1

Second order effects

Background to chapters 5.8, 5.9 and Annex H in EN 1992-1-1

Bo Westerberg, May 2002

 2

Foreword
This report gives background information about chapter 5.8 and Annex H of EN 1992-1-1.
Particular attention is paid to areas where chapter 5.8 differs much from chapter 4.3.5 in ENV
1992-1-1, and areas that have been much discussed during the conversion from ENV to EN.

Chapters in this report are numbered according to the third level of headings in chapter 5.8;
thus for example chapter 7 in the report deals with 5.8.7 in the EN etc. All references to
clauses, figures and expressions starting with 5.8 concern chapter 5.8 in EN 1992-1-1,
and references starting with H refer to Annex H in the EN. Other references concern
clauses in this report.

During the development of chapter 5.8 and Annex H, there have been extensive discussions
within TC250/SC2 and the Project Team for EN 1992-1-1, but also with persons not directly
involved in the CEN groups. These external discussions have forced me to go deeper into
some areas in order to justify, and sometimes modify, the rules. This has also given valuable
material for this background document.. I want to mention in particular Mr Alasdair Beal
(UK), Prof Jostein Hellesland (Norway) and Prof Ulrich Quast (Germany) for their valuable
comments and discussions.

Stockholm, May 2002

Bo Westerberg

Member of the Project Team for EN 1992-1-1

Note. In the final draft 5.8.10 has become 5.9; this section is also covered here.
 3

Contents
1. Definitions (5.8.1) 4
2. Basic criterion for neglecting second order effects (5.8.2 (6)) 5
3. Simplified criteria for neglecting 2nd order effects (5.8.3) 6
3.1 Slenderness limit for isolated members (5.8.3.1) 6
3.1.1 General 6
3.1.2 History of the slenderness limit in prEN 1992 8
3.1.3 Background to new proposal 9
3.1.4 Comparisons with previous model for slenderness limit 10
3.1.5 Comparisons with new model 11
3.2 Effective length 12
3.3 Global second order effects in structures (5.8.3.3) 16
3.3.1 Background 16
3.3.2 No significant shear deformations, rigid moment restraint 17
3.3.3 Effect of flexible moment restraint 19
3.3.4 Effect of global shear deformations 20
4. Effective creep ratio (5.8.4) 22
4.1 General 22
4.2 Effect of creep in cross sections 23
4.2.1 Uncracked unreinforced cross section 23
4.2.2 Uncracked reinforced cross section 23
4.2.3 Cracked reinforced cross section 25
4.2.4 Conclusions concerning cross sections 26
4.3 Effect of creep in slender columns 27
4.3.1 General 27
4.3.2 Comparison between one- and two-step calculations 28
4.3.3 The effect of creep on the slenderness limit 30
4.3.4 Safety under long-term load only 31
5. Methods of analysis (5.8.5) 33
6. General method (5.8.6) 34
6.1 General 34
6.2 Safety format 35
6.3 Interaction diagrams 36
6.4 The effect of creep 36
6.5 Simplified methods and their common basis 38
7. Method based on stiffness (5.8.7) 40
7.1 Basic equations 40
7.2 Moment distribution 40
7.3 Estimation of stiffness 42
7.4 Linear analysis of structures 43
8. Method based on curvature (5.8.8) 47
8.1 Basic relationships 47
8.2 Comparison with general method and stiffness method 48
8.3 Using the curvature method for structures 48
9. Biaxial bending (5.8.9) 49
10. Lateral instability of slender beams (5.9 ) 50
11. References 51
Appendix 1. Verification of new model for slenderness limit 52
Appendix 2. Calibration of simplified methods 55
A2.1 Main calculations and results 55
A2.2 Discussion 58
 4

1. Definitions (5.8.1)
Definitions specific to chapter 5.8 are listed in 5.8.1. Some comments are given below.

Braced bracing
The distinction braced bracing is simple: units or systems that are assumed to contribute to
the stabilization of the structure are bracing, the others are braced. Bracing units/systems
should be designed so that they, together, have the necessary stiffness and resistance to stabi-
lization forces. The braced ones, by definition, do not need to resist such forces.

Buckling
The word buckling has been reserved for the pure, hypothetical buckling of an initially
straight member or structure, without load eccentricities or transverse loading. It is pointed
out in a note that pure buckling is not a relevant limit state in real structures, due to the pres-
ence of imperfections, eccentricities and/or transverse loads. This is also a reason why the
word buckling is avoided in the title of 5.8. In the text, buckling is mentioned only when a
nominal buckling load is used as a parameter in certain calculation methods.

First order effects

First order effects are defined to include the effect of imperfections. Thus, imperfections are
interpreted as physical deviations in the form of inclinations or eccentricities. The ENV is
ambiguous in this respect; see also the background to document to 5.2.

Nominal second order moment

The nominal second order moment is used in certain simplified methods, to obtain a total
moment for design of cross sections to their ultimate moment resistance. It can be defined as
the difference between the ultimate moment resistance and the first order moment, see 6.3 in
this report. If the ultimate load is governed by instability before reaching the cross section
resistance, then the nominal second order moment is greater than the true one; this is the rea-
son for using the word nominal.

Sway non-sway and global second order effects

The terms sway non-sway have been omitted in the final draft, after many comments for or
against. The words in themselves are misleading, since all structures are more or less sway;
a structure that would be classified as sway could be just as stiff as one classified as non-
sway. These terms are now replaced by unbraced braced.

In the ENV the concept of sway non-sway was linked to the criterion for neglecting global
second order effects in structures. The classification of structures from this point of view re-
mains in the EN, but without using the sway / non-sway terminology.

A stiffness criterion like that in ENV-A.3.2 was avoided in earlier drafts of the EN, since it
was considered as too rough, and in some cases misleading. However, during the conversion
process there were many requests to include some simple criterion for evaluating the signifi-
cance of global second order effects, without the need for calculating them. This lead to the
present rules in 5.8.3.3 and Informative Annex H, which are more general than ENV-A.3.2.
The details are given in clause 3.3 of this report.
 5

2. Basic criterion for neglecting second order effects (5.8.2 (6))

Two basic criteria for neglecting second order effects have been discussed during the conver-
sion process, namely:

1. 10 % increase of the corresponding first order effect,

2. 10 % reduction of the load capacity, assuming a constant eccentricity of the axial force.

The first criterion is the one stated in 5.8.2 (6), and in the ENV, 4.3.5.1 (5). The second one
has been claimed by some to be the true, hidden criterion behind the ENV-rules.

Figure 2-1 illustrates the consequences of these two criteria in an interaction diagram for axial
force and bending moment. Their effects on the slenderness limit are discussed in chapter 3.

In a column or a structure it is the bending moment that is influenced by second order effects.
The axial force is governed by vertical loads, and is not significantly affected by second order
effects. Most design methods are based on calculating a bending moment, including a second
order moment if it is significant. From this point of view, criterion 1 is the most logical and
natural one.

The basic criterion is further discussed in chapter 3 in connection with slenderness limits.

0,5
M MNm
0,45

2
0,4
Nu; Mu 1 10% reduction of N and M
0,35
2 10% increase of M
0,3
1
0,25
Nu; Mu
0,2 1. Nu; Mu/1,1
2. 0,9Nu; 0,9Mu
0,15

0,1

0,05
N MN
0
0 0,5 1 1,5 2 2,5 3 3,5 4 4,5 5 5,5

Figure 2-1. Two different ways of defining the basic 10%-criterion for neglecting second or-
der effects, see text above. (The cross section resistance was calculated for rectangular cross
section 400 x 600 mm, concrete C35, = 0,1 (total mechanical reinforcement ratio), edge
distance of reinforcement 60 mm.)
 6

3. Simplified criteria for neglecting 2nd order effects (5.8.3)

3.1 Slenderness limit for isolated members (5.8.3.1)
3.1.1 General
The load bearing capacity of a compression member for low slenderness ratios is illustrated in
figure 3-1 by means of interaction curves, calculated according to the general method in 5.8.6.
(See chapter 6 in this report for more details about interaction curves and the general method.)

0,5
M0 kNm
0,45
=0
0,4

0,35

Figure 3-1. Interaction 0,3

curves for low slender- = 10
ness ratios, calculated 0,25 21
for a pin-ended column 31
0,2 42
with cross section as
for figure 2-1 and sub- 0,15
jected to a constant
first order moment. The 0,1

effective creep ratio ef 0,05

= 0. Equal first order N kN
end moments M01 = 0
0 0,5 1 1,5 2 2,5 3 3,5 4 4,5 5 5,5
M02 = M0.

By combining figures 3-1 and 2-1, one can find the slenderness ratios for which the basic
10%-criterion is fulfilled; see figure 3-2.
lim
60

50
Figure 3-2. Effect of normal force
(here n = N/fcdAc) on slenderness 2
limit, depending on the interpretation 40
of the 10 % criterion:
30
1. 10 % increase of bending mo- 1
ments for a given normal force 20

2. 10 % reduction of load capac-

10
ity for a given eccentricity
n
( = 0,3, ef = 0, M01/M02 = 1). 0
0 0,2 0,4 0,6 0,8 1 1,2 1,4
 7

Depending on which of the two basic criteria is chosen, see chapter 2, increasing axial force
will either decrease (1) or increase (2) the slenderness limit as shown in figure 3-2. Criterion
1 will be more severe for high axial loads, when there is little room for bending moments.
Criterion 2, on the other hand, will allow very high slenderness ratios for high axial loads.

In earlier drafts, including the final draft of October 2001, a slenderness limit independent
of the axial force was chosen as a compromise between the two basic criteria; for = 0,1 the
criterion was then identical to expression (4.62) in the ENV.

In a comment to the final draft it was pointed out that a limit independent of the axial force
could be much on the unsafe side in certain cases (see 3.1.2). Therefore, a new model was
developed. The following is quoted from 5.8.3.1:

(1) As an alternative to 5.8.2 (6), second order effects may be ignored if the slenderness is
below a certain value lim. The following may be used:

lim = 20ABC (5.13)

where:
slenderness ratio as defined in 5.8.3.2

A = 1 / (1+0,2ef)

B = (1 + 2 ) / n
C = 1,7 - rm
ef effective creep ratio; see 5.8.4;
if ef is not known, A = 0,7 may be used
= Asfyd / (Acfcd); mechanical reinforcement ratio;
if is not known, B = 1,2 / n may be used
As total area of longitudinal reinforcement
n = NEd / (Acfcd); relative normal force
rm = M01/M02; moment ratio
M01, M02 first order end moments, M02 M01

(2) If the end moments M01 and M02 give tension on the same side, rm should be taken positive
(i.e. C < 1,7), otherwise negative (i.e. C > 1,7).

In the following cases, rm should be taken as 1,0 (i.e. C = 0,7):

- for braced members with first order moments only or predominantly due to imperfections
or transverse loading
- for unbraced members in general

The background to the new criterion is presented in the following.

 8

3.1.2 History of the slenderness limit in prEN 1992

At an early stage of the conversion process (spring 1999), a slenderness limit was proposed in
which the effective creep ratio ef and the relative normal force n were included as parame-
ters. The reinforcement ratio was not included then, since it was considered unpractical.

Most people also considered it unpractical and unnecessary to include creep; there was a
widespread opinion that creep would have little effect at these low slenderness ratios.

There was also much discussion about the interpretation of the basic 10 % criterion for
neglecting second order effects, with the two main alternatives (cf chapter 2):

1. 10 % increase of bending moments due to second order effects

2. 10 % reduction of the load capacity for a given eccentricity

It was then demonstrated that the effect of the normal force on the slenderness limit was dif-
ferent depending on which alternative was used, see figure 3-2.

However, there was no agreement as to which alternative to base the slenderness limit on, and
therefore the ENV criterion (4.62), independent of the normal force, was used in draft 1.

An addition was made in draft 2, allowing the constant 25 to be increased to 35 if the rein-
forcement ratio is at least 0,5. In the final draft October 2001 an interpolation was intro-
duced to avoid discontinuity (expression (5.13)).

In November 2001, shortly after this draft had been distributed, comments and examples were
presented by Prof. J. Hellesland, showing that (5.13) could be severely unconservative in
some cases, e.g. a column bent in double curvature (end moments of different directions, fig-
ure 3-3), combineed with a high effect of creep and a moderate or high normal force.

Comments on earlier versions of prEN 1992-1-1, together with a general treatment of the
slenderness limit, are given in [8].

M02

Figure 3-3. Column bent in

double curvature. M01
 9

3.1.3 Background to new proposal

After receiving the above-mentioned comments and examples from Prof Hellesland, a sys-
tematic investigation of the slenderness limit was made, with focus on the effects of rein-
forcement, normal force, creep and moment ratio (different end moments).

Figure 3-4 shows examples of the effect of a rather moderate effective creep ratio, ef = 1.
Curves according to both basic 10 % criteria are shown. Table 3-1 shows some values read
from the figures.

lim lim
200 200
180 C40 = 0,3 ef = 0 180 C40 = 0,3 ef = 1
160 160
140 140
-0,9 -0,9
120 120
100 100
0 0
80 80
60 60
M 01/M 02 = 1 M 01/M 02 = 1
40 40
20 20
n n
0 0
0 0,5 1 1,5 0 0,5 1 1,5

Figure 3-4. Example of the slenderness limit as a function of the relative normal force and the
moment distribution. Concrete C40, = 0,3. (Solid lines = 10%-criterion alt. 1, dashed = alt. 2.)

Table 3-1. Values of slenderness limit taken from figure 3-4, alternative 1 ( = 0,3).
lim Average reduction due to
n M01/ M02
ef = 0 ef = 1 creep, %
1,0 30 25
0,5 0 70 60 16
-0,9 110 90
1,0 20 10
1,0 0 45 30 37
-0,9 70 50

For n = 1,0, the average reduction due to creep is considerable, having in mind that ef = 1
represents a rather moderate effect of creep. With a higher value of ef the reduction is more
severe (most of the comparisons were made with ef = 0 and 2 respectively, see Appendix 1).
 10

lim lim
200 200

180
C40 ef = 2 = 0,1 180
C40 ef = 2 = 0,5

160 160
-0,9
140 140
120 120
-0,9
100 100
0
80 80
0
60 60
M 01/M 02 = 1
40 40
M 01/M 02 = 1
20 20
0 0
0,0 0,4 0,8 n 1,2 0 0,4 0,8 n 1,2

Figure 3-5. Effect of reinforcement ratio on slenderness limit (only criterion 1is shown here).

Table 3-2. Values of slenderness limit taken from figure 3-5. (ef = 2)
lim Average increase from
n M01/ M02
= 0,1 = 0,5 = 0,1 to = 0,5, %
1,0 17 27
0,4 0 47 70 75
-0,9 74 107
1,0 11 20
0,8 0 27 43 56
-0,9 41 66

The effect of is considerable, as can be seen from both figure 3-5 and table 3-2. Without
as a parameter, the slenderness limit would have to be either very conservative for high to
moderate values of , or on the unsafe side for low values. However, since the reinforcement
is normally not known when the slenderness criterion is checked, a default value based on a
low value of has also been given. This can be used for a conservative estimation, or as a
starting value in an iterative process.

There have been national comments proposing to include an effect of n in the slenderness
limit. These proposals were rejected until and including the October 2001 draft, referring to
figure 3-2 and the disagreement concerning the 10%-criterion. However, as can be seen from
figures 3-4 and 3-5, with different end moments there is a strong reduction of the slenderness
limit with increasing normal force, and that is true for both 10%-criteria. The only exception
is when criterion 2 is applied to columns with equal end moments, see figure 3-4.

3.1.4 Comparisons with previous model for slenderness limit

In tables 3-3 and 3-4, values from tables 3-1 and 3-2 are compared with values according to
the slenderness criterion in the October 2001 draft (expression 5.13):
 11

lim = 25( + 0,9)(2 - M01/M02)

In both cases the 10 % criterion is according to alternative 1.

Table 3-3. Slenderness limit in draft Oct. 2001 compared to values from table 3-1 ( = 0,3).
lim
n M01/ M02 10%-criterion (alt. 1)
Draft Oct. 2001
ef = 0 ef = 1
1,0 30 25 30
0,5 0 70 60 60
-0,9 110 90 87
1,0 20 10 30
1,0 0 45 30 60
-0,9 70 50 87

For n = 0,5 and ef = 0 the values according to draft October 2001 are reasonable (table 3-3).
For n = 1,0, however, they overestimate the slenderness limit, particularly for ef = 1 and most
particularly for M01/M02 = 1, where it gives a 3 times too high value.

Table 3-4. Slenderness limit in draft Oct. 2001 compared to values from table 3-2 (ef = 2).
lim
n M01/ M02 = 0,1 = 0,5
10%-criterion Oct. 2001 10%-criterion Oct. 2001
1,0 17 25 27 35
0,4 0 47 50 70 70
-0,9 74 73 107 102
1,0 11 25 20 35
0,8 0 27 50 43 70
-0,9 41 73 66 102

Draft October 2001 gives a fairly correct influence of , but it severely overestimates the
slenderness limit for a high normal force combined with a high creep effect, see table 3-4
(here the values are based on ef = 2). The omission of the effects of both normal force and
creep are the main disadvantages of this model.

3.1.5 Comparisons with new model

In tables 3-5 and 3-6 the new model (see p. 4) is compared to the same data as in tables 3-1
and 3-2 respectively.
 12

Table 3-5. Slenderness limit according to new model compared with table 3-1 ( = 0,3).
lim
n M01/ M02 ef = 0 ef = 1
10%-criterion New model 10%-criterion New model
1,0 30 26 25 21
0,5 0 70 63 60 52
-0,9 110 96 90 80
1,0 20 18 10 15
1,0 0 45 44 30 37
-0,9 70 68 50 56

Table 6. Slenderness limit according to new model compared with table 3-2 (ef = 2).
lim
n M01/ M02 = 0,1 = 0,5
10%-criterion New model 10%-criterion New model
1,0 17 17 27 24
0,4 0 47 42 70 58
-0,9 74 65 107 88
1,0 11 12 20 17
0,8 0 27 30 43 41
-0,9 41 46 66 62

Discussion
On the whole, the new model gives good agreement with the 10%-criterion (alt. 1), and the
main parameters are taken well into account. There is a slight overestimation of the slen-
derness limit for n = 1,0 and ef = 1, table 3-5, also for = 0,1 and n = 0,8, table 3-6. How-
ever, the overestimations are small compared to the old model (see 3.1.4) and in both cases
the values are conservative compared to 10%-criterion alt. 2.

A complete verification of the new model is given in Appendix 1. A more sophisticated

model could of course give even even better agreement, e.g. by also including the concrete
grade, but the present model is considered to be good enough.

The importance of considering creep in the slenderness limit is further substantiated in 4.3.3.

3.2 Effective length

New expressions for the effective length of isolated members in frames, expressions (5.15)
and (5.16), were introduced in the second draft. They have been derived to give an accurate
estimation, based on the definition of effective length in 5.8.1. They replace figure 4.27 in the
ENV, as well as expressions (5.22) and (5.23) in draft 1 of the prEN, December 1999.
 13

The expressions in draft 1 were taken from UK proposals, included in comments on the ENV
and on earlier drafts of the EN. It was found that they are very conservative in some cases,
giving up to 40 % overestimation of the effective length for braced members. They were also
found to be on the unsafe side in other cases, giving up to 20 % underestimation of the effec-
tive length for unbraced members.

It has been claimed that the conservativeness was deliberate, in order to cover certain unfa-
vourable non-linear effects. However, the effective length is by definition based on a linear
behaviour, and the present models are aimed at giving an accurate estimation according to
this, without including some hidden allowance for possible unfavourable effects. Such effects
are instead explicitly addressed in 5.8.3.2 (5) and in 5.8.7.2 (4). The new expressions also
avoid unsafe estimations, like for unbraced members with the previous expressions.

Figure 3-6a and b show a comparison between an accurate numerical calculation of the effec-
tive length and estimations according to draft 1 (a) and the final draft (b) respectively.

Braced Draft 1 Unbraced

1,0 10
l 0/l k 2=10 l 0/l
k 2=1000
0,9 k 2=3

0,8 k 2=0,4 Dashed = draft 1

Solid = accurate
k 2=0 0,4
0,7

0
0,6
Dashed = draft 1
Solid = accurate k1
k1
0,5 1
0,001 0,01 0,1 1 10 100 0,001 0,01 0,1 1 10 100

Figure 3-6a. Effective length according to accurate and simplified calculations, draft 1.

Braced Final draft Unbraced

1 10
l 0/l
l 0/l k 2=1000
k 2=1000 k 2=1
0,9 3
Dashed = draft 2 0,4
Solid = accurate 0,2
0,8 1
0,1 Dashed = draft 2
0 Solid = accurate 0,4
0,7
0,2
0
0,6

k1 k1
0,5 1
0,001 0,01 0,1 1 10 100 0,001 0,01 0,1 1 10 100

Figure 3-6. Effective length according to accurate and simplified calculations, final version.
 14

The present k-factors are defined differently compared to the corresponding factors in draft 1
and ENV, and are called k1 and k2 to avoid confusion with the previous factors ka and kb. The
present k-factors express the relative flexibility of the restraint according to the definition in
figures 3-7 and 3-8. They are applicable to different types of flexible moment restraint, such
as beams with different boundary conditions, flexible foundations etc.

M
lc
= M k1 1
EI c

EI c
k1 =
M lc
EIc, lc lc

k2 = 0 (rigid restraint)

Figure 3-7. Definition of flexibility of 2

restraint, example.

Draft 1 (and ENV) Final draft

EIb, lb
a, 1
EI c l c EI c l c
ka = k1 =
EI b l b 4 EI b l b
EIc, lc
kb = 0 k2 = 0

b, 2

EIb, lb
a, 1
EI c l c EI c l c
ka = k1 =
EI b l b 3EI b l b
EIc, lc
kb = k2 =

b, 2

Figure 3-8. Comparison between different definitions of k-factors; examples.

 15

5.8.3.2 (4) addresses the question whether an adjacent column (in a storey above or below) in
a node should be considered as using the same restraint as the column considered, or as con-
tributing to the restraint. This will depend on the magnitude of the axial force in the adjacent
column. If both columns connected to the node will reach their respective buckling load at the
same time (under proportional increase of loads), they will both have to share the restraint
provided by other connected members (beams), and k should then be defined as

EI aEI
k= + c (3-1)
M la lc

Here subscripts a and c refer to the adjacent column and the one considered respectively; see
Figure 3-9.

In the opposite case, where the adjacent column has a relatively low axial load, it can be in-
cluded among the members which resist the moment M, i.e. it will contribute to the restraint.
A reasonable model for the transition between the two limiting cases is the following:

EI EI
k= a + c (3-2)
M 1 + M 2 + ... + (1 ) M a la lc

where M1, M2 are restraining moments in members 1, 2, see Figure 3-9

Ma is restraining moment in the adjacent column, see Figure 3-9,
calculated without taking into account the axial force Na
= Na/NBa
Na is the axial force on the adjacent column
NBa is the buckling load of the adjacent column (can be estimated approximately,
e.g. taking into account only the horizontal members adjacent to its nodes)

Na
EIa adjacent
la column
Ma EI2, l2
M1
M2
EI1, l1
node
Nc
EIc column
lc considered

Figure 3-9. Illustration of node

with adjacent members.
 16

3.3 Global second order effects in structures (5.8.3.3)

3.3.1 Background
a) Model in ENV, A.3.2

According to the ENV, global second order effects may be neglected, and the structure may
be considered as non-sway, if

L FV / E cm I c (3-3)

where
L total height of building (htot in the ENV)
FV total vertical load (FV in the ENV)
EcmIc sum of bending stiffnesses in bracing members
0,2 + 0,1ns 0,6 (in the ENV, no particular symbol is used for this parameter)
ns number of stories (n in the ENV)

This criterion is valid only on certain conditions (which are not stated in the ENV):
1. No significant rotation at the base (rigid restraint / stiff foundation)
2. No significant global shear deformations (e.g. no significant openings in shear walls)

These conditions are not fulfilled for e.g. a bracing system including frames, nor for shear
walls with large openings and/or flexible foundations.

The criterion also explicitly requires that bracing members are shown to be uncracked. In
practice, bracing members are often more or less cracked in ULS, due to high lateral loading
and low vertical loading (most of the vertical load is often carried by the braced members).

For the above reasons, ENV A.3.2 has a very limited field of application. Since the limitations
of the applicability are not stated, and no information is given for the cracked stage, there is a
also risk that it is used outside the scope, giving unsafe results.

b) New proposals
Due to the above shortcomings, the ENV criterion was not included in earlier drafts of the
EN. After many requests to include something similar, two alternative proposals were pre-
sented to CEN TC250/SC2 (Berlin, May 2000), a mini-version and a full version:

Mini-version: same scope as in ENV-A.3.2, and a criterion given in a similar closed

form. However, the conditions and restrictions are clearly stated, and the criterion is im-
proved to be less conservative and to take into account cracking in a simple way.

Full version: formulated in a more general and transparent way. Detailed information is
given only for regular cases, but the formulation opens for general cases. A simple exten-
sion is given to cover the effect of global shear deformations.

In the final draft a somewhat extended mini-version is given in the main text, and further
extension to the full version are given in Annex H.
 17

3.3.2 No significant shear deformations, rigid moment restraint

The basic criterion second order effects 10% of first order effects gives, together with the
simplified magnification factor for bending moment in 5.8.7.3 (3):

M 0Ed
M Ed 1,1 M 0Ed (3-4)
1 FV,Ed / FV,BB

This gives the following criterion for the vertical load, cf. expression (H.1):

FV,Ed FV,BB 0,1 / 1,1 0,1 FV,BB (3-5)

Here
FV,Ed total vertical load
FV,BB nominal buckling load for global bending (no shear deformations)
M0Ed first order moment
MEd design moment

The global buckling load for bending can be written

FV,BB = 0
EI (3-6)
L2
where
0 coefficient depending on number of storeys, distribution of vertical load etc.
EI total bending stiffness of bracing members; to account for cracking in a simplified
way EI is based on 0,4EcdIc; for uncracked section 0,8 may be used instead of 0,4
L total height

Note that FV,BB is a nominal buckling load, calculated for a secant stiffness representing the relevant
ULS conditions (including lateral loading). Thus, it is not a load for which pure buckling (without
eccentricities or lateral loading) would occur.

The coefficient 0,4 (or 0,8) for estimating the stiffness (see H.1.2 (3)) can be compared to
0,3/(1+ef) in expression (5.26). Expression (5.26) is valid for isolated members, where all the
vertical load considered acts on the member itself. Then there is an effect not only of crack-
ing, but also of non-linearity in compression. This effect can be strong, particularly in cases
where the section is uncracked, usually associated with high vertical load. For the same rea-
son, a higher stiffness value for uncracked section is not given in 5.8.7.2. In a structure, on the
other hand, most of the vertical load is normally on the braced units, which means that there
is less effect of compression non-linearity on the bracing units, in which case a particular
value for uncracked section (0,8) is justified1. A further difference is that the bending moment
normally has a more favourable distribution in a bracing unit than in isolated members, which

1
The ratio 0,5 between the stiffnesses for cracked and uncracked sections is of course a rough simplification.
The ratio should depend on the reinforcement and the normal force, and with a normal force there is a more or
less smooth transition between the two stages. However, since this is about cases where second order effects are
more or less negligible, a simple rule is acceptable.
 18

gives less overall effect of cracking. These circumstances together justify the use of 0,4/0,8
instead of 0,3/(1+ef). Creep is not included in the criterion for neglecting second order ef-
fects in structures (as it is for isolated members). The reason is that for global second order
effects in structures, the dominating first order effect is wind. Then there is little effect of
creep, and consequently, the effective creep ratio according to 5.8.4 will be low.

The coefficient 0 in expression (3-6) depends on various parameters. For constant stiffness,
equal load increment per storey and rigid moment restraint at the base, 0 will depend on the
number of storeys and (to some extent) on the distribution of vertical load between braced and
bracing members according to Figure 3-9 (the buckling load has been calculated numerically
by Vianellos method, and 0 has then been evaluated according to expression (3-6)).

8
0
7
6 P = 100 %
of FV
5
L 4 N = 100 % of FV
3
2
1
N u m b e r o f s to rie s , n s
N P 0
on bracing on braced
members members 1 2 3 4 5 6 7 8 9 10

FV = N + P

Figure 3-9. Global buckling due to bending and coefficient for buckling load. Constant stiff-
ness and equal increment of vertical load per storey.

The coefficient 0 according to the upper curve in Figure 3-9 can be approximated by

nS
0 7,8 (3-7)
nS + 1,6

where ns = number of storeys

Combining expressions (3-4) to (3-6) gives

0,4 E cd I c ns E I E I
FV 0,1 0 = 0,312 cd2 c = cd2 c (3-8)
L2
ns + 1,6 L L

where 0,312 = 7,80,10,4 and = 0,312ns/( ns+1,6) (3-9)

This is the background to expression (5.18). Compare the ENV formulation (see 3.3.1 above):
 19

L FV / E cm I c (3-10)

Expression (5.18) can be formulated in the same way (substituting Ecd with Ecm and explicitly
including partial factor cE = 1,2 on the right hand side):

L FV / E cm I c / 1,2 (3-11)

In Figure 3-10 the two corresponding parameters /1,2 (EN) and 2 (ENV) are compared.
For the EN, curves for both cracked and uncracked sections are shown. The ENV gives no
values for cracked section, therefore there is no comparison for this case.

The comparison shows that for uncracked section, the two models give rather similar results,
although the ENV is often much more conservative.

0,5
2, /1,2 EN ( )
uncracked
0,4

ENV (2)
0,3 (uncracked)
EN ( )
cracked
0,2

0,1
Figure 3-10.
Comparison be- Number of stories, ns
tween EN and 0,0
ENV criteria.
1 2 3 4 5 6 7 8 9 10

3.3.3 Effect of flexible moment restraint

Flexible moment restraint at the base will reduce the buckling load. This effect can be directly
included in the factor for the buckling load. (For isolated members this effect is normally
taken into account by increasing the effective length. In the global analysis of structures, how-
ever, a direct reduction of the buckling load is more convenient, since the effective length is
not a practical parameter for bracing members and systems, having varying axial load.)

For bracing units with two or more stories, reasonably equal increment of the vertical load per
storey and constant cross section, the isolated effect of flexible end restraint can be accurately
modelled by the factor
 20

1
1 (3-12)
1 + 0,7 k
where

/M
k= (same definition as for isolated members, see 3.2)
L / EI

= rotation for bending moment M (compare Figure 3-7)

The factor 1 is an approximation, which has been derived by calibration against accurate
numerical calculations for different numbers of storeys, see Figure 3-11. The product 01
corresponds to in expression (H.2) in Annex H of the EN.

1,0
1
0,9

0,8
L
0,7

0,6 number of storeys:

ns=10
0,5
ns=2-4

M 0,4
1
0,3
Figure 3-11. Effect of ns=1
flexibility of end restraint 0,2
for bracing units. The solid
curves represent the 0,1
exact solution. The
dashed curve represents 0,0
the approximation accor- 0 1 2 3 4 5 k 6
ding to expression (3-12).

The effect of flexible moment restraint is not covered in the ENV, therefore no comparison
can be made.

3.3.4 Effect of global shear deformations

a) Shear deformations only (see Figure 3-12):

The (hypothetical) buckling load for shear deformations only is:

 21

FV,BS = S (or S for more than one bracing unit) (3-13)

Here S is the shear stiffness (= shear force giving a shear angle = 1; see figure H-1 in the EN).

Figure 3-12. Hypothetical

buckling due to global shear
deformations only. N P

b) Combined bending and shear:

The combined buckling load, taking into account bending and shear deformations, can be
expressed as
1 FV,BB FV,BB
FVB = = (3-14)
1 / FV,BB + 1 / FV,BS 1 + FV,BB / FV,BS 1 + FV,BB / S
Expression (3-14) can be derived analytically for simple cases like isolated members with
constant normal force. By numerical calculations, it can be verified also for bracing units with
vertically increasing axial load and significant global shear deformations (e.g. shear walls
with large openings).

The basic criterion for neglecting second order effects is the same as before:

FV 0,1 FV,B (3-15)

which leads to expression (H.6) in Annex H.

This case is not covered in the ENV, therefore no comparison can be made.
 22

4. Effective creep ratio (5.8.4)

4.1 General
The ENV stated that creep should be considered in connection with second order effects, but
gave no information on how. In the EN, on the other hand, practical models for taking into
account creep are given, based on the so called effective creep ratio.

A general approach would be to first calculate creep deformations under long-term load, then
to analyse the structure for the additional load up to design load. With the effective creep ra-
tio, the analysis can instead be made directly for the design load in one step.

Figure 4-1 illustrates a hypothetical load history and the corresponding deformations. The
total load is assumed to consist of one Long-term part QL (corresponding to the quasi-perma-
nent combination) and one additional short-term part up to the Design load QD, applied after a
long time.2 The total load history can then be divided into three parts:

1. AB - long-term load QL giving an elastic deformation

2. BC - constant load QL giving a creep deformation based on full creep coefficient
3. CD - additional load (QD - QL) giving an additional elastic deformation

Load Creep deformation

.
QD
D

QL
B C

Figure 4-1. Illustration of load

history and deformations.
Time A Deformation

The total deformation under long-term load can also be calculated directly using an equiva-
lent E-modulus3 for the concrete, Ee = Ec/(1+). This corresponds to line AC in figure 4-1.4

The total deformation under design load can be calculated in a similar way if an effective
creep ratio ef is used, line AD in figure 4-1. The effective equivalent concrete modulus
would then be

Eef = Ec/(1+ ef) where ef is the effective creep ratio

2
Subscripts L and D are used in this chapter for simplicity; they correspond to Eqp and Ed in 5.8.4.
3
Non-linear effect will be dealt with later, see clause 4.3.
4
Theoretically, this is not fully correct, since concrete stresses will decrease and reinforcement stresses increase
with time. However, it is a reasonable approximation in most cases.
 23

4.2 Effect of creep in cross sections

In the following, three examples are used to derive and illustrate the effective creep ratio ef.
The examples deal with bending moment and curvature in the following cases, assuming lin-
ear elastic material behaviour:

a) uncracked, unreinforced cross section (4.2.1)

b) uncracked reinforced section (4.2.2)
c) cracked reinforced section (4.2.3)

4.2.1 Uncracked unreinforced cross section

This is the simplest case for demonstrating the idea behind the effective creep ratio. The total
curvature under a long-term bending moment ML is (cf. line AC in figure 4-1):

1 ML
= (1 + ) (4-1)
r L Ec I c

The part that is caused by creep can be separated:

1 ML
= (4-2)
r C Ec I c

Under design load with total bending moment MD, part of which is a long-term moment ML
with a load history according to figure 4-1, the total curvature will be

1 MD 1 MD ML MD M MD
= + = + = 1 + L = (1 + ef ) (4-3)
r D Ec I c r C Ec I c Ec I c Ec I c M D Ec I c

Thus, the effective creep ratio is (cf. expression (5.19), with different notation):

ML
ef = (4-4)
MD

4.2.2 Uncracked reinforced cross section

The total curvature under long-term bending moment ML can be expressed in the following
simplified way, using an equivalent E-modulus for the concrete (see second footnote in 4.1):

1 M ML
= L = (4-5)
r L EI Ec
I c + Es I s
1+

The part of this curvature that is caused by creep can be separated:

 24

1 ML ML
= (4-6)
r C Ec Ec I c + Es I s
I c + Es I s
1+

Introduce the following parameters:

= (is/ic)2
= (Es/Ec)(As/Ac)
ic = radius of gyration of concrete area
is = radius of gyration of reinforcement area

The curvatures can then be expressed in the following way:

1 ML 1+
= (4-7)
r L E c I c 1 + (1 + )

1 ML 1+ 1
= (4-8)
r C Ec I c 1 + (1 + ) 1 +

Under design load with total bending moment MD the total curvature will be, including creep
due to a long-term bending moment ML:

1 MD 1 1 MD 1 ML 1+ 1
= + = + (4-9)
r D Ec I c 1 + r C Ec I c 1 + Ec I c 1 + (1 + ) 1 +

The same curvature expressed with the effective creep ratio would be, cf. expression (4-7):

1 MD 1 + ef
= (4-10)
r D E c I c 1 + (1 + ef )

Combining expressions (4-9) and (4-10) gives, after simplification:

1 + ef 1 M 1+ 1
= + L (4-11)
1 + (1 + ef ) 1 + M D 1 + (1 + ) 1 +

From this the effective creep ratio can be solved:

A(1 + ) 1
ef = (4-12)
1 A

where

1 M 1+ 1
A= + L
1+ M D 1 + (1 + ) 1 +
 25

Figure 4-3 shows the relationship between ef and ML/MD for = 3 and different values of
for the uncracked reinforced cross section.

3
ef Uncracked section
= total reinforcement
2,5

Eq (4-4) and (5.19)

2
=0
0,01
0,03
1,5
Figure 4-2. Effec-
tive creep ratio as
a function of the
1
ratio ML/MD for
= 3, uncracked
rectangular cross
0,5
section with sym-
metric reinforce-
ML/MD
ment, edge distance
t = 0,1h and = 0
0 0,2 0,4 0,6 0,8 1
Es/Ec = 6.

The straight line for = 0 is identical with expressions (4-4) and (5.19). With reinforcement,
i.e. > 0, expression (5.19) becomes more or less conservative.

4.2.3 Cracked reinforced cross section

The cracked cross section can be treated analogously, although it is a little more complicated.
As the simplest case, consider a rectangular cross section with bending moment only and ten-
sile reinforcement only. The flexural stiffness in the cracked stage (ignoring any contributions
from concrete in tension) can then be expressed as

EI = Es As d 2 (1 )(1 / 3) (4-13)

where
As = area of reinforcement
d = effective depth
= x/d
x = depth of compression zone

The relative depth of compression zone for a certain creep coefficient can be obtained from
 26

2
= (1 + ) 1 + 1 (4-14)
(1 + )

where = (Es/Ec)(As/Ac)

To simplify expressions, introduce the symbol

1
B = (4-15)
(1 )(1 / 3)

The total curvature under design load can then be written:

1 MD ML MD
= 2
B0 + 2
( B B0 ) = Bef (4-16)
r D Es As d Es As d Es As d 2

where B0 is parameter according to expr. (4-15) for = 0 and Bef is the same for = ef.

Values of ef for which expression (4-16) is satisfied can be found by iteration (direct solution
is not possible in this case). Figure 4-3 shows the result for = 3.

3
ef Cracked section
= tensile reinforcement
2,5

= 0,04
2 0,02
0,01
0,005
Figure 4-3. Effec- 1,5
tive creep ratio as
a function of ratio Eq (4-4) and(5.19)
ML/MD for a crac- 1
ked rectangular
cross section with
tensile reinforce- 0,5
ment only, based
on d = 0,9h and ML/MD
= 6. Basic creep 0
coefficient = 3. 0 0,2 0,4 0,6 0,8 1

In this case the curves will approach the straight line according to expression (5.19) the higher
the reinforcement ratio is. However, curves for low and moderate ratios are also quite close.

4.2.4 Conclusions concerning cross sections

The idea behind the effective creep ratio in 5.8.4 is illustrated in figure 4-1 and demonstrated
in three examples. The simple linear relationship according to expression (5.19) is always
 27

more or less conservative, but deviations are generally small. Furthermore, in a reinforced
section the overall effect of creep on the stiffness is reduced with increasing reinforcement,
since creep only affects the concrete contribution to the stiffness. Therefore, the effect of de-
viations on the stiffness will not be as strong as it may appear from the above figures.

4.3 Effect of creep in slender columns

4.3.1 General
The above derivations concern only cross sections and are based on linear material behaviour.
In this clause the relevance of the effective creep ratio for slender columns will be examined.
A slender column behaves in a non-linear way, due to both material and geometrical non-line-
arity. A non-linear behaviour similar to the linear one in figure 4-1 is outlined in figure 4-4.

Load

QD
D

QL 3
B 2
C

Figure 4-4. Illus-

tration of load his- Time Deformation
tory and deforma- t t0 A y1 E y2 y3
tions with non-
y0 = y 2 - y1
linear behaviour.

The load history can be divided into three steps:

1. Application of long-term load QL, immediate deformation y1, calculated for ef = 0

2. Long-term load QL during time t-t0, total deformation y2, calculated for ef =
3. Load increase up to design load QD, additional deform. y3 - y2, calculated for ef = 0

A realistic calculation representing this load history should involve these three steps, includ-
ing the relevant first order moments or eccentricities for each step. As a simplification, steps 1
and 2 can be combined into one, using a stress-strain diagram with the strains multiplied by
(1+), see 6.4. This corresponds to line AC, and the calculation is then reduced to two steps.

The last step can be calculated in two alternative ways:

a. After calculating point C, the additional load QD QL is added, with a deformation start-
ing from y2. See line CD in figure 4-4.
b. After calculating point C, the total load QD is applied from scratch, but with y0 = y2 - y1
as an initial deflection, added to other first order effects. See line ED in figure 4-4.

Alternative b will be used in two-step calculations in the following. The distribution of y0

along the column should in principle be the same as the distribution of y2 y1. For a pin-ended
column, however, a sine-shaped or parabolic distribution will do as a simplification.
 28

A further simplification is a one-step calculation, using an effective creep ratio ef; line AD in
the figure. For the definition of ef there are two main options:

a) based on first order moments M0L and M0D, i.e. ef = M0L / M0D
b) based on total moments ML and MD, including 2nd order moments, i.e. ef = ML / MD

The relevant deformation parameter in second order analysis is curvature, which depends
primarily on bending moment. Therefore, the axial load should not be included in the defini-
tion of effective creep ratio.

Alternative b) is the most realistic one, since creep deformations will mainly be governed by
total moments. With this alternative, however, iteration is inevitable since second order mo-
ments depend on stiffness, which depends on effective creep ratio, which depends on total
moments etc. Therefore, alternative a) will be the normal choice in practical design.

Alternative a) is always more or less on the safe side. The reason is that the second order mo-
ment is a non-linear function of the axial load. Therefore, the moment increase due to second
order effects will be greater under design load than under long-term load, and the ratio ML/MD
will be lower if second order moments are included. This is easy to verify with a magnifica-
tion factor based on linear material behaviour (see chapter 7); this tendency will be even
stronger in a non-linear analysis. 5

4.3.2 Comparison between one- and two-step calculations

An example will be used to compare the one-step calculation, using the effective creep ratio,
with the more realistic two-step calculation. A high slenderness ratio has been chosen, in or-
der to emphasize the effects considered. All calculations below have been done with the gen-
eral method. (For a general description and discussion of this method, see chapter 6.)

- Concrete C40
- Reinforcement S500
- Rectangular cross section with reinforcement concentrated to opposite sides
- Mechanical reinforcement ratio = 0,15 (total reinforcement)
- Edge distance of reinforcement 0,1h
- Eccentricity e0 = 0,08h (same for long-term and design load; no other first order effect)
- Slenderness l/h = 40 ( = 139)
- Basic creep coefficient = 3

In the following, all axial loads and bending moments will be expressed in relative terms, i.e.
n = N / Acfcd and m = M / hAcfcd. Therefore, no absolute dimensions are given.

a): nL = 0,100 (long-term axial load)

1. Immediate deformation, calculated with = 0: y1/h = 0,0173
2. Total deformation, calculated with = 3: y2/h = 0,0819

5
A "curvature method", giving a fixed 2nd order moment (see ch. 8), would lead to the wrong conclusion here.
 29

3. Creep deformation: y0/h = 0,0819 0,0173 = 0,0646

y0 is taken as an initial deflection with parabolic distribution, and is added to the constant
first order eccentricity e0 given above. The load capacity with this total first order effect
and with no creep ( = 0) is calculated. The result is nRd = 0,235

b): nL = 0,125
1. y1/h = 0,023 ( = 0)
2. y2/h = 0,134 ( = 3)
3. y0/h = 0,134 0,023 = 0,111 (creep deformation)
The load capacity calculated with e0+ y0 and with = 0 is nRd = 0,189

These values will be compared to the result of a one-step calculation, using an effective creep
ratio based on first order moments.

a) nL = 0,100
ef = M0L / M0D = NLe0 / (NDe0) = nL / nD = 30,100/0,235 = 1,28 6
The load capacity with ef = 1,28 is nRd = 0,198
Cf. 0,235 in two-step calculation; thus the result is 16 % conservative

b) nL = 0,125
ef = 30,125/0,189 = 1,98 nRd = 0,166
Cf. 0,189 in two-step calculation; thus the result is 12 % conservative

These results are a bit conservative, as could be expected (the reason is explained above).

Next is a one-step calculation with ef based on total moments.

a) nL = 0,100
Total moment under nL is mL = nL(e0 + y2) = 0,100(0,08 + 0,0819) = 0,0162
After iteration the following values are found:
Total moment under design load mD = 0,0618
Effective creep ratio ef = mL / mD = 30,0162/0,0618 = 0,786
Load capacity with ef = 0,786 is nRd = 0,224 (total moment for this load is mD = 0,0618)
This is within 5 % of the two-step calculation (which gave nRd = 0,235)

b) nL = 0,125
Total moment under nL is mL = nL(e0 + y2) = 0,125(0,08 + 0,134) = 0,0267
After iteration: mD = 0,0531, ef = mL / mD = 30,0267/0,0531 = 1,151, nRd = 0,183
This is within 3 % of the two-step calculation (0,189)

6
In this particular example the first order moment is proportional to the axial load, therefore the effective creep
ratio can be based on axial loads as well as moments. In a general case only moments should be used.
 30

Conclusions
It is conservative to use an effective creep ratio based on first order moments; total moments
will give more accurate results. In practical design total moment are much more complicated
to use, however, since iteration will be necessary. Therefore, the normal procedure will be to
use first order moments. This is further discussed below.

4.3.3 The effect of creep on the slenderness limit

The effect of creep on the slenderness limit will be further studied here, comparing the one-
and two-step methods according to 4.3.1 and 4.3.2. It is thus a complement to clause 3.1, deal-
ing with the slenderness limit in general. It is also a complement to 4.3.2, dealing with creep
combined with a high slenderness, since this clause deals with low slenderness ratios.

Table 4-1 shows the results of calculations, with a slenderness corresponding to the limit for
which second order effects may be neglected with ef = 0, see 3.1. The basic parameters are
the same as for the example in 4.3.2, except those for which different values are given.

Table 4-1. The effect of creep for columns with a low slenderness.
C e02/h e01/e02 l0/h nu0 nL/nu0 ef nu1 nu2 nu1/nu0 nu2/nu0 nu1/nu2
1,0 7,3 0,418 0,426 0,94 0,95 0,98
0,1 0,32 0,0 19,2 0,446 0,5 1,5 0,362 0,363 0,81 0,81 1,00
-0,9 31,2 0,325 0,348 0,73 0,78 0,93
20
1,0 8,1 0,702 0,697 0,97 0,96 1,01
0,5 0,32 0,0 20,0 0,725 0,67 2,0 0,619 0,616 0,85 0,85 1,00
-0,9 30,5 0,596 0,593 0,82 0,82 1,01
1,0 7,4 0,579 0,563 0,98 0,95 1,03
0,67 2,0
0,32 0,0 19,0 0,592 0,463 0,455 0,78 0,77 1,02
40 0,3 -0,9 29,4 0,5 1,5 0,453 0,498 0,77 0,84 0,91
0,08 0,0 12,9 1,033 0,849 0,949 0,82 0,92 0,89
0,67 2,0
1,28 0,0 34,7 0,122 0,111 0,095 0,91 0,78 1,17
1,0 6,8 0,393 0,404 0,91 0,94 0,97
0,5 1,5
0,1 0,32 0,0 17,2 0,432 0,314 0,304 0,73 0,70 1,03
-0,9 26,8 0,33 1,0 0,313 0,367 0,73 0,85 0,85
80
1,0 8,6 0,647 0,645 0,94 0,93 1,00
0,67 2,0
0,5 0,32 0,0 19,4 0,690 0,520 0,522 0,75 0,76 1,00
-0,9 29,4 0,5 1,5 0,518 0,604 0,75 0,88 0,86
0,83 0,85 0,98
Symbols in table:
e02 the greater of the two first order eccentricities
e01 the lesser eccentricity
l0/h slenderness corresponding to the limit for 10 % moment increase at = 0
n relative normal force N/Acfcd
nu0 load capacity for the current slenderness and = 0
nL long-term load
ef effective creep ratio = nL / n0 7; here = 3 has been assumed
nu1 load capacity including the effect of creep according to 1- step method
nu2 load capacity including the effect of creep according to 2- step method

7
In these examples M0 = Ne0, and then ef = NLe0 / Nue0 = NL / Nu. Generally, ef = ML / Mu.
 31

Comments
The agreement between the 1- and 2-step methods is in most cases good. For e01/e02 = -0,9
(double curvature bending) the 1-step method is generally slightly conservative compared to
the 2-step method. There are also a few cases where the opposite is true, but in these cases the
long-term load is close to the limit where instability would occur with ef = , and then the 2-
step method becomes uncertain. In these cases, the result would have been more representa-
tive with a somewhat lower long-term load.

The use of an extended stress-strain diagram in the 2-step method can be discussed. In princi-
ple it means that the creep deformations will correspond to the stresses in the final stage. In a
more accurate calculation they should be integrated from 0 to , with increasing second order
moment. However, the error will be small, since the stresses are normally not very high under
long-term load, and since second order moments are small at these low slenderness ratios.

In most cases a first order eccentricity e02/h = 0,32 has been used, with the aim of having a
moderate normal force. For the sake of completeness, one case with a high normal force is
also included (nu0 = 1,022, e02/h = 0,08) and one with a low normal force (nu0 = 0,122, e02/h =
1,28). Even with the low normal force, there is a significant effect of creep. (The 2-step
method here seems to give more effect of creep for low than for high normal force. This is
misleading, however; it is a consequence of the long-term load being close to the instability
load for ef = , see discussion in the first paragraph above.)

Conclusion concerning the effect of creep on the slenderness limit

In these examples, creep reduces the load capacity by 5 to 30 % (average 15 %). If second
order moments are neglected, which is allowable at these slenderness ratios, the result is in
principle already 10 % on the unsafe side. If creep would also be neglected, the results would
be another 5 to 30 % on the unsafe side.

The conclusion is that creep can not be neglected in the slenderness limit.

4.3.4 Safety under long-term load only

The effective creep ratio is based on moments under quasi-permanent load which, according
to its definition in EN 1990, is an SLS load with no load factors (except 2 < 1 for variable
loads). Thus, in the extreme case of permanent load only, assuming first order moments pro-
portional to the load as in the above examples, the highest possible effective creep ratio is

ef = M0L/M0D = 1,0/1,35 = 0,74

The following question now arises:

Can load NL together with ef = be more severe than ND = 1,35NL with ef = 0,74 ?

The example in 4.3.2 is used again. The load capacity for ef = 0,74 = 2,22 is found to be
nRd = 0,159. The corresponding long-term load is nL = 0,159 / 1,35 = 0,118. The load capac-
ity for full creep, ef = = 3, is found to be nRdL = 0,141. This is higher than the current long-
term load, and the safety factor is
 32

L = 0,141/0,118 = 1,19

This safety may be considered a bit low, although it should be observed that it is not the
whole safety; the normal material safety factors are already included in the calculated capaci-
ties. A reasonable lower limit for the load safety factor could be 1,35.

As shown above, it is conservative to use an effective creep ratio based on first order mo-
ments. The extra safety can be estimated by comparison with more accurate calculations,
e.g. a two-step calculation or a one-step calculation with ef based on total moments.

A two-step calculation according to the above scheme is done with different values of nL, un-
til a value is found for which nRd = 0,159. This happens for nL = 0,134. Thus, one could say
that the additional built-in safety is 0,134/0,118 = 1,14, and the total safety against creep
failure would be

L = 1,141,19 = 1,36

This is sufficient, and it can be shown that this safety will be higher for lower values of slen-
derness, higher first order moments and higher amounts of reinforcement. In this respect, the
current example is rather extreme, in the unfavourable direction.

Furthermore, long-term load = 74 % of design load is the worst possible case as regards the
effect of creep. In normal cases there is always some variable load. The percentage of long-
term load then decreases, since variable loads are included in QL with 2Qk, where 2 < 1 8,
whereas in QD they are included with QQk, where Q > 1. Therefore, the more variable load,
the higher the safety against creep failure will be.

The conclusion is that a one-step calculation, using an effective creep ratio based on first or-
der moments, will give sufficient safety against failure under quasi-permanent load with full
creep. Therefore, this case need not be checked separately, and it is not necessary to include
any safety factor on M0L in the definition of ef.

In 5.8.4 (3) the alternative to use total moments in the definition of effective creep ratio is
given. This is less conservative, however, and most of the extra safety against creep failure
is then lost. Therefore, 5.8.4 (3) states that a separate check should then be made for 1,35 QL
and with ef = . This may become governing in cases where the percentage of long-term
moment is moderate or high, more precisely when first order moment ratio M0L / M0D > 0,5.

8
EN 1990 gives values for 2. For some loads, e.g. wind, 2 = 0. A common value is 0,3 (office and residential
areas). The highest value given is 0,8.
 33

5. Methods of analysis (5.8.5)

Three basic types of methods are described in 5.8.5. Of the simplified methods, (b) is basi-
cally the model column method in the ENV, with some modifications. The old name is not
used here, since it tells nothing about the method (all methods are based on models). A more
suitable name is curvature method, since the method is based on the estimation of a curva-
ture. This name will be used here, together with stiffness method for method (a), which is
based on the estimation of stiffness.

There are other simplified methods than those mentioned in EC2. One such method, combin-
ing analysis and cross section design in one step, will be shortly described here as an example
(it is currently used in the Swedish code). It can be used for isolated columns with formally
centric load, i.e. no other first order effect than the prescribed imperfection. The load bearing
capacity is given as

NRd = kcfcdAc + ksfydAs (5-1)

where kc and ks are coefficients depending on slenderness ratio, imperfection, concrete grade,
effective creep ratio etc, calibrated against calculations with the general method.

A method of this type works ideally if the imperfection, an eccentricity or an initial deflection,
is proportional to the buckling length of the column. This is the case in some codes, but not in
the Eurocodes. If the imperfection is proportional to the effective length, the coefficients can
be given in one simple table or diagram with slenderness as the basic parameter.

If the imperfection is not proportional to the effective length, then the absolute value of this
length must be added as a separate parameter, which complicates the presentation (for exam-
ple, one diagram or table would only be valid for one length). However, with some simplifica-
tions this type of method could be useful also under EN 1992, particularly for storey high pin-
ended columns, which are common as interior columns in buildings.

If there are first order moments other than that due to the imperfection, a separate design for
normal force and (magnified) moment must be made. A special moment magnification factor
is included in the method for such cases, but the simplicity is lost and the method no longer
has any particular advantages over the stiffness or curvature methods in EC2.

In the following chapters, the general method and the simplified methods (a) and (b) are de-
scribed.
 34

6. General method (5.8.6)

6.1 General
The most accurate of the methods described in 5.8.5 is the general method. It is based on
non-linear analysis, including both material and geometric non-linearity (second order ef-
fects). General here refers to the fact that the method can be used for any type of cross sec-
tion, any variation of cross section, axial load and first order moment, any boundary condi-
tions, any stress-strain relations, uniaxial or biaxial bending etc. The limiting factor is the ca-
pability of the available computer program. The method rests on a few simple assumptions:

linear strain distribution

equal strains in reinforcement and concrete at the same level
stress-strain relationships for concrete and steel

Conditions of equilibrium and deformation compatibility are satisfied in a number of cross

sections, and the deflection is calculated by double integration of the curvature, having an
assumed variation between the selected sections. This may be self-evident, but it is mentioned
in 5.8.6 (6) as a reference for a simplified version, in which only one cross section (or certain
critical sections) is studied, and the curvature is pre-assumed to have a certain variation in
other parts of the member. This gives simpler computer programs and faster calculation, but
less accuracy. See figure 6-1.

assumed assumed variation

variation of of curvature, e.g.
curvature, e.g. similar to that of
linear, constant the first order
or other moment

cross section(s)
for which
equilibrium and
deformation
compatibility
are satisfied

Figure 6-1. Illustration of

accurate (left) and simplified
(right) versions of the general
method.

Any stress-strain relations can be used. A continuous curve with a descending branch is con-
sidered to be the most realistic alternative for the concrete; it is also convenient for computa-
tional reasons. Creep can be considered in different ways; the simplest way is to multiply all
concrete strains with (1+ef), see clause 6.4.
 35

Tension stiffening (i.e. the contribution from concrete in tension between cracks) can easily
be taken into account in the general method, e.g. by using a descending branch of the concrete
stress-strain curve in tension, by modifying the stress-strain curve of the reinforcement or by
any other suitable model. In the calculations presented in this report, however, all contribu-
tions from concrete in tension have been ignored; this is always more or less conservative.

6.2 Safety format

The safety format in non-linear analysis has been much debated, and different models have
been proposed. The safety format is particularly important in second order analysis, where the
absolute magnitude of deformations has a direct influence on the ultimate load.9

The safety format should satisfy two basic criteria.

1. It should be possible to use the same set of material parameters in all parts of the
member, in order to avoid discontinuities and computational problems.

The model in ENV 1992-1-1 (Appendix 2) does not comply with this, since it assumes mean values of
material parameters for the calculation of deformations and design values for the check of resistance in
critical sections. This also means that there will be no material safety at all in the calculated resistance,
in cases where failure occurs before reaching the design cross section resistance (stability failure) unless
critical section is substituted by some critical length (which then remains to be defined, however).

2. The safety format should be compatible with the general design format based on par-
tial safety factors.

The model in ENV 1992-2 (Appendix B) does not comply with this, since it uses mean values for the
analysis and a global safety factor R = 1,3 to reduce the ultimate load resulting from the analysis. This
gives the same results as using design values fcm/1,3, fyk/1,3, Ecm/1,3 and Esm/1,3. Thus, it makes no dif-
ference whether the ultimate load is governed by concrete or steel, resistance or stiffness. The reduction of
the reinforcement strength is too severe, as is also the reduction of the material stiffness parameters, parti-
cularly for reinforcement (Esm/1,3). A non-linear analysis using this safety format will be conservative,
and the potential benefits of using a refined method are lost.

The safety format defined in 5.8.6, based on using design values in the analysis, satisfies both
criteria. A design value of the ultimate load will be obtained as a direct result of the analysis,
and the problems associated with the above-mentioned safety formats are avoided. Since the
E-modulii vary less than the corresponding strengths, the partial safety factors given for E
should be lower than for f:

For concrete, c = 1,5 for strength considers not only strength variation, but also geometrical
deviations in the cross section. Assuming a factor 1,1 for these deviations, and considering the
relationship between strength and E-modulus, a reasonable value of the factor for Ec is cE =
1,1(1,5/1,1)1/3 1,2.

For steel, s = 1,15 includes a factor of about 1,05 for geometrical deviations. Thus, a design
value Esd=Esm/1,05 would be logical, considering the fact that variations in the E-modulus are
negligible. However, a factor 1,0 has been chosen as a simplification, and in order not to de-
viate from 3.2.3; differences in terms of calculated result are negligible.

9
The absolute magnitude can be of importance also in e.g. continuous beams, but only in the check of rotation
capacity, and it would normally not have the same direct influence on the ultimate load as in 2nd order analysis.
 36

6.3 Interaction diagrams

The resistance of slender columns resulting from a general analysis can be shown in a prac-
tical form with interaction curves, figure 6-2. One such curve shows the maximum first order
moment M0 (or eccentricity e0 = M0/N) for a certain axial load N.

The thin curves in figure 6-2 show the total moment M as a function of N for a given e0. The
higher the slenderness, the more the total moment M increases over the first order moment
M0. (Note that the diagram gives axial load and moment in relative terms n and m.) One point
on the interaction curve for a certain slenderness is obtained by plotting the maximum value
of n on the line representing m0 or e0. This is demonstrated in figure 6-2 for one relative ec-
centricity e0/h = 0,1 and different slenderness values = 35, 70, 105 and 140.

The difference Mu - M0 between the cross section resistance (curve = 0) and the first order
moment at maximum load represents the second order moment. However, in some cases there
is a stability failure before any cross section reaches its ultimate moment, and then the true
second order moment is less than Mu - M0. This occurs for = 105 and 140 in figure 6-2.

This nominal second order moment Mu - M0 is useful as a basis for simplified methods; see
clause 6.5 and chapters 7 and 8.

0,2
m m0 m u
mu

Figure 6-2. Interaction 0,15

curves for columns of
e/h=
different slenderness,
m0/n=
calculated with the gen- 0,1
eral method. Rectangular
cross section. n and m0 0,1
are relative axial force =0
and first order moment 35
respectively, i.e. n =
N/bhfcd, m0 = M0/bh2fcd. 0,05
70
All curves are based on
= 0,2 and ef = 0. 105
Concrete grade is C80.10 140
First order moment is
constant, e.g. caused by 0
equal end eccentricities. 0 0,2 0,4 0,6 0,8 1 n 1,2

6.4 The effect of creep

Creep can be taken into account in different ways. The most accurate model would be to in-
crease load and time in steps, for each step taking the stresses, strains (and corresponding de-
flections) from the previous step as starting values for the next increment. For each step,
strains would be calculated taking into account their time-dependence.
10
This diagram is taken from [1], which primarily deals with high strength concrete according to Swedish rules,
but this makes no difference for what the diagram is intended to show.
 37

A simplified model is to multiply all strain values in the concrete stress-strain function with
the factor (1+ef), see figure 6-3, where ef is an effective creep ratio relevant for the load
considered. With this model, the analysis can be made either in steps for loads of different
duration, or directly for the design load combination in one step, see chapter 4. For creep in
slender members in particular, see clause 4.3.

=0 = ef

Figure 6-3. Simple way of taking

into account creep in general
0 =(1+ef).0
method.

Figure 6-4 is calculated in this way, using ef = 2,0 and other parameters the same as in figure
6-2. Curves according to figure 6-2 are also included (dashed), showing the reduction of the
load capacity resulting from creep. The relative reduction increases with slenderness.

0,2
m0

=0
0,15
35

0,1
Figure 6-4. Interaction 70
curves for ef = 2. Other
parameters are the same
105
as in figure 6-2. Dashed 0,05
curves are the correspond-
ing curves from figure 6-2,
i.e. for ef = 0. The differ-
140
n
ence represents the effect 0
of creep. 0 0,2 0,4 0,6 0,8 1 1,2

Another question is whether one and the same effective creep ratio should be used along a
compression member (or in different parts of a structure), or if it should vary as the ratio
MEqp/MEd may vary. The latter would be the most correct alternative, but normally it is rea-
sonable to use one representative value of ef for a member or even a whole structure.
 38

6.5 Simplified methods and their common basis

In a simplified calculation method one can use the difference between cross section resistance
and first order moment, Mu - M0 in figure 6-2, as a nominal second order moment. When this
moment is added to the first order moment, a design moment is obtained for which the cross
section can be designed with regard to its ultimate resistance. As pointed out above, this
nominal second order moment is sometimes greater than the true second order moment.
However, it can give correct end results, even in cases where the load capacity is governed by
a stability failure before reaching the cross section resistance, if given appropriate values.

For practical design, there are two principal methods to calculate this nominal second order
moment:

1. estimation of the flexural stiffness EI to be used in a linear second order analysis (i.e. con-
sidering geometrical non-linearity but assuming linear material behaviour); this method is
here called stiffness method, see chapter 7

2. estimation of the curvature 1/r corresponding to a second order deflection for which the
second order moment is calculated; this method is here called curvature method, see
chapter 8.

Before entering into details of the two methods in chapters 7 and 8, their common basis will
be shortly described.

The total moment including second order moment for a simple isolated member is:

1 l2
M = M0 + M2 = M0 + N y = M0 + N (6-1)
r c
where (see figure 6-5)
M = total moment
M0 = first order moment
M2 = second order moment
N = axial force
y = deflection corresponding to 1/r
1/r = curvature corresponding to y
l = length N
c = factor for curvature distribution
Transverse load

r
y

Figure 6-5. Illustration of deformations and mo-

ments in a pin-ended column. (In the figure, first
order moment is exemplified as the effect of a M0 M2
transverse load. First order moment could also be
M 1/r
given by eccentricity of the axial load.)
 39

The difference between the two methods lies in the formulation of the curvature 1/r.

In the stiffness method 1/r is expressed in terms of an estimated nominal flexural stiffness EI:

1 M
= (6-2)
r EI

The stiffness EI should be defined in such a way that ULS cross section design for the total
moment M will give an acceptable end result in comparison with the general method. This
includes, among other things, taking account of cracking, creep and non-linear material prop-
erties.

In the curvature method, the curvature 1/r is estimated directly, on the basis of assuming yield
strain in tensile and compressive reinforcement:

1 2 yd
= (6-3)
r 0,9d

This model overestimates the curvature in those cases where yielding is not reached, giving a
too conservative end result. The typical example is where the ultimate load is governed by
stability failure, before reaching the cross section resistance. The model may also underesti-
mate the curvature in some cases, since it does not take into account creep. However, various
corrections can be introduced to improve the result.

In the following chapters the two simplified methods will be described and compared to the
general method.
 40

7. Method based on stiffness (5.8.7)

7.1 Basic equations
A simple isolated column is considered, e.g. pin-ended with a length l = l0; see figure 6-5. The
second order moment can be expressed in the following way, cf. equation (6-1) and fig. 6-5:

1 l2 M l 02 l2 M M
M2 = N y = N 0 = N = N 0 0 + 2 (7-1)
r c EI c EI c0 c2

With c0 and c2 it is possible to consider different distributions of first and second order mo-
ments (primarily the corresponding curvatures). Solving for M2 gives

l 02
N
c 0 EI c 2 / c0
M2 = M0 = M0 (7-2)
l 2
c 2 EI / l 02 N 1
1 N 0
c 2 EI

In many cases it is reasonable to assume that the second order moment has a sine shaped dis-
tribution. This corresponds to c2 = 2, and M2 can then be written

2 / c0
M2 = M0 = M0 (7-3)
EI / l N 1
2 2
0 NB / N 1

where NB = nominal buckling load (based on nominal stiffness)

= 2/c0, parameter taking into account the distribution of first order moment

The total moment will be

M = M 0 1 + (7-4)
N B / N 1

which corresponds to equation (5.28) in 5.8.7.

7.2 Moment distribution

In some cases the value of c0 is known, as in the examples mentioned in 5.8.7.3 (2).

The case of differing end moments will be examined more closely. A reference is made to
5.8.8.2 (2), with the well-known formula for an equivalent constant first order moment:

M0e = 0,6 M02 + 0,4 M01 0,4 M02 (7-5)

This is illustrated in figure 7-1.

 41

M02
M02

Total
moment
1st order
moment Equiva-
lent total
moment
Equivalent 1st
order moment
Figure 7-1. Illustra-
tion of equivalent mo- M01
ments in case of differ-
M01
ing end moments.

Equation (7-4) can be used with the equivalent first order moment according to (7-5). An ex-
ample of the result is shown in figure 7-2, where two different values of c0 were used: 8 and
10 respectively.
M max/M 02
2,5
Figure 7-2.
Slender member with differing end
2,0
moments according to figure 7-1
with e02/h = M02/Nh = 0,1 and c0 = 8
c 0 = 10
NB/N = 2:
1,5
Comparison between maximum mo-
Exact
ment according to exact solution and
equivalent first order moment (7-5) 1,0
with magnification factor (7-4).
Thick line = exact solution.
0,5
Upper thin line = equivalent mo-
ment with c0 = 8. M 01/M 02
Lower thin line = equivalent mo- 0,0
ment with c0 = 10. -1 -0,5 0 0,5 1

Figure 7-2 shows good agreement with the exact solution for c0 = 8, whereas c0 = 10 may give
slightly unsafe results. Therefore c0 = 8 is recommended in 5.8.7.3 (2); this is also consistent
with the assumption of a constant equivalent first order moment. The example is based on a
comparatively high second order effect (N/NB = 0,5), which enhances the differences.
 42

In many cases it is reasonable to assume that first and second order moments have similar
distributions, in which case 1. Equation (7-4) can then be simplified to

M0
M= (7-6)
1 N / NB

This corresponds to equation (5.30) in 5.8.7.3. It can be shown that this expression can be
used also for structures, provided a global buckling load can be defined. See 7.4 for global
analysis of structures.

7.3 Estimation of stiffness

The maximum first order moment M0 for different slenderness ratios, axial forces and rein-
forcement ratios can be determined by calculations with the general method, see chapter 6.
The values thus obtained can be considered as the correct ones. With the stiffness
method, on the other hand, the maximum first order moment can be expressed on the basis of
equation (7-4), assuming that the total moment is equal to the ultimate moment resistance Mu
for normal force N:

Mu Mu
M0 = = (7-7)

1+ 1+
NB / N 1 2 EI / l 2 N 1

The following is a simple model for the stiffness, expressed as the sum of separate contribu-
tions from concrete and reinforcement:

EI =K c Ec I c + K s Es I s (7-8)

where Ec/s = E-modulus of concrete and steel respectively

Ic/s = moment of inertia of of concrete and steel area respectively

The correction factors Kc and Ks can be calibrated in more or less sophisticated models, to
give the desired agreement between expression (7-7) and the general method. In 5.8.7.2 (2)
basically two alternative models are given. a) (expr (5.22)) is a more accurate alternative,
valid for reinforcement ratios down to = 0,002. b) (expr (5.26)) is a simplified alternative,
valid only for reinforcement ratios 0,01. Thus, for < 0,01 only a) may be used, but for
0,01 either method may be used.

Ks = 1
a) if 0,002: (7-9)
Kc = k1k2 / (1 + ef)

Ks = 0
b) if 0,01: (7-10)
Kc = 0,3 / (1 + 0,5ef)

where is geometrical reinforcement ratio, As/Ac

ef is effective creep ratio, see chapter 4
k1 depends on concrete strength class, see (7-11)
k2 depends on axial force and slenderness, see (7-12)
 43

k1 = f ck / 20 (fck in MPa) (7-11)

k2 = n 0,20 (7-12)
170

where n is relative axial force, NEd / (Acfcd)

Ac is area of concrete cross section
is slenderness ratio, l0/i

For cases where is not defined, a simplified alternative to (7-12) is also given (5.25):

k2 = n0,30 0,20 (7-13)

More sophisticated models for estimating the stiffness can be found in [2] and [3]. Background, see [1].

The results of calculations with stiffness according to expressions (7-9) to (7-12) are presen-
ted in Appendix 2 of this report, in the form of comparison with calculations with the general
method. The Appendix also compares the curvature method; see chapter 8.

7.4 Linear analysis of structures

Clause 5.8.7 opens for the possibility to use linear second order analysis for structures, using
reduced stiffness(es) taking into account the effect of cracking, creep and material non-
linearity in a simplified way. Without this possibility, the only alternative for second order
analysis of structures would be non-linear analysis.

When global second order effects are significant, the effects of cracking etc. may be as impor-
tant as for isolated members. It should also be kept in mind that second order effects may be
significant in a structure, even if the geometrical slenderness of individual bracing units is
small, namely if the braced units carry a comparatively high vertical load.

The paragraphs applicable to structures are 5.8.3.3 (criterion for neglecting global second or-
der effects), 5.8.7.3 (3) and Annex H. Two different approaches can be distinguished, one
based on a magnification factor for bending moments, 5.8.7.3 (3), and the other based on a
similar factor for horizontal forces, H.2.

The two approaches are basically the same, but the one based on moments is suitable mainly
for structures with bracing units consisting of shear walls without significant global shear de-
formations, or structures braced by simple cantilever columns, see examples in figure 7-3.
 44

bracing braced
member(s) members

H0Ed

MEd MEd

Figure 7-3. Example of structures where a magnification factor can be applied directly to
bending moment(s) in bracing unit(s).

The approach based on magnification of horizontal forces, on the other hand, can be used for
all kinds of structures, and it should be used for frames, shear walls with large openings etc. If
properly used, it gives the correct second order effects in structural systems like frames, shear
walls with or without openings etc; see the schematic example in figure 7-4.

Bracing members Braced members

HEd

Figure 7-4. Example of a structure

where the magnification factor
should be applied to horizontal
forces rather than to bending mo-
ments. (No deformation is shown in
this case, second order effects are
instead assumed to be included in
the fictitious, magnified horizontal VEd
force HEd.)
 45

Expressions (5.30) and (H.7) are useful if the global buckling load can be defined without
difficulty, like in certain regular structures, see e.g. 5.8.3.3 and H.1. In other cases second
order effects may be calculated step-wise as indicated for a simple frame in figure 7-5.

y0 y1 y1 y2
V V V

H0 H0
a) b) H1 c) d) e)
H2 y3 etc.

Figure 7-5. Illustration of step-wise calculation of second order effects.

a) Horizontal load H0 (without vertical load) gives deformation y0.
b) Verical load V on deformed structure gives additional deformation y1.
c) H1 is an equivalent horizontal load that would give the same deformation y1.
d) Vertical load V and deformation y1 give additional deformation y2.
e) H2 is equivalent horizontal load giving the same deformation y2 etc ....

The total equivalent horizontal force is

H = H0 + H1 + H2 + H3 + (7-14)

If ki = Hi/Hi-1 is < 1, then the sum H will be finite (i.e. the structure is stable). With increasing
number of steps, in a linear analysis, the ratio will ki sooner or later become constant. In other
words, following terms will form a geometric series.

The simplest alternative is to assume that all terms, including H0, will form a geometric se-
ries. The total equivalent horizontal force is then obtained as

H0 H0
H= = (7-15)
1 k1 1 H 1 / H 0

which is equivalent to expression (H.8).

Note. Expression (7-15), including the definition of k, can also be expressed in terms of y or a relevant M.
It can also be shown that the final value of k is equal to the ratio V/VB, where V is the total vertical load
and VB is the global buckling load. Thus, the method of stepwise calculation can be seen as just a different
formulation or derivation of the method based on a magnification factor.

If the distribution of H1 is significantly different from that of H0, the accuracy can be im-
proved by including one or more steps:

H1 H2
H = H0 + , H = H 0 + H1 + etc. (7-16)
1-H 2 /H 1 1-H 3 /H 2
 46

The simple alternative (7-15)/(H.8) is sufficiently accurate in most cases, compared to other
uncertainties like the effect of stiffness variations within and between members due to crack-
ing etc. It should be observed that variation in the degree of cracking between first and fol-
lowing steps does not have to be considered, if reduced stiffness values according to 5.8.7.2
are used; these values are intended to be valid for the final stage of deformation. However, if
values for uncracked section are used in early steps, although cracking might occur in later
steps, then more steps have to be included in the analysis, like in expression (7-16); k-values
for early steps would otherwise be too low for later steps. This would apply generally in a
more refined analysis, where gradual cracking is taken into account.

When the structure is analysed for the equivalent horizontal force HEd, the relevant second
order effects can be obtained everywhere in the structure.

To magnify all moments with the same factor, as in expression (7-6)/(5.30), would not be
correct in for instance a frame or a shear wall with large openings.
 47

8. Method based on curvature (5.8.8)

8.1 Basic relationships

This method is basically the same as the so-called model column method in the ENV. The
second order moment is expressed in the following way, cf. equation (6-1):

1 l2
M2 = N y = N 0 (8-1)
r c

As mentioned in 6.5, 1/r is estimated on the basis of reaching yield strain in tensile and com-
pressive reinforcement. Here correction factors Kr and Kf is included:

1 1 2 yd
= K r K = K r K (8-2)
r r0 0,9d

In chapter 5.8 an extended definition of the effective depth d has been introduced, in order to
cover cases where there is no unambiguous definition of d; see figure 8-1.

h/2 h/2
d d
Figure 8-1. Definition of
is
effective depth in cross is
sections with reinforce-
ment distributed in direc-
tion of bending.

In order to reduce the curvature in cases where yielding is not reached in the tensile rein-
forcement, a factor Kr is introduced (same as K2 in ENV 1992-1-1, 4.3.5.6.3):

Kr = (nu - n) / (nu - nbal) 1 (8-3)

where
n = NEd / (Ac fcd), relative normal force ( in the ENV)
NEd is design value of normal force
nu =1+
nbal is value of n at maximum moment resistance; the value 0,4 may be used
= As fyd / (Ac fcd)
As is total area of reinforcement
Ac is area of concrete cross section

There is another factor K1 in ENV 1992-1-1, 4.3.5.6.3 (2), which reduces the curvature for
values of between 15 and 35. The purpose of this factor was presumably to avoid disconti-
nuity to cases where second order effects may be neglected. However, second order effects
 48

will often be neglected for between 25 and 35 (see 5.8.3.2), so discontinuities will still oc-
cur. Furthermore, independent of method, there is always a basic discontinuity following from
the rule that second order effects may be neglected if they are below a certain limit. For these
reasons, the factor K1 has not been included in 5.8.8.

The ENV gives no indication of how to take into account creep in the model column
method. Comparisons with the general method indicate that in certain cases the method can
give unsafe results without allowance for creep, and the factor K has been introduced for this
purpose. It has been calibrated against calculations with the general method.

More sophisticated models for estimating the curvature can be found in [2] and [3]. Their background is
presented in [1].

8.2 Comparison with general method and stiffness method

The result of calculations with curvature according to expressions (8-1) to (8-3) is presented
in Appendix 2, in comparison with calculations with the general method. In the same Appen-
dix calculations with the stiffness method (chapter 7) are also presented and compared.

8.3 Using the curvature method for structures

In 5.8.5 there is an indication that the curvature method can be used also for structures, with
proper assumptions concerning the distribution of curvature. This statement is based on [4],
where a method is given by which the curvature method can be used also for second order
analysis and design of unbraced frames. For details, see [4].
 49

9. Biaxial bending (5.8.9)

The general method is suitable also for biaxial bending. The same principles as in uniaxial
bending apply, although the complexity of the problem increases.

Simplified methods like the stiffness or curvature method can also be used. They are then
used separately for each direction, and if the resulting bending moments fulfil a certain crite-
rion, expression (5.38), no further action is necessary.

The criterion in (5.38) is similar to expressions (4.74) and (4.75) in the ENV, 4.3.5.6.4, but
there is one important difference: the ENV check concerns only first order eccentricities,
whereas in 5.8.9 it concerns total eccentricities including second order effects. The reason for
including the second order effects is illustrated in figure 9-1:

Areas where biaxial bending total eccentricities

may be neglected (including 2nd order)
total eccentricities
1st order ecc.
A B 1st order eccentr.

Figure 9-1. Example of member with different slenderness in the two directions.

Assume for example = 100 in one direction and = 20 in the other. Second order effects
will then be significant in one direction but negligible in the other. A and B are two examples
of the position of the axial load, both fulfilling the criterion for separate checks according to
the ENV, based on first order eccentricities. This would be acceptable for case A, since the
second order effect will make the total eccentricities even less biaxial. It is not acceptable
for case B, however, since the second order effect will now give total eccentricities outside
the permissible area. Thus, a first order criterion can be misleading and unsafe.

If criterion (5.38) is not fulfilled, the cross section should be designed for biaxial bending. A
simple model for this, in the absence of an accurate cross section analysis, is given in 6.1:

a a
Mx My
+ 1 (9-1)
M
Rx Ry
M

where Mx/y is design moment in the resp. direction, including nominal 2nd order moment
MRx/y is corresponding moment resistance of cross section
a is exponent

The values of the exponent a are taken from a UK proposal, in turn based on [5]. The expo-
nent has been slightly adjusted according to [6]. These values can be used in the absence of
more accurate values.
 50

10. Lateral instability of slender beams (5.9 11)

Compared to the ENV, the following changes have been made:
1. It is clearly stated that the check of lateral instability of beams is relevant in situations
where lateral bracing is lacking. For beams in finished structures, lateral instability is nor-
mally prevented by lateral bracing from adjacent members (e.g. floor or roof elements).
2. A lateral deflection l/300 has been introduced as an imperfection to be used in calculations
concerning lateral instability and balance at supports.
3. The criterion for neglecting second order effects is different. It is explained and compared
to the ENV below.

Expression (5.40) is based on a numerical study [7]. It is technically equivalent with the cor-
responding criterion in the new DIN 1045 [8], but it has a different mathematical formulation
to show the main parameters l/b and h/b more clearly.

Figure 10-1 shows a comparison according to [7] between the numerical results and expres-
sion (5.40). The corresponding criterion according to the ENV is also shown. It is quite clear
that the ENV criterion does not represent the numerical results very well; it is too conserva-
tive in many cases and unsafe in other cases. The DIN model is much better.

120
l/b Second order effect > 10 %
Second order effect < 10 %
110

DIN 1045 and

100 EN, Expr (5.46)

90

80

70

60

50
ENV
Figure 10-1. Cri-
teria for neglect- 40
ing second order
effects in beams 30

according to ENV
20
and EN in com-
parison with nu- 10
merical results
[7]. 0
0 1 2 3 4 5 h/b 6

In the final version, a distinction between persistent and transient design situations has been
introduced, together with an additional criterion for h/b; this is based on national comments.

11
5.8.10 in earlier drafts.
 51

11. References
[1] Westerberg, B: Design Methods for Slender Concrete Columns. Tyrns Technical Re-
port 1997:1. Stockholm, September 1997.

[2] FIP Recommendations, Practical Design of Structural Concrete. fib (CEB-FIP), Sep-
tember 1999.

[3] Design Handbook for High Performance Concrete Structures. Handbook published in
Sweden, 1999.

[4] Beeby, A W and Narayanan, R S: Designers Handbook to Eurocode 2, Part 1.1. Tho-
mas Telford, London, 1995.

[5] Bresler, B: Design Criteria for Reinforced Columns under Axial Load and Biaxial
Bending. ACI Journal, November 1960.

[6] Whittle, R T and Lawson, R: Biaxial bending with axial compression. An investigation
into the use of Bresler coefficients for determining the capacity of reinforced concrete
sections under combined axial compression and biaxial bending. March 2000

[7] Knig, G and Pauli, W: Nachweis der Kippstabilitt von schlanken Fertigteiltrgern
aus Stahlbeton und Spannbeton. Beton- und Stahlbetonbau 87 (1992)

[8] Hellesland, J: On column slenderness limits. Mechanics Division, University of Oslo,

1999-05-28
 52

Appendix 1. Verification of new model for slenderness limit

Variables covered:
Concrete grade: C20, C40, C80
Reinforcement ratio: = 0,1 and 0,5 (for C40 also = 0,3)
Effective creep ratio: ef = 0 and 2 (for C40 and = 0,3 also ef = 1)

Explanations to diagrams:
Horizontal axis: relative normal force n
Vertical axis: slenderness limit lim
Full curves: 10%-criterion, alternative 1
Dashed curves: 10%-criterion, alternative 2
Thick grey curves: new proposal for slenderness limit

Concrete C20

200 200
180 C20 ef = 0 = 0,1 180
C20 ef = 2 = 0,1
M 01/M 02 = - 0,9
160 160 M 01/M 02 = - 0,9

140 140

120 120

100 100
0
80 80 0

60 60
1 40
40 1
20 20

0 0
0,0 0,2 0,4 0,6 0,8 1,0 1,2 0,0 0,2 0,4 0,6 0,8 1,0 1,2

200 200
C20 ef = 0 = 0,5 C20 ef = 2 = 0,5
180 180
M 01/M 02 = - 0,9
160 160 M 01/M 02 = - 0,9

140 140

120 120

100 100
0
80 80
0
60 60

40 1 40 1
20 20

0 0
0,0 0,5 1,0 1,5 0,0 0,5 1,0 1,5
 53

Concrete C40

200 200
180 C40 ef = 0 = 0,3 180 C40 ef = 1 = 0,3

160 M 01/M 02 = - 0,9 160

M 01/M 02 = - 0,9
140 140

120 120

100 100
0 0
80 80

60 60

40 1 40 1
20 20
n n
0 0
0,00 0,50 1,00 1,50 0,00 0,50 1,00 1,50

200 200

180 C40 ef = 0 = 0,1 180 C40 ef = 2 = 0,1

160 M 01/M 02 = - 0,9 160

M 01/M 02 = - 0,9
140 140

120 120

100 100
0
80 80
0
60 60

40 1 40
1
20 20

0 0
0,0 0,2 0,4 0,6 0,8 1,0 1,2 0,0 0,2 0,4 0,6 0,8 1,0 1,2

200 200

180 C40 ef = 0 = 0,5 180 C40 ef = 2 = 0,5

160 M 01/M 02 = - 0,9 160 M 01/M 02 = - 0,9

140 140
120 120

100 100
0
80 80
0
60 60
40 1 40 1

20 20
0 0
0,0 0,5 1,0 1,5 0,0 0,5 1,0 1,5
 54

Concrete C80

200 200
180 C80 ef = 0 = 0,1 C80 ef = 2 = 0,1
180
160 M 01/M 02 = - 0,9 160

140 140

120 120
M 01/M 02 = - 0,9
100 100

80 0 80

60 60 0

40 40
1 1
20 20

0 0
0,0 0,2 0,4 0,6 0,8 1,0 1,2 0,0 0,2 0,4 0,6 0,8 1,0 1,2

200 200
180 C80 ef = 0 = 0,5 180 C80 ef = 2 = 0,5
160 160
M 01/M 02 = - 0,9
140 140 M 01/M 02 = - 0,9

120 120
100 0 100
80 80 0
60 60
1 1
40 40
20 20
0 0
0,0 0,5 1,0 1,5 0,0 0,5 1,0 1,5
 55

Appendix 2. Calibration of simplified methods

A2.1 Main calculations and results
Calculations have been made for isolated columns with the following variables:

Cross section Rectangular12

Reinforcement Symmetrical, concentrated to tensile and compressive
edges, edge distance 0,1h
First order moment Constant (equal end eccentricities)6
Boundary conditions Pin-ended
Slenderness ratio = l0 / i (0) 35 70 105 140
Relative eccentricity e0/h (0) 0,1 0,2 0,4 0,6 0,8 1,2 1,7 3,0 6,0
Reinforcement ratio = Asfyd/Acfcd 0,1 0,2 0,4
Concrete grade fck 20 40 80
Effective creep ratio ef 0 1 2

The total number of individual cases is 49333 = 972 (not including = 0 and e0 = 0).

For each case, calculations have been made with the general method according to 5.8.6 and
with the simplified methods, i.e. the stiffness and curvature methods according to 5.8.7 and
5.8.8 respectively. The results are summarized in table A2-1 on the following two pages.

The vertical axes in the diagrams represent the ratio

Maximum first order moment according to simplified method

Maximum first order moment according to general method

The moment ratio is given with the mean value m and with m+s and m-s respectively, where s
is standard deviation. The mean value (and standard deviation) for a certain value of the inde-
pendent variable includes all the values for the other variables. Thus, for one value of e.g. the
slenderness, the mean value and standard deviation of the moment ratio represent 972/4 = 243
individual values; for one value of the eccentricity 972/9 = 108 values etc.

The horizontal axes represent the main variables: slenderness , eccentricity e0/h, reinforce-
ment ratio (both and ), concrete strength fck and effective creep ratio ef.

Interaction diagrams have also been prepared, covering all the above-mentioned cases and
including the different methods and alternatives. However, to present all these diagrams
would require too much space, and it would be difficult to obtain an overall view of the re-
sults. Therefore, only two such diagrams will be shown to illustrate certain aspects.

12
In [1] the simplified methods are compared with the general method also for columns with circular cross sec-
tion and with a different distribution of the first order moment. Although [1] deals with more refined versions of
the stiffness and curvature methods, the main conclusion is that the same models, as for rectangular section and
constant moment, can be used also for other cross sections and variations of the first order moment.
 56

Table A2-1. Summary of comparisons between simplified methods and general method.
Variable

Stiffness method, 5.8.7 Curvature method, 5.8.8

Expression (5.26) Expression (5.22) Expr. (5.31-5.37)
(simplified alternative) (more accurate alternative) ------- With K = 1*)
1,3 1,3 1,3
M0,simplified/M0,general
1,2 1,2 1,2
m+s
1,1 1,1 1,1
1. Slenderness

1,0 1,0 1,0

0,9 0,9 0,9
m
0,8 0,8 0,8
0,7 0,7 0,7
0,6 0,6 0,6
m-s
0,5 0,5 0,5
0,4 0,4 0,4
35 70 105 140 35 70 105 140 35 70 105 140

1,4 1,4 1,4

M0,simplified/M0,general
1,2 1,2 1,2
m+s
1,0 1,0 1,0
2. Eccentricity

0,8 0,8 0,8

m
0,6 0,6 0,6

0,4 0,4 0,4

0,2 m-s 0,2 0,2

0,0 0,0 0,0

10,0 1,0 e0/h 0,1 10,0 1,0 0,1 10,0 1,0 0,1

1,3 1,3 1,3

3. Mechanical reinforcement

M0,simplified/M0,general
1,2 1,2 1,2
m+s
1,1 1,1 1,1

1,0 1,0 1,0

0,9 0,9 0,9

i

m
0,8 0,8 0,8

0,7 0,7 0,7

m-s
0,6 0,6 0,6

0,5 0,5 0,5

0,1 0,2 0,3 0,4 0,1 0,2 0,3 0,4 0,1 0,2 0,3 0,4

*) In other words equivalent with the ENV

 57

Table A2-1, continued.

Variable

Stiffness method, 5.8.7 Curvature method, 5.8.8

Expression (5.26) Expression (5.22) Expr. (5.31-5.37)
(simplified alternative) (more accurate alternative) ------- With K = 1*)
1,4 1,4 1,4
M0,simplified/M0,general
4. Geometrical reinf. ratio

1,3 1,3 1,3

1,2 1,2 1,2
1,1 1,1 1,1
m+s
1,0 1,0 1,0
0,9 0,9 0,9
0,8 0,8 0,8
m
0,7 0,7 0,7
0,6 0,6 0,6
m-s
0,5 0,5 0,5
0,4 0,4 0,4
0 0,01 0,02 0,03 0,04 0 0,01 0,02 0,03 0,04 0,05 0 0,01 0,02 0,03 0,04 0,05

1,3 1,3 1,3

M0,simplified/M0,general
1,2 1,2 1,2
m+s
5. Concrete strength

1,1 1,1 1,1

1,0 1,0 1,0

0,9 0,9 0,9

m
0,8 0,8 0,8

0,7 0,7 0,7

0,6 m-s 0,6 0,6

0,5 0,5 0,5

20 40 60 fck 80 20 40 60 80 20 40 60 80

1,3 1,3 1,3

M0,simplified/M0,general
6. Effective creep ratio

1,2 1,2 1,2

m+s
1,1 1,1 1,1
1,0 1,0 1,0
0,9 0,9 0,9
m
0,8 0,8 0,8
0,7 0,7 0,7
m-s
0,6 0,6 0,6
0,5 0,5 0,5
0 1 ef 2 0 1 2 0 1 2

*) In other words equivalent with the ENV

 58

A2.2 Discussion

All the simplified methods show a rather wide scatter when compared to the general method.
This is inevitable; a method giving close agreement with the general method over a wide
range of parameter values would no longer be simple.

An illustration is given in figure A2-1. The curvature method gives reasonable results for
low to moderate slenderness, but becomes extremely conservative for high slenderness ratios.
This is because the factor Kr (K2 in ENV) gives no reduction of the curvature at all when n <
0,4, and for high values n is practically always < 0,4. The same is true for the simplest ver-
sion of the stiffness method (expr. 5.26). With correction of the stiffness for normal force and
slenderness, expression (5.22) to (5.24), the result is much improved.

It is difficult to calibrate a simple method so that it accurately follows the general method,
particularly for high slenderness ratios and small eccentricities. This is true for all methods,
see the first and second rows of diagrams in table A2-1.

0,2
m0
=0
0,18
35
0,16

0,14

0,12
70
0,1

0,08
105
0,06

0,04

0,02
140 n
0
0 0,2 0,4 0,6 0,8 1 1,2

Figure A2-1. Example of interaction curves. fck = 40, = 0,2, ef = 0.

general method
stiffness method, expressions (5.22)-(5.24)
stiffness method, expression (5.26)
curvature method, expressions (5.31)-(5.37)
 59

In figure A2-1 there is no effect of creep (ef = 0). Figure A2-2 shows the corresponding
curves for ef = 2 (a comparatively high value). Two sets of curves are given for the curvature
method. The upper curves are based on K = 1, corresponding to the method in ENV 1992-1-
1, 4.3.5.6.3, where there is no effect of creep. The lower curves are based on K according to
expression (5.37).

Without effect of creep (= ENV), the curvature method is consistently unsafe for low and
moderate slenderness. This can be seen also in the third column of diagrams in table A2-1.
With K according to expression (5.37), creep is well taken into account.

0,2
m0
=0
0,18
35
0,16

0,14

0,12

0,1
70
0,08
105
0,06
K = 1
0,04 K accord.
to (5.43)
0,02
140 n
0
0 0,2 0,4 0,6 0,8 1 1,2

Figure A2-2. Example of interaction curves. fck = 40, = 0,2, ef =2.

general method
stiffness method, expressions (5.22)-(5.24)
curvature method, expressions (5.31)-(5.36) (upper curve, K = 1)
curvature method, expressions (5.31)-(5.37) (lower curve, K according to (5.43))*)
* For = 105 and 140, (5.43) gives K = 1, therefore there is only one curve in these cases.
)