You are on page 1of 18

\

PERGAMON Engineering Failure Analysis 5 "0888# 002029


S02495296:88:, ! see front matter 0888 Published by Elsevier Science Ltd[ All rights reserved
PII] S0 2 4 9 5 2 9 6 " 8 7 # 9 9 9 1 4 8
Fatigue design of welded aluminum rectangular hollow
section joints
K[A[ Macdonald\ P[J[ Haagensen
Norwegian Unirersiiy of Science ana Technology, Deparineni of Siruciural Engineering, Tich. Birkelanasrei 1a\
N-7034 Tronahein\ Norway
Received 6 August 0887^ accepted 7 September 0887
Abstract
Fatigue design methods for welded aluminum joints are reviewed\ including various approaches to fatigue
life estimation currently adopted in design codes across a range of industrial applications[ The applicability
of these established methodologies to the fatigue design of automotive space frame structures is critically
assessed[ The hot spot stress method is identi_ed as the most promising in terms of providing a coherent
and comprehensive approach to design[ Particular problems related to implementation are considered such
as failure sites and determination of appropriate stress concentration factors from physical models\ _nite
element calculations or parametric equations[ Preliminary results from _nite element stress analyses and
fatigue tests are also presented for rectangular hollow sections welded in a T!joint con_guration[ Recom!
mendations are made for a design methodology for welded rectangular hollow!section joints in aluminum
space frames\ including use of a single hot spot SN curve[ 0888 Published by Elsevier Science Ltd[ All
rights reserved[
Keyworas. Automotive design^ Fatigue design^ SN curves^ Space frames^ Weld fatigue
0[ Introduction
Aluminium welded hollow section "RHS# joints are _nding increased use in crane and bridge
structures\ transport vehicles and in automotive structures[ Tubular structures are occasionally
fabricated using forged or cast nodes but these are more expensive to produce than welded inter!
tube connections which are consequently more common[ However\ the fatigue design basis for
welded RHS joints in aluminium is limited and no design recommendations currently exist for
such joints[ An appropriate starting platform from which to establish a fatigue design methodology
could be the experience recently gained with similar structures fabricated in steel[ In particular\
fatigue assessment based on the geometric hot spot stress range "or hot spot stress# concept could
Corresponding author[
K.H. Macaonala\ P.1. Haaqensen,Enqineerinq Failure Hnalysis 6 (1998) 113130 003
be adopted for aluminium structures[ However\ in automotive design the material thicknesses tend
to be smaller than those used in civil engineering design and this di}erence might limit the
possibilities for transfer of data from the steel industry[
This paper gives a brief overview of fatigue assessment procedures used in the design of steel
tubular structures[ The limited SN data available for aluminium RHS joints are presented and
the need for further experimental research is outlined[
Trends in the development of design codes are also discussed in light of the fact that international
codes are rapidly harmonising through co!operation between organisations such as ISO "Inter!
national Standards Organisation#\ CEN "Comite Europeen de Normalisation# responsible for
Eurocodes\ IIW "International Institute of Welding# and API "American Petroleum Institute#[
1[ Fatigue life assessment methods for welded tubular joints
In the majority of current fatigue design codes there are two generic types of SN curve used[
In conventional welded structures involving plates and beams\ the noninal siress approach is
normally employed where the di}erent fatigue behaviour of various structural elements or details
is described by assigning to them di}erent SN curves\ termed design categories or classes\ and
combining these with nominal stresses remote from the weld[ Since fatigue failure in welded
constructions is not only related to geometry\ the direction of loading and failure site also in~uences
joint categorisation[ Examples of design categories of such basic connections are joints with butt
or _llet weld\ having di}erent design categories dependent on the stress direction[ The main
advantage of the nominal stress approach is that the SN curves for each weld category "or weld
class# include the notch e}ect of the weld as well as the e}ect of the component geometry[ The
main sources of scatter due to fabrication variables are thus included in the test data used to create
the curve[ Nominal stresses are calculated from a structural analysis and comprise the membrane
stress range\ S
m
and the bending stress range S
b
in general SS
m
S
b
[ The complexities of and
interactions between geometry and loading found in some structures\ e[g[\ in circular section
tubular joints\ give rise to a plethora of possible failure sites and here the hoi spoi siress approach
is used to reduce the design to a common basis[ Additionally\ the di.culties encountered in de_ning
a nominal stress in complex structures also make the hot spot approach the only practicable
method[ Fatigue failure will occur at sites of high peak stress in such joints and it is assumed that
fatigue life is related to the magnitude of these peaks[ Fatigue design is accomplished by combining
knowledge of these local stress peaks*usually in the form of a stress concentration factor "SCF#
for a particular load con_guration*with an SN curve representing a simple weld without any
structural SCF[
2.1. The geoneiric hoi spoi siress approachcircular hollow-seciion (CHS) joinis
The hot spot approach has been used extensively in the o}shore industry in the analysis of
results from tests on steel tubular joints 0[ Stresses in circular hollow!section "CHS# joints arise
from three main sources]
"i# The basic stress response due to the global action of the remote applied load\ i[e[\ the stress
K.H. Macaonala\ P.1. Haaqensen,Enqineerinq Failure Hnalysis 6 (1998) 113130 004
that can be calculated from a frame analysis disregarding the stress concentrating e}ects of
both the joint and the weld[
"ii# The geometric stresses resulting from local bending of the tube walls to maintain compatibility
between the members[
"iii# The highly local stress near the intersection lines between members[ These local stresses are
strongly in~uenced by the weld shape[
The maximum stress "or notch stress#\ located at the weld toe is the sum of the nominal stress\ the
geometric stress and the local stress components[ The geometric hot spot stress range or hot spot
stress concept has evolved as the most practical basis for fatigue design of welded joints[ It captures
the _ctitious local stress that characterises the fatigue performance of the joint\ but excludes the
very local stress perturbations caused by changes in the weld toe geometry[ Also excluded are the
e}ects of any undetectable defects[ These e}ects are included in the SN curve[ The hot spot stress
concept places di}erent structural geometries on a common basis\ enabling the use of a single
SN curve[ The hot spot stress S
h
is related to the global loads in the structure through]
S
h
SCFS
nom
"0#
where S
nom
is the nominal stress range and the stress concentration factor "SCF# is normally
obtained from either _nite element analyses or from strain gauge measurements[ It is important
that consistency with the SN curve is maintained by using the same method for estimating the
hot spot in the fatigue test as used in obtaining SCFs[ The hot spot stress method for steel CHS
joints has been validated by SN data for di}erent types of joints and loading conditions that plot
into a single scatter band 0[
There is general agreement that the hot spot is located at the weld toe but there are many
opinions as to the proper method of determining the hot spot from strain gauge measurements[ In
the early US practice for o}shore structures\ the API and AWS codes de_ned the hot spot stress
range as the total stress range measured by a strain gauge placed adjacent to the weld toe\
perpendicular to the weld[ Therefore an attempt was made to measure the maximum stress at the
weld toe\ including the notch e}ect of the weld[ Typically hot spot gauges were placed within
5 mm to 9[0zri of the weld toe with a gauge length of 2 mm\ r and i referring to the outside radius
and thickness of the instrumented member[
In the European Coal and Steel Community "ECSC# method 1\ also for o}shore steel structures\
an extrapolation is made from two strain gauges placed just outside the weld notch zone and in
the region of stress linearity to determine the geometric hot spot stress range[ This method is also
used for non!linear stress distributions[ In the Det Norske Veritas "DNV# method 2 for o}shore
structures\ many strain gauges are placed near the weld to allow for a more accurate determination
of the region of hot spot stress[ The ECSC and the DNV recommended locations for the strain
gauges are shown in Fig[ 0[ The ECSC de_nition is based on the maximum principal stress\ i[e[\
the stress components are extrapolated to the weld toe and the maximumprincipal stress calculated[
The stress normal to the weld used in the US de_nition is somewhat lower than this but in the
regions of highest stress\ the crown and saddle location the two are almost identical[
In the IIW design recommendations 3\ a non!linear "quadratic# extrapolation procedure is
recommended for cases of high local shell bending stresses caused for example by eccentric
attachments in large diameter tubes or plane plates[ The quadratic extrapolation requires a
K.H. Macaonala\ P.1. Haaqensen,Enqineerinq Failure Hnalysis 6 (1998) 113130 005
Fig[ 0[ Recommended location of strain gauges to measure hot spot stress in tubular joints\ ECSC "top# and DNV
"bottom# 1\ 2[
minimum of three strain gauges and is based on points on a curve\ _tted to the strain gauge
measurements by regression analysis[
The SCF is determined by several methods]
"i# Physical models using strain gauges\ as outlined above[
"ii# Finite element methods using appropriate meshes which model the overall joint geometry
without actually modelling the weld[
"iii# Parametric formulae based on either strain gauge measurements or FEM[
In the _nite element analysis\ care must be taken to obtain stresses at positions for extrapolation
that are consistent with the de_nition of the hot spot stress used[ In the IIW recommendations 3
detailed instructions are given for the FEM analysis[
The value of the SCF is critical for the accuracy of the predicted life\ and large e}orts have
therefore been made to assess available data to de_ne sets of parametric equations for use in design
codes[ In the latest Health and Safety Executive revision of the guidance for o}shore structures
K.H. Macaonala\ P.1. Haaqensen,Enqineerinq Failure Hnalysis 6 (1998) 113130 006
4\ two sets of parametric equations are recommended for design of tubular CHS joints 5\ 6[ A
comprehensive survey of the changes in the HSE guidance has recently been presented 7[ Changes
similar to those in the HSE guidance are likely to be implemented in the API code 8 for o}shore
structures which is currently undergoing extensive revisions which are aimed at producing a new
ISO code for o}shore structures[ The new ISO code is in turn intended to become the Eurocode 2
"steel structures# fatigue clauses for o}shore structures 09[ Since there is a trend towards har!
monisation of European design rules\ the new ISO provisions in Eurocode 2 for hot spot based
fatigue assessment of tubular structures are expected to in~uence the hot spot assessment pro!
cedures in future versions of Eurocode 8 for aluminium structures 00[
The SN curve to be used with the hot spot concept is usually identical to or very similar to the
SN curve for a simple butt weld[ In the early "0873# Department of Energy guidance on which
NS 2361 01 for steel structures are based\ the T!curve for tubular joints is almost identical to the
D!curve which applies to transverse butt welds in plate structures[ This follows from the de_nition
of the hot spot stress[ For a transverse butt weld in a plate or pipe there is no geometric stress
concentration and the SCF0[ Therefore the associated hot spot stress SN curve should only
include the e}ect of the weld\ i[e[\ the SN curve for a transverse butt weld should be used[
The physical size of the joint under consideration is important because the size of the strain
gauge limits the distance from the weld to the locations of the gauge[ In thin!walled structures the
gauge has therefore to be placed farther away from the weld than has been possible in o}shore
steel joints\ and special guidance has therefore evolved for thin section welded joints[
The e}ect of plate thickness is usually handled the same way in design recommendations for the
hot spot stress method as for the nominal stress method[ A thickness penalty factor is imposed for
plate thicknesses greater than the reference thickness\ i
9
[ The penalty factor\ when applied to the
stress range\ is usually expressed as]
S
S
9

0
i
9
i 1
q
"1#
where S is the stress range at thickness i[ In most of the older rules the value of thickness exponent
q is 9[14[ However\ assessment of recent research data has indicated stronger in~uence of thickness
and in the latest HSE and API:ISO revision for o}shore structures a higher penalty factor of
q 9[29 is given[
2.2. Teciangular ana square hollow-seciion joinis
The considerable knowledge gained regarding the static response of circular hollow!section
joints is unfortunately not directly transferable to rectangular hollow!section joints because of
marked di}erences in the behaviour of the latter 02[ The sti}ness distribution in the ~at sidewalls
is di}erent from that of cylindrical shells and the brace corners near to the chord walls have a
strong in~uence on the maximum stress[ Interest in joints of this type has been con_ned to those
made in steel 03 and recommendations have been made on the selection of hot spot stress
de_nition\ parametric equations for SCFs and on appropriate SN curves for RHS joints 04[ In
contrast\ static stress and fatigue guidance for RHS joints in aluminium are absent in the literature
and little data exists[ Some fatigue test results have been reviewed 05\ but the data are not analysed
K.H. Macaonala\ P.1. Haaqensen,Enqineerinq Failure Hnalysis 6 (1998) 113130 007
Fig[ 1[ Notch hot spot stress 06[
in terms of local or hot spot stress\ but these data nonetheless indicate that a design hot spot stress
range "FAT value# of 39 MPa would be applicable to welded square hollow!section joints[
2.3. The hoi spoi siress concepi appliea io plaie joinis
Recent fatigue design guidance for welded details in steel ship structures 06\ and similar
guidance for aluminium ships 07\ have in both cases adopted a hot spot stress approach[ This is
due mainly to the geometric complexity of the internal hull structure of modern ships that makes
it di.cult to de_ne nominal stresses[ The guidance favours the maximum local stress at the weld
toe "called the notch stress# over the extrapolated geometric stress described earlier^ see Fig[ 1[ The
e}ect of the local weld geometry is therefore included in the notch stress\ dictating that the same
e}ects are excluded from the SN curve\ which is consequently elevated above that for a butt weld
by a factor of approximately 0[4[
The notch stress concept is attractive in that it is an actual stress\ in contrast to the _ctitious
extrapolated hot spot stress already described[ In both practical and numerical applications\
however\ the notch stress approach is problematic[ The two most important parameters in~uencing
the stress concentration factor for the weld are highly variable along the weld length\ the weld toe
radius\ r\ and the weld angle a\ and these parameters can only be readily described in terms of
statistical distribution[ Therefore\ in a deterministic _nite element analysis the variations in r and
a cannot be modelled\ and some descriptive values "average or maximum# are employed as default
values[ In the DNV rules for welded steel ship structures 07 these default values are r,i 9[04
and a 34> for butt welds\ where i plate thickness[ Similar problems are encountered when the
notch stress is to be determined experimentally from strain gauge measurements since it cannot be
measured directly at the weld because the strain gauge would have to straddle the weld toe[ Instead
an extrapolated hot spot stress is obtained from strain gauges placed at distances of 9[4i and 0[4i
from the weld[ To account for the stress concentration e}ect of the weld the hot spot stress is
K.H. Macaonala\ P.1. Haaqensen,Enqineerinq Failure Hnalysis 6 (1998) 113130 008
Fig[ 2[ The DNV design SN curves for aluminium ships 07[
multiplied by a default SCF of K
w
0[4 for the weld[ Thus the DNV notch stress approach is
e}ectively the same as the hot spot approach since a factor of 0[4 is _rst applied to the SN curve\
then it is accounted for by multiplying the hot spot stress by the same factor of 0[4[
2.4. Hoi spoi SN curres for aluniniun joinis
In the DNV guidance for welded aluminium ship structures 07 two SN curves are given for
welded joints "three curves including one for welded joints in a corrosive environment#\ as shown
in Fig[ 2[
The applied stress range is the noich siress ranqe as de_ned in Fig[ 1[ Curve I applies to base
material[ The highest curve for welds "Curve II# represents butt welds while the lower curve\ Curve
III\ which also has a steeper slope "n2[26#\ represents _llet welds as low strength "high severity#
joints[ Curve IV is for welds in a corrosive environment[ Since the slope of Curves II and III is
close to the value of n2 in the IIW guidance\ corresponding SN curves\ characterised by their
FAT value "FATstress range at N109
5
# can be obtained from the DNV curves[ The stress
range of curve III at 109
5
cycles is 34 MPa but this includes the K
w
factor of 0[4 so the
corresponding FAT value becomes 34:0[4 29 MPa[ Similarly\ the FAT value of Curve II is
44:0[4 26 MPa[ These curves can now be compared with other curves for aluminium welded
joints based on hot spot stresses[
SN data from a large number of welded aluminium plate specimens with a variety of geometries
and in thicknesses up to 5 mm have recently been collected 08 and analysed in terms of hot spot
stresses obtained either from strain gauge measurements or from _nite element analyses[ The
K.H. Macaonala\ P.1. Haaqensen,Enqineerinq Failure Hnalysis 6 (1998) 113130 019
Fig[ 3[ SN data for small MIG welded aluminium plate specimens 08[
results plotted as hot spot stresses are summarised in Fig[ 3[ It was concluded that a weld category
curve corresponding to FAT 39 would constitute a conservative hot spot based design curve for
welded aluminium structures[
In the US the Category B curve in the Aluminum Association Design Manual has recently been
proposed 19 for use as a general design hot spot stress based design curve for welded aluminium
joints[ This is the SN curve for longitudinal butt welds[ In Eurocode 8 00 and in BS 7007 10\
the hot spot stress method is mentioned but no advice is given on the choice of SN curve[
The DNV classi_cation note for ships 07 is the only known code that uses the hot spot
stress concept for welded aluminium structures[ However\ by analogy with design codes for steel
structures where the hot spot stress SN curves are nearly identical to the curves for two sided butt
welds\ SN curves for hot spot stress fatigue design can be obtained from other codes[ SN curves
obtained in this way are compared in Table 0[
2[ Finite element analyses
The determination of the state of stress experienced by weldments is critical in design\ but
especially so where the hot spot stress method is used[ Appropriate SCFs are pivotal in analysis
experimental fatigue data for establishing hot spot SN curves and are also critical in terms of
relating structural or nominal stresses to hot spot stresses for use in fatigue design[ The work
reported here is primarily of an exploratory nature and was _rstly aimed at determining the SCF
K.H. Macaonala\ P.1. Haaqensen,Enqineerinq Failure Hnalysis 6 (1998) 113130 010
Table 0
Possible hot spot stress SN curves for fatigue design of welded aluminium structures
Source Stress range Design Slope\ Comments
at two million category n
cycles\ S
1 mill
"MPa#
Eurocode 8 00 24 Class 24 3[9 Curve for two!sided butt welds
BS 7007 19 24 Class 24 2[1 Curve for two!sided butt welds
IIW 3 21 or 39 Fat 21 or 39 2[9 Curves for two!sided butt welds
Sharp et al[ 19 35 AA Cat[ B 3[64 Various types of welds
Partanen and Niemi 08 35 * 2[9 Various types of welds
Det Norse Varitas 07 29 or 26 Curve III or II 2[26 or 3[21 Fillet or butt welds\ respectively
Kosteas and Gietl\ 0884 05 39 * 2[4 Based on data for hollow section
joints
for welded aluminium RHS T!joints\ Fig[ 4\ under simple membrane and bending loading of the
chord member\ and secondly to establish an SN curve for this joint[
3.1. Deierninaiion of SCFs
Finite element analyses were performed using the I!Deas and Abaqus commecial codes[ Three!
dimensional solid models were used representing a quarter of the test specimen geometry by
employing symmetry planes[ Geometrically linear elastostatic analysis routines were used[ Two
sets of models were used] I!Deas based modelling of parametric variations in local weld geometry^
Abaqus based analyses of the nominal weld geometry to independently verify the SCFs determined[
The two series of analysis separately employed 09!node tetrahedral and 19!node brick elements[
Meshing rules suggested by IIW 11 were employed and the stresses for extrapolation were
also extracted in accordance with these guidelines in order to maintain consistency between FE
predictions and experimental stress analyses[
The following nominal parameters were used for the model] weld toe radius\ r 9[7 mm^ weld
angle\ a 34>^ weld throat size\ a 2 mm^ tube wall thickness\ i 2 mm[ Subjected to pure
bending of the chord member\ a weld toe SCF value of 0[82 was calculated based on extrapolation[
The site of maximum stress along the weld toe is located near to the corner of the brace member
towards the sidewall of the chord[ An entirely separate FE analysis of the same geometry using a
di}erent program and di}erent solid elements produced SCF values of 0[75 in bending and 0[40
under membrane loading[
3.2. Paraneiric siuay
In an investigation of the sensitivity of the calculated SCF to parametric variations in the weld
and section geometry\ the four parameters r\ a\ a and i were systematically varied[ Identical tube
K.H. Macaonala\ P.1. Haaqensen,Enqineerinq Failure Hnalysis 6 (1998) 113130 011
Fig[ 4[ Specimen dimensions[
geometries were used on brace and chord members in a given analysis and during variation of i
the external dimensions remained constant[ The toe radius\ r\ was varied between 9[22[9 mm^ a
was varied from 2444>^ a ranged between 13 mm^ and i was varied between 12[4 mm[ Statistical
analysis of the resulting distribution of SCF
:
"in this case de_ning the maximum stress at the weld
toe#\ based on both maximum principal stress and axial stress component\ revealed that r and i
have the strongest in~uence\ while the SCF is only a weak function of both a and a[ Examples are
given in Fig[ 5[ Models _tted to the data produced the eqns]
SCF
:
0.689.221a9.62ie
1r
9.22ia\ "2#
SCF
s0
0.069.648a9.787ie
1r
9.05a\ "3#
K.H. Macaonala\ P.1. Haaqensen,Enqineerinq Failure Hnalysis 6 (1998) 113130 012
Fig[ 5[ Examples of parametric variation of SCF "radius and toe angle#[
with a in radians and i\ r and a in mm[ Equation "2# refers to the axial "Z!direction# stress
component and eqn "3# refers to maximum principal stress at the weld toe[ The bounds investigated
"stated above# strictly apply as limits to these equations[ These models behave in a similar way to
established treatments for transverse _llet welds in plate 12 but\ as the joint under considerations
here is between hollow!sections\ there is strictly no valid basis on which to make a direct com!
parison[ The variation of SCF with both r and i is plotted in Fig[ 6 where it is clear that\ as could
be foreseen\ combinations of small weld toe radius and large weld angle\ weld throat and wall
thickness lead to high SCFs[ In particular\ combinations of r and i have most in~uence[
3[ Fatigue tests
Fatigue testing of T!joint specimens was performed on a series of four specimen groups covering
two weld metals "3932 and 4072#\ and as!welded and improved "toe!ground# weldments[ The
specimen geometry\ Fig[ 4\ is identical to that studied earlier using FE^ however\the unloaded
K.H. Macaonala\ P.1. Haaqensen,Enqineerinq Failure Hnalysis 6 (1998) 113130 013
Fig[ 6[ SCF as a function of r and i[
brace member was shortened in order to facilitate testing[ The parent material extrusions were
5971[15!T4 aluminium alloy that has speci_ed minimum tensile properties of 189 MPa yield
strength and 239 MPa ultimate tensile strength[ The specimens were loaded in 3!point bending of
the chord member[ All tests were carried out at T9[0 and under constant amplitude conditions
using a 4 Hz sinusoidal waveform[ The nominal stresses were calculated from simple elastic
bending theory for the loaded cross!section[ The hot!spot stresses were determined from a simple
multiplication of the nominal stresses by an SCF of 0[72\ derived from the FE work described
earlier[ A limited number of specimens were instrumented with strain gauges on the chord\ located
close to the anticipated failure site using the IIW guidelines 3 in order to measure the hot spot
strain\ Fig[ 7[ A polynomial curve _tted to the strain data was linearly extrapolated to the weld
toe[ Good agreement was found between the hot spot stresses based on experimental "SCF0[72#
and numerical "SCF0[82# stress analyses[ Endurance data generated in the test programme are
given in Tables 14\ presented in terms of both nominal and hot spot stress ranges\ using an SCF
of 0[72[
The data exhibited no clear dependence on the _ller metal so these data were combined[ Linear
regression analysis of the test data produced the design lines "mean minus two standard deviations#
plotted in Fig[ 8] the corresponding SN constants are noted in Table 5[
Some data are available in the literature 13 for welded T!joints in aluminium hollow pro_les[
Directly relevant data reviewed by Kosteas and Gietl 05 were already expressed as local stress
K.H. Macaonala\ P.1. Haaqensen,Enqineerinq Failure Hnalysis 6 (1998) 113130 014
Fig[ 7[ Strain gauge locations[
Table 1
Results from fatigue testing of as!welded 4072 _ller metal specimens
Nominal stress Hot spot stress Endurance
range "MPa# range "MPa# "cycles#
091[8 077[2 83[819
74[6 045[7 101[793
74[6 045[7 116[698
74[6 045[7 089[166
57[5 014[4 593[533
57[5 014[4 090[664
57[5 014[4 143[022
57[5 014[4 188[567
57[5 014[4 232[999
40[3 40[3 0068[214
ranges but the Hagstro m and Sandstro m data had to be reevaluated in terms of hot spot stress[
This involved performing an FE analysis on the test geometry used "axial loading of the brace# to
get an SCF value of 3[3[ These data are all plotted in Fig[ 09 based on hot!spot stress range with
the results of the present study for as!welded joints where it is clear that the hot spot stress approach
appears to be performing well in terms of reducing data from di}erent loading con_gurations to
a common basis[
K.H. Macaonala\ P.1. Haaqensen,Enqineerinq Failure Hnalysis 6 (1998) 113130 015
Table 2
Results from fatigue testing of toe!ground 4072 _ller metal specimens
Nominal stress Hot spot stress Endurance
range "MPa# range "MPa# "cycles#
019[9 108[5 092[150
019[9 108[5 098[075
019[9 108[5 71[953
74[6 045[7 218[867
74[6 045[7 328[130
74[6 045[7 304[113
74[6 045[7 306[681
57[5 014[4 0897[528
Table 3
Results from fatigue testing of as!welded 3932 _ller metal specimens
Nominal stress Hot spot stress Endurance
range "MPa# range "MPa# "cycles#
096[0 085[9 52[969
096[0 085[9 88[069
74[6 045[7 086[939
74[6 045[7 066[359
53[2 045[7 052[039
53[2 006[6 307[329
53[2 006[6 571[119
53[2 006[6 374[599
53[2 006[6 498[779
53[2 006[6 0601[529
Table 4
Results from fatigue testing of toe!ground 3932 _ller metal specimens
Nominal stress Hot spot stress Endurance
range "MPa# range "MPa# "cycles#
026[9 140[1 51[837
017[9 123[1 023[455
019[9 108[5 053[748
019[9 108[5 031[661
74[6 045[7 614[241
74[6 045[7 791[853
74[6 045[7 386[485
K.H. Macaonala\ P.1. Haaqensen,Enqineerinq Failure Hnalysis 6 (1998) 113130 016
Fig[ 8[ Fatigue test data from welded RHS T!joints[
Table 5
Fatigue test results\ SN constants for mean and design lines
Test series SN curves
Log C S
1 mill
"MPa# n
As!welded\ mean line 02[68 2[76
As!welded\ design line 02[31 58[0 2[76
Toe ground\ mean line 03[64 3[01
Toe ground\ design line 03[35 84[4 3[01
K.H. Macaonala\ P.1. Haaqensen,Enqineerinq Failure Hnalysis 6 (1998) 113130 017
Fig[ 09[ Comparison of test data with published results\ all are design "mean1 SD# curves[
4[ Proposed design methodology
Based on an evaluation of the published literature\ the following recommendations are made
for the _rst version of a design methodology for welded aluminium space frames made of rec!
tangular hollow section joints]
5.1. Defniiion of hoi spoi siress
Use the IIW 11 de_nition[
5.2. Deierninaiion of hoi spoi siress by sirain gauge neasurenenis
Use the IIW 11 extrapolation procedure
5.3. Finiie eleneni analysis io aeiernine SCFs
Determine stress distribution by three!dimensional analysis\ determine SCFs by linear extra!
polation from two points on the curve in accordance with the IIW procedure[ Use IIW 11
guidance for FEM analysis[
K.H. Macaonala\ P.1. Haaqensen,Enqineerinq Failure Hnalysis 6 (1998) 113130 018
5.4. Paraneiric fornulae for SCFs
Use equations proposed by van Wingerde et al[ 03 for RHS T! and K!joints as preliminary
guidance[
5.5. Hoi spoi siress aesign SN curre
Use the Partanen and Niemi curve with a stress range at 109
5
cycles of 39 MPa and a slope of
one!third[ This curve is slightly higher than the proposed DNV curves for example when corrected
for the di}erence between hot spot and notch stresses[
5[ Conclusions
Many di}erent de_nitions of hot spot stress range exist and are used in the fatigue design of
welded structures[ The guidance recently published by IIW for derivation of the hot spot stress
from strain gauge measurements and FEM analyses are recommended for use for preliminary
fatigue design of thin walled aluminium structures[
The hot spot stress SN curves for welded aluminium structures in design recommendations
that give speci_c advice are remarkably similar\ apparently converging on a design fatigue stress
range at two million cycles in the region of 2939 MPa[
The hot spot method appears well suited to welded aluminium RHS joints\ however\ the design
database needs to be expanded[ Speci_c parametric equations for SCFs need to be developed "or
adapted from existing methodologies for steel structures#[
Parametric FE analysis of a T!joint con_guration under in!plane bending of the chord showed
that combinations of small weld toe radius and large weld angle\ weld throat and wall thickness
lead to high SCFs[ In particular\ combinations of r and i have most in~uence[
Stress concentration factors determined from strain gauge measurements and FE analysis were
in good agreement[
Acknowledgements
The authors wish to acknowledge the assistance of R[ M[ Edvardsen and R[ Trandum both of
whom made contributions to the work reported here as part of their M[Sc[ thesis work[
References
0 Marshall PW[ Design of welded tubular connections[ Basis and use of AWS code provisions[ Amsterdam^ Elsevier\
0881[
1 Radenkovic D[ Stress analysis in tubular joints[ Int[ Conf[ on Steels in Marine Structures\ Paris\ 0870[
2 Gibstein MB\ Moe ET[ Fatigue of tubular joints[ In] Almar!N%ss T\ editor[ Fatigue handbook[ 0874[
3 Hobbacher A[ Fatigue design of welded joints and components[ Recommendations of IIW Joint Working Group
XIII!XV\ IIW Doc[ XIII!0428:XV!734!85[ Cambridge] Abington Publishing\ 0885[
K.H. Macaonala\ P.1. Haaqensen,Enqineerinq Failure Hnalysis 6 (1998) 113130 029
4 HSE\ Guidance on design\ construction and certi_cation[ 3th ed[ "including Feb[ 0884 amendments#[ London\
U[K[] HMSO\ 0884[
5 Efthymiou M[ Development of SCF formulae and generalised in~uence functions for use in fatigue analysis[
Proceedings of the O}shore Tubular Joint Conference[ Surrey\ U[K[\ 0877[
6 HSE\ Stress concentration factors for simple tubular joints] assessment of existing and developing of new parametric
formulae[ HSE Report OTH 80 246[ London] HMSO\ 0889[
7 Stacey A\ Sharp JV[ The revised HSE fatigue guidance[ Proceedings of the 03th International Conference on
O}shore Mechanics and Arctic Engineering[ Copenhagen\ Denmark] June 0711\ 0884[
8 API RP1A\ Recommended practice for planning\ designing and constructing _xed o}shore platforms[ 19th ed[
American Petroleum Institute\ 0882[
09 Baerheim\ M[ et al[ Proposed fatigue provisions in the new ISO code for o}shore structures[ Proceedings of the
03th International Conference on O}shore Mechanics and Arctic Engineering[ Copenhagen\ Denmark] June 07
11\ 0884[
00 CEN\ Eurocode 8] Design of aluminium structures[ Part 1] structures susceptible to fatigue[ prENV 080!1\ CEN:TC
149SC 8\ 0885[
01 Norsk standard\ NS 2361\ Prosjektering av stalkonstruksjoner[ Beregning og dimensjonering[ 1[ Utg[\ Norges
Standardiseringsforbund\ 0873[
02 van Wingerde AM et al[ Proposed revisions for fatigue design of planar welded connections made of hollow
structural sections[ Proceedings of the 4th International Symposium on Tubular Structures\ Nottingham\ 0882[
03 van Wingerde AM et al[ Criteria for fatigue assessment of hollow structural section connections[ J Constr Steel
Research 0884^24]601004[
04 van Wingerde AM et al[ The fatigue behaviour of K!joints made of square hollow sections[ IIW Doc[ XIII!0515!
85\ 0885[
05 Kosteas D\ Gietl B[ A comparative analysis of existing test data on welded aluminium tubular joints[ Fatigue and
Fracture of Engineering Materials and Structures 0885^08]620627[
06 Det Norse Varitas\ Class Note] Fatigue assessment of aluminium structures[ Report No[ LIB!J!99909\ Dec[ 0884^
revised Jan[\ 0885[
07 Det Norse Varitas\ Class Note] Fatigue assessment of ship structures[ Report No[ 82!9321\ revised Sept[\ 0885[
08 Partanen T\ Niemi E[ Hot spot S!N curves based on fatigue tests of small MIG welded aluminium specimens[ IIW
Doc[ 0525!85\ 0885[
19 Sharp ML et al[ Fatigue design of aluminium components and structures[ McGraw!Hill\ 0885[
10 BS 7007\ Structural use of aluminium[ Part 0[ Code of practice for design[ London] British Standards Institution\
0880[
11 Niemi E[ Stress determination for fatigue analysis of welded components[ Cambridge\ U[K] Abington Publishing\
0884[
12 Niu X\ Glinka G[ The weld pro_le e}ect on stress intensity factors in weldments[ Int J Fracture 0876^24]219[
13 Hagstro m J\ Sandstro m R[ Fatigue properties of welded T!joints in thin!walled aluminium pro_les[ Proceedings of
the International Conference 4th Int[ Conf[ on Aluminium Alloys\ ICAA!4\ Grenoble\ 04 July\ 0885[ Materials
Science Forum 0885^106111"2#]061621[