J. Construct.

Steel Research 3011994) 177 196

1994 Elsevier Science Limited
Printed in Malta. All rights reserved
0143-974X/94 $7.00
ELSEVIER

On the Numerical Analysis of Endplate Connections

N o r b e r t G e b b e k e n , Heinrich Rothert & Bettina Binder

Universit/it Hannover, Institut fiir Statik, Appelstr. 9a, D-30167 Hannover, Germany

Received 2 September 1992; revised version received 24 March 1993:

accepted 22 June 1993)

ABSTRACT

In this paper rowel aspects ~?]Jinite element modelling 01 bolted steehiork connec-
tions are presented in order to predict their load-earryin~t hehaHour and to cah'ulate
their limit loads. Parametric studies with respect to various ~leometrie parameters
hal,e been carried out in order to determine their inJtuenee on the.flexibility O[the
conm,etion. The samples studied are extended endplate connections as used in
beam-to-beam or heam-to-cohmm connections, bolted hy pretensioned hi~lh-stren.qth
screws. The main emphasis is given to the consideration ~?]material nonlinearity, to
the contact between d@~rmahle members, and to aecurate finite element nlodellin.q.

! INTRODUCTION

Structural connections can be designed as bolted endplate connections

(Fig. 1). The welding of stiffeners is very expensive. Therefore, unbraced
connections are applied to an increasing degree. Unfortunately, they have
the property of being much more deformable than stiffened connections.
These deformations can affect the load-carrying behaviour of frames
significantly. 14 As a consequence, the deformability has to be taken into
consideration as pointed out in Refs 1 or 2. In frame analysis, the physical
model of a connection in the sense of a macro element can be a spring
representing a bending moment. Its properties are given by the moment
versus rotation curves of the connection. The analysis of the nonlinear
load-carrying behaviour of structures with respect to the design process is
feasible only when the finite element method is applied. Consequently, it
is logically consistent to apply this powerful tool for the investigation of
all members of a structure in order to develop a conclusive design method.

177
178 N. Gebbeken, H. Rothert, B. Binder

a)

b)
a; +
'i ~--~~ tension area ? 6k

-:~ , i FB

i i J

IH ,,,,'{3

F8

Fig. 1. Beam-to-column connection: (a) type "WS" (weak flange, stiff endplate) according to
Ref. 4: {b) endplate connection.

Bolted endplate connections were first studied experimentally in order

to develop design criteria (e.g. Refs. 3 8). Simple approaches like beam
models (e.g. Ref. 6) or yield-line techniques (e.g. Ref. 8) were at first
proposed for the dimensioning of such connections ~by hand'. Also
moment versus rotation curves were developed from experimental inves-
tigations on one-to-one connections (e.g. Ref. 4). Because experiments are
time-consuming and expensive and therefore restricted to a few typical
connections, the finite element method was introduced in this field of
research. First attempts were published in Ref. 9 in 1976. The endplate and
the welded beam member were analysed purely elastically applying 8-node
isoparametric brick elements. Further finite element studies were carried
out in order to predict the elastic flush endplate behaviour. 1° The adjoined
member was assumed to be rigid, and was therefore neglected.
The studies presented in this paper focus also on the contribution of
the column flanges to the flexibility of unbraced beam-to-column con-
nections. The investigations have been carried out with two objectives.
On the one hand the load deformation behaviour is of interest in order
to provide input data for frame analyses, and on the other hand studies
 On the numerical analysis of endplate connections 179

are necessary to develop dimensioning formulas. Before going into the

details of finite element modelling, the phenomena considered are briefly
discussed.
It is known from experimental investigations that bolted connections
are plasticized in some regions even at relatively small external loads.
Consequently, the nonlinearity of the material behaviour has to be taken
into account. To model this, the J2 flow theory is adopted. In addition our
numerical calculations of connections result in strains of up to 10%.
Therefore, it is necessary to consider the material hardening. For this, a
program module has been coded for our institute's finite element program
Protech. The theory is based on a paper published by Simo and Taylor.l~
In bolted connections not all parts of the structure are connected rigidly
to one another. Owing to changes in the loading conditions, stresses will
be transferred through permanently changing contact areas. The size,
shape and location of these contact areas are not known in advance, but
the screw forces depend mainly on the contact conditions. This implies
that an incremental iterative method has to be applied. To take into
account accurately the real physical behaviour of bolted connections, the
theory for the contact between three-dimensional deformable bodies had
to be further developed. So far it has not been frequently examined for
special applications. A contact algorithm has been implemented into
Protech with respect to 8- and 20-node isoparametric brick elements for
frictionless contact.l 2
The coupling of these two nonlinear problems is absolutely necessary to
predict numerically the realistic load-carrying behaviour of bolted connec-
tions. If the required software is available, the main task for numerical
analysts as well as for structural engineers is to find a finite element model
which accurately represents the reality. In this paper the main emphasis is
given to the development of finite element models and to numerical
techniques associated with the investigation of structural connections. We
submit our experiences and results and try to formulate recommendations
for the application of the finite element method. Today's possibilities and
the limitations of the numerical calculation of steel connections are
discussed. Numerical results illustrating the methodology are presented.
In addition, a computer code has been developed on the basis of the
yield-hinge theory for the nonlinear analysis of arbitrary frames. 13't~ It
takes into account the deformations of connections using some known
moment versus rotation curves. 2'is This program is capable of calculating
the nonlinear load-carrying behaviour and limit loads of steel frames in
accordance with the German standard ~ and Eurocode 3. 18 For this, the
moment versus rotation curves are required as input data. However, we
shall not cover this topic in this paper.
180 N. Gehheken, tt. Rothert, B. Binder

2 ON THE APPLIED BOLTED ENDPLATE CONNECTION

First we chose a semi-rigid extended endplate connection, type IH

according to Ref. 7, with pretensioned high-strength screws because
Humer had presented some experimental results in Ref. 4. Using beam-
to-column joints with endplate connections, the flange forces of the beam
are transferred to the endplate and furthermore to the screws. This means
that the flange of the column and the endplate behave like plates in wide
regions. Coming from the column flange, the screw forces are transmitted
over the column web to the entire cross-section. As a result of the
deformations further stresses arise for the flange and the endplate owing to
the contact forces between them. These forces and the screw forces
produce a three-dimensional stress state in localized areas.
If one of the two connected media (endplate or flange) can be assumed
to be rigid as compared to the other one (Fig. 1), the displacement of the
rigid part is a function only of the deformable part. As a consequence, only
the deformable part of the connection needs to be considered for calcula-
lions.
According to Zoetemeijer s the connection fails when the yield lines are
formed as can be seen in Fig. 2. Their location is determined by the
principle of virtual work. This result has been transferred to a simple beam
model, which is identical to the model of Thomson and Agerskov,
presented in Ref. 6. Obviously, consideration of the prestressing effects of
the screws cannot be done with this simple approach. The ultimate limit
load of the connections, however, can be estimated roughly by applying
the proposed beam models. For finite element modelling Zoetemeijer's
investigations give valuable hints about the size and the shape of the

yield lines

Fig. 2. Yield lines according to Zt~etemcijcr, ~

 On the numerical analysis o1"endplate connections 18 l

Q) b)
I FI2
F
I
...... ........

{ ,
I

Fig. 3. (a) Simplified model restricted to areas with tension (view A A in Fig. 1); (bl
assumption of symmetry'.

portions of the connections which contribute mainly to the load deforma-

tion behaviour.
Assuming that the endplate is rigid compared to the flange, a further
simplification of the model is possible because the endplate imposes the
boundary conditions on the flange. The tension area (Fig. 1) of the
deformable part of the connection contributes mainly to the overall
deformation. As a result, it is sufficient to model only this part by finite
elements. Owing to the plastification of the flange or the endplate,
respectively, by increasing the load, the amount of the screw forces adjust
to one another. For this, a plane of symmetry results at the location of the
upper beam flange (Fig. 1). If the connection is subjected to plane bending
a further symmetry plane can be obtained as is demonstrated in Fig. 3.
Finally, only one-quarter of the tension area of the column member has to
be modelled for the finite element analysis.

3 PRELIMINARY STUDY OF ESTIMATING ENDPLATE

DEFORMATIONS

In order to check whether the assumption of rigi.2 endplates is '..,_<d or not,

the amount of the deformations of the endplates is calculated. :('t this, the
beam model of Thomson and Agerskov" is applicu, becau,,r, it ;,, :onveni-
ent for the purpose of estimation. Paying attention ~:_~ : , . endplate
stiffened by the welded I-profiles in reality, we can say that t!.,: - dculated
deformations are upper bounds.
The results of the calculation of an endplate connection attached to a
HE 300 B are shown in Fig. 4. In Ref. 7 (Standardized Connections in Steel
Construetion) an endplate thickness of 30 mm is recommended in this case.
182 N. Gehbeken, H. Rothert, B. Binder

E
E
.=_ 0,2 F - 200 kN
\\ . F - 400 kN
\\ "'".. F - 600 kN
~3 \\ , ""... .................... F - 7013 kN

\\ "~ "%.
\-.-. ~ "'%
~•-~ 01
j
~." ......
".... ~ . "'... ....

20 2'5 30 35 4;0
thickness of endplale in mm
Fig. 4. Displacements 6k of an endplate, attached at a HE 300 B, versus various plate
thicknesses dk due to different values of the load F.

It is obvious that the displacements ok are s,_._:;i ,~,:,,?ar..' . . . . . . . . . : ,

of the endplate. In addition they amount to only up to 5% of the
displacement of the flange of the connected column members. Conse-
quently, the assumption of rigid endplates is still valid.

4 A TWO-DIMENSIONAL FINITE ELEMENT MODEL

In order to obtain the most simple finite element model possible, the
connection is first reduced to a two-dimensional problem according to Ref.
17 (Fig. 5). The physical model is the plane stress approach. For this, the
finite element program Protech provides a suitable element library. Special
numerical techniques are used to take into account the pretension of bolts.
This study has been carried out in order to reproduce experimental results
which were published by Humer 4 and to provide first experiences with the
application of finite elements on this subject. Furthermore, the following
assumptions are introduced here:
t. The nonlinear material behaviour is considered by applying a
bilinear approach.
2. Nonlinear strain versus displacement relations are considered.
3. The screw geometry is considered in a simplified manner, and the
thread is ignored.
4. The friction between flange and endplate is neglected (I~ = 0).
5. For the friction between endplate and screw head only the limit cases,
completely stick (IL= ~c), and frictionless slip (l~ = 0), are considered.
 On the numerical analysis of endplate connection,s 183

Material data: Beam m e m b e r D I N 1025 Endplates D I N 17100

Steel grade St37 Yield stress or = 240 N / m m 2
Young's M o d u l u s E=210000 N/ram 2
Screws D I N 6914 Strength 10"9
Tensile strength an = 1000 N / m m 2

F]I

screw discretization

L, I i~ i!
! iJ : i! il
i j ii
l i l ! i ] i i f
i , : 7 " T i i li I

bl

,~'~ ~, ~, flange discretization

= I , i' ~ ; I i'i : ~F,
i I t I 'i

: , , , . : , !:1 . . . . !

Fig. 5. Two-dimensional finite element model of an endplate connection.

The reduction of the actual three-dimensional structure to a two-

dimensional finite element model has been carried out by applying a
'double elementation' in the bolt region. This considers the stiffness of
the screws on the one hand and the stiffness of the column flange on the
other hand. The elementation of the flange is continuous, and the finite
elements of the screw shank located behind it are connected with the
finite elements of the screw head. The elements next to the screw head
are tied to the elements of the flange. The dimensions of the flange part
were taken according to the effective width which is given by Humer. The
184 N. Gehbeken. ft. Rothert, B. Binder

structure is mainly subjected to bending. A constant bending moment

distribution along the edge of one element can be represented applying
an 8-node element which has quadratic polynomial shape functions
as an approach for the displacements. The area of the web fillet can be
well reproduced with isoparametric elements. 284 isoparametric 8-node
elements were used for this elementation. The finite element model is
subjected to the external load /:'~ acting on the web and to the force
Fv representing the pretensioned screw. These loads are not super-
posable because of the nonlinear problem. Therefore, Fv was modelled.
modifying the displacement boundary conditions for the screws. With
this, it is possible to consider prestressing exactly. These boundary
conditions were kept constant during the analysis, while the external
load FB has been increased by adding load increments. The deformed
configuration in the ultimate state is shown in Fig. 6. In this study we
applied a HE 200 B with high-strength screws M24 of quality 10"9. The
number of lines drawn on the element surfaces indicates at how many
Gaussian points plastification was observed. The black elements are
fully plasticized.
a)

t!I

Fig. 6. Two-dimensional model, deformed configurations, contact conditions between

screw head and flange: (a) completely stick (~- ~:); (b) frictionless slip 1~=0).
 On the numerical analysis qfl endplate connections 185

numerical results

/y--
600 *
+~ ~-'x"- --": ~ : ~ -x.

experimental resuffs (Humeri

400

+ . . . . + stick
x----- x slip
200
I

I
1 3
displacements in mm
Fig. 7. Load versus displacement curves of point A in direction of FB {Fig. 8).

Using the two-dimensional model and continuous elementation be-

tween screw head and the flange {this means completely stick in the
interface of the contact surfaces), the analysis of the endplate connection
led to an unsatisfactory result. When increasing the load, the elements at
the inner corner of the screw head and the flange were more and more
distorted, and finally the computer program stopped with a warning
message. In order to avoid this problem, the second limit case was
examined. Frictionless slipping between the screw head and the flange is
assumed in this case (Fig. 6(b)). A comparison of the numerical load versus
displacement curves with the experimental ones of Humer 4 is given in
Fig. 7. Material hardening was neglected in this analysis.

Remark

The analysis applying the two-dimensional model resulted in values of the

limit load that were significantly larger than the related first limit load
determined by Humer in his experimental investigations. Thus, the nu-
merical results are on the 'unsafe' side although hardening was neglected.
The two-dimensional finite element model seems to be too stilt although
the finite element grid is rather dense. When ignoring the real contact
conditions as shown in Fig. 6(a), one obtains an unreliable calculation.
This result made it advisable to develop a three-dimensional finite element
model.

5 A THREE-DIMENSIONAL FINITE ELEMENT MODEL

As already shown in Section =,

~ a complete endplate connection loaded in
the plane of symmetry can be simplified in several steps. Finally, we obtain
186 N. Gt'bbekcn, It. Rothert, B. B i m l o

the reduced model as shown in Fig. 3 which is sutticient to describe the

load-carrying behaviour. Applying the three-dimensional theory, the com-
putation requires a considerable amount of both memory, capacity and
computing time, owing to the number of degrees of freedom and the
nonlinearity of the problem studied. Except for item 1. the assumptions of
Section 4 are still valid. In order to model the complicated connection with
brick finite elements, some further simplifications are introduced:
• The circular cross-section of the screw shank is idealized to bc
square with an area identical to the cross-section. Otherwise it
would be possible to model the screw only by applying a largc
number of very small elements.
• The situation between the screw head and the endplate is assumed
to be completely stick (that means the connecting elements have
c o m m o n nodes at the interface).
• The real curve in the corner of web and flange is modelled in a
staircase-like manner.
With these assumptions a finite element model can be produced
relatively easily utilizing a preprocessor for the elementation. This pre-
processor has been written specially for this problem. Only the overall
geometry and the real diameter of the hole, the shank and the screw head
are required as input data. The transformation to a square bolt cross-
section with an identical area is carried out by the computer program.
However, the development of a preprocessor only for a single application
lacks general validity and cannot be generally recommended.
Our investigations were performed for the connection part shown in
Fig. 8. It is in accordance with the specimen studied by Humer in his first
experiment. Owing to the symmetry conditions, again only one-eighth of
the initial model and only one screw head had to be elemented. The
structure is loaded analogously to the two-dimensional model.
One objective in any finite element computation is to find an optimal
finite element grid which is as dense as necessary and as coarse as possible
to obtain the required accurate results. A minimum number of elements is
necessary to model the geometry and the discontinuities (screw, flange.
curves, screw hole). During further analyses the initial finite element model
has been improved in areas of large stress gradients or stress jumps.
respectively, until the limit load and the displacements changed their
values only very slightly (asymptotic convergence). In order to meet the
requirements of symmetry, the system is supported in the plane y = 0 in the
),-direction and in the plane z = 100 mm in the z-direction. The nodes at
the b o u n d a r y of the plane x = 0, which belong to the flange, are defined as
possible contact nodes. The external load F , is modelled by nodal forces at
 On the numerical analysis Of endplate connections 187

fff#f

100
u'~
8567 I
A FB
I
i

ISS '~-45 1
2O0 1 -

geometry and loading dimensions in mm

Fig. 8. Three-dimensional reduced structure: geometry of a HE 200 B in mm and loading.

the web. These nodal forces are equivalent to a continuously distributed

load. The length of the system was chosen according to the effective width
which is given by Humer. The real dimensions of the cross-sections can be
taken from Table 1. Figure 9(a) shows the model and its sufficiently dense
finite element grid. The deformed configuration with respect to the limit
state is plotted in Fig. 9(b).
For practice-related applications of frame analyses, moment versus
rotation curves of connections are required. For this the load versus
displacement curves which were determined with the help of the finite
element analysis have to be converted. The rotation angle 0 of the
connection approximately results from the difference in the relative dis-

TABLE I
Dimensions of the Cross-sections

Conm, ction I1 (re:n) B (ram) S (ram) T (ram) R (ram)

HE t80 B, M20 181.0 1790 9-0 t3-7 150

HE 180 B, M24 1800 180-0 81 13-2 15-0
H E 200 B, M24 202-0 200-0 8'8 14-5 18-0
H E 240 B, M27 2420 240-0 112 16-1 21-0
HE 260 B. M27 2610 260-0 t0.3 164 24.0
188 N. Gehbeken, tl. Rothert, B. Binder

o) b) >.__ ':" ,T,

~- ,q/ Fr--~_ .~ 11
~-" / /~ "~tl / t VT-'o'~U" /b
,~---_i/ ~/1 :l~ll
~:/ r~---~_>~, '/I/l~ /,'er~ -
:':/I-,/
i /[j__¢/ /,, / II/I~ " qH< ,'
f :
/5: ~i/.._ j

Y ~ Lg-_ ~ : :/./~ i

z'~-z z I

Fig. 9. (a) Finite element model: (bt deformed contiguration at the limit slate.

placements of the flanges of the beam divided by the distance between the
flanges (Fig. 1 or Ref. 3l. When assuming a rigid endplate, the relative
displacement in the compressed part equals zero. Therefore

(~K
tanO=~[~O for IOl<~ 1 i1)

holds. The related bending moment M is the product of the load F~ and
the distance between the flanges:

M=FB'h {2)

Remark

Applying the three-dimensional model, the results of Humer's experimen-

tal investigations can be well reproduced by applying brick finite elements.
This is verified by the comparison of certain moment versns rotation
curves, drawn in Figs lO(a) and (b). The results of Humer's experiments
were converted analogously in order to plot the moment versus rotation
curves. In the elastic domain, the calculated moment versus rotation
graphs agree well with those determined experimentally. The load-carrying
behaviour is still overestimated in the transition zone where the connec-
tions start to plasticize. But again, the calculated limit load is smaller than
the one determined in the experiment, if hardening is not considered in the
analysis. The numerical results agree well, if the material hardening is
considered. It is recommended to check the amount of strain during the
computation process. Plastification can be considered sufficiently only by
applying three-dimensional brick elements. The calculation of tile mon~enl
 On the numerical analysis olendplate connections 189

Q) b)
200
~125 numerical ~ "-"¢ numerical ~
.=_ c: results + ~ t~'~
x--"'~" L x
~ 100 ~//"<" experimental data
~= 150
o
*"
,/..
..~"
experimentaldata
E (Humeri E // (Humeri
75 l"
loo /
50 + + with ha[dening + + with hardening
x x rigid plastic x x rigid plastic
5O
25

0 -- I 0,'02 .... -4 O, 0"4 ~ + ~

0,02 " ~
0,0/+
rotation in radian rotation in radian
Fig. 10. Moment versus rotation curves, finite element analysis, experimental data accord-
ing to Humer: (a) HE 200 B, M24: (b) HE 240 B, M27.

versus rotation curves has been carried out for different conections that
H u m e r had investigated in his experiments.

6 A T W O - D I M E N S I O N A L FINITE ELEMENT MODEL

FOR E N D P L A T E A N D W E L D E D BEAM

In this section we study the 'SW" case (still" flange, weak endplate)
according to Ref. 4. F o r convenience the entire connection is simplified by
a two-dimensional model as shown in Fig. 1 1. This model is very similar to

Q; b)

l
~300 [UHI
4~
' 0

. . . . .11 . . . . .

~oo "~111 V
ii100 t o !i.¸.o

250 150 50 0
contact stress in N/mm 2

Fig. I1. Endplate connection (H E 300 B, M27): (a) distribution of contact stresses: (b)linite
element mesh (8-node elements), deformed contiguration at Mp1,l-7=237-2kNm: (c)
contact areas, visualized applying 'carbon paper" according to Ref. 5.
190 N. Gehheken, tl. Rothert, B. Binder

that presented in Re{. 9, but in addition we took the plastification into

account. The bending moment is represented by a linearly changing load
distribution.
First analyses have been performed with a very coarse mesh using linear
4-node elements. This finite element model proved to be sufficient for the
calculation of only the fully plastic moment of the beam member. If the
main emphasis is given to the ultimate load of a joint, the finite element
model has to be more precise so that the bending of the endplate, the
strength of the screws and the bending of the beam member also can bc
considered {Fig. 11}. The ultimate loads obtained in the present study are
given in Table 2. They agree well with the results presented in Re{. 8. If the
connection fails, and not the beam member, the limit load decisively
depends on the thickness dk of the cover plate because its deformation
influences the magnitude of the contact forces. For the purpose of
standardization the real thickness of the endplates is rounded off' and given
in steps of 5 mm. We were interested in the value of the fully plastic
moments and m answering the question whether the beam member "b" or
the connection part "c" causes the failure of the joint {last column of

TABI.E 2
Comparison of Our Cornputcd Finite Element Results with Those in Re{. 7 with
Respecl to Fully Plastic Moments and Failure ('rileria

Comle('tion & M r~' I 7 M vJ, I 7 A

{ram} {k,\" m) (kN m} ( R @ 7) ~<!i))

HE 12{1 B. M16. 20 23.5 23-o 2.35

HE 180 B, M I 6 20 55.8 53.0 5.30
ttE 180 B, M20 20 (>5.4 6S-2 4.11
lIE 200 B, MI6 20 t~l.2 5{).2 3.38
HE 200 B, M20 20 sO.5 ,~-7 3-~2
HFI 200 B. M24 25 90.7 '412 0.55
t1 E 240 B, M 16 20 76.6 -0.,; x.(14
HE 240 B, M2(} 20 105.1 I052 0-10
HE 240 B, M24 25 137.7 143.0 3.71
ItE 24{} B, M27 3(} 149.2 150.1 {).60
tIE 260 B. M2{} 2{1 1 l,R.0 lit 3 0-60
ItE 26{1 B. M24 25 !55.9 t59.2 2-07
IliE 260 B. M27 30 ISI.N 184-0 I 20
lie 300 B. M20 20 143.,} 141-I I.N4
Itt- 300 B, M24 25 194.5 191.f~ I 5I
HF. 300 B. M27 30 237.2 239.', 0.96
tlt( 300 B, M3() 30 264.l 268-,~ . 1.75
 On the numerical analysis olendplate connections 191

Table 2). The deviation A of the numerical results is less than 8% for the
fully plastic moment with regard to the results of Ref. 7. It becomes
obvious from Table 2 that the limit load of the connection is defined by
the fully plastic moment of the beam member 'b' if we concentrate on the
connection with the largest sized screws (except H E 180 B). For the other
cases the limit load is defined by the fully plastic moment of the connection
'c'. In Ref. 1 it is recommended to design connections in such a way that
the beam member fails and not the connection part. Thus this recommen-
dation is in contrast to the present results.
The deformed configuration and the associated contact stress distribu-
tion is plotted in Fig. 11 with respect to the HE 300 B, M27 connection.
The distribution of contact stresses is in good agreement with experimen-
tal results presented in Ref. 5. The location and the shape of the contact
area as well as the stress field were made visible with carbon paper as
shows in Fig. 11.

Remark

Again, the two-dimensional model is too stiff for the representation of the
real deformations. However, if only the ultimate load is of interest for the
dimensioning procedure or when the question has to be answered whether
the connection part or the beam member fails, the model chosen is efficient
and sufficient. For further studies we look forward to assembling the finite
element models presented in Sections 4 and 6 in order to study the W W
case (weak flange, weak endplate).

7 INFLUENCE OF INACCURACIES OF MANUFACTURING

Further investigations were concentrated on how the allowable inaccur-

acies of manufacturing (according to the German Standard DIN 1025)
affect the limit load behaviour of the connections. This is important
because inaccuracies in manufacturing actually enter the frame analysis by
assumed predeformations (imperfections) which represent geometrical and
material inaccuracies. On the one hand the thickness of the flange, and oil
tile other hand the thickness of the web of the connected beam member
have been varied. According to the German Standard DIN 1025 these
values may differ up to 12% from the nominal quantities. Table 3 shows
tile dimensions and the differences from the nominal quantities that were
used for the analysis of the cross-sections. The moment versus rotation
curves of the connections HE 200 B, M24 and HE 240 B, M27 are plotted
in Figs 12 and 13. The curves are plotted for the minimum, the
 192 N. Gebbeken, H. Rothert, B. Binder

TABLE 3
Dimensions of the Cross-sections for Parametric Studies:
(a) Flange, (b) Web

(a) Connection S (ram) T (ram)

HE 180 B, M20 8'5 12-5 14.0 15-5

HE 180B, M24 8-5 12.5 14.0 15-5
HE 200 B, M24 9-0 13.5 15"0 16-5
HE 240B, M27 10-0 15-0 170 19-0
HE 260B, M27 10.0 t55 17-5 t9-5

(b) Connection S (ram) T (mini

HE 180 B, M20 7-5 8 5 9,5 140

HE 180 B, M24 7-5 8 5 9.5 14.0
HE 200 B, M24 g-O 9 0 10-0 150
HE 240 B, M27 9 0 10'•) 11.0 17-0
HE 260 B, M27 94) I0-0 114) 17.5

a) b)

E E

~
Z Z
120 j j---
-~ 200
.~_

E 15o
8O

T - 15,0 mm I_ T - 17,0m m
40 T - 16,5mm T - Ig,o m m
T - 13,5mm 50 T - 15,0 m m

0,02 0,04 0 0,02 0,04

rotation in radian rotation in radian

Fig. 12. M o m e n t versus rotation curvcs, various thicknesses of the flanges: (a) HE 200 R
M24; (b) f i e 24O B, M27,
 On the numerical analysis o["endplate connections 193

a~ b)
E
Z

"~125

1oo E150

75
100
50 S - 9,0 mm S - 10,0 mm
l S - 10,0 mm S - 11,0 mm
S - 8,0 mm 50 S - g,o mm
25

0 0,02 0,0/+ 0,02 0,04 0,06

rotation in radian rotation in radian

Fig. 13. Moment versus rotation curves, various thicknesses of the webs: (a) HE 200 B,
M24; (b) HE 240 B, M27.

maximum and the nominal thicknesses of the flanges and the webs. The
calculations were carried out considering the material hardening.

Remark

It can be seen that the limit load of the connections varies by up to 7%.
This is less than the maximum variations of the thicknesses of either the
flange or the web. With these studies the influence of the inaccuracies of
manufacturing concerning the dimensions of the flange and the web
thickness on the limit loads of the connections can be estimated. It is
obvious that these imperfections have to be part of the assumed predefor-
mations in frame analyses.

8 SUMMARY AND CONCLUSIONS

As pointed out in the draft of Eurocode 318 and the associated literature, ~
a knowledge of the load-carrying behaviour of the connections will be
of more importance in the future especially with respect to stability
analyses and ultimate load analyses of steel frames, 13"15 as well as with
respect to the dimensioning of connections. Currently, one task is to
194 N. Gebbeken, ff. Rothert, B. Binder

develop methods and computer programs which provide the limit load
behaviour of complicated bolted connections accurately and effectively.
In this paper the main emphasis has been given to the questions: Ill
Which physical phenomena affect mainly the load-carrying behaviour of
structural bolted connections? (2) What are the suitable finite elements and
linite element models with application to structural bolted connections'? (3)
('an finite element computations provide moment versus rotation curves
for frame analyses considering semi-rigid .joints'? (41 How can software
tools be improved in order to reduce lime?
From our investigations we can conclude: contact phenomena and
material hardening effects have lo be considered in order to provide
accurate results. The real three-dimensional states of stress that lirst cause
plastification and softening of connections can be represented accuratel}
only when three-dimensional brick elements are applied.
Comparative analysis with the two-dimensional model and the three-
dimensional model shows clearly that the quality of the results depends a
lot on the finite element model chosen. The three-dimensional finite
element model can generally be recommended while the two-dimensional
model is advisable only for special investigations. Utilizing 8-node brick
elements, based on the displacement method, requires a very dense linite
element grid with a large number of degrees of freedom. We can recom-
mend the 20-node brick element or the 8-node enhanced assumed strain
brick element which has been developed and implemented recently.
As shown in Figs I0, 12 and 13, tlle computations provide moment
versus rotation curves which can serve as input data for nonlinear frame
analyses. Such studies have been published, for example in Ref. 2 by the
authors.
Inaccuracies in production and assembly result in limit load deviations
and have to be taken into account in frame analyses by assumed stress-free
predeformations that are affine to the first eigenmode.
Even with today's possibilities of generating the finite element model.
with the variety of finite elements and high-performance computers, the
development of finite element models remains a major task if the gener-
ation of finite element models as welt as the CPU time is not to exceed
economic limits. Therefore improved powerful preprocessors are indis-
pensable.
In order to develop design criteria as well as data-banks for moment
versus rotation curves, further investigations have to be carried out.
The results presented in this paper illustrate the suitability and the
applicability of the finite element method. Bolted extended endplate
connections subjected to plane bending were investigated. Some novel
 On the numerical analysis of endplate connections 195

aspects of the finite element modelling of c o n n e c t i o n s with nearly

rigid endplates are discussed. These and further analyses can serve as a
basis for r e c o m m e n d a t i o n s for the numerical analysis of steelwork
slructures.

ACKNOWLEDGEMENT

The a u t h o r s gratefully a c k n o w l e d g e the substantial s u p p o r t of the G e r m a n

N a t i o n a l Science F o u n d a t i o n ~Deutsche Forschungsgemeinschaft'.

REFERENCES

1. Background Documentation 6.09 to Eurocode 3, ECCS Committee 10, Beam

to Column Connections, TH Delft, March 1989.
2. Gebbeken, N., Binder, B. & Rothert, H., Zur numerischen Analyse yon
Kopfplatten-Verbindungen. Stahlhau, 61 (1992) 265 74.
3. Hedenkamp, A. & Lacher, G., EinfluB einer Zug-Normalkraft auf die
Verformbarkeit ausgesteifter Kopfplattenanschliisse. SchluBbericht zum
Forschungsvorhaben, Institut fiir Stahlbau, Hannovcr, Best.-Nr. T 2234, 1990.
4. Humer, C, Das Momenten-Rotations-Verhalten yon stcifcnlosen Rahmen-
knoten mit Kopfplattenanschltissen. Dissertation, Universit~it lnnsbruck, 1987.
5. Struik, J. H. A. & Fisher, J. W:, Guide to Desiqn Criteriafln" Bohed and Riceted
Joints. John Wiley, New York, 1974.
6. Thomson, K. & Agerskov, H., Versuche zur Ermittlung des Tragvcrhaltcns
yon KopfplattenstaBen in biegebeanspruchten IPE- und HE-B-Profil-Tr~i-
gern. Stahlbau, 42 (1973) 236 46.
7. Tvpisierte Verhindunqen im Stahlhochbau. Stahlbau-Verlags-GmbH, K61n,
1978.
8. Zoetemeijer, P., A design method for the tension side of statically loaded,
bolted beam-to-column connections. HERON, 20(1)(1974).
9. Krishnamurthy, N. & Graddy, D. E., Correlation between 2- and 3-dimen-
sional finite element analysis of steel bolted end-plate connections. Comput.
Struct., 6 (1976) 381 9,
10. Kukreti, A. R., Murray, T. M. & Abolmaali, A., End-plate connection
moment rotation relationship. J. Construct. Steel Res., 8 (1987) 137 57.
11. Simo, J. C. & Taylor, R. L., Consistent tangent operators for rate-independent
elastoptasticity. Comput. Meth. Appl. Mech. Enqnq, 48 (1985) 101 18.
12. Binder, B. & Gebbeken, N., Theoretical and numerical aspects of 3D finite
element elastoplastic contact analysis. In Computational Plasticity, ed. D.
R. J. Owen, E. Onate & E. Hinton. Pineridge Press, Swansea, 1992.
13. Gebbeken, N., Fine FlieBgelenktheorie h6herer Ordnung fiir r~,iumliche Stab-
tragwerke zugleich ein Beitrag zur historischen Entwicklung. Mitt. Nr.32/88,
Inst. for Statik, Universit/,it Hannover, 1988 and Stahlhau, 57 (19881 365 72.
196 N. Gebbeken, H. Rothert, B. Binder

14. Rothert, H., Gebbeken, N. & Binder, B., Nonlinear 3D finite element contact
analysis of bolted connections in steel frames. Int. J. Num. Meth. Engng. 34
(1992) 303 18.
15. Rothert, H., Gebbeken, N. & Binder, B., Traglastberechnung yon Stabtrag-
werken aus Stahl unter Berficksichtigung der AnschluBverformungen. Re-
search Report, Institut ffir Statik, Universit~it Hannover, 1991.
16. DIN 18800, Tell 1 und Teil 2, Stahlbauten. Beuth-Verlag, Nov. 1990.
17. Domscheit, A., Numerische Berechnung der Schraubenbolzenbelastung hoch-
fester verschraubter T-St6Be. Sonderi.ibung, Institut fiir Stahlbau, Universit~t
Hannover, unpublished, 1983.
18. Eurocode 3, Design of Steel Structures. ECCS, Draft, 1990.