Structures 20 (2019) 924–934

Contents lists available at ScienceDirect

Structures
journal homepage: www.elsevier.com/locate/structures

Resistance of slender austenitic stainless steel I- girders subjected to patch T

loading
Carlos Gracianoa, , Nelson Loaizaa, Euro Casanovab
⁎

a
Universidad Nacional de Colombia, Facultad de Minas Sede Medellín, Departamento de Ingeniería Civil, A.A. 75267 Medellín, Colombia
b
Universidad del Bío-Bío, Departamento Ingeniería Civil y Ambiental, Avenida Collao 1202, Concepción, Código Postal 4051381, Concepción, Chile

ARTICLE INFO ABSTRACT

Keywords: This paper presents a numerical investigation on the patch loading resistance of slender austenitic stainless steel
Patch loading plate girders. Current design provisions for the resistance to patch loading of stainless steel girders are based on
Resistance the plastic collapse mechanism observed in experimental and numerical studies conducted for carbon steel
Resistance function girders, disregarding the strain hardening capacities of stainless steel. At present, strength-curves approaches are
Stainless steel
used within European standards to deal with stability problems in steel plated structural elements. In this regard,
Ultimate strength
three parameters require special attention: the yield load for plastic resistance, the resistance function depending
on element slenderness, and the elastic buckling load. In this paper, an experimental dataset is firstly collected
from the literature for comparative analysis. Subsequently, an extensive parametric study is conducted through
nonlinear finite element analyses covering a wide range of slender stainless steel I-girder sections. Then, a
resistance function is calibrated throughout a statistical evaluation of experimental and numerical results.
Finally, the results show significant improvements in the predicted patch loading resistances of slender stainless
steel I-girders.

Notation v coefficient of variation v[=s/m]

αF0 imperfection factor
a length of web panel γM1 partial safety factor
bf width of flange ϵ strain
E Young's modulus F0 plateau length
Fcr critical buckling load F slenderness parameter
Fu ultimate resistance calculated with the finite element model ν Poisson's ratio
Fy yield resistance σ stress
fyf flange yield strength σ0.2 stress at 0.2% strain
fyw web yield strength σ1.0 stress at 1.0% strain
hw depth of web panel σu ultimate stress
kF buckling coefficient χF resistance function
Leff effective length for resistance
ly effective loaded length 1. Introduction
MR bending resistance
MS applied bending moment In recent years, an increasing awareness towards sustainability and
m mean value low-cost maintenance has led to the improvement and development of
n nonlinear material parameter material alloys used as construction materials. In this regard, stainless
s standard deviation steel alloys play an important role in the design of structural elements
ss length of patch load in corrosive environments, and to improve their structural performance
tf flange thickness under fire conditions.
tw web thickness Current design provisions for the resistance to patch loading of

⁎
Corresponding author at: Universidad Nacional de Colombia, Facultad de Minas, Departamento de Ingeniería Civil, A.A. 75267 Medellín, Colombia.
E-mail address: cagracianog@unal.edu.co (C. Graciano).

https://doi.org/10.1016/j.istruc.2019.07.008
Received 28 May 2019; Received in revised form 18 July 2019; Accepted 20 July 2019
2352-0124/ © 2019 Institution of Structural Engineers. Published by Elsevier Ltd. All rights reserved.
C. Graciano, et al. Structures 20 (2019) 924–934

F bf

ss
y
tw
hw
x

tf

Fig. 1. Plate girder subjected to patch loading (Notation).

stainless steel plate girders [1] are derived from those of carbon steel
plate girders [2]. The provisions in the EC3 Part 1–5 [2] are based upon
a plastic collapse mechanism proposed by Lagerqvist and Johansson [3]
that considers an idealized elastic perfectly plastic material behavior,
which is typical for carbon steel. Stainless steel exhibits nonlinear
material strain-stress curves with significant strain hardening that
should be considered in the strength calculations for economical and
efficient structural designs [4–6]. Fig. 1 illustrates a typical setup for a
plate girder subjected to patch loading.
In 2003, Unosson et al. [7] conducted an experimental investigation
on the resistance of austenitic stainless steel plate girders subjected to
patch loading. Extending the slenderness of the girders, Unosson et al.
[8] also performed a numerical study to investigate the influence of
combined patch loading and bending, as well as the influence of the
of slender stainless steel plate girders. Next, a new safety factor is de-
termined according to the guidelines provided in international design
codes. Finally, the results show a significant improvement in the pre-
dicted resistances.
2. Resistance models in European standards
2.1. Resistance according to the EC3 Part1.4 [1] - stainless steel
In accordance with EC3 Part1.4 [1], the patch loading resistance for
stainless steel plate girders without intermediate stiffener FRd is calcu-
lated by
FRd = f yw Leff tw / (1)
M1

girder slenderness. The results showed that the predicted resistances where fyw is the yield strength of the web, and Leff is the effective length
calculated with a preliminary version of the EC3 Part 1.4 [9] were for resistance to transverse loads
conservative. Recently, dos Santos et al. [10] studied experimentally
and numerically the patch loading resistance of stocky welded plate
Leff = F ly (2)
girders manufactured of stainless steel. A comparison between experi- The reduction factor χF due to local buckling is
mental and numerical resistances with theoretical predictions was
0.5
conducted, the results showed that the latter were generally rather = 1.0
(3)
F
conservative, particularly for girders with decreasing web slenderness. F

These conservative results for the stockier sections were also attributed and the effective loaded length ly is given by
to neglecting the pronounced strain hardening exhibited by stainless
steel members with stocky webs. These behaviors have also been ob-
l y = ss + 2t f (1 + m1 + m2 ) a (4)
served for stocky stainless steel cross sections under other loading where m1 and m2 are dimensionless parameters
configurations and addressed by means of the deformation based
Continuous Strength Method (CSM) [4–6]. The CSM was originally m1 = f yf bf /f yw tw m2 = 0.02(h w /t f )2 (5)
developed for stainless steel and carbon steel materials, and is a de-
In Eq. (4), m2 should be taken as zero if F < 0.5.
formation-based design framework that allows for the beneficial influ-
The slenderness parameter F is obtained from
ence of strain hardening [4].
In recent years, the plate buckling rules in the EC3 Part 1–5 [2] are Fy
=
under review for harmonization of various buckling phenomena in F
Fcr (6)
plated structures [11]. Müller [12] introduced a proposal harmonizing
the resistance function shape χF for all verifications of structural com- Fy is the yield resistance
pressed members in the EC3 Part 1–5 [2]. This formulation comprises Fy = fyw tw l y (7)
some dimensionless parameters, which can be calibrated in order to
achieve a desired level of safety. After statistical evaluations of the The buckling load Fcr is
resistance function [12], these parameters have been calibrated for Fcr = 0.9kF Et w3 / h w (8)
carbon steel plate girders subjected to patch loading [13–15], leading to
improvements in the predicted resistances. hw 2

This paper aims at investigating the resistance to patch loading of kF = 6 + 2

slender austenitic stainless steel plate girders. At first, a dataset of ex-
perimental results is collected from the literature for comparative
analysis. Secondly, an extensive parametric study is conducted through
nonlinear finite element analyses in order to expand the available da-
tabase. Thereafter, a statistical evaluation of the results is performed in
order to calibrate a resistance function for the patch loading resistance
a (9)
2.2. Resistance function update for carbon steel girders
After consensus within the TWG 8.3 [15], any modification in the
resistance function should be based on the proposal presented by Müller
[12]. This proposal harmonizes the resistance function shape for all

925
C. Graciano, et al. Structures 20 (2019) 924–934

Table 1 girders tested by dos Santos et al. [10]. All the experiments were con-
αF0, F 0 and γM1 for different resistance models. ducted using austenitic stainless steel of three different grades: Grade
Proposal αF0 γM1 EN 1.4301 for tests 1 to 5, Grade EN 1.4404 for tests 6 to 10, and Grade
F0
EN 1.4571 for tests 11 to 21.
Müller [12] 0.34 0.8 – It must be mentioned that, the investigation conducted herein
Davaine [13] 0.21 0.8 1.1 considers girders subjected to only patch loading, therefore tests with
Gozzi [14] 0.50 0.6 1.0
combined bending moment and patch loading are discarded. The in-
Chacón et al. [15] 1.00 0.5 1.0
0.75 0.5 1.1 teraction with bending is taken into account when the applied moment
MS is 50% greater than the bending resistance MR of the girder. To
verify this, MS and MR are calculated following the procedure described
verifications of structural compressed members within the EC3 Part 1–5 in EC3 Part 1–4 [1], based on the effective width concept. Therefore,
[2]. A general form of this proposal is presented as follows from the test reported in Table 2 only 6 out of the 21 experiments are
taken into account: 5 from Unosson et al. [7] and only 1 (IOF-H152-
1.0
F = 1 L150-SS30) from dos Santos et al. [10].
(10)
+ 2
F

where 4. Nonlinear finite element model

1
= [1 + F0 ( F )+ F]
2
F0
(11)
In Eq. (11), the imperfection factor αF0 and the plateau length F 0 ,
need calibration in order to achieve a desired level of safety. Chacón
et al. [15] performed a statistical evaluation of a new proposal for the
patch loading resistance of carbon steel plate girders. Accordingly, the
study concluded that a greater plateau length F 0 leads to a less con-
servative formulation, whereas a greater imperfection factor αF0 gives
safer resistance predictions.
Table 1 presents values for imperfection factors αF0, the plateau
lengths F 0 and partial safety factors γM1 available for various patch
loading resistance models. For each proposal, αF0and F 0 were attained
after calibrating Eqs. (10) and (11), while γM1 was determined em-
ploying a standard evaluation procedure provided in EN 1990 [16].
Geometrically and materially nonlinear with imperfection analyses
(GMNIA) are performed to evaluate the influence of various geome-
trical parameters on the postbuckling behavior of the slender stainless
steel girders subjected to patch loading. To achieve that, a nonlinear
finite element model is developed using ANSYS [17]. The web panel,
flanges and vertical stiffeners of the girder are modeled using the four-
node shell element S181 [17] with six degrees of freedom at each node.
Material nonlinearities are considered in the stainless steel model. In
this case, a multilinear stress-strain curve was defined using the Eq. (12)
of true stress-strain curves with strain hardening provided in the EC3
Part 1.4-Annex C.2 [1].
n
= + 0.002
E 02 (12)

To validate the numerical model, computed ultimate resistances Fu

3. Previous experimental studies
As mentioned in the introduction, experimental studies have been
conducted to evaluate the resistance of stainless steel girders subjected
to patch loading. In this section, a brief description of experimental
results found in the literature is addressed. Table 2 summarizes di-
mensions and material properties of 21 available experimental tests, of
which, 5 correspond to girders tested by Unosson et al. [7], and 16
are compared against experimental loads Fexp obtained by Unosson [7].
Fig. 2 shows the material stress-strain curves corresponding to each
specimen tested in [7].
The whole I-girders are modeled, initial geometrical imperfections
are also considered within this analysis. All nodes in the web plate are
modified resembling the first eigenmode of a plate girder subjected to
patch loading. Hence, initial curvatures in both transverse and long-
itudinal directions of the web plate are considered. In spite of the fact

Table 2
Test studies of stainless steel girders subjected to patch loading.
# Test a hw tw bf tf ss σ02w σ02f Ew Ef Fexp

[mm] [mm] [mm] [mm] [mm] [mm] [MPa] [MPa] [GPa] [GPa] [kN]

1 Pli 4301:1 998 238 4.10 118.5 11.7 40.0 297 285 200 200 176.0
2 Pli 4301:2 996 238 4.10 119.9 11.9 80.0 297 285 200 200 196.0
3 Pli 4301:3 1397 316 4.10 121.0 11.9 40.0 297 285 200 200 168.0
4 Pli 4301:4 1623 438 4.10 121.2 11.9 40.0 297 285 200 200 169.0
5 Pli 4301:5 1682 401 8.80 120.4 12.0 40.0 245 285 200 200 478.0
6 IOF-H150-L150-SS60 150 150 6.95 75.8 9.9 60.0 274 267 197 197 424.4
7 IOF-H150-L200-SS60 200 150 6.87 75.8 9.8 60.0 274 267 197 197 393.1
8 IOF-H150-L300-SS60 301 150 6.88 75.8 9.8 60.0 274 267 197 197 368.6
9 IOF-H150-L400-SS60 402 150 6.81 75.7 9.8 60.0 274 267 197 197 342.2
10 IOF-H150-L450-SS60 452 150 6.87 75.7 9.8 60.0 274 267 197 197 340.0
11 IOF-H152-L150-SS30 150 152 6.20 160.0 8.7 30.0 272 227 191 205 340.0
12 IOF-H152-L300-SS30 301 153 6.18 159.0 8.8 30.0 272 227 191 205 322.2
13 IOF-H152-L450-SS30 451 152 6.22 159.6 8.7 30.0 272 227 191 205 301.1
14 IOF-H152-L600-SS30 598 152 6.18 159.6 8.9 30.0 272 227 191 205 296.7
15 IOF-H152-L750-SS30 751 152 6.13 159.8 8.7 30.0 272 227 191 205 275.0
16 IOF-H102-L300-SS5 299 101 4.89 67.9 5.1 5.0 222 222 187 187 126.7
17 IOF-H102-L300-SS7.5 299 102 4.98 67.9 5.2 7.5 222 222 187 187 132.3
18 IOF-H102-L300-SS10 299 101 4.92 67.8 5.2 10.0 222 222 187 187 121.8
19 IOF-H102-L300-SS12.5 298 101 4.94 67.9 5.2 12.5 222 222 187 187 143.2
20 IOF-H102-L300-SS15 298 102 4.99 67.8 5.1 15.0 222 222 187 187 130.8
21 IOF-H102-L300-SS20 299 101 4.91 67.8 5.1 20.0 222 222 187 187 142.5

Tests 1–5 come from Unosson [7]; Tests 6–21 come from dos Santos et al. [10].

926
C. Graciano, et al. Structures 20 (2019) 924–934

Fig. 4. Convergence study of the finite element mesh.

Fig. 2. Stress-strain curves of the specimens tested in [7].

Overall, the numerical model accurately captures the nonlinear re-

sponses of the slender stainless steel girders subjected to patch loading.
Table 3 shows a comparison between the experimental resistances Fexp
and those computed with the numerical model Fu, a good agreement is
found with an average difference mΔ = 6.83%, a standard deviation
sΔ = 1.33 and coefficient of variation equal to vΔ = 0.19.

5. Parametric study

5.1. General

According to the review performed in Section 3, only 6 tests sa-

tisfied the patch loading case. In order to expand this database, an
extensive parametric study is carried out. Table 4 presents the fixed
geometrical and material properties used herein, it must be pointed out
that these material properties were the same for all the components
(web and flanges) of the plate girder. Moreover, a maximum geome-
Fig. 3. Adopted finite element mesh. trical web imperfection of w0 = min(hw/100, tw) was included in the
numerical model, this value meets the recommendations described in
EC3 Part 1.4 - Annex C.2 [1], where the rules design for FE simulations/
that residual stress patterns have been defined for welded austenitic
modeling should be taken from EC3 Part 1.5 - Annex C [2]. These im-
and ferritic stainless steel sections [18], residual stresses were not ac-
perfections are large enough to attain a lower bound limit for design
counted for in the analysis since these have a diminished impact on
[22].
patch loading resistance [19–21]. For stress states involving the full
Table 5 summarizes the values of the geometric variables employed
cross-section of the elements, such as bending and compression, re-
in the parametric study. For each combination of variables presented in
sidual stresses can detriment the ultimate strength [22], but this is
Table 5, the ultimate resistance of the stainless steel girder was com-
barely the case for patch loading in which the stress state is rather lo-
puted, therefore a total of 7020 numerical simulations were performed.
calized.
These dimensions are similar to the girder proportions tested by Un-
As shown in Fig. 3, two boundary conditions are applied to the
osson [7]. It is worth quoting that, this section aims at investigating the
model, first, for the support ends, simple support conditions are de-
effect of girder dimensions, covering a wide range of slenderness, on the
fined, i.e. displacements in x and y directions were restricted in the
patch loading resistance of slender stainless steel I-girders. Therefore,
nodes along support A, and the nodes on support B were free to move
based on the effective width concept [1], the cross-section for all
longitudinally (x-direction). Next, the load is applied on the nodes lo-
modeled girders classifies as slender Class 4 cross-sections.
cated over an equivalent length ss in the upper flange, in which the
displacements in the x and z directions, and all rotations are restricted,
5.2. Effect of the panel aspect ratio a/hw
allowing only the vertical displacement in the y direction. Moreover,
the nonlinear response in the postbuckling region is accurately captured
Fig. 6 shows the computed resistances Fu in terms of the panel as-
using the Riks method [23], an arc-length based incremental method.
pect ratio a/hw for various patch loading lengths ss/hw and web slen-
This method allows the capture of the structural response considering
derness hw/tw. It can be seen that, the ultimate resistance decreases
the material nonlinearities and unstable collapse of the structure.
proportionally with the panel aspect ratio. When the patch loading
After conducting a convergence study (see Fig. 4) using as a basis
support is largely distributed over the loaded flange, the ultimate re-
the geometry of the second specimen (Pli 4301:2) tested by Unosson
sistance presents a nonlinear reduction as the panel aspect ratio is in-
[7], a finite element mesh with 4352 elements was chosen as shown in
creased, while for short patch loading lengths the decrease is almost
Fig. 3, i.e. a mesh with 2112 elements in the web panel, 880 elements in
linear.
each flange, and 240 elements in each transverse stiffener. Similarly to
Moreover, for large values of patch loading lengths Fu decays con-
[24], a maximum geometric imperfection of w0 = 0.003hw was used for
siderably when a/hw is increased. As seen in Fig. 6(f), for ss/hw = 0.50
the validation of the numerical model.
and hw/tw = 200, the ultimate resistance is reduced from Fu = 72.1 to
Fig. 5 presents a comparison between experimental and numerical
60.2 kN (a 16.5% of decrease), for an increasing panel aspect ratio from
load-deflection curves corresponding to the five specimens tested in [7]
a/hw = 1 to 2, respectively. On the other hand, for a short loading
(see Table 2). For all the specimens, the numerical model describes
length ss/hw = 0.10 and hw/tw = 200 (Fig. 6(f)), Fu is only reduced from
accurately the gradual loss of resistance in the postbuckling region.
46.9 to 43.4 kN (a 7.5% of decrease) once a/hw is increased from 1 to 2,

927
C. Graciano, et al. Structures 20 (2019) 924–934

Fig. 5. Experimental and numerical load-deflection curves comparison.

Table 3 Table 5
Comparison between experimental Fexp and computed resistances Fu. Varied geometrical dimensions in the numerical database.
# Test Fexp Fu Δ Variable hw [mm] a/hw hw/tw ss/hw tf/tw

[kN] [kN] % Value 400 1.00 75 0.10 1.00

600 1.25 100 0.15 1.25
1 Pli 4301:1 176.0 186.8 6.14 1.50 125 0.20 1.50
2 Pli 4301:2 196.0 208.0 6.12 1.75 150 0.25 1.75
3 Pli 4301:3 168.0 183.4 9.17 2.00 175 0.30 2.00
4 Pli 4301:4 169.0 179.3 6.09 200 0.35 2.25
5 Pli 4301:5 478.0 446.2 6.65 0.40 2.50
0.45 2.75
0.50 3.00
3.25
Table 4
3.50
Geometric and material properties of the plate girder. 3.75
bf [mm] E [GPa] υ σ0.2 n σ1.0 σu 4.00
[MPa] [MPa] [MPa]

120 200 0.30 285 9 334 611 increment when the web slenderness is reduced. As expected, girders
with low web slenderness ratios (thicker webs) show a better perfor-
mance for patch loading. This could be observed in Fig. 7(d) where for
respectively.
any value of flange-to-web thickness ratio tf/tw the resistance of a
thicker web (hw/tw = 75) subjected to a wide loading length of ss/
5.3. Effect of the web slenderness hw/tw hw = 0.40 is around 5.6 times the resistance of a slender girder (hw/
tw = 200).
Furthermore, the resistance Fu is also affected by the slenderness of It is interesting to notice that the resistance is also influenced by the
the web panel hw/tw. Fig. 7 illustrates the behavior of Fu in terms of hw/ thickness ratio tf/tw. The results show that increasing flange thickness
tw for various flange-to-web thickness ratios tf/tw and patch loading results in an enhanced resistance, this can be appreciated in Fig. 7(a) for
lengths ss/hw. In this case, the ultimate resistance presents a nonlinear

928
C. Graciano, et al. Structures 20 (2019) 924–934

Fig. 6. Ultimate resistance Fu in terms of the panel aspect ratio a/hw (hw = 400 mm, tf/tw = 2).

Fig. 7. Ultimate resistance Fu versus the web slenderness hw/tw (hw = 400 mm, a/hw = 1.50).

929
C. Graciano, et al.
Fig. 8. Ultimate resistance Fu as function of the loading length ss/hw (tf/tw = 2, a/hw = 1.50).
short lengths ss/hw = 0.10 and thicker webs (hw/tw = 75), where Fu
rises from 201.3 to 319.6 kN (an increment of 58.8%), when the
thickness ratio is goes from tf/tw = 1 to 4. This also occurs for slender
webs (hw/tw = 200) and wide lengths ss/hw = 0.40, in which the re-
sistance increases 56.2% for an increasing ratio tf/tw going from 1 to 4.
5.4. Effect of the patch loading length ss/hw
Fig. 8 depicts the influence of the patch loading length-to-web
height ratio ss/hw on the resistance for two web heights hw, and various
web height-to-web thickness ratios hw/tw. As observed, the ultimate
resistance rises with ss/hw, in this case the increase is almost linear. For
any value of hw and hw/tw, the numerical results show that the re-
sistance of a girder subjected to ss/hw = 0.10 is approximately 1.37 to
1.47 times the resistance of a girder with ss/hw = 0.50.
In addition to the influence of the patch loading length, the varia-
tion of the ultimate resistance is also related to the height of the girder.
As seen in Fig. 8, for any value of ss/hw a significant increment of
around 110% is achieved by increasing hw from 400 to 600 mm.
6. Statistical evaluation
6.1. Calibration of the resistance function χF
As mentioned earlier, it has been demonstrated that ultimate re-
sistance predictions for carbon steel girders subject to patch loading
improve when the resistance function χF is calculated using Eq. (10)
[15]. Hence, in this section a statistical calibration of the resistance
function is conducted using the numerical database obtained for slender
930
Structures 20 (2019) 924–934
stainless steel girders in Section 5. In this calibration, the imperfection
factor αF0 and the plateau length F 0 are adjusted in order to achieve a
desired level of safety. Following the same procedure employed by
Chacón et al. [15] a sensitivity analysis is performed. The examined
variable within this evaluation is the ratio between the resistance ob-
tained with the numerical model Fu and the resistance FRd⁎ predicted
with Eq. (3) (without the partial safety factor) using the new resistance
functions given in Eqs. (10) and (11).
Table 6 summarizes basic statistics corresponding to each evaluated
case of αF0 and F 0 . As observed in Table 6, for high values of the
plateau length F 0 the formulation tends to overestimate the ultimate
resistance. This can be appreciated in the statistics, where the percen-
tage of values lower than Fu/FRd⁎ = 1 increases considerably with F 0 ,
while the mean, maximum and minimum values of Fu/FRd⁎ decrease. On
the other hand, standard deviations and coefficients of variation
slightly rise for increasing plateau lengths. Concerning the imperfection
factor αF0, the statistical evaluation shows that the theoretical model
slightly underestimates the ultimate resistance when αF0 is high, i.e. the
formulation is more conservative. The results show that the standard
deviation for the ratio Fu/FRd⁎ increases proportionally with αF0.
Overall, Table 6 shows that for high F 0 and low αF0 values (bottom
left corner) the prediction of ultimate resistance is less conservative,
while for low F 0 and high αF0 values (top right corner) there is an
overestimation of the ultimate resistance. Based on these observations,
the combination of values that best fit the available results occurs when
plateau length is F 0 = 0.1 and the imperfection factor is αF0 = 0.4. The
goodness of the fit is demonstrated by observing the statistics of the
ratio Fu/FRd⁎, in which their corresponding values are acceptable
(max = 1.52; min = 0.99), with a low standard deviation of s = 0.09.
C. Graciano, et al. Structures 20 (2019) 924–934

Table 6
Statistics for the ratio between computed and predicted resistances Fu/FRd⁎ for slender austenitic stainless steel girders.
λF0 αF0

0.20 0.40 0.60 0.80

min max m s v % <1 min max m s v % <1 min max m s v % <1 min max m s v % <1

0.10 0.83 1.28 1.09 0.07 0.07 11.00 0.99 1.52 1.30 0.09 0.07 0.07 1.13 1.75 1.49 0.11 0.07 0.00 1.27 1.96 1.67 0.12 0.07 0.00
0.20 0.81 1.27 1.08 0.08 0.07 17.05 0.97 1.50 1.27 0.09 0.07 0.40 1.10 1.72 1.45 0.11 0.08 0.00 1.20 1.93 1.62 0.13 0.08 0.00
0.30 0.80 1.26 1.06 0.08 0.08 24.91 0.94 1.48 1.25 0.10 0.08 0.87 1.04 1.69 1.41 0.12 0.08 0.00 1.13 1.90 1.57 0.14 0.09 0.00
0.40 0.79 1.25 1.04 0.08 0.08 32.56 0.90 1.47 1.22 0.10 0.09 1.68 0.98 1.67 1.38 0.13 0.09 0.04 1.05 1.87 1.52 0.15 0.10 0.00
0.50 0.76 1.24 1.03 0.09 0.09 39.23 0.85 1.45 1.19 0.11 0.09 3.33 0.91 1.64 1.34 0.14 0.10 0.41 0.97 1.83 1.47 0.16 0.11 0.04
0.60 0.74 1.23 1.01 0.09 0.09 45.21 0.80 1.43 1.16 0.12 0.10 12.95 0.84 1.62 1.29 0.15 0.11 1.65 0.87 1.80 1.42 0.18 0.12 0.60
0.70 0.70 1.22 0.99 0.10 0.10 50.44 0.73 1.41 1.13 0.13 0.12 20.90 0.74 1.59 1.25 0.16 0.13 6.37 0.74 1.77 1.37 0.19 0.14 2.49
0.80 0.67 1.21 0.97 0.11 0.11 55.06 0.67 1.40 1.09 0.14 0.13 25.64 0.67 1.57 1.21 0.18 0.15 18.13 0.67 1.73 1.31 0.21 0.16 10.04
0.90 0.62 1.20 0.95 0.12 0.12 60.71 0.62 1.38 1.06 0.16 0.15 35.31 0.62 1.54 1.16 0.20 0.17 21.88 0.62 1.70 1.25 0.24 0.19 18.69

6.2. Partial safety factor γM1 V = exp(s 2) 1 (16)

In the previous section, the resistance function χF was calibrated. In with

order to guarantee the required safety level of the patch loading re- i = ln( i ) (17)
sistance model, the next step is to determine the safety factor γM1,
n
following the guidelines provided in the EN 1990 - Annex D [16]. This 1
=
factor considers the uncertainties of the geometrical and material n
i
(18)
i=1
properties related to the resistance model. The steps to calculate the
partial safety factor are described as follows 1
n
s2 = ( i )2
n 1 i=1 (19)
- First, a design model is developed using the following expression
Additionally, a normality test should be carried out. To this purpose,
rt = grt (X )
_ (13) a Kolmogorov-Smirnoff test was performed to verify the normality of
the distribution of the errors.
where grt is the resistance function of the numerical model and X is the
_
array of values of the basic variables. - Thereafter, the coefficients of variations Vxi for the basic variables
(geometry and material) were defined. In this case, based on the
- Next, numerical resistances re and theoretical resistances rt are statistical evaluation of data collected from the steel producers and
compared (see Fig. 9) manufacturers of stainless steel sections conducted by Baddoo and
Francis [25], the coefficients used for the material and geometry
With the purpose of comparing the deviation of the predicted re- were: Vx1 = 0.105 for austenitic steel material and Vx2 = 0.0214 for
sults, the least squares best-fit to the slope (b) is calculated I-section stainless steel member subjected to compression.
re rt
b=
rt2 (14) Following the recommendations of Davaine [13], an additional
coefficient that considers the deviations between the numerical model
and the error term δi is determined for each value of numerical re- and the experimental results was included in this calibration. The
sistance rei and theoretical resistance rti procedure to obtain the coefficient of variation was the same as the one
=
rei followed to calculate the coefficient of variation Vδ. In this case, the
i
b rti (15) value obtained was VFEM = 0.0763.

- Subsequently, the coefficient of variation of the error is computed as - Then, the final coefficient of variation is computed in order to take
into account all variations in one parameter
j
Vr2 = (V 2 + 1) (Vxi2 + 1) 2
1 + VFEM
i=1 (20)
-Furthermore, the design and characteristic resistances (rd, rk) are
calculated using the following expressions

rd = bgrt ( Xm) e ( k Q 0.5Q2)

_ (21)

rk = bgrt ( Xm ) e ( k Q 0.5Q2)
_ (22)
where Xm is the mean array of values of the basic variables and Q is
_
computed as follows

Q= ln(Vr2 ) + 1 (23)
with k∞ equal to 3.04 and 1.64 for design and characteristic resistance
Fig. 9. Numerical vs. theoretical resistances. [16], respectively. The partial safety factor is obtained dividing rk by rd

931
C. Graciano, et al. Structures 20 (2019) 924–934

Table 7
αF0, F 0 and γM1 obtained from the statistical evaluation.
αF0 F0 γM1

0.40 0.10 1.05

rk
=
(24)
M1
rd

and latter corrected with the purpose of adjusting for better statistical
variations
rn
=
M1 M1
rk (25)

with rn as the nominal resistance, calculated employing Eq. (26) with

Xn known as the array of nominal values of the basic variables
_
rn = bgrt ( Xn ) e ( k Q 0.5Q 2)
_ (26)

After following the procedure presented above, the partial safety

factor obtained for the numerical database was γM1∗ = 1.05. This result
shows correspondence and reliability with the partial safety factors
provided in the EC3 Part 1.4 [1], where the resistance models for
stainless steel cross sections under local buckling employ partial factors
less or equal than 1.10.

6.3. Comparison of the revised formulation with the numerical results

A comparison between numerical results derived from the para-

metric study and ultimate resistances computed using Eq. (1), in com-
bination to the newly calibrated resistance function for stainless steel
girders developed in Section 6.1, is carried out. Table 7 summarizes the
imperfection factor αF0 and the plateau length F 0 corresponding to the
calibration of the resistance function, along with the partial safety
factor γM1 obtained for stainless steel girders subjected to patch loading.
On the one hand, Fig. 10 displays the ratio between resistances
computed numerically Fu and resistances FRd calculated using the cur-
rent provisions of the EC3 Part 1.4 [1] versus various geometrical re-
lationships. On the other hand, Fig. 11 shows the same relationship, but
FRd is calculated using the resistance function in Eq. (10) and con-
sidering the values reported in Table 7 (with the already computed
partial safety factor). The basic statistics for the ratio Fu/FRd corre-
sponding to both ultimate resistance approaches are presented in
Table 8.
As observed in Fig. 10, there is a significant scatter in the ratio Fu/
FRd (max = 2.18; min = 1.02), when the ultimate resistance is pre-
dicted with the current provisions in the EC3 Part1.4 [1]. As seen in
Fig. 10(a) and (c), the results indicate that there is no influence of the
panel aspect ratio a/hw and the patch loading length ss/hw on the ulti-
mate load ratio Fu/FRd. However, the results also show that the ratio Fu/
FRd rises for increasing values of web slenderness hw/tw and thickness
ratio tf/tw (Fig. 10(b) and (d)). Additionally, it is interesting to observe
that when the ultimate resistance is computed using the modified
proposal (see Fig. 11) the scatter in the load ratio Fu/FRd is significantly
reduced. This is clearly observed with the statistics for the ratio Fu/FRd
in Table 8 with a mean ratio m = 1.36 and a standard deviation
s = 0.09.
Once the dimensionless parameters of the resistance function have
been calibrated, the resistance curve is evaluated. To this purpose,
Fig. 12 presents a comparison of the numerical results Fu/Fy and the
Fig. 10. Fu/FRd in terms of various geometrical parameters (FRd calculated ac-
resistance curves of the current EC3 Part1.4 formulation [1] and the
cording EC3: 1–4 provisions [1]).
resistance curve calibrated herein. As expected, there is an increase of
Fu/Fy when the slenderness parameter is reduced.
It is also observed that, the resistance curve of the EC3 Part1.4 [1] subjected to patch loading. While, Eq. (10) corresponding to the re-
which is plotted using Eq. (3) and originally developed for carbon steel sistance curve calibrated in Section 6.1 presents an acceptable lower
girders underestimates the resistance of the stainless steel girders bound for the resistance values obtained numerically.

932
C. Graciano, et al. Structures 20 (2019) 924–934

Table 8
Statistical values of Fu/FRd.
Approach min max m s v

EC3 Part1.4 [1] 1.02 2.18 1.61 0.19 0.12

Revised 1.04 1.59 1.36 0.09 0.07

Fig. 12. Fu/Fy versus the slenderness parameter.

7. Conclusions

This paper presents a comprehensive evaluation of the resistance of

slender austenitic stainless steel I- girders subjected to patch loading. To
this purpose, an extensive parametrical evaluation with more than 7000
numerical simulations was carried out, in order to present some design
recommendations of the effect of several geometrical parameters on the
ultimate resistance. Based on these results, a revised formulation to
predict the ultimate resistance is discussed. From the analyses con-
ducted herein, the following conclusions are drawn:

• The ultimate resistance of slender austenitic stainless steel plate

girders enhances significantly when the patch loading length-to-web
height ratio ss/hw is incremented, and also when the web panel
height is increased.
• The results have demonstrated that is possible to improve the ulti-
mate resistance predictions when the non dimensionless parameters
used to calculate the resistance function are calibrated for stainless
steel girders subjected to patch loading. Moreover, the results also
show the reliability of the resistance model, this is established after
obtaining a partial safety factor within the limits established in the
EC3 Part1.4 [1] for stainless steel cross sections under local buck-
ling.
• Only slender stainless steel plate girders have been analyzed herein,
therefore an additional work should be performed for girders with
low web slenderness ratios.
• Finally, the resistance model was validated using a limited number
of experimental results available in the literature, therefore more
tests are necessary to fully validate it.

Declaration of Competing Interest

All authors have participated in (a) conception and design, or

Fig. 11. Fu/FRd in terms of various geometrical parameters (FRd calculated with
modified proposal).
analysis and interpretation of the data; (b) drafting the article or re-
vising it critically for important intellectual content; and (c) approval of
the final version. This manuscript has not been submitted to, nor is
under review at, another journal or other publishing venue. The authors
have no affiliation with any organization with a direct or indirect fi-
nancial interest in the subject matter discussed in the manuscript.
References

[1] EN 1993-1-4 Eurocode 3. Design of steel structures - part 1–4: general rules –
supplementary rules for stainless steels CEN. Brussels: European Committee for

933
C. Graciano, et al.
Standardization; 2006.
[2] EN 1993-1-5. Eurocode 3: design of steel structures – part 1–5: plated structural
elements. Brussels: European Committee for Standardization (CEN); 2006.
[3] Lagerqvist O, Johansson B. Resistance of I-girders to concentrated loads. J Constr
Steel Res 1996;39(2):87–119.
[4] Afshan S, Gardner L. The continuous strength method for structural stainless steel
design. Thin-Walled Struct 2013;68:42–9.
[5] Ahmed S, Ashraf M, Anwar-Us-Saadat M. The continuous strength method for
slender stainless steel cross-sections. Thin-Walled Struct 2016;107:362–76.
[6] Zhao O, Afshan S, Gardner L. Structural response and continuous strength method
design of slender stainless steel cross-sections. Eng Struct 2017;140:14–25.
[7] Unosson E. Patch loading of stainless steel girders: experiments and finite analyses
Licentiate thesis Division of steel Structures, Luleå University of Technology; 2003.
[12, February 2003].
[8] Unosson E, Olsson A. Stainless steel girders: Resistance to concentrated loads and
shear. In stainless steel in structures: International experts seminar 20/05/2003-20/
05/2003. Stainless steel in structures: International experts seminar 20th may 2003,
Ascot, UK. Proceedings. 2003. p. 123–30.
[9] prEN 1993-1-4. Eurocode 3: design of steel structures - part 1–4: general rules –
supplementary rules for stainless steels CEN. Brussels: European Committee for
Standardization; 2002.
[10] dos Santos GB, Gardner L, Kucukler M. Experimental and numerical study of
stainless steel I-sections under concentrated internal one-flange and internal two-
flange loading. Eng Struct 2018;175:355–70.
[11] Johansson B, Veljkovic M. Review of plate buckling rules in EN 1993-1-5. Steel
Constr 2009;2(4):228–34.
[12] Müller C. Zum Nachweis ebener Tragwerke aus Stahl gegen seitliches Ausweichen
Doctoral Thesis Shaker Verlag: RWTH Aachen, Lehrstuhl für Stahlbau; 2003. No.
47. [In German].
[13] Davaine L. Formulation de la résistance au lancement d'une âme métallique de pont
934
Structures 20 (2019) 924–934
raidie longitudinalement Doctoral Thesis D05–05 France: INSA de Rennes; 2005.
[In French].
[14] Gozzi J. Patch loading resistance of plated girders – ultimate and service ability
limit state Doctoral Thesis Luleå University of Technology, Division of Steel
Structures; 2007. 2007:30, Luleå, Sweden.
[15] Chacón R, Mirambell E, Kuhlmann U, Braun B. Statistical evaluation of the new
resistance model for steel plate girders subjected to patch loading. Steel Constr
2012;5(1):10–5.
[16] EN 1990. Eurocode – basis of structural design. 2002.
[17] ANSYS. Ansys release 19 elements reference. 2018.
[18] Yuan HX, Wang YQ, Shi YJ, Gardner L. Residual stress distributions in welded
stainless steel sections. Thin-Walled Struct 2014;79:38–51.
[19] Granath P. Behavior of slender plate girders subjected to patch loading. J Constr
Steel Res 1997;42:1–19.
[20] Chacon R, Mirambell E, Real E. Influence of designer-assumed initial conditions on
the numerical modelling of steel plate girders subjected to patch loading. Thin-
Walled Struct 2009;47:391–402.
[21] Chacon R, Serrat M, Real E. The influence of structural imperfections on the re-
sistance of plate girders to patch loading. Thin-Walled Struct 2012;53:15–25.
[22] Couto C, Real PV. Numerical investigation on the influence of imperfections in the
local buckling of thin-walled I-shaped sections. Thin-Walled Struct 2019;135.
(89–08).
[23] Riks E. An incremental approach to the solution of snapping and buckling problems.
Int J Solids Struct 1979;15:529–51.
[24] Unosson E, Olsson A, Lagerqvist O. A numerical study of the resistance of stainless
steel girders subjected to concentrated forces. Proc. Nordic steel const conf 18–20th
June 2001 Finland, Helsinki. 2001. p. 799–806.
[25] Baddoo NR, Francis P. Development of design rules in the AISC design guide for
structural stainless steel. Thin-Walled Struct 2014;83:200–8.