Вы находитесь на странице: 1из 8

Journal of Constructional Steel Research 66 (2010) 13661373

Contents lists available at ScienceDirect

Journal of Constructional Steel Research


journal homepage: www.elsevier.com/locate/jcsr

Compression strength of corroded steel angle members


L.-V. Beaulieu, F. Legeron , S. Langlois
Department of Civil Eng., University of Sherbrooke, 2500 boul. de lUniversit, Sherbrooke, Quebec, Canada J1K 2R1

article

info

Article history:
Received 6 November 2009
Accepted 17 May 2010
Keywords:
Corrosion
Steel structures
Compressive strength
Compression testing
Structural integrity
Residual capacity

abstract
Steel structures corrode when exposed to the environment, and their capacity is reduced accordingly. In
practice, when a member is found corroded during inspection, it is necessary to estimate the residual
capacity of corroded members in order to decide whether to change the member, repair it or just remove
corrosion and re-protect the member.
The objective of this article is to provide data to engineers on the structural behavior of corroded
steel angle members under compressive load. Sixteen angle members were corroded with an accelerated
procedure and then tested in compression. Eight uncorroded members were also tested in compression.
The influence of corrosion on compressive capacity was measured and compared to analytical methods
accounting for weight loss. Recommendations are drawn from this research to provide guidance to
engineers on how to evaluate compressive capacity of corroded members. Needs for future work are also
highlighted.
2010 Elsevier Ltd. All rights reserved.

1. Introduction
Steel structures exposed to the environment are subject to
corrosion. Even galvanized steel can experience corrosion after
galvanic protection is consumed. Corrosion can be particularly
important in transmission line foundations where steel angle
members are often placed in direct contact with the soil and
exposed to variable level of water table. It is also the case of many
industrial facilities, antenna towers, bridges, and other exposed
structures.
For utilities which manage very large structure networks, it is
very difficult to assess the residual capacity of an existing structure
and determine when it is no longer safe after corrosion has started.
One method based on qualitative parameters was developed by
Hathout [1] to evaluate the reliability of transmission lines. To
examine the problem in detail, it would be useful to develop a
model predicting the residual capacity of corroded members.
The traditional approach to evaluate residual capacity is to
perform visual inspection of the corroded members and classify
the members according to their level of damage. This method is
however highly subjective. For members loaded in compression,
the precision obtained by this method can be inadequate because
the capacity is very sensitive to geometrical imperfections.
Kayser and Nowak [2] developed a corrosion model for steel
girder bridges, which takes into account the location and rate

Corresponding author. Tel.: +1 819 821 7395; fax: +1 819 821 7974.
E-mail addresses: Louis-Vincent.Beaulieu@USherbrooke.ca (L.-V. Beaulieu),
Frederic.Legeron@USherbrooke.ca (F. Legeron),
Sebastien.Langlois@USherbrooke.ca (S. Langlois).
0143-974X/$ see front matter 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jcsr.2010.05.006

of corrosion. Sarveswaran et al. [3] evaluated structural safety


of corroded steel structures by applying the theory of structural
reliability. The model predicts an interval of residual capacity
based on an estimate of the remaining thickness of the member.
In [4], minimum capacity curves which vary with the percentage
loss of flange thickness are presented. These curves account for
the principal modes of failure. However, variation in thickness
due to corrosion and local concentration of corrosion are not
considered in this approach. Corrosion might also modify steel
from a metallurgical point of view and characteristics of remaining
steel, such as yield strength, can be altered. The phenomenon of
corrosion is well studied but, very little research has been done on
how the compressive capacity of steel members is affected once
corrosion has developed. There is a lack of experimental data on
the relation between weight loss due to corrosion and residual
strength.
The objective of this study is to provide experimental data on
compressive capacity of corroded steel angle members that could
be used by practical engineers. The simplified method to calculate
the residual capacity of members according to the average residual
thickness is evaluated.
2. Experimental program
In order to evaluate the residual capacity of corroded angle
members, an experimental program was performed. The main
parameters of this study are: (i) the overall slenderness ratio (L/r ),
(ii) the width-to-thickness ratio (b/t ), and (iii) the corrosion level.
The slenderness ratio (L/r ) is one of the main factors affecting
the capacity of steel members under compression load. This is
the ratio of the unsupported length L to the radius of gyration,
r. The length of the test specimens was selected such that two

L.-V. Beaulieu et al. / Journal of Constructional Steel Research 66 (2010) 13661373

1367

Nomenclature
A
Aeff
b
beff
b/t
E
Fa
Fe
Fcr
Fy
KL
KL/r
L
t
r

Section area
Effective area
Leg width
Effective leg width
Width-to-thickness ratio
Modulus of elasticity
Allowable compression stress
Elastic critical buckling stress
Effective yield stress
Specified minimum or measured yield stress
Effective length
Effective slenderness ratio
Length of member
Leg thickness
Radius of gyration
Flat width (b t )

levels of this variable are investigated: (i) slenderness ratio of


approximately 40, and (ii) slenderness ratio of approximately 110.
The width-to-thickness ratio (b/t ) is defined as the ratio of
the leg width to the leg thickness. This ratio affects the class
of the member based on the definitions of the CAN/CSA S1609 standard [5], and therefore the method for calculating the
compression capacity of the member. The two levels of width-tothickness ratio are 6.66 and 13.3. With b/t = 6.66, the member is
supposed to yield before local buckling in compression happens.
These members are classified as Class 1, 2 or 3 in CAN/CSA S1609. On the contrary with b/t = 13.3, the member is supposed
to be subjected to local buckling before yielding in compression.
The latter type of member is classified as Class 4 in CAN/CSA S1609. In AISC standard 2005 [6], the angles in compression only with
b/t = 6.66 is classified as non compact and angles with b/t =
13.3 is classified as slender elements. Even if the classification
follows different wording, both standards recognize that member
with b/t = 6.66 should buckled globally before local buckling
is reached whereas member with b/t = 13.3 may reach local
buckling first.
The corrosion level is measured as a mass loss percentage.
Three levels of this variable were investigated: 0% (not corroded),
25% (moderate corrosion), and 40% (severe corrosion). Galvanized
members were used in this project because most angle members
exposed to the environment are galvanized and it could have an
effect on the development of corrosion.
Table 1 shows all specimens and the level for each variable
mentioned above, as well as the section, length, S16-09 class,
and gross area. Each specimen is named after its leg thickness
(first two digits), length (L for long and S for short), approximate
corrosion level (0%, 25% or 40%), and specimen number (1 or 2).
Two replicate specimens of each combination of leg thickness,
length, and corrosion level were tested.
2.1. Accelerated corrosion procedure
The steel members were corroded by the galvanic corrosion
process [7]. The members were immersed in a conductive solution
saturated with copper sulphate, and connected to the positive node
of an electric circuit. The galvanized steel piece then plays the role
of the anode. A copper plate cathode was also immersed in the
solution and connected to the negative node of the circuit. Current
was applied with a constant intensity generator to oxidize the
steel member (from the anode to the cathode). Current intensity
and voltage were monitored. Standard assembly for the corrosion
process is shown in Fig. 1.

Fig. 1. Standard assembly for corrosion.

The steel members were submitted to the process for durations


between 250 and 1820 h. The level of corrosion was determined by
the ratio of the mass of the steel member altered during corrosion
to the mass of the intact steel member. Measurements of the leg
width (b) and leg thickness (t ) were taken during and after the
corrosion process at three or four locations along the member,
depending on the member length. The development of corrosion
was verified regularly on the members.
2.2. Experimental set-up
The experimental set-up developed at the University of
Sherbrooke to study the compression capacity of single and X
diagonal bracing members was used here to assess the strength in
compression of the corroded members. The set-up was developed
to represent the type of connection used in practice with gusset
plates. Computation of end flexibility have shown that for a wide
range of members, the boundary conditions of the experimental
set-up replicate adequately the type of flexibility found in practice
in bracings. A sketch of the test set-up is shown in Fig. 2. Further
details can be found in [8].
Compressive tests were performed under displacement control
using a 500 kN hydraulic actuator in order to develop a stable
post-buckling behavior. The angle members were connected to the
frame with bolted connections on one leg only (three bolts per
connection). Holes in the angle members were drilled after the
corrosion process.
As shown in Fig. 3, the displacement in the axial (z-axis),
in-plane buckling (x-axis) and out-of-plane buckling (y-axis)
directions were measured during the tests to observe the nonlinear behavior of the steel members. The axial displacement
was evaluated using a potentiometer fixed at the centre bolt of
both connections. The displacement in the y-axis at the centre of
the angle member was measured by a linear variable differential
transformer (LVDT) and adjusted relative to the displacement
at the connections, which were evaluated with potentiometers
aligned with the centre bolt of each connection. Similarly, the
displacement in the x-axis was measured by a LVDT at the centre
of the member and adjusted relative to the displacement at the
connections measured by a LVDT.

1368

L.-V. Beaulieu et al. / Journal of Constructional Steel Research 66 (2010) 13661373

Table 1
Test specimens.
Specimen

Section

L (mm)

Nominal % of corrosion

b/t

Class

L/r

A (mm2 )

48S-00-1
48S-00-2
48L-00-1
48L-00-2
95S-00-1
95S-00-2
95L-00-1
95L-00-2
48S-25-1
48S-25-2
95S-25-1
95S-25-2
48L-25-1
48L-25-2
48S-40-1
48S-40-2
95S-40-1
95S-40-2
95L-25-1
95L-25-2
95L-40-1
95L-40-2
48L-40-1
48L-40-2

L64 64 4.8
L64 64 4.8
L64 64 4.8
L64 64 4.8
L64 64 9.5
L64 64 9.5
L64 64 9.5
L64 64 9.5
L64 64 4.8
L64 64 4.8
L64 64 9.5
L64 64 9.5
L64 64 4.8
L64 64 4.8
L64 64 4.8
L64 64 4.8
L64 64 9.5
L64 64 9.5
L64 64 9.5
L64 64 9.5
L64 64 9.5
L64 64 9.5
L64 64 4.8
L64 64 4.8

500
500
1358
1358
500
500
1358
1358
500
500
500
500
1358
1358
500
500
500
500
1358
1358
1358
1358
1358
1358

0
0
0
0
0
0
0
0
25
25
25
25
25
25
40
40
40
40
25
25
40
40
40
40

13.3
13.3
13.3
13.3
6.66
6.66
6.66
6.66
13.3
13.3
6.66
6.66
13.3
13.3
13.3
13.3
6.66
6.66
6.66
6.66
6.66
6.66
13.3
13.3

4
4
4
4
1
1
1
1
4
4
1
1
4
4
4
4
1
1
1
1
1
1
4
4

39.7
39.7
108
108
40.3
40.3
110
110
39.7
39.7
40.3
40.3
108
108
39.7
39.7
40.3
40.3
110
110
110
110
108
108

582
582
582
582
1120
1120
1120
1120
582
582
1120
1120
582
582
582
582
1120
1120
1120
1120
1120
1120
582
582

Table 2
Results of the corrosion process and compression tests.
Specimen

Actual % of
corrosion

t (average) (mm)

t (standard
deviation)

b/t (mm) Class KL/r (S16 and


AISC)

KL/r
(ASCE)

A (mm2 )

Exp.
capacity (kN)

Failure
modea

48S-00-1
48S-00-2
48L-00-1
48L-00-2
95S-00-1
95S-00-2
95L-00-1
95L-00-2
48S-25-1
48S-25-2
95S-25-1
95S-25-2
48L-25-1
48L-25-2
48S-40-1
48S-40-2
95S-40-1
95S-40-2
95L-25-1
95L-25-2
95L-40-1
95L-40-2
48L-40-1
48L-40-2

0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
26.4
26.6
26.9
25.3
27.4
27.3
37.2
37.3
37.1
36.1
22.4
21.5
30.7
27.2
40.7
47.1

4.76
4.76
4.76
4.76
9.53
9.53
9.53
9.53
4.32
4.18
7.28
7.18
3.86
3.84
3.51
3.37
6.43
6.25
7.80
7.98
6.79
7.08
3.28
2.63

0.45
0.41
0.50
0.47
0.19
0.27
0.24
0.39
0.51
0.54
0.33
0.37
0.49
0.88
0.24
0.39

13.3
13.3
13.3
13.3
6.66
6.66
6.66
6.66
14.6
15.1
8.65
8.77
16.3
16.4
17.9
18.7
9.80
10.1
8.08
7.89
9.28
8.90
19.2
24.0

79.9
79.9
114
114
80.2
80.2
115
115
80.0
80.0
80.3
80.2
114
114
79.9
79.9
80.2
80.2
115
115
115
115
114
114

582
582
582
582
1120
1120
1120
1120
526
509
864
853
471
469
430
413
769
748
922
942
809
842
403
324

n/a
114
91
88
323
335
168
167
57
n/a
253
251
80
79
66
63
216
202
151
150
99
99
65
25

3
2
1
4
1
1
1
4
3
2
1
1
1
1
2
2
1
2
1
1
3
1b
1
1b

a
b

4
4
4
4
1
1
1
1
4
4
2
2
4
4
4
4
3
3
2
2
3
2
4
4

90.9
90.9
123.4
123.4
91.6
91.6
125.3
125.3
91.1
91.0
91.5
91.5
123.6
123.6
90.9
90.9
91.4
91.3
125.1
125.1
124.7
124.8
123.4
123.1

Legend for failure modes: 1-global buckling, 2-local buckling near connections, 3-local buckling near centre, 4-mode not clearly identified in the test.
Perforated members.

3. Experimental results
3.1. Geometry of corroded members
The steel members reached corrosion levels between 21% and
47% when submitted to the accelerated corrosion process. The
corrosion level is measured in terms of weight loss. Each member
was weighed before being corroded. During accelerated corrosion
and at the end, the member was cleaned from corrosion with a
brush and weighed. Half of the specimens are displayed in Fig. 4,
showing the samples after the compression test. Two members
were perforated and are displayed in Fig. 5. One of the legs of
member 95L-40-2 was very lightly perforated near the centre.
Member 48L-40-2 nearly lost an entire leg over a short length near

the centre. The circular hole in this member was drilled before the
corrosion process.
The main properties and the S16-09 class of the specimens after
corrosion are shown in Table 2. The leg width (b) for corroded
members was approximately constant at 63 mm for all specimens.
The thickness of the member was also measured on each leg at
three locations for short members and four locations for long
members to provide an average thickness to be used to predict
the residual strength. In practice, weight measurements are not
an option, but it is common to measure the residual thickness and
various measuring tools are available for this purpose. Hence, to
insure that the predicted strength is obtained in a realistic manner,
the parameters b/t, KL/r and A from Table 2 were calculated based
on the average measured thickness of the corroded members.
In this table r is the radius of gyration around the minor axis.

L.-V. Beaulieu et al. / Journal of Constructional Steel Research 66 (2010) 13661373

1369

Fig. 2. Elevation of the compression test assembly.

Fig. 3. Position of the instruments.


Fig. 5. Perforated members 95L-40-2 and 48L-40-2.

To provide more information on the uniformity of the corrosion,


the thickness of corroded members was measured at each point of
the grid shown in Fig. 7. The standard deviation on thickness for
all corroded members is shown on Table 2. It is interesting to note
that the standard deviation on thickness is not clearly related to
the initial member thickness and to the corrosion level, although
there is a lot of variability.
3.2. Compression tests

Fig. 4. Selection of corroded members after the compression.

The ratio of average thickness over nominal thickness is graphed


with respect to the level of corrosion in Fig. 6. Error bars are
also present in this figure to show the minimum and maximum
thicknesses measured on the members and illustrate the variability
of corrosion. Fig. 6 shows that the average leg thickness is
correlated to the level of corrosion (measured by weight loss) and
therefore it can be used as a practical measure of corrosion.

Fig. 8 shows the results of the tests in terms of ratio between


residual strength and uncorroded strength. The value of the
uncorroded strength is taken as the average of the experimental
strength of the two members of the same series that are not
corroded. Table 2 presents the capacity obtained during the tests
and the observed failure modes. It is observed in Fig. 8 that
the rate of capacity loss tends to increase with the level of
corrosion. Indeed, most points at low corrosion levels are above
the reference line whereas most points at high corrosion level
are below the reference line. The reference line supposes a linear
relation between level of corrosion and capacity loss. The two
replicate specimens generally gave very similar results. The type of
failure was observed to be either local or global and the buckling
occurred at various locations along the members. It should be
noted that the classification of failure modes is not straightforward
and therefore these observations should be analyzed with care.
Nevertheless, the following comments can be made:

48S members generally showed local buckling.


The three other types of members generally displayed global
buckling, except for some severely corroded members.

1370

L.-V. Beaulieu et al. / Journal of Constructional Steel Research 66 (2010) 13661373

parameter K . We will first present the column strength curves used


by the three codes, and after the estimation of K .
ASCE 10-97 proposes a strength equation accounting for both
local and global buckling. The allowable compression stress Fa is
given by Eqs. (1)(3).


Fa = 1

1
2

Cc =

Fig. 6. Ratio of average thickness to nominal thickness versus level of corrosion.

In two cases, as shown in Fig. 5, complete perforation of the


member was observed. In the case of member 48L-40-2, the
perforation was major and observations of the failed member
show that the perforation affected the overall compression
capacity. Specimens 48S-00-1 and 48S-25-2 did not produce any
compressive capacity results due to difficulties in the experimental
procedure.
The radius of gyration varied moderately due to the moderate
change of the leg width which affects the inertia of the member.
The slenderness ratio is directly dependent on the variation
of the radius of gyration and therefore was subject to minor
modifications only. However, the section area of the members
varied significantly, causing major changes to the compressive
capacity of the steel angle members.

2E
Fy

KL

Cc

> Cc

(1)

(2)

(3)

To account for local buckling of angle sections, Fy is replaced


in Eqs. (1)(3) by Fcr for large width-to-thickness ratios w/t. For
simplicity, the flat width w is defined in the present study as the
leg width b minus the thickness t. Eqs. (4)(6) present the various
expressions for Fcr and their corresponding width-to-thickness
ratios.

w
t

80

lim

= p

(4)

Fy


w/t
= 1.677 0.677
Fy
(w/t )lim
w
w
144
for

p


Fcr

Fcr =

lim

0.0332 2 E

for

(w/t )2

Fy

144

(5)

(6)

Fy

with = 2.62 for SI units. Similarly to ASCE10-97, AISC 2005 has


included in the column curve the leg slenderness effect through the
use of parameter Q :

(a)

KL
r

s
4.71

Fcr = Q 0.658

(b)

KL
r

E
QFy

The CAN/CSA S16-09 standard [5], AISC 2005 [6] and the ASCE
10-97 standard [9] were used to calculate the theoretical capacity
in compression of the corroded steel members based on the
average leg thickness measured in the laboratory.
The modulus of elasticity E assumed in the calculations is
200,000 MPa. The yield stress Fy is 365 MPa for L64 64 4.8
members, and 345 MPa for L64 64 9.5 members as measured on
coupons cut from the members before the corrosion process. The
yield stress was measured as per the ASTM E8-61T standard [10].
The three codes use a column strength curve where maximum
stress is related to effective slenderness KL/r. In order to account
for the way the loads are applied in a real single angle bracing
(loads not applied at centroid and retraints different from simple
support at end) the three codes propose equations to estimate

for

4. Analysis of results
4.1. Strength prediction

Fy

KL

for



Cc

2E

KL 2

Fa =

KL/r

QFy
Fe

s
> 4.71

or Fe 0.44QFy

(7a)

or Fe < 0.44QFy

(7b)


Fy
E

QFy

Fcr = 0.877Fe
where Q is calculated from the ratio b/t:

(i) When

b
t

s
0.45
s

(ii) When 0.45

Fig. 7. Sketch of thickness measurements.

E
Fy

E
Fy

<

b
t

Q = 1.

s
0.91

(8a)
E
Fy

L.-V. Beaulieu et al. / Journal of Constructional Steel Research 66 (2010) 13661373

Fig. 8. Experimental percentage of residual capacity in compression versus level of


corrosion.

Q = 1.34 0.76

(iii) When

b
t

 r

s
> 0.91

Fy

E
Fy

Q =

(8b)
0.53E
Fy

 .

b 2
t

(8c)

The CAN/CSA S16-09 provides only a strength curve accounting


for global buckling. The local buckling is accounted for through
a reduction of the member cross-section named effective area.
To calculate the capacity of the members as per CAN/CSA S16-09
standard (2005), the effective area Aeff has to be calculated for all
class 4 angle members. Please note that the area shown in Table 2
is the area evaluated from the measured leg thickness and not
the effective area. Eq. (9) below represents the limit of width-tothickness ratio that must be respected in the S16-09 standard for
class 3 members. If it is not respected, the member is class 4, and
a value of beff is determined such that beff /t respects the criteria of
Eq. (9).
b
t

200

p .

(9)

Fy

An effective area is then calculated using Eq. (10).


Aeff = A 2(b beff )t .
The capacity of the angle is therefore calculated with
Cr = AFy 1 + 2n

1/n

(11)

n = 1.34

KL
r
KL
r

= 60 + 0.5

L
r

for 0

= 46.2 + 0.615

L
r

L
r

120

for 120

L
r

(12)

250.

(13)

Both effects are found in the present study, but in agreement with
ASCE 10-97, it is assumed that for L/r below 120, the eccentricity
of connections governs and Eq. (12) is used to calculate KL/r. AISC
2005 and S16-09 have introduced such a method to calculate the
effective length. Both codes have the same equations:

L
rx

80 :

> 80 :

KL
r

KL
r

= 72 + 0.75

= 32 + 1.25

L
rx

L
rx

200.

(14a)
(14b)

Even if ASCE10-97 uses radius of gyration around the principal axis


and AISC 2005 as well as S16-09 use radius of gyration around the
geometrical axis, both approach provide values close to each other
as it is seen in Table 2, S16-09 and AISC 2005 providing slightly
larger KL/r than ASCE10-97.

Fy
Fe

E
 .
KL 2
2

Fe =

be predicted assuming simple connections and no eccentricity.


This phenomenon is due to the effect of the connections and is
currently studied at the University of Sherbrooke. Nevertheless,
because the gusset plates in the tests had flexibility very close to
the flexibility of gusset plates or angle leg of cord members used
in practice, it was considered that the codes should take this effect
into account, at least in a limited way. Indeed, ASCE 10-97 allows
modifying KL/r with Eq. (12) for members with normal framing
eccentricities at both ends of the unsupported panel and Eq. (13)
for members partially restrained against rotation at both ends of
the unsupported panel (ASCE 10-97).

rx

where

Fig. 9. In-plane and out-of-plane displacements at centre of angle member 95S25-1.

0
(10)

1371

Calculation of strength is based on the determination of variable


KL/r. The length of the member L is taken in the calculation as
the distance from centroid of connection to centroid of connection.
Experimental evidence has however shown that the effective
length KL may be different from theoretical length and that
buckling may not occur around the minor axis. For example, it
can be seen in Figs. 9 and 10 that at the time initial buckling
occurs, i.e. when the force is a maximum, the displacement in
the in-plane and out-of-plane directions are different, indicating
that the failure did not initiate in the minor axis as it would

4.2. Comparison of experimental and theoretical results


Table 3 shows the predicted capacity. The calculation of the
theoretical strength was based on an average thickness measured
in laboratory. The theoretical calculations therefore assume a
constant corroded thickness. Predictions with ASCE 10-97 are
usually better than S16-09 and AISC 2005. It is seen from Table 3
that if codes S16-09 AISC 2005 and ASCE 10-97 can capture well
the strength loss of some members, capacities are sometimes not
well predicted. This can be attributed to the lack of uniformity
of the corrosion. In fact, complete perforation was found on two
members (see Fig. 5). In particular, the capacity of member 48L40-2 was greatly reduced due to perforation because most of one
leg was lost near the centre. Another example is member 48S25-1 which is particularly weaker than its predicted capacity.

1372

L.-V. Beaulieu et al. / Journal of Constructional Steel Research 66 (2010) 13661373

Table 3
Experimental and theoretical results for the compression capacity.
Specimen

Actual % of corrosion

Exp.
capacity (kN)

Capacity
S16-09 (kN)

Capacity
ASCE (kN)

Capacity
AISC (kN)

Exp./S1609

Exp./ASCE

Exp./AISC

48S-00-1
48S-00-2
48L-00-1
48L-00-2
95S-00-1
95S-00-2
95L-00-1
95L-00-2
48S-25-1
48S-25-2
95S-25-1
95S-25-2
48L-25-1
48L-25-2
48S-40-1
48S-40-2
95S-40-1
95S-40-2
95L-25-1
95L-25-2
95L-40-1
95L-40-2
48L-40-1
48L-40-2

0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
26.4
26.6
26.9
25.3
27.4
27.3
37.2
37.3
37.1
36.1
22.4
21.5
30.7
27.2
40.7
47.1

n/a
114
91
88
323
335
168
167
57
n/a
253
251
80
79
66
63
216
202
151
150
99
99
65
25

77.2
77.2
49.6
49.6
186
186
119
119
63.4
59.4
144
142
32.5
32.2
41.9
38.7
128
125
97.9
99.9
86.4
89.7
23.5
15.2

142
142
88.6
88.6
278
278
168
168
121
114
214
212
70.4
69.9
80.6
73.0
191
186
137
140
121
126
54.3
32.1

108
108
66.1
66.1
209
209
123
123
95.3
91.6
162
160
53.4
53.2
73.1
69.1
144
140
102
104
90.1
93.6
45.2
33.2

n/a
1.48
1.84
1.77
1.74
1.81
1.42
1.41
0.89
n/a
1.77
1.77
2.44
2.46
1.57
1.62
1.69
1.62
1.54
1.50
1.15
1.10
2.76
1.65

n/a
0.80
1.03
0.99
1.16
1.21
1.00
0.99
0.47
n/a
1.18
1.19
1.13
1.13
0.82
0.86
1.13
1.09
1.10
1.07
0.82
0.79
1.20
0.78

n/a
1.06
1.38
1.32
1.55
1.60
1.36
1.35
0.59
n/a
1.57
1.57
1.49
1.49
0.90
0.91
1.50
1.44
1.48
1.44
1.10
1.06
1.44
0.76

Fig. 10. In-plane and out-of-plane displacements at centre of angle member 95S40-2.

Fig. 11. Stress versus width-to-thickness ratio.

The failure mode of this member was very localized on one leg
near the centre. The failure in this case was due to the large
b/t and is not influenced by its KL/r value. The thickness near
its failure point is approximately 3.3 mm, nearly 1 mm smaller
than its average measured thickness. The experimental capacity
of 57 kN would be obtained for a thickness of approximately
3.1 mm with ASCE 10-97. As seen in these two examples, the
compression capacity is largely dependent on the extent and the
location of the corrosion. Concentration of corrosion reduces in
an acute way the capacity of the member, specifically when such
concentration occurs near the middle of the member. Corrosion
is critical at the middle of the member because this is where the
moment due to the eccentricities of the connections is a maximum.
This non-uniformity of the phenomenon explains in large part the
differences between the experimental and theoretical results.
Study of the failure modes and comparison of the experimental
and predicted strength is best observed with the help of Fig. 11.
In this figure, the compressive stress, equals to the compressive
strength divided by the average section area, is plotted against
the width-to-thickness ratio. The 48S and 95S members should be
compared to the S16 and AISC, KL/r = 91 and ASCE, KL/r = 80
curves. On the other hand, 48L and 95L members are compared to

the S16, KL/r = 125 and ASCE, KL/r = 115 curves. The slenderness ratios used in the ASCE 10-97 are larger due to the
modifications of Eqs. (12)(14). Members 95S are found in an
area of the strength curve, which is not affected by the widthto-thickness ratio. Consequently, as observed in the tests, these
members generally display global buckling. Similar observations
can be made for 95L members. Members 48S, however, mostly
displayed local buckling, which is in accordance with their location
in the strength curve: the capacity in this area is influenced by
the width-to-thickness ratio. Except for member 48L-40-2, which
has a very large width-to-thickness ratio, 48L members are not
influenced by b/t. For this reason, the predictions of the ASCE 1097 are much more accurate than those of the S16-09 for this type
of member. AISC 2005 provides better results than S16-09 because
the b/t is better accounted for even if the larger KL/r reduces the
capacity predicted for those members as compared to ASCE10-97.
Table 4 presents various averages and standard deviations
for the ratio of experimental to theoretical capacity. ASCE 10-97
generally provides averages closer to one and smaller standard
deviations than S16-09. AISC is generally between ASCE and S16
in terms of prediction and standard deviation. As expected, the
variability in the prediction of capacity is larger for corroded
members than for uncorroded members.

L.-V. Beaulieu et al. / Journal of Constructional Steel Research 66 (2010) 13661373

1373

Table 4
Average ratio of experimental to theoretical capacity.
Characteristic of
specimens

Average ratio S16-09

Standard deviation
S16-09

Average ratio ASCE


10-97

Standard deviation
ASCE 10-97

Average ratio AISC

Standard
deviation AISC

All specimens
Uncorroded
Corroded
Corroded 4.8 mm
Corroded 9.5 mm
25% nominal
corrosion
40% nominal
corrosion

1.68
1.64
1.70
1.92
1.52
1.77

0.43
0.19
0.51
0.66
0.26
0.55

1.00
1.03
0.98
0.91
1.05
1.04

0.19
0.13
0.21
0.26
0.16
0.25

1.29
1.38
1.25
1.08
1.39
1.38

0.29
0.18
0.33
0.38
0.20
0.35

1.65

0.51

0.94

0.17

1.14

0.29

It is interesting to note in Table 4 that predictions of the capacity


of members with 9.5 mm legs are better than the predictions
on 4.8 mm leg members. The average of the predictions is close
to experimental results for 9.5 mm leg members and variability
is smaller than for 4.8 mm leg members. The scatter in the
ratio experimental/prediction observed on 4.8-mm leg thickness
members might be attributed to the irregularity of the corrosion
process when corrosion level increases.
Surprisingly, members with nominal level of corrosion 40% do
not have larger variability than 25% corroded members. This is
partly due to the large scatter obtained for some 25% corroded
members, in particular member 48S-25-1. More specimens with a
larger range of level of corrosion would be needed to further study
the effect of corrosion on the difference between predicted and
experimental capacity.
5. Conclusions and recommendations
This article presents the compression test of 24 members, 16
of them being corroded at a level corresponding to weight loss
ranging from 21% to 47%. The corroded members lost between 10%
and 70% of their uncorroded capacity.
The reduction in capacity of many members was not predicted
accurately using the S16-09, ASCE 10-97, and AISC codes. The
variation in the results can be explained by the non-uniformity of
corrosion. ASCE 10-97 predicts the strength reduction of members
more accurately because it offers a better way to predict local
buckling due to large width-to-thickness ratio.
This experimental program showed that it may be possible to
estimate the capacity of a corroded angle steel member using a
prediction method based on average residual thickness of member
and ASCE 10-97. For members with initial thickness of 9.5 mm, the
ratio of predicted/experimental capacity has an average of 1.05 and
a standard deviation of 0.16. For members with thickness 4.8 mm,
the ratio of predicted/experimental capacity has an average of 0.91
and a standard deviation of 0.26.
The method needs improvement to be able to predict accurately
residual strength. For example, it is required to take into account
the unevenness of corrosion and concentration of corrosion as well
as location of those concentrations.

To develop empirical charts that can be used in practice to


evaluate directly the residual strength, it is believed that additional
experimental data would be required, in particular for members
with level of corrosion between 0% and 20%. Additional data for
other leg thickness and KL/r would also be interesting. For utilities,
residual strength curves as a function of time are also necessary to
plan asset repairs and replacements. For this purpose, additional
research is required to transform corrosion level into remaining
life time, depending on environment aggressiveness. Finally, it is
also necessary to verify that the actual corrosion pattern found on
steel structures is similar to the corrosion pattern obtained in this
experiment through an accelerated corrosion method.
Acknowledgements
The research was performed within the Industrial Research
Chair NSERC/HQT on Mechanics and Structures of Transmission
Lines. The financial support of NSERC and Hydro-Qubec Transnergie (HQT) is therefore acknowledged. The valuable collaboration
of Maryse Lavoie and Andr Valle from HQT for this project is also
acknowledged. The help of Jonathan Dubuc, graduate student at
the University of Sherbrooke was greatly appreciated.
References
[1] Hathout I. Damage assessment and soft reliability evaluation of existing
transmission lines. In: Proceedings of the international conference on
probabilistic methods applied to power systems. IEEE; 2004. p. 9806.
[2] Kayser JR, Nowak AS. Capacity loss due to corrosion in steel-girder bridges.
J Struct Eng 1989;115:152537.
[3] Sarveswaran V, Smith JW, Blockley DI. Reliability of corrosion-damaged steel
structures using interval probability theory. Struct Safety 1998;20:23755.
[4] Sarveswaran V, Smith JW. Structural assessment of corrosion-damaged steel
beams using minimum capacity curves. Struct Eng 1999;77:1723.
[5] CAN/CSA S16-09. Limit states design of steel structures. 2005.
[6] ANSI/AISC 360-05. Specification for structural steel building. 2005.
[7] Winston Revie R. Uhligs corrosion handbook. 2nd ed. New York: John Wiley
& Sons; 2000.
[8] Morissette E. valuation des normes de calcul et du comportement des
cornires simples en compression utilises comme contreventements dans les
pylnes treillis en acier. M.Sc.A. thesis; 2008.
[9] ASCE 10-97. Design of latticed steel transmission structures. 1997.
[10] ASTM E 8-61T. Tentative methods of tension testing of metallic materials.
1961.

Вам также может понравиться