Ships and Offshore Structures

ISSN: 1744-5302 (Print) 1754-212X (Online) Journal homepage: http://www.tandfonline.com/loi/tsos20

Fatigue design of plated structures using finite

element analysis

I. Lotsberg

To cite this article: I. Lotsberg (2006) Fatigue design of plated structures using finite element
analysis, Ships and Offshore Structures, 1:1, 45-54, DOI: 10.1533/saos.2005.0006

To link to this article: https://doi.org/10.1533/saos.2005.0006

Published online: 08 Jul 2010.

Submit your article to this journal

Article views: 413

View related articles

Citing articles: 10 View citing articles

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=tsos20

Download by: [Perpustakaan Nasional Republik Indonesia (PNRI)] Date: 26 November 2017, At: 13:20
 Fatigue design of plated structures using finite
element analysis
doi:10.1533/saos.2005.0006

I. Lotsberg
Det Norske Veritas, Høvik, Norway

Abstract: Finite element analysis is being used by designers for fatigue assessment of structures. It
Downloaded by [Perpustakaan Nasional Republik Indonesia (PNRI)] at 13:20 26 November 2017

is therefore important that a proper link between calculated hot spot stress and fatigue capacity is
established, and that guidance on such analysis is included in design standards. This paper presents a
summary of the finite element analyses performed for assessment of hot spot stress with link to one hot
spot S–N curve in the FPSO Fatigue Capacity Joint Industry Project. Recommendations are presented
on how to perform fatigue assessment of plated structures based on finite element analysis combined
with one hot spot S–N curve.

Key words: Fatigue design, plated structures, finite element analysis, hot spot stress S–N curve.

INTRODUCTION THE HOT SPOT STRESS METHODOLOGY

Finite element analysis (FEA) programs have become a The hot spot stress concept
standard design tools in most design offices. By these pro- The hot spot stress concept has a long history in fatigue
grams the stresses from the dynamic loads can be rather design of tubular joints. The same concept is also becoming
accurately calculated. In a design process these stresses part of a fatigue design methodology for plated structures
are assessed against the capacity of the analysed structural in ships and FPSOs. The hot spot stress can be directly
details. The fatigue design of details in plated structures derived from the performed FEA, or it can be derived
has mainly been based on a concept of nominal stress ap- through use of appropriate stress concentration factors and
proach until the nineties. The use of more refined FEA the nominal stress from the analysis data.
resulted in a need for an assessment procedure based on S–N data based on a nominal stress concept is rather
a hot spot stress methodology as it was found difficult well defined with respect to failure criterion and associated
to define/derive nominal stresses from fine element mesh nominal stress to be used. However, a practical derivation
models. In the early 1990, the Classification companies of the nominal stress from FEA is considered to be more
introduced fatigue assessment procedures based on the difficult. The hot spot stress concept can hardly be out-
hot spot stress concept also for plated structures. With lined from a theoretical sound basis. A failure criterion
construction of Floating Production and Offshore Load- is no longer so well defined when fatigue test data based
ing Ships, the need for a reliable fatigue design proce- on nominal stress are transferred into one hot spot stress
dure for plated structures was emphasised and an industry S–N curve. Also one hot spot stress may be associated with
project was initiated. This paper presents a summary of different stress gradients through the thickness and thus
finite element analyses performed for assessment of hot different fatigue lives. However, with these approxima-
spot stress and derivation of a hot spot stress S–N curve tions in mind the hot spot stress concept is considered to
in the FPSO Fatigue Capacity JIP (Bergan and Lotsberg be an efficient engineering methodology for fatigue analysis
2004). of plated structures.

Methods for hot spot stress derivation

Because of the nature of the stress field at a hot spot re-
Corresponding Author: gion there are questions on how to establish the hot spot
I. Lotsberg stress, see Figure 1. The notch effect due to the weld is
Det Norske Veritas included in the S–N curve and the hot spot stress is de-
NO-1322, Høvik, Norway rived by extrapolation of the structural stress to the weld
Tel: +47 675 79900; Fax: +47 675 79911
Email: [email protected]
toe. It is observed that the stress used as basis for such
an extrapolation should be outside that affected by the

C Woodhead Publishing Ltd 45 SAOS 2006 Vol. 1 No. 1 pp. 45–54
 I. Lotsberg

• One 20-node isoparametric element over the thickness

Notch stress
Stress (second-order displacement function giving linear stress
Hot spot stress when calculating its derivative)
Surface stress
• Shell elements that represent axial force and bending mo-
ment over the plate thickness.
Attachment
t/2 3t/2
plate
Hot spot stress In addition, one may calculate the ‘structural stress’
Notch stress (also denoted geometric stress) from that of a linear stress
Fillet weld
distribution through the plate at the hot spot. The hot spot
t stress can be calculated directly from the nodal forces from
a FE analysis. This methodology has been presented in the
Membrane stress literature by Dong et al (2001), and the methodology has
Stress evaluation plane been assessed by Doerk et al (2003). For three-dimensional
Downloaded by [Perpustakaan Nasional Republik Indonesia (PNRI)] at 13:20 26 November 2017

Nominal stress analysis, an integration (or linearisation) of the stress may

be performed as an alternative. See also Designers Guide by
Niemi (2001).
It is questioned how to consider a varying stress distri-
Fillet weld bution along the elements width, see Figure 1. An alterna-
Attachment plate tive might be to calculate a mean stress over the element
width, e.g., by using solid elements having the width of
the attachment (incl. welds) and taking the average stress
A A
in these elements as proposed by Fricke (2001). The result
will be a function of the element width.
Figure 1 Schematic stress distribution at a hot spot.
LINK BETWEEN HOT SPOT STRESS AND S– N CURVE

weld notch, but close enough to pick up the structural Background

stress. An assessment of hot spot stress S–N curve for plated
A number of Classification Societies have recommended structures was performed by Maddox (2001) in this project.
derivation of hot spot stress from linear extrapolation using He assessed available fatigue test data at The Welding
the stress values calculated at points 0.5t and 1.5t from the Institute. A number of fatigue test data of small-scale test
weld toe (or intersection line) from FE analysis with a well specimens were derived in the first phase of the FPSO
defined mesh size at the hot spot region. This procedure is Fatigue Capacity project by Kang and Kim (2003). The
found acceptable with respect to accuracy and robustness. test data were linked to the hot spot stress methodology
However, it is considered to be time consuming in terms using FEA by Fricke (2001). FAT90 was recommended as
of engineering time. This is due to manual work involved a hot spot stress S–N curve when the hot spot stress from
in the derivation of the hot spot stress. It is also observed FEA was derived by linear extrapolation from stresses at
that engineers perform this different if it is not properly read out points 0.5t and 1.5t from the weld toe or the
described in a guideline how to establish the stress at 0.5t intersection line when the weld is not included in the FE
and 1.5t. Some derive these stresses from the Gaussian model. Then the hot spot stress is derived at the weld toe
stresses and other directly from the nodal stresses. The or at the intersection line. FAT80 was recommended as a
procedure for calculation of elements nodal stresses are hot spot stress S–N curve when the hot spot stress from
not clearly described in the computer programs and this FEA was derived from stresses at read out points 0.5t from
introduces some uncertainty when unproven programs are the weld toe or the intersection line when the weld is not
used for hot spot stress derivation. From an engineering included in the FE model.
point of view, a simple derivation of the hot spot stress at
a position 0.5t from the weld toe is attractive. It also shows Fatigue test data from small-scale specimens transferred
small scatter in the calculated results (refer, Fricke 2001, into one hot spot S–N curve
Storsul et al 2004a).
The five HHI fatigue tested specimens are shown in
For analysis by shell elements the distance to the stress
Figure 2. S–N diagrams based on nominal stress were de-
read out points is measured from the intersection lines as
rived based on these test results. Then the hot spot stress
the weld is not normally included in the finite element
range is obtained as
model. For analysis by solid elements the distance to the
stress read out points is measured from the weld toe. σHot spot = KσNominal (1)
The hot spot stress methodology implies a linear stress
distribution through the plate thickness in the way that the where K is a structural stress concentration factor for the
finite element modelling is performed: considered detail and σNominal is the nominal stress. The

SAOS 2006 Vol. 1 No. 1 46 doi:10.1533/saos.2005.0006

C Woodhead Publishing Ltd
 Fatigue design of plated structures using finite element analysis

Table 1 S--N curves

Specimen no Type of stress data Loga St. dev.
1 Nominal 12.200 0.111
2 Nominal 11.686 0.125
3 Nominal 12.186 0.072
4 Nominal as tested 12.123 0.232
4 Nominal—Corrected for bending by factor 1.61 on life 11.916 0.232
5 Nominal as tested 12.326 0.104
5 Nominal—Corrected for residual stress by factor 2.8 on life 11.879 0.104
1–5 Hot spot stress S–N curve derived from calculated and measured stress 12.565 0.167
1–5 Hot spot stress S–N curve derived from fatigue test data 12.562 0.139
Downloaded by [Perpustakaan Nasional Republik Indonesia (PNRI)] at 13:20 26 November 2017

hot spot S–N curve. The resulting S–N curve and the fa-
1 3
2 tigue test data for a high mean stress loading is shown in
Figure 3. The design S–N curve corresponds to a stress
range of 98.70 MPa at 2 Mill cycles which is above the
FAT90 design curve. The tested specimens were fabri-
cated from 10 mm thick plates. This is a lower thickness
4 than normally used as reference thickness in the S–N data.
5
Thus it is recommended to use the FAT90 curve as a hot
spot S–N curve and use the reference thickness as rec-
ommended by IIW equal 25 mm. Increasing the thickness
Figure 2 Specimens fatigue tested by HHI and analysed by from 10 to 25 mm reduces the S–N curve to that of FAT 90
FE. for a rather low thickness exponent equal 0.1. The fatigue
test data from HHI supports the FAT90 curve as a design
hot spot S–N curve. The scatter in the test data including
K-factors derived from finite element analyses for deriva- all five data sets is small. A standard deviation of 0.167 is
tion of hot spot stress are listed in Table 1. calculated. This is low even if it is compared with one of
It should be noted that the finite element modelling the data sets such as the hopper corner where the standard
might influence the calculated stress at the hot spot re- deviation is 0.232. Reference is also made to Table 1.
gion. Parameters affecting this are type and size of element
used, and how the stresses are derived from the analysis Hot spot S–N curve based on fatigue tests
(Gaussian stress, nodal stress, etc.). Therefore, the same An alternative way of assessing target hot spot stress values
procedure for analysis should be used for design as that is to base the evaluation on the fatigue test data for the five
used for assessment of a design hot spot stress S–N curve. HHI test specimens. The mean S–N curves for each of
the five details are known in terms of nominal stress S–N
Hot spot stress from FE analysis and measurements curves, see Table 1. From these, the relative distribution of
Based on the calculated stress through the plate thickness K-factors for the different details is known. Thus the main
of the hopper corner detail (specimen number 4), the fa- question is related to how to derive an absolute value of the
tigue test data is reduced by a factor 1.61 in life due to hot spot S–N curve. Thus it is sufficient to determine one of
the actual stress gradient for comparison with one hot spot these K-factors. Then the other follows. It is observed that
S–N curve (Lotsberg and Sigurdsson 2004). The fatigue detail 1 is a well-known detail frequently fatigue tested such
test results for specimen number 5 by HHI were explained that this detail is considered to be one of the most reliably
by compressive stresses at the hot spot area. Strain mea- categorized details in design standards. It is seen that the
surements showed compressive stresses at the hot spots. K-factor is equal 1.32 for specimen number 1 from Phase
The fatigue tests performed on stress-relieved specimens I based on FE analysis. The same value for this K-factor
showed that stress relieving reduced the fatigue life by a was also derived from fatigue design standards includ-
factor 2.8 (refer, Kim and Lotsberg 2004). Residual stresses ing this particular detail (refer, Lotsberg and Sigurdsson
in tension as may be possible after a construction of a more 2004). The resulting K-factors are listed in the right col-
complex structure might reduce the fatigue capacity fur- umn of Table 2. For specimen number 2 a somewhat higher
ther. The hot spot S–N curve should be applicable to K-factor value is obtained from the S–N curve than de-
welded structures that may be in a state of tensile residual rived from FE analysis. Also for specimen number 3, a
stress at the considered hot spots. Therefore, the original somewhat higher K-factor value is obtained from the S–N
fatigue test data for specimen number 5 were reduced by curve approach than that derived from FE analysis. For
a factor 2.8 on number of cycles for comparison with one specimen 4 a lower K-factor resulted from the S–N data

C Woodhead Publishing Ltd doi:10.1533/saos.2005.0006 47 SAOS 2006 Vol. 1 No. 1
 I. Lotsberg

1000

Hot spot stress range (MPa)

No 1
No 2
No 3
No 4
100
No 5
Mean
Mean minus 2 std
FAT 90

10
Downloaded by [Perpustakaan Nasional Republik Indonesia (PNRI)] at 13:20 26 November 2017

10000 100000 1000000 10000000

Number of cycles

Figure 3 Fatigue test results from specimens 1–5 plotted into one hot spot S–N diagram.

Table 2 Target K-factors for HHI specimens where

Specimen no. ref. K-factor from FE K-factor resulting
x = K-factor for detail 1 as k1 = 1.0.
Figure 2 analysis supported from fatigue S–N
with measurements data from the ki = K-factor for considered detail i relative to that for
HHI tests detail number 1.
KFEi = K-factor from FE analysis for detail number i as
1 1.32 1.32 presented in Table 1.
2 1.85 1.96
3 1.22 1.33 By putting δ = 0, a K-factor for detail number 1 equal
4 1.82 1.64 to 1.327 is calculated. This is approximately the same value
5 1.77 1.69
as derived above. This indicates that the derived K-factors
are sound.
It is observed from Table 2 that the difference be-
than from the FE analysis. This may be due to difference in
tween the target K-factor derived from analyses and mea-
failure criteria for the specimens. Failure for this specimen
surements corresponds well with the K-factors derived
is defined as crack growth through the thickness. For the
from the fatigue test data from the HHI specimens. The
HHI specimens, 1–3 failure is defined as fracture of the
difference is in the range 0–10%. The largest differ-
specimens and for HHI specimen number 5 the failure cri-
ence is derived for the hopper corner detail as discussed
terion is crack growth equal 20 mm as a mean value along
above.
the two sides. Especially for specimen number 4 there is
significant crack growth before the crack is going through
the plate thickness. An earlier stop in the fatigue testing Fatigue test data from full scale test specimens
would result in a lower nominal S–N curve and hence a The hot spot stress design curve is verified by fatigue test-
larger K-factor. Thus, it is possible that the target value ing of five full scale test specimens (Lotsberg et al 2001,
should be closer to that analysed and measured of 1.82. 2004). Four different connections between side longitu-
However, to be consistent a target spot value as derived dinals and transverse frames were fatigue tested. The fa-
from the fatigue test data of 1.64 is used for specimen 4. tigue cracks initiated at different hot spot positions. Also
For specimen 5 a target value of 1.69 came out from the the fatigue cracks grew into different areas. In spite of
S–N analysis data. This is somewhat lower than from that this difference the test data in terms of hot spot stress fit
considered to be lower bound FE analysis. well into one S–N diagram as shown in Figure 4. Crack
An alternative using least square method to assess tar- growth from a crack size equal 12–50 mm is shown. The
get K-factors is also considered. The quadratic error of scatter in the test data is small: A standard deviation of
the difference between K-factors from FEA and that from 0.17 for a crack size of 50 mm. This standard deviation is
S–N data is minimized according to the following expres- not larger than that is normal for test data from one sin-
sion gle geometry based on nominal stress. Thus, the fatigue
test data from the full scale test specimens support the
∂

5
δ= (KFEi − xki )2 (2) FAT90 curve as a design hot spot S–N curve for plated
∂ x i =1 structures.

SAOS 2006 Vol. 1 No. 1 48 doi:10.1533/saos.2005.0006

C Woodhead Publishing Ltd
 Fatigue design of plated structures using finite element analysis

IIW Mean
1000 FAT 90
IIW Design
FAT 90
Specimen1
Crack 1
Specimen1
Crack 4
Specimen 2
Stress range (MPa) Crack 5
Specimen 2
Crack 6
100 Specimen 3
Crack 1
Specimen 3
Crack 2
Specimen 4
Crack 1
Specimen 4
Downloaded by [Perpustakaan Nasional Republik Indonesia (PNRI)] at 13:20 26 November 2017

Crack 3
Specimen 5
Crack 1
Specimen 5
Crack 2
10
10000 100000 1000000 10000000

Number of cycles

Figure 4 S–N data from full scale fatigue tests compared with hot spot stress S–N curve.

FINITE ELEMENT ANALYSIS FOR ASSESSMENT OF about the methodology with finite element modelling for
METHODOLOGY such conditions. The HHI specimen 4 is considered to be
the most severe for out of plane bending.
Specimens analysed
The specimens selected for finite element analyses are
shown in Figures 2 and 5. In Phase I of the project the ISSC, Target hot spot stress values
HHI 2, HHI 4, HHI 5, and GL doubling plate specimens Target hot spot stress values are required for assessment of
were analysed. Thus, there was already a considerable data accuracy of the finite element analyses. The derived target
base with FE analyses results with mesh size t × t (t = hot spot stress values for the different specimens are shown
thickness) (refer, Fricke 2001, Fricke and Säbel 2000). It in Table 2. These target hot spot stress values were derived
was decided to supplement these data with FE analyses for based on a total assessment of the performed analyses sup-
other element meshes for the same specimens. In addition, plemented by measured stress and fatigue test data. The
it was decided to perform analyses of a specimen with a results from three-dimensional FE analyses where the fillet
very severe notch at crossing plates where measurements weld is included are given a more significant weight than
of strain at the hot spot region were performed in another the other analyses in this assessment. The target values
project (Lotsberg et al 1998), see Figure 5. This detail was refer to the CS methodology of deriving hot spot stress
fabricated with a full penetration weld without shoulder by extrapolation from 0.5t/1.5t to the intersection line for
fillets. This is a detail with a more severe notch than nor- FE-models with shell elements and to the weld toe for
mally found in ships and FPSO structures with a sound models with three-dimensional elements. The final target
design. The detail is included due to strong in-plane stress hot spot stress values for the HHI test specimens 1–5 were
gradients and will therefore provide valuable information derived from the fatigue test data as explained in section

ISSC (1997) specimen Severe notch detail (SND) DNV (1998)

Investigated
hot spot area

Figure 5 Geometry of specimens analysed.

C Woodhead Publishing Ltd doi:10.1533/saos.2005.0006 49 SAOS 2006 Vol. 1 No. 1
 I. Lotsberg

‘Fatigue test data from small scale specimens transferred the intersection line are shown in Table 4 for hot spot
into one hot spot S-N curve.’ A target hot spot stress for stress linked to the FAT80 curve. It is noted that large
the ISSC detail equals 155 MPa was used based on FE 4-node elements are acceptable for significant out of plane
analysis reported by Olafsrud (2002). The target hot spot bending such as for the hopper corner (HHI 4). Large 4-
stress value for the GL Doubling plate is the same as used node elements are less good for in plane stresses such as
by Fricke and Säbel (2000): Kg = 1.49. The target hot spot for the ISSC model.
stress value for the DNV Severe Notch Specimens was The considered element meshes are in general larger
based on a three-dimensional analysis with a fine mesh: than that of converged meshes. Also when the mesh size is
Kg = 3.13. This was in correspondence with that of the such that all the read out points are situated inside the first
measured stresses. element at the intersection point, this will give hot spot
stress values that are somewhat affected by the singularity
Accuracy of finite element analyses
at the hot spot. This effect is likely larger for the 4-node
The following definition of accuracy is used for presenta- shell elements than the 8-node shell elements. Thus, larger
Downloaded by [Perpustakaan Nasional Republik Indonesia (PNRI)] at 13:20 26 November 2017

tion of hot spot stress derived from finite element analyses: stresses are derived by the 4-node shell elements than by the
Good: Within +10 and −5% of target value. 8-node shell elements. It is observed that the methodology
Acceptable +: Within +10 and +20% of target value. for analysis of the severe notch detail with crossing plates
Acceptable −: Within −10 and −5% of target value. shown in Figure 5 in general is conservative when it is
Conservative: More than +20% (C). modelled with shell elements. This might also be expected
Non-conservative: Less than −10% (NC). as the stress in the actual connection will be distributed
over a thick crossing plate (thickness = 50 mm). Only its
The following abbreviations are used in tables: ABS: centre line is represented in the shell FE model.
American Bureau of Shipping; AMT: Aker Maritime Ten- Similar analyses were performed using 8-node shell el-
tech; BLU: Bluewater Engineering; BV: Bureau Veritas; ements and 20-node three-dimensional elements. Tables
DNV: Det Norske Veritas; GL: Germanischer Lloyd; summarizing the results are presented by Lotsberg (2004).
HHI: Hyundai Heavy Industries Co. Ltd.; ISSC: Interna- In general, the scatter in the derived hot spot data is not
tional Ship Structural Committee; LR: Lloyds’ Register; considered to be large. Some of the scatter is likely more
UM: Umoe (Now ABB). due to different techniques used to derive stresses at the
A number of different analyses for evaluation of con- read out points than that due to the analyses themselves.
vergence and divergence in hot spot stress have been per-
formed. For assessment of the results the following infor-
mation is given: RECOMMENDATIONS ON FE-MODELLING AND HOT
SPOT STRESS CALCULATION
• The DNV FE model of HHI 4 with shell elements is
made with width t for element length 0.5t. For model with Modelling
three-dimensional elements the width of the elements is t
for length 0.5t and 2t. The following guidance is made to the computation of hot
• The misalignment at the hopper corner is not included in spot stresses with potential fatigue cracking from the weld
the shell models of HHI 4 specimen using 4-node elements toe with local models using the finite element method.
by LR, HHI, and BV (2). (Number 2 in the latter means Hot spot stresses are calculated assuming linear material
4-node element with additional degrees of freedom for behaviour and using an idealized structural model with
in-plane deformation). no fabrication-related misalignment. The extent of the lo-
• By DNV, (1) is understood model for HHI 3 with in- cal model has to be chosen, such that effects due to the
creased thickness. By DNV, (2) is understood model for boundaries on the structural detail considered are suffi-
HHI 3 where the weld between the doubling plate and the ciently small and reasonable boundary conditions can be
main plate is modelled by transverse shell elements with formulated.
thickness twice that of the plates. Refer also Storsul et al In plate structures, three types of hot spots at weld toes
(2004a). can be identified as exemplified in Figure 6 (Fricke 2001):
The results from models with 4-node shell elements • at the weld toe, on the plate surface at an ending attachment
with internal degrees of freedom for proper representation • at the weld toe, around the plate edge of an ending attach-
of in-plane stress are shown in Table 3. Here, a linear ment
extrapolation of stresses at read out points 0.5t and 1.5t • along the weld of an attached plate (weld toes on both the
from the intersection line is used to derive hot spot stress plate and attachment surface).
linked to the FAT90 curve. A mesh size from 0.5t × 0.5t
up to 2t × 2t may be used. Larger mesh sizes at the hot spot Models with thin plate or shell elements or alternatively
region may provide nonconservative results. However, an with solid elements are normally used. It should be noted
element size of 4t × 4t may be considered acceptable for that on the one hand the arrangement and type of elements
fatigue screening of details. The results from the same have to allow for steep stress gradients as well as for the
models with hot spot stress from read out points 0.5t from formation of plate bending, and on the other hand, only the

SAOS 2006 Vol. 1 No. 1 50 doi:10.1533/saos.2005.0006

C Woodhead Publishing Ltd
 Fatigue design of plated structures using finite element analysis

Table 3 Assessment of hot spot stress by 0.5t/1.5t extrapolation using 4-node shell
elements linked to FAT90
Mesh size
S–N curve: FAT90
Specimen Company 0.5t × 0.5t t×t 2t × 2t 4t × 4t
ISSC NTNU Good Good NC (−17%)
ISSC LR Good
ISSC ABS Acceptable−
ISSC BV(2) Good
HHI 1 HHI Acceptable+
HHI 1 DNV C C Acceptable+
HHI 2 2 LR Good
HHI 2 HHI Good
Downloaded by [Perpustakaan Nasional Republik Indonesia (PNRI)] at 13:20 26 November 2017

HHI 2 BV(2) Good

HHI 2 DNV Good Good NC (−15.2%)
HHI 3 HHI Good
HHI 3 DNV (1) Good Good Good
HHI 3 DNV (2) Good Good Good
HHI 4 LR C
HHI 4 HHI C
HHI 4 ABS C
HHI 4 BV (2) C
HHI 4 DNV C C C Good
HHI 5 LR C
HHI 5 HHI C
HHI 5 BV (2) C
HHI 5 DNV C C Good
GL DP LR Good
GL DP BV (2) Good
DNV SND DNV C C C

shown in Figure 6b in connection with 4-noded elements,

which should contain improved in-plane bending modes.
The welds are usually not modelled except for special cases
where the results are affected by high local bending, e.g.,
due to an offset between plates or due to a small free plate
length between adjacent welds such as at lug (or collar)
c plates. Here, the weld may be included by vertical plate
elements having appropriate stiffness or by introducing
b
constrained equations for coupled node displacements. A
c thickness equal 2 times the thickness of the plates may be
a
used for modelling of the welds by transverse plates.

An alternative particularly for complex cases is offered

by solid elements which need to have a displacement func-
tion allowing steep stress gradients as well as plate bending
Figure 6 Different hot spot positions. with linear stress distribution in the plate thickness di-
rection. This is offered, e.g., by isoparametric 20-node
linear stress distribution in the plate thickness direction elements (with midside nodes at the edges) which mean
needs to be evaluated with respect to the definition of that only one element in plate thickness direction is re-
structural stress. quired. Modelling of the welds is generally recommended
The following methods of modelling are recommended: and easily possible as shown in Figure 7b.
• The simplest way of modelling is offered by thin plate and
Derivation of stress at read out points 0.5t and 1.5t
shell elements which have to be arranged in the midplane
of the structural components, see also Figure 7a. The element sizes depend on the method of stress eval-
• Eight-noded elements are recommended particularly in uation. For this reason, the evaluation methods are de-
case of steep stress gradients. Care should be given to pos- scribed before summarizing recommendations on element
sible stress underestimation especially at weld toes of type sizes. The average stress components between adjacent

C Woodhead Publishing Ltd doi:10.1533/saos.2005.0006 51 SAOS 2006 Vol. 1 No. 1
 I. Lotsberg

Table 4 Assessment of hot spot stress from 0.5t using 4-node shell elements linked to FAT80
Mesh size
S–N curve: FAT80
Specimen Company 0.5t × 0.5t t×t 2t × 2t 4t × 4t
ISSC NTNU Good Good Acceptable−
ISSC LR Good
ISSC ABS Acceptable−
ISSC BV(2) Good
HHI 1 HHI C
HHI 1 DNV C C Good
HHI 2 LR Good
HHI 2 HHI Good
HHI 2 BV(2) Good
Downloaded by [Perpustakaan Nasional Republik Indonesia (PNRI)] at 13:20 26 November 2017

HHI 2 DNV Good Good NC (−11.5%)

HHI 3 HHI Good
HHI 3 DNV (1) Good Good Good
HHI 3 DNV (2) Acceptable+ Acceptable+ Good
HHI 4 LR C
HHI 4 HHI C
HHI 4 ABS C
HHI 4 BV (2) C
HHI 4 DNV C C C C
HHI 5 LR C
HHI 5 HHI Acceptable+
HHI 5 BV (2) C
HHI 5 DNV C C Good
GL DP LR Acceptable+
GL DP BV (2) Acceptable+
DNV SND DNV Acceptable+ C C

elements are used for hot spot stress derivation. Recom- Then these stresses can be interpolated linearly to
mended stress evaluation points are located at distances the surface centre or extrapolated to the edge of
0.5t and 1.5t away from the hot spot, where t is the plate the elements if this is the line for hot spot stress
thickness at the weld toe. These locations are also denoted derivation.
as stress read out points. For meshes with 4-node shell elements larger than t ×
If the element size at a hot spot region of size t × t is t, it is recommended to fit a second-order polynomial to
used, the stresses may be evaluated as follows: the element stresses in the three first elements and derive
• In the case of plate or shell elements, the surface stress may stresses for extrapolation from the 0.5t and 1.5t points.
be evaluated at the corresponding midside points. Thus, This procedure may be used to establish stress values at
the stresses at midside nodes may be used directly as stress the 0.5t and 1.5t points. For 8-node elements, a second-
at read out points 0.5t and 1.5t. order polynomial may be fitted to the stress results at the
• In the case of solid elements, the stress may first be midside nodes of the three first elements and the stress at
extrapolated from the Gaussian points to the surface. the read out points 0.5t and 1.5t can be derived.

a b

Figure 7 Stress extrapolation in a three-dimensional FE model. (a) Shell model. (b) Three-dimensional model including
weld toe.

SAOS 2006 Vol. 1 No. 1 52 doi:10.1533/saos.2005.0006

C Woodhead Publishing Ltd
 Fatigue design of plated structures using finite element analysis

Derivation of hot spot stress where

Two alternative methods can be used for hot spot stress σa,spot = membrane stress
derivation: σb,spot = bending stress
The reason for a reduction factor on the bending stress
• A linear extrapolation of the stresses to the hot spot (weld
toe or intersection line) from the read out points at 0.5t is load-shedding effect during crack growth. The effect
and 1.5t. The principal stress at the hot spot is calculated is limited to areas with a localised stress concentration,
from the extrapolated component values. which occurs for example at a hopper corner. However, in
• The hot spot stress is taken as the stress at the read out a case where the stress variation along the weld is small,
point 0.5t away from the hot spot and multiplied by 1.12. the difference in fatigue life between axial loading and pure
bending is much smaller. Therefore, it should be noted that
Mesh size it is not correct to generally reduce the bending part of the
Shell elements
stress to 60%. This has to be restricted to cases with a
pronounced stress concentration.
Downloaded by [Perpustakaan Nasional Republik Indonesia (PNRI)] at 13:20 26 November 2017

If 4-node shell elements are used they should contain ad-

ditional internal degrees of freedom for improved in plane This procedure is supported by fatigue test data de-
behaviour. For these elements and for 8-node shell ele- rived under out of plane loading by Kang et al (2002) and
ments a mesh size from t × t up to 2t × 2t may be used. by fracture mechanics analyses (Lotsberg and Sigurdsson
Larger mesh sizes at the hot spot region may provide non- 2004).
conservative results.
Limitations for simple connections
Three-dimensional solid elements It should be noted that the definition of the stress field
Solid modelling is time consuming and more laborious through the plate thickness in section ‘Methods for hot spot
than shell modelling that is normally the preferred method stress derivation’ implies that the described hot spot stress
when a plated structure is to be analysed. However, solid el- methodology is not recommended for simple cruciform
ements are considered to give more reliable results and are joints, simple T-joints in plated structures or simple butt
recommended used in some cases of more complex struc- joints that are welded from one side only. Here the nominal
tures. For modelling with three-dimensional elements the stress approach is considered to be most reliable for analysis
dimensions of the first two or three elements in front of the of these connections. It should also be noted that fabrication
weld toe should be chosen as follows. The element length tolerances are most important for these joints and need to
may be selected to correspond to the plate thickness. In be considered in a fatigue assessment.
the transverse direction, the plate thickness may be chosen
again for the breadth of the plate elements. However, the
breadth should not exceed the ‘attachment width’, i.e., the CONCLUSIONS
thickness of the attached plate plus 2 × the weld leg length
A hot spot stress S–N curve for fatigue cracking from
(in case of type c: the thickness of the web plate behind
weld toes is derived from small-scale fatigue test data.
plus 2 × weld leg length), see Figure 7b. For 20-node hex-
This hot spot S–N curve is supported by the fatigue test
ahedral elements it is sufficient with one element over the
data from the full scale specimens. The tested specimens
thickness to pick up a linear stress distribution. For simple
were analysed by finite elements for derivation of hot spot
8-node brick elements at least four elements are required
stress. This derivation of hot spot stress was linked to a
for the same purpose.
hot spot stress S–N curve to obtain a consistent fatigue
In cases where three-dimensional elements are used for
assessment procedure. Based on the performed work, rec-
the FE modelling, it is recommended that also the fil-
ommendations with respect to finite element modelling
let weld is modelled to achieve proper local stiffness and
of fatigue sensitive structures were made. These recom-
geometry. Reference is made to Storsul et al (2004b). In
mendations are presented in section ‘Recommendations
general, finite element meshes should be made without
on FE-modelling and hot spot stress calculation’.
significant distortion of the mesh and width length width
ratio within 1:4. Including the fillet weld will limit the size
of the mesh at the hot spot region. In order to capture the ACKNOWLEDGMENT
properties of bulb sections with respect to St. Venant tor-
sion, it is recommended to use several three-dimensional The author thanks the participants of Phase II of the
elements for modelling of a bulb section. FPSO Fatigue Capacity JIP for making this work possi-
ble: Hyundai Heavy Industries (HHI), Daewoo Shipbuild-
Derivation of effective hot spot stress from FE analysis ing & Marine Engineering (DSME), Bluewater, Total,
At hot spots with significant plate bending one might derive ConocoPhillips, Shell Deepwater Develop. Systems, Sta-
an effective hot spot stress for fatigue assessment based on toil, BP Exploration Operating Company Limited, Petro-
the following equation: bras, Navion, Bureau Veritas, Det Norske Veritas, Health
and Safety Executive (HSE), MARIN, Technical Univer-
σe,spot = σa,spot + 0.60σb,spot (3) sity Hamburg-Harburg, National University of Singapore,

C Woodhead Publishing Ltd doi:10.1533/saos.2005.0006 53 SAOS 2006 Vol. 1 No. 1
 I. Lotsberg

Norwegian University of Science and Technology, and FPSO’04-0018 International Conference, Houston,
Shanghai Jiaotong University. TX.
Lotsberg I, Sigurdsson G. 2004. Hot spot S–N curve for
fatigue analysis of plated structures. In OMAE-
REFERENCES
FPSO’04-0014, International Conference, Houston,
Bergan PG, Lotsberg I. 2004. Advances in fatigue assessment of TX.
FPSOs. In OMAE-FPSO’04-0012, International Conference, Lotsberg I, Müsch K, Måseide MO et al. 1998. Tested capacity of
Houston, TX. welded connections made of high strength steel. Lisbon:
Doerk O, Fricke W, Weissenborn C et al. 2003. Comparison of OMAE.
different calculation methods for structural stresses at welded Lotsberg I. 2004. Recommended methodology for analysis of
joints. Int J Fatigue, 25:359–69. structural stress for fatigue assessment of plated structures.
Dong P, Hong JK, Cao C et al. 2001. A mesh-insensitive structural OMAE-FPSO’04-0013, International Conference, Houston,
stress procedure for fatigue evaluation of welded structures. TX.
IIW Doc. XIII-1902-01. Lotsberg I, Askheim DØ, Haavi T et al. 2001. Full scale fatigue
Downloaded by [Perpustakaan Nasional Republik Indonesia (PNRI)] at 13:20 26 November 2017

Fricke W. 2001. Recommended hot spot analysis procedure for testing of side longitudinals in FPSOs. In: Proceedings of the
structural details of FPSO’s and ships based on round-robin 11th ISOPE, Stavanger.
FE analyses. In Proceedings of the 11th ISOPE, Stavanger. Lotsberg I, Landet E. 2004. Fatigue capacity of side longitudinals
Also in Int J Offshore Polar Eng, 12(1), March 2002. in floating structures. In: OMAE-FPSO’04-0015,
Fricke W, Säbel A. 2000. Hot spot stress analysis of five structural International Conference, Houston, TX.
details and recommendations for modelling, stress evaluation Maddox S. 2001. Recommended design S–N curves for fatigue
and design S–N Curve. GL Report No FF99.188.A, Rev 02. assessment of FPSOs. In: Proceedings of the 11th ISOPE,
IIW. 1996. Fatigue design of welded joints and components. In Stavanger.
Hobbacher A, ed. Recommendations of IIW Joint Working Niemi E. 2001. Structural stress approach to fatigue analysis of
Group XIII-1539-96/XV-845-96. Cambridge, UK: Abington welded components, -Designer’s Guide. IIW Doc.
Publishing and The International Institute of Welding. XV-1090-01.
ISSC. 1997. Report of Committee II.1 ‘Quasi-static response’. In Olafsrud K. 2002. Analysis methods for stress concentration in ship
Moan T, Berge S, eds. Proceedings of the 13th International hull details. Trondheim: NTNU, Department of Marine
Ship and Structures Congress, Vol. 1. London: Elsevier Structures.
Science. Storsul R, Landet E, Lotsberg I et al. 2004a. Convergence analysis
Kang SW, Kim WS, Paik YM. 2002. Fatigue strength of fillet for welded details in ship shaped structures. In OMAE-
welded steel structure under out-of-plane bending. FPSO’04-0016, International Conference, Houston, TX.
International Welding/Joining Conference, Korea. Storsul R, Landet E, Lotsberg I et al. 2004b. Calculated and
Kang SW, Kim WS. 2003. A proposed S–N curve for welded ship measured stress at welded connections between side
structure. Welding J, 82(7):161–9. longitudinals and transverse frames in ship shaped structures.
Kim WS, Lotsberg I. 2004. Fatigue test data for welded In Proceedings of OMAE-FPSO’04-0017, International
connections in ship shaped structures. In OMAE- Conference, Houston, TX.

SAOS 2006 Vol. 1 No. 1 54 doi:10.1533/saos.2005.0006

C Woodhead Publishing Ltd