Journal of Constructional Steel Research 80 (2013) 202212

Contents lists available at SciVerse ScienceDirect

Journal of Constructional Steel Research

Steel plate girder webs under combined patch loading, bending

and shear
C. Graciano , A. Ayestarn
Universidad Simn Bolvar, Departamento de Mecnica, Apdo. 89000, Caracas 1080-A, Venezuela

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

Article history: During bridge erection employing the incremental launching method, plate girders are subjected to a com-
Received 24 April 2012 bined loading situation. Due to the support reaction, the thin webs are withstanding concentrated loads,
Accepted 24 September 2012 and due to the self weight of the launching nose and the span between the piers the web is also under the
Available online 30 October 2012
action of bending and shear force. This paper is aimed at investigating the nonlinear behavior of unstiffened
girder webs subjected to combined loading (concentrated loading, bending and shear) by using the nite el-
Keywords:
Patch loading
ement method. Firstly, the numerical models are validated against experimental results taken from the liter-
Bending resistance ature. Secondly, each individual resistance is calculated in order to normalize the applied loads. Thereafter, a
Shear resistance parametric analysis is conducted looking at the interaction between the three types of loading, and a com-
Incremental launching bined failure mode is identied. Finally, the results shows limits in the resistance when all three loads are
Finite element applied.
Modeling 2012 Elsevier Ltd. All rights reserved.
Interaction diagram

1. Introduction the collapse of girder webs with/without longitudinal stiffening under

Slender girders are often subjected to various loading conditions
namely bending, patch loading or shear. Each one of its components
are designed to withstand a particular load, the anges are capable
to withstand compressive/tensile forces resulting from bending stress
distribution, thin slender webs may be capable to resist heavy shear
loads as well as localized compressive loads due to patch loads. Verti-
cal stiffeners and thin webs should be able to undertake tension eld
actions that derivates from shear loading.
At designing these slender girders, there is a compromise between
weight-cost and strength, the webs becomes high and thin in order to
reduce weight, increasing compressive/tensile loads in anges. Con-
sequently, the slender girder webs are prompter to failure due to in-
stability or buckling phenomena.
Regarding patch loading, a large amount of research can be found
in the literature. Roberts and Rockey [1] presented a failure mecha-
nism model based on the formation of plastic hinges in the anges
and yield lines in the webs. Thereafter, a series of experimental inves-
tigations has been developed [26] in order to investigate the resis-
tance of the slender girder webs subjected to patch loading. From
the results obtained by Lagerqvist [5] a model based on strength
curves was presented and used currently in design codes [7]. Numer-
ical models have also been developed to propose improvement in
current formulations [810]. Shear loading is perhaps the less studied
case, D'Apice et al. [11] performed a series of experiments to investigate
Corresponding author.
E-mail address: cgracian@usb.ve (C. Graciano).
shear loading, and Deslesques [12] proposed a model to calculate criti-
cal shear stresses. Numerical models for transversely stiffened girder
webs under shear were investigated by Lee and Yoo [13,14] and Lee et
al. [15] and for longitudinally stiffened webs by Pavlovic et al. [16].
More experimental investigations focusing on the structural behavior
of longitudinally stiffened girders under shear have been conducted in
the last three decades [1719]. Concerning bending, D'Apice et al. [11]
also conducted an experimental investigation on the resistance of slen-
der girder under bending moments. Dubas and Tschamper [20] studied
experimentally the stability of slender girders with/without longitudi-
nal stiffeners subjected to bending, and combined patch loading and
bending.
Most experimental/numerical investigations have looked up the
interaction between two types of loading: patch loading and bending
or patch loading and shear [21,22] but few have been able to investi-
gate the interaction between the three types of loading [23,24]. Braun
and Kuhlmann [24] investigated rst the interaction between patch
loading and shear, and secondly between patch loading and bending,
thus by merging both interactions an equation for the three types of
loading was proposed. However, the analysis was mainly based on a
two dimensional interaction between each pair of loads and then
merging the equations into a three dimensional one. Therefore, it is
necessary to perform computer simulations on the basis of paramet-
ric analysis where the three loads are varying simultaneously, for a
deeper investigation of this problem. This paper is aimed at investi-
gating the interaction between patch loading, bending and shear
force in unstiffened girder webs by using nonlinear nite element
analysis.

0143-974X/$ see front matter 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.jcsr.2012.09.018
 C. Graciano, A. Ayestarn / Journal of Constructional Steel Research 80 (2013) 202212 203

P bf

ss
M
M

V V
hw

tw

tf
a

Fig. 1. A plate girder subjected to combined patch loading, bending and shear.

Patch Load (P) where P and V are the applied patch load and shear, respectively; PR
and VR are the corresponding resistances. Shahabian and Roberts
[21] also investigated the inuence of combined shear and patch
loading on the strength of slender girder webs, and proposed a model
   2
P V
1: 2
PR VR

Simmetry Khlmann and Braun [22], based on experimental and numerical

Boundary studies, proposed for the interaction between patch loading and
Conditions shear the equation
   1:6
P V
1: 3
PR VR

Regarding the interaction between patch loading and bending,

y Bergfelt [3] investigated this interaction and found that the bending
moment affects considerably the patch loading resistance if the ap-
x
plied moments M is 60% larger than the bending resistance MR. As a
z
result the following equation was proposed.
Simply Supported  8  2
P M
1: 4
PR MR
Fig. 2. Numerical model for patch loading resistance.

In 1983, Elgaaly [26] proposed the following equation for the in-
2. Literature review teraction between patch load and bending.
 3  3
At present, individual resistances have been investigated, regard- P M
1: 5
less that in most practical cases these loading conditions act in com- PR MR
bined forms. Three decades ago, Zoetemeijer [25] investigated the
inuence of shear load on the strength of rolled section under con- Later on, Ungermann [27] suggested a linear relationship between
centrated loading, as a result the following equation was proposed the two types of loading.
 2  2    
P V P M
1 1 1:4: 6
PR VR PR MR

Table 1
Dimensions and material properties tested by Roberts [23].

Girder a hw tw bf tf ss fyw fyf PREXP PRFEA

(mm) (mm) (mm) (mm) (mm) (mm) (MPa) (MPa) (kN) (kN) %

B3-3 600 500 3.05 149 3.05 50 221 221 70.56 77.1 9.3
B3-7 600 500 3.05 149 6.75 50 221 279 90.72 88.78 2.1
B3-12 600 500 3.05 149 11.75 50 221 305 111.36 115.9 4.1
B3-20 600 500 3.05 149 20.06 50 221 305 130.6 135.2 3.5
204 C. Graciano, A. Ayestarn / Journal of Constructional Steel Research 80 (2013) 202212

Shear Load (V)

x
z
Support Test Panel
100 mm
Out-of-plane restriction
In secondary panel

Simply Supported

Fig. 3. Numerical model for shear resistance.

Table 2 In an attempt to cover the gap regarding the interaction between

Dimensions and material properties for test girders used for validation of the shear re- the three types of loading, Roberts and Shahabian [23] proposed a
sistance model.
model for this interaction as follows:
Girder a hw tw bf tf fyw fyf VREXP VRFEA
(mm) (mm) (mm) (mm) (mm) (MPa) (MPa) (kN) (kN) (%)    4  4     2  2 
M V M V M P M P
LS1 1270 1270 4.76 355 38.1 323 234 808 811 0.4
1 1
MR VR MR VR MR PR MR PR
[11]    2  2     
V P V P V P M
H1-T1 3810 1270 9.98 459 24.8 745 703 2542 2413 5.0 1 2 0
[30] VR PR VR PR VR PR MR
G8 2851 1080 8.9 221 22.4 372 431 996 973 2.3 9
[31]

where represents the numerical value of the interaction function.

Finally, Lagerqvist [5] proposed a similar model More recently, Braun and Kuhlmann [24] presented a formulation
for the interaction between patch loading, bending and shear, consid-
    ering a consistent denition of the FV and MV interaction equations
P M
0:8 1:4: 7 as a continuous function
PR MR
   3:6  1:6
The relationship between shear and patch loading has also been P M V
1:0: 10
investigated separately. Roberts and Shahabian [23] proposed an PR MR VR
equation for this interaction:
General recommendations to estimate the reduction in patch load-
   4 ing resistance due to the presence of global bending combined with
V M shear for slender girders has been difcult to make due to the limited
1: 8
VR MR number of test and numerical results available in the literature.

Load (P)

Simmetry
Boundary
Conditions

x
z 100 mm
Out-of-plane restriction
In secondary panel
Simply Supported

Fig. 4. Numerical model for bending resistance.

 C. Graciano, A. Ayestarn / Journal of Constructional Steel Research 80 (2013) 202212 205

Table 3 Table 4
Dimensions and material properties tested by Dubas and Tschamper [20]. Nomenclature for girder types according to the amount of panels in the model.

Girder a hw tw bf tf fyw fyf MREXP MRFEA Girder designation N

(mm) (mm) (mm) (mm) (mm) (MPa) (MPa) (kN m) (kN m) (%) [Panels] [%Q]

T01-4 2400 990 4 150 10 360 274 535 515.8 3.6 Type N0 2 51015
T02-4 2400 990 4 150 8 349 298 480 456.3 4.9 Type N1 3 202530
T03-4 2400 990 5 150 8 317 294 500 488.3 2.3 Type N2 5 354045
VT03-4 2480 1000 5.2 150 12 305 277 715 649.8 9.1 Type N3 7 505560
VT05-4 2480 800 5 150 8.4 292 300 410 404.7 1.3 Type N4 9 657075
Type N5 11 808590

3. Computation of individual resistances

Table 5
Details of plate girders tested by Roberts and Shahabian [23].
In engineering practice, there exist various manners to determine
the individual resistances to patch loading (PRFEA), shear (VRFEA) and Girder N a hw tw bf tf ss fyw fyf
bending (MRFEA) for plate girders, i.e. by means of experimental studies, (mm) (mm) (mm) (mm) (mm) (mm) (MPa) (MPa)

computer simulations or using available formulae in design codes. PG1-2SP1 2 600 600 4.1 200 12.3 50 339 257
Herein, the individual resistances are calculated by means of nonlinear PG1-2SP2 2 600 600 4.1 200 12.3 50 339 257
PG3-3SP2 2 900 600 3.2 200 10.2 50 275 258
nite element analysis. The computer models of the plate girders were
PG4-3SP2 2 1000 500 1.9 200 10 50 236 294
developed using the nite element program ANSYS [28]. Shell elements PG5-1BSP 3 600 600 3.03 150 10.3 50 197 273
Shell 181 (4 nodes and 6 degrees of freedom at each node) from the
ANSYS library were used to model the web, and anges (top and bot-
tom). The FE analysis was performed using the modied Riks method and boundary conditions only one half of the girders (Fig. 2) was
[29] to properly trace the nonlinear path of the loaddisplacement re- modeled. Vertical stiffeners were considered at the girder ends, and
sponse of the girder. Herein, the initial geometric imperfections were simply support conditions were given for all nodes along the stiffener,
modeled as the 1st buckling mode shape with a maximum allowable i.e. only displacements in x direction and rotation around z directions
amplitude chosen equal to hw/200 according to EC3 [7]. Each numerical were allowed. Table 1 shows the dimensions, material properties and
model was validated against experimental results found in the litera- results for the girders used for validation. The material was consid-
ture. Fig. 1 shows a schematic representation of a girder subjected to ered to have a perfectly elastoplastic behavior.
patch loading P, a bending moment M and shear load V. After performing a convergence analysis on the patch loading re-
In Fig. 1, a is the length of the web panel, bf the ange width, fyf the sistance, a mesh with 3480 elements was used for this calculation.
yield stress of anges, fyw the web panel yield stress, hw is the web As seen in Table 1, the difference between numerical PRFEA and exper-
panel depth, ss the patch load length, tf the thickness of anges, and imental PREXP patch loading resistances ranged from 2.1% to 9.3%. This
tw the web panel thickness. value is suitable for this type of study according to previous re-
searches [58,11]. The difference in the results could be due to the
3.1. Patch loading resistance (PRFEA) way in which the load was applied in both cases. In the experiments
the load was transferred to the ange through rectangular blocks,
For patch loading, the model was validated with experimental re- while in the numerical model the load was applied over the patch
sults obtained by Roberts [2]. Due to symmetry in geometry, loading load length.

P V
(a) Girder type N3

Test
1 2
panel

L1 L2 L3
R1 R2

M1 M2
Mv
Mp
(b) Bending moment
Diagram
P
V

M1 Test M2
(c) panel

Girder type N3 simplifed in N1

Fig. 5. Schematic view of the analyzed girders. (a) Girder type N3 full girder; (b) bending moment diagram; and (c) girder type N1 girder type N3 simplied [33].
206 C. Graciano, A. Ayestarn / Journal of Constructional Steel Research 80 (2013) 202212

Table 6 Table 7
Comparison between experimental and computed values for patch loading resistance Computed resistances for Girder PG1-2SP1.
(FR) and shear resistance (VR).
Girder VRFEA PRFEA MRFEA
Girder PEXP PFEA |P| VEXP VFEA |V| (kN) (kN) (kN m)
(kN) (kN) % (kN) (kN) %
PG1-2SP1 350 260 434
PG1-2SP1 205 204.4 0.3 158 157.5 0.3
PG1-2SP2 208 206.5 0.7 97 96.3 0.7
PG3-3SP2 106 109.7 3.5 107 111.0 3.7
4. Combined loading model
PG4-3SP2 44 43.6 0.9 43 42.6 0.9
PG5-1BSP 79 83.7 5.9 59 62.5 5.9
4.1. Model simplication

There would be very high computational and time costs involved

3.2. Shear resistance (VRFEA) in the calculations herein if the girders are modeled with all panels,
therefore it was considered convenient to make some simplications
For shear, the model is similar to the one proposed by Estrada et in the model. Fig. 5 shows how a girder type N3 with 7 panels can be
al. [30], consisting of two panels simply supported at the ends and simplied into a girder type N1 with only 3 panels, by giving addi-
with the load applied in the area between the intermediate stiffeners tional boundary conditions [33].
(Fig. 3). As in the experiments the secondary panel, opposite to the Bending moments M1 and M2, in girder type N1, can be calculated
test panel, was restricted to move in the out-of-plane direction to according to the following procedure. At rst, the lengths L1, L2 and L3
guarantee failure in only one side. The numerical model was validated are calculated as
with three experimental results obtained by D'Apice et al. [11],
Cooper et al. [31], and Nishino et al. [32]. Table 2 shows the dimen- N1 a
L1 a ds 11
sions, material properties and results for the girders used for valida- 2 2
tion. The material was considered to have a perfectly elastoplastic
behavior. ads
After performing a convergence analysis on the shear resistance, a L2 12
2
mesh with 12200 elements was used for this calculation. The differ-
ence between numerical VRFEA and experimental VREXP resistances to  
N1 N1 d
shear ranged from 0.4% to 5% (Table 2). In spite of the difference in L3 a ds 1 s 13
2 2 2
the geometry of the girders used for validation and the lack of infor-
mation regarding initial imperfection the results are very accurate.
where N is the number of panels in the model (N = 2, 3, 5, 7, 9, 11);
and ds = 100 mm is the distance between the vertical end stiffeners.
3.3. Bending resistance (MRFEA) Let's now calculate the support reactions at the ends (R1 and R2),
and bending moments in the test panel (MP and MV) are
For bending, the model was validated with experimental results
obtained by Dubas and Tschamper [20]. Due to the symmetry in ge- P R L2 L3 V R L3
ometry, loading and boundary conditions only one half of the girders R1 14
L1 L2 L3
(Fig. 4) was modeled. As in the experiments, nodes in the secondary
panel were restricted to move in the out-of-plane direction to guar-
R2 P R V R R1 15
antee failure in the test panel. The load was applied within the area
corresponding to the intermediate stiffeners. Table 3 shows the di-
M P R1 L 1 16
mensions, material properties and results for the girders used for
validation.
A convergence analysis on the resistance to bending leads to a MV R1 L1 L2 P R L2 17
mesh with 10,600 elements for this calculation. The difference be-
tween numerical MRFEA and experimental MREXP bending resistances where MP and MV are the bending moments at the points where the
ranged from 1.3% to 9.1% (Table 3). concentrated load P and the shear load V are applied, respectively.

Patch Load (P)

Shear Load (V)

x
z
Support
100 mm
Test Panel 100 mm Out-of-plane restriction
In secondary panel
Simply Supported

Fig. 6. Representative nite element models for girder PG5-1BSP tested by Roberts and Shahabian [23].
 C. Graciano, A. Ayestarn / Journal of Constructional Steel Research 80 (2013) 202212 207

Table 8 Table 8 (continued)

Numerical results for the parametric analysis and interaction equation Girder
Girder type P V M Pn Vn Mn Eq. (10)
PG1-2SP1.
(kN) (kN) (kN m)
Girder type P V M Pn Vn Mn Eq. (10) 166.40 108.50 312.48 0.64 0.31 0.72 1.10
(kN) (kN) (kN m) 161.20 112.00 325.50 0.62 0.32 0.75 1.14
N0 208.00 157.50 43.40 0.80 0.45 0.10 1.08 156.00 115.50 338.52 0.60 0.33 0.78 1.18
208.00 161.00 47.74 0.80 0.46 0.11 1.09 148.20 122.50 351.54 0.57 0.35 0.81 1.22
208.00 168.00 47.74 0.80 0.48 0.11 1.11 140.40 126.00 364.56 0.54 0.36 0.84 1.27
208.00 175.00 47.74 0.80 0.50 0.11 1.13 117.00 115.50 338.52 0.45 0.33 0.78 1.03
208.00 185.50 52.08 0.80 0.53 0.12 1.16 114.40 126.00 373.24 0.44 0.36 0.86 1.22
208.00 192.50 56.42 0.80 0.55 0.13 1.18 104.00 129.50 390.60 0.40 0.37 0.90 1.29
208.00 203.00 60.76 0.80 0.58 0.14 1.22 91.00 133.00 399.28 0.35 0.38 0.92 1.30
205.40 213.50 65.10 0.79 0.61 0.15 1.24 78.00 136.50 412.30 0.30 0.39 0.95 1.35
205.40 227.50 73.78 0.79 0.65 0.17 1.29 62.40 133.00 412.30 0.24 0.38 0.95 1.28
202.80 245.00 86.80 0.78 0.70 0.20 1.35 44.20 126.00 399.28 0.17 0.36 0.92 1.11
200.20 262.50 95.48 0.77 0.75 0.22 1.41 31.20 133.00 416.64 0.12 0.38 0.96 1.20
192.40 308.00 125.86 0.74 0.88 0.29 1.57 15.60 129.50 412.30 0.06 0.37 0.95 1.10
184.60 336.00 147.56 0.71 0.96 0.34 1.67 N5 169.00 84.00 299.46 0.65 0.24 0.69 1.01
163.80 350.00 164.92 0.63 1.00 0.38 1.66 166.40 87.50 308.14 0.64 0.25 0.71 1.04
N1 208.00 101.50 95.48 0.80 0.29 0.22 0.94 163.80 91.00 321.16 0.63 0.26 0.74 1.08
208.00 105.00 99.82 0.80 0.30 0.23 0.95 158.60 91.00 325.50 0.61 0.26 0.75 1.08
208.00 108.50 104.16 0.80 0.31 0.24 0.96 156.00 94.50 338.52 0.60 0.27 0.78 1.13
205.40 112.00 108.50 0.79 0.32 0.25 0.96 150.80 98.00 351.54 0.58 0.28 0.81 1.18
205.40 119.00 112.84 0.79 0.34 0.26 0.98 145.60 101.50 360.22 0.56 0.29 0.83 1.21
205.40 122.50 117.18 0.79 0.35 0.27 0.99 137.80 105.00 368.90 0.53 0.30 0.85 1.23
205.40 129.50 121.52 0.79 0.37 0.28 1.00 130.00 105.00 377.58 0.50 0.30 0.87 1.25
202.80 136.50 130.20 0.78 0.39 0.30 1.01 104.00 105.00 377.58 0.40 0.30 0.87 1.15
202.80 147.00 138.88 0.78 0.42 0.32 1.05 96.20 108.50 386.26 0.37 0.31 0.89 1.18
200.20 157.50 147.56 0.77 0.45 0.34 1.07 85.80 108.50 394.94 0.33 0.31 0.91 1.20
200.20 168.00 160.58 0.77 0.48 0.37 1.11 72.80 105.00 390.60 0.28 0.30 0.90 1.11
197.60 182.00 173.60 0.76 0.52 0.40 1.15 59.80 105.00 394.94 0.23 0.30 0.91 1.09
192.40 203.00 195.30 0.74 0.58 0.45 1.21 49.40 105.00 394.94 0.19 0.30 0.91 1.05
187.20 224.00 225.68 0.72 0.64 0.52 1.30 36.40 105.00 399.28 0.14 0.30 0.92 1.03
179.40 252.00 260.40 0.69 0.72 0.60 1.44
145.60 245.00 269.08 0.56 0.70 0.62 1.30
106.60 238.00 269.08 0.41 0.68 0.62 1.13 Solving for the bending moments M1 and M2, we obtain
N2 200.20 98.00 156.24 0.77 0.28 0.36 0.93  
200.20 101.50 164.92 0.77 0.29 0.38 0.94 MP N1
197.60 108.50 169.26 0.76 0.31 0.39 0.95
M1 a ds 1 18
L1 2
197.60 112.00 177.94 0.76 0.32 0.41 0.96
195.00 115.50 186.62 0.75 0.33 0.43 0.97  
MV N1
195.00 122.50 195.30 0.75 0.35 0.45 0.99 M2 a ds 1 : 19
192.40 129.50 208.32 0.74 0.37 0.48 1.01 L3 2
189.80 136.50 221.34 0.73 0.39 0.51 1.04
187.20 147.00 234.36 0.72 0.42 0.54 1.08
Concentrated loads P and shear loads V cannot be applied sepa-
182.00 154.00 247.38 0.70 0.44 0.57 1.10
158.60 147.00 238.70 0.61 0.42 0.55 0.98
rately, as in the experiments a total load Q is applied to the test
174.20 178.50 290.78 0.67 0.51 0.67 1.25 panel. Nevertheless, P and V are related, therefore it is necessary to
166.40 192.50 321.16 0.64 0.55 0.74 1.36 dene a dimensionless parameter given by
N2 156.00 210.00 355.88 0.60 0.60 0.82 1.53
140.40 224.00 386.26 0.54 0.64 0.89 1.69 V=Q: 20
119.60 234.50 412.30 0.46 0.67 0.95 1.82
91.00 238.00 429.66 0.35 0.68 0.99 1.85
59.80 234.50 434.00 0.23 0.67 1.00 1.76 Hence, through the computational analysis the relationship be-
31.20 231.00 434.00 0.12 0.66 1.00 1.63 tween P and V is established as
N3 189.80 94.50 212.66 0.73 0.27 0.49 0.93
189.80 98.00 221.34 0.73 0.28 0.51 0.95 P 1 Q : 21
187.20 101.50 230.02 0.72 0.29 0.53 0.96
184.60 105.00 238.70 0.71 0.30 0.55 0.97
184.60 112.00 251.72 0.71 0.32 0.58 1.01 A concentrated load P was transferred into the girder by loading
179.40 115.50 260.40 0.69 0.33 0.60 1.02 all the nodes located in the patch loaded ange along the loading
176.80 122.50 277.76 0.68 0.35 0.64 1.07 length ss with an equal value force. Correspondingly, the shear load
174.20 129.50 290.78 0.67 0.37 0.67 1.11 Q was applied onto the nodes placed between the vertical stiffeners.
169.00 136.50 303.80 0.65 0.39 0.70 1.15
163.80 143.50 321.16 0.63 0.41 0.74 1.21
Displacement constraints were applied to these loaded nodes in the
156.00 150.50 338.52 0.60 0.43 0.78 1.27 out-of-plane direction and all rotations were restrained. Table 4
145.60 157.50 360.22 0.56 0.45 0.83 1.35 shows the nomenclature employed herein for the numerical models
135.20 164.50 381.92 0.52 0.47 0.88 1.45 according to the number of panels N; i.e. girder type N0 has 2 panels,
119.60 168.00 394.94 0.46 0.48 0.91 1.48
and girder N5 has 11 panels.
101.40 171.50 412.30 0.39 0.49 0.95 1.54
83.20 171.50 420.98 0.32 0.49 0.97 1.54
59.80 164.50 407.96 0.23 0.47 0.94 1.33 4.2. Geometry and material properties
41.60 168.00 420.98 0.16 0.48 0.97 1.37
20.80 164.50 420.98 0.08 0.47 0.97 1.27 Table 5 shows the dimensions of the girders used for validation of
N4 182.00 91.00 260.40 0.70 0.26 0.60 0.97
176.80 91.00 269.08 0.68 0.26 0.62 0.97
the numerical model considering the three forms of loading. In this
176.80 94.50 277.76 0.68 0.27 0.64 1.00 table, a is the length of the web panel, bf the width of anges, fyf the
174.20 101.50 290.78 0.67 0.29 0.67 1.04 yield stress of anges, fyw the web panel yield stress, hw is the web
169.00 105.00 303.80 0.65 0.30 0.70 1.07 panel depth, ss the length of patch load, tf the thickness of anges,
208 C. Graciano, A. Ayestarn / Journal of Constructional Steel Research 80 (2013) 202212
(a) Girder Type N1 ( = 0.05-Mn = 0.25)
(c) Girder Type N5 (= 0.05-Mn= 0.80)
Fig. 7. Equivalent von Mises stress distribution [MPa] at ultimate load level for = 0.05 at various bending moment levels.
and tw the web panel thickness. The material was considered to have
a perfect elastoplastic behavior. Young's modulus and Poisson's ratio
were set to 200 GPa and 0.3 respectively.
4.3. Model validation
For the validation of the numerical model of the girder subjected to
combined loading; ve girders previously tested by Roberts and
Shahabian [23] were used. As seen in Table 6, the comparison between
computed (PFEA, VFEA) and experimentally measured (PEXP, VEXP) loads is
satisfactory; the difference is only 2%. Fig. 6 shows the nite element
model subjected to concentrated loading, bending and shear force.
5. Parametric analysis
In this section a parametric analysis is conducted in order to inves-
tigate the interaction between the three types of loading:
was varied from 5% to 95% at 5% intervals (Table 4).
N was varied from 2 to 11; i.e. N = 2, 3, 7, 9 and 11, depending on
the number of panels in the model. It is important to notice that
the level of bending increases with N.
For each combination of and N, the values for concentrated load,
shear load and bending moment were computed. Hence a total of 104
girders were simulated. The applied loads were normalized with its cor-
FEA
responding resistance calculated according to: P n P =P FEA R
Vn
V FEA MFEA
=V R
FEA
Mn =M R :
FEA
(b) Girder Type N3 ( = 0.05-Mn = 0.55)
PRFEA, VRFEA, and MRFEA are the resistances calculated using the models
described in Section 3; and Pn, Vn and Mn are the normalized resistances.
In order to continue the study, the geometry of the Girder PG1-2SP1 is
used as a basis for the parametric analysis. Table 7 shows the computed
resistances for each individual load for Girder PG1-2SP1.
In addition, Table 8 shows the numerical results obtained for the
all the load combinations, besides the validity of Eq. (10) is examined.
On the one hand, the minimum patch loading resistance is observed
for Girder type N4 (Pn = 0.06, Vn = 0.37 and Mn = 0.95), showing
that bending is a dominant factor at reducing it. On the other hand,
the minimum shear resistance is achieved for Girder type N5 (Pn =
0.65, Vn = 0.24 and Mn = 0.69), this result a more balanced picture
of the interaction. Finally, for the bending resistance the minimum
value is observed for Girder type N0 (Pn = 0.80, Vn = 0.45 and Mn =
0.10), it shows that a higher applied shear load gives a smaller reduc-
tion than the one caused by bending. Regarding the validity of
Eq. (10), Table 10 shows the evaluation for the 104 load combina-
tions; the average is 1.19 with a standard deviation of 0.21. It means
that most of the load combinations lie on the safe side for Eq. (10).
In the following sections the interaction is investigated through the
analysis of observed failure mechanism and the loaddisplacement
responses.
5.1. Effects on stress distribution
Herein the structural behavior of girder webs under combined
loading is investigated in detail by studying the membrane stress dis-
tribution in the web panel.
 C. Graciano, A. Ayestarn / Journal of Constructional Steel Research 80 (2013) 202212
(a) Girder type N0 ( = 0.70-Mn = 0.38)
(b) Girder type N1 ( = 0.70-Mn = 0.58)
(b) Girder type N1 ( = 0.70-Mn = 0.58)
Fig. 8. Equivalent von Mises stress distribution [MPa] at ultimate load level for = 0.70 at various bending moment levels.
Fig. 7 shows the equivalent von Mises stress distribution for
girders type N1, N3 and N5, subjected to various level of bending mo-
ments (Mn = 0.25, 0.55 and 0.8, respectively). As expected Fig. 7a
shows that the most stressed area is that one close to the loaded
ange, beneath the patch load.
Fig. 7a also shows the membrane stress distribution for girder type
N1 (Mn = 0.25). In the tension zone the distribution is rather uniform;
in contrast, the distribution in the compression zone is inuenced by
the compressive stresses due to the patch load. For girder types N3
(Fig. 7b) and N5 (Fig. 7c) with Mn = 0.55 and 0.80, respectively, the
internal forces produced by bending cause compressive stresses in
the loaded ange. Furthermore, the stresses in the vicinity of the ver-
tical stiffener are very small. The presence of bending increases the
stressed area below the concentrated load. Still the magnitude of
the applied shear force is small ( = 0.05), therefore its contribution
on the failure mechanism is minimal.
Observing Fig. 7c, when the applied bending moment increases
(Mn = 0.80) there is a redistribution of the stresses with respect to
the neutral axis. Consequently, the patch loading resistance is signif-
icantly reduced.
Once the magnitude of the shear force increases in relation to the
concentrated load, i.e. = 0.70; the failure mechanism changes.
Fig. 8a shows a failure characteristic of shear buckling mode with
the presence of a tension eld in the test panel. When the applied
bending moment increases (Fig. 8b and c) the failure mechanism
changes, and the stresses are more localized toward the compression
ange and the vertical stiffener allowing tension eld action. After
applying the bending moment the longitudinal stresses increase
209
(c) Girder type N2 ( = 0.70-Mn = 0.92)
(c) Girder type N2 ( = 0.70-Mn = 0.92)
reducing the effect of the shear force in the stress distribution at ulti-
mate load level.
5.2. Effects on postcritical behavior
Fig. 9 shows the loaddisplacement response for girders Type N0,
N1, N2 and N3, for various levels of shear loading. The curves for
Girders Type N0 and N1 (Fig. 9a and b), the variation in patch loading
resistance for values of between 0.05 and 0.40 is very small, since
the effect of the concentrated load dominates over shear. For larger
values, the reduction in patch load resistance is more signicant
(60% for Girder type N3), and in the postultimate range after a sharp
drop all curves seem to converge.
When increasing the applied bending moment as for girders type
N2 and N3 (Fig. 9c and d) there is a substantial reduction even for
small values. It can be observed that for values larger than 0.55
the reduction in resistance is remarkable larger for all girders (Pn =
0.4 with = 0.75). This behavior is due to the presence of bending,
increasing the compressive stresses beneath the loaded anges. In
the postultimate range the curves all curves converge to different
values, for both larger shear load and bending moment the load
displacement tends to reach a plateau. A similar result was observed
by Graciano and Casanova [34] for longitudinally stiffened girders
web under combined loading and bending.
Fig. 10 shows the loaddisplacement responses for girder types
N1, N2 and N3 for a constant value of shear ( = 0.75) and various
levels of bending. It is clearer to see the effect of bending both in
patch loading resistance and structural response.
210 C. Graciano, A. Ayestarn / Journal of Constructional Steel Research 80 (2013) 202212

Fig. 9. Loaddisplacement responses for various values of . (a) Girder type N0, (b) girder type N1, (c) girder type N2 and (d) girder type N3.

6. Interaction diagrams

In order to investigate the combined action of the three loads, it is

convenient to plot interaction diagrams for various permutations. In
the numerical model the level of bending depends on the number of
panels, and the applied loads P and V.

Fig. 10. Loaddisplacement responses with = 0.75 for girders types N1, N2 and N3. Fig. 11. Interaction between patch loading and shear.
 C. Graciano, A. Ayestarn / Journal of Constructional Steel Research 80 (2013) 202212
Fig. 12. Interaction between patch loading and bending.
Fig. 11 shows the interaction between patch loading and shear for
various levels of bending. At lower values of Mn b 60%, the reduction in
patch loading resistance Pn is minimal. For larger values of Mn > 60%,
the inuence of bending in more noticeable, and the shear load Vn re-
mains practically constant. For practical purposes, it can be said that
for large bending moments the full shear resistance is unattainable.
Eqs. (1)(3) are also plotted in Fig. 11, the model presented by
Zoetemeijer [24] seems to cover a wider range of the computed
values. The models developed by Shahabian and Roberts [21] and
Khlmann and Braun [22] are very similar in nature. Comparing the
numerical results with the models found in the literature for this in-
teraction, it is observed that their use can be unsafe for most of the
computed values herein. It is important to point out that those equa-
tions were developed disregarding the effect of bending action.
Studying the interaction between patch loading and bending,
Fig. 12 shows how the resistance to patch loading decreases at bend-
ing moment increments. It seems that the shear load has little inu-
ence for values of Vn b 70%. Fig. 12 also shows the models found
in the literature for this interaction, Eqs. (4)(7) are drawn; it
seems that the model developed by Ungermann [27] ts quite well
the computed values. Once again it is shown that the use of these
two-dimensional interaction formulae for the interaction between
patch loading (Pn) and bending (Mn) may lead to unconservative re-
sults, although these formulae were developed without accounting
for the interaction with shear.
Fig. 13 shows the interaction between shear Vn and bending Mn. It
is interesting to observe that it is possible to achieve the bending re-
sistance Mn = 1 for values of Vn b 60%. Besides this, using the formula
presented by Shahabian and Roberts [21], Eq. (8), may also lead to
conservative results when comparing to the computed values.
Fig. 13. Interaction between shear and bending.
211
Fig. 14. Interaction surface between concentrated loading Pn, shear Vn and bending Mn.
After a closer look of Figs. 11, 12 and 13, it is possible to conclude
that when the three loads are applied simultaneously the current for-
mulae found in the literature for any combination of just two loading
cases tends to provide conservative results. Therefore, it is necessary
to look at the interaction in a 3D diagram, i.e. plotting Pn, Mn and Vn
in the same diagram.
Fig. 14 shows the interaction between the three types of loading
for the girder webs reported in Table 8. Representing the results in
this particular manner, it is possible to identify regions where for a
particular combination of loading the maximum resistances are
achieved, as discussed previously. It is observed that the maximum
patch loading resistance PR can be achieved when the applied shear
force is lesser than 60% of the shear resistance VR and for a bending
moments smaller than 30% of the bending resistance MR, as seen
from the results obtained for Girder Type N0 (Pn = 0.80, Vn = 0.45
and Mn = 0.10). A similar analysis can be performed for the bending
resistance MR, a thin girder web can achieved its bending resistance
if the applied patch load is kept under 15% of the patch loading resis-
tance PR and the applied shear force is lesser than 70% of the shear re-
sistance VR as the results shows for Girder Type N2 (Pn = 0.12; Vn =
0.66 and Mn = 1.00). Eq. (10) proposed by Braun and Khulmann
[24] is also plotted in Fig. 14, it is observed that this equation follows
very well the interaction between the three types of loading. At this
stage, it is noticeable that Eq. (10) was developed on the basis that
the interactions between patch loading/shear and patch loading/
bending can be overlaid.
7. Conclusions
The interaction between combined patch loading, bending and
shear for unstiffened plate girder webs has been investigated herein
by means of nonlinear nite element analysis. The numerical results
showed that the patch loading resistance can be severely reduced
due to the presence of coexisting bending and shear, nevertheless
bending action is the most inuencing factor at dropping the patch
loading resistance. Regarding failure of the girders, it is characterized
by a combination of mechanisms that strongly depends on which of
the three loads is larger. The numerical results also showed that a
three-dimensional interaction represents more appropriately the ef-
fect of these load combinations.
References
[1] Roberts TM, Rockey KC. A mechanism solution for predicting the collapse loads of
slender plate girders when subjected to in-plane patch loading. Proc Inst Civ Eng
1979;67(2):155-75.
212 C. Graciano, A. Ayestarn / Journal of Constructional Steel Research 80 (2013) 202212
[2] Roberts TM. Slender plate girders subjected to edge loading. Proc Inst Civ Eng
1981;71(2):805-19.
[3] Bergfelt A. Studies and tests on slender plate girders without stiffeners shear
strength and local web crippling. Proceedings of IABSE Colloquium; 1971.
p. 67-83.
[4] Roberts TM, Markovic N. Stocky plate girders subjected to edge loading. Proc Inst
Civ Eng 1983;75(2):539-50.
[5] Lagerqvist O, Patch loading Resistance of steel girders subjected to concentrat-
ed forces. Ph.D. thesis, Lulea University of Technology, Division of Steel Structures,
Lulea, Sweden. 1994.
[6] Gozzi J, Patch Loading Resistance of Plate Girders Ultimate and serviceability
limit state. Ph.D. thesis, Lulea University of Technology, Division of Steel Struc-
tures, Lulea, Sweden. 2007.
[7] EN 1993-1-5. Eurocode 3. Design of steel structuresPart 15: plated structural
elements, CEN; 2006.
[8] Shimizu S. The collapse behaviour of web plates on the launching shoe. J Constr
Steel Res 1994;31(1):59-72.
[9] Granath P. Behavior of slender plate girders subjected to patch loading. J Constr
Steel Res 1998;42(1):119.
[10] Chacon R. Resistance of transversally stiffened hybrid steel plate girders to con-
centrate loads. Doctoral thesis. Universitat Politcnica de Catalunya; 2009.
[11] D'Apice MA, Fielding DJ, Cooper PB. Statics tests on longitudinally stiffened plate
girders. Welding Research Council; 1966 (Bulletin No. 117).
[12] Delesques R. Rsistance des mes de poutres sans raidisseurs intermdiaires.
Const Met 1974;2:519 [in French].
[13] Lee SC, Yoo CH. Strength of plate girder web panels under pure shear. J Struct Eng
ASCE 1998;124(2):184-94.
[14] Lee SC, Yoo CH. Experimental study on ultimate shear strength of web panels.
J Struct Eng, ASCE 1999;125(8):838-46.
[15] Lee S, Yoo CH, Yoon DY. New design rule for intermediate transverse stiffeners at-
tached on web panels. J Struct Eng, ASCE 2003;129(12):1607-14.
[16] Pavlovcic L, Detzel A, Kuhlmann U, Beg D. Shear resistance of longitudinally stiff-
ened panels part 1: test and numerical analysis of imperfections. J Constr Steel Res
2007;63:337-50.
[17] Horne MR, Grayson WR. The ultimate behavior of longitudinally stiffened web
panels subjected to shear stress. Proc Inst Civ Eng 1983;75(2):175-203.
[18] Evans HR, Tang KH. Longitudinal stiffeners for girder webs; their behavior and de-
sign. J Constr Steel Res 1986;6(3):173-97.
[19] Pavlovcic L, Beg D, Kuhlmann U. Shear resistance of longitudinally stiffened
panels part 2: numerical parametric study. J Constr Steel Res 2007;63:351-64.
[20] Dubas P, Tschamper H. Stabilit des mes soumises a une charge concentre et a
une exion globale. Const Met 1990;2:25-39.
[21] Shahabian F, Roberts TM. Combined shear-and-patch loading of plate girders.
J Struct Eng, ASCE 2000;126(3):316-21.
[22] Kuhlmann U, Braun B. Combined Shear and patch loading: numerical studies and
development of an interaction equation. Combri-Report-USTUTT-002, Universitt
Stuttgart, Germany; 2007.
[23] Roberts TM, Shahabian F. Ultimate resistance of slender web panels to combined
bending shear and patch loading. J Constr Steel Res 2001;57:779-90.
[24] Braun B, Kuhlmann U. The interaction behaviour of steel plates under transverse
loading, bending moment and shear force. Proceedings of SDSS RIO 2010 stability
and ductility of steel structures; 2010. p. 731-8 [Rio de Janeiro].
[25] Zoetemeijer P. The inuence of normal, bending and shear stresses on the ulti-
mate compression force exerted laterally to European rolled sections. Report
6-80-5, Facultiet der Civiele Techniek, Technische Universiteit; 1980.
[26] Elgaaly M. Web design under compressive edge loads. Eng J AISC 1983;20(4):
153-71.
[27] Ungermann D. Bemessungsverfahren fr vollwand und kastentrger unter
besonderer bercksichtigung des stegverhaltens. Heft 17, Lehrstuhl fr Stahlbau,
RWTH Aachen, (ISSN 07221037); 1990 [in German].
[28] ANSYS, Inc., ANSYS Release 11.0 Elements Reference, USA. 2007.
[29] Riks E. An incremental approach to the solution of snapping and buckling prob-
lems. Int J Solids Struct 1979;15(7):529-51.
[30] Estrada I, Real E, Mirambell E. General behavior and effect of rigid and non-rigid
end post in stainless steel plate girders loaded in shear. Part II: extended numer-
ical study and design proposal. J Constr Steel Res 2007;63:985-96.
[31] Cooper PB, Lew BT, Yen BT. Welded constructional alloy steel plate girder. J Struct
Div ASCE 1964;90(1):136.
[32] Nishino F, Okumura T. Experimental investigation of strength of plate girders in
shear. 8 Congress AIPC, New York; 1968. p. 451-63.
[33] Ayestarn A. Interaccin de cargas de corte, cargas concentradas y momento
ector en la resistencia ltima de vigas esbeltas, Master Thesis, Coordinacin de
Postgrado en Ingeniera Mecnica, Universidad Simn Bolvar, Caracas, Venezuela
2009 [In Spanish].
[34] Graciano C, Casanova E. Ultimate strength of longitudinally stiffened I-girder
webs subjected to combined patch loading and bending. J Constr Steel Res
2005;61:93111.