Anda di halaman 1dari 13

Engineering Structures 24 (2002) 1119–1131

www.elsevier.com/locate/engstruct

A combined finite element based soil–structure interaction model


for large-scale systems and applications on parallel platforms
M.C. Genes, S. Kocak ∗
Department of Civil Engineering, Cukurova University, Adana 01330, Turkey

Received 17 September 2001; received in revised form 4 March 2002; accepted 6 March 2002

Abstract

A finite element (FE) based soil–structure interaction (SSI) model is presented and parallelized for applications on distributed
systems. The SSI model is established by combining two methods: the Consistent Infinitesimal FE Cell Method (also referred to
as ‘the scaled boundary-finite element method’) proposed by Wolf and Song (Finite-element modelling of unbounded media, Wiley,
England, 1996) for modelling the soil region extending to infinity (far-field), and the standard FE for the finite region (near-field)
and the structure. By using this combined model, a computer program for harmonic and transient analyses of soil–structure systems
is coded. The model is investigated by solving various example problems existing in the literature. The results of the model agree
with the results presented in the literature for selected problems. The advantages of the model are demonstrated through these
comparisons. In order to decrease the computation time and achieve the solution of large-scale problems, the model is parallelized.
As a result of this parallel solution, significant time is saved for large-scale problems.  2002 Elsevier Science Ltd. All rights
reserved.

Keywords: Soil–structure interaction; Transient; Impedance; Parallel; Finite element

1. Introduction Therefore, soil region, which extends to infinity, is trunc-


ated at certain regions called artificial boundaries. The
In soil–structure interaction (SSI) problems the ability waves propagating through the soil medium should be
to predict the coupled behaviour of the soil and structure dissipated at the artificial boundaries of the compu-
is necessary. Therefore, these problems require com- tational mesh. This condition somehow has to be satis-
bined soil and structure models. While structure models fied in the mathematical models. The main challenge in
are very well established in the literature, soil models SSI analyses is the representation of energy dissipated
involve complicated analysis due to their unbounded nat- at the artificial boundaries.
ure. The main difficulty in modelling the soil region Soil region, which in many cases presents a complex
arises from the propagating wave characteristics in the and non-linear behaviour, can be discretized using the
soil medium. The main goal of many related engineering Finite Element Method (FEM), the Boundary Element
studies is to develop SSI models, which are reliable and Method (BEM) or hybrid models. All of these have
easy to implement. There are numerous valuable works advantages and disadvantages depending on the
in literature, which are focused on the proposition of applied problem.
mathematical models for soil region of the SSI problems. FE models require a large-scale mesh to represent the
Most of these mathematical models require a solution surrounding soil medium (near-field), which is bounded
defined over a computational mesh. For practical pur- by the far-field that is represented by artificial bound-
poses, a limited size computational mesh is required. aries. In numerical modelling of wave propagation prob-
lems the presence of artificial boundaries introduces spu-

rious reflections, which contaminate the solution. This
Corresponding author. Present address: Scientific Computation
Research Center, Rensselaer Polytechnic Institute, CII 7309, Troy, NY error can be avoided by introducing special boundary
12180-3590, USA. Tel.:+1-518-276-6168; fax:+1-518-276-4886. conditions, which might be referred to as radiation
E-mail address: kocakFs@yahoo.com (S. Kocak). damping. These special boundary conditions such as

0141-0296/02/$ - see front matter  2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 1 4 1 - 0 2 9 6 ( 0 2 ) 0 0 0 4 2 - 1
1120 M.C. Genes, S. Kocak / Engineering Structures 24 (2002) 1119–1131

transmitting, non-reflecting and silent boundaries absorb To obtain a simple yet efficient and reliable soil model
the wave energy. The non-reflecting viscous boundaries for far-field, an FE based model proposed by Wolf and
developed by Lysmer and Kuhlemeyer [2] and White et Song, the Consistent Infinitesimal FE Cell Method
al. [3] have been widely used for various dynamic soil– (CIFEM), is used [1,15,16]. CIFEM is a BE-like method
structure interaction problems. These boundaries are that decreases the spatial dimension by one without
only capable of transmitting plane and cylindrical waves requiring a fundamental solution [1]. In their recent pap-
and, therefore, must be located far away from the struc- ers, Wolf and Song [17–19] referred to the CIFEM as
ture or source of dynamic load. This results in large- ‘The Scaled Boundary-Finite Element Method’. In this
scale FE meshes. study, however, the method will be addressed as the
Another way of treating these artificial boundaries CIFEM. The method satisfies the radiation condition and
might be introducing infinite elements for modelling the calculates the dynamic response of the region extending
far-field of SSI problems. These elements were proposed to infinity at a truncated region [1]. By using the CIFEM,
by Bettess [4] to obtain solutions to static and steady- the chosen truncated soil region representing the far-field
state dynamic problems. Since then, many authors, might be very close to the near-field or the super-struc-
including Medina and Penzien [5], Yang and Yun [6] ture. This differs from other FE based methods where
and Yerli et al. [7] have successfully extended the infi- the far-field should be kept at a significant distance from
nite element formulation to model wave propagation the near-field. Also, the free soil surface, which has to
problems in an infinite medium. Similar to the artificial be discretized in the BEM, can either be eliminated or
boundary treatment discussed in the previous paragraph, kept to minimum by using CIFEM. This results with a
a fine FE mesh is required to define the near-field for small sized computational mesh. This and other aspects
reliable predictions when infinite elements are used of the CIFEM will be demonstrated by comparisons in
along the truncated regions. the example problems section. However, the CIFEM
BEM is a semi-numerical method and requires results in a system of first-order non-linear ordinary dif-
obtaining the fundamental solutions pertaining to the ferential equations (ODE) for the dynamic-stiffness
region, provided that the radiation condition at infinity matrix with the independent variable frequency (w). In
is satisfied. It has advantages because only surface other FE or BEM based algorithms, the system equations
discretization of the body is required which decreases the can be written in discrete frequency values. Therefore,
spatial dimension by one. Two- and three-dimensional through the CIFEM, dynamic-stiffness calculations of
problems involving the dynamic analysis of surface and the infinite region are performed through an integration
underground structures have been solved by the BEM. algorithm in a pre-defined frequency interval. Also, in
This is evident in the works of Karabalis and Beskos the CIFEM formulations, the radiation condition, which
[8], Israil and Banerjee [9] and Von Estorff et al. [10]. is satisfied at high frequency values, is presented by
Multi-region Boundary Element Method is a different introducing a non-dimensional (ND) frequency. Due to
BEM approach proposed by Tanrikulu et al. [11]. Tanri- this latter assumption, the integration algorithm for the
kulu et al. solved two- and three-dimensional layered solution of the resulting ODE is started from high fre-
type problems and underground structure problems quency values, which requires some extra calculations
under harmonic and transient loading. BEM decreases until the desired frequency level is reached. Hence, for
the computational mesh significantly. However, it large-scale problems the computation time increases sig-
induces an un-symmetric equation system, and its use is nificantly because in order to reach a specific frequency
limited when non-linearity is of concern. result, the calculations have to be performed within a
Hybrid numerical schemes combine the FEM for the frequency interval until that particular frequency is
structure or the near-field with another numerical or ana- reached. The calculations made throughout the study
lytical method for the far-field. The widely used hybrid unveiled that the vast majority of the time is spent for
model is Coupled Finite Element and Boundary Element this integration procedure. This shortcoming of the
Method (FEBEM) both in time and frequency domain. model is eliminated by introducing a parallelized algor-
Two- and three-dimensional problems involving the ithm for the solution of the first-order non-linear ODE
dynamic analysis of surface and underground structures for the dynamic-stiffness matrix of the infinite region.
have been solved by the FEBEM. This is evident in the Following the substructure method CIFEM and standard
works of Von Estorff and Kausel [12], Von Estorff and FE method are combined for large-scale SSI problems
Firuziaan [13], Yu et al. [14]. Coupled FEBEM analysis and a computer program is developed. The CIFEM
exploits the advantages of the FEM and the BEM and based computer program, implemented in this study, is
forms a relatively small-size computational mesh, yet validated by solving the example problems presented by
requires the conversion of system equations into either Wolf and Song [1], for elastic soil region. Then, CIFEM
FE or BE form. Still, for many engineering problems FE is extended for viscoelastic material behaviour by introd-
based formulations are preferred and have wider appli- ucing hysteretic damping. Following, the model is inves-
cations. tigated under harmonic and transient loads. For the trans-
M.C. Genes, S. Kocak / Engineering Structures 24 (2002) 1119–1131 1121

formation between frequency and time space, Fast by length re results in the unknown dynamic-stiffness
Fourier Transformation (FFT) is used [20]. matrix of the unbounded medium with length ri. Through
the use of compatibility and equilibrium equations
between the cell and the bounded and the unbounded
2. Combined SSI model regions, the relation between the dynamic-stiffness matr-
ices of the faces of the cell are obtained in frequency
CIFEM, proposed by Wolf and Song, is based on the space in terms of static-stiffness and mass matrices of
calculation of the dynamic-stiffness matrix defined on the cell. Then by taking the limit as cell width goes to
the truncated soil region, which extends to infinity. The zero and using similarity, the formulation leads to a first-
detailed formulation of this method is given in [1]. The order non-linear ODE for the dynamic-stiffness matrix
computer program, namely SIMILAR, is also available in the independent variable w as frequency (see Eq.
to calculate only the dynamic-stiffness matrix on the (A.2) in Appendix A).
interface of the far and the near-field for elastic soil Here, some features of the CIFEM are listed. The
media [1]. In this study, a computer code is established method calculates dynamic-stiffness matrix for a discret-
by using the CIFEM for far-field and the FEM for near- ized region, which extends to infinity. It is a BE-like
field and super-structure. CIFEM used in this study is method. However, it does not require a fundamental sol-
also extended to viscoelastic materials. Here, we will ution. It is based on the FE formulation and satisfies the
discuss the basics of the method. radiation condition at infinity while also decreasing the
In this method, a discretized cell along the near- and spatial dimension of the problem by one as in BE. The
far-field interface is introduced (as in Fig. 1(b)) with its nodes on the interface coincide with those of the FE
interior boundary coinciding with the near-field and its mesh of the near-field. The dynamic-stiffness matrices
exterior boundary with the fictitious interface (see Fig. on the interface are calculated in a frequency interval
1(b)). In the figure, O is the similarity centre [1], and ri where the frequencies are automatically determined by
and re represent the characteristic length of interior and the solution algorithm, which will be discussed in Sec-
exterior boundaries of the cell, respectively. Here, the tion 3. The procedure is exact in the radial direction and
unbounded medium with the characteristic length ri is converges to the exact solution in the FE sense on the
to be analysed to calculate its dynamic-stiffness matrix. circumferential directions. It is appealing to perform the
Assembling the dynamic-stiffness matrix of the cell limit of the cell width analytically, which leads to the
(which is straightforwardly determined from its static- consistent formulation.
stiffness and mass matrices) and the unknown dynamic- In the combined SSI model the formulation of the
stiffness matrix of the unbounded medium characterised dynamic-stiffness matrix of the near-field and far-field
is derived in frequency domain. The near-field is discret-
ized by eight-node and twenty-node quadratic isopara-
metric elements, and the boundary between the structure
and the unbounded medium is discretized by three-node
and eight-node line and quadrilateral isoparametric
elements for 2D and 3D problems, respectively. Thus,
the compatibility of the displacements and equilibrium
of forces between the near-field and far-field are fully
satisfied. Static-stiffness and mass matrices pertaining to
above-mentioned finite elements are very well estab-
lished and can be found in standard FE books [21]. Here,
the general aspects of the model are discussed and well-
known formulations are not repeated. Once the dynamic-
stiffness matrix of the unbounded medium (S⬁) is calcu-
lated, it is then added to the near-field dynamic-stiffness
matrix, in frequency space, as given in Eq. (1)

冋册 冋
Fs
Fi

Sss Ssi
Sis Sii ⫹ S ⬁ 册冋 册Us
Ui
(1)

where S, F and U designate the dynamic-stiffness


matrix, force and displacement vectors, respectively. The
subscripts i and s in Eq. (1) designate the interface and
Fig. 1. Fundamental concept of FEM with the CIFEM (a) combined non-interface nodes of the FE mesh, and superscript ⬁
SSI model (b) representation of CIFEM. designates the far-field (infinite medium) as illustrated
1122 M.C. Genes, S. Kocak / Engineering Structures 24 (2002) 1119–1131

in Fig. 1(a). Here, it should be noted that the dynamic- of the model, is parallelized for applications on concur-
stiffness matrices of the finite region can be calculated rent systems [26].
at discrete frequency values. For S⬁, however, which is Bulirsch–Stoer algorithm uses rational function for
the dynamic-stiffness matrix of the infinite region, the extrapolation and yields more accurate results and faster
system of ODE given in Appendix A has to be solved. solutions for ODE. The advantages of the algorithm
In the formulations, material damping is introduced by were demonstrated in Bulirsch and Stoer [23] by com-
the use of correspondence principle by replacing the real paring it with alternatives such as: Runge–Kutta, linear
shear modulus (G) by the complex modulus (G(1 ⫹ multi-step method of Adams–Moulton–Bashforth and
2iz)), where, z is the hysteretic damping and ‘i’ is the extrapolation with polynomials. The Bulirsch–Stoer
imaginary number. algorithm self adjusts the step size and consists of six
In order to test the validity of the implemented pro- extrapolations before reaching the next calculation step.
gram, some example problems are solved to check the After performing six extrapolations, an automatic step
correctness of S⬁. The results are compared with those size correction is done, and the calculations for the next
given by Wolf and Song [1] for elastic soil media . step start.
Further, the combined SSI model, which is obtained by For the parallel solution of the ODE given in Eq.
adding a finite region and extending the CIFEM for vis- (A.2), a master–slave structure is used on multiple
coelastic soil region, is compared with other models in instruction multiple data (MIMD) concurrent systems. In
the literature. The comparisons are presented in Sec- the parallel algorithm, master processor partitions the
tion 4. matrix into submatrices, and slave processors perform
the Bulirsch–Stoer calculations in parallel. Processors
are assigned to the same amount of work by the master
3. Parallelized SSI model to avoid any delays due to unbalanced work. Therefore,
an extra dynamic load balancing process is automatically
The solution of Eq. (1) can be performed for the eliminated. Each processor deals with a portion of the
required frequencies during the calculation of the resulting matrix (S⬁). In Fig. 2, a sample matrix par-
dynamic-stiffness matrix of the far-field (S⬁) via solving titioned into four sections is illustrated. To keep inter-
the first-order non-linear ODE given in Appendix A. The processor communication to a minimum, the calculated
dynamic-stiffness matrix calculations of the far-field are matrices are updated only once at each step for the fol-
performed at a frequency interval starting from large ND lowing step. Symmetry property of S⬁ is considered, and
frequency values using a step-by-step method called the calculations are performed only for the upper portion
Bulirsch–Stoer algorithm [22,23]. Several 2D and 3D of the matrix. For message passing, Parallel Virtual
problems are solved under harmonic and transient load- Machine (PVM) software is used and the program is
ing using the computer program, and results are found tested on WINDOWS and UNIX machines. The time
to match the existing ones in the literature as presented statistics pertaining to those problems are taken and stud-
in Section 4. However, the computation time statistics ied in the following section.
taken from various examples demonstrated that the
majority of the time is spent for the calculation of S⬁.
For example, for the problem given in Section 4.5, 80% 4. Results and discussions
of the computation time is spent for the calculation of
S⬁. This example required 24 h computation time on a In this section, some example problems solved by the
PENTIUM 2 400 MHz. processor. Because of this exten- present SSI model are presented and compared with
sive time requirement, parallel solution techniques are other methods from the literature, and parallel efficiency
sought. of the model is discussed. As stated before, in the present
Parallel algorithms are widely used in engineering
applications where there are large-scale or large-time-
requiring problems. Widely used parallel algorithms are
based on partitioning a computational mesh and then
assigning each partitioned mesh to a separate processor,
thus reaching a solution using processors concurrently
[24,25]. In this study, the time requiring process is the
Bulirsch–Stoer integration algorithm applied to the inter-
face region. Our time statistics demonstrated that the
time required for the solution of near-field at each fre-
quency step is relatively less than the far-field. There-
fore, in order to decrease the solution time, the inte-
gration algorithm, which is the main time requiring part Fig. 2. Partitioned matrix for four processors.
M.C. Genes, S. Kocak / Engineering Structures 24 (2002) 1119–1131 1123

model infinite soil region is defined by line and surface


elements in 2D and 3D cases, respectively. In the
remainder, in the FE mesh figures, infinite regions,
which are calculated by CIFEM, are plotted separately
in order to demonstrate the interface between finite and
infinite regions. During the analysis, the matching nodes
of finite and infinite regions are automatically combined
by the program.

4.1. Circular cavity embedded in full-plane

To study the correctness of the developed code, in-


plane motion of circular cavity embedded in a full-plane
is considered (see Fig. 3). This problem is solved by
Wolf and Song [1] using SIMILAR for elastic soil
region. In order to validate the implementation of infinite
soil region, this problem is solved by the present code.
Due to symmetry, only one-quarter of the structure
medium interface is discretized with four three-node line
elements. In the comparisons, dynamic-stiffness coef-
ficients are used. The dynamic-stiffness coefficients are
calculated by
S⬁(w) ⫽ ⌿TS⬁(w)⌿ (2)
where ⌿ represents the rigid motion of the structure–
medium interface. The dynamic-stiffness coefficients are
represented as spring and damping coefficients as
S⬁(a0) ⫽ G(k(a0) ⫹ ia0c(a0)) (3)
The radius of the cavity is considered in ND form as Fig. 4. Dynamic-stiffness coefficient in frequency domain of circular
cavity embedded in full-plane. (a) Spring coefficients, (b) damping
r ⫽ 1. The material properties of the soil medium are coefficients.
given in ND form as: shear modulus G ⫽ 1, mass den-
sity r ⫽ 1, and Poisson’s ratio n ⫽ 1 / 3. First, the soil
material is taken as elastic, and the problem is solved
by the present code, SIMILAR and a BEM based code region. SIMILAR cannot consider material damping.
[27]. The spring and damping coefficients calculated by However, in this study CIFEM is extended to soils with
three codes for elastic soil region (0% damping ratio) material damping. Therefore, in order to compare the
are compared in Fig. 4. From the figures, one can see results of the present model for viscoelastic materials
that the present code and SIMILAR give the same only BEM results are given for 2, 5 and 10% damping
results. This validates the present code for infinite soil ratios in Fig. 4. Studying the figures, one can see that
the present model gives very close results to that of BEM
for various damping ratios.

4.2. Rigid strip foundation resting on a half-space

To study how far the cell representing the far-field


(infinite soil medium) would be located, a massless rigid
strip foundation of width 2B and thickness H on a half-
space is considered (see Fig. 5). Here, the near-field
(finite region) and the far-field (infinite region) material
properties are chosen to be the same. The constants
defining the soil material behaviour are as follows: the
Poisson’s ratio of the soil is taken as n ⫽ 0.25, in ND
form, the shear modulus and the mass density of the soil
are taken as Gs ⫽ 1 and rs ⫽ 1, respectively. ND com-
Fig. 3. Circular cavity embedded in full-plane. pliance values are studied, which are defined as
1124 M.C. Genes, S. Kocak / Engineering Structures 24 (2002) 1119–1131

Fig. 6. Vertical compliances of the rigid strip foundation.

Fig. 5. FE meshes (a) 72 elements for near-field (b) 0 element for


near-field (only foundation is discretized).

C̄vv ⫽ πGsCvv, C̄hh ⫽ πGsChh, C̄rr ⫽ πGsBCrr


(4)
a0 ⫽ wB / cs
where cs is the shear wave velocity, a0 is the ND fre-
quency, and subscripts v, h and r designate vertical, hori-
zontal and rocking motions, respectively.
In order to study the effects of the truncation for the
far-field, the interface between near- and far-field is
located closer to the foundation by simply discarding
rows and columns of the FE mesh of near-field (see Fig.
5(a)). The final mesh, which is referred to as ‘only foun- Fig. 7. Horizontal compliances of the rigid strip foundation.
dation’ in the following diagrams, is illustrated in Fig.
5(b), which corresponds to the simplest mesh where the
finite soil region is not discretized, and only foundation Temel [29] in Figs. 6–8, respectively. In Ref. [29] infi-
and far-field are discretized. Here, it is interesting to note nite elements are defined to represent the infinite soil
that the mesh given in Fig. 5(b) is similar to FEBEM region, and the mesh used consists of 58 finite elements
meshes where only the surface of the soil region is and eight infinite elements for a circular mesh. The num-
discretized in BEM analyses. However, using the present ber of elements referred to in the diagrams represents
model the free soil surface is not discretized. This should
be discretized in BEM analyses within a reasonable dis-
tance until the truncation is done. This is a significant
advantage of CIFEM over BEM. Especially in 3D prob-
lems, by eliminating or decreasing the size of the free
soil surface, a significant amount of computational effort
is saved. In CIFEM the free soil surface that has to be
discretized can be eliminated or kept to a minimum,
which will be demonstrated in the following diagrams.
A total of six FE meshes are studied having 72, 50, 32,
18, 8 and 0 eight-node quadrilateral elements for finite
region and 24, 20, 16, 12, 8 and 4 three-node line
elements for the interface representing the far-field. The
ND vertical, horizontal and rocking compliance values
of the foundation obtained by the present model for vari-
ous meshes are compared with analytical solutions given
by Luco and Westman [28] and FE solutions given by Fig. 8. Rocking compliances of the rigid strip foundation.
M.C. Genes, S. Kocak / Engineering Structures 24 (2002) 1119–1131 1125

the number of finite elements used to model the finite subscripts L and s represent layer and half-space,
region for the present algorithm. respectively. In the analyses, the Poisson’s ratio for the
As can be seen from Fig. 6 the vertical compliance whole soil region is taken as n ⫽ 0.4, and the following
distribution obtained by the present model illustrates a ND quantities are used: Gs ⫽ 1; rL ⫽ 1; rs ⫽ 1. For
good match with the others. It is interesting to note that this problem, a finite region between the soil surface and
even the simplest mesh, which represents locating the the bottom of the top layer is defined where the FE mesh
infinite soil region just underneath the foundation, gives is truncated at a 4B distance from the foundation. As a
the compliance values at a reasonable approximation. special application in this particular problem, CIFEM is
Studying the compliance values predicted by the present used both for the half-space underneath the top soil layer
model for different meshes, one can note that the finite (extending to infinity in vertical direction) and for the
region can be kept at a minimum size. As it is seen in truncated regions of the top layer (extending to infinity
Fig. 6, the results of meshes that have 72, 50, 32, 18 in horizontal) provided the related soil materials are
elements for near-field converged to eight elements. For introduced.
the sake of simplicity, they are not illustrated in the Figs. The horizontal foundation impedance values, calcu-
7 and 8. The same observation is made for the horizontal lated by the present model, are compared with the BEM
and rocking motions, except “only foundation” solutions solutions given by Tanrikulu [30] in Figs. 9 and 10 for
of the present model for horizontal and rocking did not different soil properties. The impedance and frequency
predict the infinite soil response as well as in vertical values presented in the figures have the following ND
case (see Figs. 7 and 8). Studying the figures, one can form
see that for surface foundation problems, the present
Shh wB
model gives reasonable results with a small discretized S̄hh ⫽ , a0 ⫽ (5)
finite region. Here, it should be noted that the CIFEM GsB cs
is very advantageous for embedded structures consider- From the figures, one can see that predictions of the
ing the similarity phenomenon [1]. Since, for embedded present model for the foundation impedance values are
structures the interface is located at a chosen distance in agreement with BEM values. This example is chosen
from the similarity centre. The advantages of the CIFEM to demonstrate the applicability of the present model for
for embedded structures can be seen in Fig. 6 by layered soils. Here, it should be noted that number of
investigating the good match for large meshes, which soil layers can be increased provided that each soil layer
represents the interface for embedded structures. is defined by a finite region until the last layer extending
to infinity is reached.
4.3. Rigid strip foundation resting on a layered half-
space 4.4. Distributed transient strip load on an elastic half-
space
A massless rigid strip foundation of width 2B on a
layered half-space is studied (see Fig. 9). The half-space The effect of a uniform 2B width strip transient load
consists of two layers where the top layer has thickness at the point A on the free soil surface which is 10B far
D( ⫽ 4B). The analyses are performed for two different away from the centre of the load on an elastic half-space
soil material ratios GL / Gs ⫽ 4 and GL / Gs ⫽ 0.25 where is studied (see Fig. 11). The variation of the load with

Fig. 9. Horizontal impedances of rigid strip foundation on a layer for GL / Gs ⫽ 4.


1126 M.C. Genes, S. Kocak / Engineering Structures 24 (2002) 1119–1131

Fig. 10. Horizontal impedances of rigid strip foundation on a layer for GL / Gs ⫽ 0.25.

Fig. 11. Distributed transient strip load on an elastic half-plane, geometry and loading.

respect to time is given in Fig. 11. The constants defining


the material behaviour are as follows: E ⫽ 330MPa,
n ⫽ 0.25 and r ⫽ 1.7857 × 103kg / m3. Due to sym-
metry, half of the problem is considered and the finite
region is modelled by using 72 eight-node quadrilateral
elements. The finite and infinite region interface is mod-
elled by using 18 three-node line elements.
FFT algorithm is used for the transformation between
frequency and time space. Since the frequency values
at which infinite soil responses are determined by the
Bulirsch–Stoer algorithm automatically, interpolation is
used for required values of Fourier parameters. Fig. 12
illustrates the results of the present model and Yerli et
al. [7], which is a study based on using the infinite
elements on the truncated region to represent the infinite Fig. 12. Vertical displacement at point A.
soil medium. As can be seen in Fig. 12 the present model
presents a good match with the model proposed by Yerli
et al. [7]. to study how far the truncated soil region should be
located, the size of the FE mesh given in Fig. 13(a) is
4.5. Rigid square foundation decreased as in Example 4.2. The final mesh is obtained
by introducing a single layer of soil, as presented in Fig.
A massless rigid square foundation of width 2B on a 13(b). Due to symmetry, only a quarter of the problem
layered half-space is considered. Similar to Example 4.2, is analysed, and a total of three meshes are considered
M.C. Genes, S. Kocak / Engineering Structures 24 (2002) 1119–1131 1127

Fig. 13. FE meshes (a) 27 elements for near-field (b) nine elements for near-field.

having 27, 18 and 9 twenty-node 3D solid finite elements


for the finite region and 27, 21 and 15 eight-node finite
elements for the interface.
To demonstrate the applicability of the present model
to various different problems, the following soil regions
are considered: (i) homogeneous, (ii) layered, and (iii)
top layer resting on a rigid bedrock.

(i) Homogeneous soil. In this case, the top layer has


a thickness of H ⫽ 3B and the bottom layer is
taken as half-space. The properties of the two lay-
ers are taken as identical with the following con-
stants: Gs ⫽ 24 × 109N / m2, ns ⫽ 1 / 3, rs ⫽
2400kg / m3. The compliance values of the foun-
dation calculated by the present model for various
meshes are compared with others in Figs. 14–16. Fig. 15. Horizontal compliances of the square rigid foundation.
In the diagrams the following ND compliance
definitions are used
C̄vv ⫽ GBCvv, C̄hh ⫽ GBChh, C̄rr ⫽ GB3Crr (6)
for vertical, horizontal and rocking compliances,
respectively. For comparison, the available data
are taken from different references. For vertical
compliance, BEM based solutions are taken from
Israil et al. [9] and infinite element based solutions

Fig. 16. Rocking compliances of the square rigid foundation.

are taken from Yerli [31]. For horizontal com-


pliance values, infinite element based model of
Yerli [31], and Hall–Lamp and Hall–Kitamura
point loading formulation based model by Chow
[32] are used. For rocking compliance, the results
given by Yerli [31] are used. As in Example 4.2,
Fig. 14. Vertical compliances of the square rigid foundation. the meshes studied presented close results (see
1128 M.C. Genes, S. Kocak / Engineering Structures 24 (2002) 1119–1131

Svv Shh
S̄vv ⫽ , S̄hh ⫽ (7)
GsB GsB
In Figs. 17 and 18 the results predicted by the
present model are found to be in agreement with
the BEM results [11]. The BEM mesh used in [11]
consists of 256 constant quadrilateral elements at
the surface of the layer and 64 elements at the
interface between the layer and half-space for the
full problem.
(iii) Top layer resting on a rigid bedrock. The top soil
layer has the same properties with the one given
in Case (i) and it has a thickness of H ⫽ 2.5B.
Here, to represent the rigid bedrock, the bottom
soil layer is taken as rigid. This is accomplished
by simply restraining the nodes on that surface.
Fig. 17. Vertical impedances of the square rigid foundation on a
Here, to represent the soil layer, a finite region has
layer.
to be defined until the rigid bedrock is reached.
For this problem the FE mesh given in Fig. 13(a)
Fig. 14), and for the sake of simplicity in Figs. is chosen, and in Fig.19 the ND vertical com-
15 and 16, only nine element FE mesh results are pliance values of the foundation are compared
plotted. From the Figs. 14–16 one can see that the with the infinite element results [31] and the BEM
present model demonstrates a good match with the results obtained by computer program of Ref. [27].
others. Similar to the example presented in Section In the diagram ND compliance is calculated as in
4.2, the advantage of the CIFEM for embedded Eq. (6). As can be seen from Fig. 19, the com-
structures can also be seen here in Fig. 14 by pliance values of the present model are in agree-
studying the good match between large mesh and ment with the others.
the others. Here, large mesh represents the inter-
face for embedded structures. 4.6. Parallel efficiency of the model
(ii) Layered soil. In this case, top layer has a thickness
of H ⫽ B. The Poisson’s ratio for the two layers A parallel code for Bulirsch–Stoer algorithm is
are taken as ns ⫽ 1 / 3, and the ND properties of developed for applications on distributed systems. In
the two layers are taken as follows: GL ⫽ 1, order to investigate the efficiency of the parallel algor-
Gs ⫽ 1.766, rL ⫽ 1, rs ⫽ 1.13 where subscripts ithm, several test cases are studied. The studied cases
L and s represent layer and half-space, respect- showed that the parallel algorithm is efficient and easy
ively. In the analyses, the FE mesh given in Fig. to use. In this section, for the sake of simplicity, one
13(a) is used for modelling the soil layer and the example problem is given. Three different example
impedance values of the foundation are compared meshes are chosen for structure–medium interface of
with the results given by Tanrikulu [11] in Figs. example problem given in Section 4.5, and timings of
17 and 18. The ND vertical and horizontal imped-
ances are given as follows:

Fig. 18. Horizontal impedances of the square rigid foundation on a Fig. 19. Vertical compliances of the square rigid foundation on a
layer. layer resting on rigid bedrock.
M.C. Genes, S. Kocak / Engineering Structures 24 (2002) 1119–1131 1129

those meshes are taken on WINDOWS and UNIX sys- ent analyses FFT is used. Several example problems are
tems. The number of eight-node plane elements at the solved to validate and investigate the applicability of the
structure–medium interface are chosen as 27, 32 and 48 model. The results of the present model are compared
for meshes 1, 2 and 3, respectively. Here, it should be with the data in the literature, and a good match is noted.
noted that the size of the dynamic-stiffness matrix to be By using the present model one can introduce a small
calculated by Bulirsch–Stoer algorithm increases with computational mesh for infinite soil region compared to
the number of elements defined at the structure–medium other Finite or Boundary Element based models. CIFEM
interface. In a study by Kocak and Akay [25] the direct requires the solution of a first-order non-linear ODE for
relation between the high performance parallel architec- the calculation of dynamic-stiffness matrix of the infinite
tures and the efficiency is presented. In that study, it is soil region. This differs from other FE or BE based
also shown that as the problem gets larger, higher speed- methods where the dynamic-stiffness matrices can be
ups are achieved due to low communication costs. In calculated in discrete frequency values by simply introd-
this study, the time statistics demonstrated similar results ucing inertia forces. This results in the use of an inte-
to those of Kocak and Akay [25]. As UNIX based paral- gration algorithm, which is highly expensive in compu-
lel system performed better than WINDOWS, for larger tation time. As a remedy to this problem, the integration
problems the efficiency results noted better performance. algorithm is parallelized for applications on concurrent
In Fig. 20, the elapsed speedup diagrams of three systems. Especially for large-scale problems, significant
meshes, pertaining to UNIX based system, are plotted. computational time is saved, e.g. for a moderately large
Here, it should be noted that the timings are not taken problem 83% of total elapsed time is saved by using
at reserved times, which means the processors of the sys- twelve processors.
tem were also available to other users. In order to present
more reliable results, the same problem is solved several
times to reach to less-densely used system hours. As can Acknowledgements
be seen from the diagram, for the largest mesh, due to
its being computation-bound, a maximum of ~6 speedup The authors are thankful to Dr Y. Mengi and Dr A.H.
is reached for a twelve processor case and a decrease Tanrikulu for their permission to use their BEM pro-
tendency starts after twelve processors. A similar obser- grams for comparisons.
vation is true for meshes 1 and 2 after eight processors
where they reach to their maximum efficiency. However,
here it should be noted that for mesh 3, ~83% compu- Appendix A
tation time is saved by using twelve processors. This is
a significant amount of time. By using the similarity [1], another similar fictitious
interface is defined by multiplying the coordinates of the
interface between near- and far-field according to simi-
5. Conclusions larity centre O by a similarity factor (1 ⫹ w). The simi-
larity factor is defined by dividing the distance between
A combined FE based SSI model is studied. In the the similarity centre and the exterior (fictitious) bound-
model, for infinite soil region CIFEM in frequency space ary by the distance between the similarity centre and
is used, and finite soil region is added on top. The model interior boundary.
is subjected to harmonic and transient loads. For transi- In the CIFEM, the ND cell width measured in the
radial direction is assumed to be a very small number
(see Fig. 1). Therefore, the relation between the
dynamic-stiffness matrices of the interior and exterior
boundaries will yield a limit with ND cell width (w).
This results in a derivative of the dynamic-stiffness
matrix of the unbounded medium with the characteristic
length r of the interior boundary according to similarity
centre. This derivative of the dynamic-stiffness matrix
according to r can be altered with the derivative of the
dynamic-stiffness matrix according to frequency w based
on similarity in frequency domain [1] as given in Eq.
(A.1)
rS⬁(w),r ⫽ (s⫺2)S⬁(w) ⫹ wS⬁(w),w (A.1)

where, S (w) is the dynamic-stiffness matrix of the
Fig. 20. Speedup diagram for Bulirsch–Stoer algorithm. unbounded medium and s is the spatial dimension (=2
1130 M.C. Genes, S. Kocak / Engineering Structures 24 (2002) 1119–1131

or 3). This represents the dynamic-stiffness matrix of the [7] Yerli HR, Temel B, Kiral E. Transient infinite elements for 2-D
unbounded medium as a function of frequency w, and soil–structure-interaction analysis. J Geotech Geoenviron Eng
Div ASCE 1998;124(10):976–88.
results in a system of non-linear ODE of first-order in [8] Karabalis DL, Beskos DE. Dynamic response of 3-D rigid surface
the independent variable w foundations by time domain boundary element method. Earth-
quake Eng Struct Dynamics 1984;12:73–93.
(S⬁(w) ⫹ E1) (E0)⫺1(S⬁(w) ⫹ (E1)T)⫺(s (A.2) [9] Israil AS, Banerjee PK. Effect of geometrical and material
⫺2)S⬁(w)⫺wS⬁(w),w⫺E2 ⫹ w2M0 ⫽ 0 properties on the vertical vibration of three-dimensional foun-
dations by BEM. Int J Numer Anal Methods Geomech
where, E0, E1, E2 represent the static-stiffness matrices 1990;14:49–70.
and M0 represents the mass matrix of the interior surface [10] Von Estorff O, Pais AL, Kausel E. Some observations on time
domain and frequency domain boundary elements. Int J Numer
of the cell [1]. Methods Eng 1990;29:785–800.
The solution of the system of non-linear ODE given [11] Tanrikulu AH, Yerli HR, Tanrikulu AK. Application of the multi-
in Eq. (A.2) can be performed by using Bulirsch–Stoer region boundary element method to dynamic soil–structure inter-
method [22,23], which is a powerful and efficient inte- action analysis. Comput Geotech 2001;28:289–307.
gration algorithm. For the solution of the differential [12] Von Estorff O, Kausel E. Coupling of boundary and finite
elements for soil–structure interaction problems. Earthquake Eng
equation by Bulirsch–Stoer method, the initial starting Struct Dynamics 1989;18:1065–75.
dynamic-stiffness matrix of the interface must be known. [13] Von Estorff O, Firuziaan M. Coupled BEM/FEM approach for
In order to propose an acceptable solution, the boundary non-linear soil–structure interaction. Eng Anal Boundary
condition must satisfy the radiation condition in fre- Elements 2000;24:715–25.
quency domain, and yet the static-stiffness matrix cannot [14] Yu G, Mansur WJ, Carrer JAM, Lie ST. A more stable scheme
for BEM/FEM coupling applied to two-dimensional elastodyn-
satisfy the radiation condition explicitly. The radiation amics. Comput Struct 2001;79:811–23.
condition must be formulated at an infinite distance from [15] Wolf JP, Song C. Dynamic-stiffness matrix in the time domain of
the interface, which corresponds to an infinite character- unbounded medium by infinitesimal finite-element cell method.
istic length r. However, by introducing the associated Earthquake Eng Struct Dynamics 1994;23:1181–98.
ND frequency a0 ⫽ wr / cs, infinite ND frequency can [16] Wolf JP, Song C. Unit-impulse response matrix in time domain of
unbounded medium by infinitesimal finite-element cell method.
also be achieved by an infinite w instead of an infinite Comput Methods Appl Mech Eng 1995;122:251–72.
r. In an actual calculation, the dynamic-stiffness matrix [17] Song C, Wolf JP. The scaled boundary finite-element method-
to be used in Eq.(A.2) can be calculated by choosing a alias consistent infinitesimal finite-element cell method for elasto-
large w. Thus, the high-frequency behaviour of S⬁ is dynamics. Comput Methods Appl Mech Eng 1997;147:329–55.
studied examining the asymptotic expansion of the dif- [18] Wolf JP, Song C. The scaled boundary finite-element method-a
primer: derivations. Comput Struct 2000;78:191–210.
ferential equation given in Eq. (A.3) [19] Song C, Wolf JP. The scaled boundary finite-element method—


m a primer: solution procedures. Comput Struct 2000;78:211–25.
1
S⬁(w)⬇iwC⬁ ⫹ K⬁ ⫹ j Aj (A.3) [20] Cooley JW, Lewis PAW, Welch PD. The Fast Fourier transform
j⫽1
(iw) and its applications. IEEE Trans Educ 1969;12:27–34.
[21] Bathe KJ. Finite element procedures in engineering analysis. New
where C⬁, K⬁ and Aj represent dashpot matrix, spring Jersey: Prentice-Hall, 1982.
matrix and unknown coefficient matrices of the asymp- [22] Flannery WH, Teukolsky BP, Vetterling WT. Numerical recipes.
Cambridge: Cambridge University Press, 1988.
totic expansion, respectively. m in Eq.(A.3) represents [23] Bulirsch R, Stoer J. Numerical treatment of ordinary differential
the number of terms to be considered in the analysis, equations by extrapolation methods. Numerische Mathematik
and i is the imaginary number. The details of the calcu- 1966;8:1–13.
lation of S⬁can be found in Ref. [1]. Once S⬁ at a high [24] Farhat C, Wilson E, Powell G. Solution of finite element systems
a0 (ND frequency) is calculated, it remains to solve Eq. on concurrent processing computers. Eng Comput 1987;2:157–
65.
(A.2) by Bulirsch–Stoer integration. [25] Kocak S, Akay HU. Parallel Schur complement method for large-
scale systems on distributed memory computers. Appl Math
Modelling 2001;25:873–86.
References [26] Genes MC. Soil structure interaction models for 2-D and 3-D
problems and applications on parallel platforms. PhD Disser-
[1] Wolf JP, Song C. Finite-element modelling of unbounded media. tation, Cukurova University, Department of Civil Engineering;
England: Wiley, 1996. 2001.
[2] Lysmer J, Kuhlemeyer RL. Finite dynamic model for infinite [27] Mengi Y, Tanrikulu AH, Tanrikulu AK. Boundary element
media. J Eng Mech ASCE 1969;95(EM4):859–77. method for elastic media an introduction. Ankara: Middle East
[3] White W, Valliappan S, Lee IK. Unified boundary for finite Technical University Press, 1994.
dynamic models. J Eng Mech ASCE 1977;103(5):949–64. [28] Luco JE, Westman RA. Dynamic response of a rigid footing
[4] Bettess P. Infinite elements. Int J Numer Methods Eng bounded to an elastic half-space. J Appl Mech ASME
1977;11:53–64. 1972;39(2):527–34.
[5] Medina F, Penzien J. Infinite elements for elastodynamics. Earth- [29] Temel B. Dynamic analysis of soil–structure interaction problems
quake Eng Struct Dynamics 1982;10:699–709. by using finite-infinite elements and Laplace transformation
[6] Yang S, Yun C. Axisymmetric infinite elements for soil–structure method. PhD Dissertation, Cukurova University, Department of
interaction analysis. Eng Struct 1992;14(6):361–70. Civil Engineering; 1996.
M.C. Genes, S. Kocak / Engineering Structures 24 (2002) 1119–1131 1131

[30] Tanrikulu AH. A boundary element model with nonlocal bound- Dissertation, Cukurova University, Department of Civil Engin-
ary conditions for dynamic analysis of a two-phase composite. eering; 1998.
PhD Dissertation, Cukurova University, Department of Civil [32] Chow YK. Simplified analysis of dynamic response of rigid foun-
Engineering; 1999. dations with arbitrary geometries. Earthquake Eng Struct Dynam-
[31] Yerli HR. 2D and 3D analyses of dynamic soil–structure interac- ics 1986;14:643–53.
tion problems by coupling of finite and infinite element. PhD

Anda mungkin juga menyukai