Professional Documents
Culture Documents
Abstract: A series of static strength tests were performed on full-scale precracked square-to-square hollow section welded T joints. These
joints were subjected under brace end axial load until failure using a special purpose designed test rig. An alternating current potential
drop technique was employed to monitor the crack propagation during the tests. In addition, three sets of linear variable displacement
transducers were used to record the crack mouth opening displacements. The tests were performed at room temperature; therefore, the
material is on the upper shelf of the ductile-brittle transition curve where failure occurs in a ductile manner. In all these tests, a stable
ductile tearing was observed with a small tearing region. Based on the yield line theory, the plastic collapse load is estimated, and then
applied to the failure assessment diagram 共FAD兲 of BS7910. The results show that BS7910 Level 2A FAD provides an adequate procedure
for the assessment of cracked square-to-square hollow section T joints.
DOI: 10.1061/共ASCE兲0733-9445共2006兲132:3共368兲
CE Database subject headings: Axial loads; Cracking; Fractures; Failures; Hollow sections; Joints; Strength.
Specimen Details
grade 50D, and they were available from the previous fatigue test
programs conducted earlier 共Chiew et al. 2002兲. Every specimen
contains at least one fatigue surface crack under the weld toe.
These hot finished square hollow section joints dimensions are
given in Table 1, and the mechanical properties of the BS4360-
50D steel provided by the supplier called Corus in the United
Kingdom, are given in Table 2. The weld profile and the specimen
preparation were carried out in accordance with the American
Welding Society 共AWS兲 Structure Welding Code D1.1-2000
共AWS 2000兲 specifications. They were then checked using the
ultrasonic technique to deem the welds’ quality. The two chord
ends were reinforced by stiffeners to avoid the chord end walls
Fig. 1. Test rig local buckling. The brace end was welded with the 40 mm thick
plate used to connect to the spread beam, and stiffeners are
welded to ensure the strength and even distribution of the applied
load.
Test Rig and Loading System
Weld and Fatigue Crack Details
A specially designed test rig as shown in Fig. 1 was used to test
these cracked SHS joints. The rig consists of two servo-hydraulic
The full penetration butt weld is used to connect the brace and
actuators which can provide ±1,000 kN load and ±125 mm dis-
chord members as shown in Fig. 3. The weld width, tw,
placement, respectively. The two actuators apply the loads to the
was measured after the joint was fabricated, and is tabulated in
two ends of the spread beam, and in turn these loads are trans-
Table 3. The fatigue crack length and depth were measured after
mitted to the brace end through the six bolts tightened to the
the fatigue and static tests, respectively. They are also shown in
spread beam as shown in Fig. 2. Therefore, a total maximum axial
Table 3.
load of 2,000 kN can be applied to the specimen. The actuators
were controlled using the Instron 8500 controller which can be in
load control or displacement control. The chord ends of the joints Monitoring of Crack Growth
were supported on a pin and a roller respectively placed on the
top of 600 mm⫻ 600 mm concrete supports as shown in Fig. 2, The Alternating Current Potential Drop 共ACPD兲 technique is a
creating a pinned end support condition at both ends. well-established technique 共Dover et al. 1995兲 which can be used
shape of a growing crack in the specimen. The ACPD method is the crack mouth opening displacement 共CMOD兲 along the crack.
based on the “skin effect” a characteristic of high-frequency cur- The locations of the first, second, and third set LVDTs are shown
rent flowing through a ferromagnetic material, whereby the ma- in Figs. 4 and 5 for specimens T1 and T2. Fig. 6 shows the
jority of the current is confined to a thin skin at the surface of the arrangement of the three LVDTs. One block and one bracket are
material. At a frequency of 5 kHz, ferromagnetic mild steel has a attached to the chord face and brace face, and in turn three LVDTs
skin depth on the order of approximately 0.1 mm 共Chiew et al. are fixed to the bracket. The three LVDTs are in contact with the
2004兲. In the crack specimen, the high-frequency current is in- block, as shown in Fig. 7, so that the vertical displacement of
jected at points on the end face of the specimen on either side of points 1 and 2 and the horizontal displacement of point 3 relative
the crack surfaces, and flows along a path defined by the crack to the brace can be measured accordingly.
surfaces. In principle, a linear surface potential gradient should In order to transfer the measurements obtained to the crack
exist along this current path. A reference potential may be mea- mouth opening displacement, an extrapolation approach is
sured on an uncracked region to define the strength of the poten- adopted in the test as shown in Fig. 8. If the region between the
tial gradient. The crack depth can be estimated by normalizing the crack and the measured point is treated as a rigid body, the
cross-crack measurement with respect to the reference measure- displacements at the measured points are caused by the displace-
ment 共Lie et al. 2001兲. ments and rotations of the crack faces. If the displacement
In this study, the ACPD crack depth measurements were at measured by LVDT 1, 2, and 3 are expressed as d1, d2, and l, the
fixed points around the fatigue crack location as shown in Figs. 4 following relationships can be found for:
and 5 for specimens T1 and T2, respectively. Because of the 1. rotation of crack
limitation of the channels, only the deepest crack in specimen T1
was monitored during the test. Specially manufactured probes of d2 − d1
␥= 共1兲
0.18 mm diameter and 10 mm length were spot welded on the L2
chord surface securely. All these probes were connected to the
2. vertical displacement of crack mouth
U10 Crack Microgauge 共TSC 1991兲 which provides the high-
frequency current and phase-sensitive detection. The U10 crack dc = d1 − ␥ · L1 共2兲
microgage was connected to an IBM personal computer using the
RS232 serial interfaces, and it was controlled by the Flair soft- 3. horizontal displacement of crack mouth
ware 共TSC 1998兲 that provides the automated instrument control, lc = l − l⬘ = l − ␥ · h 共3兲
data storage facilities, and dedicated graphical output under the
Windows environment. During the test, the crack was scanned at 4. and crack mouth opening displacement 共CMOD兲
every load step. Fig. 5 shows the setup of the ACPD probes along ␦CMOD = 冑d2c + l2c 共4兲
the fatigue crack.
where L1, L2, and L3 = respective distances from the LVDT 1, 2,
and 3 to the weld toe; ␥ = rotation angle of crack face which is
Measurement of Crack Mouth Opening equal to the rotation angle of the chord face; l⬘ = horizontal dis-
Displacement „CMOD… placement caused by the block rotation; and h = vertical distance
of the LVDT 3 away from the chord face, and can be calculated as Test Results and Discussion
h = h0 + d1 + ␥ · 共L3 − L1兲 共5兲
Fig. 9 illustrates the load–displacement curves for the static
strength of the two T joint specimens. The curve T1 shows that
the load is related linearly with the displacement up to a load of
Measurements of Brace End Axial Load and 600 kN. Subsequently, the slope of the load–displacement curve
Displacement changes producing an approximately linear region of up to
824 kN. After this point, there is a plateau, and the load can still
The brace end displacement was obtained from a LVDT with a increase slightly up to the maximum load 875 kN before it drops
stroke capacity of 50 mm which was calibrated and attached to down as the displacement increases rapidly. The curve T2 also
the bottom of spread beam in line with the centerline of the brace. shows a linear relationship in the elastic region, then the slope of
The load was obtained via load cell of the actuators, and the load the curve changes to another linear region up to 899 kN. Unlike
was measured using the strain gages attached to the brace to T1, after this point the load drops down to 805 kN, then it in-
confirm the load readings. creases slightly to 825 kN before it drops down dramatically.
If the twice-elastic-compliance limit 共Cheatani and Burdekin
1994兲 is used to determine the global plastic collapse load of the
Static Strength Test joints, it is obvious that the joint failed before the displacement
had reached the deformation limit, which means that the failure of
Elastic loading was carried out before the test to check the opera- the joints is fracture dominant even though the material is on the
tion of the various gages, LVDTs, and the output of the actuator. upper shelf of the transverse curve.
The specimen was loaded to 3% of the predicted total capacity, Fig. 10 shows the CMOD at the deepest point calculated from
and then unloaded to ensure that the instrumentation readings Eq. 共4兲 versus brace end axial load. Because the displacement of
returned to zero. The axial tensile load was applied at the end of point 2 is beyond the stroke of the LVDT 2 for specimen T1, the
brace using displacement control. The tests were deemed com- approximate CMOD under the maximum load is extrapolated
plete when the joint failed completely, which resulted in the loads from the last two readings, which may give a value slightly higher
than the real one.
Fig. 6. Setup of ACPD probes, and LVDTs along fatigue crack Fig. 8. Dimensions of LVDTs during test
Determination of CTOD
Fig. 9. Load–displacement curves at brace end
From the experimental tests, it is observed that the fatigue crack
surface is slightly curved. Hence, it is reasonable to use the in-
clined secant line to replace the curved crack surface cross section
Figs. 11 and 12 show the ACPD readings in test for specimens as shown in Fig. 17. Based on the assumption that the CMOD is
T1 and T2, respectively. It was found that the stable ductile tear- caused by the displacement of the ligament and rotation of the
ing exists. The ductile tearing initiates under a load much less crack face, the CTOD can be deduced from the CMOD measured
than the failure load. The recorded loads are 596 and 783 kN for as shown in Fig. 18. The relationship between CMOD and CTOD
spedimens T1 and T2, respectively, and they are shown in Fig. 9. is
But the ductile tearing region is small. A peak of load was found
␦CTOD = ␦CMOD − ␦⬘ 共6兲
at a point in which the brittle fracture took place and the crack
around the deepest point extended through the thickness of the where ␦⬘ = dIC · ␥, dIC = inclined crack depth, ␥ can be obtained
chord wall suddenly. It can be assumed that the CTOD of the from Eq. 共1兲.
deepest point was the critical parameter to assess the integrity of
the cracked T joint. As this load was reached, the crack growth
was very fast, and the load–displacement curve indicated a pla- Determination of Plastic Collapse Load
teau or a drop. Hence it is reasonable to take it as the failure load.
Figs. 13 and 14 show the first and second crack surfaces of The plastic collapse loads for the cracked geometry are evaluated
specimen T1, while Fig. 15 shows the crack surface of specimen by reducing the plastic collapse loads for the corresponding un-
T2 after the T joints had failed. A typical fracture surface is cracked geometry on the basis of the net load-bearing area for
shown in Fig. 16. This fracture surface indicates that ductile tear- axial loading, and the effect of the flaw area on the plastic col-
ing had taken place within a small region enclosing the crack tip lapse modulus for bending loads. The correction factor is given
2 which agrees with the ACPD readings. The fracture surface by the equation 共Cheatani and Burdekin 1994兲
冉 冊冉 冊
around the deepest point is shiny and smooth which indicates
crack area 1
brittle cleavage fracture. The cross section of this crack surface at FAR = 1 − ⴱ 共7兲
the deepest point is shown in Fig. 17. It can be seen that the weld length*t0 Q
fatigue crack surface is slightly curved, and a kink is very obvious where FAR = reduction factor to allow for the loss of load-bearing
at the end of the fatigue crack surface. This means the brittle cross sectional are due to the presence of the flaw. This equation
fracture is under a mixed-mode condition. was adopted by BS7910 共BSI 1999兲, and was validated by the
experimental and numerical analyses. But all these validations
were based on circular hollow section joints 共Burdekin 2001兲. For
RHS joints, the failure model is expected to be different. There-
fore, a new method should be investigated to calculate the plastic
Fig. 10. Load–CMOD curves at deepest point Fig. 12. ACPD readings of specimen T2
collapse load. When the width ratio between brace and chord  is
less than 0.8, normally the failure model of the T joint under
brace end axial loads is chord face yielding 共Packer et al. 1992兲.
Fig. 14. Second crack surface of specimen T1 The ultimate strength of the uncracked RHS T joint can be de-
rived using the yield line theory 共Wardenier 1982兲. Based on the
yield line theory, an approach was given to predict the collapse
loads of RHS T joints with surface cracks in this study, and it is
validated numerically in Part II of this paper. This approach is
based on the following two assumptions:
1. The yield line pattern stays the same as the uncracked joint
and
2. The thickness of the crack region is t0 − a, where t0 is the
thickness of the chord face, and a is the vertical depth of the
crack under weld toe.
The yield line pattern is shown in Figs. 19 and 20, where the
Fig. 15. Crack surface of specimen T2 energy participating in yield lines 1 – 5 remains the same as the
uncracked joint, which is
4 tan ␣
· ␦ · mp
1−
2 · tan ␣
· ␦ · mp
1−
Fig. 17. Cross section of crack surface at deepest point Fig. 19. Yield-line failure model
Fig. 21. CEGB R6 Revision 2 FAD 共adapted from Milne et al. 1988兲 Fig. 22. BS7910 共1999兲 level 1 and level 2A FAD
aged RHS T joints. FADs are used to consider failure by linear Kr = 共1 − 0.14Lr2兲关0.3 + 0.7 exp共− 0.65Lr6兲兴 共14兲
elastic fractures as one limiting criteria and failure by plastic col- for Level 2A and
lapse as the second criteria. When performing a structural integ-
rity assessment of a flaw in a stressed structure, an assessment
point is derived from two different calculations and plotted on the
FAD. The structure is deemed unsafe if the point calculated lies
Kr = 冉 Eref Lr3Y
+
LrY 2Eref
冊 −0.5
共15兲
on the curve or falls outside it, and it is safe if the point is within for Level 2B. The curve for Level 2A assessment is shown in
the curve. The failure assessment curve defined by CEGB R6 is Fig. 22.
shown in Fig. 21 共Kristiansen and Turner 1993兲.
Level 3—Ductile Tearing Assessment
BS7910 This level employs a full tearing instability approach and provides
a more accurate description of ductile materials. The FADs of
The PD6493 共BSI 1991b兲 procedure for the assessment of defects Levels 3A and 3B is the same as these of Levels 2A and 2B, but
in welded components, originally published in 1980 and subse- more material information is needed. The FAD of Level 3C
quently revised in 1991, has been used with some success in the specific to a particular material and geometry is obtained by
industry, and it has been applied to offshore structures 共Burdekin determining the J integral using both elastic and elastic–plastic
2001兲. Its main applications are the fitness-for-purpose assess- analysis of the flawed structure under the loads of interest. The
ment of fabrication and in-service defects, inspection scheduling, Level 3C FAD is the most accurate and complicated, and the
and determination of whether postweld heat treatment is required. assessment curve must be generated separately for different
As a result of The Welding Institute’s work with the British geometry and materials.
Standard Institution PD6493 共BSI 1991b兲 was further revised as
BS7910 共BSI 1999兲 which refines the approaches of the earlier
edition but retains the same principles 共Wiesner et al. 2000兲. Failure Assessment Diagrams
There are three levels of assessment depending on the input
data available, the level of conservatism, and the degree of Because the maximum SIFs or CTOD are found at the deepest
accuracy required. point of the crack, the SIFs or CTOD at this point should be used
to calculate the Kr value. For the surface crack, the region of the
deepest point is always considered under plane strain state. Hence
Level 1—Simplified Assessment the critical fracture toughness must use the plane strain fracture
Level 1 is the screening level introduced into the PD6493 共BSI toughness. To satisfy the plane strain criterion in BS7448 共BSI
1991b兲. It provides a conservative estimate from its use of a sim- 1991a兲, a specimen of more than 150 mm in thickness is required
plified FAD with in-built safety factors and requires conservative for the BS4360-50D steel on the upper shelf. It was not practical
estimates of the applied stress, the residual stress, and the fracture to test such a specimen in practice 共Austin 1999兲. Hancock and
toughness input. The curve for Level 1 assessment is shown in Spurrier 共1982兲 studied the critical CTOD of BS4360-50D steel
Fig. 22. under different thickness and temperature 共Fig. 23兲. In this study,
the ␦c of 2.7 mm for the 60 mm thickness on the upper shelf is
used, and this is considered to be a reasonable approximation. The
calculated values of Kr and Lr are given in Tables 4 and 5, re-
Table 4. Results of Kr共Kr = 冑␦CTOD / ␦IC兲 spectively. Fig. 24 shows the failure assessment diagrams for the
␦CMOD Inclined crack depth ␦CTOD ␦IC specimens using BS7910 Level 2A analysis.
Specimen 共mm兲 ␥ 共mm兲 共mm兲 共mm兲 Kr In accordance with the FAD shown in Fig. 24, it can be found
that the Level 2A FAD gives a good assessment for the SHS T
T1 3.96 0.12 17.5 1.86 2.7 0.83
joint under brace end axial loads when the plastic collapse load is
T2 3.41 0.10 21.2 1.29 2.7 0.69
predicted using Eq. 共12兲.
References
tors for part-through cracks in tubular welded joints.” Proc., 12th Int. acceptability of flaws in metallic structures.” Int. J. Pressure Vessels
Conf. on Offshore Mechanics and Arctic Engineering, Glasgow, Piping, 77, 883–893.