Вы находитесь на странице: 1из 12

EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS

Earthquake Engng Struct. Dyn. 2003; 32:1431–1442 (DOI: 10.1002/eqe.285)

Evaluating foundation mass, damping and stiness


by the least-squares method

S. H. Ju∗; †
Department of Civil Engineering; National Cheng-Kung University; Tainan 70101; Taiwan

SUMMARY
This paper discusses how to use the three-dimensional (3D) time-domain nite-element method incor-
porating the least-squares method to calculate the equivalent foundation mass, damping and stiness
matrices. Numerical simulations indicate that the accuracy of these equivalent matrices is acceptable
when the applied harmonic force of 1+sine is used. Moreover, the accuracy of the least-squares method
using the 1+sine force is not sensitive to the rst time step for inclusion of data. Since the nite-element
method can model problems exibly, the equivalent mass, damping and stiness matrices of very compli-
cated soil proles and foundations can be established without diculty using this least-squares method.
Copyright ? 2003 John Wiley & Sons, Ltd.

KEY WORDS: damping; nite-element method; foundation; least-squares method; mass; soil; stiness

INTRODUCTION

For the vibration analysis of foundations placed on soft ground, it is necessary to con-
sider the dynamic eect produced from the soil. Such a soil dynamic eect can usually
be modelled by using equivalent mass, damping and stiness matrices. A substantial amount
of research on calculating these three soil parameters of embedded foundations is reported
in the literature. Wolf and Song [1] calculated the static stiness and mass matrices of a
bounded medium and the dynamic stiness and unit-impulse response matrices of an un-
bounded medium by using the scaled boundary nite-element method. Bernal [2] presented
the formulation of a frequency-domain substructure approach for the analysis of secondary
systems using a dynamic-stiness matrix in physical co-ordinates to characterize each one
of the substructures. Qian et al. [3] used the boundary-element procedure to analyse the dy-
namic response of rigid surface footings on an elastic half-space. Mulliken and Karabalis [4]
∗ Correspondence to: S. H. Ju, Department of Civil Engineering, National Cheng-Kung University, Tainan 70101,
Taiwan.
† E-mail: juju@mail.ncku.edu.tw

Contract=grant sponsor: National Science Council, Republic of China; contract=grant number: NSC-90-2211-E-
006-063.
Received 19 November 2001
Revised 14 August 2002 and 2 November 2002
Copyright ? 2003 John Wiley & Sons, Ltd. Accepted 2 November 2002
1432 S. H. JU

developed a discrete model for predicting the dynamic through-the-soil interaction between
adjacent rigid foundations supported by the homogeneous, isotropic and linear elastic half-
space. Wolf [5] developed a simple spring–dashpot–mass model with frequency-independent
coecients to present the foundation on deformable soil. Rigid foundations on the surface and
with embedment were considered for all translational and rotational motions in his study. Zhao
et al. [6] generated a method to calculate the dynamic stiness matrix for two-dimensional
(2D) foundations in a horizontally layered soil media. Their method is based on indirect
boundary-element methods. Qian and Beskos [7] used a boundary-element methodology to
study the stiness and damping of 2D rigid surface foundations of arbitrary shape perfectly
bonded to the elastic half-space. Song and Wolf [8] used the damping-solvent extraction
method to calculate the dynamic-stiness matrix of an unbounded medium. For a literature
review before 1994, the interested reader can consult the article of Qian and Beskos [7].
Most of the above references used the boundary-element method. The time-domain nite-
element method is seldom adopted to simulate the unbounded wave problem because the mesh
prole requires a large number of degrees of freedom; moreover, the boundary wave will re-
ect back if only the traditional nite-element method is used. Since the nite-element method
can model problems exibly, complicated soil proles and foundations can be modelled with-
out diculty. Thus, this paper discusses how to use the three-dimensional (3D) nite-element
method incorporating the least-squares method to calculate the equivalent foundation mass,
damping and stiness matrices.

LEAST-SQUARES METHOD FOR DYNAMIC EQUATIONS

For the dynamic analysis, the equation of motion in the matrix form at a certain time step t
and a certain force indexed as k is

[M]{kt a} + [C]{kt v} + [K]{kt x} = {kt F} (1)

where [M]; [C] and [K] are mass, damping and stiness matrices, respectively; {kt a}; {kt v}
and {kt x} are the acceleration, velocity and displacement vectors at time step t under a certain
force k, respectively; {kt F} is the applied force k at time step t. The ith term of Equation (1)
is

{kt a}T {Mi } + {kt v}T {Ci } + {kt x}T {Ki } = kt Fi (2)

where {Mi }; {Ci } and {Ki } are the ith row of [M]; [C] and [K], respectively.
Since [M]; [C] and [K] are unknowns to be calculated, it is convenient to represent them
in the vector form as {Mv }; {Cv } and {Kv }. For example, one simple arrangement of the
components of the symmetric mass matrix and the mass vector is

Mi(i−1)=2+j
v
= Mij for i¿j (3)

Then, a transformation matrix [Qi ] can be used to dene the relationship between {Mi }
and {Mv } as follows:

{Mi } = [Qi ]{Mv } (4)

Copyright ? 2003 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2003; 32:1431–1442
EVALUATING FOUNDATION MASS, DAMPING AND STIFFNESS 1433

If the scheme of Equation (3) is used for the transformation of the mass, damping and
stiness matrices and vectors, the components of the matrix [Qi ] are
Qi(i−1)=2+j; j = 1 for i¿j; Qj;j(j−1)=2+i = 1 for j ¿i and Qi; j = 0 for others (5)
Substituting Equation (4) into Equation (2), one obtains
  
[Qi ] [0] [0]  {Mv } 
[{kt a}T {kt y}T {kt x}T ]  [0] [Qi ] [0]  {Cv } = kt Fi (6a)
 
[0] [0] [Qi ] {Kv }

or written in a simplied form {kt A}T [Ti ]{B} = kt Fi (6b)


where
 k     
 {t a }  [Qi ] [0] [0]  {M v } 
k
{t A} = {kt v} ; [Ti ] =  [0] [Qi ] [0]  and {B } = {Cv } (6c)
 k   
{t x} [0] [0] [Qi ] {Kv }

The sum of the squares of the error for a number of applied forces and time steps is
k T
= ({t A} [Ti ]{B} − kt Fi )2 (7)
i k t

To minimize the sum of the squares of the error (@=@{B} = 0), one obtains the following
linear equation:
[S]{B} = {F} (8a)
where



[S] = {kt A}{kt A}T [Ti ]
[Ti ]T (8b)
i t
k

[F] = [Ti ]T {kt A}kt Fi (8c)
i k t

When the applied force k is dened, the time-dependent nodal accelerations, velocities and
displacements can be obtained using the nite-element method for a soil–structure interactive
problem. Alternatively, the nodal accelerations and velocities can also be calculated from the
nodal displacements using the central dierence method as shown in Equations (9a) and (9b);
thus, the only input is the displacement eld. Then, Equation (8a) is obtained by substituting
them into Equations (8b) and (8c). Finally, mass, damping and stiness matrices can be found
by solving Equation (8a).
{kt v} = ({kt+t x} − {kt−t x})=(2t) (9a)

{kt a} = ({kt+t x} − 2{kt x} + {kt−t x})=(t 2 ) (9b)


where t is the time-step length.

Copyright ? 2003 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2003; 32:1431–1442
1434 S. H. JU

SELECTING AN APPROPRIATE HARMONIC APPLIED FORCE {kt Fi }

A simple way to nd the exible matrix of a structure is to put a unit force at a certain degree
of freedom, and then a column of the exible matrix can be obtained. In this study, a similar
way is used. A harmonic force is arranged at the degree of freedom whose mass, damping
and stiness needs to be calculated, and zero forces are arranged at other degrees of freedom.
Thus, if the number of degrees of freedom whose dynamic parameters need to be calculated is
N , there are N dynamic analyses required for the arrangement of Equation (8a). For example,
for a foundation with six degrees of freedom, there are six nite-element analyses required
to nd the dynamic parameters. The harmonic force is applied to each degree of freedom for
each nite-element analysis. For this arrangement of applied forces, the index i of Equations
(8b) and (8c) is counted from 1 to N , and the index k is also counted from 1 to N .
A suitable harmonic function should have the features of continuity and dierentiability;
moreover, the rst and second derivatives should have the same period as the original function,
such as sine and cosine functions. However, using them as applied forces will make Equation
(8a) nearly singular. This condition can be validated from a dynamic problem with a single
degree of freedom as follows:

ma + cv + kx = sin(!t)
 (10)

where m; c; k; a; v and x are mass, damping, stiness, acceleration, velocity and displacement,
respectively. The solution of this equation is

x = xh + p sin(!t
 − ); v = ẋh + p! cos(!t
 − ) and a = xh − p! 2 sin(!t
 − ) (11)

where xh is the homogeneous solution of Equation (10), p is a factor and  is a phase angle.
Those two values are functions of m; c; k and !.
 When ! is larger than the natural frequency
of Equation (10) and the damping c is not zero, the eect of the homogeneous solution
can be minor, especially for the time after a number of cyclic periods. Thus, neglecting the
homogeneous solution and substituting Equation (11) into Equation (8a), one can obtain
    
[p sin(!t
 −)]2 p2 ! sin(!t
 −) cos(!t
 −) −p2 ! 2 sin(!t
 −) sin(!t
 −)
 t t   t 
 
 p2 ! cos(!t
 −) sin(!t
 −) [p! 2 cos(!t
 −)]2 −p2 ! 2 cos(!t  −) 
 −) sin(!t
[S]=  
 t  t t  
 
−! 2
[p sin(!t
 −)] 2
−! 2
p ! sin(!t
2
 −) cos(!t
 −) ! 2
p ! sin(!t
2 2
 −) sin(!t
 −)
t t t

(12)

The above equation indicates that the components of the third column of matrix [S] are −! 2
times those of the rst column, which means that matrix [S] is singular. If the homogeneous
solution is not neglected in Equation (8a), matrix [S] will not be singular. However, when the
eect of the homogeneous solution is minor, matrix [S] is nearly singular, and this condition
causes the solution of Equation (8a) to be unstable. This condition will be demonstrated in the
section of numerical simulations. The functions of 1+sine and 1−cosine will also be tested
in numerical simulations as they do not have the above-mentioned drawback.

Copyright ? 2003 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2003; 32:1431–1442
EVALUATING FOUNDATION MASS, DAMPING AND STIFFNESS 1435

Figure 1. Mass, damping and stiness errors vs Nt using Equation (13) for Example 1.

NUMERICAL VALIDATION

Example 1: The example with two degrees of freedom

In this numerical example, a problem with two degrees of freedom was rst solved. To realize
the accuracy of the analysis, an accuracy ratio R is dened in Equation (13). The dynamic
matrix equation of this two-degree-of-freedom problem is shown in Equation (14), where Hiv
is the actual ith component of the mass, damping or stiness vector illustrated in Equation
(3), H iv is that component calculated from Equation (8a) and NH is the total number of
components of the vector.
NH
i=1|Hiv − H iv |
R= NH (13)
i=1 |Hi |
v

        
2 −1 a1 10 10 v1 2000 −500 x1
+ + = {f } (14)
−1 3 a2 10 20 v2 −500 1000 x2

First, the accuracy of the harmonic forces including sine, 1+sine and 1−cosine with the
period T of 2 s was investigated. The nite-element method was used to nd the solution of
Equation (14) for 500 time steps with the time-step length of T=100 = 0:02 s. The summations
of t in Equations (8b) and (8c) were arranged from time step Nt to time step 500. Figure 1
shows the error ratio of the calculated mass and stiness matrices changing with Nt . This
gure indicates that the accuracy of the sine function is highly dependent on Nt , but the
accuracy of those under the other two functions is almost independent of Nt . For the sine
function, the calculation error increases when Nt increases. This is because the eect of
the homogeneous solution is minor for a large Nt . At this time, the sine function causes
Equation (8a) to be unstable. This condition has been explained in the previous section.
Then, the periods T of the harmonic forces were changed from 0.1 to 4 s in order to
investigate the variation of the error ratio. The results are shown in Figure 2, which demon-
strates that the calculated mass is inaccurate for the sine function when the period of the sine
function increases. For the other two harmonic forces, the results are accurate enough.

Copyright ? 2003 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2003; 32:1431–1442
1436 S. H. JU

Figure 2. Mass, damping and stiness errors vs the period of applied force
using Equation (13) for Example 1.

Example 2: A rigid circular footing resting on an elastic half-space

A rigid circular footing of radius a resting on an elastic half-space investigated by Luco and
Mita [9] was compared. The footing radius a, shear modulus G, Poisson’s ratio  and mass
density  of the homogeneous soil are 10 m, 8e4 kN=m2 , 31 and 2 t=m3 , respectively. Within
the circular footing area, a master–slave node scheme [10] was used. The master node was
arranged at the area centre and other nodes within this area were set to slaved nodes, whose
three degrees of freedom are dominated by the master node which has vertical, horizontal,
torsional and rocking degrees of freedom. Luco and Mita [9] obtained the complex stiness
matrix [KLuco ] with the following relationship of real [M]; [C] and [K] in Equation (1).
1 1
[Ke ] = Re([KLuco ]) = ([K] − !2 [M]) (15)
Ga1 Ga1
[Ce ] = Im([KLuco ])=(a0 Ga1 ) = ![C]=(a0 Ga1 ) (16)
where [Ke ] is the equivalent dimensionless stiness, [Ce ] is the equivalent dimensionless
damping, ! is the wave frequency, a1 equals the footing radius a for horizontal and vertical
translation terms, equals a3 for the rocking term and equals a2 for the horizontal-rocking
coupled term and a0 is the dimensionless frequency as follows:
a0 = !vs =a (17)

where vs (= G=) is the shear velocity.
The nite-element mesh is shown in Figure 3, where the absorbing boundary conditions
[11; 12] were used along the ve boundaries, except the top surface, to simulate semi-innite
soil stratum. Three nite-element analyses were performed to nd the vertical, horizontal and
rocking terms in Equations (15) and (16) using a unit 1 + sine load with the frequencies of
15, 22.5, 30, 45 and 60 rad=s, and the applied force was subjected to each degree of freedom
at the master node in each analysis. The Newmark direct integration method was used in
the time-history nite-element analysis with Newmark parameters  and  of 0.25 and 0.5,
respectively. Eight-node 3D isoparametric elements were used to simulate the soil prole.
The solution method of the nite-element analysis was the conjugate gradient method with

Copyright ? 2003 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2003; 32:1431–1442
EVALUATING FOUNDATION MASS, DAMPING AND STIFFNESS 1437

Figure 3. Finite-element mesh of Example 3 (model dimensions = 290 m × 290 m × 113 m,


471240 8-node solid elements).

Table I. Horizontal, rocking, vertical and horizontal–rocking stiness and damping values of Example 3.
a0 (Equation (17)) Horizontal Rocking Vertical Horizontal–rocking

Ke Ce Ke Ce Ke Ce Ke Ce
0.775 4.943 3.567 4.106 0.517 5.899 5.976 −0:538 0.016
1.162 4.958 3.333 3.769 0.821 5.568 5.674 −0:556 0.110
1.549 4.705 3.317 3.412 1.020 4.779 5.679 −0:552 0.138
2.324 4.413 3.376 2.745 1.265 3.405 6.224 −0:478 0.187
3.098 4.210 3.351 2.058 1.419 2.824 6.611 −0:261 0.193

1
the SSOR scheme [13]. The time-step length was set to 100 of the loading period and a total
of 500 time steps were used. Table I shows the horizontal, rocking, vertical and horizontal-
rocking stiness and damping (Equations (15) and (16)) of the least-squares result. The
least-squares and the Luco and Mita [9] results are shown in Figure 4, which indicates an
acceptable agreement.

Example 3: Flexible foundation and soil–structure interactive analyses of buildings

A 14-storey reinforced concrete building with the storey height of 3:2 m and the column
spacing of 7:5 m was analysed. Each column is connected to a pile cap with four reinforced
concrete piles as shown in Figure 5, in which the pile diameter is 1:2 m and the length is
50:4 m. The top surface of piles is connected to a cap with thickness of 2:2 m. Within the
area of the 1:4 m diameter disk at the pile cap surface center, a master–slave node scheme
[10] was used to model the connection eect of the column. The master node was arranged
at the area centre, and other nodes within this area were set to slaved nodes, whose three

Copyright ? 2003 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2003; 32:1431–1442
1438 S. H. JU

Figure 4. Comparison between least-squares and theoretical [9] results for Example 2.
(a) K e (Equation (15)) changing with a0 (Equation (17)), (b) C e (Equation (16))
changing with a0 (Equation (17)).

Figure 5. Foundation dimensions and nite-element mesh of the foundation of Example 3.

degrees of freedom are dominated by master node. The master node having six degrees
of freedom, three translations and three rotations, represents the structural behaviour of the
foundation. First, the least-squares method was used to nd the 6 by 6 mass, damping and
stiness matrices at the master node. The Young’s modulus, Poisson’s ratio and mass density
of the concrete are 2e7 kN=m2 , 0.15 and 2:4 t=m3 , respectively. The Young’s modulus of the
surface soil is 0:75 e5 kN=m2 and more than 50 m under the ground is 10e5 kN=m2 . The
linear interpolation is applied to determine the Young’s modulus between these two depths.
The mass density and Poisson’s ratio of the soil are 2 t=m3 and 0.48, respectively. The two
factors of Rayleigh damping ([Damping] =  [Mass] +  [Stiness])  and  equal 0:4=s and

Copyright ? 2003 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2003; 32:1431–1442
EVALUATING FOUNDATION MASS, DAMPING AND STIFFNESS 1439

Figure 6. The plane view of the nite-element mesh for Example 3. (The foundation of Figure 5 is
located at the mesh centre, and along the ve mesh surfaces except the top surface, absorbing boundary
conditions were set. The Z-direction contains 32 element layers.)

Table II. Residual ratios of Example 3 at three dierent frequencies using Equation (18).
Frequency (rad=s) 15 30 60
Re , Residual ratio (Equation (18)) 0.0029 0.0025 0.0066

7:3 e4 s, respectively, which provides approximately 2% soil damping ratio at a frequency of


7 Hz. The nite-element mesh is shown in Figure 6, where the absorbing boundary conditions
[11; 12] were used along the boundaries shown in Figure 6 to simulate semi-innite soil
stratum. Six nite-element analyses were performed to nd the three dynamic matrices for a
unit 1 + sine load with the frequencies of 15, 30 and 60 rad=s (periods of 0.419, 0.209 and
0:105 s), and the applied force was subjected to each degree of freedom in each analysis.
The Newmark direct integration method was used in the time-history nite-element analysis
with Newmark parameters  and  of 0.25 and 0.5, respectively. Eight-node 3D isoparametric
elements were used to simulate the soil prole. The solution method of the nite-element
analysis was the conjugate gradient method with the SSOR scheme [13]. The time-step length
1
was set to 100 of the loading period and a total of 500 time steps were used. Since there is
no exact solution of this example, Equation (7) was rearranged to obtain the residual ratio
(Re ) as shown in Equation (18). Table II shows the residual ratios of the 1+sine loads at
three dierent frequencies using Equation (18). Since the residues are small, the least-squares
method obtains accurate simulations. Figure 7 shows the wave propagation at 2:17 s after
applying the vertical loading under the frequency of 30 rad=s, which shows the reasonable
wave propagation without the error reection from nite-element boundaries.

k T k k
Re = |{t A} [Ti ]{B} − t F| |t F| (18)
i k t i k t

Copyright ? 2003 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2003; 32:1431–1442
1440 S. H. JU

Figure 7. Wave propagation at 2:17 s after applying the vertical loading under the frequency of 30 rad=s
of Example 3 (magnifying factor = 4e7).

Then, the 14-storey building was analysed using the same numerical parameters mentioned
above. The square column size is 1m, the rectangular foundation beam size is 1m by 2:8m and
the other rectangular beam size is 0:5 m by 0:7 m. In the nite-element analysis, the building
oors were assumed to be rigid and the numerical scheme was used from Ju and Lin [14].
There are two types of nite-element analyses including the exible-foundation analysis and
the soil–structure interactive analysis. The exible-foundation analysis assumes the foundation
is exible, and 6 by 6 mass, damping and stiness matrices are arranged at each connection
of the foundation beam and the column end. These matrices were averaged from the 30 and
60 rad=s matrices. The soil–structure interactive analysis models the building, soils and pile
foundations using a whole nite-element mesh and absorbing boundary conditions. Figure 8
shows the mesh of the building and pile foundations, in which the soil mesh with dimensions
of 216 m × 216 m × 92:2 m (length × width × depth) is not shown in this gure. The applied
force (N ) of sin(40t) is subjected to the mass centre of the building roof in the X and Y
directions, and a torsion (T-m) of 2 sin(40t) is also subjected to the roof mass centre, where
t is time (with unit = second). Figure 9 shows the wave propagation at 1:25 s after applying
the 40 rad=s sine loads, which indicates reasonable wave propagation results. Figure 10 shows
the displacements and rotation of the foundation cap centre (point A in Figure 8). Figure
10 indicates that similar results are obtained from the exible-foundation analysis and the
soil–structure interactive analysis. A major dierence of the two methods is that the soil–
structure interactive analysis includes the coupled eect of all the foundations but the exible
foundation does not. This example indicates that the least-squares method can obtain accurate
equivalent matrices of the foundation.

CONCLUSION

In this study, a least-squares method was developed to calculate the equivalent foundation
mass, damping and stiness matrices using the time-domain nite-element results. Numerical
simulations indicate that the accuracy of these equivalent matrices is acceptable when the

Copyright ? 2003 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2003; 32:1431–1442
EVALUATING FOUNDATION MASS, DAMPING AND STIFFNESS 1441

Figure 8. Finite-element mesh of the building and piles of Example 3 (the soil mesh
is not shown in this gure).

Figure 9. Wave propagation of Example 3 at 1:25 s after applying the 40 rad=s sine
loading (magnifying factor = 3e8).

applied harmonic force of 1+sine is used. Moreover, the accuracy of the least-squares method
is not sensitive to the rst time step (Nt ) for inclusion of data. Using the least-squares
formulation deduced in this paper is not complex. Only the nodal displacements calculated
from time-domain nite-element analyses are used to nd a linear matrix equation. After
solving this matrix equation, one obtains the equivalent mass, damping and stiness matrices.
Since the nite-element method can model problems exibly, the equivalent mass, damping

Copyright ? 2003 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2003; 32:1431–1442
1442 S. H. JU

Figure 10. X displacement, Z displacement and X rotation at the foundation cap


centre (point A in Figure 8) of Example 3.

and stiness matrices of very complicated soil proles and foundations can be established
without diculty using this least-squares method.

ACKNOWLEDGEMENT

This study was supported by the National Science Council, Republic of China, under the contract
number: NSC-90-221l-E-006-063.

REFERENCES

1. Wolf JP, Song CM. The scaled boundary nite-element method—a fundamental solution-less boundary-element
method. Computer Methods in Applied Mechanics and Engineering 2001; 190:5551– 5568.
2. Bernal D. A dynamic stiness formulation for the analysis of secondary systems. Earthquake Engineering and
Structural Dynamics 1999; 28:1295 – 1308.
3. Qian J, Tham LG, Chung YK. Dynamic analysis of rigid surface footings by boundary element method. Journal
of Sound and Vibration 1998; 214:747–759.
4. Mulliken JS, Karabolis DL. Discrete model for dynamic through-the-soil coupling of 3-D foundations and
structures. Earthquake Engineering and Structural Dynamics 1998; 27:687–710.
5. Wolf JP. Spring-dashpot-mass models for foundation vibrations. Earthquake Engineering and Structural
Dynamics 1997; 26:931–849.
6. Zhao JX, Carr AJ, Moss PJ. Calculating the dynamic stiness matrix of 2-D foundations by discrete wave number
indirect boundary element methods. Earthquake Engineering and Structural Dynamics 1997; 26:115–133.
7. Qian J, Beskos DE. Dynamic interaction between 3-D rigid surface foundations and comparison with the ATC-3
provisions. Earthquake Engineering and Structural Dynamics 1995; 24:419 – 437.
8. Song C, Wolf JP. Dynamic stiness of unbounded medium based on damping-solvent extraction. Earthquake
Engineering and Structural Dynamics 1994; 23: 169 – 182.
9. Luco JE, Mita A. Response of a circular foundation on a uniform half-space to elastic waves. Earthquake
Engineering and Structural Dynamics 1987; 15:105 – 118.
10. Ju SH. Investigating contact stresses on articular surfaces by 3-D rigid links. Journal of Engineering Mechanics
(ASCE) 1997; 123:1253 – 1259.
11. Ju SH, Wang YM. Time-dependent absorbing boundary conditions for elastic wave propagation. International
Journal for Numerical Methods in Engineering 2001; 50:2159 – 2174.
12. Ju SH. Finite element analyses of wave propagations due to high-speed train across bridges. International
Journal for Numerical Methods in Engineering, 2002; 54:1391–1408.
13. Ju SH, Kung KS. Mass types, element orders and solving schemes for the Richards equation. Computers and
Geosciences 1997; 23:175 –187.
14. Ju SH, Lin MC. Building analyses comparisons assuming rigid or exible oors. Journal of Structural
Engineering (ASCE) 1999; 125:25 – 31.

Copyright ? 2003 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2003; 32:1431–1442

Вам также может понравиться