Professional Documents
Culture Documents
Abstract
Unbolted base plates are often used in civil engineering as structural connections in storage racks construction. The aim of this work is
to describe the development of a numerical model to simulate accurately the connection between columns and foundation in metallic
structures, which constitute any frame in automated storage systems. In this way, the bending stiffness of the column can be modeled in
the analysis of these structures, in order to approach the real behavior in service, and these values can be included in linear beams of
structural analysis programs, such as ESCAL3D [del Coz Diaz JJ, Ordieres Mere B, Siare Dominguez FJ, Bello Garcı́a A, Felgueros
Fernández D. J Constr Res 1998; 46:273–5]. In this study, a non-linear structural behavior of the model occurs due to the changing status
of the contact surfaces and point-to-point contacts, the geometric non-linearities of the model and the material non-linearities, such as
plasticity and surface friction. The finite element method is a general technique for numerical solution of differential and integral
equations in science and engineering. Thus the finite element approach has been carried out in two phases. Firstly, a pre-buckling analysis
has been accomplished and secondly, the above-mentioned non-linear analysis has been performed, updating the geometry of the finite
element model to the deformed configuration for the first mode buckling. A total of four load cases were analyzed, with different
compressive load and imposed lateral displacement. In order to validate the results some experimental models were tested to compare
with the numerical model, so that a good agreement and better correlations were obtained between both.
r 2006 Elsevier Ltd. All rights reserved.
Keywords: Finite element modeling; Numerical methods; Base plates; Plasticity; Large displacements; Contact problems
0263-8231/$ - see front matter r 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.tws.2006.04.008
ARTICLE IN PRESS
530 J.J. del Coz Dı´az et al. / Thin-Walled Structures 44 (2006) 529–541
designer and manufacturer of structural storage systems in numbers). The functions n in V often represent continu-
Gijón (Spain). ously varying quantities such as a displacement in an
The analysis was performed through a sequence of elastoplastic body, F ðvÞ is the total energy associated with v
multiple steps. A half-symmetry three-dimensional surface and (M) corresponds to an equivalent characterization of
model of the assembly was initially created in SolidWorks the solution of the differential equation as the function in V
2001. The model was then imported to ANSYS 9.0 via that minimizes the total energy of the considered system. In
IGES file. Once in ANSYS the geometry was completed, general, the dimension of V is infinite and thus in general
some details were added and, after a geometry checking, the problem (M) cannot be solved exactly. To obtain a
the model was meshed. Boundary conditions were included problem that can be solved on a computer the idea in the
to suitably reproduce the laboratory tests and the load was FEM is to replace V by a set V h consisting of a simple
applied in increasing levels, from 25% to 100% the beam functions only depending on finitely many parameters.
ultimate load. A modal analysis was subsequently per- This leads to a finite-dimensional minimization problem of
formed to obtain the mode shapes and frequencies. The the form: (M h ) find uh 2 V h such that F ðuh Þ F ðvÞ for
first mode shape was included to the model and a non- allv 2 V h .
linear analysis was accomplished regarding the material In practical situations, the determination of suitable
and geometric non-linearities. Finally, the results were basis functions for use in the Galerkin method [5] can be
reviewed and the simulation was compared and validated extremely difficult, especially in cases for which the domain
with the experimental tests. The discussion of the results O does not have simple shape. The FEM overcomes this
will be focused on the values obtained for the stiffness in difficulty by providing a systematic means for generating
each model and in the failure modes. basis functions on domains of fairly arbitrary shape. What
The use of FEM [3] shows innumerable advantages of makes the method especially attractive is the fact that these
economical and practical order due, on the one hand, to basis functions are piecewise polynomials that are non-zero
the cost that the realization of real tests plays, and on the only on a relatively small part of O, so that computations
other hand, to the technical difficulty, since the elements of may be carried out in a modular fashion, which is well
the present study are big in size. suited to computer-based approaches. As we know, the
In general, engineering problems are mathematical family of spaces V h defined by the finite element procedure
models of physical situations. Mathematical models are possesses the property that V h approaches V as h
differential equations with a set of corresponding boundary approaches zero, in an appropriate sense. This is, of
and initial conditions. The differential equations are course, an indispensable property for convergence of the
derived by applying the fundamental laws and principles Galerkin method.
of nature to a system or a control volume. These governing
equations represent conservation balance of mass, force, or
2. Geometrical model
energy [4].
FEM is a numerical procedure than can be used to
To define the geometry of a specimen is complex using
obtain solutions to a large class of engineering problems
an analysis program by finite elements. For this reason, a
involving stress analysis, heat transfer, electromagnetism,
three-dimensional parametric design program was used. As
and in our case a structural problem. The method was
it was mentioned above the geometric model was initially
introduced by engineers in the late 1950’s and early 1960’s
designed in SolidWorks 2001. Symmetry, drills, holes and
for the numerical solution of partial differential equations
other construction details were added once the model was
in structural engineering (elasticity equations, plate equa-
imported to the finite element program.
tions, etc.). At this point the method was thought of as a
The dimensions of the specimen are the following:
generalization of earlier methods in structural engineering
for beams, frames and plates, where the structure was
subdivided into small parts, so-called finite elements, with
unbolted base plate: 175 mm of width, 207 mm in length
and 8 mm thick.
known simple behavior.
The basic idea in any numerical method for a differential
angles: 65 mm long, 40 mm of width and 3.5 mm thick.
equation is to discretize the given continuous problem with
perforated beam section: 500 mm full-length, 112 mm of
width, 127 mm in length and 3 mm thick.
infinitely many degrees of freedom to obtain a discrete
problem or system of equations with only finitely many
force plate: 122 mm width, 133.5 mm large, 10 mm thick.
unknowns that may be solved using a computer. In this
The assembly consists of different parts, named force plate,
case, we start from a reformulation of the given differential
perforated beam, angles and base plate, as shown in Fig. 1.
equation as an equivalent variational problem. In the case of
elliptic equations this variational problem in many cases is
a minimization problem of the form: (M) find u 2 V such 3. Mathematical model
that F ðuÞ F ðvÞ for allv 2 V , where V is a given set of
admissible functions and F : V ! < is a functional (that is The resolution of this problem implies the simultaneous
F ðvÞ 2 < for all v 2 V with < denoting the set of real study of three non-linearities: (1) material non-linearity
ARTICLE IN PRESS
J.J. del Coz Dı´az et al. / Thin-Walled Structures 44 (2006) 529–541 531
qF
n;p ¼ l
d~ . (3)
qsn
Angles
In this expression l is a proportionality constant, as yet
undetermined. The rule is known as the normality principle
Fig. 1. Geometric model.
because relation (2) can be interpreted as requiring the
normality of the plastic strain increment ‘vector’ to the
yield surface in the space of n stress and strain dimensions
(plastic behavior in this case), (2) geometric non-linearity [6].
or large displacements, and (3) contact non-linearity. We Restrictions of the above rule can be removed by
will describe the equations that govern the behavior of specifying separately a plastic potential Q ¼ Qð~ s; kÞ which
these non-linearities in the following section. defines the plastic strain increment similarly to Eq. (2), i.e,
qQ
3.1. Plasticity p ¼ l
d~ . (4)
s
q~
‘Plastic’ behavior of solids is characterized by a non- The particular case of F ¼ Q is known as associated
unique stress–strain relationship as opposed to that of non- plasticity. When this relation is not satisfied (F aQ), the
linear elasticity. Indeed, one definition of plasticity may be plasticity is non-associated. In what follows this more
the presence of irrecoverable strains on load removal. If general form will be considered. During an infinitesimal
uniaxial behaviour of a material is considered, a non-linear increment of stress, changes of strain are assumed to be
relationship on loading alone does not determine whether divisible into elastic and plastic parts. Thus,
non-linear elastic or plastic behavior is exhibited. Unload-
ing will immediately discover the difference, with the elastic ¼ d~
d~ e þ d~
p . (5)
material following the same path and the plastic material
The elastic strain increments are related to stress incre-
showing a history-dependent, different, path.
ments by a symmetric matrix of constants D as usual. We
Many materials show an ideal plastic behavior in which
can thus write Eq. (5) incorporating the plastic relation (4)
a limiting yield stress, sy, exists at which the strains are
as
indeterminate. For all stresses below such yield a linear (or
non-linear) elasticity relationship is assumed. A further qQ
refinement of this model is one of hardening/softening ¼ D1 d~
d~ sþ l. (6)
s
q~
plastic material in which the yield stress depends on some
parameter k (such as plastic strain ~p ). It is with such kinds When plastic loading is occurring the stresses are on the
of plasticity that it is concerned and for which much theory yield surface given by Eq. (1). Differentiating this we can
has been developed. therefore write
It is quite generally postulated, as an experimental fact,
qF qF qF
that yielding can occur only if the stresses s satisfy the dF ¼ ds1 þ ds2 þ . . . þ dk ¼ 0 (7)
general yield criterion qs1 qs2 qk
F ðs; kÞ ¼ 0, (1) or
where k is a ‘hardening’ parameter. This yield condition T
qF
can be visualized as a surface in n-dimensional space of s Al ¼ 0
d~
s
q~
stress with the position of the surface F (yield surface)
dependent on the instantaneous value of the state para- in which we make the substitution
meter k.
Von Mises first suggested that basic behavior defining qF 1
A¼ dk . (8)
the plastic strain increments is related to the yield surface qk l
ARTICLE IN PRESS
532 J.J. del Coz Dı´az et al. / Thin-Walled Structures 44 (2006) 529–541
Eq. (6) and (7), can now be written in a single matrix form 1. Tresca:
as F ¼ 2s cos y Y ðkÞ ¼ 0, (13)
2 3
1 qQ where Y ðkÞis the yield stress from uniaxial tests.
6 D
d~ 6 s7
q~ 7 d~ s 2. Huber-von Mises:
¼6 T 7 . (9) pffiffiffi
0 4 qF 5 l
A F ¼ 3s Y ðkÞ ¼ 0. (14)
s
q~
The indeterminate constant l can now be eliminated The main difference between the problems of plasticity
(taking care not to multiply or divide by A which may be and non-linear elasticity is that there is no an explicit
zero in ideal plasticity). This results in an explicit expansion relationship equivalent to ~ s¼~sð~
Þ. Although for a given
that determines the stress changes in terms of imposed strain state, the stresses have to be on the yield surface or
strain changes with inside it, it is impossible to determine the exact value of
s ¼ Dep d~
d~ (10) each one of their components. The solution is obtained
taking into account that the tangent matrix DT ¼ Dep is
and known for a defined value of stress and the direction of the
T " T #1 applied forces, and that the stresses can be integrated using
qQ qF qF qQ the expression d~ s ¼ DT d~
.
Dep ¼ D D D Aþ D .
s
q~ s
q~ s
q~ s
q~
(11) 3.2. Large displacements
The elastoplastic matrix Dep
takes the place of the elasticity
matrix DT in incremental analysis. This matrix is symmetric In many problems, it has been implicitly assumed that
only when the plasticity is associated. The non-associated both displacements and strains developed in the structure
material will present special difficulties if tangent modulus are small. In practical terms this means that geometry of
procedures other than the modified Newton–Raphson are the elements remains basically unchanged during the
used. The matrix is defined even for ideal plasticity when loading process and that first-order, infinitesimal, linear
A ¼ 0. Explicit formulation of plasticity in this form was strain approximations can be used.
first introduced by Zienkiewicz and Taylor [7]. In practice such assumptions fail frequently even though
Clearly the general procedures outlined allow determina- actual strains may be small and elastic limits of ordinary
tion of the tangent matrices for almost any field surface structural materials are not exceeded. If accurate determi-
applicable in practice. If the yield surface (and the material) nation of the displacements is needed, geometric non-
is isotropic it is convenient to express it in terms of the linearity may have to be considered in some structures.
three stress invariants. A particularly useful form of these is Conversely, it may be found that a load is reached where
given below, introducing here also the indicial notation, deflections increase more rapidly that predicted by a linear
solution and indeed a state may be attained where load-
J1 sx þ sy þ sz sii carrying capacity decreases with continuing deformation.
sm ¼ ¼ , This classic problem is that of structural stability and
3 3 3
obviously has many practical implications. The applica-
1=2 rffiffiffiffiffiffiffiffiffiffiffiffi tions of such an analysis are clearly of considerable
1=2 1 2 1
s ¼ J2 ¼ sx þ s2y þ s2z þ t2xy þ t2yz þ t2zx sij sij , importance in aerospace engineering, design of radio
2 2
telescopes, cooling towers, box girders bridges and other
(12) relatively slender structures.
pffiffiffi
In many cases very large displacements may occur
1 3 3 J3 p p without causing large strains. Typical in this context is
y ¼ arcsen with oyo ,
3 2 s 6 6 the classical problem of the ‘elastica’ of which an example
is a watch spring.
where
Whether the displacements (or strains) are large or small,
J 3 ¼ sx sy sz þ 2txy tyz tzx sx t2yz sy t2xz sz t2xy det ~
s equilibrium conditions between internal and external
‘forces’ have to be satisfied. Thus, if the displacements
and are prescribed in the usual manner by a finite number of
nodal parameters ~ a, we can obtain the necessary equili-
s x ¼ s x s m ; s y ¼ s y s m ; s z ¼ s z sm
brium equations using the virtual work principle [6]:
or Z
T
Cð~aÞ ¼ B ~ s dV f~ ¼ 0, (15)
dij sii V
sij ¼ sij .
3 where C once again represents the sum of external and
The yield surfaces in metal plasticity can be given as internal generalized forces, and in which B is defined from
ARTICLE IN PRESS
J.J. del Coz Dı´az et al. / Thin-Walled Structures 44 (2006) 529–541 533
s ¼ D d~
and using equation d~ and Eq. (16) we obtain where the part given in brackets corresponds to the usual
s ¼ D d~
d~ a
¼ DB d~ terms:
ttij is the Cartesian components of the Cauchy stress
and taking into account the Eq. (17), it is verified that tensor (forces per unit areas in the deformed geometry).
dB ¼ dBL . Therefore, dt eij is the strain tensor corresponding to virtual
Z
displacements.
dC ¼ dBTL ~
s dV þ K d~
a ¼ K s d~
a þ K d~
a, (19) dui are the components of virtual displacement vector
V
imposed on configuration at time t, a function of
where xtj ; j ¼ 1; 2; 3 . . ..
Z
T xti is the Cartesian coordinates of material point at time
K¼ B DB dV ¼ K 0 þ K L
V t.
t
in which K0 represents the usual, small displacements V Bis t the volume at time t.
f i are the components of externally applied forces per
stiffness matrix and the matrix KL is due to the large
unit Svolume
t at time t.
displacements, and are given by
Z f i are the components of externally applied surface
tractions per unit surface area at time t.
K0 ¼ BT0 DB0 dV ,
V S tf is the surface at time t on which external tractions are
Z
T applied.
KL ¼ B0 DBL þ BTL DBL duSi is the dui evaluated on the surface Stf (the dui
V
components are zero and corresponding to the prescribed
þ BTL DB0 dV . ð20Þ
displacements on the surface S tu ).
To summarize, Eq. (19) can be expressed globally as The last summation sign in Eq. (22) gives the contribu-
dC ¼ ðK 0 þ K s þ K L Þ d~
a ¼ K T d~
a, (21) tion of the contact forces. The contact force effect is
included as a contribution in the externally applied
where KT represents the total, tangential stiffness, matrix. tractions. The tcomponents of the contact tractions are
Newton-type iteration can once more be applied precisely denoted as f ci and act over the areas S tc (the actual area
in order to solve the final non-linear problem. of contact for body at time t), and the components of the
t
known externally applied tractions are denoted as f Si
3.3. Contact conditions and act over the areas S tf . It is possible to assume that the
areas S tf are not part of the areas Stc , although such an
A particularly difficult non-linear behavior to analyze is assumption is not necessary.
the contact between two or more bodies [8,9]. Contact Fig. 2 (upper-left) illustrates schematically the case of
problems range from frictionless contact in small displace- two bodies, which are now considered in greater detail. In
ments to contact with friction in general large strain this paper, the two bodies in contact are denoted as body I
inelastic conditions. Although the formulation of the and body J. Note that each body is supported such that
contact conditions is the same in all these cases, the without contact no rigid body motion is possible. Let
ARTICLE IN PRESS
534 J.J. del Coz Dı´az et al. / Thin-Walled Structures 44 (2006) 529–541
Fig. 2. Bodies in contact at time t (upper-left) and contact elements (upper-right between angles and beam, lower-left between base plate and foundation
and lower-right between beam and base plate).
IJ t
f~ be the vector of contact surface tractions on body I necessarily of equal size. However, the actual area
IJ t JI t
due to contact with body J, then f~ ¼ f~ . Hence, of contact at time t for body I is S tc of body I, and for
body J it is S tc of body J, and in each case this area is
the virtual work due to the contact tractions in (22) can be part of SIJ and SJI. It is convenient to call SIJ the
written as ‘contactor surface’ and SJI the ‘target surface’. There-
Z Z
IJ t IJ t JI fore, the right-hand side of (23) can be interpreted as
I
dui f i dS þ duJi f JI
i dS
S IJ
S JI the virtual work that the contact tractions produce over
Z the virtual relative displacements on the contact surface
IJ t IJ
¼ duIJ
i fi dS , ð23Þ pair.
S IJ
n be the unit outward normal to SJI and let ~
Let ~ s be a
where duIi and duJi are the components of the virtual vector such that ~n and ~
s form a right-hand basis (Fig.
t 3). It
displacements on the contact surfaces of bodies I and J, is possible to decompose the contact tractions f IJ acting
respectively, and on SIJ into normal and tangential components correspond-
duJI I J ing to ~
n and ~s on SJI
i ¼ dui dui . (24)
IJ JI
The pair of surfaces S and S are termed a ‘contact IJ t
surface pair’ and note that these surfaces are not f n þ t~
¼ l~ s, (25)
ARTICLE IN PRESS
J.J. del Coz Dı´az et al. / Thin-Walled Structures 44 (2006) 529–541 535
To solve a given differential equation approximately The base plate, beam, force plate and angles were
using FEM, one has to go through basically the following modeled using SHELL181. CONTA174 and TARGE170
steps: were utilized in different contact pairs throughout the
model, such as the frontal and lateral contact between
beam and angles or between base plate and foundation.
variational formulation of the given problem.
The frontal and lateral contact between angles and beam
discretization using FEM: construction of the finite-
(see Fig. 2 upper-right) was modeled as flexible–flexible
dimensional space Vh.
standard type, with beam/shell thickness effect, and ALM
solution of the discrete problem.
as the numerical algorithm.
implementation of the method on a computer: program-
For the ALM and penalty method, normal and
ming.
tangential contact stiffnesses are required, so that we have
to define a normal contact stiffness factor named normal
The advantage of FEM as compared with finite penalty stiffness (FKN). The usual factor range is from
difference methods is that complicated geometry, general 0.01 to 1.0. For bulk deformation the best value is 1.0 and,
boundary conditions and variable or non-linear material if bending deformation dominates, we recommend using a
properties can be handled relatively easily. smaller value (0.1). In this study, the value of normal
Based on the geometric model previously described, the penalty stiffness (FKN) used was 1.0.
finite element model was built, following a four-step For ALM it is also necessary to define a tolerance factor
process. Step one was the definition of material properties. to be applied in the direction of the surface normal, named
Step two was selecting the element types, formulations and penetration tolerance (FTOLN). This factor is used to
real constants. The model was meshed in a third step and determine if penetration compatibility is satisfied. Contact
finally, in a fourth step, loads and boundary conditions compatibility is satisfied if penetration is within an
were applied. allowable tolerance (FTOLN times the depth of underlying
elements). The depth is defined by the average depth of
each individual contact element in the pair. If the program
4.1. Material properties detects any penetration larger than this tolerance, the
global solution is still considered unconverged, even
Different material properties and constitutive laws were though the residual forces and displacement increments
defined in the model for each part of the assembly. Pure have met convergence criteria. The range for this factor is
perfect isotropic elastic material behavior was assumed for less than 1.0 (usually less than 0.2), with a default of 0.1,
the force plate of the specimen. However, the rest of the and is based on the depth of the underlying solid, shell, or
parts are subjected to plasticity. Thus the multilinear beam element. In this study, the value of penetration
kinematic hardening option was selected to describe the tolerance used was 0.1.
material behavior and the data provided by the experi- The contact between base plate and foundation was
mental tests (in form of stress–strain curves) was curve- modeled as rigid–flexible standard type, with beam/shell
fitted to a multilinear representation for beam, angles and thickness effect (see Fig. 2 lower-left), and ALM as the
base plate. Fig. 5 shows the data curve-fitting employed for numerical algorithm. The values of normal penalty stiffness
the beam. Finally, for the contact pairs, a static coulomb (FKN) and penetration tolerance (FTOLN) were 1 and 0.1,
friction coefficient of 0.3 was adopted for the material. respectively.
Tension test
600
500
Stress [N/mm ]
2
400
300
200
100
0
0.000 0.005 0.010 0.015
Strain [%]
Fig. 5. Data curve-fitting for the multilinear kinematic hardening values. Fig. 6. Bolt and coupled degrees of freedom.
ARTICLE IN PRESS
J.J. del Coz Dı´az et al. / Thin-Walled Structures 44 (2006) 529–541 537
The contact between beam and base plate was modeled chosen for mode extraction was Block Lanczos [11], with
using CONTA178. It was assumed as point-to-point 10 modes.
contact (see Fig. 2 lower-right), with zero-penetration Once the mode shapes were reviewed, the geometry of
value, -Z- global direction, and pure LMM as the the finite element model was updated according to the
numerical algorithm. The weak spring option was also displacement results of the previous analysis for the first
considered for preventing rigid body motion, with a mode shape. A displacement multiplier of 2 mm was
maximum tensile force of 1 N. applied.
The 8 mm diameter bolts were modeled using
BEAM188. The nodal displacements were coupled in the
three directions (UX, UY and UZ) (Fig. 6) so that they 5.2. Non-linear analysis
move accordingly to the nodes on the beam holes.
The following step consisted of applying on the model a
4.3. Boundary conditions compressive solicitation and a transversal displacement, in
the direction which was important to measure the value of
The boundary conditions should reproduce quite accu- the stiffness.
rately those of the experimental tests performed on the Firstly, four compressive loads were applied on the
laboratory. Therefore, X and Y displacements were specimen: 55,000, 110,000, 165,000, and 220,000 N. Next,
constrained in the point of application of the axial load the imposed displacement was varied from 18 to 35 mm.
and rotations around the longitudinal axis of the bolts were In order to avoid convergence problems the analysis was
also constrained. The rest of the model remained un- accomplished in two load steps. Firstly, the compressive
constrained. load was applied ramped, together with a small displace-
ment of 0.001 mm. Next, in a second load step, the
corresponding displacement was imposed.
5. Analysis The solution controls were also adjusted to improve
convergence. Thus the parameter time was set to the value
The analysis was carried out in two phases. Firstly, a of the maximum displacement obtained (in postprocessor),
pre-buckling analysis was accomplished and then, a non- the geometric non-linearities were activated, the inertial
linear analysis was performed updating the geometry of the effects were not included, the number of equilibrium
finite element model to the deformed shape for the first iterations was specified and the tolerance convergence
mode buckling. values of forces were delimited and so were the time step
sizes for each load step.
5.1. Pre-buckling analysis
A buckling analysis was carried out with the purpose of 6. Analysis results and discussion
determining the eventual influence of geometric imperfec-
tions in the behavior of the model. Thus a compressive The analysis results were retrieved and displayed by
load, Nu, was applied on the force plate. The method means of the time-history postprocessor, POST26 [2–9,11].
Fig. 7. Results of pre-buckling analysis: (a) first mode bucking—17.844; (b) second mode bucking —18.311.
ARTICLE IN PRESS
538 J.J. del Coz Dı´az et al. / Thin-Walled Structures 44 (2006) 529–541
6.1. Pre-buckling analysis In the numerical model, the magnitude of reactions and
displacements were obtained from the ANSYS time-history
The results of the buckling analysis show the mode results postprocessor. The relative displacements (d1,y,d4)
shapes observed in the specimens subjected to compression. were calculated as an average value resultant from the
The imperfections derived from the first mode have been displacement results obtained for a set of five nodes, each
reflected in the numerical model, through a factor of set positioned on the beam external borders, at similar
imperfection of 2 mm. positions on which the experimental measurement is
As shown in Fig. 7, the multipliers for the first and the usually performed.
second mode are 17.844 and 18.311, respectively. The value of the ultimate moment was estimated from
the relationship
Table 1
Simulation results obtained at different load cases
Fig. 11. Displacements in the Z direction (upper) and stress plasticization ratio (lower) in the lower part of the specimen for the load value of 55,000 N.
A summary of the results obtained in the different load 6.3. Comparison with experimental results
cases is tabulated below provided in Table 1, where the
values of the estimated maximum design moment and The previous information was contrasted with the
bending stiffness are displayed for each load case. Besides experimental results obtained in the laboratory of the
the specimen’s failure mode is also shown for each load Technological Center of ESMENA, located in the Scien-
case. tific and Technological Park of Gijón, as shown in Fig. 9.
Figs. 11 and 12 display several results plots retrieved Graphs representing both the numerical and experimen-
from the analyses in different load cases. tal results obtained were drawn (Fig. 13) in order to get a
ARTICLE IN PRESS
540 J.J. del Coz Dı´az et al. / Thin-Walled Structures 44 (2006) 529–541
Fig. 12. Displacements in the Z direction and stress plasticization ratio in the lower part of the specimen for the load value of 220,000 N.
good agreement for the estimated stiffness and the ultimate unbolted base plates. A computational procedure has been
moment. developed based on the general-purpose finite element code
ANSYS, for modeling and simulating the non-linear
analysis of unbolted base plates. FEM and data from
7. Conclusions experimental tests are used for comparison purposes. The
findings of this study suggest that it may be possible to
The method of the finite elements (FEM) has been devise a practical procedure for establishing an unbolted
shown as a suitable tool in the modeling and analysis of base plate model by using a combined experimental/
singular structures, such as the analysis of the behavior of computational approach.
ARTICLE IN PRESS
J.J. del Coz Dı´az et al. / Thin-Walled Structures 44 (2006) 529–541 541
Acknowledgments
References
Fig. 13. Comparison between numerical simulation results (FEM) and [1] FEM 10.2.02-03-RV11, European Federation of Materials Handling
experimental data. and Storage Equipment. Sección X. Specifiers Code for the safe
provision of racking and shelving. 1998–1999.
After examining the numerical and experimental results [2] ANSYS, Users manual: procedures, commands and elements, vols.
obtained it can be assumed that the computer-aided I–III, Swanson Analysis System, 2004.
[3] Bathe K. Finite element procedures. Englewood Cliffs, NJ: Prentice-
simulation constitute a reasonable approach to describe Hall; 1996.
the behavior of the system. The FE model may reproduce [4] Brenner SC, Scott LR. The mathematical theory of finite element
quite accurately the local buckling failure forms. Moreover method. New York: Springer; 2002.
the estimated bending stiffness values are similar to both [5] Chandrupatla T, Belegundu A. Introduction to finite elements in
cases (10–15% deviation). However, the magnitudes of engineering. Prentice-Hall, NJ: Englewood Cliffs; 1991.
[6] Cook RD, Malkus DS, Plesha ME, Witt RJ. Concepts and
maximum design moments provided by ANSYS are very applications of finite element analysis. New York: Wiley; 2001.
similar (about 1%) to those obtained in the CTE labs. [7] Zienkiewicz OC, Taylor RL. The finite element method: solid and
The comparison between the two methods prove the fluid mechanics and non-linearity, vol. 2. London: McGraw-Hill;
finite element analysis as a reliable tool to get quite 1991.
accurate results in a reasonable amount of time, which [8] Simo JC, Laursen TA. An augmented Lagrangian treatment of
contact problems including friction. Comp. Struct. 1992;42:527–634.
allows the designer of the assemblies to evaluate and [9] del Coz Dı́az JJ, Garcı́a Nieto PJ, Rodrı́guez Mazón F, Suárez
optimize the design prior to manufacture and prototype Domı́nguez FJ. Design and finite element analysis of a wet cycle
testing. cement rotary kiln. Finite Elem Anal Des 2002;39:17–42.
For a second order boundary value problem like this [10] del Coz Dı́az JJ, Ordieres Mere B, Suárez Domı́nguez FJ, Bello
Garcı́a A, Felgueros Fernández D. Aprendizaje interactivo mediante
one, the space of admissible functions V consists of all
programa de análisis estructural avanzado ESCAL3D. J Constr Res
those functions in H 1 ðOÞ that satisfy the essential boundary 1998;46:273–5.
conditions. A Galerkin approximation uh to the solution [11] Moaveny S. Finite element analysis: theory and application with
of this problem was obtained by constructing a finite- ANSYS. Englewood Cliffs, NJ: Prentice-Hall; 1999.