Вы находитесь на странице: 1из 15

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/263466234

Fatigue Design and Structural Hot-Spot Stress Determination for Welded


Joints

Conference Paper · January 2004

CITATIONS READS

0 338

1 author:

Wolfgang Fricke
Technische Universität Hamburg
337 PUBLICATIONS   2,364 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Fatigue strength of post-weld improved high strength steel joints for high performance sailing yachts View project

All content following this page was uploaded by Wolfgang Fricke on 03 November 2015.

The user has requested enhancement of the downloaded file.


9th PORTUGUESE CONFERENCE ON FRACTURE - 2004

FATIGUE DESIGN AND STRUCTURAL HOT-SPOT STRESS DETERMINATION


FOR WELDED JOINTS

W. FRICKE

Ship Structural Design and Analysis


Technical University Hamburg-Harburg
21071 Hamburg, Germany

Abstract. During the past years, several fatigue failures occurred in welded structures such as ships,
offshore structures, bridges, wind energy converters etc., causing high costs for repair and suspended
service. The frequent failures during the first years of service indicate that fatigue has not been thor-
oughly considered during design and fabrication. The paper gives an overview about current fatigue
design procedures and their practical application based on numerical methods as well as developments
to improve the situation. Particularly the structural hot-spot stress approach is well-suited for practical
application because it allows the effects of the structural configuration to be taken into account. How-
ever, comparisons between fatigue tests and analyses show that fatigue life predictions are sometimes
wrong. Especially fabrications effects, which are considered only in a coarse way or not at all, have
shown to be responsible for that. In this connection, the fabrication-related residual stresses play an
important role. Their determination is at present limited, however it becomes clear that the numerical
simulation of fabrication will in future enable not only the welding shrinkage and distortions to be
computed, but also their effects on fatigue strength.

1. INTRODUCTION In spite of powerful computers and effective


numerical methods such as the finite element
Several types of welded metal structures are method (FEM) it is still a problem today to
subjected to high cyclic loading, which may identify all fatigue-prone details in a complex
cause fatigue cracking. In most cases these structure (Fig. 1) and to perform a comprehen-
cracks appear at welded joints, owing to the sive fatigue analysis. This is partly due to the
sharp notches of the weld profile, inherent high effort needed to analyse local stresses at
weld imperfections and unfavourable residual welds, but also to relatively complicated pro-
stresses. Furthermore, the structural configura- cedures for the fatigue analysis.
tion plays a major role by creating an addition-
al stress concentration. Consequently, the risk is rather high that fa-
tigue failures occur during the service period at
locations which have not been analysed with
sufficient care. This is illustrated by a fatigue
failure in a containership, which occurred after
a few years in service [1]. Fig. 2 shows a finite
element model of the forebody structure of the
ship. The coarse FEM model was created at the
design stage. The fatigue failure occurred at
the knuckled longitudinal coaming (circled in
Fig. 2) in the elevation of the upper deck.
There, a localized stress peak occurs, which is
only visible in a very fine mesh having been
analysed after the failure occurred (Figs. 2 and
3). Such a mesh refinement was not performed
during the design stage as there was no indica-
tion to the critical “hot-spot”.

Such failures occur frequently in new and un-


conventional designs, as these differ from
proven structures, e. g. in geometry, material
and/or loading. Particularly in these cases, ex-
Fig. 1 – Steel structure of a large double hull tensive fatigue analyses are necessary, taking
tanker (cargo tank area) into account the local structural geometry.

1
9th PORTUGUESE CONFERENCE ON FRACTURE - 2004

from the weld depending on the plate or shell


thickness. This development, which was re-
viewed a. o. by van Wingerde et al. [4], was
particularly successful for the fatigue strength
assessment of tubular joints due to their com-
plex joint geometry and high local bending of
the tubular walls.

First attempts to apply the approach to welded


joints at plates were already seen in the early
1980's. Remarkable are investigations per-
formed in Japan to analyse the stress concen-
tration due to the local structural geometry of
ship hull details, which were summarized a. o.
by Matoba et al. [5]. The design stress was
obtained from FEM analyses by linearization
of the stress through the plate thickness.
Fig. 2 – Coarse finite element mesh of a
Radaj [6] summarized these and other investi-
containership (forward part) and mesh re-
gations and defined the structural stress at the
finement of the critical area
hot spot (weld toe) as the surface stress which
can be calculated in accordance with structural
theories used in engineering. He demonstrated
that the structural stress can be analysed either
by surface extrapolation or by linearization,
e. g. through the wall thickness, in order to
exclude the local non-linear stress peak caused
by the weld toe.

In the early 1990's, Petershagen et al. [7] de-


rived a generalized hot-spot stress approach for
plate structures using Radaj's effective notch
stress approach [6] and applied it to complex
welded structures [8]. Detailed recommenda-
Fig. 3 – Distribution of longitudinal stresses tions concerning stress determination for fa-
in the knuckled area tigue analysis of welded components were
given by Niemi [9].
Local approaches to the fatigue assessment of
welded structures have already a relatively However, several applications showed that the
long history. According to [2], first investiga- stress results are still affected by the finite el-
tions were performed in the 1960's by several ement meshing and element properties. Addi-
researchers, including Peterson, Manson and tional recommendations for finite element
Haibach, who related the fatigue strength to a modelling and hot-spot stress evaluation were
local stress or strain measured at a certain given by Huther et al. [10] and by Fricke [11],
point close to the weld toe, for example at a the latter based on extensive round-robin stress
distance of 2 mm [3]. Although the character- analyses of several details. An alternative pro-
istic fatigue strength related to this local stress cedure for the evaluation of structural stresses
shows fairly small scatter, it is still affected by has been proposed by Dong [12]. A compre-
the local notch of the weld toe and, therefore, hensive IIW-guidance for the structural hot-
not independent from local notch geometry. spot stress approach is presently under prepa-
ration [13].
Investigations of relatively thick tubular joints
have shown that the local notch effect of the In this paper, the current procedures of the
weld toe affects the stress in the region up to analysis and assessment of structural hot-spot
0.3 - 0.4⋅t away from the weld toe (t = plate stresses are reviewed and illustrated by some
thickness). This resulted in the 1970's in the examples, showing the possibilities and limita-
development of the well-known hot-spot stress tions of the approach. It will be shown that the
approach with the definition of reference fatigue life is strongly affected by fabrication
points for stress evaluation and extrapolation, effects, which have to be rationally considered
which are located at certain distances away in the life prediction.

2
9th PORTUGUESE CONFERENCE ON FRACTURE - 2004

2. THE STRUCTURAL HOT-SPOT thickness is no more a relevant parameter for


STRESS APPROACH the definition of the reference points for stress
evaluation. In this case, absolute distances for
As mentioned above, the basic idea behind the the reference points and quadratic extrapola-
structural hot-spot stress approach is the con- tion are proposed.
sideration of the stress at the welded connec-
tion excluding the local stress peak caused by This means, that we have to distinguish be-
the weld toe. This is normally done by measur- tween different types of weld toes, which are
ing or computing the surface stresses at refer- shown in Fig. 5:
ence points outside the area affected by the
weld toe and extrapolating them to the weld a) weld toe on the plate surface at the end of
toe (Fig. 4). The resulting structural hot-spot an attachment
stress σs is then considered to determine the b) weld toe at the plate edge at the end of an
fatigue life of the welded connection. attachment
c) weld toe on the plate surface along the
weld of an attachment

The fatigue strength is assessed with design


S-N curves derived from fatigue tests, where
the structural hot-spot stress has been meas-
ured. A recent evaluation of such tests has
been performed by Maddox [15] for type a)
and c) weld toes at steel.

Fig. 4 – Local stress increase at a structural


detail (bracket)

It should be kept in mind that the approach can


be applied only to fatigue failures starting from
the weld toe. In welds with incomplete root
penetration, fatigue cracks may also grow from
the root gap, which requires an additional fa-
tigue assessment based, e. g., on the nominal
stress in the weld throat area. Fig. 5 – Types of weld toes

Refs. [13] and [14] recommend linear extrapo- Fig. 6 shows the results for type a) weld toes
lation from reference points at 0.4t and 1.0t together with the 95% confidence intervals and
distance from the weld toe in normal cases, the FAT 90 mean and design S-N curves ac-
i. e. plate structures with thickness t. If, how- cording to [14]. The results justify the classifi-
ever, the weld toe considered is not located on cation of this detail into the FAT 100 class or -
a plate surface, but at a plate edge, the plate with more conservatism - into FAT 90.

Fig. 6 – Fatigue test results for type a) weld toes [15]

3
9th PORTUGUESE CONFERENCE ON FRACTURE - 2004

Fig. 7 – Fatigue test results for type c) weld toes [15]

Corresponding results for type c) weld toes are Tab. 1 – FAT classes for structural hot-spot
shown in Fig. 7. The fatigue test results indi- stresses at different types of joints [13]
cate that FAT 90 would be appropriate. How-
Joint Description Quality FAT
ever, if the results of non-load carrying and
Butt joint As-welded,
load carrying fillet welds are analysed sepa- NDT 100
rately, it can be seen that the non-load carrying
Cruciform or K-butt
fillet welds (full symbols) show a higher fa-
T-joint with welds, no
tigue strength, which justifies FAT 100. The full penetra- lamellar 100
reason for the different behaviour is seen in a tion welds tearing
more concentrated stress flow around the weld
toe in case of load-carrying fillet welds, which Transverse attachment
non-load not thicker
results in a shorter fatigue life. This local stress
carrying than the 100
concentration does not change the structural fillet welds main plate,
hot-spot stress, so that this has to be considered as-welded
by different FAT classes. Bracket end,
welds either
For type b) weld toes, i. e. at plate edges, com- welded Fillet
parisons between hot-spot and notch stresses in around or not weld(s) 100
[16] have shown that a classification according Cover plate as-welded
to FAT 90 - 100 is appropriate. The lower ends and
class should be used for longer side attach- similar joints
ments (e. g. l > 100 mm). Cruciform
joint with
load-carrying
Table 1 summarizes the FAT classes derived. fillet welds Fillet
The 6th, 7th and 9th case include the more high- weld(s) 90
ly concentrated stress flow at the weld toe, Lap joint as-welded
resulting in the reduced FAT 90 class. In the with load-
carrying
4th and 5th case it is, on the other hand, as-
fillet welds
sumed that only part of the stress in the base Type “b” Fillet or full
plate is transferred through the fillet weld at joint with penetration
the end of a bracket or cover plate, making 100
short attach- weld,
these cases comparable to those with non-load ment as-welded
carrying fillet welds. Type “b” Fillet or full
joint with penetration
90
The FAT classes appear rather high in compar- long attach- weld,
ison with those in the nominal stress approach. ment as-welded
It should, however, be kept in mind that any
additional stress due to axial or angular misa- As misalignment is usually not considered in
lignment was included in the measured hot- numerical stress analyses, stress magnification
spot stress. This means that such stress magni- factors Km should be used in the following
fications have to be taken into account in the cases, where misalignment plays a significant
calculation of structural hot-spot stresses. role (a. o. in the cases in Fig. 8a - c):

4
9th PORTUGUESE CONFERENCE ON FRACTURE - 2004

• Butt joints made in shop in 2. using solid elements allowing the weld to
flat position and laser welds Km = 1.10 be easily modelled with prismatic ele-
• Other butt joints Km = 1,25 ments. If isoparametric 20-node elements
• Cruciform joints Km = 1.40 are applied, one element is sufficient in
• Transverse, one-sided thickness direction due to the quadratic
fillet welds Km = 1,20 displacement function and linear stress
distribution. In connection with reduced
In the nominal stress approach, a correspond- integration, the linear part of the stresses
ing stress magnification is already considered can directly be evaluated.
in the FAT classes, so that the approaches are
now compatible (e. g. a transverse stiffener
with non-load carrying fillet welds is FAT 80).

Stress magnifications due to the structural con-


figuration, e. g. the “designed misalignment”
in Fig. 8d, are generally to be considered in the
structural hot-spot stress.

Fig. 9 – Typical finite element models and


Fig. 8 – Typical misalignments: a) axial stress evaluation paths
misalignment; b) and c) angular misalign-
ment; d) designed axial misalignment The 'classical' way of evaluating the structural
stress at the hot spot is the linear or quadratic
extrapolation from two or three reference
3. FEM ANALYSIS OF STRUCTURAL points in a similar way as done experimentally
HOT-SPOT STRESSES with strain gauges. The arrows in Fig. 9 show
typical stress evaluation paths. In case of shell
In practice, relatively simple models and models without weld representation it is rec-
coarse meshes are preferred in order to limit ommended to extrapolate the stress to the
the computational effort. Basically, two types structural intersection point as modelled in
of finite element modelling are usual, which order to avoid stress under-estimation due to
are illustrated in Fig. 9 by the example shown the decreased stiffness of the model [11].
above:
For type a) and c) weld toes, the IIW recom-
1. using plate or shell elements which are mendations [13], [14] propose a linear extrapo-
arranged in the middle plane of the plates. lation over two reference points, which are
The weld is frequently omitted, except in located 0.4t and 1.0t away from the hot spot,
cases with plate offsets (e. g. doubling where t is the thickness of the adjacent plate
plates, Fig. 11a) or welds close to each (Fig. 10.1). The stresses are typically evaluated
other, where interaction effects occur. In at nodal points, so that the length of the first
such cases the weld can be modelled by element is 0.4t and the second 0.6t. A different
vertical or inclined plate elements or by stress extrapolation scheme was proposed by
rigid links (constrained equations). The some ship classification societies in connection
plate or shell elements should generally with coarser meshes and reference points 0.5t
contain improved in-plane behaviour to and 1.5t away from the hot-spot (Fig. 10.2).
model steep stress gradients.

5
9th PORTUGUESE CONFERENCE ON FRACTURE - 2004

Results obtained from the surface and through-


thickness stress evaluation are compared in
[18] and [19] for different types of weld. Fur-
ther examples showing the procedures as well
as the limitations are described below.

In case of multi-axial stresses, the largest prin-


cipal stress in a sector of ±60° from the normal
to the weld toe is assessed, if larger than the
stress component normal to the weld toe.

a) Doubling Plate

Fig. 10 – Extrapolation of surface stresses to


the hot-spot (acc. to [13]) rigid links or
plate elements
At type a) weld toes, the width of the solid b) Hopper Corner
element or the two shell elements in front of
the hot spot should not exceed either two times
the plate thickness t or the attachment width
w (= attachment thickness plus two weld leg
lengths, see Fig. 9).

The situation is different for type b) weld toes,


Fig. 11 – Modelling of eccentricities in plate
i. e. at plate edges. As plate thickness is not
element models
relevant for the element size nor the location of
the reference points, fixed reference points are
proposed. Following the proposal in [17] to
apply quadratic extrapolation over three points, 4. ROUND-ROBIN FEM ANALYSES
4 mm, 8 mm and 12 mm away from the hot OF DIFFERENT DETAILS
spot, element lengths of 4 mm or even better
2 mm are required to obtain stresses at nodal Within a Joint Industry Project, several struc-
points not affected by the stress singularity tural details were investigated in a round-robin
(Fig. 10.3). An alternative proposal with coars- study [11] by applying different finite element
er meshes was presented in [16], using mid- programs and modelling techniques. Two of
side stresses of 10 mm long elements (Fig. the details are shown here to illustrate the scat-
10.4). ter of the analysis results.

An alternative stress evaluation method, which The first is a hopper corner model, sketched in
is particularly suitable for meshes with several Fig. 12, where the marked weld toe at the
solid elements over the plate thickness, is the knuckle belongs to type c) in Fig. 5. In total,
analysis of the linear stress distribution over 13 finite element analyses were performed,
plate thickness directly at the weld toe section, most of them using meshes with element size
which is supported by some finite element t x t. The legend in Fig. 12 gives some indica-
codes. Dong [12] proposed a special procedure tions about the types of elements used in the
to derive the linear part of the stress distribu- finite element analyses (shell elements with 4
tion in the through-thickness direction. The or 8 nodes; solid elements with 8 or 20 nodes
structural stress is calculated using local stress or with higher-order p-elements; the prefix
outputs from solid elements and enforcing gives the number of solid elements in thickness
through-thickness equilibrium for the section direction and the suffix 'w' indicates the weld
at the weld toe. If shell/plate elements are ap- modelled).
plied, the distribution of the structural stress
along the weld toe can be calculated directly The right part of Fig. 12 displays the computed
from the nodal forces and moments in the ele- stresses (based on unit nominal stresses) in
ments in front of the weld toe. With nodal front of the hot spot at read-out points (ROP's)
forces and moments, greater accuracy may be selected by the participants. The hot spot is the
achieved compared to procedures using ele- weld toe or the structural intersection point if
ment stresses. the weld is not modelled.

6
9th PORTUGUESE CONFERENCE ON FRACTURE - 2004

The stress values, which are connected by If the models neglecting the offset as well as
straight lines, show a relatively large scatter. the model with p-elements are excluded, the
Additionally, the mean values of measure- scatter of the results is fairly small, i. e. ±5%
ments are included, which also showed some from the mean (2,06 ± 0,10).
scatter.
Another example from the round-robin study
The scatter in the calculation is partly caused concerns a specimen with edge gussets, as
by different element types and partly by an sketched in Fig. 13, where the marked weld toe
offset in the test model between the intersec- belongs to type b) in Fig. 5. Again 13 analyses
tion points of the horizontal plate with the have been performed. The element size was
sloped plate and the vertical plate below (Fig. governed by the plate thickness t = 10 mm,
11b), which was neglected in some calcula- except for some finer meshes.
tions with shell elements.
The computed stresses (ref. to nominal stress)
The left part of Fig. 12 shows the stresses after in the right part of Fig. 13 show again a large
linear extrapolation over 0.5t/1.5t and 0.4t/1.0t scatter, particularly close to the hot spot mod-
and also directly at a point 0.5t away from the elled, where the effect of the stress singularity
hot spot modelled. The results for the extrapo- is higher than in the first example. Again plot-
lation over 0.5t/1.5t can be divided into three ted is the mean of some measurements.
groups:
After extrapolation over reference points 5 mm
• five relatively coarse shell models yield and 15 mm away from the hot spot modelled,
the highest results between 2.28 and 2.46; most of the results are fairly close together,
the offset was neglected in these models i. e. between 1.85 and 2.08 (±6%). Smaller
which causes a stress increase by approx. values come again from finer meshes
10 %. ('Solidpw' with p-elements and fine-mesh
model '2Solid20w(f)') as well as from the mesh
• six relatively coarse shell and solid models
'Shell4(css)' with 4-node shell elements having
yield results between 1.96 and 2.16; the
only constant stress state (css). These elements
shell models consider the offset by the
are not able to model the steep stress increase
plate connection, marked by the suffix 'p'
and should therefore be avoided as recom-
in the legend of Fig. 12.
mended before.
• two models yield smaller results, i.e. the
model using p-elements ('Solidpw') and a The examples show that the extrapolation from
fine mesh solid model ('2Solid20w(f)'); the reference points acc. to the left part of Fig.
the p-elements seem to give inconsistent 10 really requires finer meshing. The use of
results, while the fine-mesh model obvi- different finite element programs, element
ously yields consistent results if the stress- types and stress evaluation procedures may
es are extrapolated over 0.4t/1.0t. result in a scatter of hot-spot stresses between
±5% and ±10% from a mean value.

Fig. 12 – Results of round-robin study for hopper corner model [11]

7
9th PORTUGUESE CONFERENCE ON FRACTURE - 2004

Fig. 13 – Results of round-robin study for hopper corner model [11]

5. EXAMPLE WITH TYPE A)


WELD TOE

The next example is taken from an investiga-


tion of different bracket connections [20]. Fig.
14 shows the test model with a diagonally act-
ing hydraulic cylinder, which produces a com-
bination of axial force, shear force and bending
moment in two horizontal and vertical I-
beams. The flange thickness is t = 20 mm.

Fig. 15 – Different bracket types and hot-


spot locations

Fig. 16 shows different finite element models


of the critical area based on above described
recommendations, where the element length in
front of the bracket corresponds to the flange
thickness t. Fig. 17 compares the computed
longitudinal stress in front of the hot spot with
measured values. Apart from the measured
stress very close to the hot spot, which is af-
fected by the local notch, the agreement is ac-
ceptable. The stresses computed with the solid
model are above those from the shell model,
however, it was found that the geometry of the
actual weld differed from the nominal values
Fig. 14 – Test arrangement for the bracket which influenced the results.
investigation [20]
The results obtained from the actual weld ge-
Three different versions of 12 mm thick brack- ometry as well as the measured values yield
ets in the frame corner were investigated, see structural hot-spot stresses which are close to
Fig. 15. In the following, version 1 is consid- each other for linear as well as quadratic ex-
ered in detail, having a rounded face bar (120 x trapolation. It should be noted that the shell
15 mm) with sniped ends at the bracket toe. model yields slightly smaller values although
The critical position is the full-penetration their distance from the hot spot refers to the
weld at the toe, which exists four times in each intersection point modelled and not to the ac-
test model. tual weld toe location.

8
9th PORTUGUESE CONFERENCE ON FRACTURE - 2004

6. NOTCHED FLAT BAR WITH


TYPE B) WELD TOE

The second example is taken from an investi-


gation of specimens shown in Fig. 18 [21],
which contain a cruciform joint with full pene-
tration welds, where the critical point is locat-
ed at the transition to the rounded edge be-
cause an additional stress concentration occurs
there. Such cases can be found in practice at
stiffener ends with cut-outs for welds.

Two series with plate thickness of 10 and


20 mm were investigated, where the transverse
plate was 20 mm thick.

Fig. 18 – Specimen with type b) weld toes


[21]
Fig. 16 – Shell and solid finite element
models of bracket version 1 with longitudi- Fig. 19 shows the finite element model for the
nal stress distribution plate thickness of 10 mm. According to the
above recommendations, the element length is
10 mm as well. The computed structural stress
concentration factor was Ks = 2.2 referring to
the nominal stress in the net section (120 mm x
10 mm). Assuming FAT 100, the nominal fa-
tigue class for this detail would be FAT 45,
where additional stresses due to misalignment
are not yet considered.

Fig. 17 – Stress distribution in front of the


hot spot of bracket version 1

The fatigue tests performed at a stress ratio R =


0 [20] showed that a full penetration weld is
necessary in this case to avoid cracks from the
weld root. However, the cracks did not start
from the weld toe, but from the weld surface.
If failure is defined when the crack reaches the
lower side of the flange (through thickness
crack), the fatigue lives were shown to be
above the FAT 100 S-N curve for all four
bracket toes so that this fatigue class is appli- Fig. 19 – Finite element model for structural
cable also to this kind of detail. hot-spot stress determination (1/4-model)

9
9th PORTUGUESE CONFERENCE ON FRACTURE - 2004

For this case, also the effective notch stress


approach [2], [14] was applied, using an ex-
tremely fine mesh to evaluate the stress in the
fictitious radius of 1 mm at the weld toe, see
Fig. 20. From the highest stress at the edge of
the rounded transition, a fatigue notch factor of
approx. Kf = 5.0 was derived, again referring
to the nominal stress in the net section. The
assumption of FAT 225 for the notch stress
approach [14] yields again a nominal fatigue
class FAT 45 for this detail. So far, the differ-
ent approaches are compatible.

Fig. 21 – Weld toe radii reproduced from


silicon casts

It was not possible to measure residual stresses


in the specimens, so that their beneficial effect
is only assumed here. However, compressive
residual stresses were measured in a similar
case in another investigation [22]. Fig. 22
shows the attachment investigated together
with the results of residual stress measure-
ments and calculations. In the as-welded state
(case 1), substantial compressive residual
stresses of approx. -200 MPa are present at the
plate edge (main plate), while tensile residual
Fig. 20 – Finite element model for effective
stresses are present at the other weld toe on the
notch stress determination (detail)
surface of the base plate.

It is interesting to note that the fatigue tests An explanation for the compressive residual
performed with pulsating stress (R = 0) stresses van be found in the longitudinal
showed a much better fatigue behaviour than shrinkage of the fillet welds at the attachment,
expected from the computations. Based on which creates in-plane bending stresses due to
nominal stresses, both test series showed char- the restraint by the relatively stiff I-beam.
acteristic fatigue strengths of FAT 90 and 100,
which were already reduced by a factor of 0.8 Resulting from this, an unexpected high fa-
according to [14] to account for unfavourable tigue strength was observed also in the fatigue
residual stresses in large structures. Unexpect- tests [22] with bending of the attachment. Even
edly the cracks appeared almost simultaneous- with high R-ratio, the observed fatigue life was
ly at the plate edge and approximately in the far above the predicted life based on the struc-
middle of the specimens. Possible causes for tural hot-spot stress.
the very high fatigue strength are seen in:

• relatively large weld toe radii at the plate


edge, where values up to 3 mm were
measured,
• favourable compressive residual stresses
at the plate edges.

This means that fabrication effects seem to


play a significant role. The different weld toe
radii are shown in Fig. 21. As the notch factor
is approximately proportional to the squareroot
of the inverse notch radius, an increase by a Fig. 22 – Measured and calculated residual
factor of two or three may result in a 40% or stresses at an attachment (case 1 - as welded;
70% higher fatigue strength. Hence, the worst case 2 - after pre-stress with 0.5 ReH; case 3 -
case assumption can be very conservative. same with 0.85 ReH) [22]

10
9th PORTUGUESE CONFERENCE ON FRACTURE - 2004

7. COMPLEX STRUCTURAL DETAILS The calculation of the structural hot-spot


WITH DIFFERENT HOT SPOTS stresses showed for lateral loads on the stiffen-
er the values given in Fig. 25. The highest
The applicability of the structural hot-spot structural hot-spot stress occurs at the vertical
stress approach needs also to be validated by flat bar connected to the T-bar (89 MPa),
the investigation of complex structural details which carries part of the load. The structural
with different, competing hot spots, for which hot-spot stresses at the bracket toes of the L-
fatigue test data are available. Two such inves- bar model are smaller (66 MPa).
tigations have been performed in the recent
past with surprising results.

The first investigation concerns the third


bracket version shown already in Fig. 15,
where another critical hot spot exists at the end
of the buckling stiffener (pos. 2) in addition to
the bracket toe. The computed structural hot-
spot stresses in Fig. 23 show that the hot-spot
stress at pos. 2 is about three times higher than
at pos. 1. The measured strains gave similar
results [20]. The high stresses are mainly due
to high local plate bending.

Fig. 24 – Intersections between stiffeners


and transverse web [23]

Fig. 23 – Structural hot-spot stresses for the Fig. 25 – Computed structural hot-spot
different bracket toes (solid models, nom. stresses for the intersections between stiff-
weld dimensions, linear stress extrapolation) eners and transverse web [23]

The load for test model 3 was accordingly re- Here again, the fatigue tests showed a different
duced. However, no crack appeared after the picture. The L-bar model had the shortest fa-
anticipated fatigue life. The test was continued, tigue life, while the T-bar model showed a
and after a long period the first crack appeared better fatigue behaviour. In the latter, the first
at one of the four bracket toes (pos. 1). It is crack appeared in the flange of the T-bar and
still not fully clear why pos. 2 was not critical. not in the vertical flat bar.
However, in view with the aforementioned
results, compressive residual stresses might This means that the application of the structur-
again be a major reason for this. al hot-spot stress approach would lead to a
wrong conclusion - with the possible conse-
The other investigation concerns a comparison quence of arranging thousands of unnecessary
of different intersections between stiffeners brackets in a tanker structure. Again, further
and a transverse web [23], see Fig. 24. This influence factors seem to exist which are not
detail is particularly well-known from tanker considered in the structural hot-spot stress ap-
structures. The theoretical and experimental proach, for instance, favourable residual
investigation should answer the question, if the stresses.
connection with T-bar (lower part of Fig. 24)
needs additional brackets which were shown to The connection with the T-bar is very similar
be necessary for L-bars (upper part of Fig. 24) to the attachment shown in the previous chap-
due to unfavourable secondary bending. ter, where compressive residual stresses were
found at the weld toe on the edge of the at-
tachment.

11
9th PORTUGUESE CONFERENCE ON FRACTURE - 2004

8. PROBLEMS AND LIMITATIONS The round robin analyses described in 4. have


OF THE STRUCTURAL HOT-SPOT shown that the related uncertainties are in the
STRESS APPROACH order of ±5% to ±10%. However, not only the
meshing in front of the hot spot, but also in the
The stress computation of several structural vicinity of this area affects the results. This is
details and the comparison with tests have re- illustrated by the bracket example described in
vealed problems which are discussed in the 5. As shown in Fig. 26, the meshing of the
following in more detail. bracket itself affects the stresses in the hot-spot
area. A relatively coarse mesh yields an ap-
Neglected Influence Factors prox. 15% higher structural hot-spot stress.
This means that an additional scatter in the
Particularly some of the examples described finite element results due to the meshing
above have shown that the life prediction can around the critical area may occur.
be wrong compared to fatigue tests. The fol-
lowing influence factors on fatigue, which are
usually neglected or only roughly considered,
play certainly a role:

• Differences between the assumed and the


actual weld geometry, e. g. increased weld
leg length or weld toe radii in the exam-
ples mentioned in 5. and 6. The influence
of the actual weld geometry can quite well
be taken into account by the effective
notch stress approach, which has been
demonstrated a. o. in [2]. Fig. 26 – Stress distribution in front of the
• Differences between the assumed and the bracket toe for two different meshes of the
actual residual stress. Normally, tensile bracket
stresses up to the yield point are assumed,
which means that in case of compressive Still unclear is the procedure in connection
residual stresses the fatigue life can be with very thick components, e. g. bulbs of pro-
considerably longer. files. Here, a plate or shell model seems to be
• The effect of plate thickness, which has not suitable. But also for solid models it is not
been proven by experiments for butt yet clear how the element size and reference
welds, cruciform joints and transverse points for stress evaluation shall be defined.
stiffeners. However, not much is known One possibility would be the utilisation of the
about this effect in complex structures. attachment width w acc. to Fig. 9.
• The effect of the stress gradient in plate
It can be expected that this will be investigated
thickness direction, which is neglected in
further in the near future. In particular, refined
the structural hot-spot stress and in other
approaches will offer new possibilities in this
approaches, although it affects the fatigue
respect.
life. The stress gradient can, for example,
be considered by fracture mechanics,
Joints with Fillet and Partial Penetration
which was applied e. g. in [12].
Welds
How far these factors affect the life prediction
The examples did only cover cracks starting
was demonstrated in the examples in 7., where
from the weld toe, for which the structural
completely wrong conclusions were drawn
stress approach is applicable. As mentioned
regarding the assessment of the whole detail.
before, cracks at joints with fillet or partial
penetration welds might alternatively start
Modelling Aspects
from the weld root. The fatigue assessment of
such cracks is possible either with the nominal
Differences between the idealized and the ac-
stress in the weld and a corresponding fatigue
tual structure can affect the results as well. For
class (FAT 45 acc. to [14]) or with a refined
instance, plate modelling according to the up-
approach such as the effective notch stress
per part of Fig. 9 yields the same result for
approach or the crack propagation approach.
different weld throat thicknesses. Also the
modelling of plate offsets as shown in Fig. 11
is possible only with limitations.

12
9th PORTUGUESE CONFERENCE ON FRACTURE - 2004

In any case, the models shown in Fig. 9 are not [5] Matoba, M.; Kawasaki, T.; Fujii, T. and
suitable for the determination of the relevant Yamauchi, T.: “Evaluation of fatigue
stress. A complex stress distribution occurs in strength of welded structures - hull's
particular at fillet-welded attachment ends. members, hollow section joints, piping
Further investigations about this case are cur- and vessel joints.” IIW-Doc. XIII-1082-
rently in progress and will be described else- 83, International Institute of Welding,
where. 1983.
[6] Radaj, D.: “Design and analysis of fa-
tigue-resistant welded structures.”
9. CONCLUDING REMARKS
Abington Publ., Cambridge, 1990
At present much effort is spent on the determi- [7] Petershagen, H.; Fricke, W. and Massel,
nation of stresses in relatively complex struc- T.: “Application of the Local Approach
tures with sufficient accuracy and their as- to the Fatigue Strength Assessment of
sessment with respect to fatigue. The structural Welded Structures in Ships.” IIW Doc.
hot-spot stress approach is applied to an in- XIII-1409-91, International Institute of
creasing extent, however, in several cases with Welding, 1991.
relatively simple and coarse models. In the
[8] Fricke, W. and Petershagen, H.: “Detail
paper, modelling aspects, the stress evaluation
design of welded ship structures based on
and extrapolation as well as the fatigue as-
hot spot stresses.” In: Practical Design of
sessment are illustrated by several examples.
Ships and Mobile Units, Eds: Caldwell,
Also some problems and limitations of the J.B. and Ward, G., Elsevier Science,
1992.
fatigue prediction with this approach are dis-
cussed. Neglected influence factors and simpli- [9] Niemi, E., Ed.: “Recommendations con-
fied modelling can cause large deviations from cerning stress determination for fatigue
the actual lifetime. analysis of welded components.” Abing-
ton Publ., Cambridge, 1995.
Another problem is the assessment of cracks
starting from unwelded root gaps, which would [10] Huther, I., Gorski, S., Lieurade, H.P.,
require additional effort in modelling of com- Laborde, S. and Recho, N.: “Longitudi-
plex welded structures. Here, practical proce- nal non loaded welded joints geometrical
dures are under development. stress approach.” Welding in the World,
43:3 (1999), pp. 20 -26.
[11] Fricke, W.: “Recommended hot spot
10. REFERENCES analysis procedure for structural details
of ships and FPSOs based on round-robin
[1] Fricke, W. and von Selle, H.: “Verifica- FE analyses.” Int. J. of Offshore and Po-
tion of the Integrated Fatigue Analysis by lar Engng., Vol. 12 (2002), No. 1, pp 40 -
Selected Failure Cases (in German).” In: 47.
„Entwicklungen in der Schiffstechnik“,
Statusseminar 1999, TÜV-Verlag GmbH, [12] Dong, P.: “A structural stress definition
Köln 1999. and numerical implementation for fatigue
analyses.” Int. J. Fatigue, 23:10 (2001),
[2] Radaj, D. and Sonsino, C.M.: “Fatigue pp. 865 - 876.
assessment of welded joints by local ap-
proaches.” Abington Publ., Cambridge, [13] Niemi, E.: “Structural Stress Approach to
1998. Fatigue Analysis of Welded Components
- Designer’s Guide.” IIW-Doc. XIII-
[3] Haibach E.: “Fatigue Strength of Welded 1819-00/XV-1090-01 (Final Draft), In-
Joints from Viewpoint of Local Strain ternational Institute of Welding, 2001.
Measurement (in German).” Report FB-
77, Fraunhofer-Institut für Betriebsfes- [14] Hobbacher, A.: “Fatigue Design of
tigkeit (LBF), Darmstadt 1968. Welded Joints and Components.” Abing-
ton Publishing, Cambridge (UK), 1996.
[4] van Wingerde, A.M.; Packer, J.A. and
Wardenier, J.: “Criteria for the fatigue [15] Maddox, S.J.: “Hot-Spot Stress Design
assessment of hollow structural section Curves for Fatigue Assessment of Weld-
connections.” J. Construct. Steel Res., 35 ed Structures.” Intl. J. Offshore and Polar
(1995), pp. 71 - 115. Engng. 12 (2002), pp. 134 - 141.

13
9th PORTUGUESE CONFERENCE ON FRACTURE - 2004

[16] Fricke, W. and Bogdan, R.: “Determina- [20] Paetzold, H.; Doerk, O. and Kierkegaard,
tion of hot spot stress in structural mem- H.: “Fatigue Behaviour of Different
bers with in-plane notches using a coarse Bracket Connections.” In: Practical De-
element mesh.” IIW-Doc. XIII-1870-01, sign of Ships and Other Floating Struc-
International Institute of Welding, 2001. tures (Ed. Y.-S. Wu, W.-C. Cui and G.-J.
Zhou), Elsevier 2001.
[17] Niemi, E. and Tanskanen, P.: “Hot spot
stress determination for welded edge [21] H. Petershagen: The Hot Spot Stress Ap-
gussets.” Welding in the World, 44:5 proach - a Case Study (in German).
(2000), pp. 31 - 37. Schiffbauforschung 39 (2000), Issue 3.
[18] Doerk, O.; Fricke, W. and Weissenborn, [22] W.S. Kim, D.H. Kim, S.G. Lee and Y.K.
C.: “Comparison of Different Calculation Lee: “Fatigue Strength of Load-Carrying
Methods for Structural Stresses at Weld- Box Fillet Weldment in Ship Structure.”
ed Joints.” Int. J. of Fatigue 25 (2003), In: Practical Design of Ships and Other
pp. 359 - 369. Floating Structures (Ed. Y.-S. Wu, W.-C.
Cui and G.-J. Zhou), Elsevier 2001.
[19] Poutianen, I.; Tanskanen, P.; Martinsson,
J. and Byggnevi, P. : “Determination of [23] Park, J.; Kim, K.B.; Kim, W.S. and Kim,
the Structural Hot-Spot Stress using the D.H.: “An Experimental Investigation on
Finite Element Method - a Comparison Fatigue Behaviour of Inverted Angle and
of Current Procedures.” IIW-Doc. XIII- T-Type Side Longitudinals in Tankers.”
1991-03 / XV-1448-03, International In- In: Practical Design of Ships and Other
stitute of Welding, 2003. Floating Structures (Ed. Y.-S. Wu, W.-C.
Cui and G.-J. Zhou), Elsevier 2001.

14

View publication stats

Вам также может понравиться