Professional Documents
Culture Documents
(1996) 3D Simulation of Bolted Connections To Unstiffened Columns I. T-Stub Connections PDF
(1996) 3D Simulation of Bolted Connections To Unstiffened Columns I. T-Stub Connections PDF
(1996) 3D Simulation of Bolted Connections To Unstiffened Columns I. T-Stub Connections PDF
169-187, 1996
© 1997 Elsevier Science Ltd. All rights reserved
Printed in Great Britain
PII: S0143-974X(96)0~48-X 0143-974X/96 $15.00 + 0.00
ELSEVIER
A r c h i b a l d N. S h e r b o u r n e a & M o h a m m e d R. Bahaari b
ABSTRACT
NOTATION
169
170 A. N. Sherbourne, M. R. Bahaari
1 INTRODUCTION
i
~
Interface Elemmmts
f
I \ T-web
.
/ ~ T-flange
======it= Column - : Z
x Hans*
analysis. In his approach the column flange is treated as a T-stub flange with
specified effective length and both components can be designed independent
of each other. His proposal is not applicable for beams with spans larger than
30 times the beam depth. In its Appendix J, Eurocode No.33 accepted the
latter approach to calculate the tension resistance of the column flange, either
stiffened or unstiffened, and the beam endplate in terms of equivalent T-stubs,
independently. The force distribution among the bolts at the ultimate limit
state is taken as proportional to the distance from the center of rotation, in
general, or else plastic. The code, in fact, uses the component method (see
Ref. 4 and also Ref. 5) in which depending on the individual stiffness of
various components, the applied load is distributed among them.
An external tensile load on the connection will reduce the pressure between
the T-stub flanges caused by bolt prestressing. However, depending on the
flexural rigidity of the T-stub and the size of bolt, additional forces may be
developed near the flange tip. This phenomenon is referred to as prying and
172 A. N. Sherbourne, M. R. Bahaari
; Elements
V'///////////ffL b
T.
\
of I-
T-web
T-flange
I
(a) X-Section Symnt Ca) Elevati~
Line
v
it increases the bolt force and may be detrimental to the strength and perform-
ance of the fasteners. 6 Agerskov 7 considered three basic cases of forces acting
on endplates and suggested corresponding design equations. He found that the
plate thickness and bolt diameter have the greatest influence on the prying
action. This reaction, which affects the bolt and, consequently, endplate
design, cannot be exactly allocated through analytical models, other than the
finite element, not even when supported with experiment.
T-stub to column connections usually fail due to plate plasticity in bending
and shear, bolt or weld fracture, excessive plasticity of the column flange in
bending and/or column web shear failure. The bolts are loaded primarily in
tension; however, in thin plates, a considerable amount of biaxial bending is
induced by their deformation. 8 In such connections, the axes of rotation of
the T-flange and column flange are at right angles; the planes containing the
tensile forces are also perpendicular to each other. Therefore, only a three-
dimensional (3D) model can handle the effects of so many parameters on the
behavior and load-carrying capacity of the connection. A two-dimensional
(2D) analysis is, by its very nature, incapable of representing the variation in
the transverse direction. Some researchers have laded to resolve this dilemma
3D simulation of bolted connections~I 173
One of the interesting and, at the same time, difficult aspects of bolted
connection analysis is the unpredictability of the actual support conditions at
the back of the endplate. Obviously, the T-flange would pull away from the
adjoining flange around the T-web to a varying extent depending upon the
flange dimensions, bolt size and position, material properties and, especially,
the load level. At the same time, when it tends to bear against the other flange,
it should not move freely through the adjacent component. Therefore, the
condition at the back of the T-flange constitutes a 'variable boundary value
problem' that can be solved only by an iterative approach. The appropriate
element available in ANSYS is STIF52, a 3D interface, modeling the contact
between the two surfaces. 11 The inter-surface stiffness is defined as one order
of magnitude larger than the T-stub or column flange axial stiffness. A coef-
ficient of friction equal to 0.5 is defined for sliding resistance while the inter-
face is closed.
The stress-strain relationship for the elements of the web and flange(s) is
taken as elastic-strain hardening as shown in Fig. 3(a). The tangential stiffness
L
~uzn
/ Op
/
8y lley 21ey 8p 0.006 8ep
(a) (b)
Fig. 3. Idealized non-linear tr-E curves: (a) column and T-stub elements; (b) bolt components.
3D simulation of bolted connections--I 175
3 ANALYSIS
One of the main advantages of the finite element model over other models,
such as mechanical or mathematical types, is that the complete force-deflec-
tion relationship can be monitored very easily. Tests provided by Back and
Zoetemeijer, ]2 are used to examine the reliability of this model. T-stub flanges
were 17, 20, 25 and 32 mm, among which the first two are high strength Fe
52 steel with yield stress equal to 357 and 364 MPa, respectively, and the last
two are of grade Fe 37 with yield stress equal to 282 and 272 MPa, respect-
ively.
Depending on the flexural rigidity of the flange and the properties of the
fasteners, three possible failure modes can be expected. The first mode, (A),
176 A. N. Sherbourne, M. R. Bahaari
is dominant when the flanges are heavy in comparison with the bolts. Then,
the flanges of the T-stubs separate from each other due to plastic deformation
of the bolts. The failure load is equal to the sum of the failure loads of the
bolts. In the second mode, i.e. mode (B), yield lines develop in the flanges
along the fillet between the flange and the web of the T-stubs. The bolts may
fail simultaneously with the formation of flange yield lines. The third mode,
(C), relates to yield line development in the flanges near the bolts and the
fillet. In fact, the connection may experience more than two modes through
its load history; however, the deformations just prior to the ultimate capacity
are of interest in limit state design. With a very thick flange plate, for instance,
the initial opening of the plates might be larger than the elongation of the
bolt shanks and the prying force comes into effect (Mechanism B). However,
around final loading, the elongation of the bolts is equal or larger than the
deflection of the flange plates and, therefore, no prying force persists and,
eventually, the connection fails through Mechanism A.
In all the cases analyzed, initially the external load reduced the contact
pressure between the flanges until separation at the bolt line occurred.
Depending on the flange thickness, bending in the outer portions of the flanges
develops prying forces acting between the bolt line and the edge of the flange.
Figure 4 shows the variation of the prying force ratio, Q/TR, with increasing
applied tensile force. In the 17 mm flange, prying was about 30% of the
applied tension at a value of applied load equal to bolt pretension, Bpl, which
O
0.2
t~ 0.15
0.1
0.05
0 ¢~-'u--'~- ~ I I I I I I I I
0 20 40 60 80 100 120 140 160 180
Applied tension per bolt, Ta (kN)
gradually reduced to 25% near ultimate. For this thickness, the model is more
flexible than the test data, but the bolt force converged to the test value (Fig.
5). Although the flange was made of high strength steel, the test data is stiffer
than the 32 m m T-flange; this seems to be erroneous. In the 32 m m T-stub,
the first mode (A) occurred in the model involving complete separation of the
flanges at loads greater than 140 kN, and thereafter, the bolt force was equal
in magnitude to the applied load per bolt, i.e. the failure mode changed from
(B) to (A). The bolt force diagram is flatter and the first and complete separ-
ation loads of the flanges are very close to each other (Fig. 6). In fact, much
of the increase in applied load is spent to loosen contact and does not change
the curvature of the plate. The remaining T-stubs of lesser thicknesses perfor-
med according to the second mode with different prying actions. The situation
for 20 mm, with a higher yield stress, is very similar to a 25 m m T-flange
with normal steel (Figs 7 and 8, respectively) which validates the importance
of material properties. The prying ratio for both these connections is about
0.10. A summary of the test and computational results are shown in Table 1
for the T-hanger. The applied load per bolt, bolt force and prying ratio
resulting from the tests for various T-stubs are shown in columns (3), (4) and
(5), respectively. Columns (6) to (8) list the corresponding values resulting
from computation.
In a separate case study, it was assumed intentionally that the 17 m m T-
~ JsJ
I
16 ..~--. s 3.6
ET-17mm Gap(Test)
--li.-
Gap(Model)
120"1 _. __~ ,~'
Tb (Test) 2.4
8 0 ~ ............ f Tb (Model)
E
E
I m, ,, m" I / 1.2
40~ ~.6
1", / bolt, kN
Fig. 5. Test vs 3D model: bolt force and maximum separation, gap, of a 17 mm T-hanger.
178 A. N. Sherbourne, M. R. Bahaari
I sf
2.4
s ~
5
15C
12C ___ ~_
~ '7
•"
.,""
Gap (Model)
Tb (Test)
a
Tb (Model)
-1.8
-1.5
-
E
E
I'
-1.2 [~
f
J • (_9
6(
~ 4T~ .."
"0,9
•0,6
31: -0.3
C~ , , , 4~0 ' ' ' gO ' ' ' l:b0 ' ' ' 1~0' ' '2
T, /bolt, kN
Fig. 6. Test vs 3D model: bolt force and maximum separation, gap, of a 32 mm T-hanger.
210.I 4
• (.9
•1.5
'1
'0.5
~'g 4,0 ' ' ' go ' ' 'lt, O' ' 'lt, o' ' '260
0
T, / b o l t , k N
Fig. 7. Test vs 3D model: bolt force and maximum separation, gap, of a 20 mm T-hanger.
3D simulation of bolted connections--I 179
210"
M el)
.2.5
120-
-2 ~;
9(
-1.5
60-
30 -0.5
s /
) T , 4~0 ' ' ' 8b ' ' ' 1~0 ' ' ' 1t~0 2b0
Ta /bolt, kN
Fig. 8. Test vs 3D model: bolt force and maximum separation, gap, of a 25 mm T-hanger.
TABLE 1
Summary of Test and Analytical Results Corresponding to T-stub Connections
flange was made of mild steel. There was almost a 10% increase in prying
forces at both yield and ultimate load levels. However, the behavior mode did
not change. Thus, it is extremely important that, in any experimental study,
material properties are tested individually and key values, like yield and ulti-
mate stresses and strains, i.e. strain hardening, reported. The correlation
between the model and test data is very good, in general.
16 I
16
18 18
t
I
:it
.
, m
I
*l',~
18
-~+
18
-~
T
, / r=21
HE240B HE 160 M
'
i i!i
I
![i
,..,
I
I,
, ill I il,
it
I
I
¢o
144 ¢---160--~
+-- 160 --~
+1 ÷
+ +
t" 480-- I
Fig. 9. Geometric dimensions of the test specimens12for T-stub to unstiffened column connec-
tions.
o'y = 275 MPa, can be compared to each other. The first pair, initially, behaved
exactly the same until Ta = Bpl, from which they deviated and the prying ratio,
QITs, changed significantly [Table 1, column (8)]. The second pair retained
almost the same variation over the entire loading.
It is seen from the pattern of deformation of a 17 m m T-section to a wide
flange HE240B column connection, [Fig. 1 l(a)], that the T-flange deforms
more or less as a T-hanger about its toe line; however, the column f a n g e
bends about its web plane. Since their stiffness properties are similar, they
make contact at the comer of the T-stub. In a 32 m m T-stub, on the other hand,
the flexural deformation of the T-flange is small compared to the elongation of
the bolt and the tee part behaves much like a single bolt in tension. The
column flange, which is now much more flexible than the T-flange, deforms
as a cantilever plate about the web plane, with its maximum between the
horizontal bolt line and T-web location [see Fig. 11 (b)]. Therefore, their con-
tact remains along the side edge, between the bolt and T-web, because of
prying of the column flange on the T-section. In the same way, the 32 m m
182 A. N. Sherbourne, M. R. Bahaari
Tsh= 17ram
0.4 ]- = All bolts 16ram;Bpl = 110 kN
0.35 ~- ~ Tsh= 20ram
~5 0.25 =
0.2
et0 0.15
0.1
0.05
0 I
0 20 40 60 80 100 120 140 160 180
Applied tension per bolt, Ta (kN)
Fig. 10. Variation of prying force ratio for T-stub to unstiffened column connections.
T-flange separates from the HE160M column flange, which is stiffer than the
previous column. The column flange does not deform that much to keep its
contact with the T-flange at the ultimate load level, whereas the position of
the prying force around yield is at the same place as before.
The locations of prying force resultants under the applied tension causing
first yield together with ultimate capacity of the connections are illustrated in
Fig. 12. If the column flanges are not reinforced by adequate stiffeners, the
pattern of deformation is quite different from the T-hanger and the location
of the prying forces shift from the toe lines to the side edges of the flange.
Therefore, as a general rule and based on Fig. 12, where the locations of
prying forces for a variety of T-stub connections are illustrated, when the
relative stiffness of the column flange is much less than that of the T-section,
the prying force develops due to column flange deformation and is located
along the side edge around the bolt horizontal line [Fig. 12(e, f and i)]. When
the relative stiffnesses are similar, it still stays on the side edge with a tendency
toward the corner of the T-section [Fig. 12(d and h)]. When the T-stub is
much more flexible than the column flange, since the base is cantilevering
about its web, the prying force location remains near the corner but along the
toe edge [Fig. 12(g)].
In terms of bolt force, and with the same reasoning as above, its variation
depends mostly on the relative thicknesses of T-flange and column flange
together with the stiffness of the bolt. The 25 mm T-flange when connected
to the HE240B, with 17 m m as the thickness of its column flange, shows a
3D simulation of bolted connections--I 183
ANSYS 4.4A
D E C 27 1992
**:51:35
P L O T NO. 2
P O S T 1 STRESS
STEP=7
ITER=30
UZ
D GLOBAL
D M X =6.348
SMN =-0.086265
S M X =6.348
-0.086265
l~m 0.628697
1 344
2 059
2 774
3 489
4 204
4 918
5 633
6 348
ZV =I
DIST=334.51
XF =20
YF =86
ZF =-Ii
ANGY=40
CONE=30
FACE HIDDEN
-0.06181
0 527067
l
-- 1 116
1 705
2 294
2 883
3 471
mmm 4 06
4 649
5 238
4 o I t°!
(a) Ts. = 17mm
I
(b) Tsh = 25ram
IO
(c) T.h = 32ram
0 O,I
©
I
(d) T= = 17mm
-'-"
(e) T,o = 2 5 m m
I (f) %° = 32ram
Fig. 12. Locations of prying force resultants for (a)-(c) T-stub hangers; (d)--if) T-stub to
HE240B column flange; (g)-(i) T-stub to HEI60M column flange.
12% increase in bolt force, while the HE160M, with a 24 mm column flange
thickness, almost breaks contact with the T-flange, there being little prying
effect (Table 1). Figures 13 and 14 illustrate the bolt forces together with
the maximum separation of flanges and compare with test results for these
two connections.
20I0: ~
180"t
160"
" "'" -3.6
4
-3.2 ~"
- E
-2.8 t~
" (.9
120" •" "2.4 ~
Tb [Test] e-
~_ loo. 2 ~
+
4
18 /J
,// -3.6
] /s fJ
-3.2 ~
' 2TO ' 40 ' dO ' E~O ' 160' I~0' I~K)' 11~0' 11~0' 260' 2 :8
T, / b o l t , kN
Fig. 14. Behavior of a 25 mm T-stub connected to HE160M column flange: test vs 3D model.
186 A. N. Sherbourne, M. R. Bahaari
base. The predicted results are within the range of accuracy of experimental
values and the correlation is certainly much better than that currently available
in the literature. The following general observations are in order:
(1) Since the maximum bending stresses of the column and T-stub flanges
are perpendicular to each other and also since thin column or T-flanges
are in a bi-axial stress condition, only a three-dimensional model can
satisfactorily predict their interaction.
(2) The location of the prying forces in a T-hanger is essentially along the
toe edge of the T-flange above the bolt line and shifts toward the comer
with increasing stiffness of the flange.
(3) Prying forces in the T-hanger increase with decrease in relative stiffness
of flange to bolt.
(4) In the T-stub to unstiffened column flange connection, the prying force
is essentially along the side edge. With increasing stiffness of the T-
stub, it shifts toward the T-web. In this case, the column flange deforms
as a T-stub at right angles with the original T-stub flange which acts
as a rigid base.
(5) Most properly designed T-stubs behave through the second mode of
failure; the first and third modes are extreme cases of too flexible or
too stiff flanges. Therefore, complete double curvature action along the
flange span is unlikely.
(6) The material properties of the flanges in the joint are important for
predicting the behavior and force distribution of T-stub connections.
ACKNOWLEDGEMENTS
The writers would like to thank the NSERC of Canada for direct research
assistance through grant A1582 to the primary author. The financial assistance
of the Ministry of Culture and Higher Education of Iran for support of the
second author's research work is appreciated.
REFERENCES
3. Eurocode No. 3: design of steel structures, Part 1--general rules and rules for
the buildings, ENV 1993-1-1.
4. Yee, Y. L. and Melchers, R. E., Moment-rotation curves for bolted connections.
Journal of Structural Engineering, ASCE, 1986, 112(3), 615-635.
5. Jaspart, J. P. and Maquoi, R., Prediction of the semi-rigid and partial strength
properties of structural joints. SSRC Annual Technical Session and Meeting,
Lehigh, USA, 1994.
6. Kulak, G. L., Fisher, J. W. and Struik, J. H. A., Guide to Design Criteria for
Bolted and Riveted Joints, 2nd edn. John Wiley and Sons, New York, 1987.
7. Agerskov, H., Bolted endplate connections in steel structures. In Connections in
Steel Structures, Behavior, Strength and Design, ed. R. Bjorhovde, J. Brozzetti
and A. Colson. Elsevier Applied Science Publishers, London, 1987, pp. 52-59.
8. Sherbourne, A. N. and Bahaari, M. R., 3D simulation of end-plate bolted connec-
tions. Journal of Structural Engineering, ASCE, 1994, 120(11), 3122-3136.
9. Krishnamurthy, N. and Graddy, D,, Correlation between 2 and 3 dimensional
finite element analysis of steel bolted endplate connections. Computers and Struc-
tures, 1976, 6, 381-389.
10. Bahaari, M. R. and Sherbourne, A. N., Computer modelling of extended endplate
bolted connections. Computers and Structures, 1994, 52(5), 879-893.
11. ANSYS, Structural nonlinearities seminar for revision 4.4, Swanson Analysis
System, Inc. (1989).
12. Back, J. De and Zoetemeijer, P., High strength beam-to-column connections, the
computation of bolts, t-stub flanges and column flanges. Report 6-72-13, Stevin
Laboratory, Delft University of Technology, Delft, Netherlands, 1972.