Sie sind auf Seite 1von 13

Available online at www.sciencedirect.

com

Computers and Structures 86 (2008) 1176–1188


www.elsevier.com/locate/compstruc

Finite element shakedown analysis of reinforced concrete 3D frames


Marialaura Malena a,*, Raffaele Casciaro b
a
Department of Structures, University ‘‘Roma Tre”, Rome, Italy
b
Department of Structures, University of Calabria, Rende (CS), Italy

Received 13 December 2006; accepted 21 December 2007


Available online 4 March 2008

Abstract

In the present paper, the method proposed in [Casciaro R, Garcea G. An iterative method for shakedown analysis, Comput Methods
Appl Mech Eng 2002;191:5761–92] for the shakedown analysis is used for the evaluation of the shakedown safety factor of reinforced
concrete 3D frames subjected to a combination of varying loads. The FEM discretization for 3D frames is obtained subdividing the
frame into rods and using a 3D beam finite element. Beam end-sections are used for checking plastic admissibility and for this reason
the formulation models the interaction between the axial force and biaxial bending by a piecewise linearization of the elastic domain, for
generic shape reinforced concrete sections. Concrete and steel are assumed to be elasto-plastic, and, for concrete, tensile resistance is
ignored. The algorithm is described in detail and some numerical results, that show the efficacy and effectiveness of the proposed method,
are reported and discussed.
Ó 2008 Elsevier Ltd. All rights reserved.

Keywords: Shakedown analysis; Failure surface; Biaxial bending

1. Introduction cases limit analysis does not provide a reliable safety eval-
uation, even if each possible combination of load contained
The possibility, offered by recent European codes, to in the load domain is considered separately. In fact a con-
design structures taking into account their nonlinear tinuous increase in plastic deformations, along successive
behavior has given great importance to elasto-plastic anal- plastically admissible load cycles, could lead to excessive
ysis in civil engineering practice. displacements or produce collapse due to the fatigue, for
If the external load can be assumed as monotonically values of load multiplier less than the theoretical one given
increasing, the safety factor of elasto-plastic structures by limit analysis. A further requirement has to be met in
can be evaluated efficiently by numerical implementations this case. The rising amount of plastic deformation needs
of the classical theorems of limit analysis. Many FEM to be confined to an initial phase after which the structure
codes for limit analysis, based on implementations of behavior is purely elastic.
path-following algorithms that recover the equilibrium This problem lies in the classic and well established
path of the structures, are currently available. For this rea- shakedown theory, and is governed by Melan’s and Koit-
son checking the collapse safety of structures could be con- er’s theorems [3,4]. However, the lack of effective computa-
sidered an easy task. tional procedures for shakedown analysis, that can be used
Nevertheless, the structures are generally subject to var- in FEM codes, has limited the spread of safety check tools
iable cyclic actions, often only described by their overall based on shakedown in professional fields.
envelope domain, as in the case of seismic action. In these An iterative method for shakedown analysis has been
recently proposed in [1,2]. It is based on an iterative scheme
*
Corresponding author.
similar to that used in path-following procedures for limit
E-mail addresses: mmalena@uniroma3.it (M. Malena), rcasciaro@ analysis and proves to be both numerically efficient and
labmec.unical.it (R. Casciaro). well suited to FEM implementations. In [5], the method

0045-7949/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compstruc.2007.12.001
M. Malena, R. Casciaro / Computers and Structures 86 (2008) 1176–1188 1177

was applied to the analysis of two-dimensional structures allowing for the structure shakedown. By referring to Mel-
in Mises material. In the present paper it is implemented an’s theorem, the largest multiplier for the stress domain S
for the analysis of reinforced concrete three-dimensional allowing the sum r½t :¼ re ½t þ r
 of the elastic stress solu-
frames. tion and a self-equilibrated stress field r to lie within the
The frame is discretized in beam elements and plastic elastic domain E:
admissibility is checked on the beam end-section by using
ks :¼ maxr fk : kre ½t þ r
 2 E 8re ½t 2 Sg;
a multi-surface linearization of the cross-section failure
domain for biaxial bending and axial force. Twenty-six lin-  ¼ self-equilibrated
r ð3Þ
ear surfaces are used for each section, that are obtained with kS :¼ fkre ½t : re ½t 2 Sg being the amplified elastic
automatically from concrete section geometry and the envelope, we have to find the translation r  of kS that al-
associated reinforcement bars by assuming 26 different lows the maximum amplification of kS under the con-
plastic mechanisms. Note that this model only account straint that the resulting set belongs to the elastic domain
for flexural failures, and we implicitly assume that non-flex- E :¼ fr : f ½r 6 0g; f ½r being a convex yield function (see
ural types of failure are prevented through appropriate Fig. 1).
transverse reinforcements.
This paper begins with the presentation, in Section 2, of 2.2. Shakedown admissible domain
the basic assumptions used in shakedown analysis theory
and with the description of the iterative method proposed As suggested in [1], Eq. (3) is conveniently rewritten in
in [1] for shakedown analysis; in Section 3 the FEM model the form
used is described; in Section 4 the approximate representa-
tion of elastic domain for generic shape reinforced concrete ks :¼ max fk : 9
r 2 Ek g;  ¼ self-equilibrated;
r ð4Þ
sections is presented; in Section 5 the shakedown admissi- where
ble domain is described, Section 6 regards the return map-
ping scheme; finally in Section 7 some numerical results, Ek :¼ f
r : fs ½
r; k 6 0g; ð5Þ
that show the efficacy and effectiveness of the proposed as
method, are reported and discussed.
fs ½r; k ¼ maxff ðkre þ r
Þg ð6Þ
re 2Se
2. Shakedown analysis
is the shakedown admissible domain and represents the set
of all possible translations r
 of the domain kS within E.
2.1. Basic assumption
Obviously, Ek is also convex and satisfies the condition
We assume the structure is subject to external load p½t  2 Ek () kre ½t þ r
r  2 E: ð7Þ
expressed as a combination of basic loads pi and belonging
Starting from an initial admissible state f
r0 ; e0 g and con-
to the admissible load domain P:
( ) sidering a strain increment e  e0 we can obtain the associ-
X p ated final admissible state f r; eg by the following return
min max
P : p½t :¼ ai ½tpi : ai 6 ai ½t 6 ai 8t ð1Þ mapping scheme
i¼1
r E  Eep ;
½u; k :¼ r ½u; k 2 Ek ;
r ð8Þ
closed and convex, by definition. This assumption follows
the external load description used in Eurocodes that de-
rives from the fact that the real load evolution is often un-
known while the minimum and maximum values for every
pi are available. Denoting with rei the elastic stress solution
corresponding to single basic load pi , the elastic stress do-
main S can be expressed as
( )
X p
min max
S : re ½t :¼ ai ½trei : ai 6 ai ½t 6 ai 8t ; ð2Þ
i¼1

S represents the envelope of the elastic stresses and collects


the local values of the elastic stresses produced, at different
instants, by load path p½t contained in P:S is also closed
and convex. It is worth mentioning that time is assumed
here as an evolution variable since we always consider
the dynamic effects due to external load as negligible.
Our task is to determine the shakedown safety factor,
that is the largest multiplier factor k that can be used for
amplifying the stress domain S (or the load domain P) Fig. 1. Elastic domain E and Ek .
1178 M. Malena, R. Casciaro / Computers and Structures 86 (2008) 1176–1188
 
where r 0 þ Eðe  e0 Þ is the elastic predictor (not nec-
E :¼ r os½u; k os½u; k
K j :¼ ; yj :¼ : ð15Þ
essarily contained in Ek ), related to the strain increment ou ðuj ;kj Þ ok ðuj ;kj Þ
through the elastic matrix E, and ep the associated plastic
strain defined by the Kuhn–Tucker condition: Instead of K j , the elastic stiffness matrix K e in each itera-
ep ¼ ln; n 2 ofs ½r; k; tion cycle is used. This is assembled and decomposed at
 the start of the process, once and for all, and for this reason
¼0 if f s ½
rE ; k < 0;
l: ð9Þ the iteration process proves to be computationally fast.
P 0; fs ½
r; k ¼ 0 if f s ½
rE ; k P 0: This choice is also convenient because, as shown in [1,2],
An equivalent way for obtaining r½e; k is provided by the it assures the convergence of the iterative scheme.
Haar–Karman principle: In this way the iterative corrections u_ j and k_ j are
Z obtained from the equation
1
ð
rrE ÞT E 1 ð
rr
E Þ ¼ min; r 2 Ek ; ð10Þ " #" #  8
2 B  < u_ j ¼ K 1 ðsj þ k_ j yj Þ;
K e yj u_ j sj e

which implicitly contains both Eqs. (8) and (9). ¼ ; yT K 1 s


T
yj 0 k_ j 0 : k_ j ¼  T 1
j e j
;
yj K e yj

2.3. Finite element shakedown analysis ð16Þ

The method proposed in [1] solves Eq. (4) by means of where sj ¼ s½uj ; kj  and vector yj is evaluated by the relation
an iterative-incremental process that produces a non- Z
decreasing sequence kðkÞ ; k ¼ 1; 2; . . ., of multipliers con- yj :¼ DT r
t ½ej ; kj  dv; ð17Þ
verging to shakedown factor. It can be summarized as B
follows.
Assuming the structure has been modelled by standard t ½ej ; kj  :¼ o
as r r½uj ; kj =ok.
FEM discretization, let u be the vector of free nodal dis- The interested reader is referred to [1,2] for a detailed
placements, e ¼ Du the strain field associated to u by com- presentation of the method. It is worth mentioning here
patibility matrix D and r  the stress field, a nodal force that each step is initialized by assuming
vector s½u; k can be associated to u through the virtual
works equilibrium equation k1 :¼ kðk1Þ þ bðkÞ ðkðk1Þ  kðk2Þ Þ;
Z u1 :¼ uðk1Þ þ bðkÞ ðuðk1Þ  uðk2Þ Þ; ð18Þ
deT r
½u; k ¼ duT s: ð11Þ
B
bk being an appropriate scaling factor we can relate to iter-
Self-equilibrated stresses are characterized by the condition ation loops performed in the previous step which renders
s½u; k ¼ 0, therefore Eq. (4) can be rewritten in discrete the stepping process self-adaptive. Within each step, the
terms as scheme (13) and (16) produces a stress sequence rj monot-
ks :¼ maxr fk : 9
r½u; k 2 Es ½kg : s½u; k ¼ 0; ð12Þ onously convergent ðksj k ! 0Þ to a self-equilibrated solu-
tion. This solution satisfies the plastic admissibility, by
s½u; k being defined by Eq. (11). definition, and the corresponding multiplier kðkÞ is then less
Eq. (12) is solved by an incremental iterative process than or equal to the shakedown multiplier ks . Furthermore,
that produces a sequence of admissible safe states the sequence fkðkÞ g obtained by the step process is monot-
ðkÞ ; kðkÞ ; kðkÞ 6 ks monotonous non-decreasing
xðkÞ ¼ ½uðkÞ ; r onous non-decreasing ðkðkÞ P kðk1Þ Þ and, for kðkÞ ¼ kðk1Þ ,
ðkÞ
in k , starting from the elastic limit state xð0Þ ¼ ½0; 0; ke . provides the shakedown solution ks .
In each step, the new state xðkÞ is obtained from the previ-
ous one xðk1Þ as a limit of the convergent sequence of val-
ues xj ¼ fuj ; r j ; kj g; j ¼ 1; 2; . . ., where the displacement 3. The FEM model
vector and the load multiplier are recursively updated by
(
ujþ1 ¼ uj þ u_ j ; The FEM discretization of the 3D frame is obtained
ð13Þ using the 3D beam element proposed in [6]. The element
kjþ1 ¼ kj þ k_ j : geometry is described by assigning its length, section and
Corrections u_ j and k_ j are defined to satisfy the equilibrium orientation in space, the latter being defined by the unit
condition (12) and a constraint equation: vectors fi x ; i y ; i z g of the local system fx; y; zg in the external
( (global) cartesian fx1 ; x2 ; x3 g system, as shown in Fig. 2.

s½ujþ1 ; kjþ1  ¼ 0; s½uj ; kj  þ K j u_ j þ k_ j yj ¼ 0; The element mechanics is defined by the nodal force f i ; f j
ð14Þ and couple resultants mi ; mj of the external action on the
yT ðu  uj Þ ¼ 0; yTj u_ j ¼ 0
end-sections i and j of the element and by the associated
with sj defined by (11), K j and yj being the initial tangent nodal displacements ui ; uj and rotations /i ; /j . By collect-
in ðuj ; kj Þ ing these in the ½12  1 element vectors
M. Malena, R. Casciaro / Computers and Structures 86 (2008) 1176–1188 1179

Fig. 2. 3D beam element.

8 9 8 9
> pi > > ui > We consider a generic shape reinforced concrete section
>
> > > >
<m > = </ >
> = with an arbitrary distribution of reinforcement bars. The
i i
f e :¼ ; ue :¼ ; ð19Þ section is defined by the position of its vertices in the
> pj >
> > > uj >
> >
>
: ; > >
: ; > fx; yg plane and by the position and the diameter of
mj /j the reinforcement bars. Assuming x and y axes in the prin-
the external work is directly defined as cipal directions of the concrete section, the elastic domain E
of the section is defined as interaction domain between the
T
f e ue ; ð20Þ axial load N and the biaxial bending M x ; M y (Figs. 5–7).
The following assumptions [7,8] are made in the analysis:
while the ½12  12 element stiffness matrix Ke , implicitly
defined by the Clapeyron relation (1) plane sections before deformations remain plane after
1 1 T 1 deformation;
T T 1
Ue ¼ f e ue ¼ ue K e ue ¼ f e Ke f e ð21Þ (2) full strain compatibility exists between the steel rein-
2 2 2
forcements and the surrounding concrete;
will be obtained by recovering the explicit expression for (3) concrete tensile stresses are ignored;
the element elastic strain energy (4) elastic–plastic behavior is assumed for both concrete
Z ( 2 2 2 2 2
and steel;
1 N M 2 M y vx T x þ vy T y þ vt M t (5) strains and stresses are taken into account with their
Ue ¼ þ x þ þ
2 ‘ EA EJ x EJ y GA signs (negative for compression and positive for
 tension).
vxy T x T y þ vxt T x M t þ vyt T y M t
þ2 dx; ð22Þ (6) shear and torsion failures are always prevented by the
GA
presence of appropriate transversal reinforcements.
where, using standard notations, N ; T x ; T y ; M t ; M x ; and M y
are the strengths in the generic section of the beam (ex- Eurocode 2 specifies the design material strengths, for
pressed as a function of the end-joint actions pi ; mi ; concrete rcd and steel rsd , defined as follows:
pj and mj through the equilibrium), EA; EJ x ; EJ y ; GA; 0:85  0:83Rck Rsk
and vx ; vy ; vt ; vxy ; vxt ; vyt , are the elastic constants and the rcd :¼ ; rsd :¼ ; ð23Þ
cc cs
shear and torsional factors of section. All the elastic factors
can be obtained from the section geometry through an Rck and Rsk being the reference strengths and cc and cs the
automatic process. Explicit expressions for the element associated safety factors, having the values 1.6 and 1.15,
stiffness matrix K and the nodal force vector s[u] and respectively.
further details about the element are reported in [6].
4.1. Piecewise linearization of the yield surface
4. The elastic domain for a reinforced concrete section
A convenient approximate representation of the elastic
The implementation of the solution method described in domain of the beam sections will be used, based on a
Section 2 needs the design shakedown domain Es ½k to be multi-surface piecewise linearization of its yield surface.
properly defined. This is done by determining the elastic Each flat surface is defined by referring to a possible plastic
domain E for all the reinforced concrete sections employed strain mechanism for the section, defined by the axial strain
in the analysis, once and for all before the start of the incre- ep and the bending strains vpx and vpy , collected in the vector
mental process. The shakedown domain Es ½k is then ep :¼ fep ; vpx ; vpy g: ð24Þ
updated, starting from the elastic domain and using the
current values of k, at each iteration loop of the iterative This choice allows us to determine the position of the neu-
scheme. tral axis and the boundary of compressed concrete section
1180 M. Malena, R. Casciaro / Computers and Structures 86 (2008) 1176–1188

Xr . The corresponding failure stresses are then determined and c :¼ ðcx þ cy Þ=2 is defined as the average value of
by the equilibrium equations cx and cy .
8 R This choice allows to rescaling the surfaces according to
> Pnb
>
> N ¼ f cd dX þ Asi fsdi ; the anisotropies in the resistance behavior of the section.
>
> Xr
>
> i¼1
For the isotropic case kt1  t2 k ¼ kt3  t4 k ¼ kt5  t6 k to
< R Pnb
M x ¼ Xr fcd y dX þ Asi fsdi y i ; ð25Þ take six surfaces normal to the axis, eight normals to bisec-
>
> i¼1 tions of the octant, and 12 normals to the bisections in the
>
>
>
> R P
nb x–y; y–z and z–x planes (see Fig. 3).
>
: M y ¼  Xr fcd x dX  Asi fsdi xi ;
i¼1
To better illustrate the proposed multi-surface lineariza-
tion, we consider here the three test sections in Fig. 4. The
nb being the number of reinforcement bars, and corresponding piecewise linearization of the elastic domain
fxi ; y i g and Asi the position and the area of each bar. Note is reported in Figs. 5–7 by referring to the undimensional
that the integrals appearing in Eq. (25) are easily obtained, scaled stresses:
integrating by part through the general formula:
M xu M yu Nu
Z XL Z ðxhþ1 y k Þ^
j¼N mx ¼ ; my ¼ ; n¼ ;
k k njx þ ðxh y kþ1 Þ^
njy kt1  t2 k=2 kt3  t4 k=2 kt5  t6 k=2
x y dX ¼ dC; ð26Þ
Xr j¼1 Cj hþkþ2 ð31Þ

as a sum of the explicit contributions derived from the N L


boundary sides Cj of the compressed concrete section,
njx and ^
^ njy being the components of the external normal
of the generic side.
Letting t :¼ fN ; M x ; M y g be the generalized stresses act-
ing on the section and tk the failure stresses related to
ep ¼ nk by Eq. (25), the yield surface f ½t ¼ 0 can be repre-
sented by m flat surfaces, each defined by a different direc-
tion nk , i.e.
f ½t :¼ ff1 ½t; . . . ; fm ½tg; ð27Þ
where
fk ½t :¼ fnTk t  ck g 6 0; ck ¼ tTk nk ; k ¼ 1; . . . ; m:
Fig. 3. Directions.
ð28Þ
The accuracy of this representation, obviously, derives
from the numbers of flat surfaces and on the way of defin-
ing them. A satisfactory compromise between precision
and computational complexity is to use 26 flat surfaces, de-
fined through Eq. (25) by the 26 columns nk of matrix N:

ð29Þ
where cx ; cy and cz are defined by limit stress relatively to
principal strain direction:
kt1  t2 k kt3  t4 k kt1  t2 k
cx :¼ ; cy :¼ ; cz :¼ ð30Þ
kt5  t6 k kt5  t6 k kt3  t4 k Fig. 4. Test sections.
M. Malena, R. Casciaro / Computers and Structures 86 (2008) 1176–1188 1181

Fig. 5. Admissible and linearized domains for the rectangular section.

Fig. 6. Admissible and linearized domains for the T-shaped section.

Fig. 7. Admissible linearized domains for the L-shaped section.

where M xu ; M yu and N u are the failure section stresses and 4.2. Some remarks
t1    t6 are defined as in Eq. (30). The relative errors in do-
main linearization derived from a comparison with the ex- Note that Eqs. (27) and (28) correspond to an external
act solution are reported in Table 1. tangent linearization of the admissible domain, so they will
1182 M. Malena, R. Casciaro / Computers and Structures 86 (2008) 1176–1188

Table 1 bk computed once and for all at the beginning of the pro-
Errors in piecewise linearization cess, while amin max
and tei are defined according to Eq.
i ; ai
Section Middle error (%) Quadratic middle error (%) (2).
Rectangular (Fig. 5) 1.85 2.89
T shaped (Fig. 6) 2.96 4.39 5.2. The return mapping scheme
L shaped (Fig. 7) 1.72 2.43
The plastic flow rule (9) and (8), related to the multi-sur-
face yield condition (32) and (33), can be written in the
provide an overestimate of both the collapse and shake-
compact form
down multipliers. However, in concrete applications, the
errors in the evaluation of the domain are quite small ep ¼ Nl;
_ l_ T ¼ ½l_ 1 ; . . . ; l_ m ; l_ k P 0; l_ k fsk ¼ l_ k f_ sk ¼ 0;
(see Table 1) and those actually obtained in the evaluation ð35Þ
of the multipliers are even smaller (see Section 6). So the
approximation can be considered acceptable for practical where l_ is the plastic multiplier vector and N the matrix
purposes. collecting the normals to the m flat surfaces of shakedown
Also note that the process does not check the strains, admissible domain.
but only considers the failure strengths of the section. A Starting from a known safe point ðk0 ; t0 ; u0 Þ the return
check on the total plastic strains in the section is, however, mapping scheme (8) gives the new feasible value of t 2 Ek
provided by the incremental solution process itself. In fact, corresponding to the displacements increment u  u0 and
at each new point of the path, the algorithm computes a to the assigned value of k. We obtain
solution in P terms of displacements u and total plastic t ¼ tE þ Dt; ð36Þ
strains ep ¼ k DepðkÞ . This can be assumed as reference
value for the total plastic strain needed by the adaptation where
process to obtain the shakedown solution. So, during the Dt ¼ ENl;
_ tE ¼ t0 þ EðDT u  DT u0 Þ; ð37Þ
incremental analysis both the greatest compression strain
in concrete and tensile strain in steel are monitored and matrix E being the stiffness matrix defined in Section 3.
compared with the corresponding ultimate values assigned Neglecting the torsional contribution, the feasibility condi-
by Eurocode 2 (ecu ¼ 0:0035 for the concrete and esu ¼ 0:01 tion becomes
for the steel). When these values are surpassed, the analysis NT ðtE þ DtÞ  c þ kb 6 0; cT ¼ ½c1 ; . . . ; cm ;
is stopped. bT ¼ ½b1 ; . . . ; bm : ð38Þ
Previous equations can be summarized in the Haar–Kar-
5. Implementation details man form (10). We obtain
(
5.1. The shakedown admissible domain Es ½k minimize : PHK ½Dt ¼ 12 DtT E1 Dt
ð39Þ
subject to : f s ½Dt; k  CT Dt  b P 0;
The analysis uses the shakedown admissible domain
Es ½k, defined from the elastic domain E through Eqs. (5) where
and (6). This domain should be updated, according to C ¼ N; b ¼ kb  c þ NT tE ; ð40Þ
the current value kj , at each iteration loop of the incremen-
tal process. For this reason the updating of Es ½k plays an that is, a standard strictly convex quadratic programming
important rule in the solution process. problem which is conveniently solved by the active set algo-
The updating can take advantage of the proposed piece- rithm by Goldfarb and Idnani [9], also used in [5] (the inter-
wise linearization of elastic domain (Eq. (28)). In fact, Es ½k ested reader is referred to these papers for further details).
comes to be defined by the shakedown admissibility It is worth mentioning that problem (39) has to be solved
conditions

fs ½t; k :¼ ffs1 ½t; k; . . . ; fsm ½t; kg; ð32Þ

where

fsk ½t; k ¼ fnTk t  ck þ kbk g 6 0; bk :¼ maxfnTk te ½tg ð33Þ


te 2Se

are the shakedown k-yield conditions and


*
Xp
amin T
if nTk tei < 0;
i nk tei
bk ¼ aki ; aki :¼ ð34Þ
i¼1 amax
i nTk tei if nTk tei P 0; Fig. 8. Continuous beam.
M. Malena, R. Casciaro / Computers and Structures 86 (2008) 1176–1188 1183

separately for each beam of the frame, using the stiffness According to Eqs. (36) and (37), we have
matrix E of the beam.
ot½uj ; kj 
¼ ENl;
_ k; ð42Þ
5.3. Evaluation of vector yj ok

where l;_ k can be easily obtained as a by-product of the ac-


Another quantity to be computed is the vector yj defined
tive set solution algorithm. By splitting vector l_ into the ac-
by Eq. (17)
Z Z tive part l_ a and non-active part l_ n , we have
ot½uj ; kj   T 
yj :¼ DT dv ¼ DT t;k dV : ð41Þ 
B ok B
l_ a ½k ¼ H1
a N a t E  c a þ kba ; l_ a ½k ¼ 0; ð43Þ

where H :¼ NTa ENa and NTa ; ca and ba collect the active
rows in matrix NT and vectors c and b, respectively. There-
fore, we obtain

_ k ¼ H1 b;
l; t;k ¼ ENH1 b: ð44Þ

6. Numerical results

In this section we present and discuss some numerical


implementations of the proposed shakedown solution algo-
rithm. Several structures are analyzed by comparing the
shakedown safety domains provided by our algorithm with
the corresponding limit-elastic and collapse safety ones
obtained by using the same 26-plane piecewise linearization
of the section yield surface. The collapse multipliers are
evaluated through a standard path-following analysis for
proportional loading.
Test cases have been chosen to illustrate the algorithm
results for simple problems, but also cover quite complex
real-scale examples of reinforced concrete structures, in
order to show the effectiveness of our proposal as a practi-
Fig. 9. Simple 2D frame. cal design tool.

Fig. 10. Interaction domain for continuous beam and simple 2D frame under P 1  P 2 .
1184 M. Malena, R. Casciaro / Computers and Structures 86 (2008) 1176–1188

6.1. Continuous beam

The first numerical test refers to the multibay simply


supported beam shown in Fig. 8, with constant section
characterized by ultimate strength M 0 . Two elements are
used to describe each horizontal span. External loads are
defined by
p½t :¼ a1 ½tP 1 þ a2 ½tP 2 ; 0 6 a1 ½t 6 1; 0 6 a2 ½t 6 1:
ð45Þ
The elastic, shakedown and collapse safety domains are
shown in Fig. 10, for different values of the ratio P 2 =P 1 .
Note that the shakedown domain while larger than the
elastic one is overestimated by the collapse domain for
some P 2 =P 1 ratios as expected.

6.2. Simple 2D frame

The second test refers to the simple plane frame shown


in Fig. 9. One element is used to describe each column
and two for each horizontal beam. The loads domain is
defined by
p½t :¼ a1 ½tP 1 þ a2 ½tP 2 ; 0 6 a1 ½t 6 1; 0 6 a2 ½t 6 2; Fig. 12. Interaction domain for simple 3D frame.

ð46Þ and two for each horizontal beam. The load domain is
P 1 and P 2 being the horizontal and vertical forces, respec- defined by
tively. The safety domains are shown in Fig. 10, for differ- X
3

ent values of the ratio P 2 =P 1 . Also in this case, the p½t :¼ ai ½tP i ; 1 6 a1 ½t 6 1; 1 6 a2 ½t 6 1; a3 ½t ¼ 1:
i¼1
shakedown domain is overestimated by the collapse one.
ð47Þ
6.3. Simple 3D frame The results for the elastic, shakedown and collapse safety
factors are reported in Fig. 12. Note that the shakedown
The 3D frame, shown in Fig. 11, is considered under the safety domain is quite highly overestimated by the collapse
action of varying horizontal concentrated loads ðP 1 ; P 2 Þ, domain for all P 2 =P 1 ratios.
assumed to be components of a horizontal load P making
an angle a with the global X-axis, in the presence of a con- 6.4. 3D frame 1
stant distributed vertical loading P 3 applied on the beams
that could be interpreted as dead vertical loads. As in the The test refers to the four storey 3D frame shown in
previous test, one element is used to describe each column Figs. 13 and 14, where the horizontal slabs are modeled as

Fig. 11. Simple 3D frame geometry.


M. Malena, R. Casciaro / Computers and Structures 86 (2008) 1176–1188 1185

Fig. 13. 3D frame 1 geometry and horizontal loads.

Table 2
P 3 and P 4 loads for beams [KN/m] in 3D frame 1
Level P3 P4 Level P3 P4 Level P3 P4 Level P3 P4
1 40.0 4.0 2 32.0 4.0 3 23.0 4.0 4 18.0 4.0

in-plane rigid membranes. The structure is consider under


the action of varying horizontal concentrated loads
ðP 1 ; P 2 Þ, assumed to be components of a horizontal load P
making an angle a with the global X-axis and applied in cor-
respondence to the center of gravity of each slab. These
forces act in the presence of distributed vertical loads
P 3 and P 4 applied on the beams and defined in Table 2.
The load domain is defined by
X
4
p½t :¼ ai ½tP i ; 1 6 a1 ½t; a2 ½t 6 1;
i¼1
Fig. 14. Cross-sections in 3D frame 1. a3 ½t ¼ 1; 0 6 a4 ½t 6 1; ð48Þ

Fig. 15. Interaction domain for 3D frame 1 (dotted lines include a plastic strain check).
1186 M. Malena, R. Casciaro / Computers and Structures 86 (2008) 1176–1188

Fig. 16. 3D frame 2 geometry and horizontal loads.

Table 3 Table 5
Cross-sections for columns in 3D frame 2 P 3 and P 4 loads for beams [KN/m] in 3D frame 2
Grid reference Level Section i Section j Grid reference Level P3 P4
A3 1–2–3–4 5 5 A1–A2 A2–A3 A3–A4 1 9.74 0.59
A4 B4 1–2 5 5 B1–B2 B2–B3 B3–B4 1 4.33 0.59
B3 1–2 5 5 A1–B1 1 20.75 5.0
B3 3–4 6 6 A2–B2 1 27.85 10.0
A1 1 6 5 A3–B3 1 22.85 8.0
A1 2–3–4 5 5 A4–B4 1 20.75 5.0
B1 A2 1–2–3–4 6 6 A1–A2 A2–A3 2 9.74 0.59
B2 1 7 6 B1–B2 B2–B3 B3–B4 A3–A4 2 4.33 0.59
B2 2–3–4 6 6 A1–B1 2 20.75 5.0
A2–B2 2 27.85 10.0
A3–B3 2 28.25 8.0
A4–B4 2 15.35 5.0
Table 4 A1–A2 A2–A3 3 9.74 0.59
Cross-sections for beams in 3D frame 2 B1–B2 B2–B3 3 4.33 0.59
Grid reference Level Section i Section c Section j A1–B1 A3–B3 3 20.75 5.0
A2–B2 3 27.85 10.0
A1–A2 A2–A3 A3–A4 1 3 1 3 A1–A2 A2–A3 B1–B2 B2–B3 4 4.04 0.29
A1–B1 A2–B2 A3–B3 1 3 4 3 A1–B1 4 12.85 2.5
A4–B4 1 2 1 2 A2–B2 4 22.85 5.0
B1–B2 1 2 1 3 A3–B3 4 18.85 4.0
B2–B3 B4–B3 1 3 1 1
A1–A2 A4–B4 2 3 1 3
A2–A3 B1–B2 2 3 1 1 where P 3 and P 4 could be interpreted as dead and live ver-
A3–A4 B3–B4 B2–B3 2 1 1 1
tical loads and P 1 ; P 2 roughly simulate a horizontal seismic
A1–B1 2 2 1 2
A2–B2 2 2 3 4 action.
A3–B3 2 4 3 4 Safety domains, for the elastic, shakedown and collapse
All beams 3 1 1 1 state are reported in Fig. 15, for different values of the ratio
All beams 4 1 1 1 P 2 =P 1 .

Fig. 17. Cross-section elements in 3D frame 2.


M. Malena, R. Casciaro / Computers and Structures 86 (2008) 1176–1188 1187

During both collapse and shakedown analysis a defor- 6.5. 3D frame 2


mation check has been performed at the end of each step
of the incremental process, in order to test if the admissible The test refers to multi-storey 3D frame, shown in
plastic strains in the end-element sections have been sur- Fig. 16, whose cross-sections details for the columns and
passed, according to Eurocode 2 criteria (shakedown beams are reported in Tables 3 and 4, respectively, by ref-
strains are evaluated as described in Section 4). The result- erence to Fig. 17. Note that a different reinforcement
ing safety domains are represented as dotted lines in Fig. 15 scheme is considered in the end-sections ði and jÞ of both
and appear a little more restrictive. columns and beams af the frame and in the medium span
Furthermore, in order to show the errors deriving from sections (c) of the beams.
the 26-plane piecewise linearization of the section yield As in previous test case, varying horizontal concentrated
surfaces in practical contexts, the analysis has been loads ðP 1 ; P 2 Þ are considered, assumed to be components of
repeated with a refined linearization base on 66-plane a horizontal load P making an angle a with the global
approximation which provided values for both the shake- X-axis and applied in correspondence to the center of grav-
down and collapse multipliers lower than but very near ity of the floor slabs. These forces, roughly simulating a
to those obtained with 26-planes. The maximum relative seismic action, act in the presence of distributed vertical
error for the collapse was 2.61% and 0.72% for the dead load P 3 and live load P 4 applied to the beams and
shakedown. concentrated at the top of the columns (see Tables 5 and 6).

Table 6
P 3 and P 4 loads for columns [KN] for 3D frame 2
Grid reference Level P3 P4 Grid reference Level P3 P4
A1 1 84.30 12.50 A1 2 81.50 12.50
A2 1 123.40 25.20 A2 2 120.60 25.20
A3 1 106.30 20.40 A3 2 106.50 20.40
A4 1 80.80 12.80 A4 2 46.60 12.80
B1 1 119.20 24.10 B1 2 116.40 24.10
B2 1 165.50 50.40 B2 2 162.70 50.40
B3 1 138.20 40.20 B3 2 162.20 40.20
B4 1 120.80 25.10 B4 2 85.60 25.10
A1 3 79.30 12.50 A1 4 41.20 6.30
A2 3 120.60 25.20 A2 4 75.80 12.60
A3 3 79.00 12.30 A3 4 57.40 9.90
B1 3 114.10 24.10 B1 4 70.50 12.10
B2 3 162.70 50.40 B2 4 132.40 25.20
B3 3 116.10 24.70 B3 4 103.10 19.70

Fig. 18. Load/displacement path and interaction domains for 3D frame 2.


1188 M. Malena, R. Casciaro / Computers and Structures 86 (2008) 1176–1188

Load domain is defined by does not consider the resistance deterioration due to the
X
4 brittle behavior of the concrete. However its ability to also
p½t :¼ ai ½tP i ; 1 6 a1 ½t; a2 ½t 6 1; provide an evaluation for the total plastic strains spent in
i¼1 the adaptation process, could allow, at least in principle,
a3 ½t ¼ 1; 0 6 a4 ½t 6 1: ð49Þ to also include in the analysis a safety check based on duc-
tility requirements. This extension will be necessary in prac-
The safety domains are reported in Fig. 18 (on the right). tical design, because the assumption of perfect plasticity
In the same figure, on the left, the evolution of the shake- can lead to solutions which overestimate the load capacity
down multiplier versus displacement in the X-direction of of the structure.
node B4 in correspondence of fourth storey is reported. Finally, it is worth mentioning that the ability of the
The number of iterations performed in each step is also re- proposed approaches to analyze, not only simple tests of
ported in the same figure. academic interest but also large complex frames, shows
that shakedown analysis could be considered a suitable
7. Conclusions tool to be used in practical design of reinforced concrete
structures like a possible and more appropriate substitute
A FEM procedure for the shakedown analysis of rein- for the methods which use the same structural modeling
forced concrete 3D frames subject to complex loading con- but, are based on collapse analysis approaches.
ditions has been presented, based on the iterative method
proposed in [1]. References
Only flexural plastic mechanisms have been considered
and the yield surfaces of the sections have been piecewise [1] Casciaro R, Garcea G. An iterative method for shakedown analysis.
linearized. The numerical investigation confirmed that the Comput Methods Appl Mech Eng 2002;191:5761–92.
use of the proposed 26-plane piecewise external lineariza- [2] Casciaro R, Garcea G. Errata to ‘An iterative method for shakedown
analysis’. Comput Methods Appl Mech Eng 2006;196:714–5.
tion can be considered an acceptable compromise between
[3] Melan E. Zur Plastizität des raümlichen continuum. Ing Arch
accuracy and computational efficiency. Increasing the num- 1938;9:116–26.
ber of linearizing planes is not sufficiently justified by [4] Koiter WT. General theorems for elastic plastic solids. In: Sneddon
improvements in accuracy of the safety domains of the IN, Hill R, editors. Progress in solids mechanics. Amsterdam: North-
structure, in practical contexts. Further extensions of this Holland; 1960. p. 165–221.
[5] Garcea G, Trunfio GA, Casciaro R. Mixed formulation and locking in
approach are however possible and could be convenient
path following nonlinear analysis. Comput Methods Appl Mech Eng
for a better recovering nonlinear structural response, which 1998;165(1–4):247–72.
include shear and torsion or other relevant plastic failure [6] Petrolo AS, Casciaro R. 3D beam element based on De Saint Venant’s
mechanisms. rod theory. Comput Struct 2004;82:2471–81.
Numerical results also confirmed that shakedown multi- [7] De Vivo L, Rosati L. Ultimate strength analysis of reinforced concrete
sections subject to axial force and biaxial bending. Comput Methods
pliers could be noticeably smaller than those provided by
Appl Mech Eng 1998;166:261–87.
collapse analysis. Consequently safety check procedures [8] Sfakianakis MG. Biaxial bending with axial force of reinforced,
based on nonlinear analysis considering monotonous composite and repaired concrete sections of arbitrary shape by
increasing loads should be considered unsafe in the pres- fiber model and computer graphics. Adv Eng Software 2002;33:
ence of multiple loading conditions. 227–42.
[9] Goldfarb D, Idnani A. A numerical stable dual method for solving
The proposed method is based on a crude, maybe too
strictly convex quadratic programs. Math Prog 1983;27:1–33.
crude, elastic–perfectly plastic structural modeling which

Das könnte Ihnen auch gefallen