Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Int d Fatigue 12 No 1 (1990) pp 13-23

Stress concentration factors for


tubular Y- and T-joints

A. K. Hellier, M.P. C o n n o l l y and W . D . D o v e r

A systematic study of stresses in tubular Y- and T-joints has been conducted in


which nearly 900 thin-shell finite-element analyses were performed. These cover a
wide range of joint geometries under axial loading, in-plane bending and out-of-
plane bending. For each mode of loading, and for both the chord and brace sides
of the intersection, semi-empirical equations are derived which relate the stress
concentration factors at selected locations to a parametric function of the joint
geometry. Equations are also obtained for the angular location of the hot-spot stress
site around the intersection. The accuracy of these parametric equations is then
assessed by comparing the predicted values with results from steel model tests and
also with the predictions of other previously published equations.

Key words: tubular joints; bending stress; membrane stress; stress concentration
factors; design

Steel offshore structures used for the extraction of oil and gas parametric functions of the joint geometry for different modes
are composed of tubular joints welded together to form a of loading. 1-3
three-dimensional space frame. Such a structure is susceptible Tubular joints are conventionally designed using a
to localized fatigue failure at the welded intersections as a lower bound stress-life (S-N) curve approach. S-N curves,
result of the high stresses in these regions together with the established experimentally from large-scale fatigue tests on
large number of stress cycles experienced during its operational tubular joints, have been successfully used together with
life. In order to ensure that the structure is adequately parametric stress equations for a number of years to ensure
designed against fatigue failure it is necessary to be able to that tubular joints satisfy the fatigue design requirements,
predict the stresses around the welded intersection where measured in terms of the total number of cycles to failure at
cracks are likely to initiate and grow. As a consequence of a given applied stress range. Of late, however, there has been
this, much research effort has been directed towards the an increasing realization that a need exists to perform
determination of stress concentration factors (SCFs) for simple remaining life calculations on cracked joints. These are based
tubular joints, and these results have often been expressed as on fracture mechanics techniques which utilize the stresses

Notation
geometric ratio (=d/D)
al, a2, a3, coefficients in regression analysis geometric ratio (=D/2T)
d4~ a5~ a6~ e t c T geometric ratio (=t/T)
d external diameter of brace 0 brace angle
D external diameter of chord tp angle around intersection
SCF at crown position (~=0) ~x, ~y Loof node rotations
K~.~0 SCF at crown position (~= 180°) position of hot spot
¢~HS
Ks SCF at saddle position ~fis position of positive hot spot (IPB)
KHS SCF at hot spot ¢Pfis position of negative hot spot (IPB)
Kfis SCF at positive hot spot (IPB) chord main tubular member
Kfis SCF at negative hot spot (IPB) brace tubular attachment to main tubular
L chord length member
R2 product moment correlation coefficient hot-spot peak stress in chord (or brace)
SCF stress concentration factor stress
t brace wall thickness crown weld toe regions at 0° or 1800 (see Fig. 2)
T chord wall thickness saddle weld toe regions at 90 ° or 270 ° (see Fig. 2)
tCx~ ~y~ /4z displacements in x, y and z directions IPB in-plane bending
0~ geometric ratio (=2LID) OPB out-of-plane bending

0142-1123/90/010013-11 (~ 1990 Butterworth & Co (Publishers) Ltd


Int J Fatigue January 1990 13
acting on the anticipated crack path and the material crack
growth behaviour. For this purpose, not only is information
required on the absolute maximum value of stress at the
intersection (the 'hot-spot' SCF), but also the distribution of
SCFs around the intersection as well as the stress variation
through the thickness must be available.
The problem of predicting the relative proportions of
through-thickness bending and membrane stresses in tubular
Y- and T-joints has been addressed in an earlier paper 4
published by us in this issue. This present paper is concerned
with obtaining parametric expressions for the SCFs at all
locations of significance on both the chord and brace sides
of the intersection. Since the same finite-element meshes were
used for both studies, the equations resulting from each arc
directly compatible and may be used together. Also, although
parametric equations already exist for SCFs in tubular Y- Fig. 1 Eight-noded semi-Loof thin-shell element
and T-joints, those presented in this paper arc the only
comprehensive set which include values at the crown, saddle steel model results, with the error on the brace likely to be
and hot-spot positions, as well as equations to describe the greater than that on the chord.
location of the hot-spot stress site. It is anticipated that all
this information will be required in order to fit characteristic Mesh generation
formulae for the stress distributions around the intersection
The wide range of joint geometries covered by.' this study
adequately. necessitates the use of an automatic mesh generator. This
The objective of this work, then, is to provide a full must be capable of producing fine elements near the intersec-
predictive capability for the stresses acting on the anticipated tion, where the stresses are changing very rapidly, and coarser
crack plane in a tubular joint in service. All the equations are elements in regions where the stresses are more evenly
to be incorporated into a software package s which may be distributed. Also the elements should not be excessively
used to conduct rapid fracture mechanics analyses and elongated or otherwise distorted. A suitable program was
remanent life calculations on cracked tubular members. developed at UCL (Ref. 7) and modified for the present work
so that it would reliably generate meshes for tubular Y- and
Finite-element analyses T-joints having widely differing geometric parameters cl, [3,
y, T and 0, with an absolute minimum of user input.
Owing to the complex geometrical nature of tubular joints, The mesh generation program was written in FORTRAN
analytical solutions for their stress distributions are not to run on a MicroVAX II machine and produces a data filc
feasible. Instead these must be determined by numerical suitable for immediate analysis by PAFEC. A typical Y-joint
methods such as finite-element analysis or by experimental mesh comprising 1478 nodes and 477 elements is shown in
tests on strain-gauged steel or acrylic models. A finite-element Fig. 2. It took approximately five minutes of CPU time to
approach was adopted in this study, and since a detailed generate and a further 90 CPU minutes for the analysis to be
description of this has already been given in a previous paper 4 performed. Only one half of the tubular joint needed to bc
it will only be summarized here. meshed and further details of the procedure used and mesh
Element type validation may be found in Ref. 4.

The stress analyses were performed with the 'PAFEC' finite- Boundary conditions
element package 6 using semi-Loof thin-shell elements. These The boundary" conditions and applied loads are sumrnarizcd
elements are not capable of transmitting shear forces, but diagrammatically in Figs 3a,b respectively. All degrces of
displacements normal to the elements and rotations about freedom (u,, Uy, u~, q:~, %) were fixed at the chord ends,
their edges are allowed. The elements used were mostly where te, corresponds to the displacement in the x direction
quadrilaterals with eight nodes, denoted '43210', together
with a smaller number of their triangular six-noded counter-
parts known as type '43110'. A diagram of the eight-noded
element is shown in Fig. 1. All eight nodes possess three
translational degrees of freedom ux, u), and u~, and a rotational
degree of freedom also exists at each of the Loof nodes
marked with an 'x' in Fig. 1. These rotations are tangential
to the element side but are considered for the purposes of
input and output as ~. and %, rotations associated with the
nearest midside node.
In a shell analysis of tubular joints, the junction between
brace and chord is modelled as the intersection of their
midplanes. This is because shell elements are really two
dimensional, having thickness only in a mathematical sense
(necessary to define the element stiffness). Furthermore, the
weld cannot be incorporated in a model which consists purely
of shell elements. For both these reasons the SCF values
./J. I
.JJ
SaddJe
obtained from thin-shell finite-element analyses are subject to Fig. 2 Typical e x a m p l e of finite-element mesh used to model
some error, and are usually found to overestimate equivalent t u b u l a r joint (c~ = 8.97, !3 = 0.6, ~, = 14.5, T = 0.8, 0 = 60°)

14 Int J Fatigue January 1990


in offshore structures. Values for the brace angle 0 of 36 °,
45°, 60°, 75° and 90° were used. The minimum value of O was
restricted to 35 ° since it was not possible to generate
topologically sound meshes having smaller brace angles. For
each load case and angle 0 approximately 60 analyses were
conducted. The nominal stress for each joint and loading
mode was calculated automatically by the mesh generation
program and written to a data file. This was subsequently
read, along with the PAFEC stress output file, by a
postprocessor program and used to calculate SCF values at
AI the crowns, saddle and hot spot(s) on both brace and chord.
The SCFs were obtained by dividing the numerically greatest
principal stress at a given point around the intersection by
the appropriate nominal stress for that mode of loading.
Nominal stresses were calculated by dividing the total
applied load by the cross-sectional area of the brace for axial
loading. For moment loading, the nominal stresses were
derived from simple beam theory using a moment arm
measured from the brace end along its outer surface to the
(q0=0) crown position for IPB, and to the saddle position
(~=90 °) for OPB. ~=0 refers to the crown acute angle
a
BB u z c~x £by U x Uy location for a Y-joint (see Fig. 2). In all cases it is assumed
that the free length of brace or chord is at least three times
the diameter of the relevant member.

Parametric equations
From the raw database of SCF values, parametric equations
were obtained for the SCF at a number of locations around
the intersection, both chordside and braceside, under axial
Axial IPB OPB
loading, in-plane bending and out-of-plane bending. The
b parametric equations were derived using a statistical regression
Fig. 3 Details of: (a) boundary conditions; (b) modes of loading
package known as 'MINITAB' (Ref. 8) which is capable of
used for finite-element joint analyses
performing multiple regression and correlation analysis. The
methodology used in deriving the equations was as follows.
and ¢Px and ~y are the Loof node rotations. Note that these
(a) The variations of the SCF at the point in question were
are not rotations about the x and y axes (refer back to Fig.
plotted for each brace angle 0 as a function of the
1). Under axial loading or in-plane bending (IPB), symmetry
parameters c~, [3, ~/and • in order to determine the best
exists about the xy plane containing the chord and brace
forms of the terms required, and also to ascertain if any
centre-lines. A half-joint mesh may therefore be used provided
cross-correlation existed between the terms.
the displacements uz and rotations q~x and % are restrained
at all the nodes which lie on this plane. Since the restraints (b) A first attempt at the equation was made using the simple
form
are identical for both these modes of loading, they were
analysed consecutively as two load cases of the same finite- SCF = al cl'~ [3"3y'4 ,r~5 0'~ (6)
element run, without the need for recomputing the element
stiffness matrix. where al to a6 were determined from the regression
For out-of-plane bending (OPB) this symmetry about analysis.
the xy plane no longer exists. However, it was found that (c) Equation (6) was then modified by using other (eg
results of acceptable accuracy could be obtained by using a exponential) terms, and numerous regressions performed
half-joint mesh with the displacement components Ux and uy until a suitable equation with a large product moment
fixed at all points on the xy plane. 4 correlation coefficient was obtained.
(d) Having obtained the basic form of the final equation, the
Extraction of results exponent for the oe-variation (a2 in Equation (6)) was
The tubular joints analysed in this study encompassed the adjusted to reflect the observation that increasing o~ (ie
following ranges of geometries, expressed in terms of the the chord length) beyond the limit of 13.10 for this study
dimensionless parameters el, 13, ~/, "r and the brace angle 0: is known to have little effect on SCF values. By taking
into account this far-field behaviour it has been deemed
6.21 ~< ~x ~< 13.10 (1) possible to remove the upper validity limit on o~for these
0.20 ~< [3 ~< 0.80 (2) equations.
7.60 ~< ~/~< 32.0 (3) A summary of the parametric equations obtained together
0.20 ~< $ ~< 1.00 (4) with their correlation coefficient R 2 is given in Table 1. A
value of R 2 = 100% would imply that the fitted equation
35 ° ~< 0 ~< 90 ° (5) explains all the variations in SCF. It is apparent from Table
With the exception of a, which will be discussed in the 1 that a correlation of better than 90% has been achieved in
next section, these include the majority of tubular joints used most cases. All the equations derived in this study are set out

Int J Fatigue January 1990 15


Table 1. Summary of the parametric equations general lie close to the saddle point and so for simplicity were
obtained always taken to occur at ~p-90 °.
The parametric equations derived for in-plane bending
Type of Chord/ Quantity R 2 (%) Equation are more numerous than those for either of the other load
loading brace cases. This is due to the presence of two hot spots under this
type of loading, one having a positive SCF and the other
Axial Chord K° 97.2 (A1) having a negative value. For a T-joint these are equal and
Axial Chord K(!8° 97.4 (A2) opposite. For the loading direction shown in Fig. 3, the
Axial Chord Ks 99.2 (A3) positive hot spot lies at or near the ~p-0 crown position while
Axial Chord Kas 98.8 (A4) the negative hot spot has a ~p-value of 180° or thereabouts.
Axial Chord ~PHS 76.0 (A5) The positive hot spot for a Y-joint is usually, but not
Axial Brace K° 85.5 (A6) invariably, of greater magnitude than the negative onc.
Axial Brace K~8° 93.2 (A7) However, when dealing with a mixed-mode loading condition
Axial Brace Ks 98.8 (A8) and, in particular, attempting to fit a stress distribution curve
Axial Brace KHs 96.7 (A9) to Y-joints under in-plane bending, knowledge of the
Axial Brace ~HS 84.9 (A10) magnitudes and locations of both hot spots is necessary.
IPB Chord K° 99.4 (All) Simplified expressions were considered sufficient to descrii~e
IPB Chord K18° 98.1 (A12) the hot-spot locations in this case.
IPB Chord Ks -- (A13) An attempt will now be made to assess the accuracy ot
IPB Chord KGs 99.5 (A14) these parametric formulae by comparing them with other
IPB Chord KHS 98.4 (A15) parametric equations from the literature and finally with
IPB Chord ~s -- (A16) results obtained from steel model tests.
IPB Chord q;HS -- (A17)
IPB Brace K° 86.4 (A18)
IPB Brace K18° 91.9 (A19) Comparison with other equations
IPB Brace Ks -- (A20)
As a check on the general applicability of the parametric
IPB Brace KGs 89.7 (A21)
equations given in Appendices 1-6 a series of comparisons
IPB Brace KHs 91.2 (A22)
are now made, where possible, with existing parametric
IPB Brace ~;4s -- (A23)
formulae given in the literature. All of these formulae predict
IPB Brace ~HS -- (A24)
SCF values for both the chord and brace sides of the
OPB Chord K° -- (A25)
intersection under each of axial, IPB and OPB loading. The
OPB Chord K~8° -- (A26)
particular equations considered are as follows.
OPB Chord Ks 98.8 (A27)
OPB Chord KHS 98.8 (A28) (a) Kuang's formulae for T- and Y-joints.' These predict
OPB Chord ~HS -- (A29) hot-spot SCFs and are based on thin-shell finite-element
OPB Brace Kc° -- (A30) (FE) analyses of 46 different geometries.
OPB Brace K~8° -- (A31) (b) Wordsworth and Smedley's formulae for T- and Y-
OPB Brace Ks 96.4 (A32) joints.-' These were obtained from the results of an
OPB Brace KHs 96.8 (A33) unspecified number of acrylic model tests and give the
OPB Brace ~HS -- SCF at the saddle and crown positions.
(c) Gibstein's formulae for T-joints) These arc for hot-spot
SCFs and were derived from a study of 17 geometries,
fully in Appendices 1-6. A consequence of this approach is again using thin-shell FE analysis.
that for some particular geometries more than one equation
can apply. These equations should give very similar answers. Comparison of the 'UCL' equations (which are based on
If the calculated values are slightly different one should take thin-shell FE analyses of 291 T- and Y-joint geometries, each
the hot-spot value in preference to crown or saddle values, subject to all three modes ot loading) with these other
or if choosing between two crown or saddle values take the parametric formulae is not intended to be exhaustive, but
highest calculated value. rather to give a flavour of the relative variations in hot-spot
For axial loading, equations were fitted to the SCF SCF when one of the geometric parameters is allowed to
variations at all significant locations, namely the two crown change. Fig. 4 shows the variation in chord SCF for all the
positions (at ~p=0 and ~p=180°), the saddle point (~p=90°) and equations as a function of [3 for axially loaded T-joints. It
the hot-spot stress site where the greatest value of SCF occurs. can be seen that the UCL equation is a curve of similar shape
An equation was also developed to describe the angular location to the others but predicts values which are nearest those given
of the hot spot, ~pHS.This information is used more extensively by the highest of the existing curves, lying closest to the
in the development of parametric equations describing the Kuang equation at low !3 and approaching Smedley's equation
distribution of stresses around the connection. The equations at high 13. The peak value in the SCF is at a slightly lower
will be reported in the third paper in this series (Ref. 21, value of [3 for the UCL equation when compared with
published in this issue). Since the finite-element results are Gibstein and Smedley, but higher when compared with
sampled at a limited number of nodes around the intersection, Kuang.
the raw data for ~PHSare discrete, and a degree of circumspec- The chord SCF variation with y for T-joints under in-
tion is required in the use of this equation. A similar set of plane bending is shown for all four equations in Fig. 5. All
equations was also derived for out-of-plane bending with the curves show a steady increase in SCF as y becomes larger.
the exception of the crown positions. Here the SCFs on both Of the previously published formulae, Smedley's equation
chord and brace are very small in most cases, and were always yields the highest value in this case, and the UCL
therefore taken to be zero. Also, the hot-spot locations in equation predicts slightly higher values than Smedley.

16 Int J Fatigue January 1990


10 15_

UCL
..... Kuang .,~:
12 ..... Smedley ~ " , ~ I .~'
- . - O, te,n

..'5- /

~° ~
UCL
. . . . Kuang
..... Smedley
~-= Gibstein

o 1,,,lILl,li, 0
0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 .0

T
Fig. 4 C h o r d SCF v a r i a t i o n w i t h 13 f o r axially loaded T-joints (cx Fig. 6 Chord SCF v a r i a t i o n w i t h ~ f o r OPB-Ioaded T-joints (cx =
= 24.0, ~/ = 16.0, T = 0.5) 24.0, 13 = 0.5, 3' = 16.0)

joints and measuring the changes in strain due to the applied


UCL loads. The experimental procedures required to determine the
..... Kuang
..... Smedley
so-called extrapolated weld toe stresses have been well defined
m . m Gibstein and are summarised in Ref. 9. A body of data is therefore
available which relates the measured SCFs to the tubular joint
geometric parameters. Unfortunately, many of these tests
were conducted on 'unrealistic' joints, in the sense that their
geometric parameters have values unlikely to be found in
-- ,O@e o
practice, and also lie outside the limits of the parametric
equations.
In attempting to assess the accuracy of the various
parametric equations, a database of comparable steel model
results was compiled and is given in Tables 2, 3 and 4 for
the axial, in-plane bending and out-of-plane bending cases
i i i i I i i n J I i n n J I = i J n I = a a I respectively. Many of these were obtained from Ref. 9 but
6 12 18 24 30
do not include results where (x < 5.0, ~ > 32.0 or 13 = 1.0.
Tubular joint geometries with cl < 5.0 were ignored since it
Fig. 5 Chord SCF v a r i a t i o n w i t h 3, f o r IPB-Ioaded T-joints (cx = was felt that for these cases the end conditions applied to the
24.0, ~ = 0.5, T = 0.5) chord became too influential. Joints having - / > 32.0 are not
common in practice and also lie beyond the validity range of
the parametric equations. The 13 = 1.0 case (where both brace
Fig. 6 shows the chord SCF variation with T for OPB- and chord are of the same diameter) has a very complex
loaded T-joints. The UCL equation gives a linear increase in intersection geometry which places it in a category of its own
SCF with increasing $ in similar fashion to the existing as far as numerical analysis is concerned. Although quite often
equations, but is a shade more conservative than the highest found in offshore structures, this geometry is beyond the
prediction obtained using any of these. scope of the present study and has therefore been excluded
The graphs given in Figs 4-6 show firstly that the UCL from consideration. The remaining results are now considered
equations exhibit similar trends to those displayed by the for axial loading, in-plane bending and out-of-plane bending
established parametric formulae, and secondly that the UCL in turn.
formulae predict SCF values which appear to be consistently
higher than those given by these other equations. This Axial loading
tendency toward conservative predictions is an important The comparisons between the steel model results and the SCF
attribute for design purposes. The FE-based studies of Kuang, values produced by the UCL, Kuang, Wordsworth/Smedley
Gibstein and UCL show similar trends but always differ. and Gibstein formulae are given in Table 2 for axial loading.
This could be a consequence of the elements chosen and the All results are for the chord unless otherwise stated. The total
number of cases studied. number of steel model results is 24, and for convenience the
comparisons between the predicted and actual values are
Comparison with steel models expressed as the ratio of predicted SCF to actual SCF for
each of the four formulae. A ratio of 1.0 indicates an exact
In order to test the relative accuracy of the parametric fit. From this table it can be seen that the UCL formula
equations in predicting hot-spot SCFs, the UCL equations predictions are generally conservative, with only one result
together with the Kuang, 1 Wordsworth and Smedley, 2 and having a ratio less than 1.0 (Ref. 13, Spec. No 44). The Kuang
Gibstein, 3 equations are compared with steel model test and Gibstein equations underpredict the steel model results
results. These steel model results are normally obtained by seven and five times respectively, whereas the results from
carrying out large-scale loading tests on strain-gauged tubular the Smedley formulae are generally similar to those for the

Int J Fatigue January 1990 17


Table 2. Comparison between predicted and measured SCFs for axially loaded T- and Y-joints

Ref. Spec- c, 13 e Steel UCL UCL/ Kuang Kuang/ Smedley Smedley/ Gibstein Gibstein/
No imen (deg) SCF steel steel steel steel
No

10 1 10.0 0.50 13.4 0.50 90 5.7 7.5 1.32 6.3 1.11 7.5 1.32 6.1 1.07
4 10.0 0.50 14.3 0.50 90 6.7 7.8 1.16 6.6 0.99 8.0 1.19 6.5 0.97
11 10.0 0.25 14.3 0.39 90 4.7 5.7 1.21 5.4 1.15 5.0 1.06 4.0 0.85
13 10.0 0.50 14.3 0.50 90 7.7 7.8 1.01 6.6 0.86 8.0 1.04 6.5 0.84
11 II 5.0 0.46 20.0 1.00 60 11.0 16.8 1.53 17.0 1.55 17.3 1.57 -- --
12 1(c) 6.9 0.66 23.1 0.91 45 8.6 10.8 1.26 9.7 1.13 11.2 1.30 -- --
l(b) 6.9 0.66 23.1 0.91 45 6.5 8.7 1.34 8.4 1.29 8.1 1.25 -- --
13 19 10.5 0.53 13,4 0.86 90 11.4 14,5 1.27 12.6 1.11 12.9 1.13 12.8 1.12
41 5.0 0.50 14.4 1.00 90 12.0 18.0 1.50 16.1 1.34 16.1 1.34 16.1 1.34
42 5.0 0.50 14.4 0.28 90 3.3 3.7 1.12 3.0 0.91 4.5 1.36 2.8 0.85
43 5.0 0.24 14.4 0.47 90 4.8 7.0 1.46 6.7 1.40 5.9 1.23 4.9 1.02
44 5.0 0.24 14.4 0.25 90 3.7 3.2 0.86 2.9 0.78 3.1 0.84 2.1 0.57
14 UCL 7.3 0.71 12.0 1.00 90 10.0 13.7 1.37 10.8 1.08 11.7 1.17 12.0 1.20
TW2,4 7.3 0.71 12.0 1.00 90 10.5 13.7 1.30 10.8 1.03 tl.7 1.11 12.0 1.14
TW3 7.3 0.71 12,0 1.00 90 10.3 13.7 1,33 10.8 1.05 11.7 1.14 12.0 1.17
15 CP1 7.3 0.71 12.0 1.00 90 9.3 13.7 1.47 10.8 1.16 11.7 1.26 12.0 1.29
CP2 7.3 0.71 14.3 1.00 90 13.4 15.6 1.16 12.4 0.93 13.9 1.04 14.0 1.04
CP3 7.3 0.71 14.3 1.00 90 13.7 15.6 1.14 12.4 0.91 13.9 1,01 14.0 1,02
CP4 7.3 0.71 14.3 1.00 90 13.1 15.6 1.19 12.4 0.95 13.9 1.06 14.0 1.07
16 SHl(c) 7.3 0.71 14.4 0.80 90 9.3 12.0 1.29 9.3 1.00 11.2 1,20 10.4 1,12
SHl(b) 7.3 0.71 14.4 0.80 90 6.1 10.5 1.72 10.2 1.67 8.1 1,33 7.8 1.28
17 GCI(c) 7.2 0,71 14.3 0.79 90 8.7 11.7 1.34 9.0 1.03 11.0 1,26 10.1 1,16
GCI(b) 7.2 0.71 14.3 0.79 90 6.0 10.4 1.73 10.0 1.67 7.9 1.32 7.7 1,28
18 Y 13.0 0,48 15.9 0.63 45 5.2 6.1 1.17 5.6 1.08 6.1 1.17 -- --

(c) = chord, (b) = brace. All others are chord values.

Table 3. Comparison between predicted and measured SCFs for IPB-Ioaded T- and Y-joints

Ref. Spec- (x 13 ~/ T e Steel UCL UCL/ Kuang Kuang/ Smedley Srnedley/ Gibstein Gibstein/
No imen (deg) SCF steel steel steel steel
No

9 1 10.0 0.50 13.4 0.50 90 1.1 2.6 2.36 1.9 1.73 2.4 2.18 2.1 1.91

12 1(c) 6.9 0.66 23.1 0.91 45 3.3 5.7 1.73 3.6 1.09 4.5 1.36 -- --
l(b) 6.9 0.66 23.1 0.91 45 2.7 4.0 1.48 2.8 1.04 3.8 1.41 -- --

13 41 5.0 0.50 14.4 1.00 90 4.9 5.2 1.06 3.6 0.73 4.3 0.88 4.5 0.92
42 5.0 0.50 14.4 0.28 90 1.3 1.6 1.23 1.2 0.92 1.6 1.23 1.2 0.92
43 5.0 0.24 14.4 0.47 90 1.5 2.2 1.47 1.9 1.27 2.2 1.47 2.0 1.33
44(b) 5.0 0.24 14.4 0.25 90 2.0 2.2 1.10 2.4 1.20 1.8 0.90 1.8 0.90

14 UCL 7.3 0.71 12.0 1.00 90 3.0 5.0 1.67 3.2 1.07 3.7 1.23 4.0 1.33

17 G C I ( c ) 7 . 2 0.71 14.3 0.79 90 3.1 4.2 1.35 2.9 0.94 3.4 1.10 3.3 1.06
GCI(b) 7.2 0.71 14.3 0.79 90 2.0 3.3 1.65 2.5 1.25 3.2 1.60 2.3 1.15
GC2(c) 7.2 0.71 14.3 0.79 45 2.7 4.0 1.48 2.4 0.89 3.0 1.11 -- --
GC2(b) 7.2 0.71 14.3 0.79 45 2.5 3.6 1.44 2.3 0.92 2.9 1.16 -- --

(c) = chord, (b) = brace. All others are chord values.

18 Int J F a t i g u e J a n u a r y 1990
T a b l e 4. C o m p a r i s o n between predicted and measured SCFs for OPB-Ioaded T-joints

Ref. Spec- e<* 13 ~/ • 0 Steel UCL UCL/ Kuang Kuang/ S m e d l e y S m e d l e y / Gibstein Gibstein!
No imen (deg) SCF steel steel steel steel
No

19 23/1(c) -- 0.25 14.3 0.39 90 2.3 3.2 1.39 2.2 0.96 2.2 0.96 2.1 0.91
23/2(c) -- 0.25 14.3 0.39 90 2.9 3.2 1.10 2.2 0.76 2.2 0.76 2.1 0.72
23/3(c) -- 0.25 14.3 0.39 90 2.6 3.2 1.23 2.2 0.85 2.2 0.85 2.1 0.81
23/1(b) -- 0.25 14.3 0.39 90 2.0 2.8 1.40 2.9 1.45 2.4 1.20 2.3 1.15
23/2(b) -- 0.25 14.3 0.39 90 2.2 2.8 1.27 2.9 1.32 2.4 1.09 2.3 1.05
23/3(b) -- 0.25 14.3 0.39 90 1.9 2.8 1.47 2.9 1.53 2.4 1.26 2.3 1.21
24/1(c) -- 0.25 14.3 0.28 90 1.9 2.3 1.21 1.6 0.84 1.6 0.84 1.4 0.74
24/2(c) -- 0.25 14.3 0.28 90 1.6 2.3 1.43 1.6 1.00 1.6 1.00 1.4 0.88
24/3(c) -- 0.25 14.3 0.28 90 1.8 2.3 1.28 1.6 0.89 1.6 0.89 1.4 0.78
24/1(b) -- 0.25 14.3 0.28 90 1.7 2.4 1.41 2.4 1.41 2.0 1.18 2.0 1.18
24/2(b) -- 0.25 14.3 0.28 90 1.6 2.4 1.50 2.4 1.50 2.0 1.25 2.0 1.25
24/3(b) -- 0.25 14.3 0.28 90 1.8 2.4 1.33 2.4 1.33 2.0 1.11 2.0 1.11
19 703/1 -- 0.53 13.4 0.86 90 8.0 9.6 1.20 7.5 0.94 9.5 1.19 9.5 1.19
703/2 -- 0.53 13.4 0.86 90 7.7 9.6 1.25 7.5 0.97 9.5 1.23 9.5 1.23
703/3 -- 0.53 13.4 0.86 90 8.4 9.6 1.14 7.5 0.89 9.5 1.13 9.5 1.13
704/1 -- 0.53 13.4 0.51 90 5.4 5.7 1.05 4.7 0.87 5.6 1.04 5.2 0.96
704/2 -- 0.53 13.4 0.51 90 5.0 5.7 1.14 4.7 0.94 5.6 1.12 5.2 1.04
704/3 -- 0.53 13.4 0.51 90 5.4 5.7 1.05 4.7 0.87 5.6 1.04 5.2 0.96
14 UCL 7~3 0.71 12.0 1.00 90 8.5 10.0 1.17 7.1 0.84 11.9 1.40 10.5 1.24
16 SHI(c) 7.3 0.71 14.4 0.80 90 9.7 9,8 1.01 7.0 0.72 11.4 1.18 9.6 0.99
20 SH2(c) 7.3 0.71 14.4 0.80 90 10.6 9.8 0.92 7.0 0.66 11.4 1.08 9.6 0.91
SHl(b) 7.3 0.71 14.4 0.80 90 6.2 8.2 1.32 7.5 1.21 8.2 1.32 7.4 1.19
SH2(b) 7.3 0.71 14.4 0.80 90 6.0 8.2 1.37 7.5 1.25 8.2 1.37 7.4 1.23
17 GCI(c) 7.2 0.71 14.3 0.79 90 9.1 9.6 1.05 6.9 0.76 11.2 1.23 9.4 1.03
GCI(b) 7.2 0.71 14.3 0.79 90 5.1 8.1 1.59 7.4 1.45 8.2 1.61 7.3 1.43

(c) = chord, (b) = brace. All others are chord values.


*a assumed to be 7.0 where no value given.

UCL formulae, with only one underprediction. The UCL Summary


formulae are a little more conservative than Smedley's but
occasionally give a fairly high level of overprediction (for The total number of steel model results given in the SCF
example a ratio of 1.73 for Ref. 17, Spec. No GCI(b)). database of Tables 2-4 for axial loading, in-plane bending and
In-plane bending out-of-plane bending is 61. Fig. 7 shows the SCF values
predicted by the UCL equations plotted against the experimen-
Table 3 contains the comparisons between the steel model tal results for all these cases together, and serves to confirm
results and the SCF values predicted by the UCL, Kuang, the inherently conservative nature of the equations.
Wordsworth/Smedley and Gibstein formulae for in-plane If all the results for the three modes of loading are now
bending. The total number of steel model results is 12 in this broken down into intervals depending upon the ratios of
case. Once again the UCL equations are the most conservative, predicted SCF to actual SCF for each equation in turn, then
providing predictions which are consistently higher than the the histograms of Fig. 8 are obtained. The tendency of the
steel model results. The Kuang, Smedley and Gibstein UCL formulae to overpredict SCF values is again clear to
equations underpredict the steel model results 5, 2 and 3 see, with a mere 3% of the values being underpredictions. This
times respectively. should be contrasted with the Kuang, Wordsworth/Smedley
Out-of-plane bending
and Gibstein formulae which underpredict 44%, 13% and 34%
of the results respectively. Of these, the Wordsworth/Smedley
The steel model results and the SCFs obtained from the UCL, equations are generally considered the most appropriate for
Kuang, Wordsworth/Smedley and Gibstein formulae are design 9 because of their relatively low level of underprediction.
compared in Table 4 for out-of-plane bending. There are 25 On this basis, the UCL equations proposed in this paper
results for this mode of loading, all of which were obtained would appear to offer an even more reliably conservative
from tests on T-joints. It is apparent that in every case except prediction of SCF when designing a tubular joint.
one (Ref. 16, Spec. No SH1 (c)) the UCL equations overpredict It can be seen that under some circumstances the equations
the steel model results. For the other equations the numbers will predict a SCF of zero. In these cases consideration should
of cases where the SCF is underestimated are considerable, be given to using a minimum value instead of zero. The
with 15, 5 and 10 underpredictions for the Kuang, Smedley minimum value would be determined from previous design/
and Gibstein equations respectively. service experience.

Int J F a t i g u e J a n u a r y 1990 19
20
locations of significance on both the chord and brace sides
of the intersection under axial loading, in-plane bending and
out-of-plane bending. Equations have also been obtained
16
which describe the positions of the hot-spot stress sites around
LL
the intersection. It is anticipated that these might be used in
I.,3 12 conjunction with characteristic distribution formulae to pro-
vide a full description of the SCF variation around a tubular
CJ joint.
8 The equations for the hot-spot SCF are found generallx
Q. to overestimate the measured SCFs from steel model tests.
They have also been compared with existing parametric
4 formulae due to Kuang, Wordsworth/Smedley and Gibstein,
and found to be generally more conservative than anx of
these. It is therefore concluded that the hot-spot S('F
o ,, , I ,,, ,I ,,, , I ,,, , I,,,, equations presented in this paper are the most reliable in
4 8 12 15 20 predicting a conservative value of SCF which could bc used
Experimental SCF in design.
Fig. 7 SCF predicted by UCL equations plotted against exper-
imental SCF for all modes of loading
Future work
Several areas may be identified in which further work could
profitably be undertaken. The first of these, which is
Conclusions considered in the next article in this issue, 2~ is the generation
A thin-shell finite-element study has been performed in order of characteristic stress distribution formulae for each mode
to obtain stress concentration factors for a wide range of of loading which utilize the equations presented in this paper
tubular Y- and T-joint geometries. Parametric formulae or as input parameters, thus providing a full description of the
simple expressions have been derived for the SCF at all SCF variation around the intersection. This should enable

UCL 40 - I Smedley
40 -- I
I
I

30 - - 30 -
>~ >.
u L)
r- r"

20 -- 20
U"

U_ U_

10 10

a C

40 40 - Gibstein

30I Kuang
30

>, >,
(J ~J
c"
20 20
EY

LL LL

10 10

/ / / / A
0
0 0.75 I .0 I .25 I .5 0.75 1.0 1.25 1.5

b Predicted SCFltest result d Predicted SCF/test result

Fig. 8 SCF histograms for all load cases combined: (a) UCL formulae; (b) Kuang's formulae; (c) Wordsworth and Smedley's
formulae; (d) Gibstein's formulae

20 Int J Fatigue January 1990


more refined fracture mechanics analyses of cracked tubular (0 is in radians).
joints to be performed using a two-dimensional stress field Parametric equation for SCF at the brace crown position
on the anticipated crack plane. (lp=180 °) under axial loading:
Secondly, it would be desirable to extend the scope of
Klc8° = 1 9 . 0 5 5-0"035[3 (0"213/02 - 00439)5(0'48 + 000589/05)
this study to encompass higher values of 13 (the ratio of brace
diameter to chord diameter) up to and including 13 = 1.0. x exp(-0.00148~ 2 - 1.25 sin 0
Future studies to produce similar sets of parametric equations + 0.00055~2/[3 - 1.09132"r) (A7)
for X- and K-joints would also be worthwhile. (0 is in radians).
Parametric equation for SCF at the brace saddle position
Acknowledgements under axial loading:
The authors wish to acknowledge the Marine Technology Ks = 0.538 0f°°28650'963
Directorate Ltd, SERC and Sponsors of the Cohesive Fatigue
x exp(-0.987133 - 4.67~/-°.s
Programmes 1985-9 for providing some of the financial
support needed for this work. + 4.87 sin 1/2 0 - 0.1335/[3) (AS)
Parametric equation for SCF at the brace hot-spot stress site
Appendix 1 under axial loading:
KHS = 4.40 50"°s6850 ss3
Parametric equation for SCF at the chord crown position
(~0=0) under axial loading: X exp(--1.18613 s -- 4.2811-y -°.s
+ 1.987 sin 0 + 0.427[35) (Ag)
K ° = 0.575 51.°"°6313-°2°7°1/25(°"988-°'133/°)
exp(0.66513 s Parametric equation for position of brace hot-spot stress site
+ 0.0204~/+ 1.64 sin 0 - 0.469[35) (A1) under axial loading:
(0 is in radians). ¢4~HS -- 58.6 500309[30-05585(0.14402 - - 0.404)
Parametric equation for SCF at the chord crown position
x exp(1.01/~/+ 0.343/sin 0 - 0.0503[3/5) (A10)
(lp=180 °) under axial loading:
(0 is in radians, IPHS is in degrees).
K~:8° = 3.94 (~-0"03913--0'10505(0"906 + 0"055/03)exp(0.327134 Validity ranges:
+ 0.0177~ - 0.05 sin 0 - 0.422[3"0 (A2) 6.21 ~< 5
(0 is in radians). 0.20 ~< 13 ~< 0.80
Parametric equation for SCF at the chord saddle potation
7.60 ~< ~/~< 32.0
under axial loading:
0.20 ~< 5 ~< 1,00
Ks = 2.51 5°°3725112 exp(-1.96134
- 11.3/~/+ 2.47 sin 0 + [35) (A3) 35 ° ~< 0 ~< 90 °.
Parametric equation for SCF at the chord hot-spot stress site Appendix 3
under axial loading:
Parametric equation for SCF at the chord crown position
KHS = 9.92 50'029513 exp(--1.3813 s's -- 5.53~/-°-s (lp=0) under in-plane bending:
+ 2.15 sin 0 -- 0.05695/[3) (A4)
Kc° = 2.84 5 - ° °°ss~/°-2345(T M - o.171o)exp[(0.00213,
Parametric equation for position of chord hot-spot stress site
under axial loading: - 0.0373)/[32 - 0.0179/sin 0] (All)
(0 is in radians).
lPHS = 193.7 5--0"041513 (0.267 -- 0"093502)5(0"2710 -- 0.491) Parametric equation for SCF at the chord crown position
× exp(15.5/'y 2 - 0.77 sin 0 (~=180 °) under in-plane bending:
+ 0.000028~/2/[32 - 0.00932/[35) (A5) K~cs° = - 3 . 8 4 5 5--O'O29'yO'4075(O'966 -- 0.07190) (A12)

(0 is in radians, 0HS is in degrees). × exp[(0.00166~/- 0.0399)/[32 - 0.682/sin 0]


Validity ranges: (0 is in radians).
6.21 ~< 5 Parametric equation for SCF at the chord saddle position
under in-plane bending:
0.20 ~< [3 ~< 0.80
Ks = 0.0 (A13)
7.60 ~< -¢ ~< 32.0 Parametric equation for SCF at the chord positive hot spot
0.20 ~< -r ~< 1.00 under in-plane bending:

35 ° <~ 0 ~< 90 °. Kfts = 2.31 5°'°°33~/°'326'r (1"24 - 0.1870) (A14)


x exp[(0.00154~/- 0.0323)/[32 - 0.0248/sin 0]
Appendix 2 (0 is in radians).
Parametric equation for SCF at the brace crown position Parametric equation for SCF at the chord negative hot spot
(~p=0) under axial loading: under in-plane bending:
g° = 1.46 50'0098130'214/05 (0.623 - - 0'0811/02) /iris = - 8 . 9 3 0 f - - 0 " 0 1 7 6 ' ~ / 0 ' 4 5 5 (0"94 -- 0.04410) (A15)
x exp(0.87213 s - 0.00104~/2 + 1.47 sin 0 × exp[(0.00153~/- 0.0386)/[32 - 1.69/sin1/20]
+ 0.000474~/2/[3 - 1.73135) (A6) (0 is in radians).

Int J Fatigue January 1990 21


Parametric equation for position of chord positive hot spot Validity ranges:
under in-plane bending:
6.21 ~- c~
+ -- (y/0.41)]~ 2
~PHS (A16)
0.20 ~< 13 ~< 0.80
(~P~s in degrees).
Parametric equation for position of chord negative hot spot 7.60 ~< 7 ~< 32.0
under in-plane bending: 0.20 ~< T <~ 1,00
~PHS = 180 -- (y/0.45)[372 (A17) 35 ° ~< 0 <~ 90 °.
(q~fis is in degrees).
Validity ranges :
Appendix 5
6.21 ~ cx Parametric equation for SCF at the chord crown position
0.20 ~ [3 ~< 0.80 (q~=0) under out-of-plane bending:

7 . 6 0 ~< y ~< 32.0 K ° = 0.0 (A25)


0.20 ~ ~ ~< 1.00 Parametric equation for SCF at the chord crown position
(~p- 180°) under out-of-plane bending:
35 ° ~< 0 ~< 90 °.
K ~ ° = 0.0 (A26)
Appendix 4 Parametric equation for SCF at the chord saddle position
Parametric equation for SCF at the brace crown position under out-of-plane bending:
(~p=0) under in-plane bending:
Ks = 0.315 (t. 0"O54y(hI2 O'O939/O2)'1
-
K ° = 1.89 O¢--0"065%/-0"2142/O2T0"186~O'298 × exp(-0.001391134 + 0.654 sin 2 0) (A27)
× e x p ( - 5 . 8 8 / y + 1.81/sin 0
(0 is in radians).
0.0534[33y - 0.084818/7) (A18)
Parametric equation for SCF at the chord hot-spot stress site
-

(0 is in radians). under out-of-plane bending:


Parametric equation for SCF at the brace crown position
I~THS : 0.255 (~0.O13y(t.24 0.224/0)T
(lp=180 °) under in-plane bending:
x exp(-0.00135/iB 4.' + 0.923 sin 0) (A28)
K{~° = - 7 . 0 2 ~-O'O4S~O'141/O2yO'167/0270"143
(0 is in radians).
× exp(--0.015/132 -- 164.4/73 -- 0.524/sin20 Parametric equation for position of chord hot-spot stress site
- - 0.02541327) (A19) under out-of-plane bending:
Parametric equation for SCF at the brace saddle position ~Hs = 90.0 (A29)
under in-plane bending:
(tpHs is in degrees).
Ks : (-,//60.0)(1/0 2) (A20) Validity ranges:

(0 is in radians). 6.21 ~< c~


Parametric equation for SCF at the brace positive hot spot 0.20 ~< 13 ~< 0.80
under in-plane bending:
7.60 ~< 7 m 32.0
K~ts = 0.332 O~O'OOS3~/(0"432 O"161/02)70"296
--

0 . 2 0 ~< T ~< 1.00


× exp(--0.00436/lg 2 + 1.4/sin 0) (A21)
35 ° ~< 0 ~< 90 °.
(0 is in radians).
Parametric equation for SCF at the brace negative hot spot
Appendix 6
under in-plane bending:
Parametric equation for SCF at the brace crown position
Kfis = - 9 . 4 0 0c°°277(°°~82 + 0.~3o) (~p=0) under out-of-plane bending:
× exp(-0.0289/132 - 3.59/7 - 0.633/sin 0
+ 0.00558y/[8) (A22) K~ : 0.0 (A30)

(0 is in radians). Parametric equation for SCF at the brace crown position


Parametric equation for position of brace positive hot spot (%0=180°) under out-of-plane bending:
under in-plane bending: K{~° : 0.0 (A31)
~P'~s = (y/0.35)[ 32 (A23) Parametric equation for SCF at the brace saddle position
(lpfis is in degrees). under out-of-plane bending:
Parametric equation for position of brace negative hot spot Ks = 0.0788 O£°'°91~-°'132/°'~°'9°aT°'957
under in-plane bending:
× exp (-0.211/18 t2 + 2.62 sin 0) (A32)
~PF~s= 180 - (7/0.45)!B72 (A24) Parametric equation for SCF at the brace hot-spot stress site
(~pHs is in degrees), under out-of-plane bending:

22 Int J Fatigue January 1990


KHS = 0.103 ~x°°°s ~--0'227/8'~/0'848'1"0"469 11. Brandi, R. 'Behaviour of unstiffened and stiffened tubular
joints' ibid paper TS6.1
x exp(-0.232/131-2 + 2.59 sin e) (A33)
12. Kratzer et al. 'Schwingfestigkeitsuntersuchungen an
(O is in radians). Rohrstrukturelement-modellen von offshore-plattformen'
Parametric equation for position of brace hot-spot stress site Report BMFT-FB (M81) (June 1981)
under out-of-plane bending: 13. Irvine et al. 'Tubular joint fatigue data obtained at the
National Engineering Laboratory' UKOSRP Report 4/02
• HS = 90.0 (A34) (D. En., 1986)
(~ is in degrees). 14. Dover, W.D. and Wilson, T.J. 'Fatigue fracture mechanics
Validity ranges: assessment of tubular welded T-joints' Dept. of Energy
Contract OT/F/917, Final Report (Dept. of Energy, July
6.21 ~< o~ 1983-June 1985)
15. Dover, W.D., Peat, C. end Sham, W. P. 'Random load
0.20 ~< 13 ~< 0.80 corrosion fatigue of tubular joints and tee butt welds'
Dept. of Energy Contract TA 93/22/94 , Final Report
7.60 ~< ~/~< 32.0 (Dept. of Energy, December 1987)
0.20~< $~<1.00 16. Dover, W.D., Holdbrook, M.S.J., Hibberd, R.D. and Charle-
sworth, F.D.W. 'Fatigue crack growth in T-joints: out-of-
35 ° ~< 0 ~< 90 °. plane bending', lOth Ann. Offshore Technology Conf.,
Houston, Texas, May 8-11 1978, paper OTC 3252
References 17. Dover, W.D., Cheudhury, G.K. end Dharmavasan, S.
'Experimental and finite element comparisons of local
1. Potvin, A.B., Kuang, J.G., Leick, R.D. and Kahlick, J.L. stress and compliance in tubular welded T and Y joints
'Stress concentration in tubular joints' Soc Petroleum Int. Conf. on Steel in Marine Structures, Paris, 1981,
Engrs J (August 1987).
paper No 4.3
2. Wordsworth, A.C. and Smedley, G.P. 'Stress concen-
trations at unstiffened tubular joints' European Offshore 18. Gibstein, M.B. end Moe, E.T. 'Numerical and experimental
stress analysis of tubular joints with inclined braces' Int.
Steels Res. Semin., Cambridge, UK, November 1978
(Welding Institute, Cambridge, 1978) Paper 31 Conf. on Steel in Marine Structures, Paris, 1981, paper
No 6.3
3. Gibstein, M.B. 'Parametric stress analysis of T-joints' ibid
Paper 26 19. Irvine et al. 'Tubular joint fatigue data obtained at The
Welding Institute', UKOSRP Report 4/03 (D. En., 1986)
4. Connolly, M.P., Hellier, A.K., Dover, W.D. and Sutomo,
J. 'A parametric study of the ratio of bending to 20. Holdbrook, S.J. 'The application of linear elastic fracture
membrane stress in tubular Y- and T-joints' Int J Fatigue mechanics to fatigue crack growth in tubular welded
12 1 (1990) pp 3-11 joints' PhD Thesis (University College London, 1980)
5. S E R CProgramme on Fatigue Crack Growth in Offshore 21. Hellier, A.K., Connolly, M.P., Kare, R.F. end Dover, W.D.
Structures, July 1987-June 1989 (UCL Internal Report, 'Prediction of the stress distribution in tubular Y- and T-
London, 1989) joints' /nt J Fatigue 12 1 (1990) pp 25-33
6. PAFEC Data Preparation User Manual Version 6.1 (PAFEC
Ltd, Nottingham)
Authors
7. Dharmavasan, S. 'Fatigue fracture mechanics analysis of
tubular welded Y-joints' PhD Thesis (University College
London, 1983) W.D. Dover is with The Department of Mechanical Engineer-
8. M/N/TAB Reference Manual (Minitab Inc., PA, 1988) ing, University College London, Torrington Place, London
9. UEG Design of Tubular Joints for Offshore Structures WC1E 7JE, UK, where this work was carried out. A.K.
'Local behaviour of simple welded joints' (UEG, London, Hellier is now with The School of Materials Science and
Vol. 2, part C) Engineering, University of New South Wales, Kensington,
10. Dijkstra, O.D. et al. 'Fatigue strength of tubular X and T NSW 2033, Australia. M.P. Connolly is now with General
joints (Dutch) tests' Steel in Marine Structures, Paris, Electric, Aircraft Engine Business Group, Cincinnati, Ohio
October 1981, paper TS8.4 45215, USA.

Int J F a t i g u e J a n u a r y 1990 23

You might also like