Professional Documents
Culture Documents
PhD Dissertation
KATALIN OSZVALD
Budapest University of Technology and Economics
Budapest, 2014
TABLE OF CONTENT
Acknowledgement ..................................................................................................................................................................... 4
Abstract ...................................................................................................................................................................................... 5
Introduction ..................................................................................................................................................................... 6
General.................................................................................................................................................................... 6
Research strategy.................................................................................................................................................. 18
Conclusion ............................................................................................................................................................. 66
Buckling resistances of eccentrically connected non-corroded and corroded angle members under compression ..... 67
3.4.1. Effect of the disproportion of the corroded surface on the buckling ........................................................... 87
3.4.2. Effect of the initial geometrical imperfection on the buckling.................................................................... 87
3.4.3. Effect of equal volume loss due to corrosion on the bucking ...................................................................... 88
Prediction of design buckling resistance reduction .............................................................................................. 89
Conclusion ............................................................................................................................................................. 94
• This work is connected to the scientific program of the “Development of quality-oriented and harmonized R+D+I
strategy and functional model at BME” project. This project is supported by the New Hungary Development Plan
(Project ID: TÁMOP-4.2.1/B-09/1/KMR-2010-0002)
• The work reported in the thesis has been partly developed in the framework of the project „Talent care and
cultivation in the scientific workshops of BME" project. This project is supported by the grant TÁMOP-4.2.2.B-10/1-
-2010-0009.
• The work reported in the thesis has been partly developed in the framework of the project „Campus Hungary
program”. This program is executed by the financial support of the European Union in the framework of the Social
Renewal Operational Program (TÁMOP) of Hungary. This project is supported by the grant TÁMOP-4.2.4.B/1-11/1-
2012-0001.
I would like to express my especial gratitude to my supervisor Professor László Dunai who has elevated and supported me by
their advice to improve the presented research.
I would like to express my thanks to Dr. Pál Tomka for his advises and helpful remarks at the final part of the research work.
I would like to express my thanks to Dr. Attila Joó for his helps and advices at the beginning of my research period.
I express my gratitude to thank the staff of the Structural Laboratory; their huge experiences, and committed work have
helped me to execute the experimental tests.
I have a claim to tender my thanks for all kinds of help and support to the members of the Department of Structural
Engineering. I would like to thank to Levente, Ádám, Viktor, Árpi, Balázs, Piroska, Noémi and others for their help which
covered not only the professional help, but personal advices.
Special thanks are due to my family who have emotionally supported me to execute the presented research.
-4-
Abstract
The corrosion is a persistent problem in the world, which causes significant problems and costs in all industries. In spite of
the developed corrosion protection the steel structures are corroded due to weather circumstances and inadequate
maintenance. Therefore the widely used angle members in steel structures are corroded. In my research work I analysed
corroded compressive angle members supposing various levels of corrosion with the aim to determine the remaining capacity
for the refurbishment. The focus of the current research is on the buckling behaviour; the aim is to develop applicable process
to determine the ultimate buckling force of corroded angle members.
In the research program experimental and numerical studies are completed. The experimental and numerical studies are
carried out on members with different end connection, as centric hinge connection, connection with one bolt and connection
with two bolts. In the civil engineering practice these members are chord and bracing members in trussed towers.
Compressive buckling tests are carried out on corroded steel angle members, where the corrosion is modelled by artificial
reduction of the thickness. In parallel with the experimental tests finite element models are developed using solid and shell
finite elements to follow the material reduction in the compression members. By the material and geometrical nonlinear
analyses the test results are simulated and the developed model is verified and proved to be applied for further analysis.
Extended numerical parametric studies using the finite element models are completed to determine the tendencies of the
buckling phenomena of the corroded angle members. In the studies different cross-section reduction, location and extend of
corrosion are investigated. In the analysis both global and local effects of the corrosion on the stability behaviour are studied.
By the evaluation the ratio of the ultimate buckling force of the corroded and non-corroded members is determined regarding
to the connection types and the corrosion parameters. The proposed method to analyse a corroded member is easily
applicable in the practice. The evaluation method propose refurbishment or total replacement of the members taking into
account the result.
-5-
Introduction
General
Generally it can be said, that corrosion is one of the major problems of steel structures. The corrosion can appear on every
structure irrespective where these are found. Soil, water and air also can be a corrosive medium. In case of buildings and
structures the corrosion as a problem does not occur, if the ventilation is acceptable. The different weather circumstances and
the lack of the maintenance are leading to corrosion. Many different types of the corrosion occur on the structures, some of
them cause only surface changes (e.g. paint flaking off) and then the thickness or volume is reduced. Some types of corrosion
change the material properties of the element. At the same time the appearance and the measure of the corrosion can be
various and it is difficult to describe using different parameters such as, thickness reduction, position within the cross-section,
location along the element’s length and the extension. This diversity makes the analyses more difficult.
However the knowledge of the behaviour and the remaining capacity of the corroded elements are important to decide
whether the elements must be replaced or it is enough to strengthen it during the refurbishment. The recent codes of steel
structures (Eurocode [1], CAN/CSA, AISC, ASCE) give only minor recommendation [2], how to consider the effect of corrosion
in the analyses of the elements. Typically the application of average cross-section is proposed by the standards, but in general
this approximation is not accurate enough due to the diversity of the corrosion.
The corrosion appears on many structures, where the conditions are given for the development of corrosion. Each structure
has its typical and possible types of corrosion. It depends on the location and operating conditions of the structures. In the
following different types of corrosion are presented in detail.
Corrosion types
In general terms the corrosion is the following [3]: change of the surface of material due to the environment, which takes
place by chemical or electrochemical processes. Usually the chemical corrosion occurs on structures operating in high
temperatures or gas and steam can cause this type of corrosion. In this process there is no need for electrolyte, but in case of
electrochemical process it is necessary, which can be among others: water, solution or water drop. The corrosion can be
shorted by different ways due to diversity of it. Accordingly in the literature can be found different classification, one of them
is the following [4]:
-6-
The distribution of corrosion on the attacking surface:
• Local distribution
• Average distribution
Other corrosion types which are not involved in the list but appear on steel structure are the following: stress corrosion
cracking, corrosion fatigue, lamellar corrosion, intergranular corrosion, galvanic corrosion and hydrogen embrittlement.
Table 1 also contain a classification [5]. In the table the characteristic of the corrosion type is found together with examples.
Figure 1 Types of corrosion: (a) uniform or general corrosion, (ii) pitting corrosion; (iii) crevice corrosion [4]
-7-
Figure 2 Distribution of the types of corrosion [6]
The two most common corrosion types, uniform and pitting corrosion do not cause changes in material properties [7]. These
cause only surface changes and thickness or volume reduction. Summarized these types of corrosion are known as
atmospheric corrosion, because the corrosive medium is the atmosphere. These corrosion types and the crevice corrosion
appear on steel structures, as bridges, truss towers and on other structures located in the open.
Years ago the Hungarian Electricity Private Limited Company (MVM) developed a new material (KORELL) to avoid corrosion,
but this attempt was not successful. The rate of the corrosion was significant at the connections, therefore the number of
broken bolts increased [8]. Figure 4 presents an example of this problem. Here a plaque-like corroded layer developed
between the connected elements causing relatively high additional tensile stress and some bending in the bolts that led first
of all to the fracture of the bolts with relatively small shank diameter.
Réseau de Transport d'Électricité (RTE in France) also reported that the corrosion on the lattice towers in France is also
significant. Reiner [9] introduced a visual inspection process of lattice towers by a flying machine. In Figure 5a-b the flying
machine called UAV can be seen and in Figure 5c-d the pictures of corroded lattice towers are presented which are taken by
this machine. Generally, the horizontal members are more sensitive to corrosion, as it can be seen in Figure 5.
-8-
Figure 3 Corroded lattice towers - 1
(c) (d)
Figure 5 UAV flying machine and photographs of lattice towers taken by UAV [9]
-9-
Further pictures of corroded truss towers can be found in Appendix A. On the basis of the pictures it can be concluded, that
the variety of the corrosion position and extension is big. Corrosion occurs in generally on both sides of the element legs and
not only on full width of the angle legs.
Literature overview
The corrosion and the effect of corrosion on steel structures are widely studied due to the frequency of the problem and the
economic impact. Several experimental and theoretical researches are completed in the field of corroded elements and
structures. The main goals of these studies were to determine the effect of the corrosion on the structural behaviour, to
predict the ultimate failure and calculate the resistance decrease of the corroded elements. Various corrosion appearances
were analysed in the research studies as localized corrosion, pitting corrosion and uniform corrosion. In the following studies
of various corroded elements are shown, in some cases just the types of the analyses and elements are presented. In some
cases equations are determined to predict the remaining capacity or strength different. These equations regard to not angle
type structural elements; only one research study is completed in the field of corroded angle-section members previously.
A railway bridge with steel I-girder was analysed by Fukuda et al. [12]. Local corrosion damage on the upper flanges under
rail sleepers was studied. An evaluation method was presented in the study to determine the remaining strength of the
corroded plate girder. It was proposed to calculate the buckling stress of the upper flange with the reduced thickness.
Experimental and finite element analyses were carried out to verify the applicability of the proposed method. Furthermore
Fukuda and his co-workers analysed the effect of web corrosion of steel I-girder bridge on the shear buckling behaviours and
strength. In case of flange torsional buckling a model is supposed in the analysis, as it can be seen in Figure 7b. Based on test
and analytical results Eqs. 1a-b are supposed to calculate the ultimate stress (σu). In Figure 7a the test result and in Figure 7c
the results by the proposed equations are plotted comparing to analytical results.
- 10 -
Figure 6 Procedures for predicting the remaining capacity (%RMC) [11]
(a)
(b) (c)
Figure 7 Flange torsional buckling [12]: (a) experiment; (b) analytical model; (c) results of the proposed equations (Eq. 1a-b)
1 0.433
౫
౯
(1a)
0.433
౫ . .
(1b)
౯
Where,
and σy: the yield stress; k=0.43: buckling coefficient.
మ ౯
Kim et al. [13] and Ahn et al. [14] completed experimental analyses and based on the test results numerical analyses on the
locally corroded web of the girder were carried out. As a result an equation was derived to calculate the shear buckling
strength ratio (RSf ) of the corroded and non-corroded members. The proposed equation (Eq. (2)), the test results, and the FE
analysis results are shown in Figure 8, where Cv = Cc/Cs, Cc is the volume of the corroded web and Cs is the volume of non-
corroded web.
- 11 -
= 0.7368 +
.
షఴ.ఴయభఴ (2)
షቀ ౬ ቁ
షయ.రలలల
Figure 8 (a) Remaining shear strength reduction factor of web panel with local corrosion; (b) Test specimen [14]
The same researchers presented in [15] a practically applicable method (CFRP - Carbon-fiber-reinforced polymer) to repair
the locally corroded plate girder around its support, on the bases of experiments and FE analyses. The researchers pointed
out that the effect of web corrosion of the plate girder on the residual bearing stress is less than the effect of the local corrosion
on the stiffener at the support. It means that by repairing plate girder with local corrosion at its support, the corroded stiffener
should be repaired before the corroded web.
ి =
1
(3)
బ 1.25 − 0.0144 − 0.336 − 0.166 − 0.434
ి =
1
(4)
బ 1.43 − 0.0414 − 0.603 − 0.220 − 0.576
- 12 -
In Figures 9a-b the numerical model is shown, the purple colour marks the corroded area. The results using the proposed
equation are shown in Figure 9c.
(a)
(b) (c)
Figure 9 Locally corroded unstiffened plates under uniaxial compression: (a)-(b) model; (c) results by proposed equations Eqs. (3)-(4)
[16]
Later Khedmati et al. [17] and Silva et al. [18] also studied the effect of pitting corrosion on the ultimate strength. Khedmati
supposed random corrosion, but the position of the pitting was regular while Silva generated the corrosion by Monte Carlo
simulation. Khedmati noted that the reduction in the buckling strength of both sides randomly corroded steel plates is
generally higher, than the reduction in their ultimate strength. Teixera et al. [19] took into consideration the initial distortion
as well. The ultimate strength of a corroded steel plate with random initial distortions and random material and geometrical
properties predicted by semi-empirical design equations or by means of non-linear finite element analysis. It was adopted as
a background of case study to demonstrate the accuracy of the proposed approach, executed on corroded marine structural
elements in the last few years. Previously Nakai et al. [20] [21] and [22], carried out a comprehensive study on different
structural elements assuming pitting corrosion. Its effect on tensile strength of members was analysed by tensile tests.
Buckling tests were also carried out to examine the effect on buckling behaviour of members. The authors made the
conclusion that the tensile strength was reduced gradually and the total elongation was extremely reduced due to the
increased thickness loss. On the basis of the results of tensile and buckling tests, they concluded that the tensile strength and
the compressive buckling strength of pitted members was smaller or equal to that members with uniform thickness loss in
terms of average thickness loss. Furthermore 3- and 4-point-bending tests were completed on corroded beam elements. The
conclusion was the following: the effect of the web plate pitting on the lateral-distortional buckling strength was almost
negligible, but the ultimate strength decreases with the increasing corrosion. In the 3-point-bending tests, the local plate
buckling occurred just after reaching the ultimate strength. The authors investigated the effect of pitting corrosion on the
strength of web plates subjected to patch loading. They pointed out that the web crippling behaviour is strongly affected by
the pit distribution on the web plates. In this comprehensive study FE analyses were completed for each experimental test.
In the literature further studies can be found in this research field: Jiang and Soares [23] proposed a closed formula to
determine the ultimate capacity of corroded plate due to pitting corrosion under biaxial loading. The results of FE analyses
- 13 -
showed that volume loss has more influence on the decrease of the compressive capacity of pitted mild steel plates than the
loading ratio. Plate slenderness has significant effect on biaxial interaction curve shape. Both the volume loss and the plate
slenderness are taken into account in the proposed formulae Eq. (5).
= 0.42 + 1 −
౮౫ . ౢ౩౩ ((.
⁄).)
(5)
౯ బ
Where σxu is the remaining strength of plates under uniaxial compression, σy is the material’s yield strength. β is the plate
slenderness (Eq. (6)), V0 and Vloss are the original plate volume and total volume loss due to pitting corrosion, respectively.
= బ
(6)
Where; t is the plate thickness; b is the width of plate, E is the Young’s modulus.
In addition to these studies, many others were executed. Plate elements under tension were investigated by Appuhamy and
Kaita. The first author analysed the effect of corroded surface measurement on the strength reduction [24]. The latter
analysed specimens under tensile loading which were prepared from real corroded structures [25]. The conclusions were the
following: a minimum average thickness has to apply to estimate the remaining yield strength of corroded steel plates with
high accuracy and using the representative effective thickness the remaining tensile strength can be well estimated. Box
girders and other structural models were analysed in [26-29] and the corrosion was taken into consideration.
• The buckling resistance is decreased with an increase of mass loss parameter and decrease of the length parameter.
• The decrease of design buckling resistance due to uniform corrosion is less than 10% for up to 5% of characteristic
value of relative corrosion mass loss.
• If the corrosion is localised down to one tenth of plate length and Gck /G %=5 the design resistance is 53.4% of the
non-corroded plate (Figure 11).
Figure 11 The design values Rd and the mean values of buckling resistance of corroded plate [30]
Finally the authors pointed out that in case of localised corrosion the standard deviation of buckling loads is greater, than in
the cases of overall instability resulting in a significant decrease of the design buckling resistance of the studied plate.
- 15 -
the results of the tests and the analytical calculation the authors proposed a method using an average residual thickness of
corroded member to estimate the buckling capacity according to the ASCE 10-97 code. This method is not able to consider
the localized corrosion. The results of the test and analytical results are presented in Figure 14. In the conclusion of the study
the necessity of further studies is emphasized. Recently Japanese researcher published research study of severely corroded
angle members [32]. In the experimental program 17 corroded angle and 10 corroded channel section specimens were
analysed; the elements were taken from real structures. The support condition was fix on both ends and axial compression
was applied trough end plates. In the evaluation the different cross-sectional specimens were not separated. They presented
the failure modes and applied a prediction method.
- 16 -
(a) (b)
Figure 14 Results of (a) test and (b) the stress and width-to thickness ratio diagrams according to codes and test results [31]
Aims of the research
During reconstructions or reinforcements it is important to determine the resistance of the elements damaged by corrosion.
The standards and design codes do not give any applicable recommendations; just propose to apply reduced cross-sectional
properties. Typically the average cross-section reduction is proposed, but in general this approximation is not accurate
enough due to the diversity of the corrosion. Another - time-consuming - opportunity is to determine the ultimate buckling
force of an artificially corroded member and analyse it as proposed in standards [33, 34].
In the previous sections the existing studies of corroded structures are summarized. Note that only two study of corroded
angle member was found in the literature. The aim of the current research is to significantly extend the previous research on
this field by studying the stability behaviour of corroded compression angle members by comprehensive experimental and
numerical studies. In the research equal-leg angle members are analysed, because this is the most common structural
element type of lattice towers.
In the research it is aimed to consider the diversity of the corrosion applying various corrosion patterns, thickness reductions,
extension, and position of the corrosion. The relevant behaviour modes and the decrease of the ultimate buckling force are
to be determined in the function of the introduced corrosion parameters. The aim is to identify the different local and global
mode and determine the limit values of the corrosion parameters to separate the behaviour modes.
Further goal is to analyse the effect of the different end supports on the behaviour and also on the ultimate buckling force of
corroded angle members.
All of the investigated angle members of the current study represent structural elements of lattice towers, such as chord
member or bracing member (see Figure 15). The analysis of lattice towers are required due to corrosion as it is shown in
Section 1.3.
Finally the fundamental aim of the research is to develop practically applicable design procedure for the buckling resistance
of corroded angle structural members.
- 17 -
Bracing
member
Chord
member
In the parametric study the joint influence of the three main various parameters is analysed, such as the thickness reduction,
the extension, and the position of corrosion. Two different corrosion patterns are applied and the analysis is completed with
three different basic cross-sections and three different slendernesses. Simplifying the diversity of corrosion in case “A”
corrosion pattern just one leg is weakened and in case of “B” pattern, both of the legs are corroded.
Based on the results of the parametric study the design buckling resistance of corroded members is determined (Section 2.5)
and a design method is proposed on the prediction of the reduced design buckling resistance (Section 2.6).
The above mentioned procedure - parallel experimental tests and numerical study - are executed on eccentrically connect
corroded angle members (Chapter 3). The eccentrically connected angle members are bracing members in lattice towers (see
Figure 15). Two different number of bolts are applied in the connection, as one bolt and two bolts. As a first step experiments
- 18 -
are carried out and then a numerical parametric study is completed. The reduced design buckling resistances due to corrosion
are determined in terms of the parameters, what are the same as in axial loading cases.
In the course of the study the applicability of the standards are examined regarding the corroded angle members. Chapter 4
contains the evaluation of the proposed design method and through an example the applicability of method in practice is
illustrated.
The possible buckling modes in case of compressive members are the followings:
• Flexural buckling (FB)
• Torsional buckling (TB)
• Torsional-flexural buckling (TFB),
• General case of torsional-flexural buckling (GTFB)
• Local (plate) buckling (LPB)
In the following sections the buckling modes and the background of them in the different codes are shown.
, ; ̅
∙∙౯ ∙౯
(7)
భ ౙ౨
- 19 -
Where, χ is the reduction factor for the relevant buckling mode and cross-section (see Table 6.2 in EC3-1-1), for angles curve
“b”, with imperfection factor of 0.34), A is the cross-sectional area and Aeff is the effective cross-sectional area. Eq. (7) regards
to Class 1-3 and Eq. (8) to Class 4 cross-sections.
The chord members of lattice towers can be regarded as axially loaded members, in this case Eqs. (7)-(8) are applicable.
Usually the angle bracing members are connected by one leg, and the EC3-1-1 gives order about these elements specific and
separates the angles into two groups on the basis of the connection type. In the first group the elements are connected with
bolted connection using only one bolt. In this case the element has to be calculated as an eccentrically loaded element, by
the interaction of compression and bending. If the connection is welded or bolted with minimum two bolts, the buckling
resistance can be calculated as in centric loading cases, but using an effective relative slenderness ratio. In this case the
combined effect (compression and bending) is traced back to the case of axially loaded member.
The particular codes of lattice towers, EC3-3-1 and EN 50341 give order about the angle members with some differences from
EC3-1-1. The important, fundamental identities and differences among the codes are as follows:
• In all cases an effective relative slenderness is used in the design method. In Table 3 the calculation of the effective
relative slenderness ratio is detailed, relating to the relevant axis and to the design code. The table contains those
cases that relate to the current study.
• EC3-3-1 and EN 50341 distinguish the chord and the primary and secondary bracing members. Effective slenderness
factor is determined to calculate an effective slenderness ratio; it depends on the position of the angle members in
the structures (chord or bracing member).
• EN 50341 divides the bracing members on the basis of the relative slenderness ratio into two groups ( ̅ < √2 and
̅ >√2). If ̅ < √2 the dominant effect is the eccentricity of the end connection and as the slenderness ratio
increases the effect of the eccentricity diminishes. Therefore for higher slenderness ratio the end restraint has more
effect, the effective slenderness ratio is determined in the function of the number of bolts.
• EC3-3-1 recommends a reduction factor η = 0.8 for single angle members connected by one bolt at both ends. The
design buckling resistance defined in Eq. (7) should be reduced by this factor.
• If the cross-section is in Class 4 and the members connected by one leg, reduction applies only to the connected leg
by EN 50341 and EC3-3-1 codes.
• The cross-section classification is different in EC3-1-1 and EC3-3-1 in respect of angle members. The / have to
be calculated according to Table 22 by the different codes. In further recommendation of EC3-3-1 is used.
ഥ=ࢎ
࢈
t EC3-1-1
ഥ/࢚ = ࢎ/࢚
࢈
h
ഥ = (ࢎ − ∙ ࢚)
࢈
EC3-3-1 / EN 50341
b ഥ/࢚ = (ࢎ − ∙ ࢚)/࢚
• ࢈
/ to cross-section classification
Table 2
- 20 -
̅ EC3-1-1 EC3-3-1 EN 50341
̅
Chord v-v* ̅୴ (0.8 + ) ∙ ̅ ̅
10
member
y-y** ̅୷ ̅ ̅
Table 3 Equations to calculate the relative slenderness ratio; * - v-v minor axis; ** - y-y axis parallel to the leg,
1.7.2. Torsional buckling (TB) and Torsional-flexural buckling (TFB)
Torsional buckling (TB) appears in case of double symmetric elements. Because equal-leg angel members - either in non-
corroded or corroded case - have at least only one symmetric axis, this buckling mode is not possible in case of equal-leg
angle members.
Torsional-flexural buckling (TFB) is a possible buckling mode in the case of members of open cross-section with one axis of
symmetry. The displacement is a combination of a torsion about the shear centre and a displacement to symmetric axis. Table
4 contains the in-plane displacement of the cross-section in both cases as TB and TFB.
- 21 -
, = మ 1 + −
1 − +4
ౙ౨,౯ ౙ౨, ౙ౨, ౭ ౙ౨,
(9)
౭ ౙ౨,౯ ౙ౨,౯ ౭ ౙ౨,౯
౭
The slenderness ratio may be approximately calculated using Eq. (11) according to previous version of EN 50341 (2011).
̅# =
%౯ &
(11)
$ &
+ +
' = ∙ (12)
= ̅# > 0.748
)
(౦ .
) మ ≤1 (14)
(౦
̅# =
=
.∙+,-
%౯ *⁄
(15)
ౙ౨ ಚ
Where A is the area of gross cross-section, kσ = 0.43 is the buckling factor corresponding to the stress ratio and boundary
conditions; further details can be found in [37]. Figure 18 presents some possible in-plane deformation of the cross-section
in case of local plate buckling.
- 22 -
Buckling resistance of corroded angle members under axial compression
In this chapter the analyses and results of equal-leg angle chord members are summarized. First the tests are shown than
the numerical and analytical results.
Experimental study
Mechanical process is applied to simulate corroded angle members in the experimental study, detailed in this section. As it
mentioned the atmospheric corrosion does not cause changes in the material properties, therefore the corrosion might be
made by mechanical process instead of artificial corrosion process.
The corrosion is modelled as thickness reduction and milling process is applied to reduce the thickness of members at the
preparation of the specimens. In several previous experiments and also in numerical studies [20, 21] the corrosion was
modelled as thickness reduction. Usually parts of the thickness are eliminated artificially. In the current research, in the
applied milling process the machine has a spinning mill head and it does not change the material properties of the steel
member.
The details of the specimens are summarized in Tables 5-7 according to the above mentioned three groups. The corrosion is
marked with black colour in the tables on the schematic figure of the specimens. In the case of specimen L5 the rate of the
corrosion is just moderate, but this specimen is analysed and evaluated compare to highly deteriorated because the total
volume loss caused by corrosion is the same as in case of L1 - L4 specimens.
Tensile coupon tests are carried out on 9 specimens to determine the material properties of the specimens. The measured
average material properties are as follows: fy = 345 N/mm2, fu = 470 N/mm2. The standard deviation of the measured values
are 9.2 and 9.8 alternatively.
The level of the corrosion is described by the following geometric properties: tcorr, tred, ∆V /V0 and Mcorr. The thickness reduction
of the leg is nominated by tcorr and the remaining/reduced thickness by tred.
The effective specimen dimensions and the thickness of the members are measured by mortise gauge. The measurement
executed on each leg of the specimens at ten locations. Initial geometrical imperfection is not measured before the tests.
- 23 -
The ∆V /V0 characteristic is introduced as the corrosion volume reduction; it is calculated as the ratio of the (∆V ) volume loss
compared to the total original volume (V0) of the specimen. The maximal cross-section reduction (Mcorr) is interpreted as the
ratio of the maximal cross-section reduction along the member compared to the gross area. ∆V /V0 and Mcorr values are
calculated on the basis of the measured dimensions and detailed in Tables 5-7.
The classification of the cross-section is determined based on the width-to-thickness ratio according to EC3-3-1. In the case
of corroded members it is calculated by the width-to-reduced thickness ratio (/). The non-corroded specimens are in Class
1 and all of the corroded members are in Class 3 and 4.
Members from A1 to A11 simulate the uniform corrosion. It means that the level of the corrosion is the same along the length
of the specimen (total or partial length). Because of the diversity of corrosion and corrosion appearance the analysed test
specimens are simplified compare to the real elements. The reason of the simplicity is the execution and the subsequent
numerical modelling of the specimens. The specimens thus are also able to model the real structural corroded elements.
Corrosion on full with of both legs is applied on specimen A1, it is considered as a basic element.
Specimen Location of corrosion tcorr /tred [mm] ∆V /V0 [%] Mcorr [%] ܾത/ݐ
- 24 -
In case of A2 - A6 the position of corrosion is different within the cross-section, but the level of corrosion (∆V /V0) is similar.
These specimens might be termed slightly corroded elements.
At specimen A7 the corrosion is on the ends of the member and in specimen A8 it is in the middle of the member. In the last
two cases (A7, A8) the ∆V /V0 is similar to A2 - A6 specimens but the Mcorr value is different. In specimens A9 - A11 the corroded
surface is the same but their arrangement is various.
In Appendix A pictures about corroded structure are presented, what present partially corroded angle members. For example
A5, A10 and A11 are partially corroded, they model not uniform corrosion along the length or within the cross-section as in
case of specimens A1 or A9. The corrosion might be different location at the same time as the real examples show. Figure 19
presents an example for A5 and A11 specimens.
Specimen Location of corrosion tcorr /tred [mm] ∆V /V0 [%] Mcorr [%]
- 25 -
pit accurately arrangements are not found, but these might be followed and modelled and in other researches [21] [22]
similar arrangements were applied.
The local and high (Mcorr ≥ 50%) deterioration (deep penetration corrosion) is analysed by specimens L1 - L4. The measure of
corrosion (∆V /V0 and Mcorr) is similar in case of L1 - L3, but the position of the corrosion is different. In the case of L1, L4 and
L5 specimens the ∆V /V0 is similar with different Mcorr parameter. The details of the specimens are summarized in Table 7.
Specimen Location of corrosion tcorr /tred [mm] ∆V /V0 [%] Mcorr [%] /
- 26 -
Figure 21 Specimens in the following order: A3, A4, A8, A11 and P2
In the test set-up the loading is applied in the centre of gravity of the non-corroded cross-section, with some unavoidable,
random eccentricities. Additional eccentricity is resulted from the difference between the corroded and original cross-section.
The support is hinge connection and bearing balls are applied at the end plates of the specimens.
A load cell is applied to measure the load. Horizontal and vertical displacements are measured by a pulley systems with
displacement transducers between the head of the loading machine and in the half of the member length, respectively. In
Figure 22 a schematic figure of the test set-up is shown. In Figure 24 the parts of the end set-up are shown.
F
F
25 790/2
Figure 23 Test set up: (a) End plate bottom-view; (b) View from above of end plate; (c) End set-up
- 27 -
2.1.2. Experimental behaviour
In the tests the ultimate behaviour of the specimens is obtained. The results are compared to the original, non-corroded
members and to the results of other specimens. The results are evaluated also in terms of corrosion.
In Section 1.7 the possible buckling modes are detailed, these nominations are used in the evaluation. In the test just the
absolute displacement of the corner of the angle members is measured, therefore the flexural buckling (FB) and the general
case of torsional-flexural buckling (GTFB) mode cannot clearly distinguish in these cases. In the further evaluation of the test
results the “flexural buckling” nomination is used.
In the case of non-corroded (O1, O2) members the observed failure mode is flexural buckling (FB) about the weak axis.
According to the expectation, a yield mechanism is developed in the middle of the specimens. In Figure 24a the test-set up
with the deformed shape of O1 specimen can be seen. The typical load-displacement curves of the non-corroded members
are also presented in Figure 24b.
(a) (b)
60
50
40 O2
Load [kN]
30
20 O1
10
0
0 1 2 3 4 5
Vertical discplacement [mm]
Figure 24 (a) Flexural buckling mode; (b) Load-displacement curves of non-corroded elements
(a) (b)
60
50
30
40
Load [kN]
Load [mm]
20
30 L1
O2
L4
20
A10 10
L5
10 P1
0
0
-5 0 5 10
0 2 4 6 8 10
Figure 25 Typical load-displacement curves: (a) O2, A10 and P1 specimens; (b) L1, L4, and L5 specimens
- 28 -
Flexural buckling is observed in all cases of uniform corroded members (A1 - A11), irrespective of the corrosion position. In
these cases the maximal cross-section reduction (Mcorr) is less than 50%. The pitting corrosion members had also flexural
buckling mode irrespective of the corrosion position and pit diameter. The load - horizontal displacement curves of uniform
and pitting corrosion specimens with global buckling mode are shown in Figure 25a. The shape of the curves is similar to the
original non-corroded members having smaller ultimate buckling force in the case of the corroded specimen.
The effect of same ∆V /V0 on the behaviour is studied by L1, L4 and L5 specimens. The observed failure modes are flexural
buckling in L4 and L5 specimens and local plate buckling (LPB) in the case of L1 specimen. The initial behaviour is the same
in all cases what can be seen on the load - horizontal displacement curves in Figure 25b.
Local failure mode is observed in case of L1 - L3 specimens. In these cases the maximal cross-section reduction is more than
60%. The failure mode is local plate buckling (LPB) in the corroded part with initial global behaviour in L1 - L3 specimens
irrespective of the corrosion position. The character of the load - horizontal displacement curves is the same, as it is illustrated
in Figure 26; in the cases of L1 and L2 some hardening is observed. In Figure 27 the shape of the failure modes of three
specimens P4, L2 and L4 are presented.
16
L1
L2 12
Load [kN]
L3
8
0
-8 -6 -4 -2 0 2
Horizontal displacement [mm]
P4 L2 L4
- 29 -
2.1.3. Measured ultimate buckling forces
The experimental ultimate buckling forces of the members are interpreted as the plateau of the load - displacement curve.
The ultimate buckling forces are summarized in Tables 8-11; these contain the measured ultimate force (Nb,m) and the ratio
of the Nb,m and Nb0,m where Nb0,m denotes the average of the measured ultimate buckling force of the non-corroded members.
In tables FB notes flexural buckling and LPB notes local (plate) buckling. The ultimate buckling force of the non-corroded
members are presented in Table 8. Note that the Euler critical force is 54.63 kN (calculated by L = 840 mm, the distance
between the bearing balls).
O2 FB 50.33 1.00
A9 FB 43.68 0.87 3 13 13
On the basis of the measured ultimate buckling forces the following observations are made:
• Corrosion on the inner side of the leg (A3) causes bigger ultimate buckling force reduction than corrosion on the
outer side (A2).
• Specimens A4 - A6 have same corroded surfaces on different parts of the legs. Corrosion position on the inner part
of the legs with symmetrical arrangement (A5) causes approximately 30% ultimate buckling force decreasing
compared to the non-corroded member. This is significantly different from A4 specimen where the corrosion is on
the outer part of the legs and the reduction of the ultimate buckling force is almost negligible. Unsymmetrical
corrosion pattern (A6) causes even more reduction in ultimate buckling force (~ 33%).
• Corrosion in the middle of the specimen (A8) causes bigger decrease in ultimate buckling force compared to
corrosion close to the supports (A7). The difference is about 12% compared to the non-corroded members.
• A9 - A11 have the same corroded surface with different positions in the specimens. When corrosion is only on one
of the legs (A9), the ultimate buckling force value is bigger, compared to the corrosion on both of the legs. When
corrosion less continuous the ultimate buckling force decrease is bigger, as it is observed in A10 and A11 specimens.
Generally the reasons of the various rate of ultimate buckling force reduction are the followings:
• Slightly different thickness reduction (tcorr).
• Different moment of inertia, if the corrosion is different within the cross-section.
• The moment of inertia of the corroded sections different from each other regards to the axes of the original gross
cross-section.
• Centre of gravity of the cross-section causes different eccentricity direction and additional moment.
• In each cases the corrosion position has unfavourable influence to the buckling mode.
Table 10 summarizes the results of pitting corroded specimens. The following observation can be done on the basis of the
results:
• If smaller pit diameter is applied (P1, P3) the ultimate buckling force value is lower, compared to specimens having
bigger diameters (P2, P4).
• Uniform pit arrangement (P1) cause bigger reduction in ultimate buckling force, comparing to non-uniform
arrangement (P3).
• The decrease in ultimate buckling force by pitting corrosion is smaller comparing to the uniform specimens.
- 31 -
tred ∆V /V0
ID Location of corrosion Behaviour Nb,m [kN] Nb,m /Nb0,m Mcorr [%]
mode [mm] [%]
• Corrosion position in the middle of the specimen (L1) causes smaller decreasing in ultimate buckling force, than
corrosion location near to the support (L3). The maximal cross-section reduction in these cases appr. 70%, and the
governing behaviour is local buckling (LPB). It should be noted, that in case of A7 - A8 and L1 - L3 specimens the
object of the analysis is the same, the effect of the location of the corrosion on the buckling behaviour. The influence
of the location is different in these cases, because the rate of the corrosion is significantly different and the ultimate
behaviour mode is also different.
• The tendency of the reduction of ultimate buckling force is approximately linear as the position of the corrosion is
closer to the support.
Widely spread corroded surface with small cross-section reduction (L5) causes smaller reduction in the ultimate buckling
force compared to localized corrosion with bigger cross-section value (L4 and L1). In the cases the ∆V /V0 is same. The
tendency of the decrease is not linear; when the specimen has local buckling behaviour, the value of the decrease is bigger.
- 32 -
2.1.4. Evaluation of the results
The experimental results of the specimens in three groups are evaluated in this section jointly. The evaluation is carried out
on the basis of two parameters. One of them is the level of corrosion, the volume loss (∆V /V0) and the other is the maximal
cross-section reduction (Mcorr). Beaulieu et al. [31] also used ∆V /V0 parameter for the comparison of their results. They
concluded that their method cannot take into consideration the local corrosion.
Comparing the ultimate forces from the tests, in the case of members with 9% ∆V /V0 reduction the mean value of the
ultimate buckling force is 28.5 kN, but in some cases significant differences can be observed. In the case of specimens with
~13% ∆V /V0 the mean value of the ultimate buckling force is 40.1 kN. The mentioned mean values are different by more
than 25%, even though the associated ∆V /V0 different from each other by only 5%.
Figure 28 presents the ratio of the ultimate buckling force of corroded (Nb,m) members to those of non-corroded (Nb0,m)
members plotted in the function of volume reduction (∆V /V0). The equation of the solid line is Nb,m /Nb0,m = 1 - ∆V /V0. This
line supposes same percent reduction of ultimate buckling force of the corroded members as the ∆V /V0. This is a reference
line. The locally corroded (deep penetration corrosion) members (L1 - L3; the lowest-lying three points) show a high erosion
of ultimate buckling force, whereas the maximal cross-section decrease is high but the volume (∆V /V0) decrease is much
smaller compared to the uniform specimens. On the other hand the difference in ultimate buckling force of A specimens is
much smaller, but even the maximal decrease is 27% (A11).
This observation shows that the estimation of the ultimate buckling force by only volume loss (solid line) might be uncertain
and the ∆V /V0 parameter alone is not reliable to determine the resistance reduction. This conclusion is the same that
Beaulieu et al. [31] made.
1,2
0,8
Nb,m/Nb0,m
0,6
0,4
0,2
0
0 5 10 15 20 25 30 35
∆V/V0 [%]
- 33 -
For better approximation the ultimate buckling force is compared to the maximal cross-section reduction (Mcorr) in Figure 29.
This line supposes same percent ultimate buckling force of the corroded members as the maximal cross-section reduction of
the members. The equation of the solid line is Nb,m /Nb0,m = 1 – 0.01Mcorr. The deviation of the observed ultimate bucking
forces from the solid line here again is significant, thus even this assumed approximation is not acceptable. The determined
equivalent ultimate buckling force decrease is not acceptable, particularly in case of the localized corroded members with
large discrete cross-section reduction (L1 - L3).
Application of a maximal cross-sectional reduction for the approximation of the decrease of the ultimate buckling force is
better than using volume loss (∆V /V0). The members with deep penetration corrosion cannot be at all evaluated by volume
loss. However taking into consideration Figure 29. It is evident that even the Mcorr parameter cannot be used as a governing
parameter of the resistance reduction. The final conclusion is that only one parameter is not enough to predict the ultimate
buckling force of corroded members.
1,2
0,8
Nb,m/Nb0,m
0,6
0,4
0,2
0
0 20 40 60 80 100
Mcorr [%]
- 34 -
ID Ncr [kN] Ncorr /N0 Nb,m [kN] Ncorr /N0
O 54.63 - 50.5* -
A1 36.95 0.68 33.2 0.65
A2 49.82 0.91 51.02 1.00
A3 48.76 0.89 41 0.81
A4 48.54 0.89 50.95 1.00
A5 44.06 0.81 35.5 0.70
A6 41.99 0.77 33.6 0.66
Table 12 Critical force of A1 - A6 specimens (* - average value)
In the other cases, if the corrosion is extended to one certain part of the specimen the previous method does not give
acceptable results, because the cross-section properties are not constant along the whole length. Therefore in these cases
the following two methods might be used: Timoshenko and Lindner.
2.2.2. Columns with varying cross-section - Method by Timoshenko for critical force
If the corrosion is not on the whole length, but it is symmetrical to the half of the member length, the Eq. (16) by Timoshenko
and Gere [38] can be applied to determine the critical force. Figure 30 presents the basis of the equation. In the current
calculation the length of the element (l ) on the figure is the half of the analysed corroded member’s length.
ଶ
ଶ 1
ୡ୰ =
4 ଶ + ଵ ଶ − 1
ଶ − 1 ଶ
ଶ (16)
ଵ ଵ
2.2.3. Columns with varying cross-section - Method by Lindner for critical force
Lindner [39] developed an approximate method for determining the critical force of tapered column members. The critical
force is calculated according to Euler formulae, but using a µ modification factor, which takes into account the change of the
cross-section. The effect of extension of corrosion on the critical force is not included in this method, it follows from the Eq.
(17), where Imax and Imin is the moment of inertia in order of non-corroded cross-section and corroded cross-section, L is the
length of the element.
- 35 -
Table 13 contains the critical force results of O1, A7, A8, L1, L4, and L5 by the mentioned two methods.
Ncr [kN]
ID Timoshenko Lindner
O1 55.21 55.21
A7 51.66 46.28
A8 43.76 46.28
L1 33.01 29.11
L4 38.78 37.17
L5 42.37 45.29
Table 13 Critical force of O1, A7, A8, L1, L4, L5 specimens
In the case of specimens with pitting corrosion similar calculation might not be executed. In the cases of L2 and L3 specimens
just a modified and complicated Timoshenko model can be used, therefore it is not practical for engineers.
In the calculations the yield strength of test specimens is applied, fy = 345 N/mm2.
In the case of Lindner method (varying cross-section) an Amod modified area is applied, it takes into account the effect of the
extension of corrosion on the buckling force with the following Eq. (18):
Where Acorr: the area of the corroded cross-section; lcorr: length of the corroded zone; l: total length of the specimen. The Amod
is to apply in the calculation of the relative slenderness ratio and the ultimate buckling force also in case if the critical force is
obtained by the Timoshenko method.
In Tables 14-15 and Figures 31-32 the ultimate force results, the ratios of the corroded (Ncorr), non-corroded elements (N0)
and the test results are presented.
- 36 -
almost every case lower based on the test, than the calculated values. In case of partial corrosion along the length the
Timoshenko method gives better results than the Lindner method. The cause is obvious: in the reality the specimens are not
tapered as it is assumed in the Lindner method.
Euler Test
ID Ncorr Ncorr /N0 Nb,m Nb,m /Nb0,m
O1 41.14 50.5*
A1 27.72 0.67 33.2 0.66
A2 36.70 0.89 51.02 1.01
A3 36.23 0.88 41 0.81
A4 35.79 0.87 50.95 1.01
A5 33.42 0.81 35.5 0.70
A6 31.65 0.77 33.6 0.66
Table 15 Ultimate buckling forces using critical force according to Euler and EC3-1-1 (* - average value)
0,50
0,28
0,00
A7 A8 L1 L4 L5
Figure 31 Ratio of ultimate buckling force of corroded elements comparing to non-corroded members I
1,50
Euler Test
1,01 1,01
0,89 0,88 0,81 0,87 0,81
Ncorr/N0
0,00
A1 A2 A3 A4 A5 A6
Figure 32 Ratio of ultimate buckling force of corroded elements comparing to non-corroded members II
2.2.5. Design on the basis of test
The Eurocode enable for the engineer to determine the design resistance (Rd) assisting by test. In the following the steps of
the calculation (Eqs. (19)-(22)) is described according to EN 1993-1-3 [40].
The results of the calculations are summarized in Table 16. The values show that the design buckling resistances assisting by
test appr. 50% less than the measured ultimate buckling force in test. The reason of the low values is that from each specimen
is just one sample, therefore the code specifies significant decrease of the measured value in favour of the safety. Therefore
the results and the resistance by different method is not comparable.
- 37 -
678 = 9:8⁄; (19)
; = ∙
=
%౯ౘ,ౘ౩ ౘ౩
(21)
%౯ౘ ౣ
7 = >!> @ ౡ
?
(22)
In the calculation tobs /tnom equal to 1, because the real properties of the corroded cross-section are used and β = 1.
Robs µR Radj Rk Rd
ID
[kN] [kN] [kN] [kN]
O1 50,65 1,25 40,37 25,44 25,44
A1 33,2 1,25 26,46 16,67 16,67
A2 51,02 1,25 40,67 25,62 25,62
A3 41 1,25 32,68 20,59 20,59
A4 50,95 1,25 40,61 25,59 25,59
A5 35,5 1,25 28,30 17,83 17,83
A6 33,6 1,25 26,78 16,87 16,87
A7 43,8 1,25 34,91 22,00 22,00
A8 37,7 1,25 30,05 18,93 18,93
A9 43,68 1,25 34,82 21,93 21,93
A10 38,5 1,25 30,69 19,33 19,33
A11 36,93 1,25 29,44 18,55 18,55
L1 14 1,25 11,16 8,03 8,03
L2 9,6 1,25 7,65 5,51 5,51
L3 4,9 1,25 3,91 2,81 2,81
L4 28,3 1,25 22,56 14,21 14,21
L5 34 1,25 27,10 17,07 17,07
Table 16 Design buckling resistance calculated from test result according to EC3-1-3
- 38 -
2.2.6. Predicting the ultimate buckling forces according to the Eurocode specifications (O, A and L specimens)
If there is no other tool in the hand of the engineer, the standard can be easily applied, but this may necessarily contain many
approximations. In Table 3 in Section 1.7 the effective slenderness ratios can be found. In the next calculation the previously
mentioned specifications ([1], [35] and [36]) for angle members are used with the following assumptions:
• The thickness reduction is supposed to be on both of the member legs along the whole length.
• Effective cross-section is applied to determine the buckling resistance according to the codes (Eqs. (12)-(15)). In
almost every case, the corrosion causes change in the classification of the cross-section. Therefore, in the ultimate
buckling force calculation the effective cross-section is used, along the whole length.
• The measured yield stress fy = 345 N/mm2.
In Figure 33 the results by three codes ([1], [35] and [36]) are plotted in the function of / ratio. The solid lines present the
codes and the dots present the test values; with marking the global behaviour mode with blue colour and local with yellow
colour.
The results according to the codes are different from each other by a maximum of 10%. Comparing the test results to the
results according the codes the latters are on the safe side in all cases. In some cases the test values are significantly higher
than the results according to the codes, but in some cases there is no relevant difference between them.
60
50
Ultimate buckling force [kN]
40 (h-2t)/t
h/t
30 EC 1993-1-1
EC 1993-3-1
EN 50341
20
10
0
0 10 20 30 40 50
ܾത/ݐ
Figure 33 Measured and EN-based calculated ultimate buckling forces in the function of ܾത/ݐ
In Figure 33 the dash dot lines mark the bound of the Class 3 to Class 4 cross-section according to EC3-1-1 and EC3-3-1. As it
can be seen in the figure (by the blue and yellow dots) the lines cannot separate the specimens into two groups according to
the ultimate behaviour. It means using the cross-section classification according to the codes the prospective ultimate
behaviour might not be determined.
It should be noted, that these values are not exactly design buckling resistances, because the test results are not reduced and
in the calculation according to codes the real yield strength is applied contrary to the nominal yield strength.
- 39 -
From the above evaluation of the results it can be concluded that the complex buckling phenomena of the corroded angle
members can be described by many parameters, as cross-section reduction, volume decrease, position of corrosion and
extension. The completed tests show the tendencies of the behaviour, but these are not enough to derive practically
applicable design recommendations. The standards can predict the ultimate buckling force on the safe side, but the diversity
of corrosion cannot be taken into consideration.
Numerical study
The executed tests are extended by a numerical study. Applying a numerical finite element model a comprehensive study is
carried out. In this study the corrosion parameters are changed and therefore the joint influence of the three main corrosion
parameters on the buckling could be analysed.
To facilitate the numerical analyses, a parametric code is programmed in ANSYS. In the algorithm the location, the extension
and the thickness reduction as the parameters of the corrosion are considered. In the algorithm of the pitting corrosion the
distribution of the pitting can be completed randomly or uniformly, using solid elements. A similar algorithm is developed
and programmed for uniform corrosion using shell elements. Figure 34 illustrates the different corrosion models, (a) solid
model with rough corrosion, (b) pitting corrosion, and (c) shell model with uniform corrosion.
- 40 -
In case of solid model, if the corrosion occurs on part of the element along the length or within the cross-section, there is no
transition zone between the corroded and non-corroded parts, the boundary is sharp. This is the same shaping as in case of
test specimens.
The steel is modelled by bilinear material model (Figure 35), as linear elastic - perfectly plastic material; fy = 235 N/mm2; the
Young’s modulus 210000 N/mm2 and the Poisson’s ratio 0.3. At the ends of the element rigid region (CERIG - Ansys command)
is defined and in the centre of the gravity of non-corroded cross-section hinge connection is applied as support.
In case of shell model in the plane of cross-section the legs are divided to 5 and 10 parts. The width-to-high ratio is from 1 to
10. On the basis of results (Figure 37) the applied division is 10 and the width-to-high ratio is 2.5. The difference from the
analytical Euler critical force is 2.2%.
- 41 -
Element type Element size Ncr [kN] ∆ [%]
4.0 56.93 4.26
3.5 56.60 3.70
Solid45 3.0 56.43 3.42
2.5 54.84 0.61
2.0 54.40 -0.19
Solid186 4.0 56.82 4.01
Table 17 Convergence analysis - Solid finite element model
57,50 8:24:00
57,00 7:12:00
Critical force [kN]
56,50 6:00:00
Running time[s]
56,00 4:48:00
55,50 Critical load 3:36:00
Running time
55,00 2:24:00
54,50 1:12:00
54,00 0:00:00
1,5 2 2,5 3 3,5 4
Finite element size [mm]
55
54,5 10
54
Critical force [kN]
53,5
53
52,5
52
51,5
5
51
50,5
50
0 10 20 30 40
Finite element size [mm]
- 42 -
Mcorr Nb,m Nb,SHELL Nb,SHELL Nb,SOLID Nb,SOLID
ID Location of corrosion
[%] [kN] [kN] /Nb,m [kN] /Nb,m
O2 ― 0.0 50.33 48.73 0.97 50.75 1.01
A11 195 200 200 195 13 36.93 37.04 1.02 36.88 1.00
- 43 -
measured ultimate buckling force of the test and the ultimate force of the numerical models: Nb,SHELL and Nb,SOLID. In the
notation the subscript marks the finite element types applied in the numerical analyses. The ratios of the numerical analysis
to the test results are also presented in Table 18. In the case of the shell model the applied magnitude of geometrical
imperfection is between L/400 - L/2000 (2.1 - 0.42 mm) and for solid model it is between L/100 - L/1500 (8.4 - 0.56 mm),
where L is the length of the compression angle member.
(a) 35 (b)
30
25
20 Test
Load [kN]
15 SHELL
10 SOLID
5
0
0 5 10
Horizontal displacement [mm]
Figure 38 Results of preliminary study: (a) load-displacement curves of specimen L4; (b) local buckling failure mode of specimen L3
The results of the preliminary study are illustrated by the load-displacement curves of specimen L4 in Figure 38a and Figure
38b shows the local buckling failure mode of specimen L3. In Figure 38a the results show that both shell and solid models
are in good agreement with the test.
- 44 -
In the table the measured ultimate buckling force of the tests RM and the results of the numerical analyses, applying the
introduced numerical model in Section 2.2.1, are detailed. The ratio of the numerical and the test results and the mean value
and the coefficient of variation of the comparative ratios are also presented. The ultimate behaviour modes obtained by the
FEM analyses correspond to the test results, the observed behaviours are flexural buckling and torsional-flexural buckling.
The executed analyses and the results confirm the applicability of the developed models.
No. Specimen ID ex [mm] ey [mm] RM [kN] RSOLID [kN] RSOLID/RM RSHELL [kN] RSHELL/RM
1 E900-1 0 0 117 124.35 1.06 105.34 0.90
2 E900-2 4.2 0 110.9 118.68 1.07 107.29 0.97
3 E900-3 10 0 105.3 106.06 1.01 102.56 0.97
4 E900-4 16.8 0 84.5 80.71 0.96 99.48 1.18
5 E900-5 29.4 0 59 58.86 1.00 60.32 1.02
6 E900-6 50.4 0 46.1 41.55 0.90 40.42 0.88
7 E900-7 0 8.4 65.6 66.19 1.01 61.46 0.94
8 E900-8 0 16.8 46.4 49.65 1.07 46.09 0.99
9 E900-9 0 29.4 34.4 36.45 1.06 33.92 0.99
10 E900-10 0 50.4 23.3 25.6 1.10 23.65 1.02
11 E1200-1a 4.2 0 68.1 73.65 1.08 67.84 1.00
12 E1200-1b 4.2 0 69.7 73.65 1.08 67.84 0.97
13 E1200-2 10 0 66.7 73.56 1.10 66.27 0.99
14 E1200-3a 16.8 0 66.4 73.29 1.10 66.27 1.00
15 E1200-3b 16.8 0 71.3 72.36 1.01 65.03 0.91
16 E1200-4 29.4 0 57 53.36 0.94 57.17 1.00
17 E1200-5 50.4 0 42.5 39.47 0.93 40.73 0.96
18 E1500-1a 0 0 45.3 50.98 1.13 46.61 1.03
19 E1500-1b 0 0 41.1 50.98 1.13 46.61 1.13
20 E1500-1c 0 0 42.8 50.98 1.13 46.61 1.09
21 E1500-2 8.4 0 42.9 50.99 1.19 46.36 1.08
22 E1500-3 16.8 0 41.3 50.81 1.23 45.55 1.10
23 E1500-4 29.4 0 40.1 48.89 1.22 42.59 1.06
24 E1500-5 50.4 0 37 36.66 0.99 35.72 0.97
25 E1500-6 0 8.4 35.2 36.97 1.05 34.78 0.99
26 E1500-7 0 16.8 27.8 30.82 1.11 28.54 1.03
27 E1500-7 0 29.4 23.4 24.82 1.06 22.93 0.98
28 E1500-7 0 50.4 18 18.72 1.04 17.44 0.97
Mean value 1.06 1.00
Average deviation [%] 7.6 6.8
- 45 -
2.3.4. Evaluation of the model validation results
The results show that both shell and solid models can be used to predict the behaviour of the centric compressive corroded
chord members. The disadvantage of the solid model is the small finite element size in (i) large number of model degrees of
freedom’s in the model and (ii) time consuming analysis. The results of the test can be more accurately estimated by solid
finite element model, but the shell model also gives acceptable results.
In the eccentric tests it is observed that the shell model approximates the test results by higher accuracy: the mean value of
the comparative ratio is 1.0, while this value is 1.06 in the case of solid model. The average deviation is also less in the case
of the shell model. It can be concluded that the shell model gives an acceptable approach to estimate the ultimate buckling
force of the eccentrically compressed angle specimen and it follows the observed behaviour. Having these findings in the
further analysis the shell model is applied.
- 46 -
The applied yield strength is fy = 235 N/mm2. This steel grade is chosen because this is the most often applied steel grade of
structures. In the GMNI analyses the ultimate behaviour and load are determined. The determined maximal load is a design
buckling resistance according to EC3 design method.
Various size and patterns of corrosion are assumed, as detailed in the followings. Eleven basic elements are investigated with
relative column slenderness between 0.7 - 1.5, as shown in Table 21. The initial / (width-to-thickness) ratio is given, as
the basis of the cross-section classification of angle members by EC3-3-1. In the further evaluation of the corroded elements
the / is the ratio of the reduced thickness and the width of the legs according to EC3-3-1, as a corrosion parameter.
Two typical corrosion patterns are applied in the numerical study: pattern “A”: the corrosion occurs on one leg, and pattern
“B”: both of the legs are corroded. The corrosion is marked by black on the simplified figure of the patterns in Table 22. The
parameters which describe the corrosion in the study are as follows: the thickness reduction (Tred), the extension of corrosion
(Ext) and the position of the corrosion (pc).
The Ext parameter is equal to the ratio of the length of the corroded area to the total length. The pc parameter is defined as
the ratio of the distance between the centroid of the corroded area and the near end of the member to the half of the length.
(Note that previously tred used as the reduced thickness in the test, but with small letter, in this numerical study Tred with
capital letter is value of the thickness reduction in percentage.)
In the further evaluation the / ratio is an important parameter of the comparison. The same Tred parameter does not result
the same / on elements RF-1, RF-6, RF-9 because of the initial / ratio of the non-corroded elements is different.
Therefore, consecutive GMNI analyses are carried out on further corroded elements, where the Tred is modified to have equal
/ ratio. In the further the / ratio always regards to the corroded zone, it means t is tred.
- 47 -
Pattern Tred [%] Ext [%] Corrosion position (pc)
80
“B” pattern 100 1.00
In order to identify clearly the buckling modes GBTUL program (Generalised Beam Theory at the University of Lisabon) [44]
program is used. This program performs elastic buckling (bifurcation) analysis of prismatic thin-walled members, by a thin-
walled bar theory that (i) accounts for local deformation and (ii) provides an advantageous representation of the deformation
field, as a combination of structurally meaningful cross-section deformation modes [44]. Note that as an alternative solution
the constrained CUFSM program can be also used for the same purpose [45]. Camotim et al. made research studies in the
field of stability analysis of thin-walled angle section members using GBTUL and Ansys programs. The results are detailed in
[46-48].
In the current study the analyses are performed with the following parameters and conditions:
• Cross-section: L40×40×4 mm, length 100 mm - 10000 mm. Mesh options: 10 intermediate nodes per leg, and 20
finite elements along the length.
• Simply supported ends.
• Loading - axial, in the centre of the gravity. Eccentricity due to corrosion is taken into consideration with additional
end moments.
• The first 10 deformation modes are taken into consideration in the modal participation.
• Corrosion: pattern “A” - Tred: 40, 60 80%; pattern “B” - Tred: 50, 70%. Corrosion is along the whole length.
In the verification the non-corroded member is analysed first. The Euler critical force is 54.26 kN from GBTUL and by analytical
method is 54.65 kN, therefore it is applicable and accurate enough.
- 48 -
In the first row of Table 23 the four dominant deformation modes can be seen, the ID of them is used in the following. The
buckling modes are combined from these clear cases. The ID number of the deformation modes is used in Tables 24-25, too,
the modes are marked by colour.
In the second row of Table 23 the result of non-corroded element is presented. In the first curve the critical forces are plotted
in the function of the length in logarithmic scale. In the second curve of the tables, the horizontal axis is also the length in
logarithmic scale and the vertical is the deformation mode participation of the buckling modes. In case of non-corroded
members and less than 54.50 column slenderness, the dominant mode of buckling is the torsion (4) - secondary warping and
in small percentage major axis bending (2); this is the torsional-flexural buckling (TFB). If the column slenderness is higher
the buckling mode is clearly minor axis bending (3) - flexural buckling (FB).
In Tables 24-25 the first column presents the applied pattern and the Tred thickness reduction. In the second and third columns
the axes of the curves are the same as in Table 23.
In case of “A” pattern corrosion, two typical modes can be identified. Dominant local plate buckling (LPB) is observed with
small percentage torsion (4). As the level of corrosion increase the percentage of the torsion is decreased. The other type of
buckling is the general case of buckling, dominant minor bending, and constant percentage of major bending by increasing
corrosion level. In Table 24 the results of elements with pattern “A” corrosion are summarized.
In case of “B” pattern corrosion the buckling modes are the same as in case of non-corroded members as it can be seen in
Table 25. As the corrosion increases the change in dominant deformation mode is occurred by higher column slenderness, as
in case of “A” pattern. The results of further analyses also lead to this conclusion.
It can be concluded, that in both cases of corrosion pattern two buckling modes can be identified. The buckling mode depends
on corrosion, because the column or plate slenderness of the element is modified due to corrosion.
non-
corroded 4 3
Tred = 0 %
2
- 49 -
“A”
pattern 5 3
Tred =
40% 4 2
“A”
pattern 5 3
Tred =
60% 2
“A”
pattern
5
Tred =
3
80%
2
“B”
pattern
4 3
Tred =
50% 2
“B”
pattern
4 3
Tred =
70% 2
- 50 -
wave is approximately two times the free width of the outstanding plates if the corrosion is on one leg (pattern “A”). In the
case of pattern “B” the length of the buckling wave is usually the length of the corroded zone if the buckling mode is torsional-
flexural buckling. There is some exception by elements with 0.7 relative column slenderness ratio. In these cases the buckling
of the legs is along to the whole length (Table 26 second row). Table 26 contains the different typical buckling modes,
thickness reduction and the extension of corrosion are detailed in the first column. In the second column the corrosion
position is presented marking red colour. In the third and fourth columns the overall displacement and in-plane displacement
can be seen.
L = 840 mm
Tred = 70%
Ext = 20 %
L = 510 mm
Tred = 30%
Ext = 20 %
L = 840 mm
Tred = 60%
Ext = 30 %
- 51 -
L = 840 mm
Tred = 50%
Ext = 50 %
L = 840 mm
Tred = 80%
Ext = 100 %
L = 840 mm
Tred = 20%
Ext = 100 %
- 52 -
corrosion. Figure 40 presents a schematic figure of the zones depending on the relative column slenderness and corrosion
pattern, considering the relative column slenderness. The typical buckling modes are marked in the figure. Table 27
summarizes the / limit values for the zones.
“A” pattern
̅ = 1.5
̅ = 1.15
̅ = 0.7
“B” pattern
̅ = 1,5
̅ = 1.15
̅ = 0.7
/
Figure 40 Schematic figure of I. - III. Zones and the relevant buckling modes
Applying / as the basis of the classification is a practical choice, because calculation of the relative column slenderness of
corroded elements is difficult by analytical method. The Tred parameter from describing parameter of corrosion is not enough,
because the same Tred is not equal with the same / by different non-corroded sections. Therefore already in the
classification of the cross-section may have been different result.
In Figure 41 the change of the relative column slenderness regards to flexural buckling (FB) due to corrosion can be seen in
terms of the / ratio. It is marked by blue line, furthermore the change of the relative slenderness of torsional-flexural
buckling using Eq. (9) is presented also, by green line. In Section 1.7.2 approximation equation (Eq. (11)) of relative
slenderness regards to torsional-flexural buckling is mentioned and the results by it can be seen in the figure; by red line. In
the calculation the following assumptions are used:
- 53 -
The classification limit / regards to Class 4 cross-section according to EC3-3-1 is marked; the determined bounds of I. - II.
and II. - III. zones also marked by dash dot lines. The results by numerical model, it means the critical force from GNB analyses
is used to calculate the column slenderness by Eqs. (7) or (8), are marked by red dots. On the basis of the results it can be
concluded, that the numerical results follows the analytical results. Based on the relative slenderness the corroded element
cannot be classified, because the partial corroded element and assuming total corrosion is not accurate enough to determine
the buckling mode. In Figure 42 a detail of Figure 41 can be seen about the II. zone, where the single dots marks the results
of partial corroded members by numerical model; nominating the buckling modes by black dot and red triangle. The lines
are the same as in Figure 41 and the calculation method is the same as in the case of Figure 41. The figure presents that the
relative slenderness of the partial corroded members cannot be determined according to standards in zone II. The accurate
calculation of it would be difficult, the undamaged parts of the corroded member give additional end stiffness depending on
the length of them, and it should be taken into consideration.
1,2
III. zone
1,0
̅
0,8
0,6
Class 4
0,4
0 5 10 15 20 25 30 35 40
/
Figure 41 Relative column slendernesses regard to FB and TFB modes, and calculated from numerical analysis - “B” pattern
- 54 -
FB TFB TFB appx. FB TFB
1,2
1,1
1,1
1,0
̅
1,0
0,9
0,9
0,8
17,5 17,7 17,9 18,1 18,3 18,5
/
Figure 42 Relative slendernesses in II. zone - calculated from the FE analysis (Tred = 50%)
It can be concluded, that it is important to get knowledge about the effect of the change of the corrosion parameters on the
buckling mode in the zone II. It is necessary because in the proposed checking procedure to calculate the remaining capacity
the buckling mode it is to be applied.
In Table 28 the various buckling modes are presented in the function of the extension (Ext ) and the position of corrosion (pc),
the / ratio and the relative slenderness ( ̅). If the / belongs to II. zone, the following observations can be done:
• The number of cases of the LPB and TFB modes is increased as / approaches the III. zone as shown in Table 28
Part 1. The results belong to elements with pattern “B” and ̅=0.7 relative slenderness ratio. In these cases the
decisive parameters are the position and the extension of corrosion to determine the buckling mode.
• Corrosion next to the support changes the buckling mode from flexural buckling to torsional-flexural buckling if the
extension of corrosion is 20%. In case of more than 20% extension of corrosion causes change of buckling mode,
from FB to TFB. The maximal value of the extension for local mode depends on the / ratio and on the relative
column slenderness.
• In case of pattern “B” the same buckling modes are observed (Table 28 Part 3) irrespectively of the initial / ratio
of the non-corroded element, beside the same parameters as /, Ext and pc.
• In case of pattern “A” the previous statement is not valid; in these cases the same /, Ext and pc parameters do
not result in the same buckling mode (Table 28 Part 2). The reason is that the same / ratio belongs to different
Tred because of the different initial / ratio of the non-corroded elements. Therefore Tred is the governing
parameter beside the / to determine the buckling mode.
- 55 -
̅ = 1.5 / = 23
Part 1 Part 2 Part 3
̅ = 0.7 ̅ = 1.15 / = 23
Pattern “B” Pattern “A” Pattern “B”
/ Tred Tred
Ext pc 13,2 13.9 14.3 14.6 15 60 70 66.7 60 70 66.7
20 0.20 TFB TFB TFB TFB TFB LPB LPB LPB TFB TFB TFB
20 0.47 FB FB FB FB FB GTFB LPB LPB TFB TFB TFB
20 0.73 FB FB FB FB FB GTFB LPB GTFB FB FB FB
20 1.00 FB FB FB FB FB GTFB GTFB GTFB FB FB FB
30 0.30 TFB TFB TFB TFB TFB LPB LPB LPB TFB TFB TFB
30 0.53 FB FB FB TFB TFB GTFB LPB GTFB TFB TFB TFB
30 0.77 FB FB FB FB TFB GTFB GTFB GTFB FB FB FB
30 1.00 FB FB FB FB FB GTFB GTFB GTFB FB FB FB
40 0.40 TFB TFB TFB TFB TFB GTFB LPB LPB TFB TFB TFB
40 0.70 FB FB FB TFB TFB GTFB GTFB GTFB FB FB FB
40 1.00 FB FB FB FB TFB GTFB GTFB GTFB FB FB FB
50 0.50 FB TFB TFB TFB TFB GTFB LPB GTFB FB FB FB
50 0.75 FB FB TFB TFB TFB GTFB GTFB GTFB FB FB FB
50 1.00 FB FB FB TFB TFB GTFB GTFB GTFB FB FB FB
70 0.70 FB TFB TFB TFB TFB GTFB GTFB GTFB FB FB FB
70 1.00 FB FB TFB TFB TFB GTFB GTFB GTFB FB FB FB
100 1.00 TFB TFB TFB TFB TFB GTFB GTFB GTFB FB FB FB
Table 28 Changes of buckling modes in terms of corrosion parameters
2.3.8. Ultimate behaviour
In the next phase of the numerical study GMNI simulations are carried out on the corroded elements. The observed behaviour
modes are generally determined by the applied buckling shapes (Table 28), as geometrical imperfections. Flexural buckling
is observed in the case of global buckling mode about the weak axis. Yield mechanism developed in the half of the member
on non-corroded member, but it shifted on the corroded members towards the damaged region. If the corrosion is
unsymmetrical within the cross-section, the axis of the buckling is rotated from the axis of the non-corroded case.
In the case of corrosion pattern “A” local buckling occurred, if the corrosion is significant. When the corrosion is not close to
the support a general case of torsional-flexural buckling mode is perceptible after the occurrence of the yield mechanism.
In the case of pattern “B” the typical behaviour is torsional-flexural buckling; similarly to the local plate buckling cases, global
flexural buckling is observed after the ultimate load is reached.
Figure 43 presents the results of three elements with pattern “A” corrosion and in Figure 44 the behaviour modes of pattern
“B” are presented. In the cases of local plate buckling behaviour one wave is dominant (as shown in Figure 43a), if the
corrosion is close to the support.
- 56 -
Figure 43 Behaviour modes - corrosion pattern “A”
As in the GNB analysis the values of / ratio are determined to predict the behaviour mode. On the basis of the GMNI results
it is concluded that the same values are valid as presented in Table 27.
The typical load-displacement curves are presented in Figure 45; the displacements of the middle point of the corroded part
are plotted on the load-horizontal displacement curves. Sudden decreasing can be observed on the curves after the ultimate
load is reached in the cases of local and torsional buckling.
- 57 -
45 45
40 FB 40
35 35
FB
Load [kN]
Load [kN]
30 30
25 TFB 25
20 20
15 15
10 10 TFB
5 5
LPB LPB
0 0
0 0,5 1 1,5 0 1 2 3 4
Vertical displacement [mm] Horizontal displacement [mm]
In Figure 46 the results of the RF-5 elements are shown; the resistances of the two patterns are plotted separately, the
different behaviour modes of the elements are signed. The tendency is different in the cases of the two patterns. By the
increasing of the / ratio non-linear decrease in the case of pattern “A” and exponential decrease in the case of pattern “B”
are observed. For pattern “A” the elements with local buckling are clearly separated, but it is not exact for pattern “B”. Large
standard deviation can be observed in the results for the same / ratio; the difference can be more than 20% in the
resistance.
(a) (b)
1,20 1,20
∆ - GTFB ∆ - FB
1,00 1,00
o - LPB o - TFB
Nb,Rd,S/Nb0,Rd,S
Nb,Rd,S/Nb0,Rd,S
0,80 0,80
0,60 0,60
0,40 0,40
0,20 “A” pattern 0,20 “B” pattern
0,00 0,00
0 20 40 60 0 20 40 60
/ /
Figure 46 Resistances of RF-5 elements in the function of the / parameter: (a) pattern “A”; (b) pattern “B”
The / ratios of the zones can be followed in Figures 47-48a, where the results of the elements with different relative
slenderness values are plotted. In these figures the necessary value of / is marked by red line and the sufficient value of
/ by green line. The sudden decrease in the resistance means that the behaviour is changing. The number of these cases
is decreasing as the relative slenderness is increasing.
- 58 -
(a) (b)
1,2 1,20
̅ = 0.7 1,00
1 ̅ = 1.15
0,8 0,80
Nb,Rd,S/Nb0,Rd,S
Nb,Rd,S/Nb0,Rd,S
0,6 0,60
0,4 0,40
0,2 0,20
0 0,00
0 20 40 60 0,0 20,0 40,0 60,0
ܾത/ݐ ܾത/ݐ
Figure 47 Resistances of RF-1, RF-3 elements (pattern “A”) in the function of the / parameter: (a) RF-1; (b) RF-3
(a) (b)
RF-1 20% RF-9 60%
1,20
RF-6 40% RF-11 20%
1,00 ̅ = 1.5 1,00
Nb,Rd,S/Nb0,Rd,S
0,80 0,80
Nb,Rd,S/Nb0,Rd,S
0,60 0,60
0,40 0,40
0,20
0,20
0,00
0,0 20,0 40,0 60,0 0,00
ܾത/ݐ 0 20 40 60 80 100 120
Ext [%]
Figure 48 Resistances: (a) RF-5 elements (pattern “A”) in the function of the / parameter, and (b) the effect of extension of
corrosion on the resistance decrease of pattern “B” elements
The effect of the increasing corrosion extension is different on the resistance of the analysed elements. In the case of pattern
“A” (for both behaviour modes GTFB and LPB) the increasing extension causes decrease in resistance, while by pattern “B”
the increasing extension causes decrease in resistance if the behaviour is FB and increase if it is TFB. Figure 48b presents the
results of four different elements with various Tred values and ”B” corrosion pattern in the function of the extension of
corrosion.
The effect of the position of the corrosion on the resistances is presented in Figure 49. The curves show the results of the
elements for two different / ratios and five relative slenderness values, where the extension is the same in all cases, just
the corrosion position is changed. The various colourful curves represent the different relative slenderness.
The tendencies are different, non-linear increasing or decreasing in terms of the relative slenderness. The / condition of
the behaviour mode can be followed e.g. in Figure 49a: in the case of relative slenderness 1.15. If the corrosion is close to the
support (pc =0.2) the decrease is greater than in the case when the corrosion is in the mid-length (pc =1.0). In general it can
be observed that the change in the corrosion position parameter causes non-linear decrease in resistances, as shown in Figure
- 59 -
49. If the thickness reduction and extension of corrosion are constant parameters, the change of corrosion position causes
modification in the behaviour mode and in the tendencies of the resistances. The different corrosion position may cause a
maximum of 17% difference in resistance.
(a) (b)
1,20 1,10
/ = 23 / = 18
1,00 1,00
0,7 0,90 0,7
Nb,Rd,S/Nb0,Rd,S
0,80
Nb,Rd,S/Nb0,Rd,S
(a) (b)
1,20 1,20
1,00 1,00
Nb,Rd,S/Nb0,Rd,S
0,80 0,80
Nb,Rd,S/Nb0,Rd,S
0,60 0,60
0,40 0,40
0,20 0,20
0,00 0,00
0 50 100 0 50 100
Tred [%] Mcorr [%]
Figure 50 Resistances of RF-1 elements in the function of the corrosion parameters: (a) Tred ; (b) Mcorr
In Figure 51a the results belonging to elements having pattern “A” and the solid line marks the lowest design buckling
resistance values of all (both of pattern “A” and “B”) of the analysed RF-1 elements in the function of the thickness reduction.
- 60 -
If the estimation is determined by this tendency, the difference between the calculated and the estimated values can be
significant. In Figure 51b the curves are determined by the same process as in Figure 51a, where the solid lines for all of the
investigated elements RF-1-11 are shown. The tendency is different by the different elements, therefore simplified formula
cannot be determined.
(a) (b)
1,20 0,9
0,8
1,00
0,7
Nb,Rd,S/Nb0,Rd,S
0,80 0,6
Nb,Rd,S/Nb0,Rd,S
0,5
0,60
0,4
0,40 0,3
0,2
0,20
0,1
0,00 0
0 50 100 0 50 100
Tred [%] Tred [%]
Figure 51 Approximation and tendencies of resistance reduction for pattern “A”: (a) RF-1 element; (b) all of the elements
Regarding the worst cases, a conservative estimation can be done in terms of the ultimate behaviour. Two corrosion
parameters, as the position of corrosion and the extension of corrosion are left out of the estimation. For flexural buckling
and general case of torsional-flexural buckling the behaviour of either pattern “A” or “B” elements, the resistance decrease
is approximately equal to the cross-section reduction factor (Nb,Rd,S /Nb0,Rd,S ≈ 0.01Mcorr). In the cases of local plate or torsional-
flexural buckling, the tendencies are different from global cases and depend on the relative column slenderness. Figure 52a
presents the estimation (red line) in the cases of pattern “A” and ̅=0.7, while in Figure 52b the results of pattern “B” and
̅=0.7 are shown. The estimation is linear in all cases.
(a) (b)
1,20 1,20
1,00 1,00
Nb,Rd,S/Nb0,Rd,S
Nb,Rd,S/Nb0,Rd,S
0,80 0,80
0,60 0,60
0,40 0,40
0,20 0,20
0,00 0,00
0 20 40 60 0 50 100
Mcorr [%] Mcorr [%]
Figure 52 Conservative estimation: (a) pattern “A” and ̅=0.7; (b) pattern “B” and ̅=0.7
Table 29 contains the equation of the linear approximation of the local plate buckling and the torsional-flexural buckling
cases, the dimension of Mcorr is percentage [%] in the equations.
- 61 -
Nb,Rd,S/Nb0,Rd,S
Pattern “A” Pattern “B”
0.7 0.995 0.023 ∙ corr 1.098 0.013 ∙ corr
1.15 0.73 0.016 ∙ corr 1.53 0.018 ∙ corr
1.5 1 0.0225 ∙ corr 0.79 0.009 ∙ corr
Table 29 Equations of the conservative approximation LPB and the TFB mode
If the behaviour mode is flexural buckling or general case of torsional-flexural buckling a more accurate approximation
formula can be determined, what can be calculated from the thickness reduction (Tred), the ratio of a modified length (Lmod)
and the total length (L ), as it is shown in Figure 53. The Lmod parameter can be calculated by Eq. (23). Part of the results is
shown in Figure 54 in the function of Lmod /L. The effect of the various extension and corrosion position can be considered by
this parameter.
∙ ∙
∙ 0.01
(23)
,,
⁄
,,
∙ ∙ ∙
⁄
∙ ∙
⁄
(24)
The applied a, b, c and d constants are dependent on the relative (non-corroded) column slenderness and on the pattern. The
values of the constants are given in Appendix B Table 1.
(a) (b)
1,05
1,00
Nb,Rd,S/Nb0,Rd,S
0,95
0,90 Tred [%]
Nb,Rd,S/Nb0,Rd,S
0,85 20%
30%
0,80
40%
0,75
50%
0,70 Lmod/L
0,00 0,50 1,00 Tred [%]
Lmod/L
Figure 54 Resistance reduction results in the function of the modified length parameter: (a) RF-1 elements; (b) the resistance surface
- 62 -
Comparison of the design codes and the proposed methods
The chord members of transmission columns are loaded through on both legs of the cross-section, therefore they might be
consider as a centrically loaded compressive members. The codes [1], [35] and [36] contain the design method of leg
members. In this section the results of numerical analyses are compared to the values according to the three codes
considering the corrosion patterns. The calculated design buckling resistance according to codes are determined by assuming
the followings:
In Figures 55-56 the numerical results are plotted together with the results obtained by the recommendations of codes. The
blue dots mark the specimen with general case of torsional-flexural buckling and the yellow dots mark the specimens with
local plate buckling. The dotted line marks the bound between Class 3 and Class 4 according to EC3-3-1. Note that the bound
is well predicts the bound of the ultimate behaviour modes in case of λ=0.7 relative slenderness ratio (Figure 55) but in case
of λ=1.5 relative slenderness ratio it is not acceptable (Figure 56).
The estimation of ultimate buckling according to EC3-3-1 is not suitable, because the difference is too much between the
results by codes and by FEA. In Figures 55-56 the curve called Nc means the resistance to normal forces of the corroded leg of
the members. The gross cross-section of just one leg is calculated using the effective width and the reduced thickness. This
estimation is better for corroded members having failure of local plate buckling.
70
Pattern “A” ̅ 0.7
60
EC3-1-1
50 EC3-3-1
EN 50341
Nb,Rd [kN]
40 Global
Local
30
20
10 ୡ ୰ୣୢ ∙ ୣ ∙ ୷
0
0 10 20 30 40 50 60
/
Figure 55 Results by standards and numerical analyses - pattern “A” and ̅=0.7
- 63 -
̅ = 1.5
40
Pattern “A”
35
30
EC3-1-1
25 EC3-3-1
EN 50341
Nb,Rd [kN]
20 Global
Local
15
10
5
ୡ = ୰ୣୢ ∙ ୣ ∙ ୷
0
0 10 20 30 40 50 60
ܾത/ݐ
Figure 56 Results by standards and numerical analyses - pattern “A” and ̅=1.5
The codes in general cannot estimate the value of the design buckling resistance of the corroded angle members, but in the
case of members with failure mode of flexural buckling and general case of torsional-flexural buckling, they can estimate the
design buckling resistance ratio of the corroded and the non-corroded angle members on the safe side, as it can be seen in
Figures 57-58. In these figures the vertical axis is the ratio of the ultimate buckling force of corroded (Nb,Rd) and non-corroded
members (Nb0,Rd) and the red continuous line marks the applied standard’s results. The I. - III. zones are marked by vertical
lines.
̅ = 1.15
1,20
Pattern “A”
1,00
III. zone EN 50341
0,80
GFTB
Global
Nb,Rd/Nb0,Rd
0,60 LPB
Local
0,40
0,20 I. zone
0,00
0 10 20 30 40 50 60
ܾത/ݐ
Figure 57 Results by standards and numerical analyses - pattern “A” and ̅=1.15
- 64 -
̅ = 1.15
1,20
Pattern “B”
1,00
Nb,Rd/Nb0,Rd
0,80
III. zone ECEC3-1-1
1993-1-1
FB
Global
0,60
FTB torsional
Local
0,40
0,20 I. zone
0,00
0 10 20 30 40 50 60
ܾത/ݐ
Figure 58 Results by standards and numerical analyses - pattern “B” and ̅=1.15
In the followings the comparison of the numerical result with the design buckling resistance using critical force of Timoshenko
is detailed. In Figure 59 results are plotted by the following assumptions. Corrosion is always on the middle of the length, it
means pc = 1 in all cases. Each pair of lines means different thickness reduction. The continuous lines mark the results of finite
element analyses, supposing “B” corrosion pattern. The dotted lines mark the results using the Timoshenko method for
determining the critical force to calculate the design buckling resistance according to EC3-1-1 with fy = 235N/mm2.
On the basis of Figure 59 it can be concluded, that by 20-30-40% thickness reduction the analytical results are in accordance
with the numerical results. In the case of 50% thickness reduction the difference between the numerical and analytical results
is approximately 10%. If the corrosion is significant (Tred = 60-70%), the results are different calculating by the two methods.
In these cases the ultimate behaviour mode is torsional buckling, what cannot be followed by the applied the analytical
method.
1,00
0,90 Ansys 20%
Timoshenko 20%
0,80
Ansys 30%
0,70
Timoshenko 30%
0,60 Ansys 40%
Nb,Rd/Nb0,Rd
- 65 -
Conclusion
On the basis of the experimental and numerical study on corroded angle chord member under centric compression the
following conclusions can be done:
• The ultimate behaviour mode of corroded angle members depends on the measure and the appearance of
corrosion. Four different types of ultimate behaviour modes can be distinguished: flexural buckling (FB), local plate
buckling (LPB), torsional-flexural buckling (TFB) of the corroded zone and general case of torsional-flexural
buckling (GFTB).
• The ultimate behaviour mode depends on the corrosion pattern (“A”: GTFB or LPB and “B”: FB or TFB) and on the
width-to-thickness ratio regarding to the corroded zone.
• The ultimate buckling force of corroded members is decreased due to thickness reduction caused by corrosion. The
decrease of the ultimate buckling force is more significant if the ultimate behaviour mode is plate buckling.
• The estimation of the design buckling resistance of corroded angle members is performed by using Mcorr maximal
cross-section reduction. The ratio of the design buckling resistance of corroded and non-corroded members can be
more accurate calculated by the proposed equation Eq. (24).
• The estimation of the ultimate force of corroded members by the recommendations of different codes and average
thickness reduction is not accurate enough; for the determination of the ratio, however, it is acceptable in the case
of global ultimate behaviour mode.
- 66 -
Buckling resistances of eccentrically connected non-corroded and corroded angle members
under compression
Generally the angle members are connected to other structural elements by bolted or welded connection on one of the legs.
In bolted connections usually one and two bolts are applied. In lattice towers these are bracing members. This type of
connection causes eccentricity in the loading regarding to the centre of the gravity of the angle. The eccentricity through the
connection can modify the behaviour and the resistance of the corroded elements.
The aim of the current study is to determine the measure of the change in ultimate buckling force, in design buckling
resistance and the changes in the ultimate failure mode in term of the various levels of corrosion and the applied connection
type (one bolted and two bolted).
Experimental study
Table 30 presents illustrative figures of the test specimens. The black colour marks the corrosion. Its arrangement and the
maximal cross-section reduction (Mcorr) with respect to the two groups (PI - one bolt, PII - two bolts) can be found in the table;
tred means the reduced thickness. The Mcorr is calculated on the basis of measurements in the same way as in the previously
completed tests. After the ID of the specimens in the parenthesis the numbers mean that the corrosion is on one leg or on
both of the legs.
The material properties are as follows: fy = 345 N/mm2, fu = 470 N/mm2 on the basis of the tensile coupon tests (see Section
2.1.1).
The applied arrangements of corrosion are determined similar way as in the case of chord members. Almost each element
has a counterpart in the group of analysed chord members. In specimens P-2 and P-3 the corrosion is applied in one leg and
in specimen P-2 the loading is applied on the same leg as the corrosion, while in the case of specimen P-3 on the other leg.
In the cases of P-4 - P-6 and P-8 - P-10 specimens the corroded surface is equal, but the position is different. The corrosion is
applied on both of the legs in the cases of these elements. In the case of P-7 within the cross-section only one of the legs is
corroded, while in P-8 both of the legs are damaged.
- 67 -
PI PII
ID Location of corrosion tred Mcorr tred Mcorr
[mm] [%] [mm] [%]
10
10
10
10
10
10
10
30
40
45
35
830/2 = 415
830/2 = 415
30
10 40 10
790/2 = 395
790/2 = 395
10 40 10
- 68 -
In the test set-up a welded T-section element is connected to one of the legs of the specimen by bolted connection (gusset
plate connection), using one or two bolts. The bolts are M14 and have the following nominal material properties: fyb = 640
N/mm2, fub = 800 N/mm2. The T-elements are supported by hinge edges in the plane of the gusset plate. Load cell is applied
to measure the load. Beside the load, the vertical and horizontal displacements are measured. Horizontal displacement is
measured by a pulley systems with displacement transducer in the half of the member length. The vertical displacement of
the head of the loading machine is measured by the built-in measurement system. The end connection and test set-up are
shown in Figures 60-61.
The other identified failure mode is local plate buckling with initial flexural behaviour, as shown in Figure 62c. The plate
buckling appears in the weakened area. It is also observed on P-7 - P-10 specimens. In these cases the thickness reduction is
appr. 75%, therefore the width-to-thickness ratio (/) is increasing, and the corroded cross-section is getting from Class 1
or 2 to Class 4, according to EC3-3-1.
The typical load and horizontal/vertical displacement curves can be seen in Figure 63. The behaviour modes can be followed
on the load-displacement curves: in the case of local plate buckling a sudden decrease appears in the descending brunch.
During the test slip is observed, especially in elements with two bolts. It can be detected in the curves of PII-3 and PII-8; after
- 69 -
the local slip the initial stiffness is recovered. This slip can be seen on the load-vertical displacement curves, due to the
redistribution of the forces in the bolts.
On the basis of the test results it can be concluded, that the ultimate failure mode depends on the rate of corrosion but
irrespective of the number of bolts in the connection.
Figure 62 Failure modes (a)-(b) general case of torsional-flexural buckling; (c) local plate buckling
(a) (b)
35 35
30 30
PI-3 PII-3 PII-3
25 25 PI-3
Load [kN]
Load [kN]
20 20
PI-3 PI-3
15 15
PI-8 PII-8 PII-3 PII-8 PII-3
10 10
PI-8 PI-8
5 5 PI-8
PII-8 PII-8
0 0
0 2 4 6 8 0 10 20
Vertical displacement [mm] Horizontal displacement [mm]
- 70 -
Table 31 contains the results of the non-corroded specimens, the calculated mean values and standard deviation. In the case
of non-corroded PI elements the ultimate buckling force is lower, than that of PII elements.
• Almost in every cases the reduction of the ultimate buckling force comparing to the non-corroded elements is
larger applying two bolts in the connection; the difference in decrease is ~12-15%.
• There is almost no difference between the ultimate buckling forces if the corrosion is on the same leg as the bolted
connection (P-2) and if the corrosion is on the other leg (P-3).
• Localized corrosion close to the support (P-8) causes bigger decrease in ultimate buckling force, than corrosion in
the middle (P-10). The difference in ultimate buckling force is ~18% in these cases.
• In the case of P-4 - P-6 specimens, the influence of the position of corrosion is not significant (2-3%). In these cases
the cross-section reduction does not exceed 50%.
The above observations are obtained by one-one sample tests. Further analysis are required, because standard deviation and
mean value cannot be calculated from one sample and the differences can be due to the inaccuracy of the test arrangement
and the preparation of specimen.
1 35
1 bolt 30 1 bolt
0,8
2 bolts 25 2 bolts
Nb,m/Nb0,m
Nb.m [kN]
0,6 20
0,4 15
10
0,2
5
0 0
P-2 P-3 P-4 P-5 P-6 P-7 P-8 P-9 P-10 P-2 P-3 P-4 P-5 P-6 P-7 P-8 P-9 P-10
100
100
100
100
247.5
147.5
147.5
147.5
345
395
485
590
590
Specimen
295
295
295
295
590
590
100
100
295
147.5
147.5
147.5
345
100
100
247.5
205
100
100
100
100
100
100
100
100
P-1 P-2 P-3 P-4 P-5 P-6 P-7 P-8 P-9 P-10
ID
PI GTFB GTFB GTFB GTFB GTFB GTFB LPB LPB LPB LPB
PII GFTB GTFB GTFB GTFB GTFB GTFB LPB LPB LPB LPB
tred [mm]
PI 4 3.32 3.45 2.52 2.56 2.56 1.85 1.43 1.69 1.89
PII 4 3.72 3.33 2.30 2.16 2.55 1.15 1.68 1.53 1.61
Nb,m [kN]
PI 30.7 29.1 28.8 19.9 19.1 20.1 15.8 11.2 14.7 11.6
PII 34.8 28.9 27.8 17.4 16.6 21.9 16.4 14.2 11.0 7.8
Nb,m / Nb0,m
PI 1 0.95 0.94 0.65 0.62 0.66 0.52 0.37 0.48 0.38
PII 1 0.83 0.8 0.5 0.48 0.63 0.47 0.41 0.32 0.22
- 72 -
In Figures 65-66 the black continuous line marks this approximation. The results are plotted in the function of the maximal
cross-section decrease and the applied connection is marked in Figure 65. On the left curve the specimens having GTFB mode
are plotted and on the right curve specimens with LPB mode are presented. All of the results are plotted in Figure 67 and the
different failure modes are marked. The following conclusion can be done based on the curves:
• The approximation is on the safe side if the specimen is connected by one bolt and the failure mode is GTFB.
• In the cases of PII specimens based on this approximation the calculated reduction is lower value then the reduction
of ultimate buckling force in test. Almost all the result is under the line of the approximation by Eq. (25).The
maximal difference is 10%.
• The difference from the approximation is bigger in the case of local failure. The standard deviation of the ultimate
buckling forces is also bigger. Irrespectively of the connection type the maximal difference is ~15-20%.
(a) (b)
0,6
0,4
0,4
PI 0,2 PI
0,2 PII PII
0 0
0 20 40 60 0 20 40 60 80
Mcorr [%] Mcorr [%]
Figure 65 Results of the test (a) with global mode and (b) with local failure mode and the approximation by Eq. (25)
• Comparing the results of test and approximation by Eq. (25) it can be concluded that this estimation is not generally
applicable to predict accurately the ratio of the ultimate buckling forces. The separation of the specimens by
connection type is not enough, the failure mode is also to be considered. On the basis of Figure 66 the specimens
cannot be divided into groups according to Mcorr. It is not possible to determine a boundary line between the
specimens with LPB or GTFB ultimate behaviour modes using the Mcorr parameter. At the same time based on the
/ the specimens can be classified into groups according to the ultimate behaviour mode, as it can be seen in
Figure 67.
• In term of / ratio the failure mode can be determined and the corroded specimens can be divided into groups in
accordance with the prospective ultimate behaviour mode. A necessary limit value of / ratio for the classification
is 18, based on the results of the tests. If the specimen has lower / value than the limit value the ultimate failure
is global, otherwise it is local plate buckling, the dash dot line marks the determined necessary limit value. The
dotted line marks the bound of Class 3 to Class 4 (see Figure 67), which is not separate the specimens appropriately
according to ultimate behaviour modes.
- 73 -
1
0,8
Nb,m/Nb0,m
0,6
0,4
GTFB
0,2 LPB
0
0 20 40 60 80
Mcorr [%]
1,2
Based on tests
1
LPB
Local
0,8 GTFB
Global
0,6
Nb,m/Nb0,m
0,4
EC3-3-1
0,2
0
0 10 20 30 40
ܾത/ݐ
Figure 67 Results in the function of the / ratio with the bounds of classifications
3.1.5. Ultimate buckling force according to the different Eurocode specifications
The results assisting by test are compared to the design buckling resistance according to the specification. In the calculation
EC3-3-1 [35] and EN 50341 [36] codes are applied by the following assumptions:
• The thickness reduction is supposed to be on both of the member legs along the whole length.
• Effective characteristics of the cross-section are applied to determine the buckling resistance according to the codes
(Eqs. (12)-(15)). The applied effective relative slenderness ratios are summarized in Table 3, in Section 1.7. In the
calculation of design buckling resistance the effective cross-section is used, along the whole length.
• The yield strength: fy = 345 N/mm2.
• The y-y axis is parallel to the connected leg.
In the first step the non-corroded elements are analysed. In Figure 68 the ultimate buckling forces are calculated as design
buckling resistances according to design codes [35-36] and the test results are shown and compared to each other. The
following can be concluded on the basis of Figure 68:
- 74 -
• Generally the design buckling force for one bolted connection is lower, than in the case of two bolted connection
according to EC3-3-1.
• In the calculation according to EN 50341 if the value of the relative slenderness is lower than √2 the same effective
relative slenderness must be used in one and two bolted cases, too. Therefore the design buckling force is
independent of the number of the bolts.
• The best approach of the values of the design buckling force is given by EN 50341 for non-corroded PI specimens,
the difference is 5.7%.
• In case of non-corroded PII specimen the EN 50341 and EC3-3-1 give almost the same value, the difference is 6%
in average.
• In the case of non-corroded PI the design value according to EC3-3-1 is rather small compared to the others.
• In the case of one bolted connection EC3-3-1 specifies a 0.8 factor for the design buckling resistance comparing to
two bolted connection. On the basis of the test results such difference is not observed between these cases, the
observed difference is 12%.
40
Ultimate buckling
30
force [kN]
20
10
0
Test EC3-3-1 EN 50341
PI - 1 30,70 19,95 32,45
PII - 2 34,80 32,98 32,45
Figure 68 Design buckling resistance of non-corroded members by codes and assisting by test results
In Figures 69-70 the values of design buckling resistance of the corroded angle members are plotted. The result marked by
Test is the design buckling resistance assisting by test. On the horizontal axis is the / ratio. The results are separated
according to corrosion patterns and the number of bolts. The dashed line marks the bound of Class 3 to Class 4 in Figures 69-
70 according to EC3-3-1. On the basis of the figures the followings can be concluded:
• The EN 50341 code approaches the test results better than the others in the case of angle members connected by
one bolt (see Figure 69).
• In the case of angle members connected by two bolts the codes give acceptable results.
• In Figures 69-70 all of the results by test are in the right side of the dashed line, it means all of them are in Class 4.
On the basis of the test some of them have global buckling behaviour mode. It means that using the classification
method the corroded members cannot divided into groups according to the ultimate behaviour modes.
- 75 -
45 45
1 Bolt "A" pattern 1 Bolt "B" pattern
40
35 EC3-3-1 35
30 EC3-3-1
30 Test
25 Test
25 EN 50341
EN 50341
20 20
15 15
10 10
5 5
0 0
0 10 20 30 40 0 10 20 30 40
/ /
Figure 69 Design buckling resistances by codes and assisting by test result - 1 bolt, “A” - “B” patterns
50 45
2 Bolt "A" pattern 2 Bolt "B" pattern
45 40
Ultimate buckling force [kN]
Ultimate buckling force [kN]
40 EC3-3-1 35 EC3-3-1
35 30
Test Test
30
EN 50341 25 EN 50341
25
20
20
15 15
10 10
5 5
0 0
0 10 20 30 40 0 10 20 30 40
/ /
Figure 70 Design buckling resistances by codes and assisting by test result - 2 bolts, “A” - “B” patterns
Numerical study
The aim of the numerical study is to extend the test results and to analyse other corroded angle members with various relative
slenderness’s. In this section the applied models, a parametric study, and the result of the study are detailed. Two numerical
models are used to model the eccentrically connected corroded and non-corroded members.
- 76 -
Workbench model - FE model 1
The first model is a solid model in Workbench using brick finite elements; it is shown in Figure 71. This model contains the
full test set-up, the specimen, the bolts, and the T supporting element. The corrosion is modelled by thickness reduction. In
the model the mean value of the thickness reduction is applied on the basis of the measurement, which is completed before
the tests. Standard contact elements are used between the angle member and bolts; angle member and T-elements; bolts
and T-elements. The standard (unilateral) contact means that normal pressure equals to zero if separation occurs. The
coefficient of friction is 0.2 in the model. The material model of steel is the same as used previously: linear elastic - perfectly
plastic. The measured material properties of specimens are used in the model. The mesh is denser around the bolts (see Figure
72). The support is parallel to the bolted leg and kinematic load is applied in the direction of z-axis. The rotation of x-axis is
released, but rotation of the other axes are fixed.
Figure 71 FE model 1
Mixed model - FE model 2
This model is built-up in Ansys APDL. The following elements are used in the model: (i) SHELL181, (ii) BEAM188 and (iii)
BEAM44. SHELL181 finite element is used for the angle and support plates. The segments of bolts, as head, stud, and nut are
modelled by beam elements with equivalent properties. In the case of the stud the beam element (BEAM188) has the same
cross-section properties as the applied bolts in the tests. BEAM188 [41] is suitable for analysing slender to moderately
stubby/thick beam structures. This element is based on Timoshenko beam theory and shear deformation effects are included.
In the case of nut and head BEAM44 [41] is applied, which is a uniaxial element with tension, compression, torsion, and
bending capabilities. The element has six degrees of freedom at each node: translations in the nodal x, y, and z directions and
rotations about the nodal x, y, and z-axes. These beam elements are connected by joint end to the shell elements and by fix
end to the stud. The parameters of the cross-section are determined to get the same cross-section properties as the real bolt.
The model just contains the leg of the T support element, which is parallel to the z-axis, which is the axis of the analysed
element. At the ends of the support the CERIG Ansys command is applied. This command defines a rigid region. The centre of
this rigid region is marked in Figure 72a. The contact behaviour is taken into account by surface contact elements (CONTA173
and TARGE170). This mixed model is shown in Figures 72a-b. The middle of the shell elements is shifted in the corroded
region, a transition element is used to keep the continuity (see Figures 72c-d). It is justified to follow the tests accurately.
- 77 -
(a) (b)
(c) (d)
Figure 72 FE model 2: (a) element with one bolt; (b) element with two bolts, (c)-(d) transition in the model
3.2.2. Calibration and validation
The calibration is carried out in two steps. In the process the measured yield stress fy = 345 N/mm2 is used in both models. As
a first step the magnitude of the geometrical initial imperfections of test specimen are determined to get the same ultimate
buckling force in the numerical analysis by FE model 1. This is a solid model and therefore it is more accurate. After that, the
FE model 2 is validated applying the previously determined magnitude of initial geometrical imperfections. In some cases
the initial imperfection had to be modified.
The same levels of analysis are used in the verification as in the case of chord members:
In GMNI analyses the result of GNB is applied to model the geometrical imperfection.
- 78 -
In the first step in GMNI analysis by FE model 1 the initial magnitude of initial geometrical imperfection is assumed as L/200
in accordance with the recommendation of the EC3-1-1 standard. The results of the analyses follow the test results in ultimate
behaviour, but there are differences in the ultimate buckling forces. Global flexural buckling with small rotation is observed
in P-1 - P-6 elements. After an initial global behaviour phase, plate buckling is occurred in the case of P-7 - P-10 elements.
Part of the results of the calibration is the following. The buckling modes from GNB analyses and the ultimate behaviour
modes of PII-5, PI-5 and PII-8 elements are presented in Figures 73-74. The figures present the results of both FE model 1
and FE model 2. In the case of PII-5 specimen, torsional-flexural buckling in corroded zone is observed in the stability analysis,
but the ultimate behaviour mode is general case of torsional-flexural in the GMNI analysis such as in the tests. Both the
buckling mode and the behaviour mode are general case of torsional-flexural buckling in the case of PI-5. This occurrence is
observed on specimens PI-4 and PII-4, but not in the case of PI-6 and PII-6. The difference between the cross-section
reductions may lead to the different stability mode in these two cases. Specimen PII-8 is an example of local plate buckling
ultimate behaviour mode. Figure 74 contains not only the results of FE model 1 but also the ultimate behaviour modes by FE
model 2.
On the basis of the results it can be concluded, that FE model 2 can also follow the ultimate behaviour, similarly to the FE
model 1, as it can be seen in Figures 73-74. Table 33 contains the ultimate buckling forces of the analyses of PI members
using the two different FE models in the table Nb,m is the test results and Ncorr is the results of the FE analysis. The magnitudes
of the initial geometrical imperfection and the relative magnitudes of PI elements are detailed in Table 33. Table 34
summarizes the data of the analyses and results of PII members.
ID FE model 1 FE model 2
PII-5
PI-5
PII-8
- 79 -
ID FE model 1 FE model 1
PII-5
PI-5
PII-8
Figure 74 GMNI analysis - ultimate behaviour modes of PII-5, PI-5 and PII-8
The typical load-horizontal displacement curves also follow the characteristics of the curves of the tests. The horizontal
displacement perpendicular to the axis of the angle member is measured at the same place such as in the test. Figure 75
presents the curves of the test and of the FE model 1, while in Figure 76 the curve of FE model 2 can be seen.
PI elements
FE model 1 FE model 2
ID Nb,m [kN] Imp Ncorr /Nb,m Imp Ncorr /Nb,m
Rel. Imp Ncorr [kN] Rel. Imp NS [kN]
[mm] [%] [mm] [%]
PI - 1 30.7 2.1 L/400 29.6 3.4 4 L/200 30.7 0.1
PI - 2 29.1 0.5 L/1660 28.3 3.0 2.5 L/320 29.1 -0.2
PI - 3 28.8 0.5 L/1660 27.9 3.1 0.53 L/1500 27 6.8
PI - 4 20.0 1.0 L/800 19.3 3.3 4 L/200 20.9 4.4
PI - 5 19.1 1.0 L/830 19.0 0.5 4 L/200 20.0 4.5
PI - 6 20.1 2.0 L/420 20.5 -2 4 L/200 19.1 5.0
PI - 7 15.8 4.0 L/210 16.7 -5.9 4 L/200 15.5 -2
PI - 8 11.2 4.0 L/210 11.3 -1.5 1.5 L/530 11.5 2.5
PI - 9 14.7 2.0 L/420 14.1 3.9 2 L/400 14.8 1.1
PI - 10 11.6 4.0 L/210 11.4 2.3 4 L/200 11.6 -0.2
Table 33 Results by FE model 1 and FE model 2 of PI elements
- 80 -
PII elements
FE model 1 FE model 2
ID Nb,m [kN] Imp Imp
Rel. Imp Ncorr [kN] Ncorr/Nb,m [%] Rel. Imp Ncorr [kN] Ncorr/Nb,m [%]
[mm] [mm]
PII - 1 34.8 1.7 L/490 34.1 2.1 1.7 L/470 33.7 -3.3
PII - 2 28.9 4.2 L/200 28.8 0.2 4 L/200 29.4 1.7
PII - 3 27.8 3.1 L/270 27.6 0.8 4 L/200 27 2.9
PII - 4 17.4 4.2 L/200 18.0 -3.8 4.2 L/200 18.7 7.1
PII - 5 16.6 3.3 L/250 16.3 2.2 3.3 L/240 17.5 4.8
PII - 6 21.9 2.1 L/200 21.7 1.1 4 L/200 23.6 7.7
PII - 7 16.4 4.0 L/210 19.4 -18.4 4 L/200 19.9 -21.3
PII - 8 14.2 2.0 L/420 14.6 -3.1 2 L/200 14.3 0.7
PII - 9 11.0 4.0 L/210 11.9 -8.5 2.5 L/320 10.8 -1.7
PII - 10 7.8 4.0 L/210 8.9 -13.8 4 L/400 8.2 4.0
Table 34 Results by FE model 1 and FE model 2 of PII elements
35 12
PI-1 PI-8
Ultimate buckling force [kN]
30
Ultimate buckling force [kN]
10
25
8
20
6
15
10 Test 4 Test
5 FE model 1 2 FE model 1
0 0
0 10 20 0 2 4 6
Horizontal displacement [mm] Horizontal displacement [mm]
35
PII - 2
Ultiamte buckling force [kN]
30
25
20
15
TEST
Test
10 FE model
MIXED 2
Model
5
0
0 10 20 30
Horizontal displacement [mm]
- 81 -
buckling resistance. This numerical study program is similar to the previous which is completed on axially loading angle
members namely on chord members (Section 2.3.5). The same analyses are used, bifurcation stability (GNB) and
geometrically and material non-linear imperfect analyses (GMNI). The analyses are carried out by the following assumptions
and parameters:
• In the GMNI analysis the applied initial geometrical imperfection is derived from GNB analysis with the magnitude
of L/200 in all cases, irrespectively of the buckling mode. This is the same as in the previous numerical study of
chord members (Section 2.3.5).
• The corrosion parameters, corrosion patterns, thickness reduction (Tred), extension of corrosion (Ext ) and corrosion
position (pc), are the same as previously in Section 2.3.5.
• The parameters are slightly different, since the corroded connection is not the object of this study. Therefore an
original connection is assumed in the parametric study and the corrosion starts from the end of the gusset plate, as
it is shown in Figure 77. Table 35 contains the values of the corrosion parameters.
The initial / (width-to-thickness) ratio is 8, as the basis of the cross-section classification of angle members by EC3-3-1. In
the further evaluation of the corroded elements the / is the ratio of the reduced thickness and the width of the legs, as a
corrosion parameter. It should be noted, that the class of the cross-cross section is 4 from 50% thickness reduction.
General case of torsional-flexural buckling and torsional-flexural buckling on the corroded zone is observed in the case of
pattern “B”. From this point of view the results of GNB analyses showed the same results as in the case of chord members.
There is no difference in the buckling modes regardless of the connection type. In Figure 78 the observed buckling shapes are
presented: (a) local plate buckling, (b) torsional-flexural buckling and (c) general case of torsional-flexural buckling.
The GBTUL analysis cannot be used for this purpose, because the program is not able to consider accurately the bolted support
conditions.
The I. - III. zones are determined as in the case of chord members, and the determined values are detailed in Table 36
regarding to the slenderness and the applied number of bolts. It can be concluded that the values are higher than in the case
of chord members. The reduced cross-sections are in Class 4 if / is greater than 15 according to EC3-3-1. On the basis of
Table 36 it can be concluded, that the limit value of / is higher than the value by code, to separate the elements according
to the ultimate behaviour mode. The difference between the values of / of numerical analyses and of code is significant
especially elements with high slenderness ratio.
λത = 1,15
“A” pattern
λത = 1,5
1 bolted λത = 1,92
connection
λത = 1,15
”B” pattern
λത = 1,5
λത = 1,92
ܾത/ݐ
Figure 79 Schematic figure of I. - III. Zones and the relevant buckling modes - 1 bolted connection
λത = 1,15
“A” pattern
λത = 1,5
2 bolted λത = 1,92
connection
λത = 1,15
”B” pattern λത = 1,5
λത = 1,92
ܾത/ݐ
Figure 80 Schematic figure of I. - III. Zones and the relevant buckling modes - 2 bolted connection
- 84 -
(a) (b) (c)
Figure 81 Typical behaviour modes (a) local plate buckling; (b) torsional flexural buckling; (c) general case of torsional flexural buckling
Buckling resistances
In this section the tendencies of reduction of the design buckling resistance in term of the parameters are presented. In
Figures 82-84 the tendencies of the results are shown. In these figures the vertical axis is the ratio of the design buckling
resistances of the corroded and the non-corroded angle members. The horizontal axes are the three corrosion parameters:
the ⁄ ratio, the extension, and the position of the corrosion.
On the left side of the figures the results of members connected by two bolts and assuming “A” corrosion pattern are plotted,
while on the right side results of members connected by one bolt and “B” pattern can be seen. The relative column
slenderness ratio of the non-corroded angle members is 1.5, in the case of the plotted results.
On the basis of the results it can be concluded that the characteristics of the results are the same if the relative column
slenderness ratio is lower (̅=1.15) or higher (̅=1.92). The main difference between the characteristics of the results is
observed taken as a function of the corrosion pattern. The reduction of the ultimate buckling force is significant if the corrosion
occurs on one leg. This observation is found in the case of chord members, too. The sudden decrease is attributable to the
consequences of the change of the ultimate behaviour mode.
As a functions of extension of corrosion and corrosion position the tendencies of the decrease are non-linear. The results are
separated in the case of pattern “A”, as it can be seen on the left side in Figures 83-84, due to the changing of the behaviour
mode. In these figures in the columns, the parameter of the axis is fixed and the two others is changing. The change of the
behaviour, however, slightly modified the tendencies of the decrease of the design buckling resistance.
Assuming that the corrosion is on the middle of the length the effect of the extension on the design buckling resistance can
be concluded as follows:
- 85 -
• In the case of pattern “A” the average decrease is approximately 4% by increasing extension, if the ultimate
behaviour mode is global.
• If the ultimate behaviour is local plate buckling the difference is almost negligible by increasing extension.
• These conclusions are valid to both types of connections.
• In the case of pattern “B” the decrease of the ultimate buckling force by increasing extension is higher than in the
case of pattern “A”, but the measure depends on the connection type. In the case of connection with one bolt, the
decrease is 5% and with two bolts it is 7.5%. The measure of the decrease is slightly changing by increasing
thickness reduction.
1,2 1,0
“A” pattern "B" pattern
1 0,8
2 bolted 1 bolted
0,8
Nb,Rd,S/Nb0,Rd,S
0,6
Nb,Rd,S/Nb0,Rd,S
= .
= .
0,6
0,4
0,4
0,2 0,2
0 0,0
0 10 20 30 40 0 10 20 30 40
ܾത/ݐ ܾത/ݐ
Figure 82 Ultimate buckling force of elements with ̅=1.5 relative slenderness ratio - in the function of ⁄
1,2 1,2
= .
“A” pattern, 2 bolted, = .
"B" pattern, 1 bolted,
1 1,0
0,8 0,8
Nb,Rd,S/Nb0,rd,S
Nb,Rd,S/Nb0,Rd,S
0,6 0,6
0,4 0,4
0,2 0,2
0 0,0
0 20 40 60 80 0 20 40 60 80
Ext [%] Ext [%]
Figure 83 Ultimate buckling force of elements with ̅=1.5 relative slenderness ratio - in the function of Ext
1,2 1,2
= .
“A” pattern, 2 bolted, = .
"B" pattern, 1 bolted,
1 1,0
0,8 0,8
Nb,Rd.S/Nb0,rd,S
Nb,Rd,S/Nb0,Rd,S
0,6 0,6
0,4 0,4
0,2 0,2
0 0,0
0,2 0,7 1,2 0,2 0,7 1,2
pc pc
Figure 84 Ultimate buckling force of elements with ̅=1.5 relative slenderness ratio - in the function of pc
- 86 -
Sensitivity of the resistance on the structural parameters
In addition to numerical study using FE model 2, parametric study using FE model 1 is also carried out on corroded angle
members with different parameters. The sensitivity of the ultimate bucking force is analysed in order to find the effect of
different governing parameters on the accuracy.
In the followings the effect of the different thickness reduction on the ultimate buckling force and ultimate behaviour are
presented. In the analyses the first buckling mode is used as initial geometrical imperfection. Figure 85 presents the results
of PII-9 and PII-10 elements, the measured corroded thickness values of these elements are detailed in Table 30. It can be
concluded that using the minimum remaining thickness (tred) results in significant decrease in the resistance comparing to
using the mean value. In the case of PII-9 element the difference is 10%, as shown in Table 37. If the corroded thickness in
the corroded area changes linearly, the decrease of resistance is not significant. The ultimate behaviour mode of the elements
is not modified in the course of changing the thickness reduction in these cases.
(a) (b)
15
14
12
Ncorr [kN]
12
9
Ncorr [kN]
10
6
8 t=1.53 t=1.4 t=1.67 t=1.62 t=1.22
t=1.22 t=1.84 t=1.22 3
t=1.22-2.18 t=1.45-1.75
6 t=1.22-1.84
0
0 2 4 0 1 2 3
Imperfection [mm] Imperfection [mm]
Figure 85 Sensitivity of the resistance model - variable thickness: (a) PII-9; (b) PII-10
tred [mm] PII - 9
ID
min mean value max tred [mm] Tred [%] Ncorr /N0 [%]
PII - 9 1.22 1.53 1.84 1.22 71.2 29.3
PII - 10 1.22 1.61 2.18 1.53 63.8 39
- 87 -
imperfection, applying the second mode results in larger decrease in the ultimate buckling force. The governing buckling
mode, with regard to the ultimate buckling force, depends on the location and extension of the corrosion. Further study is
needed to determine the governing initial geometrical imperfection in general cases.
1
25
20
15
Nb,Rd,S [kN]
10 1st mode 2
2nd mode
5
3rd mode
0
0 2 4 6
Imperfection [mm]
3
• The volume of the corrosion are the same in every cases, but the thickness reduction (Tred) and the extension of
corrosion are different. Part 1 of Table 38 summarizes the applied corrosion parameters.
• The corrosion is applied in one leg of the specimen. In Table 38“SL” marks the cases if the corrosion is in the same
leg as the connection and “OL” marks the opposite.
• The magnitude of the initial geometrical imperfection (Imp) is 2 mm if the buckling mode is global and 0.2 mm if
it is local plate buckling.
• The measured yield stress from test is applied, fy=345 N/mm2.
The effect of the relative position of corrosion and the supported leg on the ultimate behaviour and ultimate buckling force
are analysed. It can be concluded that if the corrosion is on the opposite leg as the connection local ultimate failure mode is
not occurred. If the corrosion is in the same leg as the connection the local mode is observed by different thickness reduction.
Applying one bolted connection the local mode occurs beside more than 50% thickness reduction, while in the case of two
bolted connection it already occurs by 45% thickness reduction.
- 88 -
In Table 38 Ncorr/N0 marks the ratio of the ultimate buckling force of corroded and non-corroded elements and marks the
difference between “SL” and “SO” cases in percentage. The tendency of the differences is presented in Figure 87. In the first
region of the parameters (Tred = 25-45%) the difference between “SL” and “OP” is increasing but when the local behaviour
mode is occurred it is decreasing.
Comparing the results of the different end connections, in the case of “OL” pattern, the tendency of the difference is
increasing. In the case of pattern “SL” explicit tendency cannot be determined, because of the change of the behaviour mode
by increasing thickness reduction.
Table 38 Results of parametric study assuming same volume loss due to corrosion
6
0,4 3
PII - "A" 0
0,2
PII - "B" -3 25 35 45 55 65 75
0,0 -6
10 60 110
-9 Tred [%]
Tred [%]
- 89 -
method cannot follow the change in the buckling behaviour. The best approach is given by EN 50341 in all cases irrespectively
of the corrosion pattern and the connection type. In the case of pattern “A” the calculated ⁄ according to code is acceptable
for dividing the corroded elements into two groups, while in the case of pattern “B” it is less than it could be. In Figures 90-
91 it can be seen that the class border cane move. It means, that if the ⁄ is greater than the ⁄ determined by code the
ultimate behaviour is not consequential torsional-flexural buckling but flexural buckling. The same sudden decrease of the
resistances as in the case of pattern “A” is not observed in pattern “B”. The measure of the decrease is increasing with
changing the ultimate behaviour. The results and curves in Figures 88-91 belongs to ̅=1.15, the characteristics of the curves
are similar in the case of higher relative slenderness ratio as 1.5 and 1.92. In that cases the sudden decrease is observed by
higher ⁄, as it is determined in Section 3.2.4. In the following a practically applicable prediction method is developed and
detailed to determine the ratio of the ultimate buckling force of corroded and non-corroded members in all cases.
35
Pattern “A” − 1 bolt − =1.15
30
25 NUM
Nb,Rd [kN]
20 1993-3-1 Y-Y
15
50341 Y-Y
10
0
0 10 20 30 40
ܾത/ݐ
Figure 88 Ultimate buckling forces by codes and numerical study; Pattern “A”, 1 bolted connection
35
Pattern “A” − 2 bolt − =1.15
30
Nb,Rd [kN]
25 NUM
20
1993-3-1 Y-Y
15
50341 Y-Y
10
0
0 5 10 15 20 25 30 35 40
ܾത/ݐ
Figure 89 Ultimate buckling forces by codes and numerical study; Pattern “A”, 2 bolted connection
- 90 -
35
Pattern "B" − 1 bolt − =1.15
30
20 1993-3-1 Y-Y
15
50341 Y - Y
10
0
0 10 20 30 40
ܾത/ݐ
Figure 90 Ultimate buckling forces by codes and numerical study; Pattern “B”, 1 bolted connection
35
Pattern "B" − 2 bolt − =1.15
30
NUM
Nb,Rd [kN]
25
20 1993-3-1 Y-Y
15 50341 Y-Y
10
0
0 5 10 15 20 25 30 35 40
ܾത/ݐ
Figure 91 Ultimate buckling forces by codes and numerical study; Pattern “B”, 2 bolted connection
Whereas the corrosion lead to change in cross-section class therefore effective cross-section properties should be calculated.
In order to avoid the use and calculation of the effective cross-section properties the proposed method is given as a function
of cross-section or thickness reduction. As a first approximation the ratio of the ultimate buckling force of the corroded and
non-corroded angles is determined using just only one parameter, the thickness reduction. On the basis of the numerical
study the worst cases are selected. Table 39 contains the approximations for the ratio of the design buckling resistances of
corroded and non-corroded bracing members.
- 91 -
In some cases the estimated value significantly different (more than 15%) from the real value due to applying only one
parameter (Tred) in the estimation. Therefore more accurate estimation is determined applying a new parameter which is
already used in chord members. The calculated parameter Lmod /L takes account the effect of the corrosion position (pc) and
extension of corrosion (Ext ) on the ultimate buckling force. In Figure 53 the meaning of Lmod is presented and it can be
calculated by Eq. (23), see Section 2.5.
The estimation of Eq. (26) is the same as in axially loaded (chord member) case if the ultimate behaviour mode is flexural
buckling. In Appendix B Table 2 contains the value of the parameters considering the connection type and the corrosion
patterns.
The determined c and d parameters are plotted as a function of slenderness in Figure 92. The changes of the parameters are
different in the case of corrosion pattern “A” and “B”. The characteristics of the curves are similar for both types of connection.
In terms of corrosion pattern, for pattern “A” parameter d decreases and c increases, while for pattern “B” the parameter c
increases and d decreases.
- 92 -
1,05
0,95
0,9
Ncorr/N0
0,85
NUM 40%
0,8 Appx. 40%
Appx. 20%
0,75
NUM 20%
0,7
0,0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8
Lmod/L
Figure 93 Part of the results by numerical analyses and accurate approximation Eq. (26), “A” corrosion pattern, one bolted connection
and ̅=1.15
If the ultimate behaviour mode is local buckling or torsional buckling, the approximation in Table 40 can be used to estimate
the ratio of the ultimate buckling force of the corroded and non-corroded elements in the function of the thickness reduction
(Tred). Table 40 contains the equations, in the case of blank cells there are just one data therefore and it is not possible fitting
a function.
- 93 -
0,45
0,4
0,35
Ncorr/N0
0,3
Figure 94 Results of elements with ̅=1.15, “A” corrosion pattern and one bolted connection
Conclusion
On the basis of the experimental and numerical study on corroded angle members with eccentrically connected leg under
compression the following conclusions can be done:
• The ultimate behaviour mode of corroded angle members depends on the measure and the appearance of
corrosion. As in the case of chord members, three different types of ultimate behaviour modes can be distinguished:
general case of torsional-flexural buckling, local plate buckling, and torsional-flexural buckling.
• The ultimate behaviour mode depends on the corrosion pattern (“A” or “B”) and on the / regarding to the
corroded zone. The type of connection has no effect on the ultimate behaviour mode. The bound of the change in
behaviour from global to local can be described by the value of /. It separates the corroded angle members
according to the ultimate behaviour mode, depending on the slenderness, the corrosion patterns and the
connection type.
• The decrease of the ultimate buckling force is significant if the ultimate behaviour mode is plate buckling, which
belongs to “A” corrosion pattern.
• The conservative approximations of the design buckling resistance of corroded angle members are performed by
using thickness reduction. The ratio of the ultimate force of corroded and non-corroded members can be calculated
by the proposed equations detailed in Tables 39-40.
• The estimation on the basis of only thickness reduction is conservative. The used Lmod /L is an applicable parameter
to get a better estimation for ultimate buckling force. This parameter takes into consideration the corrosion position
and the extension of the corrosion, too as detailed in Eqs. (26)-(27).
• The approximation of the design buckling resistance of corroded members by codes is accurate for practical
purposes, but only in the case of global ultimate behaviour mode. In the case of local plate buckling behaviour
mode it is not applicable.
Note that the above conclusions are valid for the analysed corrosion patterns; in other cases, further analyses are necessary.
- 94 -
Evaluation of the proposed methods
Comparison of the proposed methods
On the basis of the executed analyses the following statements can be done:
• The tests on corroded angle members can be described by finite element analyses. The ultimate behaviour mode
and the ultimate buckling force of test can be followed by the numerical model.
• The ratio of the design buckling resistance of the corroded and non-corroded members cannot be estimated by only
one parameters. This statement regards to both of chord and bracing angle members.
• The corroded angle members have to be classified based on the prospective behaviour mode by / to predict the
remaining capacity.
• The Lmod /L calculated corrosion parameter is applicable together with the thickness reduction to get acceptable
accurate estimation for the ratio of the design resistances.
Furthermore, it can be concluded that in all cases, such as chord or bracing members, using the maximal cross-section
reduction or the thickness reduction a rough approximation in the safe side can be done for the ratio of the design buckling
resistance of the corroded and non-corroded elements in the case of flexural buckling ultimate behaviour. In some cases this
approximation, however can differ from the numerical value by about 15%.
In the case of more accurate estimation the characteristic of the proposed equation to determine the remaining capacity is
the same irrespectively of the type of members. The parameters (a, b, c, d) of the equations are different from each other
depending on the types of the angle members, such as chord or bracing members.
In the cases of local plate and torsional-flexural buckling (appr. 50-60-70 % cross-section reduction), the constants of the
equations (Table 29; Table 40) of the proposed conservative approximation are different. In Figure 95 the tendencies are
plotted regarding to pattern “A” corrosion with ̅=1.15 and ̅=1.5 relative slenderness ratios. On the basis of the curves it
can be concluded that the decrease of the ultimate buckling force due to corrosion is the smallest in the case of chord
members. The difference from the values of the chord members are different in the function of the relative slenderness. If the
slenderness is ̅=1.15 the difference is few percent, but if the slenderness is ̅=1.5 the difference is greater than 7%.
Even greater differences can be experienced in the case of pattern “B” corrosion (Figure 96). In the case of ̅=1.15 relative
slenderness ratio the difference is appr. 20% between the decrease of the design buckling resistance of chord members and
bracing members; the decrease is greater for bracing members. The rate is changed when the slenderness is ̅=1.5, in this
case the decrease of the ultimate buckling force is greater for chord members.
The tendencies show that the equations of approximations cannot be combined into one general equation with fix
parameters for chord and bracing members.
- 95 -
0,5 0,5
Bracing - 1 bolt
0,45 ̅ 1.15 Bracing - 1 bolt 0,45 ̅ 1.5
0,4 0,4 Bracing - 2 bolts
Bracing - 2 bolts
0,35
Ncorr/N0
0,35 Chord - centric
Ncorr/N0
0,5 0,5
Bracing - 1 bolt
0,45 0,45 Bracing - 1 bolt
Bracing - 2 bolts
0,4 0,4 Bracing - 2 bolts
Chord - centric
0,35 0,35 Chord - centric
Ncorr/N0
Ncorr/N0
0,3 0,3
0,25 0,25
0,2 0,2
0,15 ̅ 1.15 0,15 ̅ 1.5
0,1 0,1
55 60 65 70 75 55 60 65 70 75
Tred [%] Tred [%]
- 96 -
Application in practice
The reduced thickness of the elements can be measured by different way. The most common and easy types are the mortise
gauge and the thickness gauge using ultrasound (ultrasonic thickness gauge). Many types of them are available.
Yamasawa et. al [52] are investigated corroded specimens, which are clipped part of the panel point from 40-years-old steel
truss bridge. The corroded specimens are measured by laser equipment precisely and corroded thickness and curvature radius
are calculated from data measured at 1 mm intervals. The soundness of the joint contact surface are evaluated by using those
curvature radius is characteristic of the study. This kind of measurement method is a bit difficult to implement for real
structures.
- 97 -
Visual inspection Measurement Original cross-section
Length
Slenderness
Connection-type
Pattern “A” Tred
Pattern “B” Extension (Ext )
Position (pc)
/, Lmod /L
Table 27 or Table 36
Approximation
ୡ୭୰୰ ⁄ = Table 29 or Table 40
More accurate
It is recommended to replace the member.
estimation
- 98 -
4.3.3. Practical structural example
The chosen structure is found in Hungary and it is a lattice tower. The assessment of the structure is carried out by a Hungarian
company. The first step of the assessment is the visual inspection, than the cleaning is carried out with grinding and with
emery-paper. An ultrasonic thickness gauge is used to measure the reduced thickness caused by corrosion. In each location
three parallel measurements were executed.
The analysed element is a leg (chord) member, part of the member can be seen in Figure 98. The geometric characteristics of
the member are the following: length: 1500 mm, cross-section: 90×90×9 mm. The nominal yield stress is 275 N/mm2 and
the real yield stress is 327 N/mm2, the latter is derived from the result of tensile tests. On the basis of Figure 98 it can be
determined that the corrosion pattern is “A” and the extension of the corrosion is maximum 100 mm. The reduced thickness
is 6.09 mm measured by ultrasonic thickness gauge. The calculated Tred is 32%. The slenderness is 85 and the relative
slenderness is 1.07 of the non-corroded member. The calculated / is 12.78 and the bound of Class 4 is 12.71 according to
EC3-3-1 using the measured fy. On the basis of / and applying Table 27 it can be determined that the prospective ultimate
behaviour mode is general case of torsional-flexural buckling. The approximate estimation of the ratio of the corroded and
non-corroded member is 0.84. It means 16% decrease of ultimate buckling force. It is, however, a rough approximation (see
Section 2). On the basis of the numerical results in Section 2, 20-30% thickness reduction, pattern “A” and less than 20%
extension the decrease of the ultimate buckling force is 1-2%.
To present a more considerable decrease of the ultimate buckling force, the extension of corrosion is modified to 40%, but
the other parameters were the same. The parameters are calculated from Table 1 in Appendix B by interpolation and
supposing 100×100×12 cross-section. The result calculated in this way is Ncorr /N0 = 0.934. The result was checked by
numerical simulation, too, and it resulted in N0 = 228.03 kN, Ncorr = 211.12 kN and Ncorr /N0 = 0.926 (during the calculation
L/200 and fy = 327 N/mm2 were supposed). Figure 99 presents the step of the checking in this example.
Using the code (EN 50341) to predict the ultimate buckling forces, the result is 0.824 for the ratio of the corroded and non-
corroded ultimate buckling forces. Comparing the results of the code and the proposed estimation the difference is 10%.
Table 27 or Table 36
Approximation
It is recommended to replace the member.
More accurate
estimation
- 100 -
Summary and conclusions
The theses of the PhD dissertation in English
1. I worked out and completed a novel experimental test program on 46 chord and bracing angle members of trussed
towers under compression, having various corrosion damages.
a. I developed equivalent corrosion test models for different types of corrosion. I designed the specimens
in which the corrosion damage is created artificially, especially by milling process. I determined the
typical corrosion patterns and parameters on the basis of practical experiences.
b. With technical assistance, I carried out experiments on the corroded angle chord and bracing members.
As a result I determined the effect of the corrosion parameters (thickness reduction, corrosion position,
extension of corrosion) on the buckling behaviour and on the ultimate buckling force.
c. I developed and verified applicable finite element models, which follow the experimental behaviour of
chord and bracing angle members. Comparing the experimental and computed behaviour modes and
ultimate buckling forces I found that they are in good agreement.
2. I completed parametric study #1 on the corroded angle members under axial compression, namely on chord
members of trussed towers by the verified finite element model. The applied corrosion parameters are as follows:
(i) thickness reduction (Tred), (ii) position of corrosion along the length (pc), (iii) extension of corrosion (Ext ) and (iv)
position of corrosion on the legs (corrosion on only one and on both legs of the angle). In the study three different
cross-sections and three relative slenderness ratios are applied. On the basis of the results:
a. I determined zones I. - III. to classify the corroded angle bracing members according to the prospective
ultimate behaviour mode. I determined the values of width-to-corroded thickness / as limits of zones.
If the corroded member is classified into zones I. or III., the ultimate failure mode is clearly defined; in
case of zone II., due to the complexity of the problem investigated, the behaviour mode cannot be clearly
determined. I recommend to use the lower value of the zone.
b. I identified the ultimate behaviour modes, determined the design buckling resistances and defined the
tendencies of the reduction of the resistance comparing to the non-corroded member in terms of the
studied parameters.
3. I completed parametric study #2 on eccentrically connected corroded angle members under compression, namely
on corroded bracing members of trussed towers, by the verified finite element model. The applied corrosion
parameters are as follows: (i) thickness reduction (Tred), (ii) position of corrosion along the length (pc), (iii) extension
of corrosion (Ext ) and (iv) position of corrosion on the legs (corrosion on only one and on both legs of the angle). In
- 101 -
the study three different relative slenderness ratios and two types of end connections are applied. On the bases of
the results:
a. I determined zones I. - III. to classify the corroded angle bracing members according to the prospective
ultimate behaviour mode. I determined the values of width-to-corroded thickness / as limits of zones.
If the corroded member is classified into zones I. or III., the ultimate failure mode is clearly defined; in
case of zone II., due to the complexity of the problem investigated, the behaviour mode cannot be clearly
determined. I recommend to use the lower value of the zone.
b. I identified the ultimate behaviour modes, determined the design buckling resistances, and defined the
tendencies of the reduction of the resistance comparing to the non-corroded member in terms of the
studied parameters.
c. I determined the effect of the number of bolts (one or two) in the end connection on the buckling
resistance of corroded bracing angle members.
Publication links to this thesis: [8]
4. I developed design methods for the resistance calculation of corroded angle members and I showed the practical
application of them via real structural example. I analysed the applicability of different design codes for corroded
chord and bracing angle members.
a. I developed a simplified and a more accurate estimation for the ratio of the design buckling resistances
of corroded and non-corroded members. I determined the constants of the design equations in the range
of the studied parameters.
b. I proposed the codes which are applicable to approximate the design buckling resistance of corroded
chord and bracing members by conservative way. I pointed out, however, that the recommendation of
the standard for the trussed towers in the case of single angle bracing members having only one bolt at
the ends seems to be over-conservative.
- 102 -
The theses of the PhD dissertation in Hungarian
1. Megterveztem és végrehajtottam egy 46 nyomott próbatestből álló újszerű kísérleti programot rácsos
torony/oszlop szerkezetek öv-, illetve rácsrúd elemeinek használt egyenlőszárú szögacélokon, különböző korróziós
károsodásokat feltételezve.
a. Egyenértékű korróziós kísérleti modelleket dolgoztam ki különböző típusú korróziókra. Megterveztem a
próbatesteket, melyeken a korróziós károsodást mesterségesen, marási folyamattal hozták létre.
Gyakorlati tapasztalatok alapján meghatároztam a tipikus korróziós mintákat és a korróziós
paramétereket.
b. Technikai segítséggel végrehajtottam a kísérleteket korrodált öv-, illetve rácsrúd szögacélokon.
Eredményként meghatároztam a korróziós paraméterek (úgymint (i) a korrózió hosszirányú
elhelyezkedése, (ii) a korrózió kiterjedése és (iii) a vastagság-csökkenés) hatását a tönkremeneteli módra
és a kihajlási határerőre.
c. Végeselemes modelleket fejlesztettem és verifikáltam, amelyek megfelelően követik az öv- és rácsrúd
szögacélok kísérleti viselkedését. Összehasonlítva a kísérleti és a számított viselkedéseket, valamint a
kihajlási határerőket, jó egyezéseket kaptam.
2. A verifikált numerikus modellt alkalmazva elvégeztem egy #1 jelű paraméteres vizsgálatsorozatot, korrodált,
központosan nyomott szögacélokon (rácsos toronyszerkezetek korrodált övrúdjain). Az alkalmazott paraméterek a
következők voltak: (i) vastagság-csökkenés (Tred), (ii) korrózió hosszirányú elhelyezkedése (pc), (iii) korrózió
kiterjedése (Ext ) és (iv) korrózió szögacél szárain való elhelyezkedése (korrózió egyik, illetve mindkét szögacél
száron). A vizsgálat során három különböző keresztmetszeti méretet és három relatív karcsúságot alkalmaztam. Az
eredmények alapján:
a. Meghatároztam a lemezelem szélesség-korrodált vastagság / paraméter értékeit a korrodált öv
szögacélok tönkremeneteli mód alapján történő csoportosításához ezáltal három zónába (I. - III.)
soroltam a korrodált szögacélokat. Az I. és III. zónában, a tönkremeneteli mód egyértelműen
meghatározott, a II. zóna esetében a probléma komplexitása miatt e paraméter alapján a tönkremeneteli
mód nem határozható meg. Ezek alapján az zóna alsó határértékének használatát javaslom.
b. Azonosítottam a tönkremeneteli módokat és meghatároztam a tervezési ellenállásokat, valamint a
teherbírás csökkenés mértékét a különböző vizsgált paraméterek függvényében.
3. A verifikált numerikus modellt alkalmazva elvégeztem egy #2 jelű paraméteres vizsgálatsorozatot, korrodált,
nyomott szögacélokon (rácsos toronyszerkezetek korrodált rácsrúdjain). Az alkalmazott paraméterek a következők
voltak: (i) vastagság-csökkenés (Tred), (ii) korrózió hosszirányú elhelyezkedése (pc), (iii) korrózió kiterjedése (Ext ) és
- 103 -
(iv) korrózió szögacél szárain való elhelyezkedése (korrózió egyik, illetve mindkét szögacél száron). A vizsgálat
során három relatív karcsúságot és két végkapcsolati típust (egy, illetve két csavaros) alkalmaztam. Az eredmények
alapján:
a. Meghatároztam a lemezelem szélesség-korrodált vastagság / paraméter értékeit a korrodált öv
szögacélok tönkremeneteli mód alapján történő csoportosításához ezáltal három zónába (I. - III.)
soroltam a korrodált szögacélokat. Az I. és III. zónában, a tönkremeneteli mód egyértelműen
meghatározott, a II. zóna esetében a probléma komplexitása miatt e paraméter alapján a tönkremeneteli
mód nem határozható meg. Ezek alapján az zóna alsó határértékének használatát javaslom.
b. Azonosítottam a tönkremeneteli módokat és meghatároztam a tervezési ellenállásokat és a teherbírás
csökkenés mértékét a különböző vizsgált paraméterek függvényében.
c. Meghatároztam a végkapcsolatban alkalmazott csavarok számának (egy vagy kettő) hatását a korrodált
rácsrúd szögacélok teherbírására vonatkozóan.
4. Kidolgoztam egy méretezési eljárást a korrodált nyomott szögacélok vizsgálatára vonatkozóan és megmutattam
az eljárás alkalmazhatóságát egy valós szerkezeti példán keresztül. Megvizsgáltam a különböző szabványok
használhatóságát rácsos toronyszerkezetek korrodált öv- és rácsrúd szögacéljainak vizsgálatára vonatkozóan.
a. Meghatároztam egy közelítő és egy pontosabb becslést a korrodált és a nem korrodált nyomott
szögacélok maximális teherbírás hányadosára vonatkozóan. Meghatároztam az egyenletekben szereplő
konstansokat a vizsgált paraméterek tartományában.
b. Meghatároztam, hogy mely szabványok alkalmasak korrodált öv- és rácsrúd elemek teherbírásának
konzervatív módon történő közelítő meghatározásához. Mindemellett rámutattam, hogy a rácsos
toronyszerkezetek egy csavarral bekötött rácsrúdjainak tervezésére vonatkozó szabványos ajánlás túl
konzervatív.
- 104 -
Publications on the subject of the thesis
1. Oszvald, K, Dunai, L, “Effect of corrosion on the buckling of steel angle elements”, Proceedings of the 8th fib PhD
Symposium in Civil Engineering, June 20-23, 2010,Lyngby, Denmark, pp. 549–554, 2010.
2. Oszvald K, Dunai L, “Korrózióval gyengített acélrudak stabilitás vizsgálata - Kísérlet, VEM analízis”, XI. Magyar
Mechanikai Konferencia, 2011. augusztus 29-31. Miskolc-Egyetemváros, Paper 144, 6 oldal, 2011.
3. Oszvald, K, Dunai, L, “Design buckling resistance of corroded members: Determining the reduction factor for angle
section elements”, Proceedings of EUROSTEEL 2011 - 6th European Conference on Steel and Composite Structures, August
31 - September 2, 2011, Budapest, Hungary, vol. A, pp. 627-632, 2011.
4. Oszvald, K, Dunai, L, “Finite element analysis of corroded steel angles under compression, Proceedings of the Conference
of Junior Researchers in Civil Engineering, June 19-20, 2012, BME, Budapest, Hungary, pp. 163-171, 2012.
5. Oszvald, K, Dunai, L, “Effect of corrosion on the buckling of steel angle members - experimental study”, Periodica
Polytechnica - Civil Engineering, vol. 56, no. 2, pp. 175-183, 2012.
6. Oszvald, K, Dunai, L, “Behaviour of corroded steel angle compression members - numerical study”, Periodica
Polytechnica - Civil Engineering, vol. 57, no 1. pp. 63-75, 2013.
7. Oszvald, K, Dunai, L, “Experiments of corroded angle-section compressive members with bolted connections”, Design
Fabrication and Economy of Metal Structures International Conference Proceedings, April 24-26, 2012, Miskolc,
Hungary, pp. 216-222, 2013.
8. Oszvald, K, Dunai, L, “Analysis 1of corroded steel angles under eccentric loading”, Proceedings of the Second Conference
of Junior Researchers in Civil Engineering, June 17-18, 2013, BME, Budapest, Hungary, pp. 148-154, 2013.
- 105 -
References
[1] EN 1993-1-1:2005, Eurocode 3: Design of steel structures. Part 1-1: General rules and rules for buildings.
[2] ANSI/AISC 360-05, Specification for structural steel building, 2005.
[3] Balázs, Gy, “Építőanyagok és kémia”, Műegyetemi Kiadó, Budapest, 1983.
[4] Kovács, K, “A korrózió megjelenési formái és okai”, Tankönyvkiadó, Budapest, 1968.
[5] Richardson, J A T, “Shreir’s Corrosion”, Elsevier Science, ISBN 9780444527882, 2009.
[6] http://www.nace.org/
[7] Szunyogh, L, “Hegesztés és rokon technológiák, Kézikönyv” Gépipari Tudományos Egyesület, Műegyetemi Kiadó,
Budapest, 2007.
[8] Szabó, J, Simon, S, “Korróziós károsodások kezelése az MVM Rt hálózati tartószerkezetein”, A Magyar
Villamosművek Közleményei, vol. 2000/3, pp. 47-49, 2000.
[9] Reiner, F, “Lattice tower corrosion inspection with UAV”, CIGRE 2012, B2 - 303, http://www.cigre.org
[10] Rahgozar, R, “Remaining capacity assessment of corrosion damaged beams using minimum curves”, Journal of
Constructional Steel Research, vol. 65, pp. 299-307, 2009.
[11] Sharifi, Y, Rahgozar, R, “Remaining moment capacity of corroded steel beams”, International Journal of Steel
Structures, vol. 10, pp. 165-176, 2010.
[12] Fukuda, M, Fujii, K, Nakayama, T, Matsui, S, “An evaluation method for the remaining strength of a plate girder
with local corrosion under sleepers”, Procedia Engineering, vol. 14., pp. 2285-2293, 2011.
[13] Kim, I T, Lee, M J, Ahn, J H, Kainuma, S, “Experimental evaluation of shear buckling behaviours and strength of
locally corroded web”, Journal of Constructional Steel Research, vol. 83, pp. 75-89, 2013.
[14] Ahn, J H, Kim, I T, Kainuma, S, Lee, M J, “Residual shear strength of steel plate girder due to web corrosion”,
Journal of Constructional Steel Research, vol. 89, pp. 198-212, 2013.
[15] Ahn, J H, Kainuma, S, Yasuo, F, Takehiro, I, “Repair method and residual bearing strength evaluation of a locally
corroded plate girder at support”, Engineering Failure Analysis, vol. 33, pp. 398-418, 2013.
[16] Ok, D, Pu, Y, Incecik, A, “Computation of ultimate strength of locally corroded unstiffened plates under uniaxial
compression”, Marine Structures, vol. 20, pp. 100-114, 2007.
[17] Khedmati, M R, Roshanali, M M, Nouri, Z H M E, “Strength of steel plates with both-sides randomly distributed
with corrosion wastage under uniaxial compression”, Thin-Walled Structures, vol. 49, pp. 325-342, 2011.
[18] Silva, J E, Garbatov, Y, Guedes Soares, C, “Ultimate strength assessment of rectangular steel plates subjected to a
random localised corrosion degradation”, Engineering Structures, vol. 52, pp. 295-305, 2013.
[19] Teixeira, A P, Ivanov, L D, Guedes Soares, C, “Assessment of characteristic values of ultimate strength of corroded
steel plates with initial imperfections”, Engineering Structures, vol. 56, pp. 517-527, 2013.
[20] Nakai, T, Matsushita, H, Yamamoto, N, Arai, H, “Effect of pitting corrosion on local strength of hold frames of bulk
carriers (1st report)”, Marine Structures, vol. 17, pp. 403-432, 2004.
- 106 -
[21] Nakai, T, Matsushita, H, Yamamoto, N, “Effect of pitting corrosion on local strength of hol frames of bulk carriers
(2nd Report) - Lateral-distortional buckling and local face buckling”, Marine Structures, vol. 17, pp. 612-641,
2006.
[22] Nakai, T, Matsushita, H, Yamamoto, N, “Effect of pitting corrosion on strength of web plates subjected to patch
loading”, Thin-Walled Structures, vol. 44, pp. 10-19, 2006.
[23] Jiang, X, Guerdes Soares, C, “A closed form formula to predict the ultimate capacity of pitted mild steel plate under
biaxial compression”, Thin-Walled Structures, vol. 59, pp. 27-34, 2012.
[24] Appuhamy, J M R S, Kaita, T, Ohga, M, Fujii, K, “Analytical study on significance of corroded surface measurement
on residual strength prediction”, Procedia Engineering, vol. 14, pp. 2260-2268, 2011.
[25] Kaita, T, Appuhamy, J M R S, Itogawa, K, Ohga, M, Fujii, K, “Experimental study on remaining strength estimation
of corroded wide steel plates under tensile force”, Procedia Engineering, vol. 14, pp. 2707-2713, 2011.
[26] Dunbar, T E, Pegg, N, Taheri, F, Jiang, L, “A computational investigation of the effects of localized corrosion on
plates and stiffened panels”, Marine Structures, vol. 10, pp. 385-402, 2004.
[27] Jeong, Y S, Kainuma, S, Ahn, J H, “Structural response of orthotropic bridge deck depending on the corroded deck
surface”, Construction and Building Materials, vol. 43, pp. 87-97, 2013.
[28] Saad-Eldeen, S, Garbatov, Y, Guedes Soares, C, “Analysis of plate deflections during ultimate strength
experiments of corroded box girders”, Thin-Walled Structures, vol. 54, pp. 164-176, 2012.
[29] Saad-Eldeen, S, Garbatov, Y, Guedes Soares, C, “Ultimate strength assessment of corroded box girders”, Ocean
Engineering, vol. 58, pp. 35-47, 2013.
[30] Sadovsky, Z, Drdacky, M, “Buckling of plate strip subjected to localised corrosion - a stochastic model”, Thin-
Walled Structures, vol. 39, pp. 247-259, 2001.
[31] Beaulieu, L V, Legeron, F, Langlois, S, “Compression strength of corroded steel angle members”, Journal of
Constructional Steel Research, vol. 66, pp. 1366-1373, 2010.
[32] Hisazumi, K, Kanno, R, Tominaga, T, Imafaku, K, “Axial compressive strength of severely corroded channel and
angle members used in truss structures”, 7th European Conference on Steel and Composite Structures,
EUROSTEEL’2014, Naples, Italy, vol. A, pp. 393-398, 2014.
[33] MSZ EN ISO 16701:2008 Corrosion of metals and alloys. Corrosion in artificial atmosphere. Accelerated corrosion
test involving exposure under controlled conditions of humidity cycling and intermittent spraying of a salt
solution.
[34] MSZ EN ISO 11130:2011 Corrosion of metals and alloys. Alternate immersion test in salt solution.
[35] MSZ EN 1993-3-1:2007 - Eurocode 3: Design of steel structures. Part 3-1: Towers, masts and chimneys. Towers
and masts.
[36] MSZ EN 50341-1:2013 - Overhead electrical lines exceeding AC 1 kV. Part 1: General requirements. Common
specifications.
[37] MSZ EN 1993-1-5:2006 Eurocode 3: Design of steel structures. Part 1-3: Plated structural elements.
- 107 -
[38] Timoshenko, S P, Gere, J M, “Theory of elastic stability”, Chapter 5, McGraw-Hill International Book Company,
1963.
[39] Petersen, C, “Statik und Stabilitat der Baukonstruktionen” Vieweg, Wiesbaden, 1982.
[40] MSZ EN 1993-1-3:2006 - Eurocode 3: Design of steel structures. Part 1-3: General rules, Supplementary rules for
cold-formed members and sheeting.
[41] ANSYS® v11.0, Canonsburg, Pennsylvania, USA.
[42] Liu, Y, Hui, L, “Finite element study of steel single angle beam-columns”, Engineering Structures, vol. 32, pp.
2087-2095, 2010.
[43] MSZ EN 10056-2:1994: Egyenlő és egyenlőtlen szárú szögacél. 2. rész: Alak- és mérettűrések.
[44] GBTUL http://www.civil.ist.utl.pt/gbt/
[45] Adany, S, Schafer, B, “Buckling mode decomposition of single-branched open cross-section members via finite
strip method: Derivation”, Thin-Walled Structures, vol. 44, pp. 563-584, 2006.
[46] Dinis, P B, Camotim, D, Silvestre, N, “On the local and global buckling behaviour of angle, T-section and cruciform
thin-walled members”, Thin-Walled Structures, vol. 48, pp. 786-797, 2010.
[47] Dinis, P B, Camotim, D, Silvestre, N, “On the mechanics of thin-walled angle column instability”, Thin-Walled
Structures, vol. 52, pp. 80-89, 2012.
[48] Mesacasa, E Jr, Dinis, P B, Camotim, D, Malite, M, “Mode interaction in thin-walled equal-leg angle columns”,
Thin-Walled Structures, vol. 81, pp. 138-149, 2014.
[49] MSZ 4388-1:1994 Corrosion ratio determination. Part 1: Corrosion stability evaluation in the case of uniform
surface corrosion.
[50] MSZ 4388-2:1994 Corrosion ratio determination. Part 2: Corrosion stability evaluation in the case of non-uniform
surface corrosion.
[51] MSZ EN ISO 11463:2008 Corrosion of metals and alloys. Evaluation of pitting corrosion.
[52] Yamasawa, T, Nogami, K, Koyama, T, “An evaluation of soundness using corrosion curvature”, 6th European
Conference on Steel and Composite Structures, EUROSTEEL’2011, Budapest, Hungary, vol. A, pp. 393-398, 2011.
- 108 -
Appendix A - Corroded lattice towers - Examples
A-1
A-2
A-3
A-4
Appendix B – Constants of Eq. (24)-(26)
B-1
Pattern ̅ 1.15 1.5 1.92 Mean value
a 1.015 1.014 1.013 1.014
b -0.4826 -0.4697 -0.4802 -0.478
1 bolted
c -0.807 -0.6116 -0.519
d 1.737 1.501 1.479
“A”
a 1.014 1.012 1.012 1.013
b -0.4596 -0.457 -0.4901 -0.469
2 bolted
c -0.8395 -0.5812 -0.4305
d 1.751 1.445 1.352
B-2