consuming. To convert the service service life of the structure. This is a
load histories into more usable form, the useful measure only if the fatigue damage
programmed spectrum type of fatigue produced by the test spectrum is equal
test has been widely adopted. or relatable to the damage produced by
In a spectrum fatigue test, the speci- the service loads. Thus, the conversion
men is subjected to repeated applica- of service load histories into test spec-
tions of "blocks," a block representing trums is the critical feature of this tech-
the load history experienced by an air- nique. For reasons of economy and
craft during a typical interval of its convenience, it is desirable to devise the
service life. The load levels and the simplest programmed test spectrum that
number of cycles at each load level in the will adequately simulate the fatigue
160
\
\ \ / B e a m tpecimtns (ref«r«nce (I) )
120 \
40
FIG. 1.—Comparison of Fatigue Properties of Beams Tested by Howard and Katz (1) witli
Those of Various Aircraft Structures and Components; Various Minimum Loads.
block are collectively termed the test damage of the service loads. One of the
spectrum. Depending upon the size of questions which must be resolved in the
the service life interval being represented, process concerns the relative effects of
it is common to speak of, say, a 20-hr the various discrete load levels in the
block or a 1000-hr block and so on. If a programmed spectrum.
random test spectrum is used, the se- This paper describes an experimental
quence of application of the various program that was undertaken to evalu-
load levels in the block is approximately
ate the effects of various load levels in a
random. More commonly, however, a
given aircraft spectrum and to examine
programmed test spectrum is used in
the capabilities of several cumulative
which all of the cycles at each load level
in the block are consecutively applied fatigue damage theories for this evalua-
before the load is changed to the next tion. Spectrum fatigue tests of built-up
level. beam specimens were carried out as
The number of consecutive blocks re- well as several series of fatigue tests
quired to produce failure of the test involving one, two, or three load levels.
MORDFIN AND H A L S E Y ON FATIGUE T E S T S OF B E A M SPECIMENS 253
limit load factor was arbitrarily taken as 2-in. strips, all of 7075-T6 aluminum
7.00, which approximates the values com- alloy. A f by |-in. steel splice bar sepa-
monly used for fighter planes and advanced rates the two channels and is symmetri-
trainers. Thus, for this investigation, the cally positioned with respect to the
\-g load is equal to 4, or 14.3 per cent of the median plane of the specimen. The
limit load.
splice bar diffuses the shear load applied
Symbols: at the center of the specimen; it adds
d, constant in Liu and Corten's damage little to the bending strength. The
theory, channels and the splice bar are joined
K, constant in Valluri's damage theory, with i^-in. steel aircraft bolts spaced at
N, cycles to failure, 3-in. intervals. The strips are fastened
S, load level, to the channel flanges with 3^-in. brazier
Se, fatigue limit.
head rivets of 2024-T31 aluminum alloy.
Si , highest load level in spectrum,
A 1-in. rivet spacing is used throughout
/3, constant in Munse's damage theory
except in the last 3^ in. at either end of
SPECIMENS the specimen, in which a |-in. spacing is
In an investigation at this laboratory used.
several years ago (1),^ fatigue tests were The beam specimens were fabricated
in accordance with conventional air-
^ The boldface numbers in parentheses refer
to the Ust of references appended to this paper. craft manufacturing practices by the
254 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
lieved to have been fabricated from dif- simply supported at its ends and loaded
ferent batches of materials. Specimens transversely at its center by a pneumatic
from the three shipments were desig- cylinder, B.
nated 101 to 140, 201 to 220, and 301 to A close-up of the specimen under a
305, respectively. small applied load is given in Fig. 4.
MORDFIN AND HAtSEY ON F A T I G U E T E S T S OF B E A M SPECIMENS 255
End fixtures, C, are fastened to the one, / , a hollow tubular member hav-
specimens with j^-in. bolts, and simple ing two wire-resistance strain gages
support is achieved by means of bearings, mounted on its reduced section, serves as
D, and flexure plates, £, which provide the load-sensing element for the spectrum
rotational and translational freedom, programmer. The two dynamometers,
respectively. The flexure plates intro- one for controlling and one for reading
duce a longitudinal stress of 100 psi into loads, are isolated electrically from
the specimen when a transverse load of one another so that independent checks
5000 lb is applied. The effective length on the control system can be made.
of the beam specimen, from bearing to Cyclic loads were applied at rates of
bearing, is 53 in. about 15 to 40 cpm, depending on the
100-120 PSI
EXHAUST
minimum loads that are zero or negative, are opened as needed by signals from
regulating valves 10 and 11 are set for the the spectrum programmer.
respective pressures and valves 6 and 7 The exhaust line from valve 12 leads
TABLE II.—RESULTS OF BASIC CONSTANT-LOAD-AMPLITUDE FATIGUE TESTS.
Loads cycled between 14.3 per cent of limit load and indicated load levels.
Load Level, per Cycles to Failure Deviation from
Test Series Specimen cent of limit load Mean, per cent
105 80 5 955 -0
B-2. 110 80 6 660 6
Geometric mean 80 6 310
104 60 11 915 0
B-3. HI 60 14 672 23
211 60 9 765 -18
Geometric mean 60 11 900
106 40 87 967 3
B-4. 112 40 83 538 -3
Geometric mean 40 85 700
120
100
Cycles to foilure
FIG. 6.—Basic Constant-Load-Amplitude S-N Curve for Beam Specimens. Minimum Load:
14.3 Per Cent of Limit Load.
258 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
outdoors and is muffled to reduce noise. circuit and a counting circuit. The
The hand pump connected to the lower bridge is the heart of the unit. Its opera-
end of the cyHnder is used to raise the tion, described in reference (3), is based
piston for mounting and removing test on the action of dynamometer, / , Fig. 4.
specimens. The two strain gages on this dynamome-
Static tests were conducted by opening ter are connected into the bridge. The
valve 5 and gradually increasing the pres- bridge is deliberately unbalanced at zero
sure on the top side of the piston with load so that it becomes balanced when
regulator valve 9 while simultaneously the desired load is reached. This null
reading the load. condition triggers a sensitive switching
6000
D , CD D
o Toble II
4000
• • 4D
3000
2000
1000
FIG. 7.—Basic Fatigue Properties of the Beam Specimens as Reported by Several Investigators;
Minimum Loads Between 529 and 780 lb.
spectrum to the next involves the opera- for each shipment by applying a 1.5
tion of valves 6, 7, and 8, Fig. 5, plus the factor of safety to the measured static
selection of the proper potentiometers in strength values (Table I).
the bridge circuit. This function is auto-
BASIC CONSTANT-LOAD-AMPLITUDE
matically controlled by a six-channel pre-
FATIGUE PROPERTIES
determining counter which is equipped
with automatic reset. Constant-load-amplitude fatigue tests
The primary weakness of the program- were conducted to establish the basic
mer is the dependence of the bridge fatigue properties of the beam specimens.
circuit on the exact values of the re- In each test the load was cycled between
Test Series Specimen Load Level, per Cycles to Failure Deviation from
cent of limit load Mean, per cent
115 60 27 970
P-3. 116 60 24 054 -7
Geometric m e a n 60 25 900
sistances involved. Slight variations in a designated load level and 14.3 per
these resistances occur as a result of cent of limit load. The results of these
normal variations in room temperature. tests are shown in Table II with the
Therefore, the potentiometer settings percentage deviations from the geometric
have to be readjusted slightly to compen- mean at each load level. The results for
sate for these variations about three positive load levels are also plotted in
times a day during a test. By making Fig. 6.
these readjustments, the load levels Figure 7 compares these results with
were maintained constant within 12 lb. results obtained by Howard and Katz
(1) and by the Aeronautical Structures
STATIC TESTS
Laboratory (4) on nominally identical
Static tests were made on two speci- specimens. This comparison is based on
mens from each of the three specimen load, in pounds, rather than on per cent
shipments. Limit load was calculated of limit load. The figure shows that the
260 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
data from the three investigations are reduction in fatigue life is obtained.
consistent despite differences in testing These conclusions were based on uni-
technique and variations in static directional bending tests of unnotched
strength. The fatigue testing machine dad 2024-T3 sheet.
used by Howard and Katz was a deflec- Templin (6) then showed that much
tion-controlled machine rather than a greater increases in life are obtained as a
load-controlled machine as was used in result of prestressing if the specimens
the other two investigations. Static (7075-T6 rod) are notched.
strengths varied from 6615 to 7316 lb, Finally, Payne (7) applied preloads to
and the minimum loads of the fatigue 2024-T3 airplane wings prior to fatigue
cycles varied from 529 to 780 lb. testing and obtained remarkable in-
In view of this consistency, and the creases in fatigue life.
TABLE IV.—RESULTS OF CONSTANT-LOAD-AMPLITUDE FATIGUE TESTS WITH
PERIODIC OVERLOADS TO 100 PER CENT OF LIMIT LOAD.
Loads cycled between 14.3 per cent of limit load and indicated load levels.
Load Level,
Test Series Specimen per cent of Interval, cycles Fatigue Cycles to Deviation from
Failure Mean, per cent
limit load
fact that the fatigue properties of the Tests were conducted in the present
beam specimens are similar to those of investigation to evaluate the effect of
full-scale aircraft structures (Fig. 1), preloading on the basic fatigue proper-
it appears that this specimen design has ties of the beam specimens. Each speci-
merit for studying the fatigue behavior men was subjected to one cycle of 100
of aircraft structures. per cent of limit load immediately prior
to the start of a constant-load-amplitude
EFFECTS OF PRELOAD fatigue test. In the fatigue test the load
Bennett and Baker (5) showed that the was cycled between a preselected load
application of a prestress prior to fatigue level and 14.3 per cent of limit load. The
testing produces an increase in fatigue results of these tests are given in Table
life providing the prestress is con- III together with the results for test
siderably greater than the maximum series B-1 (from Table II), in which the
stress of the fatigue cycle. However, if first load cycle applied to each specimen
the prestress is only slightly larger than is considered a preload.
the maximum fatigue stress, a slight The effects of preloading may be eval-
MORDFIN AND H A L S E Y ON F A T I G U E T E S T S OF BEAM SPECIMENS 261
uated by comparing Tables II and III. load was applied immediately prior to
These effects are in agreement with those the start of the fatigue test, and addi-
found by the other investigators. Fur- tional single overloads were applied at
thermore, the comparison shows that periodic intervals throughout the fatigue
when the preload is applied in a direc- test. The interval was approximately
tion opposite to the maximum fatigue equal to 10 per cent of the basic fatigue
load, a significant reduction of the fatigue life of the specimen at the fatigue load
life takes place. Bennett and Baker (s) level under consideration. The fatigue
and Templin (6) obtained similar results. loads were cycled between the preselected
load levels and 14.3 per cent of limit load.
EFFECTS OF PERIODIC OVERLOADS
Table IV lists the fatigue load levels,
Heywood (8) attributed the effects of the intervals between overloads, the
prestressing to the introduction of com- number of fatigue cycles to failure, the
Load Level,
Test Series Specimen per cent of Interval, cycles Fatigue Cycles
to Failure
Deviation from
Mean, per cent
limit load
100
90
80
v^
a
o
'^^^
^ 70
\ x^^ Overlooded
" 60
\ >.
C
u ^ N^v
fc.
o. \\ x
\
Bosic \ . Prelooded
30
10'
10*
Cycles to failure
FIG. 8.—Comparison of S-N Curves for Beam Specimens Subjected to Various Treatments.
Minimum Load: 14.3 Per Cent of Limit Load.
f 100
85
3
17 Spectrum Specimen
Blocks
to
Deviation
from
1 Failure Mean,
70 65 per cent
55 172
i [ 128 361 35
1 100 3
1 129 220 -18
85 17 130 242 -10
70 65 Geometric m e a n 268
55 172 i
i 40 283
f 134 141 -13
f
I
100 3 r 139 170 40
85 17
' 1
4 220 86 -29
70 65 Geometric m e a n 121
55 172 i
1 0 1000
-14.3 280
f 138 129 34
I 5 304
305
68
102
-29
6
100 3 Geometric m e a n 96
85 17 i
5 , 70 65 r 131 193 -13
55 172
-14.3 280
6 J 132 282 27
133 202 -9
-28.6 3 Geometric m e a n 222
i
i 115 1»
100 3
6 85 17 craft during a typical 20-hr interval of
70 65 its service life. This spectrum is now
55 172
considered obsolete.
" Prior to start of test. The spectrum fatigue tests were used
to study the effects of the various steps
II to V. Also shown in the figure are the in the spectrum in order to determine
maximum scatter observed in life (test which of them must be included in spec-
series P-4, Table III) and the maximum trum fatigue tests that are intended to
uncertainty in load (6 per cent). For provide estimates of aircraft service life.
most of the test data the scatter and The load levels and the cycles per
the uncertainty were considerably block at each load level for the six spec-
smaller than these maximum values. trums used are given in Table VII. Load
264 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
was cycled between the designated load 2 except that it includes still another
levels and 14.3 per cent of limit load. load level, 25 per cent of limit load.
Within each block the load levels were Table VIII shows that the mean life
applied in descending order from the under spectrum 3 is further reduced to
highest to the lowest (or most negative) only 49 per cent of the mean life under
level. Some effects of this sequence of spectrum 1. When this result is viewed
load application are discussed in Ap- in terms of the scatter in the test data,
pendix I.' however, it appears that the additional
The results of the spectrum tests are decrement in life may be a manifestation
given in Table VIII. Each specimen of the scatter rather than a real attribute
failed at 100 per cent of limit load within of the 25 per cent of limit load step.
the first three cycles of the block fol- Additional test data would be required
lowing the completed number of blocks to make this determination with cer-
to failure shown in the table. tainty. Even if the 25 per cent of limit
load step does reduce the spectrum fa-
EFFECTS OF VARIOUS SPECTRUM tigue life, the degree of reduction may
LOAD LEVELS not be sufficiently serious to warrant
The desire to eliminate unnecessary the additional testing time required by
load levels from a test spectrum is based its inclusion in the test spectrum.
upon the need to minimize the time re-
quired for testing. The apparatus re- Zero and Negative Load Levels:
quired to conduct programmed tests Table VII shows that spectrums 4
involving, say, eight load levels is not and 5 are the same as spectrum 1 except
significantly more complex nor more that they include load cycles to zero
costly than that needed for four load and negative load levels. Table VIII
levels. shows that the inclusion of these steps
The basic test spectrum (spectrum 1, reduced the life of the beam specimens
Table VII) consists of four load levels to 45 and 36 per cent of the mean life
ranging from 100 to 55 per cent of limit under spectrum 1.
load; the mean fatigue life of the beam If these results are considered together
specimens under this spectrum was 268 with the results of the periodic over-
blocks. loading tests, it is seen that when high
positive load levels and low negative
The 40 and 25 Per Cent of Limit Load load levels are both present in a test
Levels: spectrum, each acts to reduce the fatigue
Spectrum 2 is identical with spectrum life under the other. In test series 0-5
1 except that it includes one more load (Table IV) occasional loadings to 100
level, 40 per cent of limit load. The mean per cent of limit load reduced the fa-
fatigue life under spectrum 2 was found tigue life at —28.6 per cent of limit load
to be 162 blocks, 60 per cent of the life by introducing detrimental residual ten-
under spectrum 1. This indicates that sile stresses into the subsequent failure
the 40 per cent of limit load step is surface. Under spectrums 4 and 5 the
rather damaging and that its omission regular application of load cycles to
from the test spectrum results in un- small negative load levels reduced the
realistically high estimates of service life. fatigue life under high positive loads,
Spectrum 3 is identical with spectrum presumably by causing the beneficial
residual compressive stresses, which were
' See p . 270. introduced into the subsequent failure
MoRDriN AND HALSEY ON FATIGUE TESTS or BEAM SPECIMENS 265
surface by the high positive loads, to de- The relative closeness of the lifetime
cay more rapidly. reductions obtained with spectrums 4
The latter result is similar to that and 5 suggests that the major portion of
which is obtained with gust-load spec- the damage done by the zero and nega-
trums and with ground-to-air-to-ground tive load levels is due to the —14.3 per
cycles, which include zero and negative cent of limit load level, which is common
load levels. Wallgren (11), for example, to both spectrums. However, the test
showed that when positive and negative results are inadequate to confirm this
gust loads below the nominal fatigue premise.
FIG. 9.—Typical Failures of Beam Specimens Tested at (o) Higher Load Levels and (6) Lower
Load Levels.
limit are added to an otherwise complete The 115 Per Cent of Limit Load Preload:
maneuver and gust loads spectrum, the
fatigue life is reduced by about 20 per The effect of a preload of 115 per cent
cent. Schijve and Jacobs (12) believe that of limit load on the fatigue life under
this effect is, on the average, closer to the basic 4-level spectrum may be evalu-
50 per cent. Payne (7) showed that the ated by comparing the results for spec-
addition of periodic negative loads re- trums 1 and 6. It is seen that the preload
duced spectrum fatigue life significantly. reduced the mean fatigue life from 268
These results support the recommen- to 222 blocks (17 per cent); this reduc-
dation made by Lundberg and Eggwirtz tion loses meaning, however, when com-
(13) that, although maneuver loads are pared with the variations in fatigue life
the most important for fighter aircraft, obtained from specimen to specimen.
gust loads should be included in the Swartz and Rosenfeld (14) tested full-
test spectrum. scale 7075-T6 aircraft structures under
266 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
a similar 4-level spectrum with and loading, (joing a step further, Schijve
without a preload to 115 per cent of and Jacobs (12) feel that high loads oc-
limit load. Their results show a 16.5 curring with a frequency of less than
per cent increase in fatigue life with the about ten in the anticipated life should
preload, but here, too, the variations in not be included in the test spectrum.
lifetime from specimen to specimen make
this conclusion questionable. DESCRIPTION OF FAILURES
These findings are in agreement with Typical failed specimens are shown
the effects of preloading discussed earlier. in Fig. 9. Failures at the higher load
The 4-level spectrum is essentially a levels (150, 100, and 80 per cent of limit
maneuver-loads spectrum which consists load) were similar to that of Fig. 9(a),
of load levels that are relatively high while those at the lower load levels were
compared with the preload. Under these similar to that of Fig. 9(6). Nearly all
failures occurred at the center of the
TABLE IX.—COMPARISON OF TEST specimen or at one rivet spacing away
RESULTS WITH P R E D I C T E D FATIGUE
LIVES UNDER SPECTRUMS 1 AND 2.
from the center.
Frequent inspections usually did not
Blocks to reveal the presence of fatigue cracks
Failure
Reference prior to failure. In the few cases where
Spec- Spec- cracks were detected, at least 93 per cent
trum 1 trum 2
of the ultimate life had elapsed.
Test results Table VIII 268 162 COMPARISON OF TEST RESULTS WITH
Linear rule 43 37 PREDICTED FATIGUE LIVES
Tangent (2) 60 60
M u n s e et al (17) 23 21 Numerous theories of cumulative fa-
Gatts (18) 39 37 tigue damage have been proposed. Sev-
Valluri (19) 48 44
66 60
eral of these were used to predict the
Liu-Corten (20)
Henry (21) 64 62 fatigue lives of the beam specimens un-
Smith (22) 76 71 der spectrums 1 and 2. Some details of
Modified Liu-
Corten Appendix I I 181 178
the calculations involved in making the
predictions are discussed in Appendix
11.^ Comparisons of the predicted lives
conditions preloading has a relatively with the test results are given in Table
small effect; however, if the spectrum IX.
contained lower load levels, of the order Insufficient data were available for
of gust loads, or if a higher preload had predictions based on the theories of
been employed, a distinct improvement Richart and Newmark (23), Marco and
in life probably would have been ob- Starkey (24), and Freudenthal and Heller
served as a result of preloading. (25). The method proposed by Manson,
Nicole (15) has expressed the convic- Nachtigall, and Freche (26) is inappli-
tion that the demonstration of sufficient cable.
fatigue resistance should be made with- With the exception of the modified
out the benefits of preloading. Shanley Liu-Corten method, all of the theories
(16) recommended that any beneficial are highly conservative with respect to
loading which is expected to occur but spectrums 1 and 2. Even more significant
once in the life of an aircraft should be is the fact that none of the theories, in-
ignored because of the possibility that
certain aircraft will not experience this ' See p. 270.
MORDFIN AND H A L S E Y ON F A T I G U E T E S T S OF B E A M SPECIMENS 267
eluding the modified Liu-Corten, prop- amplitude fatigue properties of the beam
erly accounts for the damage of the 40 specimens:
per cent of limit load step, which is the 1. Preloading at 100 per cent of limit
difference between spectrums 1 and 2. load improved the fatigue life of the
The test results show a reduction in life specimens at fatigue load levels of 60
of 40 per cent as a result of this step, per cent of limit load or less, provided
while the best prediction of this reduc- the preload was applied in the same di-
tion (linear rule) is only 14 per cent. rection as the subsequent fatigue loads.
These findings should not be sur- The effect of this preloading was negli-
prising. Most damage theories do not gible or slightly detrimental to fatigue
properly account for the beneficial ef- life under load levels of 80 per cent of
fects of the residual compressive stresses limit load or higher. When the preload
that are introduced by high positive load and the fatigue loads were applied in
levels. Therefore, the theories are con- opposite directions a reduction in life
servative when applied to spectrums 1
resulted.
and 2, which consist solely of high posi-
2. Periodic overloading at 100 per cent
tive load levels.
of limit load affected the fatigue prop-
As suggested earlier, the addition of
erties in the same way as preloading did,
negative load levels to a spectrum causes
only more so. Specifically, under condi-
the residual stresses to decay and thereby
reduces their beneficial effects. Hence, tions in which preloading was beneficial,
the theories might be expected to pro- periodic overloading was more beneficial,
vide more accurate life predictions when and where preloading was ineffective or
applied to spectrums consisting of both detrimental, so was periodic overloading.
positive and negative load levels. The 3. Periodic underloading at 25 per
drawback is that most theories, including cent of limit load with a relatively small
the Liu-Corten, do not contain provision number of cycles produced no significant
for handling both positive and negative changes in the fatigue life under fatigue
load levels. loads applied in the same direction as
These considerations suggest that, al- the underloading.
though a theory may give a safe pre- Spectrum fatigue tests were carried
diction for fatigue life under spectrum out on the beam specimens using several
loading, it is not necessarily capable of variations of the comprehensive spec-
evaluating the relative effects of the in- trum in Table VI. The test results justify
dividual load levels in the spectrum. the following conclusions regarding the
necessity of including all of the load
CONCLUSIONS levels in spectrum fatigue tests which
The bending fatigue properties of a are intended to provide estimates of
7075-T6 built-up beam specimen were service life.
found to bear similarities to the fatigue 1. The 40 per cent of limit load step
properties of certain typical full-scale caused significant damage, and its in-
aircraft structures. Also, the fatigue be- clusion in the test spectrum is necessary.
havior of the beam specimen is notably 2. The 25 per cent of limit load step
insensitive to minor variations in testing was less damaging, and under certain
technique and material. conditions its omission from the test
The test results warrant the following spectrum may be justified.
conclusions regarding the constant-load- 3. The zero and negative load levels
268 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
APPENDIX I
EFFECTS OF LOADING SEQUENCE IN PROGRAMMED SPECTRUM
FATIGUE TESTS
In what sequence should the load levels order. Schijve and Jacobs (12) found the
in each block of a test spectrum be applied descending sequence to be more damaging
in order to simulate service fatigue most than the ascending sequence. Freudenthal
closely? Probably a random sequence is best. and Heller (28) reached the same conclusion
However, limitations of testing equipment and also showed that both of these sequences
frequently require that a programmed were more damaging than their random
sequence be used. In this case, the problem sequences. On the other hand, Naumann,
is reduced to selecting a programmed se- Hardrath, and Guthrie (29) found the ascend-
quence that produces fatigue damage at a ing sequence more damaging than the de-
rate approximately equal to the rate at which scending one, with their random sequences
a random sequence produces damage. being intermediately damaging. All of these
Selecting a programmed sequence that investigators worked with gust-load spec-
meets this requirment is extremely complex, trums and specimens fabricated from 7075-
if not impossible, with the present state of T6 aluminum alloy. The most likely con-
knowledge. Approaching this problem on an clusion is that the disparity in their results
analytical basis, Rosenthal (27) examined the stemmed from differences in their spectrums
and specimen designs.
residual stresses that are introduced by the
high load levels of the spectrum and the Payne (7), Kuhn (30), and others have
decay of these stresses at the other load expressed the viewpoint that whatever the
levels. His findings indicate that the desired effect of sequence may be, it would become
smaller as the block size is reduced. Un-
sequence depends on the exact spectrum and
fortunately the available facts do not sup-
test specimen under consideration.
port this. The test results reported by
This unhappy conclusion has inadvert- Schijve and Jacobs (12), Naumann, Hard-
ently been confirmed experimentally. Con- rath, and Guthrie (29), and Gassner (31) show
sider, for example, the extreme types of no systematic variation of the sequence
programmed sequences, namely, those in effect with block size.
which the loads in each block are applied A sizeable effort, both analytical and ex-
in ascending order of magnitude and those perimental, would be required to establish a
in which they are applied in descending practical understanding of the sequence
MORDFIN AND H A L S E Y ON F A T I G U E TESTS OF B E A M SPECIMENS 269
40 60
Blocks to failure
FIG. 10.—Effect of Assumed Value of Fatigue Limit on Predicted Life Under Spectrum 1 Accord-
ing to Valluri (19).
FIG. 11.—Effect of Assumed Value of d on Predicted Fatigue Life Under Spectrum 1 According
to Liu and Corten (20).
270 SYMPOSIUM ON FATIGUE OF AIRCRAFT STRUCTURES
APPENDIX II
CALCULATION OF PREDICTED FATIGUE LIVES
These comments refer to the predictions where:
of fatigue life given in Table IX. The follow-
ing interpolations of Fig. 6 were used in 5 = load level,
applying the linear damage rule. .S'l = the highest load level in spectrum,
Sg = the endurance limit, and
K = a. constant.
Load Level, per cent Fatigue Life,
of limit load cycles In the calculations this term was set equal
85. 5 200 to unity, resulting in predicated fatigue lives
70. 8 300 which are closer to the actual ones.
55. 15 600 The Liu-Corten method (20) contains a
constant, d, which, according to its de-
The following interpolations of the pre- velopers, ". . . may be obtained most simply
loaded curve in Fig. 8 were used in applying from a two-stress-level repeated block
Smith's method (22). fatigue experiment of the component or
full-size structure." Accordingly, the con-
Load Level, per cent Fatigue Life, stant, d, was computed from the results of
of limit load cycles
test series 0-2, 0-3, and U-1, and produced
85 4 350 the following values:
70 17 000
55 43 000 Test Series
0-2 4.17
The constant fi for Munse's method (17) 0-3 8.25
U-1 1.58
was calculated to be 0.51 for spectrum 1 and Average. 4.67
0.56 for spectrum 2.
The methods of Gatts (18), Valluri (19), A value of 4.7 was used to predict the fa-
and Henry (21) require use of a fatigue limit. tigue lives shown in Table IX for the Liu-
The effect of fatigue limit on the fatigue life Corten method.
under spectrum 1 as predicted by one of The feeling has been expressed, however,
these methods (19) is shown in Fig. 10. It is that a two-stress-level experiment may not
seen that the prediction improves as the provide the best value of d for spectrum
fatigue limit increases. While it is doubtful fatigue life predictions. In fact, Swartz and
that a true fatigue limit exists for the beam Rosenfeld (32) showed that a value of 8.25
specimens, a value of 35 per cent of limit gave reasonable predictions for a variety of
load was used on the basis of Fig. 8. full-scale aircraft structures under spectrum
The damage function in Valluri's method loadings. The predictions shown in Table I X
(19) contains the following term among for the modified Liu-Corten method were
others: calculated by using a value of 8.25 instead of
4.7 for d.
5, - S, The variation of the predicted fatigue
In life with d is given in Fig. 11 for spectrum 1.
K
For a predicted life of 268 blocks under
5 - Se spectrum \, d = 10.3; with this value the
In
K predicted life under spectrum 2 is 266 blocks.
MORDFIN AND H A L S E Y ON F A T I G U E T E S T S OF BEAM SPECIMENS 271
REFERENCES
(1) D. M. Howard and S. Katz, "Repeated (14) R. P. Swartz and M. S. Rosenfeld, "The
Load Tests of Aircraft Wing Beam Speci- Effect of Preloading on the Variable
mens Under Bending and Bending-Torsion Amplitude Fatigue Characteristics of a
Loads," Nat. Bureau of Standards, un- Slab Horizontal Tail for a Typical Fighter
published report. Airplane," Naval Air Material Center,
(2) ARTC/W-76 Aircraft Structural Fatigue Report No. NAMATCEN-ASL-W23, Part
Panel, "Aircraft Fatigue Handbook, Vol. III, Jan. 18, 1961.
II—Design and Analysis," Aircraft In- (15) W. Nicole, "Some Results of Fatigue Tests
dustries Assn., Washington, D. C. (1957). with Parts of Vital Importance of the
(3) L. Mordfin and R, L. Bloss, "Simple Ground Attacker P-16," in Full-Scale
Strain Gauge-Based Load Controller," Fatigue Testing of Aircraft Structures,
Review oj Scientific Instruments, Vol. 33, edited by Plantema and Schijve, Pergamon
No. 7, p. 772 (1962). Press, New York, N. Y., p. 364 (1961)
(4) M. S. Rosenfeld, private communication. (16) V. R. Shanley, "Discussion of Methods of
(5) J. A. Bennett and J. L. Baker, "Effects of Fatigue Analysis," Proceedings, Symposium
Prior Static and Dynamic Stresses on the on Fatigue of Aircraft Structures, WADC
Fatigue Strength of Aluminum Alloys," TR 59-507, p. 182, Aug., 1959.
Journal of Research, Nat. Bureau of (17) W. H. Munse, J. R. Fuller, and K. S.
Standards, Vol. 45, No. 6, p. 449 (1950). Petersen, "Cumulative Damage in Struc-
(6) R. L. Templin, "Fatigue of Aluminum," tural Joints," Bulletin 544, Am. Railway
Proceedings, Am. Soc. Testing Mats., Vol. Engineering Assn., June-July, 1958, p. 67.
54, p. 641 (1954). (18) R. R. Gatts, "Application of a Cumulative
(7) A. O. Payne, "Determination of the Damage Concept to Fati~ue," Transactions,
Fatigue Resistance of Aircraft Wings by .Am. Soc. Mechanic-1 Engrs., Vol. 83,
Full-Scale Testing," in Full-Scale Fatigue Series D, p. 527 (1961).
Testing of Aircraft Structures, edited by (19) S. R. Valluri, "A U.^ified Engineering
Plantema and Schijve, Pergamon Press, Theory of Hi. h Stress Level Fatigue,"
New York, N. Y., p. 76 (1961). Aerospace Engineering, Vol. 20, No. 10,
(8) R. B. Heywood, "The Influence of Pre- pp. 18, 19, 68-69, Oct., 1961.
Loading on the Fatigue Life of Aircraft (20) H. W. Liu and H. T. Corten, "Fatigue
Components and Structures," Current Damage Under Varying Stress Ampli-
Paper No. 232, Aeronautical Research tudes," NASA TN D-647, Nov., 1960.
Council (London) (1956). (21) D. L. Henry, "A Theory of Fatigue-
(9) T. J. Dolan, "Concepts of Fatigue Damage Damage Accumulation in Steel," Transac-
in Metals," in Fatigue, Am. Soc. Metals, tions, Am. Soc. Mechanical Engrs., Vol.
p. 1 (1953). 77, p. 913 (1955).
(10) G. M. Sinclair, "An Investigation of the (22) C. R. Smith, "Fatigue-Service Life Pre-
Coaxing Efifect in Fatigue of Metals," diction Based on Tests at Constant Stress
Proceedings, Am. Soc. Testing Mats., Vol. Levels," Proceedings, Soc. Experimental
52, p. 743 (1952). Stress Analysis, Vol. XVI, No. 1, p. 9
(11) G. Wallgren, "Fatigue Tests with Stress (1958).
Cycles of Varying Amplitude," Aero- (23) F. E. Richart, Jr. and N. M. Newmark,
nautical Research Institute of Sweden, "An Hypothesis for the Determination of
Report No. 28 (1949). Cumulative Damage in Fatigue," Pro-
(12) J. Schijve and F. A. Jacobs, "Program- ceedings, Am. Soc. Testing Mats., Vol. 48
Fatigue Tests on Notched Light Alloy p. 766 (1948).
Specimens of 2024 and 7075 Material," (24) S. M. Marco and W. L. Starkey, "A Con-
Reports and Transactions, Nat. Aero- cept of Fatigue Damage," Transactions,
nautical Research Inst. (Amsterdam), Vol. Am. Soc. Mechanical Engrs., Vol. 76,
XXIV (1960). p. 627 (1954).
(13) B. Lundberg and S. Eggwirtz, "The Re-
lationship Between Load Spectra and (25) A. M. Freudenthal and R. A. Heller, "On
Fatigue Life," in Fatigue in Aircraft Stress Interaction in Fatigue and a Cumu-
Structures, edited by A. M. Freudenthal, lative Damage Rule," Journal of the
Academic Press, New York, N. Y., p. 255 Aero/Space Sciences, Vol. 26, No. 7, p. 431
(1956). (1959).
272 DISCUSSION ON FATIGUE T E S T S OF BEAM SPECIMENS
DISCUSSION