You are on page 1of 6

ABS TECHNICAL PAPERS 2006

Proceedings of OMAE 2006


25th International Conference on Offshore Mechanics and Arctic Engineering
June 4 - 9, 2006, Hamburg, Germany

OMAE 2006 - 92268


LOW CYCLE FATIGUE ANALYSIS OF MARINE STRUCTURES

Xiaozhi Wang Joong-Kyoo Kang Yooil Kim Paul H. Wirsching


American Bureau of Shipping Daewoo Shipbuilding and Daewoo Shipbuilding and Aerospace and
16855 Northchase Drive Marine Engineering Co., LTD Marine Engineering Co., LTD Mechanical Engineering
Houston, TX 77060 1 Aju-dong, Geoje-si, 1 Aju-dong, Geoje-si, University of Arizona
USA Gyeongsangnam-do, 656-714 Gyeongsangnam-do, 656-714 Tucson, AZ 85721, USA
cwang@eagle.org KOREA KOREA phw@ame.arizona.edu
jgkang3@dsme.co.kr yikim@dsme.co.kr

Originally published by American Society of Mechanical Engineers (ASME), New York, NY,
and reprinted with their kind permission.

ABSTRACT In this paper, a literature review of material


behavior and strength of marine steels is first presented.
There are situations where a marine structure is Characteristic parameter values of cyclic stress-strain
subjected to stress cycles of such large magnitude that curve and strain-life curve are established based on the
small, but significant, parts of the structural component literature study, experimental testing and nonlinear
in question experiences cyclic plasticity. Welded joints FEA. An S-N curve is then proposed in order to define
are particularly vulnerable because of high local stress the fatigue strength in the low cycle regime. Finally, a
concentrations. Fatigue caused by oscillating strain in fatigue damage calculation method is developed based
the plastic range is called “low cycle fatigue”. Cycles to on a hot spot stress approach.
failure are typically below 104. Traditional welded joint
S-N curves do not describe the fatigue strength in the MATERIAL BEHAVIOR AND STRENGTH
low cycle region (< 104 number of cycles). Typical UNDER LOW CYCLE CYCLIC LOADS
Class Society Rules do not directly address the low
cycle fatigue problem. It is therefore the objective of For life prediction in welded joints, it is necessary
this paper to present a credible fatigue damage to define fatigue strength. For fatigue strength within
prediction method of welded joints in the low cycle low cycle regime, the general strain-life curve will be
fatigue regime. employed.
The general strain-life curve has the form, See
INTRODUCTION Dowling (1999),
ε a = ε ea + ε pa
Certain duty cycles associated with operations of a
The elastic strain-life curve is defined as
ship may produce oscillatory stresses whose magnitudes
exceed the yield strength of the material. For example, σ 'f
ε ea = ( 2 N f )b
the welded joints in certain members of tankers and E
FPSO’s during the loading/offloading process for which And the plastic strain-life curve is defined as
the total number of cycles during the service life is
expected to be less than 104. Fatigue associated with
ε pa = ε 'f ( 2 N f )c
cyclic plasticity (“low cycle fatigue”) must be
considered as a principal failure mode, yet the design S- where
N curves specified in typical class society rules are not Nf = cycles to failure
defined below 104 cycles. To perform a safety check for εa = notch strain amplitude
low cycle fatigue, it is necessary to define the S-N
curve, define the stress associated with the S-N curve ε ea = elastic notch strain amplitude
(here the hot spot stress is used), and the process by ε pa = plastic notch strain amplitude
which nominal stress is transformed to hot spot stress.
E = modulus of elasticity

Low Cycle Fatigue Analysis of Marine Structures 95


ABS TECHNICAL PAPERS 2006

σ 'f = fatigue strength coefficient This is convenient when dealing with both high and low
cycle fatigue as high cycle fatigue analysis is
b = fatigue strength exponent conventionally performed in terms of stress.
ε 'f = fatigue ductility coefficient
104
c = fatigue ductility exponent
Open circles: base metal, Grade A
Heo et al (2004)

Thus the total strain-life curve can be expressed as Solid circles: welded joint
Heo et al (2004)

σ 'f
( ) ( )

Pseudo Stress Range, SR, (MPa)


εa = 2 N f b + ε 'f 2 N f c
E Total strain-life
The strain-life curve is defined by the last five Elastic strain-life
SSC-346
SSC-346

1000
parameters of the list. Because E is well known, the
focus of this study will be on the last four. It is expected
that the fatigue crack will form in the heat affected zone
(HAZ) so that special attention will be given to the Plastic strain-life

HAZ. SSC-346

To obtain the strain-life curve, three approaches are


Cycles to Failure, N
possible: 100
- Direct measurement from testing. 100 1000 104 105

Unfortunately there is little data available in


the general literature, see Park and Figure 1: Strain-life curve
Lawrence(1998), and therefore other
methods may be employed Heo et al (2004) report the results of a fatigue test
- Use of published fatigue parameters of steels on 11 dogbone specimens of Grade A steel (base metal)
having similar monotonic properties. There having an upper bound yield and ultimate strength of
exists a large catalog of fatigue parameters 320 and 460 MPa. These data are plotted as the open
for a wide variety of steels. It is argued that circles in Figure 1. It should be noted that base metal
parameters for welded joints should be data is being compared to the HAZ curves of SSC-346.
similar to those of steels having roughly the Heo et al (2004) report the results of a fatigue test on 16
same monotonic properties welded specimens; non load carrying partially
- Empirical relationships for parameters based penetrated cruciform fillet welded joints. Stress-life data
on monotonic tests. Experience from is shown in Figure 1. Assuming that the fatigue cracks
extensive fatigue testing of steels has led to originated in the HAZ, this data is comparable to the
empirical forms in which the parameters can SSC-346 total strain-life curve. On the basis of this
be established from such monotonic data, it appears that the SSC-346 HAZ curves are non-
properties as ultimate strength and Brinnel conservative.
hardness While it is argued by Boardman (1982) and
Dowling (1999) that estimates of fatigue properties by
Park and Lawrence (1988) reporting in SSC-346 empirical forms should never be substituted for full
provide the strain-life parameters for HAZ and for weld scale testing of actual parts under service conditions,
material as relating to a specific detail. One of the the fact remains that these forms are useful in those
details consists of a center plate and two loading plates cases where data collection is impractical.
welded to the center plate by all around fillet welds. The Empirical relationships that will be useful are
base material is ASTM A-36. The Shielded Metal Arc summarized as follows:
Welding (SMAW) process and E7018 electrodes were - The fatigue strength coefficient is
used. Another series of tests were made on this approximately equal to the true fracture
cruciform joint using the Gas Metal Arc Welding stress from a tension test
(GMAW) process. The base metal was 12.7 mm plates
of ASTM A441 Grade 50 steel. σ ' ≈ σ~ f f
The Park and Lawrence parameters for HAZ - The fatigue ductility coefficient is
(SMAW; 12.7 mm plate) produce the strain-life curves approximately equal to the true fracture
shown in Figure 1. strain from a tension test
Note that the S-N curve of Figure 1 is given in ε ' ≈ ε~ f f
terms of pseudo stress. Low cycle fatigue involves
strain cycling and a strain-life relationship to define - Strain-life curves for a wide variety of
fatigue strength. However for engineering purposes it is engineering metals tend to all pass near the
useful to define pseudo stress range, SPR, as the modulus strain εa = 0.01 for a life Nf = 1000 cycles
of elasticity, E, times strain range, εR, - Strain-life curves for a wide variety of
S PR = Eε R engineering metals tend to all pass near the
strain εa = 0.01 for a life Nf = 1000 cycles

96 Low Cycle Fatigue Analysis of Marine Structures


ABS TECHNICAL PAPERS 2006

- Ultimate strength can be approximated from where, for this application, σ = hot spot stress, ε = hot
Brinell hardness spot strain, K’ = cyclic strength coefficient, n’ = cyclic
σu= 3.45 BHN (MPa) hardening exponent, E = modulus of elasticity; for steel,
- The fatigue strength coefficient can be E = 206,850 MPa
estimated from ultimate strength The parameters K’ and n’ are provided in Table 1 for
σ 'f ≈ σ u + 345 MPa four types of ship steel, based on DSME testing
results.
- For steels having an ultimate strength below
about σu = 1400 MPa, a fatigue limit occurs Table 1: Parameters for the cyclic stress-strain curve
near 106 cycles at a stress amplitude around
σa = σu/2. From the elastic strain-life curve, it Material A AH32 AH36 DH36
follows that K’ (MPa) 592 669 694 739
⎛ 2σ 'f ⎞ N’ 0.114 0.108 0.112 0.106
1
b=− log ⎜ ⎟
6.3 ⎜ σ u ⎟ The cyclic stress-strain curves are plotted in Figure 2.
⎝ ⎠
- A typical value of the fatigue strength
500
exponent b is -0.085. For soft metals, values
of around b equals to -0.12 are common as DH36
AH36
are values of b equals to -0.05 for hardened 400 AH32
materials A
- Values around c = -0.60 are common for the

Stress (MPa)
fatigue ductility exponent. A relatively 300
narrow range of c in the range of -0.50 to -
0.80 appears to include most engineering
materials 200

- The transition fatigue life can be


approximated from the Brinnel hardness
100
2Nt = exp(13.6 – 0.0185 BHN)
This is the value of life, N, for which the elastic
and plastic-strain life curves are the same, i.e., the point 0
where the curves cross. This relationship can be used to 0 0.005 0.01 0.015 0.02
Strain
estimate the fatigue ductility exponent
Figure 2: Cyclic stress-strain curves for different steel
PSEUDO HOT SPOT STRESS RANGE grades
CALCULATION
The second step is to employ Neuber’s rule which
Both ABS existing ship rules and upcoming IACS relates the actual stress σ and strain ε in the material, in
Common Structural Rules for tankers use the hot spot both the elastic and plastic states, to the nominal or
stress approach for fatigue assessment for only high elastic stress S. In terms of stress amplitude:
cycle fatigue. It is therefore consistent to develop low S a2
cycle fatigue assessment procedure based on hot spot σ aε a =
E
stress approach. Although some fatigue testing
where σa and εa are stress and strain amplitudes
measurements, as shown in Figure 1, are based on notch
respectively. The elastic stress amplitude is Sa = SL/2
stress, the geometry of the local notch at a weld varies
along the weld profile, and it may be difficult to find a σa and εa is then be determined based on the
geometry on which to base the analysis. simultaneous solution of the above two equations.
The transformation from elastic hot spot stress range to Material strain range, εR, is then computed as:
pseudo hot spot stress range is now considered. The εR = 2 εa
procedure is described by Dowling (1999). and the pseudo hot spot stress range is obtained by
The stress range SE for any of the j loading/offloading SL = E εR
cycles is assumed to be constant amplitude. The The subscript “L” implies stress associated with low
following discussion applies to any of the loading cycle fatigue. Using this procedure, the relationship
cycles. between the elastic hot spot stress range and pseudo hot
First define the cyclic stress-strain curve the form of spot stress range is derived and presented in Figure 3
which is [Dowling (1999)], for the four materials under consideration.
1 / n'
σ ⎛σ ⎞
ε = + ⎜⎜ ' ⎟⎟
E ⎝K ⎠

Low Cycle Fatigue Analysis of Marine Structures 97


ABS TECHNICAL PAPERS 2006

S-N CURVE APPLIED FOR LOW CYCLE 100 mm Profile flame cut, ground
and polished, corners
gauge
FATIGUE length radiused to 1.5 mm

12 mm
Figure 4 shows pseudo hot spot stress vs. number 102 mm
152
mm
of cycles to failure, with TWI (1974) and Heo et. al.
403 mm
(2004) data based on a Neuber correction. The 900 mm
specimen used in TWI (1974) is shown in Figure 5 with Ground
38 mm
a longitudinal non-load carrying fillet welds. As stated end

in TWI (1974), the final failure which was taken being c) 8 mm fillet weld 12 mm
152 mm
the point at which a sudden drop occurred in the cyclic
tensile load. A SCF of 1.55 is applied.
Figure 5: Test specimen from TWI (1974)

AH32 AH36 Heo et. al. (2004) test data is based on fatigue
6000
testing of a non-load-carrying partially penetrated
cruciform fillet joint, as shown in Figure 6. Test was
5000 carried out under stain control condition and strain ratio
Pseudo Hot Spot Stress Range, SL (MPa)

was set to be zero which means strain value fluctuates


4000 between zero and specified maximum value. Test was
DH36
topped when the load dropped down to 50% of initial
3000
A value which corresponded to small amount of crack
propagation. A SCF of 1.28 is applied.
2000 Elastic
Behavior 20 mm

1000
11 mm
Elastic Hot Spot Stress Range, SE (MPa)
0 20 mm
0 500 1000 1500 2000 2500

400 mm
Figure 3: Pseudo hot spot stress range as a function
of elastic hot spot stress range
Figure 6: Testing specimen presented in Heo et. al.
(2004)
The D curve is also plotted in Figure 4 for
reference. The median of the pooled TWI and DSME
The inverse slope of the median-2 standard
data is calculated based on least square fit. A design
deviation curve is 2.43. It is observed from Figure 1
curve is normally defined as the median curve minus
showing the Park-Lawrence model and the
two standard deviations. It is seen that for low cycle
experimental data in Figure 4, that there is a tendency
region, N < 104, using D curve, as a design S-N curve
for the S-N curve to have a curvature that bends
for low cycle fatigue, will yield conservative results.
upwards in the area where cycle to failure is below
1000. It will be ideal that this tendency be reflected in
the design S-N curve for low cycle fatigue, although
104
Median (least squares line)
TWI and DSME
based on Neuber analysis
using D curve would be conservative. However,
m = 2.43
modifying D curve for cycles to failure less than 1000
triangles = TWI
circles = DSME may complicate the damage model calculation.
Pseudo Hot Spot Stress Range, (MPa)

Extended
D-Curve
m=3
FATIGUE DAMAGE CALCULATION
1000 SUBJECTED TO LOW CYCLE LOADS

In the following, the assumptions are made that the


linear damage accumulation rule (Miner’s rule) applies,
Median - 2*sigma
that rainflow analysis is used to identify stress cycles,
m = 2.43
and that the material at the hot spot of the weld will
experience cyclic plasticity under stress cycles SLj as
100
100 1000 104 105
shown in Figure 7. General methods of such analyses
Cycles to Failure, N to produce a damage index are described in detail in the
books by Dowling (1999) and Lee et al (2005). The
Figure 4: S-N curve in low cycle region damage model proposed here uses linear damage
accumulation and implicitly, rainflow analysis, and is

98 Low Cycle Fatigue Analysis of Marine Structures


ABS TECHNICAL PAPERS 2006

based on the assumption that static and wave stresses REFERENCES


are constant amplitude. Damage due to the low cycle
static stresses is described in the following. Boardman, B.E., (1982), “Crack Initiation Fatigue –
Data, Analysis, Trends, and Estimation”, Proceedings
Basic application of Miner’s rule produces the of the SAE Fatigue Conference, P-109, SAE,
expression of static stress damage, Warrendale, PA.
k
1
DB =
K ∑ n j S Ljq DEn, (1995), “Offshore Installations, Guidance on
Design, Construction and Certification”, Department of
j =1
Let, Energy, UK, Amendment to the Fourth Edition,
nj London, HMSO.
fj =
n Dowling, N.E., (1999), Mechanical Behavior of
Materials, Prentice-Hall, Upper Saddle River, NJ.
fi = (number of the jth loading-offloading cycles)/
(number of wave induced cycles) at any life. Then the Heo, J.H., Kang, J.K, Kim, Y., Yoo, Y.S., Kim, K.S.,
total damage associated with the high stress duty cycles and Urm, H.S., (2004), “A Study on the Design
is, Guidance for Low Cycle Fatigue in Ship Structure”,
n
k Proceedings of the 9th Symposium of Practical Design
DB =
K j =1∑ q
f j S Lj of Ships and Other Floating Structures, Luebeck-
Travemunde, Germany.

Lee, Y.L., Pan, J., Hathaway, R., and Barkey, M,


Spj
Sj(t) (2005), “Fatigue Testing and Analysis”, Elsevier
Butterworth – Heinemann, Oxford, UK.

Park, S.K. and Lawrence, F.V., (1988), “Fatigue


SLj tcj
SBj
Characterization of Fabricated Ship Details for Design
– Phase II”, Ship Structures Committee, SSC-346.

TWI (1974), “Fatigue Performance of Welded High


ttj
Strength Steels”, A compendium of reports from a
sponsored research programme, The Welding Institute,
Svj Abington Hall, Abington, Cambridge CBI 6AL,
England.
Figure 7: A single loading/offloading cycle of the jth
type

CONCLUSIONS

Low cycle fatigue failure is a relatively new area of


concern in modern marine industry, especially the
application of FPSOs with frequent loading/unloading
operations. In this paper, typical material behavior
under low cycle large stress range is first investigated.
Characteristic material parameters are recommended
based on experimental test data. The pseudo hot spot
stress range can be calculated based on elastic hot spot
stress range and material stress-strain curve with the
application of Neuber’s rule. A suitable design S-N
curve is derived with reference to the available test data.
Fatigue damage can then be expressed in terms of
Miner’s rule. The procedure in this paper could be used
as a basis for authorities to establish the design criteria
for better control marine safety due to cracks from low
cycle fatigue and eventually the combination of low
cycle fatigue and high cycle (considering dynamic hull
girder loads, dynamic wave pressure and dynamic tank
pressure loads resulting from ship motions) fatigue.

Low Cycle Fatigue Analysis of Marine Structures 99