Sunteți pe pagina 1din 12

WAVE-PROPAGATION FORMULATION OF SEISMIC RESPONSE OF

MULTISTORY BUILDINGS
By Erdal Şafak,1 Member, ASCE

ABSTRACT: This paper presents a discrete-time wave-propagation method to calculate the seismic response of
multistory buildings, founded on layered soil media and subjected to vertically propagating shear waves. Build-
ings are modeled as an extension of the layered soil media by considering each story as another layer in the
wave-propagation path. The seismic response is expressed in terms of wave travel times between the layers and
wave reflection and transmission coefficients at layer interfaces. The method accounts for the filtering effects of
the concentrated foundation and floor masses. Compared with commonly used vibration formulation, the wave-
propagation formulation provides several advantages, including simplicity, improved accuracy, better represen-
tation of damping, the ability to incorporate the soil layers under the foundation, and providing better tools for
Downloaded from ascelibrary.org by Ryerson University on 01/05/19. Copyright ASCE. For personal use only; all rights reserved.

identification and damage detection from seismic records. Examples are presented to show the versatility and
the superiority of the method.

INTRODUCTION masses. For simplicity, we assume that the buildings are sym-
metric, deform only in shear, and do not slide or rock. There-
During an earthquake, fault rupture initiates seismic waves fore, there are no torsional motions, and the floors are rigid
that are transmitted outward in all directions by the Earth’s and do not rotate. With respect to soil-structure interaction, the
layers. The waves that reach to the surface are reflected back model incorporates the effect of finite rigidity in the soil (i.e.,
into the Earth when the surface is free. If there is a structure the effect of energy radiated back into the soil) but not the
on the surface, however, the waves continue to propagate into effects of rocking and the 2D scattering.
the structure causing the structure to vibrate. In other words, The simplifying assumptions made above limit the appli-
the vibrations of a structure during an earthquake are caused cability of the method to only certain types of buildings and
by the propagation of seismic waves into the structure. responses. However, the assumptions do not necessarily rep-
In general, seismologists use the wave-propagation ap- resent a limitation of the methodology and are made primarily
proach to calculate ground motions generated by earthquakes, to simplify the equations. The main objective of this paper is
whereas engineers prefer the vibration approach to calculate to introduce the basic methodology for a simple case, so that
the response of structures induced by earthquakes. In theory, it can be expanded to more complicated and realistic cases.
the vibration and wave-propagation approaches represent two
alternative solution schemes of the same problem. In structural WAVE PROPAGATION IN MULTISTORY BUILDINGS
engineering, vibration methods, such as the modal analysis,
have been the standard approach to calculate the seismic re- Fig. 1(a) shows a multistory building resting on a multilay-
sponse of buildings. Wave-propagation methods are used ered soil medium and subjected to vertically propagating shear
mainly for structures that can be modeled as a continuous me- waves. The seismic waves propagating through the soil and
dium [e.g., Clough and Penzien (1975), Todorovska and Lee the building can be separated into the upgoing and the down-
(1989), Todorovska and Trifunac (1989), Hall et al. (1995), going components, as shown in the figure. Whenever the up-
and Iwan (1997)]. There is a small number of references avail- going and downgoing waves cross a layer interface they are
able on applications of wave-propagation methods to discrete partly reflected and partly transmitted into the next layer. The
structures. In a series of papers, Uzgider and Aydog̃an (1986) reflections and transmissions are characterized by the reflec-
and Aydog̃an and Uzgider (1988) presented a simplified wave- tion and transmission coefficients. This is schematically shown
propagation methodology to calculate the undamped response in Fig. 1(b), where u denotes the upgoing wave, and Ru and
of frame and masonry structures. The theory of wave propa- Tu denote the reflection and transmission coefficients for the
gation and scattering for slender and truss-type structural net- upgoing wave, respectively. Similarly, d denotes the down-
works is given in Cai and Lin (1991a,b) and Yong and Lin going wave, and Rd and Td denote the corresponding reflection
(1992). The text by Doyle (1989) presents a rigorous analysis and transmission coefficients for the downgoing wave, respec-
of wave propagation in structures by using fast Fourier trans- tively.
form-based spectral techniques. We derive the wave-propagation equations by using the dis-
This paper presents a discrete-time domain formulation of crete-time formulation that is commonly used by researchers
the propagation of seismic waves in multistory buildings that in oil exploration [e.g., Claerbout (1976) and Robinson and
are founded on layered soil media. We model the building as Treitel (1980)]. Consider the three consecutive layers, layers j
an extension of the layered soil medium by considering each ⫺ 1, j, and j ⫹ 1, as schematically shown in Fig. 1(c). Let
story in the building as another layer and derive the equations uj (t) and dj (t) represent the amplitudes at discrete time t of the
for vertically propagating plane shear waves. The equations upgoing wave at the top and the downgoing wave at the bot-
incorporate the damping in the soil and the building, as well tom, respectively, of layer j. We can show from Fig. 1(c) that
as the filtering effects of the concentrated foundation and floor the upgoing wave uj (t) in layer j is composed of the reflected
portion of the downgoing wave from the bottom, plus the
1
Res. Struct. Engr., U.S. Geological Survey, Box 25046, MS 966, Den- transmitted portion of the upgoing wave from the layer below.
ver, CO 80225. This relationship can be expressed by the following equation:
Note. Associate Editor: Chia-Ming Uang. Discussion open until Sep-
tember 1, 1999. To extend the closing date one month, a written request
must be filed with the ASCE Manager of Journals. The manuscript for uj (t) = R d, j⫺1 ⭈ dj (t ⫺ ␶j ) ⫹ Tu, j⫺1 ⭈ uj⫺1(t ⫺ ␶j ) (1a)
this paper was submitted for review and possible publication on Novem-
ber 17, 1997. This paper is part of the Journal of Structural Engineer- where R d, j⫺1 and Tu, j⫺1 denote the reflection coefficient for the
ing, Vol. 125, No. 4, April, 1999. 䉷ASCE, ISSN 0733-9445/99/0004- downgoing waves and the transmission coefficient for the up-
0426 – 0437/$8.00 ⫹ $.50 per page. Paper No. 17011. going waves, respectively, at interface j ⫺ 1. ␶j is the one-way
426 / JOURNAL OF STRUCTURAL ENGINEERING / APRIL 1999

J. Struct. Eng., 1999, 125(4): 426-437


Downloaded from ascelibrary.org by Ryerson University on 01/05/19. Copyright ASCE. For personal use only; all rights reserved.

FIG. 1. Bedrock-Soil-Building System: (a) Layers, Interfaces, and Upgoing and Downgoing Waves; (b) Reflection and Transmission
of Upgoing and Downgoing Waves; (c) Three Consecutive Layers with Upgoing and Downgoing Waves

travel time of the waves in layer j. Similarly, we note that the The continuity of displacements at the floor slab requires that
downgoing wave dj (t) in layer j is composed of the reflected y j,top(t) = y j,bot(t); therefore, either (4a) or (4b) can be used to
portion of the upgoing wave from the top, plus the transmitted represent the motions of the floors.
portion of the downgoing wave from the layer above. This can If the sampling interval of the incident bedrock motion x0(t)
be expressed by is small enough such that the travel times ␶j in the layers are
integer multiples of the sampling interval, the wave-propaga-
dj (t) = R u, j ⭈ uj (t ⫺ ␶j ) ⫹ Td, j ⭈ dj⫹1(t ⫺ ␶j ) (1b) tion equations all become finite-difference (i.e., discrete-time
filtering) equations. The soil-building system is defined in the
where R u, j and Td, j denote the reflection coefficient for the up- equations by the wave travel times in the layers, and the wave
going waves and the transmission coefficient for the down- reflection and transmission coefficients at the layer interfaces,
going waves, respectively, at interface j. Eqs. (1a) and (1b) which can all be determined uniquely from the physical char-
are valid for all intermediate soil and building layers. For the acteristics of the building and the soil (we will add the damp-
first and the last layers, we modify the equations to incorporate ing in the equations later). For a given soil-building system
the boundary conditions. For the first layer, which is the bot- and the bedrock motion, the equations can be solved recur-
tom soil layer next to the bedrock [Fig. 1(a)], we can show sively starting from the bedrock [i.e., with (2a)] and continu-
that ing upward. Note that the initial values of uj (t) and dj (t) are
zero until the first arrival of the waves in layer j. That is
u1(t) = R d,0 ⭈ d1(t ⫺ ␶1) ⫹ Tu,0 ⭈ x0(t ⫺ ␶1) (2a)


j

d1(t) = R u,1 ⭈ u1(t ⫺ ␶1) ⫹ Td,1 ⭈ d2(t ⫺ ␶1) (2b) uj (t) = 0 for t < ␶k (5a)
k=1


where x0(t) = input seismic wave at the bedrock-soil interface. j
It is assumed that there are no reflections from the bottom of
dj (t) = 0 for t < ␶j ⫹ ␶k (5b)
bedrock. For the last layer [layer m ⫹ N in Fig. 1(a)], which k=1
is the top story of the building, we note that there are no
downgoing waves transmitted from the layer above. Thus We should mention that in the wave-propagation formulation
either accelerations, velocities, or displacements can be used
um⫹N(t) = R d,m⫹N⫺1 ⭈ dm⫹N (t ⫺ ␶m⫹N ) ⫹ Tu,m⫹N⫺1 ⭈ um⫹N⫺1(t ⫺ ␶m⫹N ) as input. Calculated response is of the same type as the input,
(3a) and the response in the building represents the absolute re-
sponse, not the relative response with respect to ground.
dm⫹N (t) = R u,m⫹N ⭈ um⫹N (t ⫺ ␶m⫹N) (3b)
WAVE TRAVEL TIMES
Eqs. (1) – (3) provide a complete description of the wave prop-
agation in a soil-building system. Wave travel times in the layers are calculated by simply
The actual response of any floor in the building, or any dividing the layer thickness with the wave velocity in each
interface in the soil medium, can be calculated by combining layer. The shear-wave velocities for soil are generally deter-
the upgoing and downgoing waves. For interface j in the build- mined by field tests and are available in the literature for a
ing, for example, the response at the top and the bottom of large number of different soil types. Theoretically, the 1D
the floor slab is shear-wave velocity v s in an elastic soil medium is v s =
兹G/␳, where G and ␳ are the shear modulus and the mass
y j,top(t) = uj⫹1(t ⫺ ␶j⫹1) ⫹ dj⫹1(t) (4a) density, respectively, of the soil. In buildings, if we assume
that the vertical segment between the two floor slabs is fairly
y j,bot(t) = uj (t) ⫹ dj (t ⫺ ␶j ) (4b) uniform (in terms of its mass and stiffness), we can approxi-
JOURNAL OF STRUCTURAL ENGINEERING / APRIL 1999 / 427

J. Struct. Eng., 1999, 125(4): 426-437


mate the shear-wave velocity v b in that story as v b = 兹Sh/␥ transmitted components by using the conditions that the dis-
(Clough and Penzien 1975), where S, h, and ␥ are the shear placements above and below the interface are equal and the
stiffness, height, and the mass per unit length of the story, shear forces are in equilibrium. The inertia forces acting on
respectively. ␥ represents the mass per unit length of the ver- the concentrated masses of the foundation and the building
tical segment between the floors but does not include any of floors should be included in the shear equilibrium equations
the concentrated floor masses at the top or the bottom of the for those interfaces. More detail of these calculations can be
story. If a vertical segment between the floors is not uniform found in Şafak (1997) and in textbooks on wave propagation
(e.g., a large window opening, a sharp stiffness change at [e.g., Ewing et al. (1957) and Doyle (1989)]. The coefficients
midstory, etc.), we can always divide it into smaller vertical for all interfaces can be summarized by the following equa-
segments until each segment is approximately uniform. In tions:
other words, we represent such stories with more than one
layer. More precise methods of estimating shear-wave veloc- Ij ⫺ Ij⫹1 ⫺ i ⭈ 2␲ f ⭈ mj
R u, j = , Tu, j = 1 ⫹ R u, j (6a)
ities in reinforced concrete frame buildings can be found in Ij ⫹ Ij⫹1 ⫹ i ⭈ 2␲ f ⭈ mj
Todorovska et al. (1988).
Ij⫹1 ⫺ Ij ⫺ i ⭈ 2␲ f ⭈ mj
Td, j = 1 ⫹ R d, j
Downloaded from ascelibrary.org by Ryerson University on 01/05/19. Copyright ASCE. For personal use only; all rights reserved.

R d, j = , (6b)
REFLECTION AND TRANSMISSION COEFFICIENTS Ij ⫹ Ij⫹1 ⫹ i ⭈ 2␲ f ⭈ mj
When the upgoing and downgoing waves cross from one where Ij = impedance of layer j; and mj denotes the concen-
layer to another, their amplitudes and phases are altered be- trated mass at interface j. For soil layers, Ij = ␳j v j , where ␳j
cause of the reflections at the layer interface. These alterations and v j are the mass density and the shear-wave velocity in the
are characterized by the reflection and transmission coeffi- layer, respectively; and mj ⬅ 0. For building layers, Ij = ␥ j v j ,
cients and are different for upgoing and downgoing waves. If with ␥ j and v j denoting the mass per unit length and the wave
the interface has no concentrated mass, such as the bedrock- velocity, respectively, in layer j. At the soil-foundation inter-
soil or soil-soil interfaces, the coefficients are constants. When face [interface m in Fig. 1(a)], we take Im = ␳ m v m Afn, where
the interface has a concentrated mass, such as the foundation Afn is the contact area of the foundation. Note for the bedrock-
or the building floors, the coefficients become functions of soil and soil-soil interfaces that since mj ⬅ 0 the reflection and
frequency. transmission coefficients are constants. For the foundation and
To derive the equations for the reflection and transmission the building floors the coefficients are frequency dependent
coefficients, we subject the layer interface to an incident plane because mj ≠ 0.
wave of a specified frequency and calculate its reflected and Fig. 2 shows the amplitude and the phase of the reflection

FIG. 2. Amplitudes and Phases of Reflection Coefficients for Upgoing and Downgoing Waves at Foundation-Soil Interface for Typical
Bounds of a ⴝ Imⴙ1 /Im and b ⴝ mf /Im

428 / JOURNAL OF STRUCTURAL ENGINEERING / APRIL 1999

J. Struct. Eng., 1999, 125(4): 426-437


coefficients for upgoing and downgoing waves at the soil- presence of the building does not influence the motions at the
foundation interface for typical bounds of a = Im⫹1 /Im and b = top of the soil layer, and they are practically equal to those at
mf /Im. For small a and b values (i.e., small first-story stiffness, a free field. For large a values, the reflection coefficient be-
small foundation mass, large contact area, or hard soil) the comes much less than one at low frequencies and, depending
amplitude and the phase of the reflection coefficient for the on the value of b, gradually approaches one as the frequency
upgoing waves approach one and zero, respectively, which are increases. In other words, the low frequency components are
the values for a free surface reflection. In other words, the transmitted to the building, while most of the high-frequency
Downloaded from ascelibrary.org by Ryerson University on 01/05/19. Copyright ASCE. For personal use only; all rights reserved.

FIG. 3. Amplitudes and Phases of Reflection Coefficients for Upgoing and Downgoing Waves at Building Floor for Typical Bounds of
a ⴝ Ijⴙ1 /Ij and b ⴝ mj /Ij

TABLE 1. Characteristics of 10-Story Building and Soil Layers Used in Examples


Story mass
(10⫺4 ⭈ kg) Floor (or foundation) Story height or
Building story or Shear stiffness or soil density mass above Wave velocity layer thickness
soil layer (10⫺7 ⭈ N/m) (kg/m3) (10⫺4 ⭈ kg) (m/s) (m)
(1) (2) (3) (4) (5) (6)
Story 10 4.0 1.000 2.000 181 3.5
Story 9 4.5 1.875 1.750 177 3.5
Story 8 4.5 1.750 1.750 177 3.5
Story 7 5.5 1.750 1.750 196 3.5
Story 6 5.5 1.750 1.750 196 3.5
Story 5 6.5 1.750 1.750 213 3.5
Story 4 6.5 1.750 1.750 213 3.5
Story 3 7.5 1.750 1.750 229 3.5
Story 2 7.5 1.750 1.750 229 3.5
Story 1 5.0 1.625 2.250 225 4.5
Soil layer 1 — 2,200 8.634a 200 20.0
Soil layer 2 — 2,500 — 500 30.0
Bedrock — 2,800 — 2,000 —
a
Foundation.

JOURNAL OF STRUCTURAL ENGINEERING / APRIL 1999 / 429

J. Struct. Eng., 1999, 125(4): 426-437


components are blocked. Fig. 3 gives the amplitude and the DAMPING
phase of the reflection coefficients in a building floor for typ-
Damping gives a measure of energy loss per cycle and is
ical bounds of a = Ij⫹1 /Ij and b = mj /Ij . Again, the figure shows
present in all oscillatory systems. For soils the energy loss due
that the concentrated floor masses have the effect of a low-
to damping can be represented by the following simple atten-
pass filter on the upgoing and downgoing waves crossing the
uation equation [see Aki and Richards (1980, Chapter 5.5)]:
interface.
The reflection and transmission coefficients derived above As( f ) = e⫺␲␶ f /Q (7)
are for shear waves, because we assumed that the building
deforms mainly in shear. If the bending deformations are sig- where As ( f ) represents the reduction in the amplitude of a
nificant, we have to use the reflection and transmission coef- sinusoidal wave of frequency f when it travels a distance of
ficients calculated for flexural waves. Flexural waves include travel time ␶. The damping is defined by Q, known as the
four terms, two complex-valued propagating terms (same as quality factor (Knopoff 1964). 1/Q gives the measure of en-
the shear waves) plus two real-valued ringing terms. Four co- ergy loss per cycle and can be assumed to be independent of
efficients require four equations to determine the reflection and frequency for many cases (Kjartansson 1979). The logarithmic
transmission coefficients. The equations are obtained by writ- decrement (i.e., the natural logarithm of the ratio of amplitudes
Downloaded from ascelibrary.org by Ryerson University on 01/05/19. Copyright ASCE. For personal use only; all rights reserved.

ing the equality of the displacements and rotations and the of two successive maxima) for the solid friction model is ␲/
equilibrium of the shear forces and the bending moments at Q. By comparing this with the logarithmic decrement of free
the interface. vibrations under viscous damping, we can show that ␰ = 1/
2Q, where ␰ is the viscous damping ratio. The exponent of the
TABLE 2. Modal Frequencies and Damping Ratios of Fixed- attenuation function for viscous damping is proportional to f 2,
Based Building whereas for solid friction it is proportional to f, as seen in (7).
In this paper, the damping in each soil layer will be defined
Modal frequency
by its Q value.
Mode number (Hz) Modal damping ratio
For buildings, the damping has generally been assumed to
(1) (2) (3)
be of viscous type and proportional to the mass or the stiffness
1 0.97 0.05 of the building. The assumption of proportionality is mainly
2 2.66 0.05
3 4.37 0.07
for mathematical convenience so that we can have real-valued
4 6.00 0.09 frequencies and mode shapes when using modal analysis. In
5 7.42 0.11 the wave-propagation formulation, these assumptions are not
6 8.82 0.13 needed because we can specify the damping independently for
7 9.84 0.14 each story. Experiments show that for cyclically stressed ma-
8 10.81 0.15 terials, the energy loss per cycle is independent of frequency
9 11.89 0.17
10 13.18 0.18
and proportional to the square of vibration amplitude (Thom-
son 1972). This type of damping is called the structural damp-

FIG. 4. Impulsive and Recorded Ground Motions Used as Inputs in Examples

430 / JOURNAL OF STRUCTURAL ENGINEERING / APRIL 1999

J. Struct. Eng., 1999, 125(4): 426-437


ing and is represented by the structural damping factor ␭. The shifts. A frequency-domain technique might be more appro-
attenuation function for structural damping is identical to that priate to solve such equations. The frequency-domain solution
for solid friction; except for notation, we use ␭ instead of 1/ methods will be the subject of another paper. In this paper, we
Q in the exponent of (7). It can be shown by comparing the will solve the equations by using recursive, discrete-time do-
resonance amplitudes of an oscillator under harmonic loads main methods, which are simpler and easier to program than
that at the resonance frequency the structural damping factor the frequency-domain methods.
is equal to twice the viscous damping factor. It should be re- First, we need to write the damped version of the wave-
membered, however, that for viscous damping the energy loss propagation equations given by (1) – (3). To incorporate damp-
per cycle depends on the frequency, whereas for structural ing, we multiply the upgoing and downgoing waves in each
damping it is independent of frequency. layer with the corresponding attenuation function for damping
The simplest way to incorporate damping in the equations in the layer. When this is done, the equations for layer j be-
is to make shear stiffnesses of the layers complex quantities. come
For soil layers, we can approximate the damping by multiply-
ing the wave velocities with (1 ⫺ 0.5i/Q) (Şafak 1995). For uj (t) = Aj ( f ) ⭈ [R d, j⫺1 ⭈ dj (t ⫺ ␶j ) ⫹Tu, j⫺1 ⭈ uj⫺1(t ⫺ ␶j )] (8a)
building layers, we make the stiffness complex by multiplying dj (t) = Aj ( f ) ⭈ [R u, j ⭈ uj (t ⫺ ␶j ) ⫹ Td, j ⭈ dj⫹1(t ⫺ ␶j )]
Downloaded from ascelibrary.org by Ryerson University on 01/05/19. Copyright ASCE. For personal use only; all rights reserved.

(8b)
the original stiffness with (1 ⫹ i ⭈ ␭) in each layer (Thomson
1972). Complex stiffnesses result in complex wave velocities. where Aj ( f ) denotes the attenuation function for layer j. Al-
though these equations are in the discrete-time domain, the
SOLUTION OF WAVE-PROPAGATION EQUATIONS reflection and transmission coefficients for the foundation and
the building, as well as the attenuation function due to damp-
As stated earlier, for a given building and bedrock motion, ing, are all frequency dependent. Therefore, to solve the equa-
the wave propagation equations [(1) – (3)] can be solved re- tions recursively in the time domain, we should first convert
cursively starting from the bedrock [i.e., with (2a)] and con- the frequency-dependent parameters into discrete-time filters.
tinuing upward. The initial values of u(t) and d(t) to start the Because the reflection and transmission coefficients, given
recursion are given by (5). If we choose to incorporate the by (6), include the term i(2␲ f ) explicitly, they can easily be
damping in the equations by making the shear stiffnesses com- converted into discrete-time filters by using the bilinear trans-
plex, the wave velocities and the travel times all become com- formation (Cadzow 1973). To do this we simply replace i(2␲f)
plex resulting in finite-difference equations with complex time with

FIG. 5. Calculated Upgoing and Downgoing Waves in Building and Soil Layers

JOURNAL OF STRUCTURAL ENGINEERING / APRIL 1999 / 431

J. Struct. Eng., 1999, 125(4): 426-437


2 1 ⫺ q⫺1 Similar conversions can be made for R d ( f ), Tu( f ), and Td ( f ).
i(2␲ f ) = ⭈ (9) The attenuation function for damping [(7)] does not have
⌬ 1 ⫹ q⫺1
the term i(2␲ f ) explicitly. However, it has the form of a low
where ⌬ = sampling interval in seconds; and q⫺1 = backward- pass filter and can be matched by a simple discrete-time filter
shift operator, defined as q⫺j x(t) = x(t ⫺ j). With the trans- of the following form (Şafak 1995):
formation, we can write each reflection and transmission co-
efficient as a first-order discrete-time filter. For example, R u, j 1 ⫺ ␣ 1 ⫹ q⫺1
A(q) = ⭈ (11a)
(q), the discrete-time equivalent of R u, j ( f ), is 2 1 ⫺ ␣q⫺1

␤0j ⫹ ␤1 j q⫺1 where


R u, j (q) = ⫺1 (10a)
1 ⫹ ␣1 j q
1 ⫺ 兹1 ⫺ cos2␪ Q⌬
␣= with ␪ = ln(2) (11b)
cos ␪ ␶
where the filter coefficients ␤0 j , ␤1 j , and ␣1 j are found to be
Downloaded from ascelibrary.org by Ryerson University on 01/05/19. Copyright ASCE. For personal use only; all rights reserved.

The parameter Q in (11b) should be replaced with 1/␭ if the


⌬(Ij ⫹ Ij⫹1) ⫺ 2mj ⌬(Ij ⫺ Ij⫹1) ⫺ 2mj layer belongs to the building. The derivation of (11) can be
␣1 j = ; ␤0 j = ;
⌬(Ij ⫹ Ij⫹1) ⫹ 2mj ⌬(Ij ⫹ Ij⫹1) ⫹ 2mj found in Appendix I in Şafak (1995).
By incorporating (10) and (11) in (8), we obtain the dis-
⌬(Ij ⫺ Ij⫹1) ⫹ 2mj crete-time forms of the wave-propagation equations as fol-
␤1 j =
⌬(Ij ⫹ Ij⫹1) ⫹ 2mj (10b) lows:

FIG. 6. Comparison of Foundation and Top-Floor Accelerations Calculated by Using Wave-Propagation and Modal Analyses for Stiff-
Soil, Light Foundation and Soft-Soil, Heavy Foundation Systems Subjected to Velocity Impulse

432 / JOURNAL OF STRUCTURAL ENGINEERING / APRIL 1999

J. Struct. Eng., 1999, 125(4): 426-437


冘 冘
2 2
EXAMPLES
uj (t) = A1k uj (t ⫺ k) ⫹ A2l dj (t ⫺ ␶j ⫺ l)
k=1 l=0 To illustrate an application of the methodology introduced


2
above, we will calculate the seismic response of a 10-story
building founded on a two-layered soil medium over the bed-
⫹ A3s uj⫺1(t ⫺ ␶j ⫺ s) rock. The characteristics of the building and the soil layers
s=0 (12a)
used in the calculations are given in Tables 1 and 2. The soil

冘 冘 layers selected represent a firm site. The mass of the building


2 2

dj (t) = B1k dj (t ⫺ k) ⫹ B2l uj (t ⫺ ␶j ⫺ l) is assumed to be divided equally between the floors and the
k=1 l=0 walls. The contact area of the foundation is Af = 400 m2, and


2 the structural damping factor for all stories of the building is
⫹ B3s dj⫹1(t ⫺ ␶j ⫺ s) ␭ = 0.10, which is equivalent to 5% viscous damping at res-
s=0 (12b) onant frequencies. The damping in the soil layers is defined
by Q = 30 and Q = 50 for the upper and the lower layers,
The coefficients A1k, . . . , B3s can all be calculated in terms of respectively.
Downloaded from ascelibrary.org by Ryerson University on 01/05/19. Copyright ASCE. For personal use only; all rights reserved.

the coefficients of the filters given by (10) and (11). Similar The building is first subjected to a seismic impulse at the
discrete forms can be written for the first and the top layers. bedrock-soil interface. The selected impulse is a one cycle,
There are several software packages with routines to convert two-s long, sinusoidal velocity pulse with a peak amplitude of
frequency domain filters into discrete-time filters [e.g., the m- 1.0 m/s. The velocity impulse, along with the corresponding
file invfreqz.m in MATLAB’s Signal Processing Toolbox (Sig- acceleration and displacement impulses are plotted in Fig. 4(a).
nal 1997a)]. The utilization of such routines greatly simplifies As stated earlier, the calculations can be performed by taking
the calculations. either accelerations, velocities, or displacements as the input.

FIG. 7. Comparison of Foundation and Top-Floor Accelerations Calculated by Using Wave-Propagation and Modal Analyses for Stiff-
Soil, Light Foundation and Soft-Soil, Heavy Foundation Systems Subjected to Recorded Earthquake

JOURNAL OF STRUCTURAL ENGINEERING / APRIL 1999 / 433

J. Struct. Eng., 1999, 125(4): 426-437


In the examples, we used the accelerations. The calculated dation by a factor of 10. This comparison is also given in Fig.
response represents the absolute accelerations. Fig. 5 shows 6(b). The figure shows that for soft soils and heavy founda-
the calculated upgoing and downgoing acceleration waves in tions the wave-propagation and modal analyses results differ
the building and soil layers. significantly. The primary reason for the differences is that the
To compare the results with that of the standard vibration modal analysis does not account for the energy loss due to
approach, we also calculated the response by using modal wave transmission from the building back into the soil.
analysis. For this, we first calculated the free-field ground mo- We made similar comparisons by changing the input to a
tions at the surface of the two-layer soil medium for the same recorded ground motion. Fig. 4(b) shows the input acceleration
bedrock motion. We then used the free-field motion as the time histories applied at the bedrock-soil interface. The com-
input to determine the fixed-based response of the same build- parisons of the foundation and top-floor accelerations calcu-
ing by using modal analysis. This is an approach that is com- lated from the wave-propagation analysis with those calculated
monly used in practice. The comparisons of the foundation from the modal analysis are given in Fig. 7 for both the stiff-
and top-floor accelerations calculated from the wave-propa- soil, light-foundation and the soft-soil, heavy-foundation cases.
gation analysis with those from the modal analysis are given Similar conclusions can be drawn from Fig. 7. The differences
in Fig. 6(a). Since the soil-structure system chosen for the between the wave-propagation and modal analyses are more
Downloaded from ascelibrary.org by Ryerson University on 01/05/19. Copyright ASCE. For personal use only; all rights reserved.

example represents a stiff-soil, light-foundation system, there significant for the impulsive input.
is not much difference between the results of the wave-prop- We also compared the foundation to top-floor transfer func-
agation and the modal analyses. The accelerations at the foun- tions, calculated by simply taking the top-floor-to-foundation
dation are basically identical to those at a free field. Recall ratio of the Fourier amplitude spectra of the response for all
that in modal analysis, the foundation accelerations are the four cases. Theoretically, the changes in the foundation and
free-field accelerations at the top of the two-layer soil medium. soil properties and the bedrock motions should not affect the
We made a similar comparison for a soft-soil, heavy-founda- foundation-to-top-floor transfer function. Therefore, the trans-
tion system, where we reduced the shear-wave velocity of the fer functions for all four cases should be identical. The transfer
top soil layer to 20 m/s and increased the mass of the foun- functions calculated from the modal analysis are plotted in Fig.

FIG. 8. Comparisons of Foundation-to-Roof Transfer Functions Calculated from Wave-Propagation and Modal Analyses for Four Dif-
ferent Systems

434 / JOURNAL OF STRUCTURAL ENGINEERING / APRIL 1999

J. Struct. Eng., 1999, 125(4): 426-437


8(a), and those from the wave-propagation analysis are plotted wall case match better with those from the modal analysis,
in Fig. 8(b). When calculating the transfer functions for the because in modal analysis the wall and floor masses are as-
modal analysis, we had to use a smoothing window to obtain sumed to be concentrated at the floor levels. As the walls be-
realistic amplitudes, because the high-frequency errors in the come heavier relative to the floors, more energy is put into the
response resulted in spurious large peaks. The transfer func- higher modes.
tions for the wave-propagation analysis did not need any
smoothing. Fig. 8 shows that the transfer functions obtained OTHER APPLICATIONS OF WAVE-PROPAGATION
from the wave-propagation analysis are much more accurate FORMULATION
than those obtained from the modal analysis, particularly at
higher modes. This confirms the superiority of the wave-prop- One of the important applications of the wave-propagation
agation analysis in higher modes. formulation presented above is the identification and damage
Another advantage of the wave-propagation analysis over detection in multistory buildings from vibration recordings.
the modal analysis is that it can account for the distribution Most of the methods suggested for identification and damage
of the building’s mass between the walls and the floors. In the detection are based on the identification and detection of
examples above we assumed that the story masses in the build- changes in modal characteristics, such as the modal frequen-
Downloaded from ascelibrary.org by Ryerson University on 01/05/19. Copyright ASCE. For personal use only; all rights reserved.

ing are equally divided between the walls and the floors. To cies, damping ratios, mode shapes, and modal participation
see the influence of the mass distribution, we calculated the factors. An extensive literature review of such methods can be
response assuming the floor masses are 20% (light floor, heavy found in Doebling et al. (1996). It is well known that the
walls) and 80% (heavy floor, light walls) of the total story modal parameters of multistory buildings are not good indi-
mass used in modal analysis. We considered the stiff-soil, cators for damage, because they are not sensitive to small and
light-foundation system and used the velocity impulse as the local changes in the building’s characteristics. The wave-prop-
input. The comparisons of the time histories and the Fourier agation approach can overcome some of the deficiencies of the
amplitude spectra of the calculated top-story accelerations, modal approach. Although the wave-propagation methods
along with those from the modal analysis, are given in Fig. 9. have been used to detect damages in small specimens (e.g.,
The figure shows that the results change significantly depend- plates and beams) in the laboratory, they have not been applied
ing on how the story masses are distributed between the floors to large, civil-engineering-type structures. The discrete-time
and walls. As expected, the results for the heavy-floor, light- formulation presented above provides convenient tools for

FIG. 9. Comparisons of Time Histories and Fourier Amplitudes of Top-Floor Accelerations Calculated by Assuming Different Mass
Distributions between Walls and Floors

JOURNAL OF STRUCTURAL ENGINEERING / APRIL 1999 / 435

J. Struct. Eng., 1999, 125(4): 426-437


identification and damage detection. In the formulation, we use waves and the other for the downgoing waves. They are solved
the wave travel times, reflection and transmission coefficients, recursively, starting with the first soil layer next to the bedrock.
and the structural damping factors to characterize the building. When compared to the commonly used vibration formulation,
These parameters can be identified from the recorded vibra- the wave-propagation formulation (1) is simpler to implement;
tions of the building if they are available from each floor. The (2) is more accurate at high frequencies; (3) incorporates the
damage is detected by investigating the changes in the param- damping and mass distribution more accurately; (4) can ac-
eters. When compared to modal analysis, the main advantages count for the energy absorptions by the soil layers under the
of this new methodology are that it is more accurate and much foundation; (5) provides better tools for identification and
simpler, the parameters used are more sensitive to damage, the damage detection; (6) allows utilization of digital simulation
identification and damage detection are most robust, and the software; and (7) can be extended to nonlinear systems.
method can account for the effects of soil layers underlying
the foundation. More detail on the methodology can be found
in Şafak (1998a). ACKNOWLEDGMENTS
In recent years, several powerful digital simulation soft-
The writer would like to thank Dr. W. B. Joyner of the U.S. Geological
Downloaded from ascelibrary.org by Ryerson University on 01/05/19. Copyright ASCE. For personal use only; all rights reserved.

ware packages have become available commercially [e.g., Survey, Menlo Park, Calif., Prof. R. Zhang of the Colorado School of
SIMULINK (1997b)]. These programs allow real-time simu- Mines, Golden, Colo., and the anonymous reviewers of ASCE for their
lation of linear and nonlinear systems under dynamic loads. reviews and comments.
The systems are modeled in the form of block diagrams, simi-
lar to those for electrical circuits, and simulated in the discrete-
time domain. The presented wave-propagation formulation APPENDIX. REFERENCES
provides a convenient form to use these programs to study
Aki, K., and Richards, P. G. (1980). Quantitative seismology: Theory and
seismic response of structures. We can observe the changes in methods, Vol. I, W. H. Freeman and Co., San Francisco.
response instantaneously (i.e., in real time) for any changes in ˜
Aydogan, M., and Uzgider, E. (1988). ‘‘A numerical method for the earth-
structural characteristics or loads. Complex problems, such as quake response of brick masonry structures.’’ Bull., Technical Univer-
finding the optimum design for the building, testing what-if sity of Istanbul, Istanbul, Turkey, 41(3), 415 – 431.
scenarios, or determining collapse mechanisms can all be in- Cadzow, A. J. (1973). Discrete-time systems. Prentice-Hall, Englewood
vestigated easily using such software. Cliffs, N.J.
The formulation presented above can also be used for non- Cai, G. Q., and Lin, Y. K. (1991a). ‘‘Localization of wave propagation
in disordered periodic structures.’’ AIAA J., 29(3), 450 – 456.
linear structures. Any nonlinear structure can be modeled as a Cai, G. Q., and Lin, Y. K. (1991b). ‘‘Wave propagation and scattering in
linear structure whose parameters change with time. If any of structural networks.’’ J. Engrg., Mech., ASCE, 117(7), 1555 – 1574.
the soil or building layers is nonlinear, the shear stiffness (and Claerbout, J. F. (1976). Fundamentals of geophysical data processing:
consequently the wave travel time) would change with fre- With applications to petroleum prospecting. McGraw-Hills, New York.
quency, amplitude, or time depending on the type of nonlin- Clough, R. W., and Penzien, J. (1975). Dynamics of structures. McGraw-
earity. The frequency-dependent nonlinearities can be con- Hill, New York.
Doebling, S. W., Farrar, C. R., Prime, M. B., and Shevitz, D. W. (1996).
verted into discrete-time filters, as we did for (7). Because the
‘‘Damage identification and health monitoring of structural and me-
equations are solved recursively in the time domain, we can chanical systems from changes in their vibration characteristics: A lit-
check the deformations and stresses in the layers at each time erature review.’’ Rep. No. LA-13070-MS, Los Alamos National Labo-
step to see if there is any nonlinearity, and if there is we can ratory, Los Alamos, N.M.
update the travel times and the reflection and transmission co- Doyle, J. F. (1989). Wave propagation in structures: An FFT-based spec-
efficients accordingly. tral analysis methodology. Springer, New York.
Ewing, W. M., Jardetzky, W. S., and Press, F. (1957). Elastic waves in
Finally, the methodology can also be used to investigate the
layered media. McGraw-Hill, New York.
flow of seismic energy in multistory buildings (Şafak 1998b). Hall, J., Heaton, T. H., Haling, M. W., and Wald, D. J. (1995). ‘‘Near-
We define the energy flow in terms of energy flux, which is source ground motion and its effects on flexible buildings.’’ Earth-
the amount of energy transmitted per unit time through the quake Spectra, 11(4), 569.
cross section of a layer. We select the state variables uj (t) and Iwan, W. D. (1997). ‘‘Drift spectrum: Measure of demand for earthquake
dj (t) in (1) – (3) to represent the upgoing and the downgoing ground motions.’’ J. Struct. Engrg., ASCE, 123(4), 397 – 404.
energy fluxes in the layers. Energy flux gives a dynamic mea- Kjartansson, E. (1979). ‘‘Constant Q-wave propagation and attenuation.’’
J. Geophys. Res., 84(B9), 4737 – 4748.
sure of energy demand by the earthquake at each story and Knopoff, L. (1964). ‘‘Q.’’ Rev. of Geophysics, 2, 625 – 660.
provides important new guidelines for design. Robinson, E. A., and Treitel, S. (1980). Geophysical signal analysis,
Prentice-Hall, Englewood Cliffs, N.J.
Şafak, E. (1995). ‘‘Discrete-time analysis of seismic site amplification.’’
CONCLUSIONS J. Engrg. Mech., ASCE, 121(7), 801 – 809.
Şafak, E. (1997). ‘‘Propagation of seismic waves in tall buildings.’’ Proc.,
The wave-propagation formulation has the potential, when 4th Conf. on Tall Build. in Seismic Regions, Los Angeles Tall Buildings
fully developed, to become a simpler and more powerful al- and Structural Design Council, Los Angeles, Calif., 139 – 153.
ternative to modal analysis to calculate the seismic response Şafak, E. (1998a). ‘‘New approach to identification and damage detection
in multi-story buildings subjected to seismic loads.’’ Proc., Struct.
of multistory buildings. This paper introduces the general con- Engrs. World Congr., ASCE, Reston, Va. (in CD Rom).
cept by considering very simple buildings and response modes. Şafak, E. (1998b). ‘‘Seismic energy flux in multi-story buildings.’’ Proc.,
For buildings founded on layered soil media and subjected to Struct. Engrs. World Congr., Reston, Va. (in CD Rom).
vertically propagating seismic shear waves, the buildings are Signal processing toolbox for MATLAB, version 4.0.1. (1997a). Math-
modeled as an extension of the soil medium, considering each works Inc., Natick, Mass.
story as another layer. The response is calculated in terms of SIMULINK, version 2.1 (1997b). Mathworks Inc., Natick, Mass.
Thomson, W. T. (1972). Theory of vibration with applications. Prentice-
the wave travel times at each layer and the wave reflection
Hall, Englewood Cliffs, N.J.
and transmission coefficients at layer interfaces. The wave- Todorovska, M. I., and Lee, V. W. (1989). ‘‘Seismic waves in buildings
propagation equations are obtained in the form of two simple with shear walls or central core.’’ J. Engrg. Mech., ASCE, 115(12),
finite-difference equations at each layer, one for the upgoing 2669 – 2686.

436 / JOURNAL OF STRUCTURAL ENGINEERING / APRIL 1999

J. Struct. Eng., 1999, 125(4): 426-437


Todorovska, M. I., Lee, V. W., and Trifunac, M. D. (1988). ‘‘Investigation the dynamic response of 2D frames subject to ground motions.’’ Proc.,
of earthquake response of long buildings.’’ Rep. No. 88-02, Dept. of 8th Eur. Conf. on Earthquake Engrg., Vol. 3, Laboratorio Nacional de
Civ. Engrg., University of Southern California, Calif. Engenharia Civil, Lisboa, Portugal, 6.2/17.
Todorovska, M. I., and Trifunac, M. D. (1989). ‘‘Antiplane earthquake Yong, Y., and Lin, Y. K. (1992). ‘‘Dynamic response analysis of truss-
waves in long structures.’’ J. Engrg. Mech., ASCE, 115(12), 2687 – 2708. type structural networks: A wave propagation approach.’’ J. Sound and
˜
Uzgider, E., and Aydogan, M. (1986). ‘‘Simple and efficient method for Vibration, 156(1), 27 – 35.
Downloaded from ascelibrary.org by Ryerson University on 01/05/19. Copyright ASCE. For personal use only; all rights reserved.

JOURNAL OF STRUCTURAL ENGINEERING / APRIL 1999 / 437

J. Struct. Eng., 1999, 125(4): 426-437

S-ar putea să vă placă și