Professional Documents
Culture Documents
1. Basic objectives
When modelling and analysing a structure we try to achieve the following:
Ensure that the structure as a whole will be stable under load (no global buckling,
overturning, etc).
Ensure that the structure will behave adequately under load (no excessive deflections,
vibrations, etc).
Obtain the forces to be used in the design of the elements and connections. The strength
and stability of the members have, of course, a profound effect on the stability of the
whole structure.
Ultimately, of course, modelling, analysis, member sizing and design of connections are all
part of the process of obtaining a structure that meet all the requirements under all the
conditions it may be subjected to, at minimum cost (however we may define ‘cost’).
1
Figure 1 – Floor bracing
The factors that can affect the stability of a structure are listed in the following two tables:
Table 1 –Physical attributes of structure and loading affecting steel frame stability
2
Table 2 – Modelling parameters and behavioural assumptions affecting steel frame stability
The range of things that can play a role is clearly huge. Some of these are so intractable that
we have to play safe with loads, load factors and resistance factors. Others are addressed
in the process of member and connection design. Many are handled by a ‘forgiving’ structure
that settles into its own state of equilibrium, deformation and stress, regardless of what the
calculations wanted these to be (and sometimes by leaning on non-structural elements). An
important factor is that we assume that the structure is built within certain parameters.
SANS 10162-1 assumes that the structure as built will meet the stipulations of SANS 2001-
CS1.
For the purpose of this chapter we distinguish between four different approaches to
structural analysis: first order elastic, second order elastic, first order inelastic, and second
order inelastic. How these approaches yield different results is illustrated in Figure 2.
First order elastic analysis assumes the material stays elastic, whatever the stresses, and
the deflections of the structure are so small that they won’t affect the moments in the
members. The only limit on the resistance of the structure is the elastic stability limit, when
the structure buckles elastically, similar to the Euler buckling of a strut.
With second order elastic analysis the structure is assumed to remain elastic, but
displacements cause an increase in the moments and in deflections. The subject is
discussed in some detail in 3. below. The structure will reach a limit similar to (though not
identical to) the elastic stability limit.
3
Under ‘first order inelastic’ analysis we understand an approach in which the plasticity of
the material is taken into account, but the deflections are assumed to be negligibly small.
This is done in Plastic Analysis, discussed under 4. below.
Finally, we can do a second order inelastic analysis, in which both plasticity and
displacements are taken into account. This gets the ‘real’ ultimate load of the structure, the
‘inelastic limit’ shown in Figure 2.
After covering (though not exhausting) the complex subject of the behaviour of structures
in the region of the ultimate limit state in the following sections, we will return to practical,
design office analysis in 5.
Here we consider the second order effects resulting from the famous P and P effects.
Consider the column in Figure 3:
From a linear elastic analysis we will only get that the bending moment at the bottom will
be equal to HL . However, the force H causes a deflection and thus the force P causes
an additional moment P . As the column bends, it will deflect by an amount from a
straight line between the top and the bottom, and that will cause a moment P . Thus the
actual moment diagram will look as shown in Figure 3(e). The deflection at the top will also
be more than what a linear elastic analysis would say.
Figure 4 shows a cantilever column at the right, plus three other columns that are hinged
both top and bottom. All the columns are connected by beams with pinned ends and each
column carries a load P . The force H causes a deflection , but now all of the P -forces
are displaced, causing a moment 4P at the bottom of the cantilever. Thus the lean-to
columns also need to be taken into account.
4
Figure 4 – Lean-to columns
Note that if the structure was not erected perfectly vertically, the initial deflection will yield
a P bending moment even if H 0 . Equally, if the column is not perfectly straight there
will be a P moment along its length even if H 0 . To limit these values (and for other
reasons, of course) SANS 2001:CS1 specifies that must be less than 1/500 times the
height of the column (and not more than 5 mm), and less than 1/1000 times the length
(and not more than 3 mm).
The first measure SANS 10162-1 specifies in clause 8.4.1 to deal with these effects is that
notional loads equal to 0,005 times the factored gravity loads on each storey must be applied
horizontally at the level of the storey (this includes the roof). This causes a deflection which
simulates any out-of-plumbness of the structure, but it is clearly much higher than what is
required for the P effect. The reason is that it covers more sins:
It allows us to get a reasonable factor of safety in those cases where the vertical load is
high and the horizontal load very small, such as for a tank on a stand.
It allows beam-columns to be designed more correctly (much more about this later in
the course).
It compensates for the inaccuracies that sneak in because of modelling errors,
deformations in joints, etc.
Note that the notional loads must be applied in every loadcase, and in all directions in which
instability needs to be considered. Normally, it makes sense to apply the notional load in
the same direction as the wind load. Special thought should be given to cases where
torsional instability can occur, for example if the core of a building is eccentric, or collapse
can occur by rotation around a central core.
Consider the beam in Figure 5. Both ends are restrained against rotation. The shear force
V causes, according to linear elastic analysis, a deflection 0 and a moment VL / 2 at each
end. The axial force P will, as a result of the deflection, cause an additional moment P / 2
at each end.
5
Figure 5 – Beam under lateral displacement and axial force
The sum of the moments at the ends of the beam is equal to:
M A M B VL P0
P 0
1 0
VL
And this additional deflection 1 will cause an additional deflection equal to:
P1 P 0
2
2 0 0
VL VL
P 0 P 0 2
n 0 1 2 ... 0 1 ...
VL VL
Which, in the limit, is identically equal to:
0
U 20
P 0
1
VL
Note that this only works when:
P 0
0, 7
VL
6
The amplification factor U 2 can now be applied to the moments in the beam. Note that the
shear force in the beam remains equal to V . Note also that 0 / L , if applied to a column,
is equal to the storey drift.
We find the amplification factor U 2 in clause 8.7 of SANS 10162-1. To apply it, you have to
first provide lateral supports to the structure to prevent sway, then apply all the gravity
loads, and determine the member moments M ug and the reactions at the lateral supports.
Then you remove the lateral supports and apply the lateral loads (including the notional
loads) as well as loads equal to but opposite in direction to the reactions calculated in the
first step, and calculate the moments M ut . Calculate U 2 . The final moments are given by:
M u M ug U 2 M ut
If we define:
P
u
PE
L3V
12 EI
EI
PE 2 2
L
We can write:
1
U2
1 u 2 /12
Note that the shear forces are not amplified: equilibrium requires them to remain equal to
V . What happens to the axial forces in a structure is more complex than can be handled
by means of a simple factor.
The differential equation governing the deflection of a prismatic beam with an axial load P
but without lateral loads along its span is:
d4y P d2y
0
dx 4 EI dx 2
The general solution of this equation is:
7
ux ux
y A1Sin A2Cos A3 x A4
L L
Where
P P
uL
EI PE
EI
PE 2
L2
By using the boundary conditions we can get the expressions for moment and shear at the
beam ends:
EI u 2 1 Cos(u )
M
L2 2 2Cos(u ) uSin(u )
EI u 3 Sin(u )
V
L3 2 2Cos(u ) uSin(u )
L(1 Cos(u ))
M V
uSin(u )
But V 2M e / L where M e is the linear elastic moment, thus we get:
2(1 Cos(u ))
M M e U 2*M e
uSin(u )
Thus the moments in the beam subject to axial load can be obtained by multiplying the
linear elastic moments by U 2* .
Note how very different this expression for U 2 is from that in SANS 10162-1 Annex B. And
yet, Figure 7 shows that they yield identical results over a range of values of u .
2.5
2.0
1.5
U2
1.0 U2*
0.5
0.0
0 0.5 1 1.5 2 2.5
8
Note further that u is a measure of the ratio of the axial force in the member to the Euler
buckling load. As the axial force in the member approaches the Euler buckling load, the
second order effect gets more pronounced.
Figure 8 shows that this approach amounts to amending the stiffness of the member in
compression. For linear elastic analysis we can say:
6EI EI
M 2
, so let’s say M 2
L L
The graph shows that, if u 0, 0 the value of 6, 0 , and it reduces to zero when .
Similarly, if:
EI
V Then 12, 0 when u 0, 0 and decreases with increasing u .
L3
14
12
Value of ω or ρ
10
8
6 ρ
4 ω
2
0
0 1 2 3 4
Value of u
Besides linear elastic analysis, Prokon allows two alternative settings: “Second Order” and
“Non-linear”.
The Second Order analysis is based on the stiffness adaptation approach just discussed. It
starts with a linear elastic analysis to determine the axial forces in all the members, and
then it calculates a new stiffness for each member, depending on the magnitude of the axial
force in the member. A new stiffness matrix is assembled for the whole structure and then
it is re-analysed, with no change in the geometry. This gives new member forces, a new
stiffness matrix and new results. The iteration is continued until the changes in the results
are minimal. The member forces are calculated using the amended stiffness matrix.
In the Non-Linear analysis the forces are applied in a predetermined number of steps. At
the beginning of each step a tangent stiffness is determined for the structure under that
condition of loading, taking geometric and material non-linearity (if specified) into account,
including the adaptation of the member stiffnesses as with the Second Order analysis, and
a Newton-Raphson iteration is performed as shown in Figure 9 to obtain equilibrium. This
approach can yield very accurate results, especially if several nodes are placed along the
9
length of each member, but it is certainly more refined than needed for ordinary steel
structures and will not be further discussed.
The question to be answered is how Prokon’s Second Order analysis compares with the
amplification factor approach of clause 8.7 of SANS 10162-1. Table 3 below shows, for a
column consisting of a circular hollow section (177,6x6), built in at the bottom and free to
move sideways but not to rotate at the top, for different lengths and values of the axial force
Cu , the value of the end moments, the deflection, and the value of the interaction equation
for beam-columns according to clause 8.13.
10
6 200 4 12 30.5 0.57 16.4 43.9 0.62 16.1 0.61 -0.9% -8.1%
6 300 4 12 30.5 0.83 20.4 56.3 0.94 19.4 0.93 -1.2% -11.7%
6 400 4 12 30.5 1.12 27.7 78.2 1.4 24.4 1.34 -4.2% -20.0%
6 500 4 12 30.5 1.48 44.1 128 2.43 32.9 2.10 -13.7% -39.1%
4 100 4 8 9.04 0.2 8.48 9.7 0.21 8.5 0.21 -0.7% -4.8%
4 200 4 8 9.04 0.35 9.05 10.5 0.36 9.0 0.36 0.7% -2.8%
4 300 4 8 9.04 0.5 9.7 11.4 0.52 9.6 0.52 -0.3% -3.8%
4 400 4 8 9.04 0.65 10.5 12.4 0.68 10.3 0.68 -0.4% -4.4%
4 500 4 8 9.04 0.81 11.4 13.7 0.84 11.1 0.84 0.1% -3.6%
Let us also discuss Prokon’s performance on the AISC benchmark problems. These
problems from the AISC Manual are given with results to allow engineers to check whether
the software they want to use for steel design yields the correct results. The results obtained
are listed below.
11
Prokon 5.0 5.7 6.6 7.9
The match is just about perfect. This is especially remarkable when considering that all the
members in these problems are loaded beyond their capacity as beam-columns unless they
are laterally supported along their lengths. Prokon thus meets the requirements set by AISC.
(Just for interest: AISC also shows the values for when shear deformations are taken into
account, and the deflections and moments are typically less than 1% higher than when
shear deformation is ignored. To take shear deformation into account in Prokon, enter a
value for the shear area, which is otherwise effectively taken as infinite.)
Because we are talking about non-linear behaviour, we cannot analyse the structure for the
basic loads for each loadcase and then proceed to apply the load factors and do the load
combinations and expect to get the right results. So Prokon proceeds to first calculate the
combinations of factored loads before doing a complete second order (or non-linear) analysis
for each load combination.
We mentioned the issue of lean-to columns above (see Figure 4 above). What this means is
that, if not all the columns on a level contribute to resisting the lateral forces, those that do
have to also account for the P effects of the others. With the amplification factor approach
in the code this is not a problem: the value Cu takes the axial loads in all the columns into
account. The Prokon approach also has no problem with this situation – the second order
stiffness matrix gives the correct values for the forces in all the members. The forces in the
columns that are pinned top and bottom have axial forces equal to the vertical load acting
on them, but there are small horizontal reactions at their bases, equal to the forces in the
connecting members. Note that, from a second order point of view, the horizontal reaction
at the bottom of the pin-ended column is not unrelated to the horizontal displacement at
the top, as it would be with elastic analysis (see Figure 10). Thus the leaning columns will
put more lateral load on the fixed-base column. The moment and shear force in the fixed-
base column are significantly larger, but the axial force only infinitesimally.
12
Figure 10 – Horizontal reaction in pinned element because of lateral deflection
In summary: Prokon’s second order analysis is a quick and easy approach to analyse a steel
structure as intended in SANS 10162-1. For very unusual and slender structures, a more
comprehensive analysis, taking material non-linearity and other effects into account, may
be called for.
13
Figure 12 – Bending up to full plasticity
For the elastic section we have:
bh3 I bh2
I ; Ze ; M y Ze f y
12 h/2 6
For the plastic case the moment is:
h h bh2 bh 2
Mp b fy fy f y Z pl f y ; Z pl 1,5Z e
2 2 4 4
From a study of the tables in the Red Book, we can say that, for an I-section, it is
approximately correct to say that Z pl 1,15Z e .
Despite the fact that we assume the material to exhibit bilinear behaviour, the moment-
rotation curve for the section will be nonlinear, as shown in Figure 12. For the purpose of
plastic analysis, however, we assume rigid-plastic behaviour, as shown in Figure 13.
Furthermore, we assume the region of plasticity to be infinitely short, although this is not
actually the case (see 4.2 below), so we talk about a plastic hinge (dotted line in Figure 13).
14
Figure 14 – Cantilever beam
The external virtual work is:
We P P L
And the internal virtual work is:
Wi M p
Mp Z pl f y
Letting We Wi we get: Fu
L L
Next we look at a built-in beam with a distributed load – Figure 16.
Looking at the mechanism, we can say that the internal virtual work is:
Wi 4 M p
And the external virtual work:
We w 0,5 L / 2
wL2
Thus the section required must have a plastic moment of M p . If we depended on the
16
elastic moment diagram, we would have needed a section with either a yield moment or a
plastic moment of wL2 / 12 , some 33% higher.
15
To illustrate a different matter, we place the central plastic hinge at a third point of the span
– see ‘wrong assumption’ in Figure 15. Now the internal virtual work is:
Wi M p ( / 2 1,5 ) 3M p
And the external virtual work:
1 L wL2
We wL
2 3 6
Thus the M p required equals wL2 / 18 , which is 12,5% less than the correct value. Thus, by
selecting the wrong mechanism, we overestimate the resistance of the structure. This
observation can be generalised:
Lower Bound Theorem (Static Limit): The resistance of a structure to an external force
calculated on the basis of an assumed distribution of stress that satisfied equilibrium and
the yield condition is always less than or equal to the true collapse load.
Next, we consider the structure in Figure 16. We must determine the value of F that will
cause collapse of the structure. The dimensions are in mm and M p 4500 N.mm. (It’s a
very small portal frame.)
Let us now see what happens if we apply the forces in a specific sequence. Firstly, we apply
the full 122 N vertical force and then we do a linear elastic analysis (see (c)) which tells us
that the moment in the centre of the span just reaches the plastic moment of the section,
4500 N.mm. Next we apply (only) a lateral force of 60,8 N to the frame, and get the moment
diagram in (d). Combining (c) and (d) yields (e) and we can see that we now have two plastic
hinges. Considering these now as real hinges, we apply a further 6.83 N, as shown in (f),
and if we combine this with (e) we get (g), with 3 plastic hinges. Applying 20,3 N to the
resulting structure yields (h), and combining this with (g) results in (i), with four hinges and
the forces and moments as determined by the virtual work method.
16
Figure 16 – Portal frame loaded to plastic collapse
According to the Upper Bound Theorem the collapse load we have determined by way of the
mechanism approach must be equal to or higher than the actual collapse load. However,
studying the moment diagram we see that the moment nowhere exceeds 4500 N.mm, so the
Lower Bound Theorem tells us that the collapse load we found is lower than or equal to the
real collapse load. We can conclude that we have determined the real collapse load.
The code allows plastic analysis and design (Clause 8.6). The requirements to be satisfied
are as follows:
The steel’s ultimate strength f u is at least f y / 0,85 , which will ensure that the hinge will
retain its resistance so that redistribution of moment can occur.
Plastic hinges are only allowed in Class 1 sections, as defined in Clause 11.2. These
sections are so stocky that they will not suffer local buckling, even if the rotation is
sufficient to be regarded as a plastic hinge.
Lateral buckling is prevented at plastic hinges by bracing, as specified in Clause 13.7,
and at additional points near hinges.
17
Web buckling is prevented by the provision of web stiffeners whenever a lateral load acts
at a hinge location.
Splices in members have a minimum bending resistance.
The structure will not be subject to impact or fatigue loading.
The shear force at a plastic hinge position may not exceed 0,55 tw hf y (see clause 13.4.1.2)
The serviceability limit state requirements must be satisfied.
The last requirement under 8.6 is: “the influence of inelastic deformation on the strength
of the structure is taken into account”. This points to the fact that plastic design does
not help us much when P effects are large.
In practice, plastic design is not often used, for the following reasons:
Few engineers are familiar with it.
With complex structures the solution becomes very complex.
Serviceability limit state requirements often control.
Residual stresses in the steel will increase the range of yielding much further, as can be
concluded from a study of Figure 18. Because of uneven cooling during production, the
section has residual stresses fres, with maximum compressive values at the edges of the
flanges. When an axial compressive stress (because of axial load or bending) is applied, the
combined stress can go to f y near the flange tips, and noting that whatever is plastic has a
value of E = 0,0, this means that the value of EI of the section is reduced. Residual stresses
18
don’t have an effect on Mp or Cy, but the value of My is reduced and the member’s resistance
to deformation and buckling will be reduced (more on this much later in this course).
Member imperfections (lack of straightness) tend not to be as big a factor as one might
expect, but they tend to exacerbate the effects of distributed plasticity.
The analysis and design procedure that will satisfy the code is as follows:
Determine that loads and load combinations according to SANS 10160, for both the
serviceability and ultimate limit state. Add to each loadcase the notional loads as
discussed under 3.1. Note: with second order analysis we cannot add together the effects
of different loads and loadcases – we have to calculate the full load that will act on the
structure and analyse the structure under the action of this load.
Do a second order elastic analysis of the structure under the influence of the load
combinations. One way of doing this is simply to select ‘second order’ under Settings in
Prokon or another programme that can do a satisfactory second order analysis (see 3.4
19
above). Another is to use the moment amplification factor U2 in clause 8.7 of the code
(see 3.2 above).
For a planar structure, make sure that the structure will not become unstable out of
its plane (by checking for stability and providing support or bracing when required).
Design (size and detail) all members and connections according to the code. Note that
we use the ultimate resistance (multiplied by ) of the members, for example the plastic
moment of a beam.
Check that the serviceability limit state requirements (deflection and vibration) are met.
Assure all other requirements are met – fire resistance, seismic, durability, etc.
In essence, this approach amounts to a second order elastic analysis with a first hinge limit.
As a consequence, we lose the advantages coming from redistribution of tresses – in the
case of the beam depicted in Figure 15, for example, we require M p wL2 /12 , rather than
M p wL2 /16 , an increase of 33%. In the case of the portal frame in Figure 16 we will need
to design for a moment of 5226 N.mm, rather than 4500 N.mm; an increase of 16%.
The second order inelastic curve can be regarded as the ‘real’ curve for the behaviour of the
structure. Second order elastic analysis with a first hinge limit gives us point X, where the
onset of plasticity causes the inelastic curve to deviate from the elastic one. The calculated
resistance is less than the actual one – the inelastic limit – and thus conservative. This is
true for any structure such as that shown in (b), or for almost any structure we encounter
in practice. However, for the structure shown in (c) we may not be that safe. Distributed
plasticity may cause the structure to start deviating from the elastic curve before the first
hinge forms, and the structure may collapse more quickly, after passing a reduced inelastic
limit along the dotted light green line. Note that this portal frame has a low value of H/L
(say less than 0,1), and a low level of static indeterminacy, so there is little opportunity for
redistribution of moments.
It may sound scary to know that there are structures for which the standard design
approach may falsely say that they are strong enough to resist the loads they are subjected
20
to, but the code contains no limit on or warning about the applicability of the method. The
reason is that it has been proven that actual structures designed according to this approach
are safe. Nevertheless, an engineer should not be happy to put a heavy weight associated
with small lateral loads (e.g. a silo) on a structure like that in Figure 19(c) – it’s just too easy
to visualise how it can come crashing down if only one thing goes slightly wrong.
The situation becomes a bit more complex if the structure is statically indeterminate, as in
(b), (c) and (d). We have two concerns:
i) We do not want the vertical forces to go down the bracing – it is more economical to let
them go down the columns and it would be unwise to design the columns for anything
less than the full vertical load, as one can imagine slippage in the connections of the
bracing, resulting in load being shed to the columns. The bracing is there purely to resist
the lateral loads. In Figure 10(b), (c) and (d) the vertical forces will cause compressive
forces in the bracing, thus resulting in either bigger or smaller bracing elements than if
the bracing does only the job of resisting the lateral loads.
ii) If cross bracing is used, we only want the bracing elements acting in tension to be active,
as slender members are good at resisting tension but poor in compression. (Of course,
the engineer is free to adopt an entirely different strategy than discussed here – it is but
only necessary to follow through once you have decided on an approach.)
Both concerns can be addressed by making the bracing statically determinate, by deleting
those bracing elements that will tend to go into compression. For example, in Figure 19(b)
or (c) we can delete any bracing element that goes from a top left to a bottom right corner.
21
In (d) we can delete one whole vertical row of bracing, but then the bracing may not be
sufficient. Each structure is then effectively changed into one looking like the one in (a). The
problem with this approach is that for a loadcase with the wind blowing from the other side
the bracing elements we kept will tend to be in compression.
Another option is to remove all the bracing and provide lateral supports to keep the structure
standing up, as shown in Figure 21(b). This structure can then be analysed under the
influence of all the loads to obtain the first set of member forces and the horizontal reactions.
Then the bracing elements we want to consider active can be put back and the lateral
supports removed, and forces equal and opposite to the reactions at these supports can be
imposed on the structure, to get a second set of member forces. The two sets of member
forces can then be combined to get the final result. This approach is acceptable, except for
the fact that with second order analysis we should actually not use superposition at all.
Note that vertical bracing may be required not only to support the structure under the action
of lateral forces, but also when it is subjected to gravity loads only, as illustrated in Figure
22. Even a symmetric structure may have a tendency to lean sideways if the loading is not
symmetric.
6. Advanced analysis
The following draft was submitted to the committee responsible for the 2019 version of S16,
the Canadian standard governing the design of steel structures:
22
(Note that what follows has no legal standing and may be entirely changed before the
updated version of the code is accepted by the committee. The only reason for including the
draft clause is to demonstrate in which direction thinking about the design of steel
structures is developing.)
Inelastic material behaviour of members and connections of the structure is permitted to be modelled in the
analysis for the purpose of determining member forces and deformations in the structure, including local
deformations in the modelled inelastic components (i.e. members and connections for which inelastic behaviour
is modelled in the analysis). Inherent uncertainty and variability in the resistance and stiffness properties
assigned to the inelastic components shall be accounted for to produce the most adverse effects on the
structure, including the inelastic components.
Note: Uncertainty and variability can be accounted for by using lower or upper bound resistance and stiffness
properties in the analysis, as appropriate.
The design of the structure shall satisfy all requirements of this Standard, except that the member ultimate
limit states that are modelled in the analysis need not satisfy the strength requirements of this Standard,
provided that:
a) the resistance and stiffness properties assigned to the modelled ultimate limit states are determined in
accordance with the provisions of this Standard;
b) the resistance and stiffness properties assigned to the modelled ultimate limits states account for the effects
of residual stresses and geometric imperfections in a manner consistent with this Standard;
c) the material properties governing the strength of the modelled ultimate limit states are reduced by
multiplying by a number no larger than the appropriate resistance factor specified in Clause 13.1; and
d) all ultimate limit states not modelled in the analysis satisfy the requirements of this Standard.
Components subject to inelastic deformations shall be designed to have sufficient deformation capacity to
accommodate without failure the expected inelastic deformation demand.
Inelastic analysis used for the design of a structure may be performed in accordance with the guidelines given
in Annex O or by using the plastic analysis method as described in Clause 8.3.3.
Inelastic analysis used for the seismic design of a structure shall be performed in accordance with Clause 27.12.
Rationale:
All the modifications / additions in Clause 8.3 are proposed in order to link to the new Annex O on advanced
analysis. These are technical changes.
23
The main observations to be made from this draft clause are the following:
The clause may be included in the code to allow and entirely different approach to steel
design: to analyse and design the structure without using Chapter 13 of the code which
defines the resistances of beams, columns etc.
The analysis the clause has in mind is of the most advanced nature, taking everything
into account: material nonlinearity, geometric nonlinearity, P and P effects,
residual stresses, inaccuracies in fabrication and erection, distributed plasticity, etc.
If the software being used or other factors make it impossible to take any factor, for
example buckling out of the plane of the structure, into account, that factor should be
dealt with in accordance with the code. Matters like connections have to satisfy the code
anyway.
The clause should be regarded as enabling direct analysis; the details of how it should
be employed will be found in Annex O, which still has to be drafted.
For the time being, and certainly for the coming 5 years, direct analysis is unlikely to make
its appearance in South African codes.
24