Component-based model versus stress-resultant plasticity modelling of bolted

end-plate connection: Numerical implementation

Anas Alhasawi, Samy Guezouli, Maël Couchaux

PII: S2352-0124(17)30031-0
DOI: doi:10.1016/j.istruc.2017.05.004
Reference: ISTRUC 195

To appear in:

Received date: 6 January 2017

Revised date: 18 May 2017
Accepted date: 22 May 2017

Please cite this article as: Alhasawi Anas, Guezouli Samy, Couchaux Maël, Component-
based model versus stress-resultant plasticity modelling of bolted end-plate connection:
Numerical implementation, (2017), doi:10.1016/j.istruc.2017.05.004

This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
 ACCEPTED MANUSCRIPT

Component-based model versus stress-resultant plasticity modelling

of bolted end-plate connection: Numerical implementation

T
Anas Alhasawi1 , Samy Guezouli1 and Maël Couchaux 1

P
1
INSA de Rennes, LGCGM/Structural Engineering Research Group, 20 avenue des Buttes de Coësmes,

RI
CS 70839, 35708 Rennes Cedex 7, France

SC
Abstract

NU
This paper deals with the analysis of a steel beam-to-column bolted end-plate connection
subjected to cyclic loading. The proposed model consists of an improved component-based
approach that closely follows the joint behavior. The study reminds the component-based
MA
analysis and shows how to implement two proposed improvements by the use of an elasto-
plastic formulation and discuss their influences on the joint behavior. The first modification
concerns possible separation between the end-plate and the column flange on which it is
bolted (gap effect) and the second one concerns the group of two or more than two bolt-
ED

rows (group effect). Examples are subsequently detailed for validation and highlight the
robustness of the proposed model. These modifications allows to the well-known component
based model a new variant that gives more accurate results and remains easy-to-implement
PT

in structural analysis programs.

Keywords : Steel joint connection, cyclic load, component-based model, gap problem, group effect, Finite Element Method
CE

1. Introduction
Safe and economic design of steel and composite structures requires a deep understand-
AC

ing of the joint response. Semi-rigid connections can provide several advantages including :
economy and fabrication costs, robustness of the frames, ... Two different approaches have
been adopted to model the behavior of semi-rigid connections, one can distinguish between:

- Theoretical models: these models propose empirical or semi-empirical Moment-Rotation

curves generally fitted of experimental test data. Parameters of these models are often
related to material/geometrical characteristics of the joint. They are formulated in a
way to ease their implementation in a standard displacement-based analysis of frame.
A nonlinear finite element analysis of frames considering the actual joint behavior
provides a more accurate representation of the structure deformation and the corre-
sponding internal forces. Significant improvements have been made to this approach
since the 1980s with: Richard et al. [1] proposed to include experimental curves di-
rectly in a finite element procedure; several authors proposed multi-linear curves still
dependent on a mathematical curve fitting such as: Moncarz and Gerstle [2] in 1981,
Preprint submitted to Elsevier May 23, 2017
 ACCEPTED MANUSCRIPT

Poggi and Zandonini [3] in 1985, Nethercot [4] in 1989, ... . The FEM has been used
to investigate the joint behavior; 2D and 3D models have been developed by different
authors such as: Patel and Chen [5] in 1985 for bolted connections, Bursi and Jaspart

T
[6] in 1997, Yang et al. [7] in 2000, Maggi et al. [8] in 2002 and more recently, Con-
cepcion Diaz et al. [9] in 2011, Bo Yang and Kang Hai Tan [10] in 2012. All these

P
results were used to obtain mathematical equations for the moment-rotation curves.

RI
- Mechanical models: These models well-known as the component-based models appear as a
viable alternative between semi-empirical models and complicated 3D ones. The basic

SC
idea is to distinguish within a joint a set of individual components. Each component
is characterized by its own mechanical behavior (stiffness, strength, deformability, ...).
Main developments have been proposed by Jaspart [11] in 1996, Jaspart et al. [12]

NU
in 1999, Cerfontaine [13] in 2004, Del Savio et al. [14] in 2009, Bayo et al. [15] in
2006, Minas et al. [16] in 2009 and Chang et al. [17] in 2015. Component-based
mechanical models use tension/compression springs connected to rigid links. Each
MA
spring reproduces the behavior of a component inside the joint resisting either in
tension or in compression. The models proposed by different authors concern various
types of connections (welded, double seats, bolted, ..).
Most of the above described models have been developed to describe the joint behavior under
ED

monotonic loading. Generally, these studies focused mainly on the evaluation of the resist-
ing bending moment, the rotational stiffness and the rotational capacity (ductility) of the
joint. However, some of these theoretical models were adapted to reproduce the mechanical
PT

response of connection under cyclic loading: Moncarz and Gerstle [2] in 1981 and Mazzolani
[18] in 1988. In this paper, The mechanical response of joints under cyclic loading has been
investigated using a component-based model. Gang Shi et al. [19] undertook several ex-
CE

perimental tests on different typologies of bolted end-plate connection. This investigation

has been focused on the comparison between flush and extended end-plate configurations:
considering different end-plate thicknesses, bolt diameters, number of bolts and including
AC

or not stiffeners. Some specimens have been tested under monotonic and cyclic loading.
Under cyclic loading, one can observe that in case of a flush end-plate joint, the gap be-
tween the column flange and the end-plate is more significant than in case of an extended
end-plate connection (specimen JD1). Pu Yang et al. [20] proposed a model for bolted ex-
tended en-plate connection to describe the joint behavior under cyclic loadings. The model
includes linear springs connected to a rotational spring for beam plastic hinge on one side
and a rotational spring for the column panel in shear on the other side. The simulations
were compared to six full-scale experiments. One can observe that the experimental results
(specimen ES-1-1/2-24a) show evidence of an end-plate separation at the bottom of the
beam flange. It appears that this separation (gap) has a significant influence on the bending
moment. More recently, Da Silva et al. [21] in 2016 proposed a cyclic component-based
model. The proposed model concerns steel joints subjected to a bending moment. Under
cyclic loading, the proposed model included the possibility of load reversal for any rotation
magnitude. The objective is to reproduce the hysteretic behavior with degradation of per-
formance. Possible separation (gap) is not discussed. To conclude for the bolted end-plate
2
 ACCEPTED MANUSCRIPT

connection, it appears that the literature does not give efficient mechanical models able to

T
accurately reproduce the joint behavior under cyclic loading. The newly proposed model
concerning this paper includes two aspects that significantly influence the joint behavior:

P
- The separation (gap) between the column flange and the end-plate and its effect on the

RI
global behavior of the joint. This phenomenon should take into account the collapse
models of the T-stub as defined in Eurocode 3 [23].

SC
- The group effect of some bolt-rows and how to include it within the plasticity algorithm.
We remind that Eurocode 3 [23] takes into account the group effect through the limit
resistance of the bolt-rows only.

NU
From a computational view point, the constitutive equations for each component are dis-
cretized using an implicit scheme and the consistent stiffness matrix for the joint is derived
MA
using a standard assemblage procedure.

2. Component-based analysis
ED

The main idea of this approach is to reflect each source of deformation within the joint
by a nonlinear spring and combine them within an arrangement to best reproduce the
mechanical response of the connection. This method requires the following steps:
PT

• Identification of active component,

• Evaluation of the force deformation response of each component,

CE

• Assembly of the active components for the evaluation of the whole joint response.
Fig. 1 shows the connection and its corresponding mechanical model. The components
AC

which contribute to the deformation of the bolted beam-to-column connection are defined
in Table 1.
Notice: The component BWT does not appear in the Fig. (1) because this component
is considered only for the bolt-rows located between the beam flanges. Indeed, in Fig. (1),

Table 1: List and effect of different components

Symbols Components
CWT Column Web in Tension
CFB Column Flange in Bending
EPB End-Plate in Bending
BWT Beam Web in Tension
BT Bolt in Tension
CWC Column Web in transverse Compression
BFWC Beam Flange and Web in Compression
3
 ACCEPTED MANUSCRIPT

(c)
D
CFB CWT BFWC
BT EPB CWC

T
Outer bolt-row

P
CWT CFB EPB BT
T31
CWC BFWC
T2

RI
(b)
D

SC
CWC BFWC
T4
CWT CFB EPB BT
T32

NU
Outer bolt-row

Column Beam
a) End-plate connection b) Mechanical Model
MA
Figure 1: Components effects and corresponding springs - equivalent springs ”T ypes, r”
ED

the joint configuration is double extended end-plate without bolt-rows between the beam
flanges, so no component “BWT” in this configuration.

2.1. Joint Type characterisation

PT

Each set of individual springs in series is replaced by an equivalent spring denoted

”Type”, in order to distinguish with individual components denoted ”component” (see Fig.
CE

1). Each equivalent spring is identified with a label depending on its location. The corre-
sponding activation mode (tension or compression) is highlighted. Each Type can be defined
as follows:
AC

- T 2 and T 4 are the compression equivalent springs located at the beam flanges. They both
contain the same components. T 2 is located at the top beam flange level and T 4 at
the bottom one.

- T 3i concerns bolt-rows i = 1, ..., m that work only in tension. During the cyclic loading
only the bolt-rows being in tension are activated, the others are temporarily disabled.

The force-displacement relationship of each Type within the joint may be characterized by a
bi-linear, tri-linear or non-linear curve. In this study, we consider a bi-linear elastic perfectly
plastic model obtained by assembling all relevant components. Each component (individ-
ual spring) behave elastically except the end-plate and the column flange in bending which
have an elastic-perfectly plastic behaviour. For each component ”α” the initial stiffness k α
α
and the plastic resistance FRd are calculated according to Eurocode 3 [23]. Considering
the index ”T ” related to ”Type” and ”r” to its number, the plastic strength for each Type
FT r,Rd is the minimum between individual plastic resistances of the concerned components:

4
 ACCEPTED MANUSCRIPT

α
FT r, = M in (FRd ), where nc is the number of components in the corresponding Type.

T
Rd
α=1,...,nc
The stiffness of the Type kT r can be calculated as follows:

P
RI
• The stiffness of T 2 (or T 4) is obtained by condensing out the internal degree of freedom
(Fig. 2).

SC
X &:& X7  X7 

N &:& N %):&
NU N7 
MA
Figure 2: Equivalent component - T2 (or T4)

1 1 k BF W C k CW C
ED

kT 2/T 4 = P
nc = 1 = (1)
1 kCW C
+ kBF1W C k BF W C + k CW C
kα
α
PT

• The stiffness of T 3i is obtained by condensing out internal degrees of freedom (Fig. 3).
The index ”i” is added for each Type 3 including bolt-rows and corresponding number
in the joint (i = 1 for the top bolt-row, i = m for the last bottom bolt-row).
CE

1 1
kT 3i = P
nc = 1 1 1 1 (2)
1 kCW T
+ kCF B + kEP B
+ kBT
kα
AC

X &:7 X &)% X (3% X7  X 7

N &:7 N &)% N (3% N %7 N 7

Figure 3: Equivalent component - T 3i

3. Mechanical model formulation

The proposed mechanical model consists of two rigid links, the left one represents the
column web centerline and the right one represents the end-beam line. These lines are
5
 ACCEPTED MANUSCRIPT

uT 2

T
M N
R.A. u

P
Column Web
End beam
uT 33
centerline

RI
uT 34
T31
T2

dT 31
b – Positive bending

dT 2
T32
dT 33 dT 32

SC
R.A.
dT 4
dT 34
T33
T4
T34 q

NU
uT 31
a – Initial configuration
uT 32
R.A. u N
y
M
MA
uT 4
q
x
c – Negative bending
ED

Figure 4: Proposed mechanical model

PT

connected by series of springs that are replaced by spring Types: Tr . The details of the
proposed mechanical model are given in Fig. 4 and an example of a deformed state in both
cases negative and positive bending (Fig. 4b and Fig. 4c). The index i of the bolt-row
CE

begins from the top. In case of negative bending, only T 31 , T 32 and T 4 are activated and
the others are temporarily disabled. At each ”r” level, the lengthening (shortening) ūT r are
highlighted and at the reference axis (R.A.) the global variables of the joint are represented
AC

by the couple (ū, θ): axial displacement and relative rotation. These variables are related
to the applied loads (N, M ): axial force and bending moment, respectively. In the case
of positive bending, Fig. (4b), T 33 , T 34 and T 2 will be activated and the others will be
disabled. The global force vector is defined as follows:
 T
F= N M (3)

and corresponding global displacement vector is:

 T
U = ū θ̄ (4)

The displacement of each T r can be geometrically calculated assuming small local rotations,
as follows:

ūT 3i = ū − dT 3i θ̄ ⇔ ūT 3i = 1 −dT 3i U (5)

ūT 2 = ū − dT 2 θ̄ ⇔ ūT 2 = 1 −dT 2 U (6)
6
 ACCEPTED MANUSCRIPT

In the above equations, dT r are the vertical coordinates of the corresponding Type r, positive
value for the Type above the reference axis and negative value for those under the reference
axis.

T
The stiffness matrix of the mechanical model is obtained by the use of the principle of virtual
work. The equilibrium equation between the variation of the internal virtual work, δW int ,

P
and external one , δW ext is defined by:

RI
δW int = δW ext (7)

SC
where: n
X X
int
δW = δ ūTT r kT r ūT r + δ ūTT 3i kT 3i ūT 3i (8)

NU
r i

The first term of Eq (8) corresponds to the virtual work of activated top or bottom beam
flange rows (r = 2 or 4), whilst the second term represents the virtual work of all activated
bolt-rows.
MA
External virtual work is given by:

δW ext = δUT F (9)

ED

replacing (5) into (8) and substituting it with (9) into (7), it leads to:
    
C11 C12 ū N
= (10)
C12 C22 θ̄ M
PT

where:
CE

X m
X
C11 = kT r + k T 3i
r i
m
AC

X X
C12 = − kT r d T r − k T 3 i d T 3i
r i
X m
X
C22 = kT r d2T r + kT 3i d2T 3i
r i

The model uses a displacement control procedure to solve the non-linear system of equa-
tions, Eqs. (10), (mechanical non-linearity).

4. Cyclic behaviour of bolted end-plate connection

A general model of a beam-to-column joint should include a component related to the
column shear deformation, Fig. 5. The total rotation of the end-beam cross-section is
therefore the sum of the column centerline rotation γ and the end-beam cross-section rotation
relatively to column centerline θ. During cycling loading, each activated bolt-row is loaded in
7
 ACCEPTED MANUSCRIPT

tension and it elongates according to its own stiffness. In the elastic domain, the gap vanishes

T
once the equivalent spring is unloaded. By increasing the load, if the end-plate or the column
flange component is in plastic range, a permanent gap appears. An accurate component-

P
based model should be able to follow the evolution of this gap and the deformation during

RI
the loading history.
The gap rotation is provided by relative deformation of the end-pate in bending, the column

SC
R.A. N
M NU g
q
f M
N
MA
Connection
Web panel in shear
ED

Figure 5: General mechanical model for the joint

flange in bending and the bolt in tension. This individual components that related to the
PT

gap rotation are characterized in the component method based on the T-stub model. An
example of 3D finite element simulation
Printed using Abaqus/CAE on: Wed Sep 07show clearly
17:06:56 Paris, Madrid (heurethat
d’été) 2016a gap might occur between the
column flange and the end-plate due to plastic deformations (Fig. 6).
CE
AC

Figure 6: The gap - 3D finite element model

4.1. Joint mechanical response during cyclic loading

In the aim to simplify the analysis of the joint behaviour under cyclic loading, the case
of two bolt-rows is considered. This analysis could be easily extended to the case of more
then two bolt-rows.
It is worth to mention that the specimen is loaded firstly by a negative rotation Fig. 7
followed by a positive rotation Fig. 8. The following stages can be distinguish:
8
 ACCEPTED MANUSCRIPT

- Stage 1 - loading in negative bending: Top bolt-row T 31 is in tension. Assuming that the

T
gap at this top level gt occurs due to plastic deformations of one or both components
(EPB and/or CFB see Table 1) of T 31 .

P
- Stage 2 - Unloading from negative bending: Previous plastic deformation remain perma-

RI
nent in corresponding component(s). In order to close the gap gt , a slip must occur
in the behaviour curve (FT 31 − ūT 31 ), Fig. 7(c). This slip allows the bottom bolt-row

SC
T 32 to be activated.

- Stage 3 - Loading in positive bending: Bottom bolt-row T 32 is in tension and the top one

NU
T 31 is disabled. Assuming that the gap at this bottom level gb occurs due to plastic
deformations of one or both components (EPB and/or CFB see Table 1).

- Stage 4 - Unloading from positive bending (and finish one cycle): In order to close the
MA
gap gb , a slip must occur in the behaviour curve (FT 32 − ūT 32 ), Fig. 8(f).

Starting the next cycle loading both gaps ( gt and gb ) must be closed before to allow the
activation of concerned Type. This remark is represented as slips in the behaviour model.
ED

Following Loading-Unloading cycles use the same procedure taking care to close previously
the gaps.
PT

Loading negative Unloading –

Initial State Unloading Close top Gap
bending
T31
T2
CE

q
T4 Top Gap( g t )
(a) T32 (b) (c) (d)
T31
AC

T2
Fr Fr Fr Fr T32
T4

ur
JW

(Force – displacement) curves for each Type

Figure 7: First Loading-Unloading stage (half cycle)

9
 ACCEPTED MANUSCRIPT

Loading positive Reloading negative bending after

Unloading
bending closing top and bottom gaps

T
q

P
q

RI
(e) (f) (g)
T31

SC
T2
Bottom Gap ( g b )
T32
Fr Fr Fr T4

NU
JW ur
JE
MA
Point of loading after
closing the gap

(Force – displacement) curves for each Type

ED

Figure 8: Second Loading-Unloading stage (complete cycle) - Beginning of second cycle

PT

5. Group of bolt-rows effect

CE

Cerfontaine [13] proposed to consider the group of bolt-rows effect by the use of a multi-
linear failure criterion based on the lower bound theorem of limit analysis applied to the
joint. In this section, it is proposed to show how to take into account this effect through
AC

plasticity modeling.
Assuming normality and associate rules, the plastic flow direction is given by the gradient
to the yield surface. Plastic elongation and normal force are evaluated for each bolt-row
within a group and the model provides automatically plastic redistribution of this forces
within the group of bolt-rows during the loading. It is worth to mention that, the theory
that developed for the case of 2 bolt-rows per group is presented in section 5.1 and that one
for more than two bolt-rows per group is demonstrated in section 5.2.

5.1. Group of two bolt-rows

The proposed formulation is detailed for the case of the group of two bolt-rows (T 31 and
T 32 )(Fig. 9) and will be generalised for groups with more than two bolt-rows.

5.1.1. Yield surface definition

The yield criteria depend on three lines (see Fig. 10). Horizontal and vertical yield
lines (Φ1 and Φ2 ) correspond individual strength of each bolt-rows and third line (Φ3 )

10
 ACCEPTED MANUSCRIPT

T
D(0)
q

P
RI
SC
End-column Beam

Figure 9: Extended end-plate connection

NU
corresponds to the group strength. Eurocode 3 part 1.8 provides the method to calculate
specific strengths. We denote FT 31 ,Rd and FT 32 ,Rd : individual strength of the bolt-rows T 31
MA
and T 32 respectively, and F12,Rd is the strength of the group of these bolt-rows.
The group effect of two bolt-rows creates five zones that can be distinguished as shown in
ED

FT 32
D2
DB
2
PT

5
DA
3
CE

FT 32 ,Rd B 4
F2
F3 D1
Elastic zone A
AC

F1
1

FT 31,Rd FT 31

Figure 10: Yield surface for 2 bolt-rows per group

Fig. 10:
• The lines (Φ1 , Φ2 and Φ3 ) that allow to respect the yield criteria are defined by the
following equations:

Φ1 (FT 31 , FT 31 , Rd ) = FT 31 − FT 31 ,Rd
Φ2 (FT 32 , FT 32 , Rd ) = FT 32 − FT 32 ,Rd . (11)
Φ3 (FT 31 , FT 32 , F12, Rd ) = FT 31 + FT 32 − F12,Rd

• and the lines that allow to respect normal projections:

11
 ACCEPTED MANUSCRIPT

- The lines DA and DB separate respectively the zones (3 and 4) and the zones (3 and
5).
- The lines D1 and D2 separate respectively the zones (1 and 4) and the zones (2 and

T
5). As-soon-as the trial force is known, appropriate projection is automatically

P
detected. For this aim, firstly the coordinates of the points A and B are obtained
using:

RI
F1A = FT 31 ,Rd F1B = F12,Rd − FT 32 ,Rd

SC
F2A = F12,Rd − FT 31 ,Rd F2B = FT 32 ,Rd

Therefore, the lines (DA , DB , D1 and D2 ) are defined as follows:

NU
DA (FT 31 , FT 32 ) = FT 31 − FT 32 − (2FT 31 , Rd − F12, Rd )
DB (FT 31 , FT 32 ) = FT 31 − FT 32 + (2FT 32 , Rd − F12, Rd )
(12)
(13)
MA
D1 (FT 32 ) = FT 32 − (F12, Rd − FT 31 ,Rd ) (14)
D2 (FT 31 ) = FT 31 − (F12, Rd − FT 32 , Rd ) (15)

Finally, the different zones are defined as follows:

ED

• Elastic zone: The following inequations respect the behaviour of the bolt-rows remain
to in the elastic range.
PT


 Φ1 (FT 31 , FT 31 , Rd ) ≤ 0 
∆ūT 31 , p = 0
Φ2 (FT 32 , FT 32 , Rd ) ≤ 0 ⇒ (16)
∆ūT 32 , p = 0
CE

Φ3 (FT 31 , FT 32 , F12, Rd ) ≤ 0


Herein ∆ūT 31 , p and ∆ūT 32 , p represent respectively the plastic deformation for the first
and the second bolt-rows.
AC

• Zone 1: It is limited by the lines Φ1 and D1 . The corresponding criterion must respect
the following inequations:
(
∆ūT 31 , p = ∆λ1 ∂F∂ΦT 13

Φ1 (FT 31 , FT 31 , Rd ) ≥ 0
⇒ 1 (17)
D1 (FT 32 ) ≤ 0 ∆ūT 32 , p = 0

• Zone 2: It is limited by the lines Φ2 and D2 . The corresponding criterion must respect
the following inequations:
(
∆ūT 31 , p = 0

Φ2 (FT 32 , FT 32 , Rd ) ≥ 0
⇒ ∆ūT 32 , p = ∆λ2 ∂F∂ΦT 23 (18)
D2 (FT 31 ) ≤ 0 2

12
 ACCEPTED MANUSCRIPT

• Zone 3: It is limited by the lines Φ3 , DA and DB . The corresponding criterion must

respect the following inequations:

 Φ3 (FT 31 , FT 32 , F12,Rd ) ≥ 0

T
∆ūT 31 , p = ∆λ3 ∂F∂ΦT 33
(
DA (FT 31 , FT 32 , FT 31 , Rd , F12, Rd ) ≤ 0 ⇒ 1 (19)
∆ūT 32 , p = ∆λ3 ∂F∂ΦT 33

P
DB (FT 31 , FT 32 , FT 32 , Rd , F12, Rd ) ≥ 0

2

RI
• Zone 4: The projection reaches directly the point A, this zone is limited by the lines Φ1 ,
Φ3 , DA and D1 . The corresponding criterion must respect the following inequations:

SC

 Φ1 (FT 31 , FT 31 , Rd ) > 0

∆ūT 31 , p = ∆λ1 ∂F∂ΦT 13 + ∆λ3 ∂F∂ΦT 33
(
Φ3 (FT 31 , FT 32 , F12, Rd ) > 0

⇒ ∂Φ3
1 1
D (F , F , F , F ) > 0 ∆ū = ∆λ

NU

 A T 3 1 T 3 2 T 3 1 , Rd 12, Rd T 3 2 , p 3 ∂FT 32
D1 (FT 32 ) > 0

(20)
MA
• Zone 5: The projection reaches directly the point B, this zone is limited by the lines Φ2 ,
Φ3 , DB and D2 . The corresponding criterion must respect the following inequations:

 Φ2 (FT 32 , FT 32 , Rd ) > 0
∆ūT 31 , p = ∆λ3 ∂F∂ΦT 33
 (
Φ3 (FT 31 , FT 32 , F12, Rd ) > 0

ED

⇒ 1
 DB (FT 31 , FT 32 , FT 32 , Rd , F12, Rd ) < 0 ∆ūT 32 , p = ∆λ2 ∂F∂ΦT 23 + ∆λ3 ∂F∂ΦT 33
2 2

D2 (FT 31 ) > 0

(21)
PT

5.1.2. Incremental algorithm of projection

In accordance with the previous definition of different zones, incremental plasticity algo-
CE

rithm [25] can be established for zone 1 to zone 5:

• Zone 1: In this case, T 31 is in the plastic range:
AC

Φn+1
1 = FTn+1
31 − FT 31 ,Rd > 0 (22)
whilst T 32 as-well-as the group of (T 31 and T 32 ) are in the elastic range:
Φn+1
2 = FTn+1 n+1
32 − FT 32 ,Rd ≤ 0 ⇒ kT 32 = ke2
(23)
Φn+1
3 = FTn+1 n+1
31 + FT 32 − F12, Rd ≤ 0

Herein ke2 is the elastic stiffness of T 32 , Eq. (2).

The current force in the T 31 at (n+1) increment is given as:
FTn+1 n+1 n+1


31 = ke1 ūT 31 − ūT 31 , p
(24)
= FTn+1
31 , trial − ke1 ∆λ1

where ke1 is the elastic stiffness of T 31 , Eq. (2).

The yield function becomes:
Φn+1 = FTn+1


1 31 , trial − ke1 ∆λ1 − FT 31 , Rd = 0 (25)
13
 ACCEPTED MANUSCRIPT

- The increment of plastic multiplier:

Φn+1
1, trial
∆λ1 = (26)

T
ke1

P
- The tangent stiffness

RI
Derivation of (24) gives:

∂FTn+1 ∂FTn+1

SC
31 31 , trial ∂∆λ1
kTn+1
31 = n+1 = n+1 − n+1 ke1 (27)
∂ ūT 31 ∂ ūT 31 ∂ ūT 31
Substituting Eq. (26) into (27) leads to:

NU
kTn+1
31 = 0 (28)
MA
• Zone 2: In this case, T 31 and the group of (T 31 and T 32 ) are in the elastic domain:
Φn+1
1 = FTn+1 n+1
31 − FT 31 ,Rd ≤ 0 ⇒ kT 31 = ke1
n+1 n+1 n+1 (29)
Φ3 = FT 31 + FT 32 − F12,Rd ≤ 0
ED

whilst T 32 is in the plastic range:

Φn+1
2 = FTn+1
32 − FT 32 ,Rd > 0 (30)
PT

The same procedure as for the first bolt-row is followed for the second bolt-row, there-
fore we can define
CE

- The increment of plastic multiplier:

Φn+1
2, trial
∆λ2 = (31)
ke2
AC

- The tangent stiffness

kTn+1
32 = 0 (32)
• Zone 3: The group criterion is activated:
Φn+1
3 = FTn+1 n+1
31 + FT 32 − F12, Rd > 0 (33)
Trial force of the group (T 31 and T 32 )
n+1 n+1 n+1
F12, trial = FT 31 , trial + FT 32 , trial (34)
and corrected force is:
n+1
F12 = FTn+1 n+1
31 + F T 3 2
= ke1 ūn+1 n+1 n+1 n+1



T 31 − ū T 31 ,p + k e2 ū T 32 − ū T 32 ,p (35)
n+1 n+1
⇒ F12 = F12, trial − ∆λ3 (ke1 + ke2 )
14
 ACCEPTED MANUSCRIPT

n+1 n+1 n+1

where: F12, trial = FT 31 , trial + FT 32 , trial
The yield function can be written as follows:

T
n+1
Φn+1 = F12, T rial − ∆λ3 (ke1 + ke2 ) − F12, Rd = 0

P
3

RI
- The increment of plastic multiplier:

Φn+1
3, T rial

SC
∆λ3 = (36)
ke1 + ke2

where: Φn+1 n+1 n+1

3, trial = FT 31 , trial + FT 32 , trial − F12, Rd

- The tangent stiffness:

NU
i. Tangent stiffness for T 31 is given by:
MA
∂FTn+131 ∂FTn+1
31 , trial ∂ (∆λ3 ) ke1 ke2
kTn+1
31 = n+1 = n+1 − ke1 n+1 = (37)
∂ ūT 31 ∂ ūT 31 ∂ ūT 31 ke1 + ke2
ED

ii. Tangent stiffness for T 32 is given by:

∂FTn+132 ∂FTn+1
32 , trial ∂ (∆λ3 ) ke1 ke2
kTn+1
32 = n+1 = n+1 − ke2 n+1 = (38)
∂ ūT 32 ∂ ūT 32 ∂ ūT 32 ke1 + ke2
PT

• Zone 4: In this case we have:

CE

Φn+1
1 = FTn+1
31 − FT 31 , Rd > 0
(39)
Φn+1
3 = FTn+1 n+1
31 + FT 32 − F12, Rd > 0
AC

The yield function is written as:

Φn+1
1 = FTn+1
31 , trial − ke1 ∆λ1 − ke1 ∆λ3 − FT 31 , Rd = 0
⇒ ke1 ∆λ1 + ke1 ∆λ3 = Φn+1
1, trial (40)

On the other side, the yield surface function Φ3 can be defined as follows:

Φn+1
3 = FTn+1 n+1
31 , trial − ke1 ∆λ1 − ke1 ∆λ3 + FT 32 , trial − ke2 ∆λ3 − F12, Rd
⇒ ke1 ∆λ1 + ∆λ3 (ke1 + ke2 ) = Φn+1
3, trial (41)

Combining the last equations of (40) and (41), we obtain:

    n+1 
ke1 ke1 ∆λ1 Φ1, trial
= (42)
ke1 ke1 + ke2 ∆λ3 Φn+1
3, trial

Solving (42) leads to:

15
 ACCEPTED MANUSCRIPT

- Plastic multipliers:
i.

T
Φn+1 n+1
1, trial (ke1 + ke2 ) − Φ3, trial ke1
∆λ1 = (43)

P
ke1 ke2

RI
ii.
Φn+1 n+1
3, trial − Φ1, trial

SC
∆λ3 = (44)
ke2

- Tangent stiffness:

NU
i. Tangent stiffness for T 31 is defined by:

∂FTn+1 ∂FTn+1 ∂ (∆λ3 ) ∂ (∆λ1 )

MA
31 31 , trial
kTn+1
31 = n+1 = n+1 − ke1 n+1 − ke1 (45)
∂ ūT 31 ∂ ūT 31 ∂ ūT 31 ∂ ūn+1
T 31

where
∂ (∆λ1 )
ke1 = ke1
ED

∂ ūn+1
T 31
(46)
∂ (∆λ3 )
ke1 =0
∂ ūn+1
PT

T 31

Replace (46) into (45) gives

kTn+1
31 = 0 (47)
CE

ii. Tangent stiffness for T 32 is defined by:

∂F12n+1 ∂FTn+1
32 , trial ∂ (∆λ3 )
AC

kTn+1
32 = n+1 = n+1 − ke2 (48)
∂ ūT 32 ∂ ūT 32 ∂ ūn+1
T 32

where
∂ (∆λ3 )
ke2 = ke2 (49)
∂ ūn+1
T 32

Combining (48) and (49) gives

kTn+1
32 = 0 (50)

• Zone 5:
In this case we have two activated yield surfaces: one concerns T 32 and the other
corresponds to the group of (T 31 and T 32 ).

Φn+1
2 = FTn+1
32 − FT 32 , Rd > 0
Φn+1
3 = FTn+1 n+1
31 + FT 32 − F12, Rd > 0

16
 ACCEPTED MANUSCRIPT

i. The increment of plastic multiplier can be defined by:

    n+1 
ke2 ke2 ∆λ2 Φ2, trial

T
= (51)
ke2 ke1 + ke2 ∆λ3 Φn+1
3, trial

P
RI
ii. The tangent stiffnesses of the (T 31 and T 32 ) are:

kTn+1
31 = 0 (52)

SC
kTn+1
32 =0 (53)

5.2. General formulation for the group effect criterion

NU
• The general formulation for the bolt-rows interaction criterion can be written as fol-
lows:  
q−1
! "
MA
X
FT 3q + FT 3s ≤ M in Fqs,Rd , Fks,Rd − F T 3i  (54)
i=k k=1,..,(q−1)

where q = 2,...m − 1; and s= q + 1,....., m, in which m is the total number of the

bolt-rows in the joint. Fqs and Fks are the group resistances including the bolt-rows
ED

from (q to s) or (k to s), respectively.

• Example of a group of three bolt-rows: In the aim to take into account the effect of
PT

three bolt-rows (Fig. 11), Successive group effects of two bolt-rows are proposed. The
steps of the group effect occurrence in the case of positive bending can be described.
The method remains available in case of negative bending.
CE
AC

T 31

T 32

T 33
R.A
T 34

T 35

Figure 11: End-plate connection with several bolt-rows

It can be distinguished:

- Group effect 1 (T 31 , T 32 ).
17
 ACCEPTED MANUSCRIPT

T
- Group effect 2 (T 32 , T 33 ).

P
According to EC3, individual and group resistances can be easily obtained and denoted

RI
as follows:

- Individual: (FT 31 ,Rd , FT 32 ,Rd ).

SC
- Group 1: (F12,Rd ).
- Group 2: (F23,Rd ).

NU
In case of negative bending:

- Group effect 1 is controlled by following criterion:

MA
FT 31 + FT 32 ≤ F12,Rd (55)

- Group effect 2 is controlled by the combination of both following criteria:

ED

FT 32 + FT 33 ≤ F23,Rd and FT 31 + FT 32 + FT 33 ≤ F13,Rd (56)

The first equation of Eqs. (56) corresponds to the group effect of the two bolt-rows T 32
PT

and T 33 and the second equation corresponds the group effect of the three bolt-rows
T 31 , T 32 and T 33 . The combination of these equations leads to the following criterion:
CE

FT 32 + FT 33 ≤ M in [F23,Rd , (F13,Rd − FT 31 )] . (57)

AC

6. Numerical applications
Two numerical simulations were performed to highlight both improvements proposed to
the joint response obtained by the component-based model. The first one concerns the gap
effect and the second one concerns the group effect.

6.1. Example for the gap effect

The simple example of two bolt-rows (Fig. 12) is used to investigate the influence of the
gap effect on the joint behavior. Two cycles of rotations are applied to this joint according
to the diagram given in Fig. 13. Corresponding mechanical model is given in Fig. 14. The
steel grade that adopted is S355 and the Young’s Modulus is 210 GPa. This example has
been solved using proposed mechanical model analysis and 3D finite element modelling.

18
 ACCEPTED MANUSCRIPT

160

HEB 240
M20 cl10.9 30 32 96 32 30

P T 30 32
IPE 240
aw = 5

RI
R.A.

240
364
q
af = 8 S355

SC

32 30
15

S355

NU
Figure 12: Example for the gap effect
MA

q ( mrad)
ED

23.05
8.65
PT

0
-8.65 Time
CE

-23.05
AC

Figure 13: Cyclic loading history

T31
T2

R.A.

T4
T32

Figure 14: Mechanical model

19
 ACCEPTED MANUSCRIPT

Table 2: Initial stiffness and resistance force (kN/mm and kN)

T 31 and T 32 T2 and T4
Component
FRd kini FRd kini

T
CFB 352 8498 - -

P
CWT 508 1475 - -
EPB 258 4221 - -

RI
CWC - - 642 2150
BFWC - - 565 ∞

SC
BT 441 1630 - -

NU
6.1.1. Proposed mechanical model analysis
In this section the proposed mechanical model will be used to capture the cyclic response
of the connection. It is worth to point out that the strength of equivalent Type spring is
MA
obtained as the minimum of the strengths of included components (highlighted values in
Table 2).
For each component within the joint, material hardening has been considered very light
(keq,r /1000) but not equal to zero. In accordance with the cyclic loading history (Fig. 13),
ED

the (Moment - Rotation) curve obtained as the response of the joint behavior is given in
Fig. 15. It appears clearly that the gaps obtained at both top and bottom levels of the
joint generate slipping at both sides of the zero-rotation axis of the curve (horizontal dashed
PT

arrows). Alternatively between top and bottom bolt-rows, the gap obtained by the tension
of the top bolt-row for example, must be first recovered before beginning the tension at the
bottom bolt-row. The proposed algorithm follows these sequences very rigorously in order
CE

to insure an accurate solution of the proposed model (the index t corresponds to the T 31 in
tension and the index b to the T 32 in tension).
AC

• First loading cycle (−8.65 mrad ≤ θ ≤ 8.65 mrad):

(O − At − Bt ) Loading in negative bending (1 - 2)
(Bt − Ct ) Unloading (3)
(Ct − O) Slipping (4)
(O − Ab − Bb ) Loading in positive bending
(Bb − Cb ) Unloading
(Cb − O) Slipping

• Second loading cycle (−23.05 mrad ≤ θ ≤ 23.05 mrad):

(Ct − Bt − Dt ) Reloading in negative bending (5 - 6)
(Dt − Et ) Unloading (7)
(Et − Ct − O − Cb ) Slipping
(Cb − Bb − Db ) Reloading in positive bending
(Db − Eb ) Unloading

20
 ACCEPTED MANUSCRIPT

P T
RI
SC
NU
MA
ED

Figure 15: Moment vs. Rotation curve

After each load cycle, the plastic deformation of the components within each equivalent
PT

spring are cumulated. At the end of the unloading stage, permanent deformation defining
the so-called gap needs to be closed to allow the activation of the others rows. In accordance
with Fig. 15, following values of couples (Moment - Rotation) of the joints are obtained
CE

when the top bolt-row is active (same results when the bottom bolt-row is active because
of symmetry):
• First loading cycle (−10 mrad ≤ θ ≤ 10 mrad):
AC

(O − At − Bt ) (0 - 0), (77.47 - 4.64), (77.88 - 10)

(Bt − Ct ) (77.88 - 10), (0 - 5.36)
(Ct − O) (0 - 5.36), (0 - 0)
• Second loading cycle (−20 mrad ≤ θ ≤ 20 mrad):
(O − Ct ) (0 - 0), (0 - 5.36)
(Ct − Bt − Dt ) (0 - 5.36), (77.88 - 10), (78.55 - 20)
(Dt − Et ) (78.55 - 20), (0 - 15.34)
(Et − O) (0 - 15.34) - (0 - 0)
It worth to precise that the couple of springs (T 31 – T 4) are active in same time and (T 32
– T 2) also, and both couples are active alternatively (Fig. 16); therefore:
- Negative displacement obtained in the spring T 4 (bottom beam-flange row) is due to a
compression force while the tension is active in the spring T 31 (top bolt-row).
21
 ACCEPTED MANUSCRIPT

- Negative displacement obtained in the spring T 32 (bottom bolt-row) is due to a slipping

(the spring does not support any compression force) while the positive displacement

T
obtained in the spring T2 (top beam-flange row) is also a slipping (this row does not
support any tension).

P
It is clear that the behavior of each equivalent spring ”Type” (Fig. 16) follows both sleeps

RI
occurred during these two loading cycles. The slipping values obtained from these curves
are:

SC
- (Ct - O) = 1.208 mm

- (Et - O) = 5.056 mm

NU
MA
ED
PT
CE

(a) Top bolt-row T 31 (b) Bottom beam-flange row T 4

AC

(c) Bottom bolt-row T 32 (d) Top beam-flange row T 2

Figure 16: Force-displacement curves for each row of the joint

22
 ACCEPTED MANUSCRIPT

6.1.2. 3D Finite element modelling

T
In order to validate our proposed mechanical model, a 3D model of the previous end-
plate connection, Fig. 12, is investigated with the finite element codes ABAQUS. The

P
three-dimensional solid element type C3D8R have been used. The contact element between

RI
the end-plate and the column flange surfaces as well as between the end-plate, the column
flange and the bolts surface are used with friction coefficient of 0.3. All the nodes of the
top and bottom section of the column are restrained in all directions. In order to achieve

SC
reliable results,a fine mesh was employed in the contact zone as shown in the general view
of the mesh pattern for the connection, Fig. 17.

NU
MA
ED
PT
CE
AC



Figure 17: Mesh pattern of the finite element model

In the other hand, the steel material, type S355, for the components involved the creation
of the gap; Column-flange, end-plate and bolts, is considered as ductile and behave as elastic-
perfectly plastic. Whilst, the other components have an elastic behavior.
In this analysis displacement control was applied, rotation, to a rigid surface at the end
of the beam. The applied rotation was cyclic and controlled by amplitude function.

• Results and discussions The moment-rotation curve results from FE model is com-
pared to the mechanical model response developed in section 3 and implemented in
MATLAB, the two responses are demonstrated in Fig. 18. It can be observed that the
connection response from the developed mechanical model has a good agreement with
that from 3D FE mainly in the initial stiffness and moment resistance. It is worth to
23
 ACCEPTED MANUSCRIPT

P T
RI
SC
100
80
60
NU
MA
Bending moment (kN.m)

40
20
ED

0
-20
PT

-40
-60
CE

-80 Proposed model

Numerical model ABAQUS
-100
AC

-30 -20 -10 0 10 20 30

Rotation (mrad)
Figure 18: Moment vs. rotation curves

24
 ACCEPTED MANUSCRIPT

mention that in this example there is no group effect because the two bolt-rows do not

T
work together under bending moment.

P
The creation of the gap is very clear in the 3D finite element simulation and this

RI
separation, as expected, comes from plastic deformation of the end-plate in bending
(see Fig. 19).

SC
The gap effect is demonstrated in the moment-rotation curve as the following:

i In the proposed model a slipping phenomenon is adopted on the force-displacement

curve of the equivalent spring therefore, the benching-effect is significant in the

NU
cyclic response.
ii For 3D FE model the benching –effect is not significant as in the response of the
proposed model, in contrast a degradation of the stiffness is observed during
MA
closing the gap.
ED
PT
CE

AC

Figure 19: Plastic deformation in the connection at the end of the 3D FE simulation

6.2. Example for the group effect

The following application concerns an example of a semi-rigid connection that has been
analyzed by Cerfontaine [13] (Fig. 20). This example consists of an IPE600 steel beam
connected to a HEB400 steel column towards an end-plate with 5 bolt-rows denoted T 3i
(i = 1, .., 5). Table 3 summarizes the resistances and equivalent stiffnesses of each individual
and group of bolt-rows. In addition, the same data are given for the Types T 2 and T 4
located at the top and the bottom beam flange levels, respectively. According to the Eq.
(54) (where m = 5) the yield surface criterion for different combination of group of bolt-rows
are obtained and reported in Table 4. Two calculations have been performed:

- Calculation 1: The data given in Table 3 are the same as those of Cerfontaine [13]. In
this conditions, it appears that no group effect is activated (Fig. 21 ).

25
 ACCEPTED MANUSCRIPT

T
M24 HR 8.8 140

P
T31 120 60
T2 140

RI
af= 16
T32 IPE 600 120
T33 S355
a= 20o

SC
240
aw= 3
T34
120
T35
T4
300X800X20

NU
HEB 400
S355
MA
Figure 20: Bolted end-plate configuration
ED

Table 3: Types and Groupes characterisation

PT

Individual rows (Types)

Row (Type) FTr, Rd(kN) kr,eq(kN\mm)
Top beam flange (T2) 1011 2394
CE

Compression zone
Bottom beam flange (T4) 1011 2394
First bolt-row (T31) 329 574.75
Second bolt-row (T32) 407 414.85
Tension zone Third bolt-row (T33) 388 420.75
AC

Fourth bolt-row (T34) 388 420.75

Fifth bolt-row (T35) 407 512.63
Group of bolt-rows
Group of tow bolt-rows Resistance (kN)
T31+ T32 813
T32+ T33 705
T33+ T34 728
T34+ T35 706
Group of three bolt-rows Resistance (kN)
T31+ T32+T33 1200
T32+ T33+T34 1044
T33+ T34+T35 1046
Group of four bolt-rows Resistance (kN)
T31+ T32+T33+T34 1627
T32+ T33+T34+T35 1363
Group of five bolt-rows Resistance (kN)
T31+ T32+T33+T34+T35 1772

26
 ACCEPTED MANUSCRIPT

P T
- Calculation 2: In order to activate the group effects, it is proposed to increase (arbitrary)
the resistance of Type T 4 from 1011 kN to 1500 kN (Fig. 22).

RI
Table 4: Yield surface criterion for this example(5 bolt-rows)

SC
q s F T 3 q + F T 3s ≤
2 3 M in [F23,Rd , (F13,Rd − FT 31 )]

NU
3 4 M in [F34,Rd , (F14,Rd − FT 31 − FT 32 ), (F24,Rd − FT 32 )]
4 5 M in [F45,Rd , (F15,Rd − FT 31 − FT 32 − FT 33 ), (F25,Rd − FT 32 − FT 33 ), (F35,Rd − FT 33 )]
MA
Both figures represent the (Moment-Rotation) curve of the joint corresponding to each cal-
culation. It is worth to mention that in order to reach a significant rotation of the joint, for
both calculations an elastic- perfectly plastic behavior is adopted for all the components of
the mechanical model. The algorithm is able to give the force at each bolt-row during the
ED

loading history. It can be easily observed that the equilibrium is always satisfied. These
forces are highlighted at each point where the slop is changing.

For the first calculation (Fig. 21), it can observed that:

PT

- At point ”1” (Fig. 21) and at the point ”A” (Fig.22), same distribution of forces is
obtained. This step corresponds to the case where T 31 reaches its individual resistance.
CE

All other bolt-rows are still in elastic range.

- From point ”2” (Fig. 21), the force at the Type T 4 reaches its individual resistance (1011
AC

kN ). Therefore, with respect to the equilibrium between tension and compression

zones, the tension force is too low to activate a group effect.

- Point ”3” (Fig. 21) corresponds to the maximum of the bending moment in the connection
(Mj,max = 567.5 kN.m). This obtained value is in accordance with the maximum value
of the bending moment (Mj,max = 567 kN.m) that obtained by Cerfontaine [13]. At
this point an unloading of the force at T 34 to zero leads to a redistribution of this
force to the upper bolt-rows which have not reaching their individual resistance yet
(T 32 and T 33 ). It appears that with this redistribution, T 32 reaches its individual
resistance (407 kN ).

For the second calculation (Fig. 22), it can be observed that:

- At point B (Fig. 22), the force at Type T 4 (1111,1 kN ) is less than its resistance (1500
kN ). However, a group resistance is activated by T 32 and T 33 . In accordance with
Table 5 where q = 2 and s = 3, it follows:
27
 ACCEPTED MANUSCRIPT

P T
RI
SC
NU
MA
ED

Figure 21: Moment vs. rotation curve

PT
CE
AC

Figure 22: Moment vs. rotation curve

28
 ACCEPTED MANUSCRIPT

400

T
350

P
302

300

298

RI
300
296

294
250
292

SC
[KN]

404 405 406 407

200
3
FT3

150

NU
100

50
MA
0
0 100 200 300 400 500
FT3 [KN]
2
ED

Figure 23: (FT 32 − FT 33 ) interaction diagram

PT

450
CE

400

350
AC

300
[KN]

140
250
120
4
FT3

100
200
80

60
150
290 295 300

100

50

0
0 50 100 150 200 250 300 350
FT3 [KN]
3

Figure 24: (FT 33 − FT 34 ) interaction diagram

29
 ACCEPTED MANUSCRIPT

405, 3 + 299, 7 = 705 kN = M in [705 , (1200 − 329 = 871 kN )] = 705 kN

T
- At point C (Fig. 22), the force at Type T 4 (1425,8 kN ) is also less than its resistance
(1500 kN ). On one hand, it can be observed that the activation of the previous group

P
effect leads to a first influence on the redistribution force respecting the criterion of

RI
this group (407 + 298 = 705 kN ) (see Fig. 23 and Fig. 24). On the second hand, the
group of 3 bolt-rows is activated by T 32 , T 33 and T 34 . In accordance with Table 5

SC
where q = 3 and s = 4, it follows:
 
728, (1627 − 329 − 407 = 891 kN ),
298 + 339 = 637 kN = M in = 637 kN
(1044 − 407 = 637 kN )

NU
It can be observed that the group of 3 bolt-rows is activated at this point.

- At point D (Fig. 22), the force at Type T 4 reaches its resistance (1500 kN ). The bolt-rows
MA
from 1 to 4 are in plastic range (individual or group limit resistance). The force at
the last bolt-row T 35 reaches the value (127 kN ) that insures the equilibrium between
compression and tension zones. On one hand, it can be observed that the activation of
the previous group effects leads to a first influence on the redistribution force respecting
ED

following criteria (407 + 298 = 705 kN ) and (407 + 298 + 339 = 1044 kN ). On the
second hand, there is no other group effect activated. In accordance with Table 5
where q = 4 and s = 5, it follows:
PT

 
706, (1772 − 329 − 407 − 298 = 738 kN ),
339 + 127 = 466 kN < M in = 658 kN
(1363 − 407 − 298 = 658 kN ), (1046 − 298 = 748 kN
CE

7. Conclusion
AC

The proposed beam-to-column joint model appears as an easy-to-compute tool. It is

very useful in practice to be implemented in structural analysis code. It is based on the
component-based analysis that is nowadays well-known by the designers of joints. It is
worth to remind that its performance has been extended to solve the problem of the gap
that could appear between the column flange and the end-plate during the cyclic loading.
This problem has never been considered before, especially when it is included as a part of
the plasticity algorithm.
In addition, the group effect of two bolt-rows has been detailed to show how to implement
this effect into the plasticity algorithm. This model has been generalized to take into account
the group effect including more than two bolt-rows.
Both phenomena have been developed for simple cases of joints in order to simplify
the validation of the proposed model. Nevertheless, both improvements proposed to the
component based-model remain available for any bolted end-plate connection configuration.

30
 ACCEPTED MANUSCRIPT

i Concerning the gap effect, special care during computation is required for several
tests that have to be computed for the gap to control the plasticity of each component
within the joint. This care insures to approach the real behaviour of the joint including

T
appropriate slipping to recover different gaps.

P
ii Concerning the solution proposed for the group effect, more the number of bolt-rows
increases, more the number of criteria increases. The generalisation of the interaction

RI
formula has been easily verified for the case of a group of 3 bolt-rows and remains
available for more.

SC
Acknowledgements

NU
The authors gratefully acknowledge the financial support provided by the European
Commission (Research Fund for Coal and Steel) through the project ROBUST-IMPACT
under grant agreement RFSR-CT-2012-00029.
MA
References
[1] RICHARD, Ralph M., GILLETT, Paul E., KRIEGH, James D., et al. (1980). The analysis and design
of single plate framing connections. Engineering Journal, 1980, vol. 17, no 2.
ED

[2] Moncarz, P.D., Gerstle, K.H. (1981). Steel frames with non-linear connections. Journal of the Structural
Division, ASCE, 107, N ◦ ST8, Proc. paper 16440, 1427-1441.
[3] POGGI, C. et ZANDONINI, R. (1985) Behaviour and strength of steel frames with semi-rigid connec-
tions. In : Connection Flexibility and Steel Frames. ASCE, 1985. p. 57-76.
PT

[4] Nethercot, D. (1989). Methods of prediction of joint behaviour: beam-to-column connections. Elsevier
Applied Science, Structural Connections. Stability and Strength,, 23-62.
[5] Patel, K. V., and Chen, W. F. (1985). Analysis of a fully bolted moment connection using NONSAP.
Computers and structures, 1985, vol. 21, no 3, p. 505-511.
CE

[6] Bursi, O. S., and Jaspart, J.P. (1997). Benchmarks for finite element modelling of bolted steel connec-
tions. Journal of Constructional Steel Research, 43(1), 17-42.
[7] Yang, J. G., Murray, T. M., and Plaut, R. H. (2000). Three-dimensional finite element analysis of
AC

double angle connections under tension and shear. Journal of Constructional Steel Research, 2000, vol.
54, no 2, p. 227-244.
[8] Maggi, Y.I., Goncalves, R.M. and Malite, M. (2002). Comparative analysis of bolted T-stub and
corresponding extended endplate connections: a numerical approach European Conference on Steel
Structure, Eurosteel 2002, Coimbra, 1199-1218.
[9] DÍAZ, Concepción, MARTÍ, Pascual, VICTORIA, Mariano, et al (2011). Review on the modelling of
joint behaviour in steel frames. Journal of Constructional Steel Research, vol. 67, no 5, p. 741-758.
[10] YANG, Bo et TAN, Kang Hai. (2012). Component-based model of bolted-angle connections subjected
to catenary action. In : Proc., 10th Int. Conf. on Advances in Steel Concrete Composite and Hybrid
Structures.
[11] JASPART, J.P. (1996). Etude de la semi-rigidité des noeuds poutre-colonne et son influence sur la
résistance et la stabilité des ossatures en acier (Doctoral dissertation). Université de liège, Belgique.
[12] JASPART, J.P., BRAHAM, M., et CERFONTAINE, F. (1999). Strength of joints subjected to com-
bined action of bending moment and axial force. In : Proceedings of the Conference Eurosteel. 1999.
p. 26-29.
[13] CERFONTAINE, F (2004). Étude de l’interaction entre moment de flexion et effort normal dans les
assemblages boulonnés. Centre Technique Industriel de la Construction Métallique.

31
 ACCEPTED MANUSCRIPT

[14] DEL SAVIO A.A., NETHERCOT D.A., VELLASCO P.C.G.S., S.A.L. ANDRADA, L.F. MARTHA et
al. (2009). Generalised component-based model for beam-to-column connections including axial versus
moment interaction. Journal of Constructional Steel Research, 2009, vol. 65, no 8, p. 1876-1895.
[15] BAYO, E., CABRERO, J. M., et GIL, B. (2006). An effective component-based method to model semi-

T
rigid connections for the global analysis of steel and composite structures. Engineering Structures,vol.

P
28, no 1, p. 97-108.
[16] LEMONIS, Minas E. et GANTES, Charis J. (2009). Mechanical modeling of the nonlinear response of

RI
beam-to-column joints. Journal of Constructional Steel Research, 65(4), 879-890.
[17] LIU, Chang, TAN, Kang Hai, et FUNG, Tat Ching. (2015). Component-based steel beam–column con-
nections modelling for dynamic progressive collapse analysis. Journal of Constructional Steel Research,

SC
107, 24-36.
[18] Mazzolani, F. M. (1988). Mathematical model for semi-rigid joints under cyclic loads. Connections in
Steel Structurs: Behaviour, Strength and Design, Elsevier Applied Science Publishers, London, 112-120.
[19] Gang Shi, Yongjiu Shi, Yuanging Wang (2006). Behaviour of end-plate moment connections under

NU
earthquake loading. Engineering structures, 29(5), 703-716.
[20] YANG, Pu et EATHERTON, Matthew R. (2014). A phenomenological component-based model to
simulate seismic behavior of bolted extended end-plate connections. Engineering Structures, 2014, vol.
75, p. 11-26.
MA
[21] Da SILVA L.S., AHAHBAZIAN A., GENTILI F. and AUGUSTO H. (2016). Implementation of a
component model for the cyclic behaviour of steel joints. Eighth International Workshop on Connections
in Steel Structures, CONNECTIONS VIII, Boston, Massachusetts, USA
[22] DE LIMA, L. R. O., DA SILVA, L. Simoes, VELLASCO, PCG da S., et al. (2004). Experimental eval-
ED

uation of extended endplate beam-to-column joints subjected to bending and axial force. Engineering
Structures, 26(10), 1333-1347.
[23] Eurocode. 3 (2005). Design of steel structures, Part 1.8: Design of joints. Brussels: European Committee
for Standardization.
PT

[24] Krawinkler H. (1978). Shear design of steel frames joints AISC Engineering Journal, 15(3), 82-91.
[25] MARSDEN, SS Antman JE, WIGGINS, L. Sirovich S., GLASS, L., et al. (1993). Interdisciplinary
Applied Mathematics Springer
[26] DA SILVA, L. Simoes, DE LIMA, L. R. O., , VELLASCO, PCG da S., et al. (2004). Behaviour of flush
CE

end-plate beam-to-column joints under bending and axial force. Steel and Composite Structures, Vol.
4, N0. 2 , 77-94.
[27] BRODERICK, B. M. et THOMSON, A. W. (2002). The response of flush end-plate joints under
earthquake loading. Journal of Constructional steel research, vol. 58, no 9, p. 1161-1175.
AC

32