You are on page 1of 17

Journal of the Chinese Institute of Engineers, Vol. 24, No. 6, pp.

681-697 (2001)

681

SIMPLE LUMPED-PARAMETER MODELS OF FOUNDATION USING MASS-SPRING-DASHPOT OSCILLATORS

Wen-Hwa Wu* and Cheng-Yin Chen


Department of Construction Engineering National Yunlin University of Science and Technology Yunlin, Taiwan 640, R.O.C.

Key Words: soil-structure interaction, lumped-parameter model, foundation impedance function.

ABSTRACT
Simple lumped-parameter models are developed in this study for generally representing the horizontal, vertical, rocking and torsional vibrations of rigid foundations. A spring and a dashpot are first attached in parallel to a mass to constitute a single-degree-of-freedom oscillator. Depending on the requirement of accuracy, the lumpedparameter model is then designed to consist of one or several oscillators connected in series. In addition, a new expression for the dynamic compliance function, instead of the conventional expression for the dynamic impedance function, is adopted in this research to improve accuracy. The effectiveness of these models is demonstrated for various cases including surface square foundations, surface circular foundations and embedded square foundations. It is found that the performance of the new models is better in approximating translational foundation vibrations than in cases of rotational foundation vibration. Compared with the other existing lumped-parameter models, the models proposed in this paper have advantages in requiring fewer parameters and featuring a more systematic expansion.

I. INTRODUCTION The dynamic analysis of soil-structure interaction (SSI) systems has been well studied in the past few decades. To consider SSI effects, the analysis should be extended from the structure to include the total structure-foundation-soil system. The substructure method, in which the discrete superstructure and the unbounded continuous soil are separately modelled, is commonly adopted in the SSI analysis to take advantage of the appropriate formulations for the respective subsystems. It consists of two major steps in the substructure method: (1) solving the dynamic force-displacement relationship of a

vibrating foundation sitting on soil, i.e., the dynamic impedance function of the foundation; (2) assembling the foundation impedance function with the property matrices of the superstructure to formulate the governing equations for the whole system. Therefore, a successful SSI analysis strongly depends on the application of accurate foundation impedance functions and their efficient incorporation with the structural parameters to form a total system matrix. Since most of the analysis complexity results from the frequency dependence of the foundation impedance function of the foundation, many attempts have been made to simplify SSI analysis by

*Correspondence addressee

682

Journal of the Chinese Institute of Engineers, Vol. 24, No. 6 (2001)

representing the continuous soil with frequency-independent models. Early studies usually employed representative constant values for the foundation stiffness and damping to approximate the soil system (Perelman et al., 1968; Parmelee et al., 1969; Jennings and Bielak, 1973). While satisfactory for the analysis of some typical structures, this approximation may not be adequate for other cases, such as concrete gravity dams (Chopra and Gutierrez, 1974). Several studies (Veletsos and Nair, 1974a; Meek and Wolf, 1992; Meek and Wolf, 1993) were also devoted to developing a truncated semi-infinite cone model for general applications in foundation vibration. Even though those cone models can provide valuable insights and yield good accuracy, they are not popularly applied in engineering practice due to their inability to represent the influence of surface waves and to model the portion of soil outside the cone (Wolf, 1994). It was further shown that equivalence could be established between the cone models and certain discrete physical models (Veletsos and Nair, 1974b; Wolf, 1994), which naturally lead to simple lumped-parameter models. The advantages of easy incorporation with conventional dynamic analysis and direct applicability to nonlinear structural analysis have continued to motivate the recent development of several improved lumped-parameter models (Wolf and Somaini, 1986; Nogami and Konagai, 1986; de Barros and Luco, 1990; Jean et al ., 1990; Wolf, 1991a; Wolf, 1991b; Paronesso and Wolf, 1995). These models basically are chosen by arranging a few sets of connected springs, dashpots and masses with unknown parameters, which are determined by minimizing the total square errors between the dynamic impedance of the lumped-parameter model and that for the actual system. While these discrete models have been demonstrated to have good accuracy, the intricate arrangement of vibration units usually prevents easy and systematic extension of these models with high accuracy. In addition, difficulties may also be encountered in obtaining the property matrices associated with these models. Aimed to congregate the vibration units in a more systematic way for easy expansion, a set of consistent lumped-parameter models is developed in this study to approximate vibrating foundations sitting on or embedded in an elastic half-space. Similar configurations using mass-spring-dashpot oscillators are constructed for all four different vibrating modes; horizontal, vertical, rocking and torsional directions. Moreover, a new expression for the dynamic compliance function, instead of the conventional expression for the dynamic impedance function, is adopted in this research to further improve accuracy and reduce the parameters required

for the model. The effectiveness of these lumpedparameter models is evaluated by comparing the compliance functions for various cases of circular and square foundations. II. EXPRESSIONS FOR FOUNDATION IMPEDANCE FUNCTION The dynamic impedance function of a rigid foundation supported by unbounded soil is defined as the ratio between the exerted harmonic excitation force (or moment) and the resulting steady-state displacement (or rotation). The foundation mass is usually assumed to be zero and only its rigidity is considered in this analysis for more general applicability of the impedance function. The dynamic force and displacement are generally out of phase and consequently, the foundation impedance is represented by a complexvalued function depending on the excitation frequency . The real part of the impedance function is determined from the in-phase component of the force-displacement relationship and defines the elastic restraining action of the soil medium. On the other hand, the imaginary part of the impedance function is decided from the 90-degree-out-of-phase component of the force-displacement relationship and describes the effects of radiational energy dissipation. 1. Conventional Expression For a concise presentation, the dynamic impedance function K d ( ) is conventionally normalized with respect to its corresponding static impedance K s and expressed as Kd ( )= K s[ (a 0 )+ ia 0 (a 0 )] (1)

where ( a 0 ) and ( a 0) represent the dimensionless stiffness and damping coefficients, respectively. In Eq. (1), a 0 is a dimensionless frequency parameter defined as

d a0 = v
s

(2)

to incorporate the excitation frequency , the shear velocity of soil vs and a characteristic length of foundation d (e.g., the radius of a circular foundation or the half side-length of a square foundation). From Eq. (1),(a0) and (a0) are commonly adopted in most of the studies to express the real and imaginary parts of the impedance function, respectively. 2. Euler Expression From previous works on the impedance

W.H. Wu and C.Y. Chen: Simple Lumped-Parameter Models of Foundation Using Mass-Spring-Dashpot Oscillators

683

Fig. 1 Comparison of expressions for dynamic impedance functions (surface circular foundations, =0.5)

functions of various foundations (e.g., Gazetas, 1983; Gazetas, 1993), it is evident that the dimensionless stiffness and damping coefficients of the impedance function usually fluctuate with increasing values of the frequency parameter. This feature is the major reason why numerous artificial degrees of freedom and parameters are required in most of the lumpedparameter models already developed. To overcome this difficulty, an alternative formulation using Eulers formula to denote complex quantities was suggested (Wu, 1998; Wu et al. , 2000) for expressing the dynamic impedance function: K d( a 0)= K s( e i) where i = 1 and (3)

(a 0) =

a 0 ()2 + (a 0)2 and (a 0) = tan 1( ) (4)

In other words, the normalized dynamic impedance function is expressed in terms of its amplitude and phase angle, instead of the real and imaginary parts. Therefore, and are referred to as the amplitude and the phase coefficients, respectively. The advantages of using and over and to express the impedance function primarily reside in the fact that and are generally smoother functions of a 0 . Fig. 1 illustrates this tendency for the case of a circular foundation resting on incompressible soils (Poissons ratio =0.5). Besides, it is also shown in Fig. 1 that the stiffness coefficient may become negative in this special case, which has caused problems for a number of previously suggested models.

684

Journal of the Chinese Institute of Engineers, Vol. 24, No. 6 (2001)

The amplitude and phase coefficients, on the other hand, still hold their positive and monotonically increasing characteristics. With these properties, the errors for the lumped-parameter model to fit the actual impedance function of foundation can be considerably reduced. 3. Further Adapted Expression for the Definition of Error Function In addition to a smoother expression for the foundation impedance, the definition of error function in determining the optimal model parameters is also a crucial constituent for further improving the lumped-parameter model. As long as the complexvalued impedance function is divided into two separate components (real and imaginary parts, or amplitude and phase angle), the errors respectively corresponding to these two components have to be combined in defining the total error function. Most of the previous studies directly added these two independent errors together without exploring the possibility of a more reasonable aggregation. In general, the two components of the complex-valued foundation impedance may not be of similar magnitude, which can also be observed from Fig. 1. Accordingly, the optimization process would intend to represent the component of larger magnitude with better accuracy. In this paper, the expression for the impedance function is further modified to tackle this problem. It is suggested to synthesize the two individual components of dynamic impedance through parametric representation in the complex plane using the dimensionless frequency a0 as the tracing parameter. Moreover, since the amplitude of the foundation impedance function for unbounded soil generally increases with the frequency parameter, it is difficult to assure the high-frequency convergence of the lumped-parameter model to the actual soil system. Therefore, the foundation compliance function F d (the reciprocal function of foundation impedance), which approaches zero in the high frequency limit, is adopted in this research for defining the error function. Similar to Eq. (3), Fd can also be normalized with respect to its corresponding static compliance Fs as

is also illustrated in Fig. 1 for the impedance functions of a circular foundation. III. SYSTEMATIC LUMPED-PARAMETER MODELS The lumped-parameter model proposed in this study to approximate the dynamic impedance of a rigid foundation is illustrated in Fig. 2 for all four vibrating directions. The coupled impedance between horizontal and rocking vibrations, usually insignificant, is neglected so that the lumped-parameter models can be independently established corresponding to different modes of foundation vibration. A basic unit of this model is a single-degree-of-freedom (SDOF) oscillator, which is aggregated by attaching a spring and a dashpot in parallel to a mass. The mass of this SDOF oscillator is used for modeling the inertial effect of soil, while the spring and the dashpot are employed to simulate the elastic behavior and energy radiation of the soil, respectively. Depending on the requirement of accuracy, a set of lumpedparameter models can be systematically constructed by linking one or several SDOF oscillators in series and then connecting to a rigid end. Each oscillator is specified by three parameters (mass, damping and spring coefficient). Consequently, if the lumped-parameter model consists of N SDOF oscillators, it will generally provide 3 N parameters for minimizing the discrepancy between the compliance function for the model and that for the actual system. For the lumped-parameter model illustrated in Fig. 2, let z i denote the absolute displacement of the i-th SDOF oscillator and m i, ci and k i signify the corresponding mass, damping and spring coefficient, respectively. When subjected to a single force applied on m i , the relationship between displacement ( z 1 ) of the first oscillator and the applied force ( p ) should be able to accurately represent the corresponding relationship ( u p ) for the actual system with a massless foundation sitting on or embedded in an elastic half-space. In this case, the equations of motion for the lumped-parameter model including N SDOF oscillators are given by

F d (a 0) =

1 = 1[ 1 ] = F s[ 1 e i (a 0)] K d (a 0) K s (a 0)e i (a 0) (a 0)
(5)

m 1z 1 + c 1(z 1 z 2) + k 1(z 1 z 2) = p m 2z 2 + c 2(z 2 z 3) + k 2(z 2 z 3) c 1(z 1 z 2) k 1(z 1 z 2) = 0 m N 1z N 1 + c N 1(z N 1 z N ) + k N 1(z N 1 z N ) (6) c N 2(z N 2 z N 1) k N 2(z N 2 z N 1) = 0 m N z N + c N z N + k N z N c N 1(z N 1 z N ) k N 1(z N 1 z N ) = 0

Eq. (5) shows that the amplitude of the normalized compliance function is the reciprocal of its corresponding normalized impedance function and the phase angle holds a different sign. This expression

W.H. Wu and C.Y. Chen: Simple Lumped-Parameter Models of Foundation Using Mass-Spring-Dashpot Oscillators

685

Fig. 2 Lumped-parameter models for vibrating foundations on an elastic half-space

To insure that the compliance function of the foundation is exactly modelled in the limiting static case ( =0), the spring coefficients of the SDOF oscillators have to satisfy

normalized and listed in matrix form as

Mz + Cz + Kz = 1(
where

p ) NK s

(9)

1 + 1 + k1 k2

+ 1 = 1 kN Ks

(7)

a2 0 1 0 M = 12 0

0 a2 0 2

0 ; a2 0 N 1 0 0 a2 0 N

For convenience, all the spring coefficients of the model are taken in this study as k 1 = k 2=...= k N= NK s (8)

Restrained by Eq. (8), Eq. (6) can be further

686

Journal of the Chinese Institute of Engineers, Vol. 24, No. 6 (2001)

a 0 1 a 0 1 0 a 0 1 a 0( 1 a 0 2 + 2) 0 a 0 2 1 C= 0 0

0 ; a 0( N 2 a 0 N 1 + N 1) a 0 N 1 a 0( N 1 + N)
(10)

of the i -th SDOF oscillator, respectively. It is implied in Eq. (10) that there remain totally 2N parameters to be selected for the lumped-parameter model in matching the frequency dependency of the actual foundation compliance function. Taking the Fourier transform of Eq. (9), the transfer function from the displacement z 1 of the first oscillator to the applied force p (i.e., the compliance function of the lumped-parameter model) can be solved as

F d () =

1 = Z 1() K d () P ()
(12)

1 1 0 1 2 1 K = 0 1 0 0

0 0 ; z= 2 1 1 2

z1 z2 zN 1 zN

1 0 ; 1= 0 0 0

T = 1 1 ( 2M + i C + K ) 11 NK s

where K d ( ) denotes the dynamic impedance function of the model and P ( ) and Z 1 ( ) represent the Fourier transforms of p ( t ) and z 1 ( t ), respectively. Equation (12) can be further simplified as

and

F d () Ks T = = 1 1 ( 2M + i C + K ) 11 Fs K d () N

i =

m iv 2 cv s and i = i s 2 NK sd NK sd

(11) where

H () T = 1 1 H () 1 = 11 N N

(13)

are the dimensionless mass and damping parameters

H () = ( M + i C + K ) 1
2
1

1 a2 1 a 0 1 0 0 1 + ia 0 1 2 a2 + ia ( + ) 1 ia 0 1 0 2 0 1 2 = 0 0 0 1 ia 0 N 1

0 0 1 ia 0 N 1 2 2 a 0 N + ia 0( N 1 + N )

(14)

and H 11( ) denotes the element of the first row and the first column in the matrix H ( ). The specific formulae for the normalized compliance function in terms of the dimensionless parameters are listed in the appendix for lumped-parameter models consisting of one, two and three SDOF oscillators. From Eq. (13) and the expression for H () in Eq. (14), it is easy to show that the normalized compliance

function for the model approaches zero in the limiting high-frequency case (a0), which resembles the high-frequency behavior of the actual soil system. Combination of this high-frequency convergence and the static convergence guaranteed by Eq. (8) automatically leads to doubly asymptotic convergence for these lumped-parameter models. For corresponding to the normalized foundation compliance of the

W.H. Wu and C.Y. Chen: Simple Lumped-Parameter Models of Foundation Using Mass-Spring-Dashpot Oscillators

687

actual soil system as expressed in Eq. (5), it is appropriate to reformulate Eq. (13) into

F d (a 0) = 1 e i (a 0) Fs (a 0)
where

(15)

1 = Amp{H 11(a 0)} and (a ) = Ang{H 11(a 0)} 0 N N (a 0)


(16)

Using the modified expression proposed in this study to entirely describe the compliance function in the complex plane with diminishing high-frequency amplitude, the deviation between the model and the actual soil system can be appropriately taken as certain measures on the relative errors of their corresponding complex compliance values. Directly taking the amplitude of the relative error in the complex plane, the total error function in this study is consequently defined as
2 L L F d (a 0j ) K d (a 0j ) 1 = 1 F d (a 0j ) j = 1 K (a ) d 0j 2

and Amp { . } and Ang { . } signify the amplitude and the phase angle of a complex-valued quantity, respectively. Comparing the results from Eqs. (5) and (15), the optimization process is then applied to determine the dimensionless parameters of the model through minimizing the discrepancy between the normalized compliance function for the model and that for the actual system. With these dimensionless parameters is and is available, the correlative equation of Eq. (11)

j =1

j =1

(a 0j ) i [(a ) (a )] 0j 0j 1 e (a 0j )
L

=L +

{[ 0j ]2 2[(a 0j )]cos[ (a 0j ) (a 0j )]} j = 1 (a )


0j 0j

(a )

(a )

(18) where L symbolizes the selected number of different frequencies and a 0 j denotes the value of the dimensionless frequency parameter a 0 corresponding to the j -th selected frequency. Since L is constant, the objective function subject to minimization can be taken as

NK sd 2 NK d mi = i and c i = v s i 2 s vs

(17)

can be utilized to eventually decide the mass and damping coefficient for each SDOF oscillator of the lumped-parameter model. It should be reiterated that the models derived in this section are applicable to horizontal, vertical, rocking and torsional vibrations of foundation. Consequently, force and displacement are generalized to indicate the corresponding moment and rotational angle in rocking and torsional foundation vibrations, whereas mass implies the corresponding rocking or torsional moment of inertia in these two cases. 1. Definition of Error Function in Determining Optimal Parameters For determining the dimensionless parameters of the model, a total error function has to be defined as the objective function used in the optimization process. As deliberated in the previous section, the direct addition of the errors from the two independent components of dynamic impedance may lead to a biased optimization process emphasizing the accuracy of the component with larger magnitude. Adopting the relative error for each component, instead of the absolute error, to compose the total error function can lessen this problem. However, questions of rationality in engineering applications may also be raised with this definition where equal weightings are enforced for the two separate components.

{[ 0j ]2 2[(a 0j )]cos[ (a 0j ) (a 0j )]} j = 1 (a )


0j 0j

(a )

(a )

(19)

Eq. (19) essentially provides an appropriate solution to combine the two relative errors from the amplitude and phase coefficients with geometric rationality. Substituting Eq. (16) into Eq. (19), the objective function can be expressed as a nonlinear function of 2 N dimensionless parameters. Nonlinear optimization analysis is then applied to minimize this objective function for the determination of these 2 N parameters. IV. APPLICATIONS TO DIFFERENT TYPES OF FOUNDATION SYSTEMS The effectiveness of these new consistent lumped-parameter models is first demonstrated in this section by applying them to various cases of square and circular foundations, all resting on or embedded in an elastic half-space. By slightly modifying the elements in each SDOF oscillator, these models can also be generalized for application in cases where the material damping of soil is considered.

688

Journal of the Chinese Institute of Engineers, Vol. 24, No. 6 (2001)

Table 1 Optimal parameters for lumped-parameter model of surface square foundations Poissons ratio Parameter Horizontal N =1 N =2 Vertical N =1 N =2 N =1 .0334 .3579 . . . . Rocking N=2 N=3 .0174 .1020 .1081 .3493 . . .0625 .2269 .0201 -.3228 .1510 .5743 .0689 .2172 .0217 -.3340 .1551 .5890 N=1 .0199 .2917 . . . . .0198 .2938 . . . . Torsional N =2 N=3 .0184 .0412 .0892 .3445 . . .0182 .0429 .0899 .3449 . . .0467 .1804 .0230 -.3038 .1284 .5055 .0471 .1817 .0232 -.3047 .1295 .5077

v =1/3

1 1 2 2 3 3 1 1 2 2 2 3

.0038 .0026 .0233 -.0054 .7014 .3385 1.0830 .5894 . .2714 . .4653 . 1.0028 . 1.1623 . . . . . . . . -.0023 .6744 . . . .

v =0.45

.0030 .1160 -.0134 .0537 .0259 .3086 1.2153 1.0044 .3472 .1074 .2838 . .3253 . .1011 .9933 . .8887 . .3582 . . . . . . . . . .

Fig. 3 Comparison of compliance functions in the complex plane for a surface square foundation ( =1/3)

1. Surface Square Foundations Based on the values for the impedance functions of a surface square foundation recently calculated by Wong and Luco (2001) in the range of a0 from 0 to 5, the optimal parameters of various lumped-parameter models are determined and listed in Table 1 for the soil with Poissons ratio =1/3 and =0.45. For translational (horizontal and vertical) vibrations, two models consisting of one set of SDOF oscillators ( N =1) and two sets of SDOF oscillators ( N =2) are constructed; while three different models including

one to three sets of SDOF oscillators (N=1~3) are established for the rotational (rocking and torsional) vibrations. In discussing the following two types of foundation systems, the same group of lumped-parameter models will also be chosen for evaluation. Adopting the determined optimal parameters, the normalized compliance functions of the lumped-parameter models are plotted in Figs. 3 and 4 to compare with the corresponding values of the actual soil system. It is clearly observed that the lumped-parameter model using two sets of SDOF oscillators almost perfectly duplicates the actual foundation

W.H. Wu and C.Y. Chen: Simple Lumped-Parameter Models of Foundation Using Mass-Spring-Dashpot Oscillators

689

Table 2 Optimal parameters for lumped-parameter model of surface circular foundations Poissons ratio Parameter Horizontal N =1 N =2 .0047 .6310 . . . . -.0002 .5998 . . . . .0001 .3248 .1865 .8474 . . .0006 .2872 .2071 .8364 . . Vertical N =1 N =2 N =1 Rocking N=2 N=3 N =1 Torsional N=2 N=3 .0069 .0910 .0606 .2755 . . .0260 -.1584 .0064 .1587 .0814 .3948 .0260 -.1584 .0064 .1587 .0814 .3948

v =1/3

1 1 2 2 3 3 1 1 2 2 2 3

.0237 -.0019 .0189 .0078 1.0063 .5380 .3552 .1446 . .3933 . .0914 . 1.0068 . .2952 . . . . . . . . .1605 .0826 .0425 .0238 .8813 .4163 .2888 .0910 . .4546 . .0794 . 1.1865 . .3239 . . . . . . . .

.0427 .0115 -.2121 .2743 .0058 . .2074 . .1101 . .4612 . .0445 -.1904 .057 .1908 .1106 .4501

v =0.5

.0115 .0069 .2743 .0910 . .0606 . .2755 . . . .

Fig. 4 Comparison of compliance functions in the complex plane for a surface square foundation ( =0.45)

compliance functions in the cases of translational vibration. Moreover, even the degraded model using only one SDOF oscillator can practically represent the actual soil system in good accuracy. As for the cases of rotational vibration, the proposed models are not as effective because of the more intricate behavior inherited in the corresponding foundation compliance functions. However, acceptable accuracy is still illustrated for the model employing three sets of SDOF oscillators in these cases.

2. Surface Circular Foundations The case of circular foundations sitting on an elastic half-space is also evaluated. Numerical values for the vertical impedance functions presented by Veletsos and Tang (1987) in the range of a 0 from 0 to 6 are used as the continuum values. For the corresponding horizontal and rocking impedance functions, results listed by Veletsos and Wei (1971) in the range of a0 from 0 to 8 are adopted. In addition,

690

Journal of the Chinese Institute of Engineers, Vol. 24, No. 6 (2001)

Fig. 5 Comparison of compliance fucntions in the complex plane for a surface circular foundation ( =1/3)

Fig. 6 Comparison of compliance function sin the complex plane for a surface circular foundation ( =0.5)

the comparison for the torsional impedance functions is based on the values from Luco and Westmann (1971) in the range of a 0 from 0 to 8. The optimal parameters of different lumped-parameter models are determined and listed in Table 2 for the soil

with Poissons ratio =1/3 and =0.5. Since the torsional vibration of a circular foundation is an axisymmetric problem, its corresponding impedance function is independent of the soils Poissons ratio. Applying these determined optimal parameters,

W.H. Wu and C.Y. Chen: Simple Lumped-Parameter Models of Foundation Using Mass-Spring-Dashpot Oscillators

691

Table 3 Optimal parameters for lumped-parameter model of embedded square foundations (embedded ratio h / d =0.5) Poissons ratio Parameter Horizontal N =1 N =2 Vertical N =1 N =2 N =1 Rocking N=2 N=3 .1124 .2619 .0723 -.5076 .3023 .8572 .1061 .2483 .0766 -.5157 .2820 .8433 N =1 Torsional N=2 N=3

v =0.25

1 1 2 2 3 3 1 1 2 2 2 3

.0071 .0056 .0337 -.0153 .0544 .0473 1.0283 .5078 1.2716 .7306 .4092 .0336 . .5424 . .6116 . .1869 . 1.5615 . 1.5839 . .5425 . . . . . . . . . . . . .0069 .0071 .0612 -.0531 .0517 .0425 .9931 .4796 1.2817 .9743 .3963 .0311 . .5325 . .4638 . .1758 . 1.4958 . 1.2075 . .5195 . . . . . . . . . . . .

.0563 .0579 .0927 .3411 -.0362 .2030 . .1513 .0996 . .4993 -.5249 . . .2354 . . .7891 .0575 .0583 .0937 .3400 -.0382 .2068 . .1512 .0994 . .5013 -.5204 . . .2372 . . .7964

v =0.4

Table 4 Optimal parameters for lumped-parameter model of embedded square foundations (embedded ratio h / d =1.5) Poissons ratio Parameter Horizontal N =1 N =2 Vertical N =1 N =2 N =1 Rocking N=2 N=3 .1802 -.4527 .0257 .4532 .5454 1.2113 .1725 -.4419 .0266 .4326 .5116 1.1824 N =1 .0551 .5319 . . . . Torsional N=2 N=3 .0474 .0521 .2017 .5694 . . .1237 .2889 .0657 -.5058 .3380 .9033

v =0.25

1 1 2 2 3 3 1 1 2 2 2 3

.0360 -.0029 .0337 -.0096 .0624 .0370 1.5296 .8265 1.2716 .9288 .6388 .2181 . .9380 . 1.2705 . .3274 . 2.1130 . 2.1838 . .7544 . . . . . . . . . . . . .0936 -.0147 .0612 -.0033 .0742 .0358 1.5488 1.0498 1.2817 .9261 .6300 .2288 . .5880 . 1.1170 . .3047 . 1.8123 . 2.0284 . .7482 . . . . . . . . . . . .

v =0.4

.0569 .0493 .1229 .4299 .0495 .2806 . .2013 .0681 . .5714 -.5096 . . .3358 . . .8890

the normalized compliance functions of the lumpedparameter models are depicted in Figs. 5 and 6 to compare with the values of the corresponding continuum system. Similar accuracy is observed for various lumped-parameter models as in the case of square foundations. For translational foundation vibrations, the continuum values using the modified compliance expression virtually follow smooth halfcircle curves, which can be easily produced by the four-parameter or even two-parameter model. More elaborate continuum values for rotational vibrations, on the other hand, requires a model with more parameters to attain acceptable accuracy. This trend is particularly evident in the case of incompressible soil ( =0.5).

3. Embedded Square Foundations The effectiveness of the lumped-parameter model is further investigated for the case of square foundations embedded in an elastic half-space. The impedance functions reported by Mita and Luco (1989), which cover the values for a 0 from 0.1 to 3, are adopted as the fitting basis. The ratio between the embedded depth h and the half side-width of foundation d is usually used to characterize the embedment effects for this type of foundation. The optimal parameters of different lumped-parameter models are determined and listed in Tables 3 and 4 for the foundation with the embedment ratio h / d =0.5 and h / d = 1.5 and for the soil with Poissons ratio =0.25 and

692

Journal of the Chinese Institute of Engineers, Vol. 24, No. 6 (2001)

Fig. 7 Comparison of compliance functions in the complex plane for an embedded square foundation ( =0.25, h / d =0.5)

Fig. 8 Comparison of compliance functions in the complex plane for an embedded square foundation ( =0.4, h / d =0.5)

=0.4. Applying these determined optimal parameters, the normalized compliance functions of the lumped-parameter models are displayed in Figs. 7 to 10 to compare with the values of the corresponding continuum system. Similar accuracy is obtained for various lumped-parameter models as in the case of

surface foundations. 4. Nonnegative Parameters and Comparison with Other Models It is noteworthy from Tables 1 to 4 that

W.H. Wu and C.Y. Chen: Simple Lumped-Parameter Models of Foundation Using Mass-Spring-Dashpot Oscillators

693

Fig. 9 Comparison of compliance in the complex plane for an embedded square foundation ( =0.25, h / d =1.5)

Fig. 10 Comparison of compliance functions in the complex plane for an embedded square foundation ( =0.4, h / d =1.5)

negative optimal values of the dimensionless parameters may result and no physical realizations can be implemented using these values. Similar circumstances were found in some of the previously developed lumped-parameter models (Jean et al ., 1990;

Paronesso and Wolf, 1995). This is not surprising considering that all the parameters of the lumpedparameter model are selected for minimizing the objective function in a purely mathematical sense. Nevertheless, investigations also show that the

694

Journal of the Chinese Institute of Engineers, Vol. 24, No. 6 (2001)

Fig. 11 Comparison of the lumped-parameter models using the original and nonnegative optimal parameters for a surface circular foundation ( =1/3)

Fig. 12 Comparison of different lumped-parameter models for a surface circular foundation ( =1/3)

accuracy of this model is barely affected even if all the parameters are further restricted to nonnegative. Taking the case of a surface circular foundation sitting on the soil with =1/3 as an example, the nonnegative optimal parameters can be determined to be 1 =0, 1=0.5299, 2 =0.4063, and 2 =1.0154 for the vertical model using two SDOF oscillators and 1= 0.0079, 1=0.0364, 2=0.0331, 2=0.0995, 3=0.0934 and 3 =0.3589 for the rocking model using three SDOF oscillators. For demonstration, the lumpedparameter model adopting the original optimal parameters as listed in Table 2 is compared in Fig. 11 with the model using the nonnegative optimal parameters for the vertical and rocking vibrations. In addition, comparison with the other existing lumped-parameter models is also made in Fig. 12 for the new model to illustrate its advantages. The model developed by de Barros and Luco (1990) using five parameters is taken to compare with the new model employing two SDOF oscillators (i.e., four parameters) for the vertical compliance function of a surface circular foundation. It is clearly shown in part (a) of Fig. 12 that the new model achieves better accuracy even with fewer parameters. As for the rocking compliance function of a surface circular

foundation, the model proposed by Jean et al . (1990) requiring ten parameters is adopted in part (b) of Fig. 12 to compare with the new model employing four SDOF oscillators (i.e., eight parameters). The new model performs equally well or even better in the medium to high frequency range while minor inefficiency can be observed in the low frequency range. 5. Consideration of Material Damping for Soil In the above applications, the soil is assumed to be a perfectly elastic medium where the energy of vibration is dissipated only through radiation of waves toward infinity. To be more realistic, the material damping commonly existing in soil should be also included in the foundation vibration analysis. By application of the correspondence principle, the dynamic impedance functions for the cases considering soil damping can be computed from their corresponding functions for the elastic foundation. Rather than determining independent lumped-parameter models to represent the impedance functions of viscoelastic foundations, however, a more convenient way to construct the models for the viscoelastic foundation has been developed by Meek and Wolf (1994) by directly

W.H. Wu and C.Y. Chen: Simple Lumped-Parameter Models of Foundation Using Mass-Spring-Dashpot Oscillators

695

applying the correspondence principle to the discrete elements of the lumped-parameter models corresponding to the elastic foundation. It was shown that the Voigt viscoelasticity can be introduced through augmenting each original spring in the lumped-parameter model by a dashpot and each original dashpot by a mass, attached in a special way. Besides, more realistic nonlinear-hysteretic damping can also be characterized by replacing the augmenting dashpots and masses by frictional elements. Since the discrete elements of the new lumped-parameter model proposed in this study are no more than springs, dashpots, and masses, similar augmentation can be made to generalize the new model for its application in the foundation system incorporating soil damping. V. CONCLUSIONS

Fd

Fs H H 11 h K Kd
Kd

Ks ki Simple lumped-parameter models are developed in this study to generally represent the horizontal, vertical, rocking and torsional vibrations of rigid foundations. It is demonstrated for various surface and embedded foundations that these models can effectively and efficiently replace the frequency-dependent foundation-soil system. Compared with the other lumped-parameter models to achieve the same degree of accuracy, the models proposed in this paper have the advantages in requiring fewer parameters and featuring a more systematic expansion. Furthermore, it is also shown that the effectiveness of these models barely deteriorates for nearly incompressible soil with Poissons ratio close to 0.5, which has caused difficulties for a number of previously suggested models. As for their applications in different vibration modes, the performance of the new models is better in approximating the translational foundation vibrations than in the cases of rotational foundation vibration. ACKNOWLEDGEMENT This study was supported by the National Science Council, Republic of China, under Grant NSC 89-2211-E-224-003. This support is greatly appreciated. NOMENCLATURE a0 C ci d Fd dimensionless frequency parameter (= d / v s) normalized damping matrix of the lumped-parameter model, [ T ] damping coefficient of the i-th SDOF oscillator, [ M / T or ML 2/ T ] characteristic length of foundation [ L ] dynamic compliance function of foundation,
M

mi P p vs Zi zi z

i i

[ T 2/ M or T 2/ ML 2] dynamic compliance function of the lumpedparameter model, [ T 2/ M or T 2/ ML 2] static compliance function of foundation, [ T 2/ M or T 2/ ML 2] transfer matrix of the lumped-parameter model (=( 2M + i C + K ) 1)) element of the first row and the first column in H embedded depth of foundation, [ L ] normalized stiffness matrix of the lumped-parameter model dynamic impedance function of foundation, [ M / T 2 or ML2/ T 2] dynamic impedance function of the lumped-parameter model, [ M / T 2 or ML 2/ T 2] static impedance of foundation, [M / T 2 or ML2/ T 2] spring coefficient of the i -th SDOF oscillator, [ M / T 2 or ML2/ T 2] normalized mass matrix of the lumped-parameter model, [T 2] mass or area moment of inertia of the i -th SDOF oscillator, [ M or ML2] Fourier transform of p, [ M / T 2 or ML 2/ T ] applied force or moment, [ML / T 2 or ML2/ T 2] shear velocity of soil, [L / T ] excitation frequency, [1/ T ] Fourier transform of z i, [ LT ] absolute displacement of the i -th SDOF oscillator, [ L ] absolute displacement vector of the SDOF oscillators, [ L ] dimensionless stiffness coefficient dimensionless damping coefficient dimensionless mass parameter of the i-th SDOF m v2 oscillator (= i s 2 ) NK sd dimensionless damping parameter of the i -th cv SDOF oscillator (= i s ) NK sd dimensionless phase coefficient of foundation dimensionless phase coefficient of the lumpedparameter model Poissons ratio of soil dimensionless amplitude coefficient dimensionless amplitude coefficient of the lumped-parameter model total error function REFERENCES

1. Chopra, A. K., and Gutierrez, J. A., 1974, Earthquake Response Analysis of Multistorey Buildings Including Foundation Interaction, Earthquake Engineering and Structural Dynamics, Vol.

696

Journal of the Chinese Institute of Engineers, Vol. 24, No. 6 (2001)

3, No.1, pp. 65-77. 2. de Barros, F. C. P., and Luco, J. E., 1990, Discrete Models for Vertical Vibrations of Surface and Embedded Foundations, Earthquake Engineering and Structural Dynamics , Vol. 19, No. 2, pp. 289-303. 3. Gazetas, G., 1983, Analysis of Machine Foundation Vibrations: State of the Art, Soil Dynamics and Earthquake Engineering , Vol. 2, No. 1, pp. 2-42. 4. Gazetas, G., 1991, Formulas and Charts for Impedances of Surface and Embedded Foundations, Journal of Geotechnical Engineering , ASCE, Vol. 117, No. 9, pp. 1363-1381. 5. Jean, W. Y., Lin, T. W., and Penzien, J., 1990, System Parameters of Soil Foundations for Time Domain Dynamic Analysis, Earthquake Engineering and Structural Dynamics , Vol. 19, No. 2, pp. 541-553. 6. Jennings, P. C., and Bielak, J., 1973, Dynamics of Building Soil Interaction, Bulletin of the Seismological Society of America, Vol. 63, No. 1, pp. 9-48. 7. Luco, J. E., and Westmann, R. A., 1971, Dynamic Response of Circular Footings, J ournal of Engineering Mechanics , ASCE , Vol. 97, No. EM5, pp. 1381-1395. 8. Meek, J. W., and Wolf, J. P., 1992, Cone Models for Homogeneous Soil, Journal of Geotechnical Engineering, ASCE , Vol. 118, No. 5, pp. 667-685. 9. Meek, J. W., and Wolf, J. P., 1993, Cone Models for Nearly Incompressible Soil, Earthquake Engineering and Structural Dynamics , Vol. 22, No. 5, pp. 649-663. 10. Meek, J. W., and Wolf, J. P., 1994, Material Damping for Lumped-parameter Models of Foundations, Earthquake Engineering and Structural Dynamics, Vol. 23, No. 4, pp. 349-362. 11. Mita, A., and Luco, J. E., 1989, Impedance Functions and Input Motions for Embedded Square Foundations, Journal of Geotechnical Engineering, ASCE , Vol. 115, No. 4, pp. 491-503. 12. Nogami, T., and Konagai, K., 1986, Time Domain Axial Response of Dynamically Loaded Single Piles, Journal of Engineering Mechanics, ASCE , Vol. 112, No. 11, pp. 1241-1249. 13. Parmelee, R. A., Perelman, D. S., and Lee, S. L., 1969, Seismic Response of Multiple-story Structures on Flexible Foundations, Bulletin of the Seismological Society of America , Vol. 59, No. 7, pp. 1061-1070. 14. Paronesso, A., and Wolf, J. P., 1995, Global Lumped-parameter Model with Physical Representation for Unbounded Medium, Earthquake Engineering and Structural Dynamics , Vol. 24,

No. 5, pp. 637-654. 15. Perelman, D. S., Parmelee, R. A., and Lee, S. L., 1968, Seismic Response of Single-story Interaction Systems, Journal of Structural Division, ASCE , Vol. 94, No. 12, pp. 2597-2608. 16. Veletsos, A. S., and Nair, D. V. V., 1974a, Torsional Vibration of Viscoelastic Foundations, Journal of Geotechnical Engineering, ASCE, Vol. 100, No. GT3, pp. 225-246. 17. Veletsos, A. S., and Nair, D. V. V., 1974b, Response of Torsionally Excited Foundations, Journal of Geotechnical Engineering, ASCE, Vol. 100, No. GT4, pp. 476-482. 18. Veletsos, A. S., and Tang, Y., 1987, Vertical Vibration of Ring Foundations, Earthquake Engineering and Structural Dynamics, Vol. 15, No. 1, pp. 1-21. 19. Veletsos, A. S., and Wei, Y. T., 1971, Lateral and Rocking Vibration of Footings, Journal of Soil Mechanics and Foundation Division, ASCE, Vol. 97, No. SM9, pp. 1227-1248. 20. Wolf, J. P., 1991a, Consistent Lumped-parameter Models for Unbounded Soil: Physical Representation, Earthquake Engineering and Structural Dynamics , Vol. 20, No. 1, pp. 11-32. 21. Wolf, J. P., 1991b, Consistent Lumped-parameter Models for Unbounded Soil: Frequencyindependent Stiffness, Damping and Mass Matrices, Earthquake Engineering and Structural Dynamics , Vol. 20, No. 1, pp. 33-41. 22. Wolf, J. P., 1994, Foundation Vibration Analysis Using Simple Physical Models , Prentice-Hall, Englewood Cliffs, NJ. 23. Wolf, J. P., and Somaini, D. R., 1986, Approximate Dynamic Model of Embedded Foundation in Time Domain, Earthquake Engineering and Structural Dynamics, Vol. 14, No. 5, pp. 683-703. 24. Wong, H. L., and Luco, J. E., Impedance Functions for Rectangular Foundations on a Viscoelastic Half-space, submitted to Journal of Geotechnical and Environmental Engineering, ASCE . 25. Wu, W. H., 1998, A Uniform Representation for Dynamic Impedance Functions of Rectangular Foundations Incorporation Different Aspect Ratios, Journal of Science and Technology , Vol. 7, No. 3, pp. 259-270. 26. Wu, W. H., Ke, T. C., and Kao, H. Y., 2000, Effects of Foundation Flexibility on Dynamic Impedance Functions - Two-dimensional Analysis, Journal of the Chinese Institute of Civil and Hydraulic Engineering , Vol. 12, No. 3, pp. 475-485. APPENDIX For the lumped-parameter models consisting of

W.H. Wu and C.Y. Chen: Simple Lumped-Parameter Models of Foundation Using Mass-Spring-Dashpot Oscillators

697

different number of SDOF oscillators, the corresponding normalized compliance functions can be expressed in terms of the dimensionless parameters as the following: (a) the model including one SDOF oscillator

1 a2 Fd (a 0) 0 1 i (a 0 1) = 2 2 2 Fs 1 + a 0( 1 2 1) + a 4 0 1
(b) the model including two SDOF oscillators

( 2 + a2 Fd (a 0) 0 2) ia 0( 1 + 2) = 2 4 Fs 2{[1 a 0(2 1 + 2 + 1 2) + a 0 1 2] + i [a 0( 1 + 2) a 3 0( 1 1 + 1 2 + 2 1)]}

(c) the model including three SDOF oscillators

a4 0[ 1 2( 1 + 2 + 3) + 1 3( 1 + 2) + 2 3 1 + 2 1( 2 + 3) + 2 3] + a 6 0 1 2 3} 3i { a 0( 1 + 2 + 3) + a3 0[2 1( 1 + 2 + 3) + 2(2 1 + 2 + 3) + 3( 1 + 2) + 1 2 3] a 5 0[ 1 2( 2 + 3)
+ 1 3( 1 + 2) + 2 3 1]}
Manuscript Received: Nov. 27, 2000 Revision Received: Jan. 09, 2001 and Accepted: May 11, 2001

Fd (a 0) N = D Fs
where
4 N = [3 a 2 0(2 3 + 2 2 + 1 2 + 1 3 + 2 3) + a 0 2 3]

+ i {2a 0( 1 + 2 + 3) a 3 0[ 2( 2 + 3) + 3( 1 + 2)]}
and
D = 3{ 1 + a 2 0[3 1 + 2 2 + 3 + 1( 2 + 3) + 2 3]

!"#$%&'()*+,-./012 !
!"#$
!"#$%&'(

!"#$%&'()*+,-./01+23456789:8; !"#$%&'()*+,-./0123456789:;*<= !"#$%& '()*+,-./0 123456789:8(; !"#$%&'()*+,-./0123456789:;2<=> !"#$%&'()*+,'-./01234567894:*+, !"#$%& '()*+,-./012345 6789#:; !"#$%&'()*+ !"#$%&'()*+,-.(/0123,4

You might also like