Engineering Structures 41 (2012) 548561

Contents lists available at SciVerse ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Effective length of aluminium T-stub connections by parametric analysis

Gianfranco De Matteis , Muhammad Tayyab Naqash, Giuseppe Brando
Department of Engineering, University G. dAnnunzio of Chieti-Pescara, Viale Pindaro 42, 65127 Pescara, Italy

a r t i c l e

i n f o

Article history:
Received 5 December 2011
Revised 9 March 2012
Accepted 26 March 2012
Available online 18 May 2012
Keywords:
Aluminium T-stub
Finite Element model
Parametric analysis
Component method
Eurocode 9
Effective length

a b s t r a c t
The paper presents a parametric analysis carried out on welded aluminium T-stubs by means of Finite
Element models. The applied models are suitably calibrated on the basis of available experimental tests.
The study is carried out on a large variety of specimens with different features and different type of bolts,
in order to analyse all possible failure mechanisms. Totally, 43 models are analysed and the obtained
results are carefully elaborated in order to check the reliability of the methods presently provided by
Eurocode 9. The paper represents a signicant extension of the experimental and numerical analyses carried out by the authors in the past, which were especially devoted to analyse the denition of effective
length for aluminium T-stubs. The obtained results allow to yield interesting outcomes that should be
incorporated in future editions of relevant codes dealing with aluminium structures.
2012 Elsevier Ltd. All rights reserved.

1. Introduction
It is well known, that some aluminium alloys exhibit special
properties, such as corrosion resistance, versatility, reversibility,
reasonable ductility and lightness, especially if compared to other
conventional materials like steel. Consequently, they highly attract
designers for their employment in building projects [1].
Nevertheless, only few studies have been undertaken in the past
for the identication of the behaviour of aluminium connections
and joints for structural purposes. This leads to state that the current
Eurocode 9 [2] approach on joints design is generally not complete
enough for formulating a reliable component method, contrarily
to the one already recommended by Eurocode 3 [3] for steel joints,
this being based on the outcomes of several researches and specic
studies carried out in the last four decades.
The paramount role of T-stub in the component method formulation for dening both strength and stiffness of joints is widely
recognised. It is a typical component of bolted joints used to model
column ange in bending, end plate in bending and ange cleats in
bending, etc.
The so-called T-stub consists of two T-section elements, symmetrically connected to each other in their anges by one or more
series of bolt rows, which undergo exural deformations due to a
pulling force usually transmitted by webs transversally located at
the centre of the anges (see Fig. 1).
The T-stub behaviour is governed by various phenomena,
namely the bolts strength and deformability, the exural stiffness
Corresponding author. Tel.: +39 0854537261; fax: +39 0854537255.
E-mail address: demattei@unina.it (G. De Matteis).
0141-0296/$ - see front matter 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.engstruct.2012.03.052

of the ange, the geometrical properties that can entail different

yield lines on the connected plates when incipient collapse phenomena involve the whole system, etc.
In the last years, many researches have been devoted to enhance
the knowledge on T-stub connections. These have been developed
by means of experimental, analytical and numerical analyses. Interesting experimental tests have been provided by Girao Coelho et al.
[4], who dealt with extended end plate connections for determining
the inuence of both material grade and plate thickness. De Matteis
et al. [5,6] investigated for the rst time the possibility of extending
the provision for T-stub given by Eurocode 3 also to aluminium
joints. Piluso and Rizzano [7] performed experimental analyses on
bolted steel T-stubs under cyclic loads. Moreover, theoretical models have been provided by Lemonis and Gante [8] and Stamatopoulos and Ermopoulos [9], who investigated the inuence of the Tstub exibility. Also, numerical analyses have been developed by
Mistakidis et al. [10], who proposed a computationally non-cumbersome 2-D numerical FEM model, by Giro Coelho et al. [11],
who dealt with both rolled and welded T-stub models, by Efthymiou [12], De Matteis et al. [13] and Xu et al. [14].
With particular regard to aluminium connections, the current
version of Eurocode 9 provides formulations based on the k-method which has been proposed by the authors. The current paper represents an extension with respect to the previous researches, which
were based on the assumption of the same effective length for
aluminium T-stubs as provided by EC3-Part 1.8 for steel T-stub.
Hence, a parametric study implemented on the basis of FEM numerical aluminium T-stub models is provided in order to identify the effects of the most important geometrical and mechanical
parameters on failure modes, yield patterns and, therefore, on the

549

G. De Matteis et al. / Engineering Structures 41 (2012) 548561

F u;Rd;2a

2M u;2 n
mn

2M 0;2 n
mn
X

Bu

P
P

F u;Rd;2b
F u;Rd;3

B0

Bu

3
4

The actual collapse load of the T-stub joint is determined from the
minimum value of the resisting forces governing the failure modes.
In the above equations, Bu and Bo are the ultimate and conventional elastic tensile strength of bolts, respectively. (Mu,1)w and
(Mu,1)b are the plastic moments of the critical ange cross sections,
located close to the T-stub web and bolt rows, respectively, when a
failure mode 1 arises (Eqs. (5) and (6)). Mu,2 (Eq. (7)) is the plastic
moment of the ange when the failure type is mode 2. M0,2 (Eq.
(8)) is the elastic moment at 0.2% proof strength.

M u;1 w 0:25  t 2f  Reff ;u;1  f0;haz 

Fig. 1. Idealization and schematization of T-stub.

actual effective length which is able to return the T-stub ultimate

strength according to the simplied formulation presently adopted
in EC9.
2. Aluminium T-stub failure mechanisms (EC9 k-method)
T-stub connections must be designed taking into account all the
possible yielding mechanisms and failure modes. For aluminium
connections, EC9 (Annex-k) proposes the k-method for the calculation of ultimate strength of T-stub connections. This method also
accounts for strain hardening and reduced ductility of the base
material.
Four failure mechanisms are detected, namely mode 1 complete ange failure, mode 2a partial ange failure with the
attainment of elastic strength in bolts, mode 2b bolts failure
with the attainment of the elastic strength in anges and mode
3 which consists in the complete bolts failure (see Fig. 2).
The ultimate resistance of the T-stub failure modes 1, 2a, 2b and
3 are given in Eqs. (1)(4), respectively.

F u;Rd;1

2M u;1 w 2Mu;1 b

M u;1 b 0:25  t2f  Reff ;u;1  fu 

1
k

1
k

Mu;2 0:25  t2f  Reff ;u;2  fu 

1
k

M0:2 0:25  t 2f  eff ;u;2  f0:2 

1
k

In Eqs. (5)(8), eff,u,1 and eff,u,2 are the ange section effective
lengths, dened according to the failure mode and the yield line
developing (circular or non-circular pattern), f0.2 and fu are the conventional yield and ultimate stress, respectively of the base material, f0,haz is the ultimate strength of the heat affected zone, tf is
the ange thickness, m is the distance of the weld seams from the
centre of bolts, n is the minimum between 1.25m and the distance
e of bolts from the ange edges (see Fig. 2).
The k factor is dened as:



1 f0:2
fu  f0:2
 1w

k
fu
f0:2

where

eu  1:5  e0:2
1:5  eu  e0:2

Fig. 2. Failure modes of aluminium T-stubs prescribed by EC9.

10

550

G. De Matteis et al. / Engineering Structures 41 (2012) 548561

eu and e0.2 being the ultimate and yielding strain respectively.

The method given by EC9 properly extends the formulation proposed by Eurocode 3 for steel T-stub by using a correction factor
1/k, which allows to take into account both the different strain
hardening and the reduced ductility of aluminium materials with
respect to steel.
3. The numerical model
3.1. General
The base model used for parametrical analysis is calibrated on
the basis of available experimental tests carried out by De Matteis
and Mazzolani [6]. In particular, three welded coupons, subjected
to a monotonic pulling force up to the failure, are taken into
account. These are hereafter named as specimen Sample A, specimen Sample B and specimen Sample C.
The mechanical and geometrical features of tested specimens
are listed in Table 1, whereas the stressstrain relationship of the
related materials, including also the heat affected zone closed to
the welds, is provided in Section 3.4.
It is worthy to note that one material only, namely Aluminium
alloy AW-6082, is considered for the T-stub anges, as this parametric analysis is not devoted to analyse the effect of material
hardening.
3.2. Geometric modelling
T-stub connections are fairly complex to be modelled; in fact,
their geometry is generally three-dimensional, material and geometrical nonlinearities are strongly involved in the loading process
and different contact phenomena, due to the interaction between
anges and bolts, are present. Hence, compromises in the modelling phase are usually taken into consideration so to circumvent
these difculties.
In the case under consideration, in order to reduce the cumbersomeness of the analysis, the proposed geometry of the model (see
Fig. 3a) takes into account the T-stub symmetry. As a consequence,
a rigid body xed in the space is put below the ange of one of the
two T-sections so to simulate the presence of the other part of the
specimen. For the same reason, only half of the bolt is modelled,
with the middle plane of the shank suitably restrained.
In order to take into consideration the reduction of the diameter
of the bolts due to the threaded part, a 20% reduced area with respect to the nominal one is considered. Indeed, for the adopted
M10 bolt, the existing codes provide a reduction of almost 25%,
but in the proposed model a slightly higher resistant area of bolts
is taken into account as the threaded portion is assumed to contribute to the bolt stiffness.
The T-stub model is implemented by the Code ABAQUS 6.7 [15],
where 8 node linear brick elements with reduced integration and
hourglass control (C3D8R) are used (see Fig. 3b) for webs and
anges. The bolts are meshed with 4 node linear tetrahedron
(C3D4) elements (see Fig. 3b). This is due to their more complex
geometry, which requires tetrahedral elements in order to realise
a less rened mesh without jeopardising the accuracy of the model
by contact problems. Hex dominant meshing algorithm is used to

Fig. 3. T-stub (a) geometry and (b) FEM model.

obtain meshes with a mix of hexahedral, pyramids and tetrahedral

nite elements.
3.3. Boundary conditions, loads and interactions
The T-stub web is pulled by imposing a uniform vertical displacement applied to a reference point constrained by a rigid coupling with the top of the web itself. On the other hand, the bottom
part of the rigid body below the T-stub ange is xed by applying a
clamping boundary condition.
The bolt load option provided by Abaqus has been used to
simulate the preload force Fp,Cd, accordingly to the following
equation:

F p;Cd 0:7  fub  As

11

where As is the resisting area of the bolt and fub is the ultimate stress
of the bolt material.
This bolt load option allows to automatically adjust the length
of the bolts in order to achieve the prescribed amount of pretension.
Three contact interactions are dened, namely (i) the bearing of
the back of the T-stub section against the interface with the rigid
body, (ii) the interaction between the hole and the bolt shank
and (iii) the interaction between the bolts head and the surface
of the T-stub. The rst is dened as a penalty contact characterised
by a friction coefcient of 0.3, whereas the others are taken into account as frictionless contact.
3.4. Material modelling
The material constants used for all aluminium parts are
E = 70,000 N/mm2 (elastic modulus), m = 0.3 (Poisson ratio) and
q = 2700 kg/m3 (material density), whereas steel bolts are modelled with E = 210,000 N/mm2, m = 0.3 and q = 7600 kg/m3. The
proof strengths for all the aluminium components of the T-stub
are referred to a conventional stress of f0.2.
In order to interpret correctly the behaviour of the system also
for large deformation, the available material test data are properly
transformed in true stresstrue strain, as depicted in Fig. 4, where
the experimental curves are also provided for all the parts of the
tested specimens.

Table 1
T-stub tested specimens used for calibration of the proposed models.
Sample ID

Aluminium alloy for ange

Aluminium alloy for web

Bolt material

Flange thickness (mm)

Web thickness (mm)

Bolt diameter (mm)

Sample A
Sample B
Sample C

6082
6082
6082

7020
7020
7020

4.8
7075
10.9

10
10
10

12
12
12

10
10
10

551

G. De Matteis et al. / Engineering Structures 41 (2012) 548561

Table 2
Mesh size of T-stub model and related number of elements and nodes.

(a)

(b)

(c)

(d)

(e)

Approximate global mesh

size (mm)

No. of
nodes

No. of elements
[C3D8R]

Normalised
CPU time

3
4
5

12,738
8412
4271

9700
6344
2976

3.1
2.7
1.0

(a)

(b)

(f)

Fig. 4. Nominal and true stress/strain curves for (a) ange, (b) web, (c) HAZ, (d) 4.8
steel bolts, (e) 7075 aluminium bolts, and (f) 10.9 steel bolts.

(c)
Fig. 6. Sensitivity analysis for (a) T-stub mesh, (b) bolts mesh, and (c) contacts.

3.5. Analysis implementation

The analysis of the models is carried out by using standard multiple step (two steps) analysis. In the rst step, the bolt preload is
statically applied by means of the bolt load option, provided in
the ABAQUS library. In the second step, a static Riks analysis is applied up to the collapse of the whole system in order to reproduce
the loading process of the whole T-stub.

3.6. FEM sensitivity analysis

A preliminary sensitivity analysis is carried out on specimen
Sample A, in order to check the inuence of both mesh size, nite
element adoptions and contact typology on the proposed model response. To this purpose, in a rst stage, T-stub models with

Fig. 5. Sensitivity analysis for mesh size (a) 3 mm, (b) 4 mm, and (c) 5 mm.

552

G. De Matteis et al. / Engineering Structures 41 (2012) 548561

Table 3
Mesh size for bolt model and related number of elements and nodes.
Approximate global
mesh size (mm)

Element
type

No. of
nodes

No. of
elements

Normalised
CPU time

1
2
3
4
3
1

C3D4
C3D4
C3D4
C3D4
C3D8R
C3D8R

3136
612
249
166
372
5595

13,336
2303
860
525
248
4664

8.7
6.0
4.1
2.2
1.0
7.1
Fig. 8. Deformed shapes and stress contour of (a) FEM Sample A, (b) FEM Sample B,
and (c) FEM Sample C.

Table 4
Different contact combinations for sensitivity analysis.
Combination

T-stub to base

T-stub to bolts

Normalised CPU time

1
2
3
4

Penalty-0.3
Penalty-0.3
Penalty-0.1
Penalty-0.2

Frictionless
Rough
Frictionless
Frictionless

1.3
1.3
1.2
1.0

(a)

(b)

Fig. 7. Deformed shapes of (a) test Sample A, (b) test Sample B, and (c) test Sample
C.

approximate global mesh size of 3 mm, 4 mm and 5 mm (see

Fig. 5ac) are analysed. The obtained numbers of elements, as well
as the corresponding CPU time consuming (normalised to minimum value) are provided in Table 2.
From the analysis of the obtained response, given in Fig. 6a in
terms of pulling force F vs. vertical displacement D, it is possible to observe that the selected mesh sizes do not inuence significantly the overall result. However, since using a global mesh size
of 4 mm requires sustainable time analysis consuming, giving precisely the same results of a 3 mm mesh side length, the former is
assumed for all the implemented models.
The same type of remarks can be made for bolts nite elements
typology and size. With respect to the options listed in Table 3, it is
possible to observe that using C3D4 elements with an average
mesh size of 3 mm gives the same results of C3D8R elements with
1 mm mesh side length, requiring half of the CPU time. For this reason, the former mesh typology is adopted (see Fig. 6b).
Finally, the sensitivity of the model with respect to the four different contacts combination listed in Table 4 is investigated. All the
possible combinations need more or less the same normalised CPU
time, giving approximately the same results for small displacements. Nevertheless, contacts combination 1, with penalty coefcient of 0.3 for the T-stub to base and frictionless contacts for
bolt to T-stub, is assumed, as it gives less convergence problems
for larger displacements.

(c)
Fig. 9. Experimental vs. numerical results for: (a) Sample A, (b) Sample B, and (c)
Sample C.

Sample C failure modes 2b and 1 are recognised, respectively.

The same failure modes are evidenced by numerical model (Fig. 8).
On the other hand, the proposed numerical results are in good
agreement with the experimental ones if compared in terms of
force vs. displacement curves (Fig. 9).
The obtained models are therefore reliable enough to allow for
carrying out a parametric study.

4. Analysis of results and discussion

4.1. General

3.7. Numerical vs. experimental results

The reliability of the proposed models is proved by comparing
the experimental outcomes with the numerical ones. As shown
in Fig. 7a, in case of experimental specimen Sample A, a failure
mode 2a is detected, whereas for specimens Sample B and

In the following, the main results of a parametrical analysis, carried out by changing both the mechanical and geometrical properties of the above calibrated models, are described. Totally, 43
different geometries are managed, with the aims of proving the
reliability of the formulations currently provided by EC9.

553

G. De Matteis et al. / Engineering Structures 41 (2012) 548561

Table 5
Mechanical features of the materials used in parametric analyses.

eu (%)

Part

Material

f0 (Mpa)

fu (Mpa)

E (Mpa)

Poissons ratio

Flange/web

EN AW-6082 T6151
EN AW-6082 T6151 (HAZ)

200
125

275
185

12
12

70,000
70,000

0.3
0.3

Bolts

Grade 4.6
EN AW-6082a
Grade 10.9

240
260
900

400
310
1000

16
14
7

210,000
70,000
210,000

0.3
0.3
0.3

No temper grade designation is given in Eurocode for this material when it is used for bolts.

Table 6
Values of geometrical parameters (mm) for analysed T-stubs.
Group

S. no.

Label

tf

tw

e = e1

Bolts type

1
2
3
4
5
6
7
8
9
10
11
12

40p8-a
40p10-a
40p12-a
60p8-a
60p10-a
60p12-a
80p8-a
80p10-a
80p12-a
120p8-a
120p10-a
120p12-a

8
10
12
8
10
12
8
10
12
8
10
12

140
142
144
160
162
164
180
182
184
220
222
224

80
80
80
120
120
120
160
160
160
240
240
240

40
40
40
40
40
40
40
40
40
40
40
40

10
12
14
10
12
14
10
12
14
10
12
14

40
40
40
60
60
60
80
80
80
120
120
120

20
20
20
30
30
30
40
40
40
60
60
60

11
11
11
11
11
11
11
11
11
11
11
11

4.6
4.6
4.6
4.6
4.6
4.6
4.6
4.6
4.6
4.6
4.6
4.6

13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31

40p8-c
40p10-c
40p12-c
40p15-c
60p8-c
60p10-c
60p12-c
60p15-c
80p8-c
80p10-c
80p12-c
80p15-c
120p8-c
120p10-c
120p12-c
120p15-c
150p10-c
150p12-c
150p15-c

8
10
12
15
8
10
12
15
8
10
12
15
8
10
12
15
10
12
15

140
142
144
147
160
162
164
167
180
182
184
187
220
222
224
227
252
254
257

80
80
80
80
120
120
120
120
160
160
160
160
240
240
240
240
300
300
300

40
40
40
40
40
40
40
40
40
40
40
40
40
40
40
40
40
40
40

10
12
14
17
10
12
14
17
10
12
14
17
10
12
14
17
12
14
17

40
40
40
40
60
60
60
60
80
80
80
80
120
120
120
120
150
150
150

20
20
20
20
30
30
30
30
40
40
40
40
60
60
60
60
75
75
75

11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11

10.9
10.9
10.9
10.9
10.9
10.9
10.9
10.9
10.9
10.9
10.9
10.9
10.9
10.9
10.9
10.9
10.9
10.9
10.9

32
33
34
35
36
37
38
39
40
41
42
43

40p8-b
40p10-b
40p12-b
60p8-b
60p10-b
60p12-b
80p8-b
80p10-b
80p12-b
120p8-b
120p10-b
120p12-b

8
10
12
8
10
12
8
10
12
8
10
12

140
142
144
160
162
164
180
182
184
220
222
224

80
80
80
120
120
120
160
160
160
240
240
240

40
40
40
40
40
40
40
40
40
40
40
40

10
12
14
10
12
14
10
12
14
10
12
14

40
40
40
60
60
60
80
80
80
120
120
120

20
20
20
30
30
30
40
40
40
60
60
60

11
11
11
11
11
11
11
11
11
11
11
11

AW-6082
AW-6082
AW-6082
AW-6082
AW-6082
AW-6082
AW-6082
AW-6082
AW-6082
AW-6082
AW-6082
AW-6082

In all cases the nominal values suggested by EC9 for the

mechanical features of the used materials, as listed in Table 5,
are considered.
In particular, for aluminium parts the stress strain curves are
dened using a Ramberg Osgood law, as expressed in Eq. (12),
which is properly transformed in terms of true stressestrue
strains.

r rn
e k
E

12

In the above equation, e is the strain, r is the stress, E is the

Youngs modulus, k = 0.002 and n = 14 is the hardening constant
which depends on the material under consideration.

The material ultimate strain (eu) is evaluated based on the

approximate formulation given in EC9 Annex-E as represented by

eu 0:3  0:22

f0
400

when f 0 < 400 N=mm2

13

Models with two bolt rows and resulting by the adoption of four
thicknesses of the ange tf of the T-stub (8 mm, 10 mm, 12 mm and
15 mm), different bolt grades (Type a = grade 4.6, Type b = aluminium alloy 6082 and Type c = grade 10.9) and four geometries
that consider different bolt pitches p (40 mm, 60 mm, 80 mm and
120 mm) are proposed (see Table 6). In this table, the notations

554

G. De Matteis et al. / Engineering Structures 41 (2012) 548561

(a1)

(b1)

(a2)

(a3)

(b2)

(b3)

Fig. 10. Ultimate resistance vs. displacement graphs related to different bolt pitch: T-stub with 4.6 steel bolts: (a1) tf = 8 mm, (a2) tf = 10 mm, (a3) tf = 12 mm; T-stub with
6082 aluminium bolts: (b1) tf = 8 mm, (b2) tf = 10 mm, (b3) tf = 12 mm.

(a)

(b)

(c)

(d)

Fig. 11. Ultimate resistance vs. displacement graphs: for 10.9 bolt grades with different pitches (a) tf = 8 mm, (b) tf = 10 mm, (c) tf = 12 mm, (d) tf = 15 mm.

555

G. De Matteis et al. / Engineering Structures 41 (2012) 548561

tf, B, b, m, tw and are thickness of ange, breadth of T-stub, width

of T-stub, bolt to web-face distance, thickness of web and diameter
of hole, respectively; these notations are also indicated in Fig. 12.
4.2. Numerical results

Fig. 12. Monitored points in the model for the identication of failure modes.

(a1)

In the following numerical results are presented in term of

graphs and tables. Fig. 10 shows the force vs. displacement diagrams for all the analysed models when bolt pitch varies and when
4.6 steel bolts (graphs Type a) and aluminium alloy AW-6082
bolts (graphs Type b) are assumed. Fig. 11 shows the same
force vs. displacement graphs in the case of 10.9 grade steel bolts,
also considering a ange thickness of 15 mm.

(a2)

(a3)

(b1)

(b2)

(b3)

(c1)

(c2)

(c3)

(d1)

(d2)

(d3)

Fig. 13. Failure mechanism identication for 4.6 steel bolts: (a1) tf = 8 mm, p40; (a2) tf = 10 mm, p40; (a3) tf = 12 mm, p40; (b1) tf = 8 mm, p60; (b2) tf = 10 mm, p60; (b3)
tf = 12 mm, p60; (c1) tf = 8 mm, p80; (c2) tf = 10 mm, p80; (c3) tf = 12 mm, p80; (d1) tf = 8 mm, p120; (d2) tf = 10 mm, p120; (d3) tf = 12 mm, p120.

556

G. De Matteis et al. / Engineering Structures 41 (2012) 548561

(a1)

(a2)

(a3)

(b1)

(b2)

(b3)

(c1)

(c2)

(c3)

(d1)

(d2)

(d3)

Fig. 14. Failure mechanism identication for 6082-alloy aluminium bolts: (a1) tf = 8 mm, p40; (a2) tf = 10 mm, p40; (a3) tf = 12 mm, p40; (b1) tf = 8 mm, p60; (b2) tf = 10 mm,
p60; (b3) tf = 12 mm, p60; (c1) tf = 8 mm, p80; (c2) tf = 10 mm, p80; (c3) tf = 12 mm, p80; (d1) tf = 8 mm, p120; (d2) tf = 10 mm, p120; (d3) tf = 12 mm, p120.

From such curves, the ultimate strength can be detected as the

smaller among the peak value of the related curve and the force
corresponding to the attainment of ultimate material deformation
in anyone of the points of the T-stub in which failure may arise. To
this purpose, points p1, p2, and p3 shown in Fig. 12 are monitored, in order to identify the failure mechanisms of the models in
the FEM analyses.

The failure mechanisms characterising FEM models with 4.6

steel bolts can be identied by the graphs given in Fig. 13. It is possible to retrieve the stresses of the parts in which failure may develop vs. the global displacement of the T-stub. In all the cases, a
failure mode 2a is detectable due to the weakness of the bolts
which, nevertheless, yield only after that the ange attains the
plastic moment in the HAZ.

557

G. De Matteis et al. / Engineering Structures 41 (2012) 548561

(a1)

(a2)

(a3)

(a4)

(b1)

(b2)

(b3)

(b4)

(c1)

(c2)

(c3)

(c4)

(d1)

(d2)

(d3)

(d4)

Fig. 15. Failure mechanism identication for 10.9 steel bolts: (a1) tf = 8 mm, p40; (a2) tf = 10 mm, p40; (a3) tf = 12 mm, p40; (a4) tf = 15 mm, p40; (b1) tf = 8 mm, p60; (b2)
tf = 10 mm, p60; (b3) tf = 12 mm, p60; (b4) tf = 15 mm, p60; (c1) tf = 8 mm, p80; (c2) tf = 10 mm, p80; (c3) tf = 12 mm, p80; (c4) tf = 15 mm, p80; (d1) tf = 8 mm, p120; (d2)
tf = 10 mm, p120; (d3) tf = 12 mm, p120; (d4) tf = 15 mm, p120.

of 10.9 bolts the failure of T-stub anges takes place, giving back a
failure mode 1, as the bolts are very far by the attainment of the
elastic limit.

Table 7
Failure modes for 12 mm and 15 mm thick anges using 10.9 steel bolts.
Specimen

40p12-c
60p12-c
80p12-c
120p12-c

tf = 12 mm

Specimen

EC9

FEM

1
1
1
2a

1
1
1
1

40p15-c
60p15-c
80p15-c
120p15-c

tf = 15 mm
EC9

FEM

1
2a
2a
2b

1
1
1
1

Figs. 14 and 15 show the stress vs. displacement diagrams for

AW-6082 aluminium and 10.9 steel bolts, respectively. In the rst
case, a failure mode 2b is always detectable, whereas in the case

4.3. Comparison with Eurocode 9 results

In order to check the validity of the method proposed by Eurocode 9, the above results obtained by FEM analyses are compared
with the ones provided by the k-method. To this purpose, the codied method is applied considering a material safety factor cM = 1.0.
It should be noted that EC9 formulation gives failure mode 2a
for all cases when 4.6 steel bolts are employed. These results
are also provided by stress analyses in Fig. 13. In addition, for
all the models with 6082 aluminium bolts, a failure mode 2b is

558

G. De Matteis et al. / Engineering Structures 41 (2012) 548561

(a)

(b)

(c)

(d)

Fig. 16. Ultimate strength (Fu) obtained from FEM analyses and EC9 formulation for analysed cases: (a) p = 40 mm, (b) p = 60 mm, (c) p = 80 mm, and (d) p = 120 mm.

(a)

(b)

(c)

(d)

Fig. 17. FEM vs. EC9 results in terms of tendency lines: (a) p = 40 mm, (b) p = 60 mm, (c) p = 80 mm, and (d) p = 120 mm.

559

G. De Matteis et al. / Engineering Structures 41 (2012) 548561

Table 8
Effective lengths for T-stub from FEM and EC9 using 10.9 steel bolts.
Pitch

Label

(p40)

40p8-c
40p10-c
40p12-c
40p15-c

(p60)

60p8-c
60p10-c
60p12-c
60p15-c

eff EC9 (mm)

eff EC9 (mm)

Pitch

Label

eff FEM (mm)

eff FEM (mm)

80
80
80
80

134.8
115.1
89.2
72.0

(p80)

80p8-c
80p10-c
80p12-c
80p16-c

160
160
160
160

218.7
168.8
126.5
96.0

120
120
120
120

179.8
147.7
118.5
83.0

(p120)

120p8-c
120p10-c
120p12-c
120p15-c

240
240
240
240

275.7
201.4
153.2
109

Fig. 18. Von Mises stress contours for the T-stub anges of models with steel 10.9 bolts: (a1) tf = 8 mm, p = 40 mm; (a2) tf = 10 mm, p = 40 mm; (a3) tf = 12 mm, p = 40 mm;
(b1) tf = 8 mm , p = 60 mm; (b2) tf = 10 mm , p = 60 mm; (b3) tf = 12 mm, p = 60 mm; (c1) tf = 8 mm, p = 80 mm; (c2) tf = 10 mm , p = 80 mm; (c3) tf = 12 mm, p = 80 mm; (d1)
tf = 8 mm , p = 120 mm; (d2) tf = 10 mm , p = 120 mm; (d3) tf = 12 mm, p = 120 mm.

identied (Fig. 14). This result is given also by EC9, except for
model 120p12-b, for which a failure mode 3 can be detected.
As far as the cases related to 10.9 bolts with 12 mm thick ange
are concerned (Fig. 15), a failure mode 1 is obtained from both
EC9 and FEM analysis, apart from model 120p12-c, for which a
mode 2a is obtained using EC9 formulations. On the contrary,
in the case of 15 mm thick ange, except model 40p15-c for
which mode 1 is obtained from both EC9 and FEM analysis, all
the rest of the failure mechanisms obtained from EC9 disagree with
the one obtained from FEM analysis, as shown in Table. 7.
Fig. 16 shows that EC9 formulation is in good agreement in
cases of low pitches (40 mm and 60 mm). On the contrary, it is evi-

dent that when high pitches (80 mm and 120 mm) are assumed,
the EC9 formulation overestimates the ultimate strength, it resulting therefore not conservative, especially for 12 mm and 15 mm
thick anges.
The overestimation of the ultimate strength for high thicknesses is probably due to the fact that, in these cases, the expression of the effective length by EC9 is not reliable enough.
The overestimation of ultimate strength is evident by the comparison given in Fig. 16. In addition it can be observed that FEM
analysis results provided in Fig. 17 are quite consistent in terms
of tendency lines for all the analysed thicknesses, whereas the
curves obtained from the EC9 formulation present a different rate

560

G. De Matteis et al. / Engineering Structures 41 (2012) 548561

(a)

(b)

Fig. 19. Effective lengths for 10.9 grade steel bolts from FEM and EC9 approach (a) tendency lines (b) limit curve for safe evaluation of effective length.

with increased pitch of bolts, when high strength bolts are used.
Reminding that FEM 120p12-b presents a failure mode 2b
and FEM 120p12-c model presents a mode 1 failure mode,
whereas EC9 gives back mode 3 and mode 2a respectively, it
is apparent that the formulation given by EC9 for the effective
length must be generally recalibrated, depending on the bolts pitch
and ange thicknesses.
4.4. Effective length evaluation
The above remarks about the effective length are actually conrmed by the fact that, as the bolts pitch increases, the inactive
area of the ange contributing to the ultimate strength increases
as well, with a proportionally shorter yielding pattern which tends
to a circular shape. On the contrary, it can be observed that for the
models with a pitch of 40 mm the entire ange transversal section
contributes to the resistance of the T-stub. This is clearly shown in
Fig. 18 where the stress contours for models according to the adoption of 10.9 steel bolts is depicted.
Similarly, it is quite evident that as the thickness of the ange
increases the effective area of the ange contributing to the
strength reduces. This indicates that as the thicknesses of the
ange of T-stub increases, the effective length consequently should
decrease as well.
With reference to the code formulation, it has to be underlined
that for all the analysed cases, EC9 imposes a non-circular pattern
for the calculation of the effective lengths. This seems contradictory to the above results, proving again that EC9 rule for selecting
the effective length leff is not reliable for certain geometric congurations and needs to be revised.
In Table 8, the effective length leff computed according to the
ultimate strength obtained by both FEM analysis and EC9 formulation are listed for all the models with 10.9 steel bolts. These are
plotted in Fig. 19a in terms of tendency lines. In detail, the effective lengths from FEM analyses are obtained by using the numerical ultimate strength and by reversing the Eurocode 9
formulation. This allows the evaluation of discrepancies present
in EC9 formulation and hence an assessment of the effective
length.
It is evident that when ange thicknesses increase, FEM-tendency line and the EC9-tendency line intersect each other or tend
to intersect in one point (PI). This means that the effective lengths
given by the code are not always on the safe side as shown in
Fig. 19b. This is due to the fact that EC9 formulation does not take
into account the variation of effective length with the thickness for

the failure modes where ange is strongly involved such as mode

1 and mode 2a. The above points PI provide a limit of applicability for the present codied formulation. This is evidenced in
Fig. 19b where the line joining these points is graphed.
5. Conclusions
The presented paper dealt with parametric analyses carried out
on 43 welded aluminium T-stubs models, suitably calibrated on
the basis of available experimental tests. The obtained results have
been carefully elaborated in order to check the reliability of the
methods presently provided by Eurocode 9.
The most important outcomes reached in the study may be
summarised in the following points:
 The EC9 k-method is quite reliable for interpreting the T-stub
connections behaviour especially when weak bolts are
employed.
 The ultimate strengths according to EC9 are overestimated
when T-stubs with thick anges and large bolt pitches are used,
especially in the case of failure mode 1, which is strongly
related to the effective length concept.
 An improvement of the code for assessing the effective length
would be therefore advisable by revising the denition of
effective length formulation. In such a revision the transition
of failure pattern from non-circular to circular, which takes
place when the thickness and the pitch exceed a certain limit,
should be more carefully taken into account.
In addition, further studies could be carried out also to check
the reliability of the formulation provided by Eurocode 3 for steel
T-stub effective length.
Acknowledgements
This research is a continuation of previous studies developed
within the work for the preparation of Eurocode 9, Aluminium
Structures, coordinated by Prof. F.M. Mazzolani.
The authors also acknowledge the nancial support for PhD student Eng. Tayyab Naqash given by Reti per la Conoscenza e lOrientamento Tecnico-Scientico per lo Sviluppo della Competitivit
(Re.C.O.Te.S.S.C.). Finally, the hosting of Eng. Tayyab Naqash at the
University of Naples Federico II, within the framework activities
of Master in Design of Steel Structures, is gratefully acknowledged.

G. De Matteis et al. / Engineering Structures 41 (2012) 548561

References
[1] Mazzolani FM. Aluminium alloy structures. London: E&FN SPON; 1995.
[2] EN 1999-1-1. Eurocode 9. Design of aluminium structures Part 11: General
structural rules; February 2007.
[3] EN 1993-1-8. Eurocode 3. Design of steel structures Part 1.8: Design of joints;
May 2005.
[4] Girao Coelho MA, Bijlaard FSK, Da Silva LS. Experimental assessment of the
ductility of extended endplate connections. Eng Struct 2004;26:1185206.
[5] De Matteis G, Mandara A, Mazzolani FM. T-stub aluminium joints: inuence of
behavioural parameters. Comput Struct 2000;78(13):31127.
[6] De Matteis G, Mazzolani FM. Behaviour of welded aluminium T-stub
connections: experimental analysis and interpretative models. In:
Proceedings of the ASCE structures congress; May 2006.
[7] Piluso V, Rizzano G. Experimental analysis and modelling of bolted T-stubs
under cyclic loads. J Construct Steel Res 2008;64(6):65569.
[8] Lemonis ME, Gante CJ. Incremental modelling of T-stub connections. J Mech
Mater Struct 2006;1(7):113559.

561

[9] Stamatopoulos GN, Ermopoulos JC. Inuence of the T-stub exibility on its
strength. Int J Steel Struct 2010;10(9):739.
[10] Mistakidis ES, Baniotopoulos CC, Bisbos CD, Panagiotopoulos PD. Steel T-stub
connections under static loading: an effective 2-D numerical model. J
Construct Steel Res 1997;44(12):5167.
[11] Giro Coelho MA, Da Silva LS, Bijlaard FSK. Finite-element modelling of the
nonlinear behavior of bolted T-Stub connections. J Struct Eng.
2006;132(6):91828. http://dx.doi.org/10.1061/(ASCE)0733-9445(2006).
[12] Efthymiou E. Numerical and experimental investigation of a reference
aluminium bolted joint. Jordan J Civil Eng 2008;2(4):30723.
[13] De Matteis G, Brescia M, Formisano A, Mazzolani FM. Behaviour of welded
aluminium T-stub joints under monotonic loading. Comput Struct
2009;87:9901002.
[14] Xu H, Nong Guo X, Luo YF. The load-bearing capacity of aluminium alloy Tstub joints. Adv Mater Res 2011;261263:765.
[15] Dassault Systemes Simulia Corporation. Abaqus, Users manual-version 6.7
documentation; 2007.