Gusset Plate Connection to Round HSS Tension

Members
J. J. ROGER CHENG and GEOFFREY L. KULAK

ABSTRACT welds. This arrangement is shown pictorially in Figure 1.

Part (a) of the figure shows the slotted tube and a represen-
n experimental program and associated numerical
Aanalysis were undertaken to study the shear lag effect in tation of the gusset plate. The connected tube and gusset
plate are shown in Part (b). In addition to the four longitu-
round hollow structural section (HSS) tension members that
dinal fillet welds that will be present, there also may be fil-
are welded to gusset plates. The connection is made by slot-
let welds across the thickness of the gusset plate at the top
ting the tube longitudinally, inserting the gusset plate, and
of the slot. An alternative to this arrangement is to slot the
then placing longitudinal fillet welds at the tube-to-gusset
gusset plate rather than the tube. Fabricators prefer slotting
interface. Transverse welds at the junction of the slot and
the tube, however, since alignment of the elements and sub-
the gusset plate may or may not be present. A total of nine
sequent welding are more convenient with this
specimens with three different tube sizes and various weld
arrangement.
lengths were tested in the program. The majority of the
specimens failed by fracture of the tube somewhere Only the slotted tube option is discussed in this paper and
between the two gusset plates, and there was considerable only circular tubes are treated. The work has implications
ductility prior to fracture. Numerical analyses of the con- when square or rectangular tubes are used, but those cases
nections were carried out using an elasto-plastic model and are not treated specifically.
measured material properties. Based on the tests and the
numerical analyses, it is concluded that shear lag does not
significantly affect the ultimate strength of the slotted tube
connection, even with a weld length as little as 80 percent
of the distance between the welds. The studies showed that
the restraint provided by the gusset plate at the slotted end
effectively increases the load-carrying capacity of the tube
as compared to the unrestrained portion of the member. In
the numerical analysis, fracture is assumed to have occurred
when the equivalent plastic strain reaches a critical value.
The test results are discussed in light of the requirements in
the American Institute of Steel Construction (AISC)
Specification for Design of Steel Hollow Structural Sections
(AISC, 1997).
(a) Slotted tubular member and gusset plate
INTRODUCTION
Tubular sections are used frequently in fabricated steel
construction as tension members, for example, as bracing
members. An inexpensive and easy way of making the end
connections is to slot the tube longitudinally and insert a
gusset plate, which is then welded to the tube using fillet

J. J. Roger Cheng, Ph.D., REng. is C. W. Carry professor of

steel structures, Department of Civil & Environmental
Engineering, University of Alberta, Edmonton, Canada. (b) Assembled tubular member and gusset plate
Geoffrey L. Kulak, Ph.D., PEng. is professor emeritus at the
Department of Civil & Environmental Engineering, University
Fig. 1. Slotted tube connection.
of Alberta, Edmonton, Canada.

ENGINEERING JOURNAL / FOURTH QUARTER / 2000 133

 AISC LRFD DESIGN REQUIREMENTS There is an implication in the net area calculation that fil-
let welds will not be used where the gusset plate meets the
Only a relatively small portion of the cross-section of a
end of the slot in the tubular member (transverse welds).
tube can be connected to the gusset plate. It has to be
Indeed, the Commentary to the HSS Specification states
assumed, then, that shear lag will be present and must be
that, "welding around the end of the gusset plate is not rec-
taken into account by the designer. The effective area of
ommended." This is explored in the tests reported here. It
tension members is discussed in Section B3 of the AISC
should also be pointed out that the shear reduction factor
LRFD Specification (AISC, 1993). Calculation of the effec-
applies whether or not the cross-section has been locally
tive area according to B3 acknowledges the possibility of
reduced in some way (Salmon and Johnson, 1996).
shear lag and prescribes a reduction coefficient, U, that is to
Consider, for example, an angle welded to a gusset plate by
be applied to the calculated cross-sectional area when com-
puting the factored resistance based on fracture. However, fillet welds at the heel and toe of one of the legs. The gross
the values suggested for U in this section of the cross-section and the net cross-section are identical, yet it is
Specification do not specifically treat the type of connec- obvious that shear lag must be considered.
tion shown in Figure 1. The AISC Specification for the If only longitudinal welds are used in the slotted tube
Design of Hollow Structural Sections (AISC, 1997) does connection, then the net cross-section should be used, i.e.,
cover this case, though, and it is those rules for shear lag gross cross-section less the end area of the two slots. If fil-
that will be treated here. This specification will be referred lets welds are present where the gusset plate meets the end
to as the HSS Specification. of the slot in the tubular member (transverse welds), then
the gross cross-section of the tube should be used. In either
According to Section 2.1 of the HSS Specification, the
case, the shear lag reduction factor, U, should be applied.
effective net area for tension members is to be calculated as:

AU PREVIOUS RESEARCH
(1)
Although there are two previous studies that relate to the
where A = An. In the case of a slotted tube connected to a investigation reported here, neither of them deals with cir-
gusset plate, the net area is that of the tube as reduced by the cular tubes. One was a numerical study done by Girard,
slots. For the arrangement shown in Figure 2, the HSS Picard, and Fafard (1995). That research examined the
Specification gives the shear lag reduction factor as: effect of shear lag on a slotted rectangular tube welded into
a gusset plate. No physical tests were undertaken in this
U=l-(x/L)<0.9 (2) program. The second study was that done by Korol (1996).
The terms x and L are shown in Figure 2. The term x is This researcher tested 18 specimens, all of which were slot-
intended to be the distance from the centerline of the ted square or rectangular tubes connected to gusset plates.
gusset plate to the centroid of one-half of the tube cross-sec- Only seven of the specimens tested failed in a way that
tion. A good approximation is obtained using the recom- explored the shear lag phenomenon. In the other 11 tests,
mendation given in the HSS Specification, namely: failure was by tearing out of a block of material ("block
shear"). No analysis was included in this study, but the
Din (3) work of Girard et al. was used to provide a prediction for
one test specimen. The Girard et al. model did not give a
particularly good prediction for this case. Korol observed
that restrictions on the numerical model of Girard et al.
limit its usefulness.
Considering the lack of experimental evidence, it is
appropriate to explore whether shear lag actually exists in a
slotted tube connection and, if so, report on the suitability
of the AISC HSS Specification rules for this type of con-
nection. Both an experimental program and numerical mod-
eling were undertaken to examine the tensile capacity of
round tubular members that are slotted and then connected
by welding to gusset plates.

TEST PROGRAM
The objective of the testing program was to examine the
Fig. 2. AISC shear lag nomenclature. ductility and the capacity of the slotted tubular tension

134 ENGINEERING JOURNAL / FOURTH QUARTER / 2000

 Table 1.
Tube Test AISC LRFD Calculated
Size Fillet Results Loads
dia. x Weld Shear Ultimate
Trans-
wall Lag Yield Ultimate Ultimate Load
Specimen verse
thick- length x Factor P Load Load L o a d <!>,, Numerical
Weld ' U TEST
max
ness leg size U <])(=1.0 <>,= 1.0 l/=1.0 Analysis

mm mm kN mm kN kN kN kN
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)

S1-1 102x6.4 no 170x6 0.81 830 34 743 642 792 827

S2-1 102x6.4 yes 170x6 0.81 869 135 743 688 849 856

S2-2 102x6.4 yes 170x6 0.81 849 137 743 688 849 856

S2-3 102x6.4 yes 170x6 0.81 875 142 743 688 849 856

S3-1 102x4.8 yes 150x5 0.78 645 99 519 487 624 629

S3-2 102x4.8 yes 150x5 0.78 634 102 519 487 624 629

S3-3 102x4.8 yes 150x5 0.78 631 76 519 487 624 629

S4-1 219x8.0 yes 345x10 0.80 2160 128 1729 1713 2141 2168

S5-1 219x8.0 yes 275x10 0.75 2157 54 1729 1606 2141 2110

members. This was done using the series of tests outlined in the tube slot with the gusset plate in addition to four longi-
Table 1. tudinal fillet welds.
The HSS sections used are those produced in Canada. The specimens were loaded axially using a clevis fixture
They were hot-rolled and were Grade 350W steel. The that was accommodated in the testing machine. The length
specified yield strength for this steel is 350 MPa and for of the tube between gusset plates ranged from 330 mm to
HSS the ultimate tensile strength must fall in the range 450 350 mm. Measurements of loads, deformations and strains
MPa to 620 MPa. All calculations necessary for the speci- were made as a specimen was loaded. Loading was carried
mens in this program were made using measured dimen- out at a slow rate and in the inelastic range loading was
sions and material properties (Cheng, Kulak, and Khoo, stopped at intervals so as to ensure no dynamic effect.
1998). Further details of the tests can be found elsewhere (Cheng
As shown in Table 1, the sections used were HSS et al., 1998).
102x4.8, HSS 102x6.4 and HSS 219x8.0. The first of Seven of the nine specimens failed at the mid-length of
these is not available in US practice, but the latter two sec- the member after extensive necking had taken place. Figure
tions correspond to HSS 4x0.25 and HSS 8.625x0.313, 3 shows a typical specimen in this category (S 4-1). The two
respectively, in US customary units (AISC, 1997). The stat- exceptions were S 1-1 and S 5-1. In these cases, failure
ic yield and ultimate strengths were, respectively, 375 MPa occurred in the tube where it entered the gusset plate, that
and 451 MPa for the HSS 102x4.8, 339 and 449 MPa for is, at the stressed end of the slot. Figure 4 shows the frac-
the HSS 102x6.4, and 348 MPa and 431 MPa for the HSS ture of S 1-1. In all the tests there was extensive deforma-
219x8.0. The final elongation (50 mm gage length) ranged tion prior to fracture, even for the two tests in which frac-
from 31 to 38 percent. ture occurred at the junction of the slot and the gusset plate.
The shear lag factor calculated according to Equation 2 When fracture was at the mid-length, the failure surfaces
ranges from 0.75 to 0.81 for these specimens. All the spec- were of a "cup-cone" configuration, which is typical of a
imens except S 1-1 had a fillet weld across the junction of ductile tensile fracture.

ENGINEERING JOURNAL / FOURTH QUARTER / 2000 135

 An example of the load vs. deformation curve (S 4-1) is fracture. As discussed earlier, the reduction factor, U, must
presented in Figure 5, and the test results of all the speci- be included in the calculation of Ae in Equation 5.
mens are summarized in Table 1. Calculated values in the The calculated capacity of the members according to
table were obtained using measured dimensions and mate- Equations 4 and 5, but using t= 1.0, is shown in columns
rial properties (Cheng et al., 1998). The net cross-sectional 8 and 9 of Table 1. The value of the resistance factor is set
area was used for strength calculation of Specimen S 1-1 at unity in order that the design equations can be compared
since only longitudinal welds were present. In all other directly with the test results. Equation 5 governs for all
cases, the gross cross-sectional area was used. specimens, whether or not the resistance factors are includ-
The capacity of a tension member according to the AISC ed in the calculation.
LRFD Specification (AISC, 1993) is given in Section Dl. The test load recorded in this program is the ultimate
The limit states are yielding in the gross cross-section and (fracture) load. Thus, it is the predictions shown in column
fracture in the net section. Thus, the capacity according to 9 that should be compared with the test values, column 6. It
the LRFD Specification is the lesser of can be seen that the AISC predictor equation (Equation 5,
but with t = 1.0) significantly underestimates the actual
Pn = tFyAg (4) fracture load. The mean value of the ratio, AISC predicted
ultimate load to test ultimate load is 0.78 and the standard
and deviation is 0.02. It is observed that this under-prediction is
about equal to the HSS Specification shear lag factor for
Pn = WuAe (5) these specimens. If the shear lag factor is set at unity, then
the predicted ultimate loads are those shown in column 10
The value of the resistance factor, t, is to be taken as of Table 1. When the predicted-to-test ratio is examined
0.90 in Equation 4 and as 0.75 in Equation 5. The lower now, the mean value is 0.98 and the standard deviation is
value used for Equation 5 is in recognition that failure is by 0.01.

Fig. 3. Fracture of specimen S 4-1 at mid-length. Fig. 4. Fracture at slotted end of specimen S 1-1.

136 ENGINEERING JOURNAL / FOURTH QUARTER / 2000

 On the basis of this comparison, it can be concluded that element program ABAQUS (Hibbitt, Karlsson, and
there was negligible shear lag in these test specimens. This Sorensen, 1994). In the analyses, the materials are assumed
is discussed further below, however. to behave according to the incremental, isotropic-hardening
The information in Table 1 shows that the ductility of plasticity model and connection failure is assumed to occur
specimens that fractured at the slotted end (S 1-1 and S 5- when the equivalent plastic strain in any part of the connec-
1) is significantly less than those that fractured at their mid- tion reaches a critical value. This critical value was deter-
length (the remainder). mined from the tensile coupon tests, but in order to model
The effect of shear lag should not necessarily be dis- and predict the connection failure adequately the material
missed. Figure 6 shows measured strains at the slotted end properties beyond the initiation of necking had to be estab-
of Specimen S 4-1 for a load that is about 60 percent of the lished. The details of how this was done are available else-
fracture load (the figure also shows strains obtained from a where (Cheng et al., 1998). In all of the analyses, failure
numerical analysis; this will be discussed later). According was assumed to occur in either the gusset plate or in the
to Figure 6, the strain in the region of the tube that is in con- tube, but not in the weld.
tact with the gusset plate is about double the lowest strain. The numerical analyses provided a good prediction of the
This reflects the fact that there is shear lag in the system-not behavior of the connection. The results are shown in
all of the cross-section is at the same strain as the member Column 11 of Table 1, and the predictions are always with-
enters the gusset plate. Thus, when the weld length is short in 2 percent of the physical test values. In each case, the
relative to the member size the large stresses at this location numerical analyses also correctly predicted the location of
(i.e., shear lag) may result in fracture at the leading end of fracture, and it also predicted the reduced ductility for both
the slot-tube junction rather than at the mid-length of the S 1-1 and S 5-1 that was observed in the tests. Although the
tube. This is what was observed in Specimen S 5-1. The analysis was not able to predict the deformation at fracture
stress concentration at this location is also elevated when as accurately, it was still able to provide a reasonable pre-
there is no transverse weld: S 1-1 fractured at the slotted diction of the connection ductility. Figures 5 and 7 illustrate
end and at a reduced ductility as compared to the specimens the good agreement between the actual load vs. deformation
with the weld end return. However, even with the stress response and the numerical solution for Specimens S 4-1
concentration, the majority of the test specimens fractured and S5-1, respectively, except for the deformation at
at a location other than the slot-tube junction. This can be fracture.
attributed to the stiffening effect provided by the gusset The strain distribution shown in Figure 6 and the normal
plate. The tube is restrained from contracting circumferen- (longitudinal) stress distribution shown in Figure 8 confirm
tially by the gusset plate, whereas at mid-length the tube is the expectation that there is a significant stress concentra-
free to contract. This effect is explored further in the numer- tion at the slotted end, right under the gusset plate. The high
ical analysis. stresses reflect two features: the fact that the true stress vs.
strain material response is used and the confining (von
NUMERICAL ANALYSIS Mises) effect at the junction of the tube and the gusset plate.
As illustrated in Figure 8, a considerable increase in the
In order to study the behavior of the connections, elasto- stress concentration occurs when the weld length is reduced
plastic numerical analyses were carried out using the finite from 345 mm for S 4-1 to 275 mm for S 5-1, which are

2500 800

xrrrrTTrrrTtrxir^
2000
Load 600 -h
kN
Strain
1500
x10"6 400
1000 test test
analysis 200 analysis
500
HSS 0U
|iiii|iiiiIiiii|iiiiIiiiiJ
0
-90 -45 45 90
0 20 40 60 80 100 120 140
Displacement, mm Angle in degrees
Fig. 5. Load vs. displacement specimen S 4-1. Fig. 6. Strain Distribution in S 4-1 at 500 kN, slotted end.

ENGINEERING JOURNAL / FOURTH QUARTER / 2000 137

identical specimens otherwise. The stress concentration use of a short weld length or the absence of transverse
also increases significantly when there is no transverse welds across the thickness of the gusset plate weld may
weld. Because of these higher stress concentrations, S 1-1 increase the stress concentration sufficiently that fracture
and S 5-1 have a lower ductility and failed at the slotted will take place where the tube enters the gusset plate. In
end. such a case, ductility will be reduced. Nevertheless, in all
Notwithstanding these observations regarding the stress the configurations investigated the slotted round tubular
concentration (shear lag) in the slotted tube-to-gusset plate members exhibited considerable ductility, regardless of the
connections, in all of the physical tests the connections have location of fracture.
achieved, or came close to, the ultimate capacity based on Based on the tests and the numerical analyses reported
the gross area (except, net area for S 1-1). Furthermore, in herein, it is concluded that shear lag does not significantly
all test specimens except S 1-1 and S 5-1, necking and frac- affect the ultimate strength of slotted round tubular sections
ture occurred at the mid-length of the tube. This indicates that are welded to a gusset plate. The shear lag expression
that the effect of shear lag on the ultimate strength of the given in Section 2.1(b) of the AISC Specification for the
tube is not significant. Because of the shape of the circular Design of Steel Hollow Structural Sections does not accu-
tube and the constraint provided by the gusset plate, the rately represent the behavior of a slotted tube-to-gusset
tube is restrained from contracting circumferentially and, as plate connection. It is recommended that these connections
a result, tension hoop stress is developed in the tube at its be designed with no reduction for shear lag as long as the
junction with the gusset plate. This increases the effective weld length is at least 1.3 times the tube diameter. The rea-
stiffness of the tube at the slotted end as compared to the sons for this limitation reflect the ranges of the weld length
mid-length, where the tube is free to contract. As a result, tested in the program and the reality that the block shear
the tube at the slotted end location is capable of carrying a failure mode is likely to govern when the weld length is less
higher load than at its mid-length for the same level of the than this limitation. It is also recommended that a transverse
equivalent plastic strain. Consequently, fracture occurs at weld be used across the thickness of the gusset plate at the
the mid-length of the tube because of the higher average location where the slot contacts the gusset plate (this may
plastic strain. Of course, if the stress concentration is too not be possible if the connection is made in the field
high, the slotted end could fracture before the mid-length because it would demand close tolerances). Use of a trans-
maximum load carrying capacity is reached. verse weld increases the ductility of the slotted member sig-
nificantly without incurring much extra cost.
CONCLUSIONS AND RECOMMENDATIONS
REFERENCES
The numerical analyses were able to predict the ultimate
load and location of the fracture. In addition, they provided AISC (1993), Load and Resistance Factor Design
a reasonable estimate of the ductility of these slotted tubu- Specification for Structural Steel Buildings, American
lar members. The analyses and the physical tests showed Institute of Steel Construction, Chicago, IL.
that the slotted end of the tube is stiffened significantly as
the result of the constraint provided by the gusset plate. In AISC (1997), Specification for the Design of Steel Hollow
most of the physical tests, fracture occurred in the main por- Structural Sections, American Institute of Steel
tion of the tube and not in the connection region. However, Construction, Chicago, IL.

900

Gusset

700- -90
Load HSS
Stress -\ S4-1
kN
MPa A S5-1
nominal
500-U

300
-IiiiiIiiii'iiir
0 10 20 30 40 50 60 0 30 60 90
Angle in degrees
Displacement, mm

Fig. 7. Load vs. displacement specimen S 5-1. Fig. 8. Normal stress around tube at slotted end, load 2100 kN

138 ENGINEERING JOURNAL / FOURTH QUARTER / 2000

AISC (1997), Hollow Structural Sections Connections Engineering, Vol. 22 No. 4, pp. 651-659.
Manual, American Institute of Steel Construction, Hibbitt, Karlsson, and Sorensen (1994), ABAQUS/Standard
Chicago, IL. User's Manual, Hibbitt, Karlsson & Sorensen Inc.,
Cheng, J.J.R., Kulak, G.L., and Khoo, H.-A. (1998), ABAQUS Version 5.4, Volume 1 and 2, Pawtucket, RI.
"Strength of Slotted Tubular Tension Members," Korol, R.M. (1996), "Shear Lag in Slotted HSS Tension
Canadian Journal of Civil Engineering, Vol. 25, Members," Canadian Journal of Civil Engineering,
pp. 982-991. Vol. 22 No. 6, pp. 1350-1354.
Girard, C, Picard, A., and Fafard, M. (1995), "Finite Salmon, C.G and Johnson, J.E. (1996), Steel Structures,
Element Modeling of the Shear Lag Effects in an HSS Design and Behavior, 4th Edition, Addison-Wesley, New
Welded to a Gusset Plate," Canadian Journal of Civil York, NY.

ENGINEERING JOURNAL / FOURTH QUARTER / 2000 139