Structural Analysis: Elastic or Plastic?
Johan Blaauwendraad
Prot. Dr
Delft University of Technology
Delft, The Netherlands
Johan Blaauwendraad, born in 1940, re
ceived his civil engineering degree in 1962
and his doctorate in 1973, both from Delft
University of Technology (TU Deli). He
joined the Netherlands Organisation for
‘Applied Scientific Research (TNO) in
1964 and Rijkswaterstaat in 1971. He has
been professor at TU Delft since 1979,
Jorg Schneider
Prot.
Federal Inst. of Technology
Zurich, Switzerland
Born in 1934, Jorg Schneider received his
civil engineering degree from the Federal
Inst. of Technology (ETH) Zurich. He was
aan assistant at the ETH from 1959 to 1963,
and then worked with Stahltor/BBRV in
Zurich. Since 1967 he has been professor
‘of structural engineering at the ETH. His
research interests include safety and relia-
bility, with emphasis on human error.
A
‘Structural Engineering International
Peer-reviewed by international
‘experts and accepted by the
Publications Committee
393
Summary
‘The misuse of computer analysis in structural design is an issue that needs to be
faced. With the wide application of “solution” - oriented elastic analysis comput-
er software, thirty years of research in plastic
y theory and the knowledge
gained from experience can be ignored at the push of a button, resulting in struc-
lures of questionable quality. The paper proposes an interactive hybrid design
analysis process that exploits the strengths of both elastic analysis and plasticity
theory and calls for critical reflection and active decision-making from the user.
Introduction
Until the early 1960s, elastic analysis
was the only method available for ana-
lysing statically indeterminate struc-
tures. Later, plastic analysis also found
its way into engineering practice. Al-
though only applicable to the analysis,
of the ultimate limit state of structures,
plastic analysis proved to be a powei
ful tool, simple to use and especially
effective regarding the upper and low-
er bound theorems of the theory of
plasticity.
Plastic analysis has also demonstrated
its value in the design of slabs. A nota-
ble advantage of the plastic analysis
approach is the explanation it provides
for the relative unimportance of local
moment peaks in slabs and similar
structures.
‘Through the 1970s and into the 1980s,
‘mainframe and then desktop comput-
crs entered the engineer's toolbox. A
multitude of computer programmes
for the analysis of beams, frames, grids,
plates, shells and other three-
dimensional structures have been de:
veloped. Finite element methods
(FEM) form the basis of such pro-
grammes. In this easy-to-use form,
FEM found its way from the scientist's
laboratory to the practising designer's
desk. Today, the analysis of even very
complicated structures does not
present much of a problem for any av-
erage engineer, as long.as the right PC
tools are on hand.
The Problem
Most design analysis programmes at
the disposal of engineers are elastic in
the sense that they make use of elastic
FEMs. Such programmes are inherent
ly easier to write because they are
solution-based in their basic formula-
tion, Analysing structures with such
tools is a simple undertaking for any-
one sitting behind a PC: just type in
the data, and out comes the elastic s0-
lution. Presto, the structural analysis is
completed.
The problem is that too little thought
is given to this instant solution, Wheth-
er the impressive-looking output really
is the optimal basis for subsequent di-
mensioning, detailing and execution of
the structure is a question too rarely
asked. Just put the steel where the
FEM elastic analysis indicates, that's
all. Incredible but true, instances have
been observed where manifestly over-
stressed finite elements of a concrete
slab have been reinforced just because
the computer output showed high
bending moment peaks. The gluing of
steel plates to the upper and lower
sides of the concrete slab was some-
how seen to increase bending strength.
The Proposal
To rectify such falsely placed reliance
on computer-derived “solutions”, a
new design process must be estab-
lished. The goal is a process that ex-
Science and Technology 165ploits both the strength of plasticity
theory as well as the power and ease of
computers and, importantly, stimulates
the user really to act as a designer.
This design tool supports an interac-
tive design process demanding much
more reflection than is the norm today.
This new type of interactive pro-
‘gramme, be it for framed structures or
slabs, can be called “hybrid” in order
to express its use of both elasticity the-
ory and plasticity consid
‘An Example
A three-span steel or reinforced con-
a sm is considered (Fig. 1) with
constant rigidity over its entire length
and uniformly distributed toads on
each of the three spans. The focus is on
the bending moments in the beam in
order to determine dimensioning and
detailing.
Fig, I: Sructural system
‘The bending moments M, and Mp at
the ends of the beam are zero. At the
two inner supports, negative moments,
M, and Me can be expected. A relative
rotation gq is defined between the
beam ends at support B, and similarly
‘cat support C.
In a fully elastic solution, a require-
ment is imposed on gy and g. Both
are zero and as a result the analysis
yields an outcome for the bending mo-
‘ments My and Mc. The purpose of this
analysis is to determine whether My
and Me do not exceed a given strength
criterion,
Ina fully plastic approach, a priori val-
ues for the plastic moments My and
Mc would be expected and the analysis
would result in a value for the relative
rotations gy and ge, the distortions in
the plastic hinges. It must be deter-
mined whether gy and ¢ meet a given
deformation criterion (rotation capaci-
9),
In the proposed hybrid method, the
designer decides which solution he
wants. It can be the one or the other,
but in most cases a mix of both solu:
tions would likely be preferred. This
requires establishing a relation be-
166 Science and Technology
tween the two bending moments My
and Mo, the two distortions ¢ and ge,
and the loads q,, q> and q; on the three
spans. Initially, only the loads are
known and the bending moments and
distortions are unknown. It is the role
of the hybrid program to produce the
relations for the designer,
In this simple example, use is made of
the well-known method introduced by
Clapeyron in 1830. Clapeyron’s equa-
tions, based on a statically determinate
primary system with two hinges at the
inner supports B and C provides the
two equations which the designer
needs. Using m and kN, the Clapeyron
equations for this example are:
Ely = 13.333 My + 5.000 Me +
41.667 q, + 1125.0 q2
Ele = 5.000 My + 16.667 Me +
1125.0 q; + 333.33 qy
‘These equations may be arranged so
that the known values are assembled
in the right-hand members. The left-
hand members contain the four design
parameters to be determined.
Eligy ~ 13.333 My ~ 5.000 Mc
41.667 q, + 1125.0 q,
Elyge~ 5.000 My ~ 16.667 Mc
1125.0 qs + 333.33 gy
Thus, the two equations contain four
unknowns. Two design parameters
must be determined in order to solve
the other two. In the hybrid metho
the designer is called upon to do this
demanding task and not the software.
A possible sequence would be that the
designer first wants to see the clastic
solution. So he starts with @, = 0 and
5 = 0. Solving the equations for My
‘and Me provides the outcome:
M, =~35211 q, 66.549 g, + 84507 4,
Mc= + 1.0563 q, 47.535 q, ~22.535 q,
Bending moment
Load case iB momenis
Mp,
-1310. -1357
=2696 -2426
1046-1876
1232-1380
2588-2899
2774-2403
2511-2922
= 1124-1853,
Fig, 2: Elastic moments at B and C
From these expressions, the bending
‘moments at the two intermediate sup-
ports B and C for all relevant load cas-
es can be derived. The respective re-
sults are shown in Fig. 2 based on the
following factored loads to comply
with the ultimate limit state
=20kNim,
42 kNim.
sin
oo
Fig. 3 shows the bending moments dia-
grams for some of the above load
cases.
= 2696 2922
2 7
2
Fig, 3: Elastic moment diagrams
It is obvious that a design engineer
would choose load case 7 in Fig. 2 in
order to decide on the dimensions of
the beam and to fix the reinforcement
in section C, while he would derive
from load cases 2 and 8 the reinforce-
‘ment at the bottom of the beam in
spans 2 and 3, respectively.
But, here is the problem: Would this
be an optimal solution to the problem
at hand? Sometimes the answer is
clearly affirmative. But in many cases
it would be better to make use of what
has been learned from plasticity theo-
1. It is sufficient to assume some rota-
tional capacity for the beam and thus,
to limit the bending moment in section
C to a conveniently lower value (in an
absolute sense):
Men 2775 kNm.
‘The question now is how the bending
moments in the rest of the beam are
influenced by this decision.
‘The two preselected design parame-
ters now are:
oe =O
Mc Mos
‘The outcome is a value for the design
parameters g and My. The distortion
is gos Which is the rotation capacity
requirement. The two Clapeyron
‘equations indicate that:
Ms 125 gy ~ 84.375 as
0.375 Meu
Elegcq, == 15.625 qy + 703.13 ga +
333.33 qy + 14.792 Meg
Structural Engineering International 93From these, M, must be checked
against a strength criterion (My <
May) and gq Ought to be checked
against the available rotational capa
ty of the beam at section C. Taking
load case 5 in Fig. 2 as an example, for
instance:
2634 kNm
827 kNm?
Likewise, the choice could be instead
to limit the bending moment at section.
B. The respective equations, assuming
= except at section B ~ elastic behavi-
‘our of the beam, are as follows:
Me =~675 q;~20.000q3-
0.300 May
Elen, =41.667 q, +7875 q>—
100,000 q, + 11.833 Moy,
It is even possible to go further and.
limit the bending moments at both
supports. This results obviously in two
rotational capacity requirements:
El gyq = 41.667 q, + 1125.0 q. +
13.333 Mg + 5.000 Mess
El Gey. = 1125.0 qs + 333.33 q3 +
5.000 Muy + 16.667 Mey
‘As an example, both bending mo-
ments are set to
Mug = Meni =~ 2400 kNm,
For load case 2 this results in rotation-
al capacity requirements of
Elgg = 4083 kNm?
and
Ellgcy= 1917 kN,
respectively.
‘The corresponding bending moment
ram is shown in Fig. 4 together
with the diagram related to load case
7. It is obvious that these diagrams al-
low for smaller dimensions of the
‘beam and might well be more econom-
ical with respect to the reinforcement
than the elastic solutions shown in Fig
2. Also shown in Fig. 4 is the bending
‘moment diagram for load case 8 which
still remains in the elastic range, as do
load cases 1, 3 and 4, not shown,
It should be noted that at the interme-
diate supports B and C there is either a
bending moment in agreement with
elastic analysis or a rotational capacity
requirement. There are as many free
choices of bending moments in the
structure as the degree of statical inde
terminacy. In accordance with what is
known from plasticity theory, there
Structural Engineering International 393
2325
Fig. 4: “Hybrid” moment diagrams
might, however, be local mechanisms
that in a given area restrict the choice
to what is acceptable from the equilib-
rium conditions,
‘This is the way an engineer would han-
dle the problem, but, obviously, he is
lacking the corresponding software
support, since many, if not all, availa-
ble programmes provide the elastic so
lution only, leaving the engineer with
heaps of data that do not really help
much to reach a good design.
Slabs
What has been explained for a simple
three-span beam can be applied to
slabs as well. Again, a very simple ex-
ample is chosen, and just the proce-
dure is discussed,
‘The example considered is a flat slab
consisting of square parts supported
by columns. The slab is expected to be
t
a
Fig. 5: Slab system
of constant thickness and has infinite
length in two orthogonal directions.
‘A constant uniformly distributed load
iis applied. There is no difference be-
tween the reinforcement in both dire
tions. This means that the system lines
through the columns are lines of sym-
metry, as are the lines through the cen-
tres of the slab fields (Fig. 5). Thus,
analysis can be restricted to a quarter
of a square slab part (the shaded area
in Fig. 5).
The shaded part is shown again on a
larger scale in Fig. 6. The (quarter of
the) column is in corner A. The centre
of the slab part is in comer D. The di
Fig. 6: Moments in slab
‘Stience and Technology 167mensions of the column are realistical-
ly taken into account. The load q is ap-
plied over the square ABCD.
‘The designer starts to be interested in
the elastic solution. Such a solution
can be found applying elastic FEM
procedures, The result is sketched in
Fig. 6 for the moment ma, in all four
lines of symmetry (thin lines).
‘This result, in a conventional ap-
proach, would be used to determine
the reinforcement in the x-direction,
because no twisting moments occur in
lines of symmetry. Of course, the same
reinforcement is needed in the y-
direction. It can be seen that a high
peak occurs at the position of the col-
A next step for a capable designer
would be to choose a practical rein-
forcement system, limiting the bending
moments in some defined region
around the corner. In the example, the
designer might choose a value which is,
660 % of the elastic peak value.
Assuming that the designer has a hy:
brid programme at his disposal to
make a plastic analysis, he would find a
new distribution of bending moments
reflecting the choice of the reinforce-
ment. The result is sketched in Fig. 6,
in the thick shaded lines.
It can be seen that a strong reduction
of the peak bending moment on top of,
the column hardly influences the
bending moment distribution at some
distance from the column. In slabs, itis
hard to speak about a distinct plastic
rotation ¢ as for a beam. Rather, the
notion of plastic curvature should be
used as a deformation criterion,
A further decision could be made
specifying a limit value m, close to the
168 Science and Technology
column and a lower value ms in the re-
maining part of the slab. In all cases, it
must be determined that principal mo-
ments do not occur anywhere in the
slab which surpass the limiting values
given by the reinforcement.
This simple example has been chosen
only to show that the designer is again
called upon to make choices and deci-
sions. Only then should the software
analysis proceed, defining the conse-
quences of these decisions as a help for
further improvements.
The Development Challenge
‘The choice elements that have been
demonstrated for statically indetermi-
nate structures should be incorporated
into desktop computer programmes
which would then work interactively.
The engineer should be able
~ to input the geometrical configura-
tion and sectional properties of a
structure, 1o define the load case to
be analysed, and to see the corre-
sponding elastic bending moment
diagram
to critically study the diagram and,
if felt appropriate, make changes
somewhere to the obtained elastic
solution, The engineer should. be
able to apply a limit to the bending
‘moment at any point in the struc
ture and see the resulting changed
bending moment diagram and rota-
tional capacity requirements.
= to then proceed, starting from what
had been obtained, by limiting the
bending moments at some other
points in the structure or return to
some previous decisions. In such a
way, slowly, a preferable basis for
dimensioning, detailing and execu-
tion could be interactively obtained
by this hybrid procedure.
Clearly, this procedure must be fol-
lowed for all load cases relevant to
proper dimensioning of the structure.
Conclusion
Software developers are challenged to
provide designers with interactive hy-
brid programmes. Programming. spe-
cities are not relevant. What is hidden
within the programme does not mat-
ter. Itean be linear or nonlinear, based
‘on ineremental loading. The only re-
quirements for the designer are free-
dom of choice among clearly shown al-
ternatives, speed, ease of use and, of
course, reliable and complete informa-
tion.
It is acknowledged that serviceability
problems generally cannot be dealt
with by plastic methods, but are better
considered by basing dimensioning
and detailing on elastic solutions. Thus,
care should be taken when designing
bridges and other structures exposed
to environmental loads, or structures
where deflections, ete. are of prime im-
portance. Comments above on devia-
tions from the elastic solution taking
account of the rotational capacities of
beams or slabs should not be applied
to such structures,
The proposed hybrid approach to
structural analysis relates to problems
in which the bearing capacity of a
structure and the identification of eco-
nomical solutions are primary. con-
cerns Without doubt, these considera-
tions involve the great majority of all
structures.
Structural Engineering International 93