,@C.1w,.

THE SOCIETY OF NAVAL ARCHITECTS AND MARINE ENGINEERS

>e One World Trade Center, Suite 1369, New York, N,Y, 10048
>?
*3 PaPer$ to be presented at Extreme load. R,SPOS, Symposium
~
: Arliston, VA, October 19-20, 1981
;+
co .
5
% ,*L . ,*,,+\+
@

Fatigue Criteria for Ship Structure Details

W. H. Munse, University of Illinois, U;bana, IL

ABSTWCT (s~ax)N = The maximum stress range of

a variable Ioadig history
Fatigue criteria are presented for that is expec ed to produce
the design of ship structures. l%e failure in 10 ~ cycles
criteria take into account the ship
s tructure details, the fatigue resis- = Mean constant-cycle stress
tance of specific locations in these N
range for failure at N cycles.
details, the variable loading to which
a ship is subj ected, and the desired Io8 = The maximum stress range ex-
level of reliability (factor of safety) pected to occur once in 108
A design example indicates the simple cycles, based o a Weibull
manner in which the criteria can be distribution.
useti.
w = Characteristic value of S
NOMENCLATURE (Equation 5)

log c The life intercept of the = Scatter factor (Equation 11)

S-N curve L
r = Gamma func tio
Probability density function-
fs(s)
Weibull distribution = Mean s tress of the Weibull
(Equation 2) us
distribution (Equation 3)

k Shape parameter for the E = Random load factor (Equation

Weibull distribution 7)

L(n) The probability of m = Standard ieviation of the

failure through a life, n 5s
Weibull distribution.
(Equation 9)
= Total uncertainty in fatigue
N
m Negative slope of S-N curve life (coefficient of
variation)
N The mean life necessary to
produce a useful life n with INTRODUCTION
a reliability of L(n)
Fatigue cracking in ships has been
(PF) Probability of failure a serious problem for many years As
= [l-L(n)] noted by Vedeler (1)1 in 1962, ship-
builders in Norway and Sweden cons id-
Reliability factor ered the problem of fatigue in ships
F
(Equation 13) to be of more practical importance for
ordinary ships than the question of
s Stress range brittle fracture, He noted that fa-
tigue cracks were often found in the
Constant-cycle fatigue forepeak region, bottom amidships, at
SD
design stress range for a the bulwark at both ends of the bridge,
useful life n and reliability and in the hatch corners In a recent
L(n) stuciy it has been observed that ships
may also have cracks at crossings of
(sD)all = The maximum allowable frames, longitudinal, and girders ,
fatigue stress range and many other locations (2, 3 and 4)
(Equation 15)
1
Reference numbers are indicated
in parentheses.
&

23.1
Since such cracks may be possible Fatigue Behavior of Weldments
points of initiation for catastrophic
failures and, since the repair of fa- During the nearly 50 years of
tigue cracks can be very costly, it is laboratory studies conducted on weld-
essential that fatigue be given ade- ments , numerous papers , conference
quate consideration i the design of proceedings and books have been pub-
ship structures lished wherein detailed fatigue data
for welds and weldments may be found
In this paper criteria are pz-e- (5-11) Recently, much of this infor-
sented for the fatigue design of ship mation has been re-examined to estab.
structural details along with a brief lish basic S-N relationships for
discussion of the principal parameters numerous types of welded members and
included in the criteria. 2 An example details (12) Nearlv 1500 S-N curves
of the application of these design have beeri p;oduced, &ne example of
crireria is then presented to indicate which is presented in Fig. 1. (This
the simple manner in which the criteria is the S-N curve for axially loaded lon-
can be used to provide a verification gitudinal fdl penetration groove
of the adequacy of a ship in fatigue. welds with the reinforcement intact;
based on stress range; and for mild,
FATIGUE DESIGN PARAMETERS high strength low alloy or quenched
and tempered steels. Identified by
Fatigue design generally consists DAAAXB)
of verifying that the details of a
structure have sufficient resistance The solid line mean regression
to repeated loading to provide a fa- curve (50% reliability) of Fig. 1, as
tigue life equal to or greater than established by a least-square analysis ,
required: a wrifica, tion by checking can be given by,
Drocess To achieve this for shiu
;tructure details , criteria have ~een Log N=log C-mlog SN (1)
deeloped which takes into account,
(a) the basic fatigue behavior of ~;
welded structural details, (b) the or SN = (N) (la)
tYPes Of details found in ship struc-
tures at which cracking has been ob- where,
served, (c) the loading histories to log c = the life intercept of the
which ships may be subjected, and (d) S-N curve
the level of fatigue safety to be in-
cluded in the ship design. N= number of cycles to failure
for a constant cycle stress
There are other factors that can range of SN,
affect the fatigue behavior of a struc-
ture also ; however, their importance m= the negatiw slope of the
or effect is not as great as the effect S-N curve
o f those noted above and to include
them would have greatly complicated the constant-cycle stress
N =
the overall design process. These range for failure at N
neglected factors include, (a) the cycies.
mean stress effect, an effect that is
relatively small for welded details Using such straight-line relation-
and is now neglected in most fatigue ships , the mean fati ue stress ranges
design specifications , (b) the tensile for lives of 105, 10~, 107, 108 cycles,
strength of the steel, another factor have been established for numerous
that is generally neglected in struc- welded members and details , and are
tural fatigue design specifications presented in Table 13, These mean fa-
and considered to have relatively tigue strengths are for the numerous
little effect in long life fatieue. structural fatigue details shown in
and (c) tempera ture, -rate of lo~d Fig. 2.
application, residual stresses, and
size effect, again factors that are For each of the details shown in
Eenerally found to be of secondary Fig. 2, the location at hich the fa-
Zmport anie tigue strength applies is the point
where the greatest *tress concentration
exists and, except as noted, ca be

3
Under random loadings straight-
2 line S-N relationships are found to
The ship design criteria presented extend well beyond the fatigue limits
herein were developed in an investi- often reported for constant cycle
gation conducted at the University of tests (10) Therefore: in this devel-
Illinois and sponsored by tbe Snip opment, the straight-line S-N relation-
Structure Committee, ship has been assumed to extend to 108
cycles or more.

232
 200 1 I 1 1 1 1 ! Ill [ 1 1 1 1 1 1 111 1 1 ) 1 1 1 [ Iu J 1 1
1 , 1 I ++~
1 LB,,? TGLFF7CNU L , !, s,,. 5..,, K

(OJ 9,,: CO.r)[)[kct LCWL -

m .
60 m.,, .
=-. -
-- =-
-- =---

-.
- --------
._ . ..

4-

1
++tHtH-++ +-++-}+ H
, 5 , 10,!, : , , 5,,,,

CYCLES TO FF!lLL!RE,
IN TH12U5RNC5
Fig. 1. S-N Relationship for a Weldment Containing a Longitudinal Groove
weld in the As-Welded Condition, (Detail No. 3)
considered to be a function of the prin- been identified and will provide gui-
cipal tensile stress at that location. dance in locating the details for which
This location, for example, is at the possible fatigue failure should be
end of the cover plate for detail No. 5, checked (22) AII example of a few of
is at the toe of the butt weld for de- these families and configurations of
tail No. 10, and is at the side of the details is presented in Fig. 3.
hole in detail No. 28.
The locations in Fig. 3 at which
The data in Table I and the dia- fatigue cracking might develop are cir-
grams of Fig. 2 provide the basic fa- cled and identified by the correspond-
tigue information on which the fatigue ~g structural detail number from Fig,
design criteria presented herein are (The basic fatizue resistance for
based. the detail is provi~ed in Table 1)

Ship Details Not all details in Fig, 2 are found

in ships and not all of the details
After the basic fatigue data had used in the ships surveyed hae exhibi-
been assenbled, the second phase in ted cracks Only the thirty-seven de-
the development of tbe fatigue design tails listed in Table III are those for
procedure was the identification of which cracking was reported and then,
those ship locations at which fatigue some of the details exhibited only a
crackine mizht occur. Two recent re- very limited number of cracks Those
ports o; th~ in-service performance of for which the largest number of cracks
structural ship details (3, 4) have were found are those for which these
served to define possible fatigue criti - fstigue design criteria will be of
cal locations in &hips These-reports svceatest value, However. in design a
catalog and define the types and loca- ?atige evalwition shodd be made-of
tions of details at which failures all details for which fatigue cracking
have occured in a variety of merchant is a possibility,
and Naval vessels.
Ship Loadings
A total of 86 ships were included
in the surveys The details examined To properly evaluate the fatigue
were separated into twelve general adequacy of a ship detail requires a
families (see Table II) and these were realistic estimate of the cvclic stress
in turn divided into 56 zrouDs of 634 history to which the detail-will be sub-
separate configurations. - A ~otal of jected during its lifetime. Since the
6856 failures were found in the 607584 existing full-scale ship loading data
details observed. In the investigation are for a limited length of time. it
on which this paper is based, the loca- has been necessary to-extrapolate the
tions in each of these confi~urations available data to obtain an estimated
at which fatigue might devel~p have lifetime spectrum (13 , 14) Such

233
 TASLE I
Mean Fatigue Stress Range for Fatigue Details in Fig. 2

(Constant Cycle - 0,50 Reliability)

Detalil S-N Stress Range, ksi, for n cycles2

Slope,
(seYIJig, 2) m ~ = 105 ~ = 106 = 107 ~ = 108

1 5.73 69.4 46,5 31,1 20.8

1 (F) 4.80 67.1 41.5 25.7 15.9
2 6.05 61,5 42,0 28,7 19.62
3 5.77 44.1 29,6 19.9 13,33
3 (G) 5.94 41.2 27.9 19.0 12.87
4 6.08 41.4 28,3 19,4 13.30
5 3.25 26,4 13.02 6.42 3.16
6 6,08 41.4 28.3 19.41 13.30
4.11 39,8 22.71 12.97 7.41
6.54 55,8 39,2 27,6 19.4
9 9.64 32.6 25,7 20.2 15.92
9(s) 8.85 48 37 28.5 22.0
10 7,44 39.9 29,3 21.5 15,76
1O(G) 9.32 47.2 36.9 28.8 22.49
11 6.13 33.1 22.73 15,62 10.73
11(G) 6.65 29.4 22.5 17.26 13.23
12(G) 5.66 40,8 27,2 18,09 12.05
12 3.98 35,0 19.6 11.00 6.17
13 4.23 48.3 28.0 16.27 9.44
14 7.43 40,6 29,8 21.8 16,03
15 3.48 25,98 13.40 6,91 3.57
4.63 32,8 19.93 12,12 7.37
:! (G) 6.97 32.8 23,6 16.94 12.17
3.73 27,8 15,00 8.10 4.37
17(s) 9.52 28,9 22.7 17.81 13.99
18 4.03 20.30 11.46 6.47 3.65
18(s) 9.22 25.7 20.02 15.60 12,15
19 7.49 23.1 17,00 12.49 9.18
19(s) 7.53 27,5 20,28 14.93 10,99
20 3.94 32.9 18,36 10.23 5.70
20(s) 6.44 28.02 19.60 13.71 9.59
21 3.94 (:;;:) (~M:36) (;;:;3) (5,70)
21(s) 7.36 16.59
22 3.15 39,8 19,16 9,22 4,43
23 3.26 35.7 17.6 8,68 4,28
24 3.26 35.7 17.6 8,68 4.28
25 7.09 33.2 24.0 17.36 12.54
26A 8.53 49.9 38,1 29.1 22.20
25B 3.95 (;::;) (;:::;) (M:j:) (::~:)
26 3,46
27 4.85 22.8 14,17 8.81 5.46
27(S) 4.48 22,8 13.83 8,16 4.88
28 7.74 40,1 29.8 22,11 16.42
28(F) 4.81 (29.4) (18.21) (11,28) (6,99)
29
;;#)

29R2
30 2,83 38~0 16,83 7,46 3,31

234
 TABLE I (CONTINUED)

Detail S-N Stress Range, ksi, for n cycles2

No .1 Slope,
(See Fig. 2) m ~ = 106 7 ~ = 108
n = 105 n=10

30A (::;;) (34.o) (23.5) (16.1) (:;:;:]

31 (20,16) (11.87) (6.99)
;;A
32B
3;48
3.53
(;::g:) (;:i;y) (:;:;) (:;:;)
32c
3.66 21;34 11~38 6;07 3;23
::(s) 10.39 25,5 20,45 16.38 13.12
34 3.66 (21.34) (11.38) (6.07) (3.23)
34(s) 10.39 (:::;) (;;:$:) (1::::) (1::l..)
35 3.55
36 3.95 34.9 19.51 10,89 6.08
36A 3.95 34.9 19,51 10.89 6.08
37
37s
38 3.71 32- 17.2 9;24 4:97
:; (s) 10.23 16.27 13,00 10.37 8.28
39A
39B
39C
40 3.53 (21~50) (11:21) (5;84) (3:04)
41
42 5.02 110,2 69;8 44,08 27;88
43
43A
44
45
46 4:35 (20~16) (11~87) (6:99) (4112)
47
47A
48

48R
49
50
51
52
53

1 (S) Indicates shear stress on fasteners or welds

(F) Indicates flame cut surfaces.

(G) Indicates the surfaces have been ground flush

Estimated values are shown in parentheses ( )

A dash is provided where no data are available

235
 -==2- 29( F)-29
29RI (Radi.3 = l14~,to1/2)
29 R2(Rodi. s ,1/2 Iol )
c
w. )

3;(s)-37

C===D> 30A

38( S)-38

32
1 ,, -,
7/- L~ #

c c
1 3
e=
&&J-&.
f3==J-yp-
13

Fig. 2, Structural Fatigue - Details

236 _ .
 390 39
-... ... .
7 26
38 37 26
20,21
T T
(a) Family No. I(AI) (b) Family No,l(B4)

W_
\ ; 33
25
33s
7
b
25B 2130
Zf(s)
(c) Family No, 2 (c2) (d) Family No. 3( AI)
42
7
36 36
f

m
49 III

(e) Family No,8 (A3) (f) Family No, II(AIO)

Fig. 3. Examples of Configurations in
ship family details (3,4)

45A TABLE III

Details Exhibiting Failures i Ships

Total Total
n ct&Jl, Detail No. o Detail No. of
. No. Crack & Cracks
7 272 32B 2
*+ 9 7 33 36
L@ 51 14
17
7 33s 20
2 34 23
17s 2 34s 17
19 42 36 600
19s 40

k- -#-- 20 318 37 462

r-l
L..
52

/ 21
21s
1300
54
38
40
41
2
11
8

F
26 155 42 7

c3cJ- 47
28
28F
208
222
43
44
75
14
29 9 47 29
/ 29R
53 3
(G) - Designates a ground surface
(S) - Designates shear on weld or 29F 7 48R 25
fastener 50 2
(F) Designates flame-cut edges f.ai-
30 142 51 687
comparison with machined edges
30A 672 52 105
Fig,2, Structural Fatigue-Details (cent
53 8
i--
237

1
 ~BLE II

Summary of Data for 12 Detail Families (Ref. 4)

Totals Observed

?amily Detail Family Total No, Total No, z

No, Name Details Failures Failures

1 Beam Bracket 68,586 2,252 3.2B

2 Tripping Bracket 34,o12 1,587 4.67
3 Non-Tight Collar 20,974 33 0,16
4 Tight Collar 20,654 46 0.22
5 Gunwale Connect ion 172 5 2.91
6 Knife Edges o 0

7 Miscellaneous Cutouts 296,689 853 0.29

8 Clearance Cutouts 57,307 843 1.47

9 Deck Cutouts 7,534 29 0.38

10 Stanchion Ends 7,090 122 1.72
11 Stiffener Ends 40,729 298 0.73
12 Panel Stiffeners 53,827 788 1,46

Totals 607,584 6,856 1,13

)

238
extrapolations , plotted on a semi-log
cumulative distribution basis for large 35 >.
tankers and dry cargo vessels are pre- ...
sented in Fig, 4 and are based on the 30
wave- induced longitudinal bending F
stresses. A complete loading history
should include also the high-frequency
dynamic loadings However, since these
are generally of a relatively small
stress-range and would produce little
damage they have been neglected in the
present study. (The somewhat conserva-
tive previous assumption that the S-N
curve is linear to 108 cycles tends to
compensate for this neglect of the high- 5
frequency stresses. ) Nevertheless, if
it can be demonstrated that the stress t
ranges for the expected high-frequency % f 765432[ o
loadings will not be small (above about - LOG
6 ksi), then they should be included in
(a ) Service Stresses In Large Tankers. (13)
the total loading history,

To use the fatigue design criteria

developed herein requires that the load-
ing history be represented in probabil-
istic terms by a probability distribu-
tion func tion. A variety of distribu-
tion functions were investigated,
including the Beta, Lognormal, Weibull,
Exponential and Sayleigh distributions;
however, the evaluation clearly indica-
ted that the Weibull distribution
would most effectively define existing
ship loading data,

The two parameter Weibull probabil-

ity distribution function can provide
many shapes to model the loading his-
tory, including those of the Exponential
and Sayleigh distributions , and is given
by: .
exp [-(~)k] (2) Q(x > Xj) TOTAL PROBABILITY OF EXCEEDING Xl
fs(s) = $ ($k-1
(b 1 ServiceStressesIn Dry Cargo Vessels,(14)
where:
k = shape parameter Fig. 4. Long-Term Trends in Service
Stresses for Large Tankers and Dry
w = characteristic value of s Cargo Vessels (13,14)
s = stress

The mea stress of the distributicm is ,

us
=wr(l+ l/k) (3)
and the standard deviation is ,

w [r(l + Z/k) - rz(l + l/k) ]1f2 (4)

s =

where;
r = Ganuna function

The general configurations of sevei--

al Weibull distributions (various values
of k) are shown in Fig, 5. These are
similar to the distributions often re-
ported for strain measurements made on
ships at sea, An indication of the ex-
cellent fit of a Weibull distribution
with k = 1,2 and the data obtained from Stress, S
Sea-Lands SL-7 shipboard measurements
(15) is presented in Fig. 6. Fig. 5. General Shapes of Weibull
Distributions

239
 The above procedure has been fol-
30 lowed to establish values of k for a
variety of ships for which loading his-
tOry data ~re available.
~elg~en .Ln Table V and ra~g~
These values
from 0.7
Tbe shapes of the functions
20 are presented in Fig, 7 and are similar
to those for the ships shown in Fig. 4.
It should be noted that the larger ships
(tankers and bulk carriers) tend to have
loading shape parameters equal to or
10 greater than 1.0.

For the Weibull distribution func-

tions, the maximum stress-range expected
tO be reached once in a ship, s lifetime
00 of 108 cycles (estimated to be a 20 year
10 20 30 40
life) can be determined from the follow-
s ing equation:
(Max. Peak-to-Trough Stress, kei)
S~08 = w [ln (52000)]1/k (6)
Fig. 6. SL-7 Scratch Gage Data
with Corresponding Iieihull Dis- The predicted values of S~08 for the
tribution (16) ships for which loading history data
are available are give in Table V and
To obtain tieibull distributions range from 10 ksi to 34.1 ksi.
that represent ship loading stress his-
tories, the mean, the standard devia- Random Load Factor
tion and the coefficient of variacian
for the distributions are made equal to In the fatigue design procedure
the corresponding values for the ship presented herein tbe constant-cycle
fatigue stress range in Eq. 1 must be
data. First, the appropriate value of
modified to account for the random load-
che Weibull distribution parameter k
ing represented by the Weibull distribu-
can be obtained from Table IV by making
tion functions This can be done readily
the coefficient of variation for the
Weibull distribution (standard devia- and simply by introducing a random load
factor, c (16).
tionfmean) equal to the coefficient of
variation fcm the his trograms, Next,
the value of parameter w is obtained The random load factor is based on
from the following: the assumption that the stress range
history can be modeled by the two-para-
us meter Weibull distribution function
. (5) with a maximum stress range equal to
r(l + I/k)
S108 and a minim~ stress range equal
to zero and is given by:
TABLE IV
C = (lnN)llk [r(l + m/k)] -l/m (7)
Table of Weibull Shape Parameter
Values and Corresponding This factor, as developed by Ang and
Coefficients of Variation Munse (16) , is based on the ap lication
of Miners linear damage hypot I!
C?SIS to
the random loading represented by the
Weibull Shape Coefficient of Weibull probability density function of
Parameter Variation Eq. 2. A tabulation of the random load
k 6 factors for various values of k and m
are presented in Table VI for a life of
0,5 2,236 108 cycles. (A comparable table could
0,6 1.758 readily be prepared for any other life. )
0.7 1,462
0,8 1,261 Using the random stress factor from
0.9 1,113 Eq. 7 (Table VI) (a factor that corre-
1.0 1,000 sponds to a specific Weibull distribu-
1.1 0.910 tion and S -N curve slope) , and the mean
1.2 0.837 fatigue stress range from Table I for
1.3 0.776 specific detail, one can readily obtain
1,4 0,724 the maximum stress of the variable load-
.
1.5 0.679 ing that can be expected to produce
1,6 0,640 failure at 108 cycles This maximum
1.7 0.605 stress is :
1.8 0.575
1.9 0.547 (smax)N = SN E
2,0 0.523

p.-
240
 I.c

0,8

0,6
_s_
s.
0.4

0.2

0
-( -b -> -4 -3 -2 -1 0
Log [1- FS(S)]

Fig. 7. Ship Loading Histories Nodeled by Weibull Distributions

TABLE V

Ship Loading His tories Compared With Weibull Distributions

Stress Change at
Weibull Probability of
Type of Load EXCe dance
Ship Name of Ship NOtes Shape, k = 10 5 ,S108 (ksi)

Dry Cargo Wolverine State 1.2 16,5

California State l~; 1.0 18,0
Mormacs Can 1,5,7 1.3 12.0
Mormacscan 1,5,6 1.0 10,0

Large Tankers Idemitsu Mar 2,5 1.0 12.3

R. G. Follis 2,5 0.8 30.0
Esso Malaysia 2,5 0.8 21.8
Universe Ireland 2,3,5, 0.7 18,7

Bulk Carrier Fotini L. 2,5 0,9 29,5

SL- 7 Container- See Note 9 4,6,9 1.2 34.1

ships

Data from ref. (14)

Data from ref. (20)
Data from ref. (21)
Data from ref. (15)
Load history is for wave-induced loading with
dynamic effects filtered.
Load history is for wave-induced loading with
dynamic ef feet included,
Load history based on North Atlantic voyages
Load history based on ,South American voyages.
Load history hased on data collected from eight
SL-7 containerships

241


TABLE VI

Random Load Factors

Random Load Factors , E

m k=O.7 k=O.8 k=O .9 k=l. O k=l.1 k=l,2 k=l,3

2.0 28.63 20.93 16,17 13.02 10.83 9.24 8.05

2,5 23,12 17.49 13.86 11.39 9.63 8.33 7.33
3.0 19.23 14.96 12.12 10.14 8.69 7.60 6.75
3.5 16.35 13.04 10.77 9.14 7.93 7.00 6.27
4.0 14.15 11.53 9.68 8.32 7,30 6.50 5.86
4.5 12.41 10.31 8.79 7.64 6.76 6,07 5.52
5.0 11.01 9.31 8.04 7.07 6.31 5.71 5.21
5,5 9.87 8.48 7.41 6,58 5.92 5.39 4,95
6.0 8.91 7.77 6.87 6.15 5.58 5.10 4.71
6.5 8.11 7.16 6.40 5.78 5.27 4,85 4.50
7.0 7.42 6.64 5.99 5.45 5.00 4,63 4.31
7.5 6..S3 6.18 5.62 5.16 4,76 4,43 4.14
8.0 6.31 5.78 5.30 4,89 4.54 4.24 3.98
8.5 5,86 5.42 5.01 4.66 4.35 4.08 3.84
9.0 5.46 5,10 4.75 4.44 4.17 3.92 3.71
9.5 5.11 4,81 4.52 4.25 4.00 3.78 3.59
10.0 4.79 4.55 4,30 4,07 3.85 3.65 3.48

Values are based on a life of 108 cycles For any other life N,
the values in this table would be multiplied by:

(,n N)ljk
(18,42)1/k

The term SN is the mean constant-cycle

stress range from Table I that would be
L(n)
expected to produce a fatigue failure -1,08
in the selected detail at N cycles = exp -[~ r(l + nN108)]QN
However, the value so determined is the { 1 (9)
mean value (50 percent level of reliabil - where:
ity) and must be modified by a reliabil- N = mean life to fatigue failure
ity factor (factor of safety) to obtain = total uncertainty in fatigue
an appropriate maximum design stress N
m
value.
R~bility Factor (Factor of Safety) The total uncertainty is a func-
tion of the uncertainly in the mean
The basic relationships for the fa- fatigue strength (Table I) , the uncer-
tigue reliability design criteria devel- tainty in the stress analysis, effects
oped herein have been presented by Ang
of fabrication, workmanship, corrosion,
(17) in his development of reliability etc. The uncertainty in the mean fa-
analysis for design. In this analysis tigue strength (standard deviation
it is recognized that the fatigue life divided by the mean) varies considerably
Of a structural detail is a random vari - from one detail to the next bc aver-
able and assumed that its distribution ages about 0,50 (from 0.2 to 0.8 depend-
can be represented by an approximation ing upon the detail, the amount of data,
to the Weibull distribution (18) h
etc. ) for the details in Table 1, Rela-
this basis , the reliability function tively little is kmwn concerning tbe
(the probability of no failure through other factors, Nevertheless, until
a life n) can be given by: better values can be established it is
recommended that an estimated value of
~ uncertainty of 0.80 be used.

242
 To utilize the reliability func- stress for a specified level of relia-
tion in design, the mean life N neces.
bility will be given by:
ary to produce a useful life n with a
reliability of L(n) is obtained from SD = (f) Urn
from the fol lowing: (12)
or
N = nyL
(lo) SD = + lf~ (L)h (lza)
where yL is tbe scatter factor and is L
where:
given by equation 11 (Fig. 8)
= constant cycle design stress
D
range for a useful life ~ and
1,08,
r(l+nN a specified reliability L(n)
~L =
1.08 (11) Designating the Ias t term of equation
(1.2a)as the Reliability Factor, , RF,
N
(PF)
where :
(PF) = the probability of failure
and equal to [1 - L(II)].

In Fig. 8 it can be seen that a 99 per- and the allowable constant-cycle design
cent level of reliability (one percent stress becomes,
probability of failure) , and an uncer-
tainty in life $2N = O.50, would require =SN. RF (14)
a design for a life N equal to about 8 D
times the expected useful life n. The value of RF for three levels of
reliability, various S-N curve slopes,
Under constant-cycle stress range, and uncertainties $2N equal to O.4, 0.6
from equation (la) , the required design or O.8 can be obtained from Table VII.

FATIGUE DES IGN

I04 With the relationships presented
1 1 above a simple fatigue verification or
5 \ I design procedure that takes into ac-
/ count the principal fatigue parameters
~j= r(l+akoa) can now be provided. Table VIII shows
the six steps of the procedure to be
(PP-w // followed to verify the adequacy of a
103 1 ~f=lo-> given ship strueture detail in fatigue.
/ / In step 1 the expected loading his-
5 /[ tory for the ship detail must be estab-
/ lished (the shape factor for the Weibull
// / distribution selected) Some guidance
Y~ can be obtained from Table V, a summary
of the limited data now available. As
/ ,/ more complete data are obtained concern-
102
/ , ing ship loading histories, the values
/ in Table V can be expanded, improved
5 /[ f I and updated to provide the designex
X[ with the best possible guidance.
r / /1
In the second step the ship details
at which the fatigue resistance should
be checked must be identified. The
/i / critical locations in the ship assem-
1/ r 1/ blies , such as shown in Fig, 3, can be
5 .
/ used to identify the critical details
/
in terms of the numerous details shown
// Y / in Fig. 2, (k extensive summary of de-
tails in the twelve families of Table
II are presented in the Reference 22
I
o 0.2 0.4 0.6 0.8 1,0 Report)
fin The third step is to obtain, for
the detail, the fatigue strength and
Fig. 8. Relationship Between the slope of the S-N curve (this comes from
Scatter Faccor and the Uncertainty Table I). The fourth step is to obtain
in Fatigue Life and Various Proba- the random load factor from Table VI,
bilities of Failure (17) and the fifth step is to select the

243
 TABLE VIII

Des ign Proceaure

1. Ship Loading
Distribution Choose a loading shape parameter k,
of the Weibull distribution.

2. Ship Detail Identify the number designation of

Catalogs the critical details (Figs 2 and 3)
,+ 1

SW curve slope, m, of detail

Nean fatigue stress-range,
for detail.
I

4. Design Find random load factor, C, based on

Y
Table 6 k-value and m-alue.

5. Design Find reliability factor, Rf, baSed on

Table 6 m-value and ~N-ale for desired leel
of reliability,

6. Design Equation Cmnpte allowable stress-range (SD)=ll,

(SD)all=SN. c .RF for probability of exceedance
of 10-8.
z

appropriate reliability factor from provides only for fatigue; in addition,

Table VII. The maximum allowable fa- the maximum s tress must not exceed the
tigue stress range, (sD)all 4, is rhen nominal permissible design stress pro-
obtained from the following equation vided by the basic design rules (19)
For detail number 7, at the toe of the
(SD) =SN. L.+ (15) stiffener weld in the web and flange,
the maximum allowable fatigue stress
This relationship, using Table VI, is range is 39.7 ksi for a 20 year life,
based on a desired life of 108 cycles. Similar calculaticms can readily be
For any other life the values, as noted made for details number 37 and 38.
on Table VI, would need to be modified,
SUNMARY AND CONCLUSIONS:
A design example for a beam bracket
with four structural details is presen- This paper summarizes a simple de-
ted in Fig. 9. In this instance, the sign procedure that has been developed
Weibull shape factor far the loading to provide for a fatigue strength veri-
history was taken as 1.0, the total c- fication in ship design, The criteria
ertainty in fatigue life as O.80 and provides for:
the desired level of reliability was
assumed to be equal to 90 percent. The (a) A large variety in ship strue -
resulting maximum allowable fatigws ture details.
stress range at the deck-bulkhead inter-
section (detail 39B) is found to be (b) The basic fatigue resistance of
31.8 ksi, This maximum stress-range the numerous elded details.
.1
4.
This maximum allowable stress (c) A distribution function that
range at the point in question, is the can be used to represent the
maximum peak-to-trough stress range ex- long life loading (1o8 cycles-
pected under the most severe sea state 20 years) for various types
and during the entire life of the struc- of ships
ture.

244
 Fig. 9. Design Example

Ship Assembly IA I - Beam Bracket (Fig, 3a)

Detail No. (S)]oe

(See Fig. 2) (See Table H

7 7,41 4.11 Reliability - 90%

39 B 5.9 Est. 4.0 Est. Coef, of Var. -0.80
37 5,0 Est. 3.7 Est, Weibull Dist, k = 1,0
38 4.97 3.71
*

~e - Detail 39B

(S)loa = 5.9 ksi Est. m = 4,0 Est. (Table 1)

Reliability Factor = 0.648 (Table 7)
Random Load Factor = 8.32 (Table 6)
Max. Allawable Stress Range = 5.9x 0,648x8.32 = 31.8 ksi

Considering Stiffener Detoil - Detail 7

(S)108= 741 ksi m =4.11

RF = 0.655 [ = 8.17
Max. Allowable Stress Range = 7.41 x 0.655 x 8.17 = 39.7 ksi

i-

245
 TABLE VII

Reliability Factors

I Reliability Factors, RF

I R = 90%
I
K = 95% R = 99%

m
N N N
0,40 0.60 0.80 0,40 0,60 0.80 0.40 0.60 0.80
I

2.0 .691 .546 .420 .608 .447 .320 .451 .281 .170
2.5 .744 .616 ,500 .671 .525 .402 .528 .362 .242
3.0 .782 .668 .561 .717 .585 .468 .588 .429 .307
3.5 .810 ,708 .609 .752 .631 .521 .634 .484 .363
4.0 .831 .739 .648 .780 ,669 ,566 .671 .530 .412
4.5 .849 .764 ,680 .8o1 ,699 .603 .702 .569 .455
5.0 .863 ,785 .707 .819 .725 .634 ,727 .602 .492
5.5 ,874 .602 .730 ,834 .746 .661 .748 .630 .525
6.o .884 .817 ,749 .847 .765 .684 .767 .655 .554
6.5 .893 .830 .766 .858 .781 .705 .782 .677 .580
7.0 .910 .841 .781 .867 .795 .722 .796 .696 .603
7.5 .906 ..s51 .794 .876 .807 .738 ,808 .713 .623
8.0 .912 .860 .805 .883 ,818 .752 .819 .728 .642
8.5 .917 .867 ,815 .889 .827 ,765 .829 .742 .659
9.0 .921 .874 .825 .895 .836 ,776 .838 ,754 .675

(d) A random loading factor that the author and not necessarily those
accounts for the randomness of the Advisory Committee of the Ship
of the loading during the life Structure Connnittee under whose guid-
of the structure. ante the investigation on which this
paper is based was conducted.
(e) A reliability factor (Factor of
Safety) that accounts for the KEFEKSNCES
many uncertainties that exist
1. Vedeler, G,,, To What Extent Do
The values of maximum allowable Brittle Fracture and Fatigue
fatigue stress range obtained in the Interest Shipbuilders Today,
design example provides an excellent Houdremont Lecture 1962,
calibration of the procedure and is Sveiseteknikk 1962, No. 3.
considered very reasonable, based on
the history of such details in the 2. GLasfeLd, R. , Jordan, D. , Ken, M. ,
ships at sea. Additional evaluations Jr. , and Zoner, D. Review of
now should be made of those details at Ship Structure Deta~Ls, SSC-266,
which fatigue failures have developed 1977.
to further evaluate and calibrate the
Drocedure. After the rmocedure has 3. Jordan, C. R. , Cochran, C. S. , In-
~een further verified, it should be Service Performance of Structural
possible also to use the procedure to Detail s, SSC-272, 1978,
develop relative fatigue ratings for
the manv details used for shiD struc - 4. Jordan, C. R. and Knight, L. T.,
tures. Further Survey of In-Service
Performance of Structural De-
tails, SSC-294, 1980.

The opinions herein are those of

246

 5 Munse, W. H. , Fatigue of Welded 17 !mg, A. H-S,, A Comprehend ive
Steel Structures, We ~ng Res earth Basis for Reliability Analysis
Council, New York (1964) and Design, ~, U. S, - Japan
Joint Seminar on Reliability
6 Gurney, T. R, , Fatigue of Welded
~pPrOach in Structural Engineer-
Structures, Cambridge U. Press , ~ng, Tokyo, Japan, May 19?4.
England i968)
18. Fredenthal, A, M, , !,
Prediction of
7 B,W.R.A, , Symposium on the Fatigue Fatigue Failure, ,,J. Appl phy~
Of welded structures, March 29. (Dec. 1960) ~, No, 12,2196-2198
April 1, 1960, Bri Ei~h welding J.
(March-Sept, 1960) 19. A, B,S., Rules for Building and
Classing Steel Vessels, American
8, Muse, W. H. , Stallmeyer, J. E. , Bureau of Shipping, 1979.
and Drew, F. P, , str~wr~l
Fatigue and Steel Railroad 20. Little, R. S. , Lewis, E. V, and
Bridges , PXOC. , AREA Seminar Bailey, F, C., A Statistical
(1968) Study of Wave Induced Bending
Mcments on Large Oceangoing Tank-
9. Symposium on Structural Fatigue, ers and Bulk Carriers , Trans.
J. of the Structural Div. , ASCE SNAME, 1971.
(Dec. 1965) M, No. ST12, 2663-
2797, 21. Lewis , E. V. and Zubaly, R. B. ,
Dynamic Loadings Due to Waves
10. The Welding Ins citute, pProceedings and Ship Motions, ,,SNAME, 1975,
of the Conference on Fatigue of
Welded Structures ,,,July 6.9, 22. Munse, W, H., Investigation at the
1970, The Welding Institute, Univeristy of Illinois - Study
Cambridge, England (1971) of Fatigue Characterization of
Fabricated Ship Details j Project
11 Pollard, B, and Cover, R, J. , SR-1257 , Ship Structure Committee
Fatigue of Steel Weldments, q
Weldin J. , American Welding Soc.
*) ~, No. 11, 544s-

u, Munse, W. H. , Pierce, R. C. , Martin,

S. A,, Maybe, L, , and Mahini, B, ,
Basic Material Fatigue Properties
for Use in Freight Car Fatigue
Analysis, Civil Engineering De-
partment, University of Illinois
at Urbana-Champaign, March 1981,

13. Bridge, I. C,, Structural Stress

in an Oil Tanker under Service
Conditions, ,,Trans , I.N,A, , p,
161, 1938,

14. Hoffman, D. and Lewis , E. v, ,

Analysis and Interpretation
of Full-Scale Data on Midship
Bending Stresses of Dry Cargo
Ships,, SSC-196, June 1969.

L5. Fain, R. A. and Booth, E, T. ,

Results of the First Five Data
Years of Extreme Stress Scratch
Gauge Data Collected Aboard Sea-
Lands SL-7 S, SSC-2g6, March
1979,

E. Aw, A. H-s and Munse, w. H.

Practical Reliability Basis for
Structural Fatigue, ,Preprint
No. 2494 presented at ASCE
Nat ional S trnctural Engineering
Conference, April 14-18, 1975,

247