Escolar Documentos
Profissional Documentos
Cultura Documentos
net/publication/304395262
CITATIONS READS
0 71
1 author:
Andrei M Reinhorn
University at Buffalo, The State University of New York
325 PUBLICATIONS 9,537 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Shake table controls for testing nonlinear inelastic structures View project
All content following this page was uploaded by Andrei M Reinhorn on 04 November 2016.
Mettupalayam V. Sivaselvan
University of Colorado at Boulder, Boulder, USA
Gary F. Dargush
University at Buffalo (SUNY), Buffalo, USA
Oren Lavan
Technion - Israel Institute o(Technology, Haifa, Israel
Introduction
Nonlinear analyses of structural response to hazardous loads such as earthquake
and blast forces should include (i) the effects of significant material and geometric
nonlinearities (ii) the phenomenological models describing the behavior of struc-
tural components and (iii) the energy and momentum transfer to different parts of
the structure when structural components fracture. Computer analysis of structures
has traditionally been carried out using the displacement method, combined with an
incremental iterative scheme for nonlinear problems wherein the displacements in the
structure are treated as the primary unknowns. In this paper, an alternative method is
proposed for the analysis of structures considering both material and geometric nonlin-
earities. The formulation attempts to solve problems using a force-based approach in
which momentum appears explicitly and can be potentially used to deal with structures
where deterioration and fracture occur before collapse. In conventional formulations,
the response of the structure is considered as the solution of a set of differential equa-
tions in time. Since the differential equations hold at a particular instant of time, they
310 Computational structural dynamics and earthquake engineering
provide a temporally local description of the response and are referred to as the strong
form. In contrast, in this chapter, a time integral of functions of the response over
the duration of the response is considered. Such an approach presents a temporally
global picture of the response and is referred to as the weak form. The kernel of the
integral mentioned above consists of two functions - the Lagrangian and the dissi-
pation functions - of the response variables that describe the configuration of the
structure and their rates. The integral is called the action integral. In elastic systems,
the configuration variables are typically displacements. It is shown here, however, that
in considering elastic-plastic systems it is natural to also include the time integrals of
internal forces in the structure as configuration variables. The Lagrangian function is
energy-like and describes the conservative characteristics of the system, while the dis-
sipation function similar to a flow potential describes the dissipative characteristics. In
a conservative system, the action integral is rendered stationary (maximum, minimum
or saddle point) by the response. In analytical mechanics, this is called Hamilton's
principle or more generally the principle of least action. For non-conservative systems
such as elastic-plastic systems, such a variational statement is not possible, and only a
weak form which is not a total integral is possible. It is shown moreover that the form
of the Lagrangian is invariant under finite deformations. Such a weak formulation
enables the construction of numerical integration schemes.
(a) Mass with Kelvin type resisting (b) Mass with Maxwell type resisting (c) Dual of (b)
system and force input system and velocity input
(d) Combined Kelvin-Maxwell system (e) Elastic-viscoplastic system (f) Elastic-ideal plastic system
(a} Mass with Kelvin type Resisting System: Consider a spring-mass-damper system
with the spring and the damper in parallel (Kelvin Model shown in Figure 1(a}}
subjected to a time-varying force input P(t}. The equation of motion is given by:
where 8 denotes the variational operator, and the Lagrangian function, C, and the
dissipation function, qJ, of this system are given by:
(3)
Notice that due to the presence of the dissipation function and because the force P(t}
can in general be non-conservative, equation (2} defines 8I and not I itself.
(b) Mass with Maxwell type Resisting System: Consider on the other hand, a spring-
mass-damper system with the spring and the damper in series (Maxwell Model- shown
in Figure 1(b)) subjected to a time varying base-velocity input, vm(t}. The formulation
requires obtaining a Lagrangian function and a dissipation function for this system
that determine the equations of motion as above. Formulation of compatibility of
deformations results in:
Vjn
F
+-k +-c =U
F .
(4}
where F is the force in the spring and damper. Writing the equation of equilibrium
of the mass, mit+ F = 0, solving for the velocity and substituting in equation (4 }, we
have:
1. 1 1
-F+-F+-
k c m
f t
Fdr=-V;n-Vo (5)
0
J;
where vo is the initial velocity of the mass. Defining I= F dr (as suggested by El-
Sayed et al. 1991}, the impulse of the ford~ in the spring and damper, equation (5} can
be written as: '
1·· 1. 1
-I+ -I + -I = -V;n - Vo (6)
k c m
312 Computational structural dynamics and earthquake engineering
From the correspondence between equations (6) and (1), we conclude that the
Lagrangian function, £, the dissipation function, q, and the action integral, oi of
this system are given by:
. T T . T
and oi = -8 f
0
C(j,j) dt + f a~~)if
0
of dt + f
0
[v;n(t) + vo]o] dt (7)
(c) Mass with Combined Kelvin and Maxwell Resisting Systems: Consider now the
combined Kelvin-Maxwell system shown in Figure 1(d) subject to a Force Input. (Note
that the velocity input has been excluded for the sake of simplicity). The forces in the
springs are denoted by F1 and F2 respectively and their impulses by J1 and J 2. If we
define the flexibilities of the springs as a 1 = 1/k1 and a 2 = 1/k2, then the Lagrangian
function can be given by
(8)
or in matrix notation:
. 1 2 1-T . T T
C(J,u,J) =
2mu + 21 AJ + J B u (9)
where J = U1 J2]T, A= diag(a1,a2), the flexibility matrix and B = [11], the equilibrium
matrix. The equilibrium matrix operates on the vector of internal forces to produce the
vector of nodal forces. The compatibility matrix, BT, operates on the velocity vector
to produce the rate of change of deformation. As a consequence of the Principle of
Virtual Work, the transpose of the compatibility matrix is the equilibrium matrix, B.
The dissipation function is given by:
(10)
-d (ac)
-. -(ac) aq; P =}mu+cu+
- +---:-= .. . BJ. = P
dt au au au
(11)
The mixed Lagrangian of equation (9) and the Dissipation function of equation (10)
form the basis of further developments in this chapter. It should be noted that the
Lagrangian does not contain the displacement, u explicitly; therefore the momentum,
a.c;au, is conserved (see for example Scheck, 1994), which leads to the idea of the
generalized momentum, Pu = a.c;au = mu + ]1 + ]2.
Mixed Lagrangian formulation in analysis of collapse of structures 313
(12)
where <x> is the Macaulay Bracket, equal to the value of x for positive quantities
and equal to zero for negative values of x. and sgn(x), the signum function. The above
constitutive equation can be obtained from the following:
. 1 2 1 "2 . 1 . 2
.C(it,]) = mu + a] ; rp(]) = 277 (II - Fy} ;
2 2 1
I I a~~)if I
T T • T
(e) Elastic-Ideal Plastic Dynamic System: Figure 1(f) shows an elastic-ideal plastic
dynamic system. As noted above, this system is obtained from the viscoplastic one in
the limit of the representative viscous coefficient, 71, in Equation 18 going to zero. The
dissipation function rp of equation (13) then becomes:
Mii + Cu + Bj - F = 0 (15)
where M, C and B are the mass, damping and equilibrium matrices respectively F is the
vector of element internal forces and J, its impulse. Let A be the block diagonal assembly
of the element elastic flexibility matrices. The compatibility equation is similar to
Equation (24):
(16)
314 Computational structural dynamics and earthquake engineering
(17)
. 1 T 1 ·T . T T . 1 T .
L(J, ti,J) = :fu Mti + 2:J AJ + J B ti and q5(ti,J) = lu Cti + ~J) (18)
(19)
Since all other terms of the Euler-Lagrange equations remain unaffected, it is sufficient
to examine the resulting generalized equilibrium equations obtained from the above
Lagrangian:
(20)
Mixed Lagrangian formulation in analysis of collapse of structures 315
It can be shown (Sivaselvan and Reinhorn, 2006) that (dBjdt)- Ou (tiTB) = 0. Having
recognized the symmetry in B, the above result may also be proved using index notation
as follows:
T T
8I = -8II 0 Q
po.C dr?. dt + II :~
0 Q
p0 8uk dQ dt
T T T
+I IPo:~.
o Q 1},
II
8]i;dr?.dt-
o Q
Pofk 8uk dr?.dt- II
o r
Tk 8uk drdt (22)
Using Discrete Variational Calculus results in an integrator with vastly better perfor-
mance. It is also seen in the following that this results in a form similar to the classical
Newmark family of integration schemes. Such variational integrators are symplectic
and momentum preserving and often have excellent global energy behavior (Kane et al.
2000).
7 Time-step solution
Introducing the notation, Vn and Fn as the Central Difference approximations of the
velocity and the internal force respectively, Eq. (23) then becomes:
~ M
0
h 0
0
0
0
0
Ol{v1}
0 V2
0 V3
+ rC11
Ci2
0
0 V2 }
0 0l{v
0 0 V3
1
+ [B~]T
B2
Bj
F = {P
1
p2 }
p3 h 0
1
0 ,n}
+ ~ {Mv
r0 0 0 0 V4 0 0 0 V4 BT p4 0
4
(26)
where the partitions 1 through 4 represent respectively (i) degrees of freedom with mass,
(ii) those with damping but no mass, (iii) those with prescribed forces and (iv) those
Mixed Lagrangian formulation in analysis of collapse of structures 317
(28)
(29)
(30)
where {A.= A.t,A2, ... , ANy}T is the vector of plastic multipliers, vis a penalty parameter
and < > denotes the Macaulay Bracket. The Augmented Lagrangian regularization
is a combination of the usual Lagrangian term, A.;¢;(Fn+1) and the penalty function
v/2 < ¢;(Fn+1) > 2. The latter helps accelerate convergence while the former elim-
inates the need for the penalty parameter to be large, which leads to numerical
ill-conditioning. Both terms vanish at a feasible point. The solution is obtained in
318 Computational structural dynamics and earthquake engineering
two nested stages. In the inner stage or primal stage, the dual variables, i.e. the
plastic multipliers A. are held fixed and the primal variables, i.e. the forces Fn+l are
obtained by solving the above equality constrained problem. In the outer, or dual
stage, the forces are held fixed and the plastic multipliers are updated using the
formula:
(31)
The superscripts new and old have been used, rather that iteration indices, to
denote values at the beginning and at the end of an iteration, to avoid the prolif-
eration of subscripts and superscripts. Due to the Central Difference approximation,
(h /2)A.;( a¢;(Fn+l) jaFn+l) is the plastic strain increment. In physical terms, therefore, the
Augmented Lagrangian process is equivalent to relaxing the regularizing dashpot and
allowing the frictional slider to incrementally develop plastic strain in each iteration.
A dense matrix algorithm for the solution of equation (30) is presented in Sivaselvan
and Reinhorn (2001). When considering large displacements, as stated above when
describing the effect of geometric nonlinearity, the equilibrium matrix B depends on
displacement. It is therefore updated at every step using the newly computed displace-
ments. Strictly, this requires an iterative procedure because the matrix B has to be
evaluated at time n + 1j2.
8 Numerical example
The example structure is shown in Figure 2. It is a portal frame consisting of three
elements. The connections are assumed rigid. The stress-strain curve of the mate-
rial is assumed bilinear with the following properties: E= 199955 kN/mm2 and
ay = 248.2 kN/mm2. In order to show the feasibility of this formulation as an alter-
native to existing programs, the results are compared with the program DRAIN-2DX
(Allahabadi and Powell, 1988). The example although very simple, is used to illustrate
the method and its applicability to collapse simulations.
Node2 Node3
Properties !-section
i
i Columns
i
i
i
i ......
I
W12X40
d=203.2mm
tw = 7.2mm
HE~ ~H b1 = 203.1 mm
0
! w
0
\0
<') ft= 11.1 mm
l
i Beam
--
i
i d= 303.3mm
Node 1 i Node4 fw = 7.5mm
..... """' b1 = 203.3mm
7200 ft= 13.1 mm
0.00
f---'--'--...._ ... J •••••• L••••• I ••••• J ••••••
I I -5.00
I
--~--+--+--~
I I I
-
I I
--~--~---
I
1
- -100.0 1 I I I I I
l
1 I I I I I I
-200.0 -10.00 __ L--L--~--J--~ -~--~---
1 I I I I I I
~ -300.0 I I I I I I
1 I I I I I I
-400.0 0 -15.00 --r--r--T--1--l--~- ----
32" -500.0 '2 ' I I I I I
·-e -20.00 -
I
-
I
--
I
--,--,--,--,-
I I I I
-
:il"" -600.0 ~ -LagrangianMelhod I I I I
I I I I
-1oo.o L!::::=i:=::i:::==L__;_ _;__ _;__;_ _J -25.00
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0
Time(s) Time(s)
(a) Horizontal displacement (b) Vertical displacement
0.005 80.0
0.004
60.0
~
0.003
0.002
~ ..,
"
s 0.001
0
~
.,
."
0 0.000 ~
s0 -0.001 g 0.0
-~"
~
-0.002 -20.0
0
-0.003 :r: -40.0
-0.004
-0.005 -60.0 L'==~==::!___;___;___ _;__ _J
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 -250.0 -200.0 -150.0 -100.0 -50.0 0.0 50.0
Time(s) Horizontal Displacement (mm)
(c) Joint rotation (d) Column shear force-displacement
9 Concluding remarks
The evolution of the elastic-plastic structural state in time is provided a weak formu-
lation using Hamilton's principle. It is shown that a certain class of structures called
reciprocal structures has a mixed weak formulation in time involving Lagrangian and
dissipation functions. The new form of the Lagrangian developed in this work involves
320 Computational structural dynamics and earthquake engineering
0.6 0.6
:§: :§:
0.4 0.4
"
!
a -0.2
0.2
0 2 4 6
Time(s)
10
I 0.2
-0.2
0 2 4 6
Time(s)
10
l
:§:
0.5
g
I 0.5
10
Time(s)
not only displacements and velocities but also internal forces and their impulses leading
to the concept of the generalized momentum for framed structures. The derivative of
the compatibility operator with respect to displacements possesses a symmetry that
renders the Lagrangian invariant under finite displacements. The formulation can
therefore be used in geometric nonlinear analysis. A discrete time integrator has been
derived starting from a weak formulation. The integration of each step has been shown
to be a constrained minimization problem- the principle of incremental minimum com-
plementary potential energy. An Augmented Lagrangian algorithm and a dense matrix
implementation have been derived for the solution of this problem. Since the matrix
of the minimization problem of Eq. 27 is positive definite, the solution is globally con-
vergent, allowing for larger time steps for computation. The examples show that this
method can provide as good or better information than a widely used displacement
based inelastic analysis solutions. The examples also show that the method is stable
and efficient for large structures. However, the power of the method presented herein
is in evaluating structures where various elements collapse, and forces and momentum
have to be redistributed in the remaining system.
References
Allahabadi, R. & Powell, G.H. 1988. Drain-2DX user guide, University of California at
Berkeley.
Cadzow, J.A. 1970. Discrete Calculus of Variations. International Journal of Control, 11(3):
393-407.
Mixed Lagrangian formulation in analysis of collapse of structures 321
El-Sayed, M.E.M., Marjadi, D. & Sandgren, E. 1991. Force Method Formulations Based on
Hamilton Principle. Computers & Structures, 38(3): 301-316.
Glowinski, R. & Le Tallec, P. 1989. Augmented Lagrangian and operator-splitting methods in
nonlinear mechanics, Society for Industrial and Applied Mathematics.
Kane, C., Marsden, J.E., Ortiz, M. & West, M. 2000. Variational integrators and the New-
mark algorithm for conservative and dissipative mechanical systems. International Journal
for Numerical Methods in Engineering, 49: 1295-1325.
Jose, J.V. & Saletan, E.J. 1998. Classical dynamics : a contemporary approach, Cambridge
University Press.
Marsden, J.E. & West, M. 2001. Discrete mechanics and variational integrators. Acta Numerica,
10: 357-514.
Scheck, F. 1994. Mechanics :from Newton's laws to deterministic chaos, ed. 2nd carr. and enl.,
Springer Verlag.
Simo, J.C. & Govindjee, S. 1991. Non-linear B-stability and symmetry preserving return
mapping algorithms for plasticity and viscoplasticity. International Journal for Numerical
Methods in Engineering, 31(1): 151-176.
Sivaselvan, M.V. & Reinhorn, A.M. 2004. Nonlinear structural analysis towards collapse sim-
ulation - a dynamical systems approach, Technical Report, Multidisciplinary Center for
Earthquake Engineering Research.
Sivaselvan, M.V. & Reinhorn, A.M. 2006. Lagrangian approach to structural collapse
simulation. Journal of Engineering Mechanics-ASCE, 132(8): 795-805.
Stern, T.E. 1965. Theory of nonlinear networks and systems; an introduction, Addison-Wesley.
Computational Structural
Dynamics and Earthquake
Engineering
Edited by
Manolis Papadrakakis 1
,
Dimos C. Charmpis2 ,
Nikos D. Lagaros & 1
Yiannis Tsompanakis 3
1
National Technical University of Athens, Athens, Greece
2
University of Cyprus, Nicosia, Cyprus
3 Technical University of Crete, Chania, Greece
0 ~~~,~~:~~.,.
Boca Raton london New York leiden
.
CRC Press is an imprint of the
Taylor & Francis Group, an informa business
A BALKEMA BOOK
Colophon
Volume Editors:
Manolis Papadrakakis, Dimos C. Charmpis, Nikos D. Lagaros & Yiannis Tsompanakis
Cover illustration:
Minoan palace ruins, Knossos, Crete, Greece
Nikos D. Lagaros
September 2006
Although all care is taken to ensure integrity and the quality of this
publication and the information herein, no responsibility is
assumed by the publishers nor the author for any damage to the
property or persons as a result of operation or use of this
publication and/or the information contained herein.
Computational structural dynamics and earthquake engineering I Manolis Papadrakakis ... [et al.].
TA654.C633 2009
624.1 1 71-dc22
2008044127
ISBN 13 978-0-415-45261-8(Hbk)
ISBN 13 978-0-203-88163-7(eBook)
Structures and Infrastructures Series: ISSN 1747-7735
Volume 2
Table of Contents
Editorial XI
About the Book Series Editor XIII
Foreword XV
Preface XVII
Brief Curriculum Vitae of the Editors XXXI
Contributors List XXXIII
Author Data XXXVII
PART I
Computational Structural Dynamics
PART II
Computational Earthquake Engineering