Professional Documents
Culture Documents
Fatigue Analysis Method For LNG Membrane Tank Details: M. Huther, F. Benoit and J. Poudret, Bureau Veritas, Paris, France
Fatigue Analysis Method For LNG Membrane Tank Details: M. Huther, F. Benoit and J. Poudret, Bureau Veritas, Paris, France
0
. One WorldTradeCenler,
Suite1369,New York,N.Y.1W48
$’ ‘%
~ PaPer,
t.bePresented
atExtreme
Load,
Re,Pm,eSYrn
POsium
8 : Arlington,
VA,October
1>20,1981
?> oil! ,.
%, ~,o
% .,%**
ABSTRACT
scch calculations allow the determina-
This paper gives a description of a tion of stress le”el,s, their global Q=
fatigue analysis method based on the S-N local origin, whether they are static cm
curve and crack propagation method. dynamic, or if they are due to waves .
89
Propagation laws of limited cracks two different materials are linked
in 2 D or 3 D remain difficult to esta– together.
blish and requires experimental data
which are not available for such thin The hull transmitted loads implicate
plates . hull stress calculations.
TWO types of loads exist :
To use a crack propagation method,
it is also necessary to define the size the static component due to the
of the initial crack and to verify the loading of the ship,
initiation period, i.e. the time
necessary under cyclic load, to create a the dynamic component due to waves.
sharp crack. This is small, “ith respect
to propagation period, i.e. the time The static component is computed by
necessary under cyclic load, to propagate classical naval architecture progr.?mnes
a sharp crack until perforation of the {3).
element.
The dynamic component requires the
In practice, the determination of calculation of ship behaviour in regular
the initial crack length and verification waves, in irregular waves and then the
of the initiation period length is made development of a long term statistical
by means of laboratory tests, the same analysis { 4, 5 }. In this way, it is
tests that are “sea to determine S-N possible to calculate the boundary loads
curve. of the tanks . Finite element calc”latio”s
Taking into account the above, we apPlied to tank components permit the
decided to use the classical S-N curve determination of the resulting stresses
method , with the help of crack pr-opaga– in the most loaded areas.
tion theory to determine for example the
fatigue limit under random loads or the The same finite element calculation
influence of static stresses on fatigue permits the determination of the global
life, Of a sample submitted to cyclic effect of the cooling down between ambient
loada . temperature a.”d LNG temperature (- 160°C)
LmD$. SmE%5.
‘—-- In some cases, local finite element
calculation might be necessary to deter–
mine local thermal stresses due to
different contraction properties of
different materials.
.
of failure (Fiq”re 3) . This curve will
be the design curve.
‘q’”I ~n~~.w~..b”flm
L....
~,
““”% >.,
‘\
‘\
,.,,,
\
/..
;>.m“
.,97.,
cur”.
;>tib;l,ty
Fig. 4. Scheme of a long term ~txeSS
distribution and step by step
representation for Miner SLIM
calculation
91
Two types of cyclic stresses can Therefore one can calculate the
be considered. correcting factor Ki by :
The first one is the cycling of
thermal stresses between ambient l-a Ri
~-1/3 (7)
temperature and operational temperature. Ki=(
l–a R~
In this case, the maximum stress
is equal to the service condition stress where :
and the minimum stress is equal to 0.
Ri is the R value at step (i)
The range is :
Au=o R = aiaw– OS
(2) (8)
gth + ‘lth 1
C!i
Ow+ OS
This range is constant for each
cycle and the number of cycle n] is and RSis the R value of the S–N curve
equal to the number of tank surveys used.
during the ship life. The value of R
is 0. Therefore, if the S-N C“r”e has The step by step curve, at step (i)
been established for a “al”e R which is defines a number of CYCleSni . The stCeSs
different, the Calculation of N, require.? level Ki Au on the S-N curve leads to
a correction of Au by multiplication a number of cycles Ni .
with a factor K.
Therefore, the Miner sum components
The crack propaqatio” theory shows become
that K can be expressed as follo”s
{6,7}: “M
Z2.
Ni
!3LE
c’
(ui Ki~ ni (9)
K=(l. aFJ~3 (3)
where : where :
a is a constant which can be determined M is defined by the fatigue limit of
by tests. the S–N curve.
The Miner sum component is there– But this limit, A al, is not the
fore equal to :
experimental one.
n, n,(K Au )m
(4) If one considers the life of a
N,=C welded joint for example, the experimen–
tal fatigue limit corresponds to the
For wane stress, which is the level Aul, below which the initial defect
second cyclic type, it is necessary to crack a. does not propagate.
consider the total range 2 0 t without
taking into account any possibility of Now, during cumulative damage on
compressive “dues. This rule is the ship, this initial crack of length
necessary in order to take into account
the existing residual stresses after ao , will propagate under the various
stress levels. At a given time, it will
welding. Essentially it has been fo””d be a, and that time, it can be increased
that in complex welded structures, when by the level Au ~ . This “alue is,
submitted to cycling loads, compressive
stresses are as damaging as tensile therefore, no longer the fatigue limit.
stresses {8 }. It is then necessary for assessment of
cumulative damage, to determine another
The long term statistical fatigue limit of the S-N Curve. A
distribution of the wane stresses, is practical example will be presented
represented by a step by step curve as later.
shown in Figure 4. A step level ( i) is
defined by the stress le”el to maximum 3 LOAD AND STRESS C!ALCULATIONS
stress at 10–’ probability ratio (a: )
In order to present the method, the
NOW, for the “a”e stresses, at case of the invar membrane Gaz Transport
step i , a range is found : system for LNG carriers is considered.
—-
92
Next the short term distribution of
this bending moment for irregular sea-
states is calculated. This distribution
is defined by the spectral energy of the
response bending moment and the coeffi-
cient of the corresponding Rayleigh
distribution {4, 5) . This calculation is
done for various sea headings, normal 1y
four angles from head sea to following
sea, and for various sea-states defined
by the mean period and a significant
wave height equal to unity. The periods
are taken from observation tables. The
North Atlantic observation or the World
Wide Tables can be used {lo, 11) .
93
,.
}i
T T T T T
./
,..
. ............ .
_—.
94
l’ Since the studied connections were
lapped joints, it was ?.ss”rnedthat the
slope of all S-N curves would remain the
same.
95
In order to save time, Locati tests This probability leads to a distance
were performed. between design S-N curve and mean S–N
curve of 2 s.
For Locati tests, samples are
submitted to a series of cyclic loads, A confirmation of this assumption can
increased step by step until failure. be also found in various studies {12, 13] .
In case of cumulative damage, studies
Each step is limited to a maximum {12, 13} and regulations {15) show that
number of cycles, 100,000 for the with the S–N Curve at 2 s below the mean
concerned tests. The interpretation has S–N curve a Miner sum equal to 1 is an
been done by means of the Miner SUM and acceptable standard to assure that there
expressed in terms of log c of the s.N is minimal risk of failure.
curve. An example of results is given
in Figure 11. Rut , the tests are performed on a
m.s,l..or- limited number of samples.
, e8- Therefore the distance calculation
between the design curve and the mean
.- , ,3 ,, , , , curve can be improved by statistical
m m (l@a)3,.6 4,,7 5,., 6,., 6,., 73., ,7.6 a., laws. It is well known that the mean
value follows a Student Law and the
standard deviation a X2 Law. Takinu into
-m m -m c
m . . . account a 95 % confidence limit on-the
- . ,.13 estimation and the 98 % probability of
, .Cu2 n.,, ,4.W
failure, the distance is expressed by
m3
4 , B 58cm 2,.05 1,.43
{14} :
6 5 6 .@m n.,, 14.24
7 6 , 8,m 20.4, ,4.,,
~ = ~ t(.95, n-1)
mm, 4 , ‘mm 7.4C ,3.,, [ K
5 7, cm 12.7, ,4. u ‘o(”’8)=-
6
5 4 5 ,,cm 6.C6 ,3.7s (16)-
Gr, 5 6 35 m ,0.68 1,.03 where :
, 7 65 032 ,,. m ,,.2s
, 7 8 ,0 cm 22.5, 1,.35
t student law
4 4 , 33Cu3 ,.m ,3.7, ‘$ normal law
5 4 5 70m 7.4C 13.87 ~z X2 law
—
w :me .- ofCYdee
for . cdl step
i.,m,ccc nz number of samples
%
It is then possible to calculate the For the pipe membrane connection
corresponding crack length for z given the re.s”lt.sare :
stress range :
Miner wave sum :
K =yl?l G (17) .00006
th
The crack propagation theory shows Miner full thermal :
that the coefficient y for the studied .00o7
disposition can be taken equal to 1.0 Miner partial thermal :
{16} . .0017
number of tank surveys, two per Fatigue tests can also be limited
year. if it is possible to find enough data
In 20 years, n = 40 (fUll thermal) o“ S-N curves for similar material and
number of voyagez, cycles between joints . In particular, the constant ‘,m”
loaded (partial thermal) and of the equation of the S–N curve can be
ballast conditions n = 380 determined so long as it is a characters
(from I.M. C.O.) tic of the material. The tests are,
therefore, only necessary to determine c
wave $tres,s as given by Fiqure 4 and the scatter, which are characteris-
design s-N curve at 3 s below tics of the joint type, the weld filler
the mean experimental cur”e. material, the welding procedure and the
reference stress used for calculations.
For the corner connections, the
::und value.s are as follow, the The method is general and can be
<Tresses are given as mean stresses in easily used by designers, but it always
:?.eweld : shows some inaccuracy and needs to be
improved throuqh research work, in
. wave max. stress : particular concerning the fatigue limit
48.1 MPa (7.0 Ksi) in random cumulative fatigue calculation
thermal stress , and the probability levels for design
293. o MPa (42.5 Ksi) S-N curve determination.
k.- _- _____
REFERENCES {11} Hoqben N. , Lumb F.E.
‘-Oceanwave statistics”
{ 1} Koj i Harhda and al National Physical Laboratory,
,,
Statlstlcal estimation of tank London 1967
stress by total system analysis
for spherical tank LNG carrier,’ {12] Maddox S.J.
The Hitachi zosen Technical ‘CA fracture mechanics approach
Review, vol. 39, No 2, 1978 to service load fatigue in “elded
structures ,’
{2} Huther M. and al Welding Research International,
,,
Bureau Veritas and gas carriers Vol. 4, No 2, 1974
Rules and future developments
Structural analysis and design {13} Gmmev T.R.
principles - Free and forced ,,Fati~ue d~sign roles for welded
vibrations on gas carriers” steel joints,’
Gastech 1975, Oct. 1975 Welding Institute Research Bulletin
Vol . 17, May 1976
{3} Montoya J.
“Le Bureau Veritas et le calcul {14} Ventsel H.
‘$lectronique - Application h ,,
Th&, rle des probabilit6s,’
1-architecture navale et a“x Editions MIR, 1973
Calculs de structure,’ Bulletin Tech-
nique Bureau Veritas Ott. 1972 {15} API RP2A
{4} Planeix J.M. ,,Ap
I recommended practice fOr
,!waveloads – A correlation planing designing and construct-
between calculations and meas”re– ing fixed offshore platforms,’
ments at sea,’ American Petroleum Institute,
International Shipbuilding llth edition, Jan. 1980
Progress, Aug. 1972
{16} Bathias C. , Baillon J.P.
{5} Planeix J.M. , H“ther M, Huard G. ,,Lafatigue des mat4riaux et des
,,
Statl~tlcal a“d pseudo-statistical structures’,
calculations for liquefied gas Collection Universit& de Cmnpi&gne
carriers’, Editions Maloine and Presse
Shipbuilding and Shipping Record Universitaire de Montreal, 1980
Second LNG Transportation Conference
Oct. 1973
{7} Maddox S .J .
,,
The effect of mean stress on
fatigue crack propagation .
A literature re”iew”
International Journal of Fracture,
Vol. 11 No 3 June 1975
_—
98
Nlwn a design is verified it is
APPENDIX
assumed that tbe scantling is safe if
for an adequate S-N curve the Miner sum
CUMULATIVE FATIGUE DAMAGE
remains below a“ associated value.
when a welded joint is exposed to
a cyclic load, cyclic stresses and strains As an example, the mean S–N curve
are induced in the weld. is associated with a maximum Miner sum
equal to 0.3 {1] , or a s-N curve at
Experiments show that failure can 2 standard deviation belo” the mean
curve is associated with a maximum Miner
ke observed for stress levels. below the
=terial yield strength, if enough sum of 1.0 {2) .
~cles are imposed.
WAVE STRESS DAMAGE
The number of cycles which are
The stresses induced by “a”e bending
zecessary to failure is a function of
moment are random and are defined by
zke stress range level.
their long term distribution (Figure 6
Therefore, the resistance to of the Paper) . This curve gives the
?clic loads, i .e., fatigue resistance, number of cycles which exceed a give”
:s defined by the relationship between level .
stress range and number of cycle.? at
Such a cur”. can be represented by
:ailure.
a step by step function (Figure 4 of the
This relationship is noted as the P*Per) . Each step is defined by :
5-N curve and for steel materials it can
Oi stress range
x expressed by the following formula :
ni number of cycles
~m for R
N=C (1) Ri value of R
. .he
re
!3 stress range M total number of steps
Therefore N i is given by :
Ni =~ (3)
(Kioi)m
99
Because the scatter of fatigue results REFERENCES
the calculation of c for a given typical
weld joint requires various tests, from {1} Harris and Creede
which the mean value of C and the “Shock and Vibration Handbook”
standard deviation of log C are Edition Mac Graw Hill, 1961
calculated.
{2} Maddox S.J.
“A fracture mechanics approach
to. service load fatigue in
welded structures”
Welding Research International,
Vol. 4, No 2, ,1974
_.
100