Journal of Constructional Steel Research 133 (2017) 459–474

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

Effect of contact on the elastic behaviour of tensile bolted connections

Maël Couchaux a,⁎, Mohammed Hjiaj a, Ivor Ryan b, Alain Bureau b
a
Structural Engineering Research Group/LGCGM, INSA de Rennes/UEB, 20 avenue des Buttes de Coësmes, CS70839, F–35708 Rennes Cedex 7, France
b
Centre Technique Industriel de la Construction Métallique, Espace technologique - L'Orme des Merisiers - Immeuble Apollo, 91190 Saint-Aubin, France

a r t i c l e i n f o a b s t r a c t

Article history: Both experimental evidence and 3D finite element analyses indicate that contact forces between the connected
Received 27 June 2016 parts influence the behaviour of bolted T-stub and L-stub connections in tension. In the present paper, a mechan-
Received in revised form 13 October 2016 ical model that predict the elastic behaviour of such connections is developed. The model relies on the enhanced
Accepted 15 October 2016
beam theory, proposed by Baluch et al. [13], to describe the mechanical response of the flanges in both the contact
Available online xxxx
and the non-contact regions. The model response shows that the contact stresses distribution strongly depends
Keywords:
on the ratio between the length of the contact area and the flange thickness. Furthermore, the length of the con-
Contact tact zone in a non-preloaded connection appears to be independent from the magnitude of the external force.
Bolt Several simplifications have been proposed to facilitate the determination of the separation length which is the
Connection primary unknown of the problem. To validate the analytical and calculation models, a 3D finite element model
Prying effect has been developed. Analytical and numerical results are in good agreement which demonstrates the validity
of the proposed mechanical model. The theoretical and numerical results confirm that prying action increases
with decreasing value of the ratio between the flange stiffness and the bolt stiffness.
© 2017 Elsevier Ltd. All rights reserved.

1. Introduction shows that the force B in the bolt is equal to the sum of the external
load FT and the prying force Q (see Fig. 2).
One of the most significant behavioral characteristic of nearly all For some time now, the prying force has been represented by a con-
types of tensile bolted connection is prying action. The prying action re- centrated force acting at or near the flange plate edges of T-stubs ([2–5])
fers to secondary forces that develop in tensile bolts in addition to ten- or L-stubs (Seidel [1], Petersen [6]). In such a model, the position of the
sion caused by the externally applied forces. These secondary forces prying force does not depend on the flange geometry. Agatonovic [7]
which are given by the integral of the normal stress distribution over proposed to position an elastic support at the point of application of
the contact zone, can significantly increase the bolt force. The distribu- the prying action (located between the bolt centreline and the free
tion of the normal stresses in the contact zone not only influences the edge) with a stiffness depending on the bolt dimensions and the flange
tension force in the bolt but also the joint stiffness. The effect of prying thickness. Another approach developed by Kato & Tanaka [8] and
action is particularly important for bolted T-stub and L-stub type con- Lemonis & Gantes [9] is to consider a fixed support at the point of tran-
nections which are the basic components of a wide range of bolted con- sition between the contact and the non-contact regions. At this point,
nections for beams, columns, large span trusses, chimneys, wind the curvature is null and so is the bending moment. In the elastic
turbines and pylons. L-stubs are commonly used to model the tension range, the equilibrium position thus obtained is unique and depends
part of ring flange connections (see Fig. 1). The tube-wall welded to on the ratio between the stiffness of the flange and the stiffness of the
the flange is subdivided into nb L-shaped segments, where nb corre- bolt. This hypothesis is acceptable for a relatively small contact area.
sponds to the number of bolts. The pertinence of this approach has Senda et al. [10] proposed to consider a linear distribution of the con-
been validated by Seidel [1]. tact pressure. However it is shown here that the contact pressure distri-
L-stubs provide an illustration of the prying mechanism (see Fig. 2). bution may depend on the extent of the contact area and may not be
The externally applied force FT by the tube produces bending effects in unique in shape. Bakhiet [11] has developed a model which takes into
the flanges which in turn cause contact stresses that develop near the account the size of the contact area but the prying action predicted be-
outer edges of the flange. The force-resultant of the contact stress distri- comes null in the case of a pointwise contact at the free edge of the
bution Q is the prying force. Summing the forces acting on the flange flange. Chakhari [12] proposed an incremental model where the flange
in contact is assumed to be a Bernoulli beam resting on an elastic
Winkler foundation. The stiffness of the foundation is numerically com-
⁎ Corresponding author. puted and depends on the bolt dimensions and the flange thickness as
E-mail address: [email protected] (M. Couchaux). in the Agatonovic model. The above-mentioned analytical models rely

http://dx.doi.org/10.1016/j.jcsr.2016.10.012
0143-974X/© 2017 Elsevier Ltd. All rights reserved.
460 M. Couchaux et al. / Journal of Constructional Steel Research 133 (2017) 459–474

complexity of the solution procedure, a direct application of the above

solutions to the present engineering problem seems to be cumbersome
and difficult to apply in a day-to-day design. The second approach em-
ploys engineering beam theories (Bernoulli or Timoshenko) instead of
continuum mechanics. The main shortcoming originates from the con-
straints placed on the deformation map. Indeed, engineering beam the-
ory assumes that cross sections that are plane before deformation
remain plane after deformation (plane-sections hypothesis). An equally
important assumption is that those plane sections do not distort in their
own planes, either. This hypothesis makes the contact stresses distribu-
tion indeterminate and therefore not unique. Furthermore, if Bernoulli
kinematics is considered then the bending moment equal to zero in
the contact zone as a result of the nullity of the curvature. To ease the de-
termination of the contact stress distribution, the rigid cross-section as-
sumption must be relaxed to permit transverse deformation while
preserving the simplicity of an engineering beam theory. One possible
way is to replace the rigid foundation by a semi-infinite elastic substrate
Fig. 1. Ring flange connection and L-stubs [1]. or an elastic foundation and transfer the beam cross-section
deformability to the foundation. A second approach consists in develop-
ing an analytical model, based on the refined beam theory of Baluch et al.
on important simplifications to circumvent the complexity of contact [13] to investigate the behaviour of prismatic solid in contact with a
interaction between the flanges of tensile bolted connections. From smooth rigid foundation (see Couchaux et al. [15]). The effects of both
the above discussion, it becomes apparent that the accuracy of any me- the transverse and the shear deformations are taken into account.
chanical model for such connection is strongly affected by the predic- The main purpose of this paper is to present a general analytical
tion of the contact stress distribution between the flanges. Thus the model for non-preloaded bolted connections in which the contact inter-
main challenge of this problem is to provide an accurate yet simple action between the flanges are accurately computed and so is the prying
model that reproduces correctly the contact interaction taking place be- force. The main ingredient of the proposed connection model is the en-
tween the flanges. Exploiting the “double” symmetry of the T/L-stub as- hanced beam theory which allows to accurately reproduce the contact
sembly, the flange-to-flange contact is replaced by a flange-to-rigid interactions between the flanges [15] as well as the shear deformability
foundation contact interaction where the flange is modelled as a beam. of the flange. This mechanical model is applied to analyse the behaviour
Two main routes have been followed to solve this problem. In the first of T-stubs and L-stubs subjected to a tensile force. The position of the
approach, the elastic beam is considered as a deformable elastic body prying force, the bolt force and the evolution of the contact pressure
in smooth contact with a rigid foundation which results in a mixed are particularly studied. Simplifications are proposed to ease the deter-
boundary-value problem ([16–18]). These mixed boundary-value prob- mination of the separation length. Predictions of this model are com-
lems have been solved by means of an appropriate displacement solu- pared favourably against numerical simulations based on the code
tion to the equations of plane strain linear elasticity [18]. Due to the ANSYS V.11.

Fig. 2. L-stub in tension.

 M. Couchaux et al. / Journal of Constructional Steel Research 133 (2017) 459–474 461

2. Analytical model for T-stubs and L-stubs width of the “beam” pb is taken here as that proposed for T-stubs by
Lemonis & Gantes [9]:
Bending deformations of the flange produce contact stresses over a 8
1
certain region. The resultant of these normal stresses is the prying leff < if e1 =pb b0; 87
force, located between the free edge of the flange and the bolt axis. ¼ 0; 92 þ 0; 06=ðe1 =pb Þ2 ð1Þ
pb :
The exact location of the prying force is strongly affected by the flexural 1 if e1 =pb ≥0; 87
stiffness of the flange and the bolt stiffness. With the exception of the in-
cremental model proposed by Lemonis & Gantes [9], most existing Standard Bernoulli beam theory is used for the “tube wall” slice. The
models consider that the location of the prying force is given a priori. area of the flange extending from the bolt to the outer edge is consid-
The assumption that the prying force acts on the tip of the flange is gen- ered as the potential contact area (see Fig. 2 and Fig. 4). In the present
erally accepted and is considered accurate until the length of the flange model, the behaviour of the flange in the contact area is considered
exterior to the bolt becomes large or until the flange thickness becomes via the enhanced beam model [15] whilst the response of the lift-off re-
small. gion of the flange is described using the refined beam theory of Baluch
In this paragraph, a new model for L-stub and T-stub connections is et al. [13]. The model assumptions on both zones (lifted and contact)
presented. The main ingredient of this new model is the enhanced beam being the same, compatibility is ensured. The bolt is represented by a
theory (see Couchaux et al. [15]) used to accurately evaluate the extent linear elastic spring connected to the rigid support. The absence of a
of the contact zone, the deformation of the flange, the contact stress dis- bolt hole in the plate should compensate for any bolt bending. The dis-
tribution between the flanges and therefore the location of the prying tance between the bolt axis and the inner edge of the contact zone,
force from which the bolt force is evaluated. noted ξ, is chosen as the primary unknown of the problem. The model
for the flange in contact with a rigid foundation is presented in details
2.1. Model definition and assumptions in Section 2.1.2.
The stiffness of the bolt is defined according to the EN 1993-1-8
L-stub connections are used to study the tensile part of bolted flange rules:
connections of large diameter tubes. A typical experimental set-up, used
by Petersen [6], is presented in Fig. 3. The connection is horizontally re- EAs
kb ¼ ð2Þ
strained at the junction between the tube wall and the flange. This con- Lb =2
dition reflects the fact that for a large tube diameter, the flange behaves
as a ring and, thus, the tube is restrained in the radial direction at the where Lb is the equivalent length of the bolt similar to that in EN 1993-
tube-flange junction. The length of the “tube wall” plate, L, is evaluated 1-8 [2], As the resistance area of the bolt and E the young modulus.
such as to give the same flexural stiffness as the actual tube wall (see
Section 2.1.1). 2.1.1. Model of the tube-wall
Due to double symmetry conditions, bolted flange connections are To reproduce the load-transfer mechanism from the cylindrical shell
modelled as L- or T-stub in which the flange-to-flange contact interac- to the flange, the former is replaced by a simply supported thin plate
tion is replaced with a flange-to-rigid foundation (plane of symmetry) whose cross-section geometry (width and length) is defined so that it
contact interaction and thus a single flange is represented. The model replicates the behaviour of the cylindrical shell. The length of the
developed in this paper is valid for symmetric L-stubs. tube-wall is evaluated such as the beam flexural stiffness is equal to
The material is assumed to remain linear elastic throughout. The ef- that of the actual cylindrical shell. As a result, the bending moment mE
fective length, leff, of the L-stub which corresponds to the cross-section applied by the tube-wall to the flange is equal in the two systems for a

Fig. 3. Geometry of a L-stub connection.

462 M. Couchaux et al. / Journal of Constructional Steel Research 133 (2017) 459–474

and the equivalent length is:

 
3 1−ν2
L¼ ð6Þ
2β

2.1.2. Model of the flange

The flange is subdivided in two parts: one portion is in contact with
the rigid foundation and the other not. Both parts are modelled using an
enhanced beam theory. In the lifted zone, the model proposed by Baluch
et al. [13] is employed. The latter was further enhanced by Couchaux
et al. [15] to account for contact conditions. This model has been devel-
oped to address the limitations of the classical Bernoulli beam theory in
predicting the contact pressure distribution between a beam and a rigid
foundation. In this section, the governing equations of the enhanced
beam model used to analyse bolted connections are briefly summarized.
Fig. 4. Model of an L-stub connection with a non-preloaded bolt.
Full details on the derivation of this contact beam model can be found in
[15]. Many authors (see for instance Chakhari [12]) suggest to replace
the rigid foundation with an elastic one. However for relatively short
contact area, shear deformations are important and transverse normal
given rotation (see Fig. 5). The two elements are fixed horizontally at stresses are expected to play an important role. It is therefore essential
the junction between the flange and the tube-wall. that the model takes into account both phenomena. The refined beam
The equivalent beam being simply supported at both ends, the rota- model proposed in [15] takes into account both the influence of trans-
tion produced by the bending moment mE (see Fig. 5) is: verse normal strain as well as contact conditions. The model is con-
structed by combining in a certain way two-dimensional elasticity
L (plane stress) and the classical engineering beam theory. In the lifted
θE ¼ mE ð3Þ
3EIt area, the refined beam theory proposed by Baluch et al. [13] is
recovered.
t3 l
with I t ¼ t12eff where leff is equal to unity for the present derivation. Couchaux et al. [15] consider a beam subjected to a distributed load
The rotation and the displacement at the extremity of a simply sup- p0(x) acting in the downward direction (see Fig. 6). The distributed load,
ported semi-infinitely long cylindrical shell subjected to the action of which acts over the entire beam length, is applied to the upper face of
bending moment mE and shearing force vE, both uniformly distributed the beam where we have the following stress boundary conditions:
along the bottom edge (see Fig. 5), are given by the following expres-
sions [19]: ∀x ∈ ½0; L : σ zz ðx; t f =2Þ ¼ −σ 0 ðxÞ and τxz ðx; t f =2Þ ¼ 0 ð7Þ

1 with σ0(x) = p0(x)/leff.

θt ¼ 2
½2βmE −vE  ð4Þ On the bottom surface of the beam, the boundary conditions on the
2β Dt
stresses are:

1 ∀x ∈ ½0; L : σ zz ðx; −t f =2Þ ¼ −σ c ðxÞ and τxz ðx; −t f =2Þ ¼ 0 ð8Þ

wt ¼ − ½βmE −vE  ð5Þ
2β3 Dt
where σc(x) is the contact stress and is lied to the contact pressure dis-
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
tribution p(x) by:
4 3ð1−ν Þ
2 Et 3t
with: β ¼ and Dt ¼ :
R2 t 2t 12ð1−ν 2 Þ σ c ðxÞ ¼ pðxÞ=leff
The tube-wall being simply supported, relation (5) gives:
On this surface, the unilateral contact conditions must be specified.
vE ¼ βmE These are characterized by a geometric condition of non-penetration,

θ θ

a) Semi-infinite cylindrical shell b) Equivalent beam

Fig. 5. Model of the tube-wall.
 M. Couchaux et al. / Journal of Constructional Steel Research 133 (2017) 459–474 463

Fig. 6. Beam on a rigid foundation [15].

a static condition of no-adhesion and a mechanical complementary con- where

dition:   3
dw1 ðxÞ 1 1 dMðxÞ 1 d M ðxÞ
ϕðxÞ ¼ − þ þ þ ð14Þ
∀x ∈ ½0; L : wðx; z ¼ −t f =2Þ≥0; σ c ðxÞ ≥0; σ c ðxÞ wðx; z ¼ −t f =2Þ ¼ 0 dx α1 α2 dx α 3 dx3
ð9Þ
5Eleff t f 10Eleff t f 1120Eleff leff t 3f
α1 ¼ ; α2 ¼ ; α3 ¼ ;If ¼ ð15Þ
The non-penetration condition constraints the transverse displace- 12ð1 þ νÞ 3ν 39t f 12
ment w at z = −tf / 2 while the dual relation involves the contact stress
ϕ can be assimilated to an average cross-section rotation [13]
σc between the beam and the rigid foundation which must be positive
where the first term corresponds to the derivative of the beam
(σc N 0) where there is contact and zero where there is no contact.
axis transversal displacement while the second one corresponds
The above relations make the contact problems highly non-linear.
to the contribution of the shear deformation to the cross-section
Contact conditions between the flange and the rigid foundation re-
rotation. The previous equations can be used for both the contact
quires that the transverse displacement of lower surface of the beam
and the lifted zones.
is equal to zero everywhere on the contact zone:
Over the contact area, the displacement at the bottom surface of
wðx; −t f =2Þ ¼ 0 ð10Þ the flange is equal to zero. Inserting this kinematic constraint into
Eq. (11) and eliminating the contact pressure distribution p using
The procedure followed to develop the enhanced beam model in- Eq. (12), we obtain the expression of the mid-axis transverse dis-
volves several steps where two-dimensional elasticity (plane stress) placement in the contact area in terms of bending moment and ex-
and the classical engineering beam theory are combined together in a ternal loading:
certain way to produce an enriched displacement field permitting trans- 2
13t f d MðxÞ 3ν t
verse deformation to occur. w1 ðxÞ ¼ þ MðxÞ− f p0 ðxÞ ð16Þ
The derivation produces the following expression for the transverse 32Eleff dx2 2Eleff t f 2Eleff
displacement (see [15] for details):
Introducing Eq. (16) into Eq. (13), we get the differential equation
" # " # that relates the bending moment to the external distributed load:
pðxÞ 3z 2
z p ðxÞ
4
3z z 6νMðxÞz
2 4 2
wðx; zÞ ¼ − 2z− þ2 3 − 0 2z þ −2 3 −
4Eleff tf tf 4Eleff tf tf Eleff t 3f 4
d M ðxÞ
2
d M ðxÞ
2
d p0 ðxÞ
þ w1 ðxÞ 4
−2 α 2
þ β MðxÞ ¼ ρ 2
ð17Þ
dx dx dx
ð11Þ
42 420 35
with α ¼ ,β¼ and ρ ¼ :
where w1(x) is the transverse displacement of the beam centreline (z = 13t 2f 13 t 4f 26
0). Although the strong form of equilibrium is partly lost, the overall The solution of Eq. (17) along with the boundary conditions at the
equilibrium of the beam is satisfied: contact zone border provides the expression of the bending moment
in the contact zone from which we can deduce the contact stresses.
dV dM This model has been used to study beams in contact with rigid founda-
þ p−p0 ¼ 0; −V ¼ 0 ð12Þ
dx dx tions [15] and the results obtained were in good agreement with those
of numerical and theoretical [18] models.
It can be seen that the deflection depends on both the external load This model will be used to study the contact area of two bolted com-
p0 and the contact pressure distribution and varies continuously ponents commonly considered in structural engineering connections:
through the cross-section depth. This is in sharp contrast with the clas- the L-stub and the T-stub subjected to a tensile force.
sical beam theories based on rigid cross-section assumption. Next, an
enhanced expression for the longitudinal stresses is derived based on 2.1.3. Type of contact conditions
the expressions of both the longitudinal and the transverse displace- Three different types of contact conditions between opposite flanges
ment field (see Couchaux et al. [15]). Integrating the longitudinal stress- can be observed (see Fig. 7):
es over the cross-section depth, we obtain the following differential
equation relating the bending moment to the beam centreline deflec- Case 1. The contact area develops over a portion e2 − ξ of the flange.
tion: This case can be observed in presence of thin flange and is treated in
Section 2.3.
" #
2
dϕðxÞ 6ν d MðxÞ
MðxÞ ¼ EI f − ð13Þ Case 2. The flanges are in pointwise contact at their outer edge. This
dx 5Et f leff dx2
case is dealt with in Section 2.4.
464 M. Couchaux et al. / Journal of Constructional Steel Research 133 (2017) 459–474

a) Case 1: Surface contact b) Case 2: Single-point contact c) Case 3: complete lift-off

Fig. 7. Interactions between flanges.

Case 3. If the flange and the tube wall are sufficiently stiff, a complete From Fig. 8, it can be seen that the lever arm (distance to the inner
lift-off of the flanges can be observed. Prying effect does not take place boundary of the contact zone) of the prying force is equal to the ratio
and the adjacent flanges are fully separated. This case is investigated M0/Q:
in Section 2.5.
M0
Before the analysis of these three cases, the mechanical response of lðξÞ ¼ ð19Þ
Q
the flange in the contact area modelled using the enriched beam
model [15] is detailed in Section 2.2. It will give us a mean to accurately
characterize the position of the prying force which is useful for case 1. The boundary conditions for the bending moment and the shear
force are (see Fig. 8):
2.2. Mechanical response of the flange in the contact area
M ð0Þ ¼ 0;
In the first case, contact takes place over a certain portion of the po- dM
ð0Þ ¼ 0;
tential contact area located between the outer edge of the flange and the dx ð20Þ
M ðe2 −ξÞ ¼ −M0 ; M0 ≥0
bolt axis. Considering only the portion of the flange in contact, the load- dM
ing comprises (see Fig. 8) the contact pressure distribution p, a bending ðe2 −ξÞ ¼ −Q ; Q ≥0
dx
moment M0, and a shear force Q equal to the prying force. Both M0 and Q
are applied at the point of transition between the contact and non-
At the inner boundary of the contact area (located at x = e2 − ξ), the
contact regions (see Fig. 8). At this point, the cross-section is allowed
contact stress and thus the contact pressure distribution are equal to
to rotate, i.e. no kinematic constraints are imposed.
zero:
The prying force Q can be computed by integrating the contact pres-
sure distribution over the contact area: 
d M
2
pðe2 −ξÞ ¼ − 2  ¼0 ð21Þ
dx x¼e2 −ξ
eZ
2 −ξ

Q¼ pðxÞdx ð18Þ
0 Considering that no distributed load is applied to the upper face of
the flange (p0 = 0), equilibrium Eq. (17) becomes:

4 2
d M ðxÞ d M ðxÞ
4
−2α 2
þ βMðxÞ ¼ 0 ð22Þ
dx dx

The bending moment distribution, solution of Eq. (22), has the fol-
lowing expression:

h
MðxÞ ¼ −M0 e−a0 x ðC 1 cosðb0 xÞ þ C 2 sinðb0 xÞÞ
i
þ ea0 x ðC 3 cosðb0 xÞ þ C 4 sinðb0 xÞÞ ð23Þ

with

" !# " !#
p ffiffiffiffi arccos α p ffiffiffiffi arccos α
a0 ¼ β cos
4
pffiffiffiffi ; b0 ¼ β sin
4
pffiffiffiffi ð24Þ
2 β 2 β
Fig. 8. Area of the flange in contact.
 M. Couchaux et al. / Journal of Constructional Steel Research 133 (2017) 459–474 465

Consequently l(ξ) can be express as:

 
e2 −ξ
lðξÞ ¼ ψ t f ¼ ψðωÞ t f ð29Þ
tf

The curve ψ(ω) is depicted in Fig. 9.

An accurate approximation of the function ψ(ω) is given by

2ω=3
ψðωÞ ¼ ð30Þ
 6
1
6
ω
1 þ 1;11

Inserting Eq. (30) into Eq. (29), we obtain the following expression
for l(ξ)

2ðe2 −ξÞ 1
lðξÞ ¼ ð31Þ
3  6
1
6
e2 −ξ
Fig. 9. Function ψ(ω). 1þ 1;11t f

Introducing α and β in the expressions of a0 and b0, gives: The contact pressure distribution can be evaluated using:

2; 11 1; 1 2
a0 ¼ and b0 ¼ ð25Þ d M
tf tf p¼− 2
ð32Þ
dx

The four constants of integration Ci are evaluated via the boundary The ratio of the contact pressure and its maximum value is presented
conditions (20) and (21), their expressions are given in Appendix A. in Fig. 10 for different values of ω, which is the ratio between the length
Taking the derivative of Eq. (23) gives the shear force or equivalently of the contact area and the thickness of the flange. As can be seen, the
the prying force and therefore the distance l(ξ) between the prying shape of the contact pressure distribution is governed by this
force and the point of transition between the contact and non-contact parameter.
regions can be evaluated via Eq. (19) to give: For low values of ω, the distribution of the contact pressure is linear
whereas for larger values of this parameter, the contact stress distribu-
1 tion becomes non-linear. It has been found that the upper limit of the
lðξÞ ¼ contact area length is bounded by 2.42 × tf (ω = 2.42 on Fig. 10). For
e−a0 ðe2 −ξÞ ½b0 μ 1 ðe2 −ξÞ−a0 φ1 ðe2 −ξÞ þ ea0 ðe2 −ξÞ ½a0 φ3 ðe2 −ξÞ þ b0 μ 3 ðe2 −ξÞ
this value of ω, the contact pressure becomes equal to zero at the
ð26Þ
outer edge of the flange. For larger values of ω, the contact stresses van-
ish over a certain portion of the contact area located near the free edge
where of the flange. Finally, for a given value of the thickness, the shape of the
contact pressure distribution depends on the size of the contact area.
φi ðe2 −ξÞ ¼ C i ðe2 −ξÞ cos½b0 ðe2 −ξÞ þ C iþ1 ðe2 −ξÞ sin½b0 ðe2 −ξÞ ð27Þ This result is in sharp contrast from the assumption made by Senda
et al. [10] and Bakhiet [11] in the derivation of their model.
μ i ðe2 −ξÞ ¼ C iþ1 ðe2 −ξÞ cos½b0 ðe2 −ξÞ−C i ðe2 −ξÞ sin½b0 ðe2 −ξÞ ð28Þ
2.3. Case I: surface contact between the flanges

The full expression of l(ξ) given in Eq. (26) is not convenient for A virtual cut is performed at the point of transition between contact
practical applications. The coefficients Ci being a function of and non-contact regions thus creating two segments (see Fig. 11). The
ω = (e2 − ξ)/tf (see Appendix A) hence functions φi and μi depend also first portion of the flange corresponds to that in contact with the rigid
on ω. foundation, the length of which is equal to e2 − ξ. This segment of the
flange has been studied in Section 2.2 and all mechanical variables
have been determined based on the ratio M0/Q. The second part of the
flange lifts off from the foundation. The separation length ξ will be cal-
culated considering equilibrium of the lift-off part of the flange and
the continuity of the stress-resultants at the point of transition between
the contact and non-contact regions. To ease the derivation, continuity
of the average cross-section will not be enforced nor the mid-axis trans-
verse displacement.
The overall equilibrium conditions of lift-off portion of the flange
produce the following equations

Q ¼ B−F T ð33Þ

M0 ¼ F T ðe1 þ ξÞ−Bξ−M E ð34Þ

Inserting Eqs. (33) and (19) into Eq. (34) gives:

F T ðe1 þ nÞ−M E
B¼ ð35Þ
Fig. 10. Shape of the contact pressure distribution. n
466 M. Couchaux et al. / Journal of Constructional Steel Research 133 (2017) 459–474

Fig. 11. Surface contact: mechanical model.

where n is the distance between the location of the prying force and the Similarly, the average cross-section rotation at the junction between
bolt axis: the flange and the tube-wall ^x ¼ ξ þ e1 can be written as:

n ¼ ξ þ lðξÞ ð36Þ ϕðξ þ e1 Þ ¼ −Θ F F T þ ΘB B þ ΘME M E ð41Þ

From expression (35), it becomes obvious that the bolt force de- where
pends on the position of the prying force. 2 2
The contact pressure distribution p being equal to zero in the non- ξ þ e1 ξ ðξ þ e1 Þ
ΘME ¼ ; ΘB ¼ ; ΘF ¼ :
contact region and since no external distributed loading is applied to EI f 2EI f 2EI f
the flange, the relationship between the bending moment and the
mid-axis transverse displacement is given simply by: Cross-section continuity at the junction between the tube-wall and
the flange is enforced:
d w1 ð^xÞ
2
Mð^xÞ ¼ −EI f ð37Þ ME
d^x
2 ΘME M E þ ΘB B−Θ F F T ¼ −
kθ

t 3f leff where
with I f ¼ :
12
For 0≤ ^
x ≤ξ, we have the following expression for the bending mo- 3EI t
kθ ¼ :
ment: L

Mð^xÞ ¼ − F T ðe1 þ ξ−^

xÞ þ Bðξ−^xÞ þ ME ð38Þ Thus, we get the following expression of the bending moment ME:

where ME is the bending moment applied by the tube-wall to the flange. ME ¼ −mB B þ m F F T ð42Þ
We assume that the flange is weakly clamped at ^x ¼ 0 that is both the
transverse displacement w1 of the flange centreline and the average ro- where
tation ϕ are equal to zero at ^x ¼ 0. The average cross-section rotation ϕ ΘB ΘF
is related to the centerline deflection and the bending moment through mB ¼ ; mF ¼ :
1 1
ΘME þ ΘME þ
kθ kθ
dw1 ð^xÞ 1 dMð^xÞ
ϕð^
xÞ ¼ − þ ð39Þ
d^
x GA f d^x
Combining relation (35) and (42), we obtain the equation relating
the bolt force B to the external force applied FT:
with Af = 0 , 81tfleff.
Inserting Eq. (38) into Eq. (37), integrating twice the outcome and e1 þ n−m F
making use of the above mentioned boundary conditions, we obtain B¼ FT ð43Þ
n−mB
the following expression for the deflection of the beam centreline:
" ! ! # Expressing that the elongation of the bolt is equal to the flange cen-
1 ^x2 ^x3 ξ ^x
2
^x3 ^x2 terline displacement (z = 0) at the bolt location, one obtains:
w1 ð^
xÞ ¼ FT ðe1 þ ξÞ− −B − −M E
EI f 2 6 2 6 2
^x B
þ ð F T −BÞ ¼ F T δ F −BδB −M E δME ð44Þ
GA f kb

At the bolt axis, this displacement can be expressed by summing up Inserting Eqs. (42) and (43) in the above equation, we get:
the contribution of FT, B and ME:  
1  
gðξÞ ¼ ðe1 þ n−m F Þ þ δB −δME mB −ðn−mB Þ δ F −δME m F ¼ 0 ð45Þ
kb
w1 ðξÞ ¼ F T δ F −BδB −M E δME ð40Þ

Eq. (45) permits to determine the separation length ξ and thus the
where
extent of the contact area. This equation is non-linear and may be solved
numerically. The calculation of the bolt force can be done via relation
2 3 2
ξ ðe1 =2 þ ξ=3Þ ξ ξ ξ ξ (43). It is worth mentioning that ξ doesn't depend on the magnitude
δF ¼ þ ; δB ¼ þ ; δME ¼ :
EI f GA f 3EI f GA f 2EI f of the tube force FT. The ratio between the bolt force and the tensile
 M. Couchaux et al. / Journal of Constructional Steel Research 133 (2017) 459–474 467

force applied to the connection is noted η: Furthermore, equilibrium in the vertical direction gives:

B ¼ η FT ð46Þ B ¼ FT þ Q ð50Þ

Combining Eqs. (47) to (50), one obtains the relation between the
The size of the contact area, given by the value of ξ, lies within the
bolt force and the tensile force:
interval [0,e2]. Any solution to Eq. (45) which is found to be outside
this interval means that the contact is pointwise and therefore ξ = e2. δQ 1 −δQ 2
The next paragraph addresses this case. B¼ FT ¼ η FT ð51Þ
1=kb þ δQ 1 −δQ 2 −ðδB1 −δB2 Þ

2.4. Case 2: single-point contact between the flanges

2.5. Case 3: complete lift-off of the flanges
When the stiffness of the bolt is not large enough compared to that
of the flanges, pointwise contact will develop at the free edge of the
For T-stub, prying effect disappears when the thickness of the flange
flanges. Petersen [6] proposed a similar model where the flange de-
is important compared to that of the bolt. Hence in EN 1993-1-8 [2], this
forms according to the Bernoulli beam theory. In the present paper,
effect develops if the length of the bolt, Lb, is less than the following
the beam theory of Bernoulli is replaced with the refined beam theory
limit:
of Baluch et al. [13].
To simplify the analysis, the response of the flange is studied consid- 8; 8m3 As
ering the equivalent mechanical system presented in Fig. 12 where the Lb ¼
leff t 3f
vertical support at the free edge corresponding to a point-wise contact
with a rigid foundation has been moved to the junction between the where the bolt length, m, and the equivalent length, leff, of the T-stub are
flange and the tube and supplemented by a torsional spring that repro- calculated via EN 1993-1-8 [2]. As can be seen from the deformed shape
duces the flexural interaction between the tube-wall and the flange. The of the connection (see Fig. 13), the flanges start to be in pointwise con-
flange is subjected to the bolt force B and the prying force Q, which tact at the free edge if the following condition holds:
equilibrate the applied tensile force FT. In the meantime, compatibility
has to be respected between the bolt and the flange. The prying force δb ¼ e2 θb ð52Þ
and the bolt force act at point 1 and point 2, respectively. Under unit
prying force (Q = 1) alone, the deflections at point 1 and at point 2 A bound on the bolt length can then be determined for a L-stub
are taken to be δQ1 and δQ2, respectively. Similarly, under unit bolt which corresponds to the transition from pointwise contact to full uplift
force (B = 1), the deflections at point 1 and at point 2 are taken to be of the opposite flange (see Fig. 13). In this case, the rotation of the flange
δB1 and δB2, respectively. The total flange deflection at point 2 can be at the bolt axis is:
expressed as:
e1 e2
θb ¼ F T þ 1 ð53Þ
w2 ¼ QδQ 2 −BδB2 ð47Þ kθ 2EI f

e21 e2 e1 1 e1 þ e2 e3 1 e1 and the displacement of the flange at this point is:

with δQ 2 ¼ ð þ Þ þ e1 ð þ Þ and δB2 ¼ 1 þ e1 ð þ Þ:
EI f 2 3 GA f kθ 3EI f GA f kθ
FT
δb ¼ ð54Þ
kb
Similarly, the total flange deflection at point 1 is computed as:

Inserting Eqs. (2), (53) and (54) in Eq. (52), the limit on the bolt
w1 ¼ QδQ 1 −BδB1 ð48Þ
length is obtained for a L-stub:

ðe1 þ e2 Þ3 e1 þ e2 ðe1 þ e2 Þ2 e2 e e1 L e21

with δQ 1 ¼ þ þ and δB1 ¼ δB2 þ e2 ð 1 þ 1 Þ: Lb ¼ 2e2 As þ ð55Þ
3EI f GA f kθ 2EI f kθ 3It 2I f

Compatibility demands that the elongation of the bolt is equal to the For a length of the bolt shorter than this limit, the prying effect will
difference between w1 and w2: develop. In case the tube-wall is more flexible compared to the flange,
the length L is large and so is the limit length. Note that the same
B
w1 −w2 ¼ − ð49Þ
kb

Fig. 12. Single-point contact: mechanical model. Fig. 13. Case 3: Complete lift-off – No prying effect.
468 M. Couchaux et al. / Journal of Constructional Steel Research 133 (2017) 459–474

Fig. 14. Simplified contact model.

relation can be obtained considering that η = 1 in Eq. (51) when prying

forces vanish.

2.6. Simplifications Fig. 16. Approximation of the function l(ξ).

2.6.1. Simplified model free edge of the flange, is:

The proposed analytical model is too involved for practical applica-
tions. Indeed, the separation length ξ is determined by numerically solv- λ0 þ 1
ing Eq. (45). In the present section, simplifications of the model are α R;0 ¼ ð58Þ
λ30
suggested and closed-form expressions for the separation length ξ are
given.
Assuming that l(ξ) is equal to zero and neglecting the flexural stiff- with λ0 ¼ ee21 :
ness of the tube wall (kθ = 0) and the shear deformation of the flange Contact develops on the free edge of the flange when αR is less than
(GAf → ∞), Eq. (45) becomes: αR,0 and between the free edge and the bolt axis otherwise. The bolt
force has the following simple expression:
 3
ξ ξ
αR − −1 ¼ 0 ð56Þ
e1 e1 e1 þ ξ
B¼ FT ð59Þ
ξ
e31 kb
with α R ¼ :
6EI f In the present model, the length n (equal to ξ) is underestimated and
the bolt force is overestimated.
In the present simplification, the prying force is supposed to be
located at the point of transition between the contact and non-contact 2.6.2. Calculation method
regions (see Fig. 14). At this point, the flange is considered to be To ease the calculation of ξ, we propose to evaluate the separation
clamped. The separation length ξ is the solution of a third order equa- length via the following empirical expression:
tion involving the coefficient αR which is related to the stiffness of the
flange and the bolt. For αR ≥ 4/27, the single real solution to the above rffiffiffiffiffiffiffiffiffi
α R;0
equation has the following expression: ξ ¼ e2 ≤e2 ð60Þ
αR
ξ 1 1 e2
¼ γL þ ≤ ð57Þ
e1 3α R γL e1 As already mentioned, the separation length depends on the factor
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi αR which is the main parameter involved in the expression of ξ
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi
3 −12þ81α R (Eq. (57)). This factor depends on the stiffness of the flange and the
ð108 þ 12 αR Þα 2R
bolt which are two important parameters influencing the position of
where γL ¼ :
6α R the prying force. For αR equal to αR,0, the separation length is equal to
e2 and the prying force acts on the free edge of the flange. The flange re-
The ratio ξ/e1 depends on the coefficient αR and must be less than e2/e1. mains clamped at the lift-off point when the separation length is less
The limit value of αR, which corresponds to a prying force applied at the than e2 (see Fig. 15).

FT FT
e2 e1 ξ

Q n

a) Single-point contact: n = e2 b) Surface contact : n < e2

Fig. 15. Simplified method: position of the prying force.
 M. Couchaux et al. / Journal of Constructional Steel Research 133 (2017) 459–474 469

Fig. 17. Geometric parameters of T-stub.

To further simplify the calculation method, a bi-linear approxima- one obtains:

tion of the curve l(ξ) (see Eq. (31)) is suggested and presented in Fig. 16:

lðξÞ ¼ min½2ðe2 −ξÞ=3; 0; 74t f  ð61Þ ðm þ ξÞ=2 þ lðξÞ

B ¼ FT 2
ð65Þ
ξ
lðξÞ þ ξ−
Remember that the distance between the bolt axis and the point of 2ðξ þ mÞ
application of the prying force is:

n ¼ ξ þ lðξÞ ð62Þ Neglecting shear deformations in Eq. (45), one has:

Finally, considering Eqs. (61) and (62), it is found that the distance n  
can be computed via the following relationship: 3 m2 2 1 m2 lðξÞ m þ lðξÞ m
−ξ þξ − þξ þ ðm=2 þ lðξÞÞ ¼ 0 ð66Þ
12EI f 2kb 4EI f kb kb
n ¼ min½ð2e2 þ ξÞ=3; ξ þ 0; 74t f  ð63Þ

Again, neglecting the flexural stiffness of the tube-wall, the expres- Eq. (66) is still nonlinear. Assuming that l(ξ) is equal to zero gives:
sion of bolt force takes the following form:

e1 þ n m3 kb ξ
B ¼ FT ð64Þ λ3 −λ2 −2λ−1 ¼ 0 with λ ¼ ð67Þ
n 6EI f m

2.7. Application to T-stubs and simplifications The above equation has been previously suggested by Kato and Ta-
naka [8]. The solution of Eq. (67) is:
In the case of a T-stub, (see Fig. 17) the flange is clamped at a dis-
tance m of the bolt, defined in EN 1993-1-8, for which a fully fixed
tube end can be simulated by considering kθ goes to infinity. Consider- 1 2ð1 þ 1=ð6α R ÞÞ
λ ¼ γT þ 1þ ð68Þ
ing expression (43) and taking the limit for B as kθ approaches infinity, 3α R γT

where
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
qffiffiffiffiffiffiffiffiffiffiffi
3 12þ81α R
72α R þ108α 2R þ8þ12α 2R αR
3
α R ¼ m6EIkfb and γT ¼ 6α R :

Fig. 18. Geometry and meshing of half an L-stub connection. Fig. 19. Contact pressure distribution.
470 M. Couchaux et al. / Journal of Constructional Steel Research 133 (2017) 459–474

Table 1 configurations have been studied considering L-stubs and T-stubs. The
Geometry of joints (dimensions in mm). joint were loaded by a tensile force of 5 kN.
Joint e2 e1 pb L tt Bolt Configuration

1 50 45 80 150 5 M16 L and T 3.2.2. Evolution of the contact stress

2 50 45 80 150 5 M20 L It has been demonstrated analytically (see Section 2.3) that the posi-
3 50 45 80 150 5 M30 L
tion of the prying force depends on the ratio of the stiffness of the bolt to
4 50 45 160 150 5 M20 L
5 35 35 80 150 5 M20 L
the stiffness of the flange and thus on the thickness of the flange. For the
6 70 66 122 250 12 M30 L and T same value of the externally applied to the L-stub, the contact stresses
7 80 66 95 250 18 M36 L and T obtained for different thicknesses of the flange of joint 7 are presented
8 40 41 120 200 8 M20 L in Fig. 20. The dark blue area corresponds to the non-contact region.
For thicker flanges, 35 and 40 mm, the contact area is located near
the outer edge of the flange. The length n tends to e2. For thinner flange,
3. Numerical analysis the contact area is located near the bolt head, thus the length n de-
creases and the bolt force increases. For thinner flanges, the latter ap-
3.1. Finite element model pears to be clamped at the uplift point (see Fig. 21).
Furthermore, the width of the contact area decreases for thinner
The numerical analyses were carried out using the Finite element flange whereas for thicker flanges (35 and 40 mm), the contact area
code ANSYS V11.0. Connections are generated with three dimensional spreads across the width of the flange. This phenomenon is not consid-
elements, which are hexahedral bricks. A linear elastic model is adopted ered in the analytical model which is essentially a planar model where
for steel (E = 210 GPa, ν = 0.3). The dimensions of the bolt fulfill EN the width of the contact area is supposed equal to the width of the
product requirements for HR bolts. A constant cross-section is consid- flange whatever the thickness of the flange. For a flange of 25 mm thick-
ered over the bolt length considering the effective cross area of ness, the outer edge lifts-off.
Eurocode 3 [2]. In the case of a flange connection, only half the model
needs to be represented due to symmetry. For the T-stub, only a quarter
4. Comparison between analytical and numerical results
of the connection is required to be modeled. Two types of contact ele-
ments are also used: a) flexible contact elements between the flange
The results obtained via the numerical model (see Section 3) are
and the bolt head and b) rigid contact elements between the flange
compared against those given by the analytical model and the calcula-
and the fictitious flange. An isotropic Coulomb friction law (μ = 0.25)
tion method developed for L-stubs and T-stubs. Our attention will
is used to reproduce sliding/sticking conditions between the flange
focus on the evaluation of the size of the contact area, the contact stress,
and the bolt head. Friction is neglected between the two flanges because
the position of the prying force and the bolt force.
of the symmetry (Figs. 18 and 19).

4.1. Analysis of L-stubs

3.2. Parametric study
4.1.1. Evolution of the contact pressure
3.2.1. Geometry of the joint It is possible to determine the contact stress in the contact area via
A parametric study has been carried out to test the validity of the the analytical model using Eq. (32). The slope of the contact pressure
models proposed in Section 2. Eight types of joints have been studied distribution will be computed for joint 7 via these analytical and numer-
(see Table 1) by modifying the thickness of the flange. Joints 6 and 7 ical models along inner and outer lines of the L-stub (see Fig. 22).
have been studied by Seidel [1]. The objective of this study is to evaluate The resultant of the contact stress, which is in fact the prying force,
the influence of the diameter of the bolt, the thickness of the flange, the calculated numerically and analytically, named respectively Qnum and
pitches e1 and e2 and the width of the flange. Two types of Qana are placed at their point of application (see Fig. 23).

a) tf = 40 mm b) tf = 35 mm

c) tf = 30 mm d) tf = 25 mm

Fig. 20. Joint 7 – Contact stresses on the rigid surface (in N/mm2). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
 M. Couchaux et al. / Journal of Constructional Steel Research 133 (2017) 459–474 471

a) tf = 40 mm b) tf = 35 mm
Fig. 21. Joint 7 – Deformed shape of the flange.

As can be seen, the analytical model provides a quite good estima- For all the configurations numerically investigated, prying effect de-
tion of the slope of the contact pressure distribution for two types of velops. For L-stubs prying effect develops only if the following inequal-
thickness. Furthermore, the resultant of contact pressure, the prying ity holds:
force, calculated numerically and analytically are almost identical. The
" #
length of the contact area obtained analytically is shorter than those ob- e1 L e21
tained numerically resulting in larger contact pressure. The error in the Lb ≤2e2 As þ ð69Þ
3I t 2I f
estimation of the contact stress does not result in an error in the evalu-
ation of global mechanical characteristics of the joint such as the prying
force or the bolt force. This condition is fulfilled by all configurations analyzed.

4.1.3. Bolt force

4.1.2. Position of the prying force The ratio η between the bolt force B and the tensile force FT is evalu-
The evolution of the length n, calculated via the analytical model, the ated via the analytical model, the calculation method and the numerical
calculation method and the numerical model are presented as a func-
tion of the flange thickness for joints 1, 2, 3 and 8 in Fig. 24. σc (N/mm2)
4
Numerical and analytical results are in good agreement. The differ-
Analytical Numerical (outer line) Numerical (inner line)
ence between the numerical and analytical results increases with de-
creasing value of the flange thickness. This difference is probably due
3 Qana=4,1 kN
to the fact that the force applied by the bolt to the flange is not pointwise Qnum=4,3 kN
as assumed in the analytical model.
For a thick flange, the prying force is applied near the free edge of the
2
flange, thus n ≈ e2. The theoretical (analytical model and calculation
method) results obtained for joint 3 are less good. This difference is
probably due to the fact that, for a M30 bolt, the hypothesis of a concen-
trated force applied by the bolt to the flange is not good enough in this 1

case. An improvement could be made by considering a force applied

continuously by the washer.
The ratio n calculated theoretically and numerically for 62 configura- 0
0 10 20 30 x (mm) 40
tions is presented in Fig. 25 as a function of the flange thickness. The cal-
culation method underestimates the value of n when the flange is thin. a) tf = 35mm
The results obtained numerically and theoretically are quite similar
σc (N/mm2)
when the thickness of the flange is important because the prying force 2.5
is applied near the free edge of the flanges. Analytical Numerical (outer line) Numerical (inner line)

2.0
Qana=4,7 kN Qnum=4,7 kN

1.5

1.0

0.5

0.0
0 10 20 30 40 x (mm) 50
b) tf = 30mm
Fig. 22. Outer and inner lines of a half L-stub. Fig. 23. Evolution of the contact pressure: Joint 7.
472 M. Couchaux et al. / Journal of Constructional Steel Research 133 (2017) 459–474

1.2 1,2
n/e2 n/e2

1.0 1,0

0.8 0,8

0.6 0,6

0.4 0,4

Analytical Calculation Method Numerical Analytical Calculation Method Numerical

0.2 0,2

tf (mm) tf (mm)
0.0 0,0
5 15 25 35 5 15 25 35

a) Joint 1 b) Joint 2
1,2 1,2
n/e2 n/e2

1,0 1,0

0,8 0,8

0,6 0,6

0,4 0,4
Analytical Calculation Method Numerical
Analytical Calculation Method Numerical
0,2 0,2
tf (mm) tf (mm)
0,0 0,0
5 15 25 35 5 15 25 35

c) Joint 3 d) Joint 8
Fig. 24. Ratio n/e2 versus flange thickness.

model. The results obtained for joints 1, 2, 3 and 8 are presented in different from numerical ones for thinner flanges. Again, the model as-
Fig. 26 as a function of the flange thickness. The ratios calculated theo- sumptions are not enough representative of the real behaviour of the
retically are generally larger comparatively to those calculated numeri- joint since the bolt doesn't exert a concentrated force on the flange.
cally, which may be safe for a design check of the static or fatigue The contact area can thus be located behind the bolt head.
resistance of the bolt. The results obtained via the analytical model are
in good agreement with those obtained numerically. The calculation 4.2. Analysis of T-stubs
method generally overestimates this ratio for thin flanges. The results
obtained for the calculation of the length n of joint 3 were not so As in Section 4.1, joints 1, 6 and 7 are analyzed considering T-stubs.
good, which is not the case for the ratio η. The ratio between the analytical values of n and those computed nu-
The ratio between the theoretical estimations of η and the numerical merically for 24 configurations is presented in Fig. 28 as a function of
ones is presented in Fig. 27 as a function of the flange thickness. Both the the flange thickness. The ratio between the analytical estimates of η
analytical model and the calculation method give results that are and the numerical ones is presented in Fig. 29 as a function of the flange
thickness. As can be seen, the analytical results are in good agreement
with the numerical one's.

5. Conclusion

In the present paper, a mechanical model for non-preloaded

bolted connections has been proposed. It takes into account contact
interactions between the opposite flanges. This model has proven
its ability to reproduce accurately the contact interactions which de-
velop between the flanges. The main ingredient of this model is the
enhanced beam theory proposed, by Baluch et al. [13], adapted by
Couchaux et al. [15] to study a beam in contact with rigid foundation
and used in the present paper to describe the mechanical response of
the flanges in both the contact and the non-contact regions. Over the
contact area, the enhanced beam theory gives accurate estimations
of the extent of the contact zone, the contact stress distribution be-
tween the flanges and therefore the location of the prying force
from which the bolt force can be evaluated. It was found that the
shape of the contact stresses distribution strongly depends on the
Fig. 25. Ratio of n calculated theoretically and numerically. ratio between the length of the contact area and the flange thickness.
 M. Couchaux et al. / Journal of Constructional Steel Research 133 (2017) 459–474 473

η = B/FT η = B/FT
5 5

Analytical Calculation Method Numerical

Analytical Calculation Method Numerical
4 4

3 3

2 2

1 1

tf (mm) tf (mm)
0 0
5 15 25 35 5 15 25 35

a) Joint 1 b) Joint 2
η = B/FT η = B/FT
5 4

Analytical Calculation Method Numerical Analytical Calculation Method Numerical

4
3

1
1

tf (mm) tf (mm)
0 0
5 15 25 35 5 15 25 35

c) Joint 3 d) Joint 8
Fig. 26. Evolution of the ratio η with the flange thickness.

Furthermore, the length of the contact zone is independent from the

magnitude of the external force applied to the tube and is bounded
by 2.42 times the flange thickness. Several simplifications have
been proposed to facilitate the determination of the separation
length which is the primary unknown of the problem.
To validate the analytical and calculation models, a 3D finite ele-
ment model has been developed. Contact elements were used be-
tween the flange and the bolt and rigid areas. Analytical and
numerical results are in good agreement which demonstrate the va-
lidity of the proposed mechanical model. As expected, simplified cal-
culation methods are less accurate. The theoretical and numerical
results confirm that the prying action increases with decreasing
values of the ratio between the flange stiffness and the bolt stiffness.
Fig. 27. Ratio between η calculated theoretically and η computed numerically versus the
flange thickness.

Fig. 28. Ratio between n calculated analytically and n computed numerically versus the Fig. 29. Ratio between η calculated analytically and η computed numerically versus the
flange thickness for T-stub. flange thickness for T-stub.
474 M. Couchaux et al. / Journal of Constructional Steel Research 133 (2017) 459–474

In the case of a thicker flange, the prying force is concentrated at the elastic behaviour of bolted connections. Bolt preloading can be
free edge of the flange. While in this paper the influence of bolt taken into consideration in our model [14] and further work is
preloading has been neglected, it has a significant influence on the under way to investigate it.

Appendix A. Coefficients Ci


2 a0 ς 2 2
ea0 ς a0 sinðb0 ς Þ−2a0 b0 cosðb0 ς Þ−b0 sinðb0 ς Þ þ ðe Þ −a0 sinðb0 ς Þ−2a0 b0 cosðb0 ςÞ þ b0 sinðb0 ς Þ
2 2
C 1 ðςÞ ¼ 
2 a0 ðea0 ς Þ4 b0 −4a0 sinðb0 ς Þ cosðb0 ς Þðea0 ς Þ2 −b0


2 3 a0 ς 2 3
ea0 ς a0 cosðb0 ςÞb0 þ 2a0 b0 sinðb0 ς Þ−b0 cosðb0 ς Þ þ ðe Þ 2a0 sinðb0 ς Þ þ 3a0 b0 cosðb0 ς Þ þ b0 cosðb0 ςÞ
2 3 2
C 2 ðςÞ ¼ − 
2 a0 b0 ðea0 ς Þ4 b0 −4a0 sinðb0 ς Þ cosðb0 ς Þðea0 ς Þ2 −b0


2 a0 ς 2 2
ea0 ς a0 sinðb0 ςÞ−2a0 b0 cosðb0 ς Þ−b0 sinðb0 ς Þ þ ðe Þ −a0 sinðb0 ς Þ−2a0 b0 cosðb0 ς Þ þ b0 sinðb0 ς Þ
2 2
C 3 ðςÞ ¼ − 
2 a0 ðea0 ς Þ4 b0 −4a0 sinðb0 ς Þ cosðb0 ς Þðea0 ς Þ2 −b0


3 a0 ς 2 2 3
ea0 ς −3a0 cosðb0 ς Þb0 þ 2a0 sinðb0 ς Þ−b0 cosðb0 ς Þ þ ðe Þ 2a0 b0 sinðb0 ς Þ−a0 b0 cosðb0 ς Þ þ b0 cosðb0 ς Þ
2 3 2
C 4 ðςÞ ¼ 
2 a0 b0 ðea0 ς Þ4 b0 −4a0 sinðb0 ς Þ cosðb0 ς Þðea0 ς Þ2 −b0

ς ¼ e2 −ξ

References [10] H. Senda, T. Suzuki, T. Ogawa, Inelastic behaviour of bolted T-stub connections, Jour-
nal of structural and construction engineering, Trans. AIJ 476 (1996) 159–168 (in
[1] M. Seidel, Zur Bemessung geschraubter Ringflanschverbindungen von Japanese).
Windenergieanlangen, Institut fur Stahlbau(Dissertation, Heft 20) Universität Han- [11] E. Bakhiet, Etude des assemblages boulonnés à chargement fortement excentré
nover, 2001. soumis à des sollicitations de fatigue, Thèse N°319, INSA de Toulouse, 1994 (in
[2] EN 1993-1-8: Eurocode 3, Design of steel structures – Part 1–8: Designs of Joints, French).
May 2005. [12] J. Chakhari, Modélisation d'une fixation par éléments filetés d'une structure à forte
[3] J.P. Jaspart, Etude de la semi-rigidité des nœuds poutre-colonne et son influence sur excentration de chargement et soumise à des sollicitations en fatigue, 2007 (Thèse
la résistance et la stabilité des ossatures en acier, Thèse de doctorat en science de doctorat de l'INSAT et de l'ENIT (in French)).
appliquée, Université de Liège, 1990 (in French). [13] M.H. Baluch, A.K. Azad, M.A. Khidir, Technical Theory of Beams with normal strain, J.
[4] V. Piluso, C. Faella, G. Rizzano, Ultimate Behavior of Bolted T-stubs II: Model Valida- Eng. Mech. 110 (8) (1984) 1233–1237.
tion, J. Struct. Eng. 127 (6) (2001) 694–703. [14] M. Couchaux, Behaviour of Bolted Circular Flange Joints PhD Thesis INSA of Rennes,
[5] A.M. Girao Coehlo, Characterization of Ductility of Bolted End Plate Beam-to-column 2010 (in French).
Steel Connections(PhD dissertation) University of Coimbra, July 2004. [15] M. Couchaux, M. Hjiaj, I. Ryan, Enriched beam model for slender prismatic solids in
[6] C. Petersen, Stahlbau (Steel Construction), Vieweg-Verlag, Braunschweig, 1988. contact with a rigid foundation, Int. J. Mech. Sci. 93 (2015) 181–190.
[7] P. Agatonovic, Beam model of bolted flanged connection, Eng. Comput. 2 (1985) [16] G.G. Adams, D.B. Bogy, J. Appl. Mech. Trans. ASME E 42 (4) (1975) 800–804.
21–29. [17] G.G. Adams, Moving load on elastic strips with one-sided constraints, Int. J. Eng. Sci.
[8] B. Kato, A. Tanaka, Experimental study on tension-type high strength bolted connec- 14 (1976) 1071–1083.
tion (no. 2 prying action), Trans. Architect. Inst. Japan 72 (147) (1968) 33–41 (in [18] L.M. Keer, M.A.G. Silva, Two mixed problems for a semi-infinite layer, J. Appl. Mech.
Japanese). (1972) 1121–1124.
[9] M.E. Lemonis, C.J. Gantes, Incremental modelling of T-stub connections, J. Mech. [19] S.P. Timoshenko, S. Woinowsky-Krieger, Theory of Plates and Shells, second ed.
Mater. Struct. 1 (7) (2006) 1135–1157. McGraw-Hill International Editions engineering Mechanics Edition, 1959.