Professional Documents
Culture Documents
Abstract: Strength-reduction 共R兲 and displacement-modification 共C兲 factors have been extensively studied in the past for firm ground,
and even for soft soils considering site effects, but always excluding soil–structure interaction. In this work, they are investigated for a
single elastoplastic structure with flexible foundation excited by vertically propagating shear waves. The concepts developed earlier for
fixed-base yielding systems are extended to account for soil–structure interaction. This is done by use of the simplified reference model
and a nonlinear replacement oscillator recently proposed by the writers. Numerical evaluations are conducted for typical system configu-
rations, using the great 1985 Michoacan earthquake recorded at one site representative of the lakebed zone in Mexico City. Results are
compared with those corresponding to the fixed-base case. It is found that the site effects observed by other writers for the rigid-base
condition are increased or decreased by soil–structure interaction, depending on the period ratio of the structure and site. Finally, it is
shown how a site-dependent reduction rule proposed elsewhere for fixed-base systems should be adjusted for interacting systems using the
information presented.
DOI: 10.1061/(ASCE)0733-9445(2005)131:2(221)
CE Database subject headings: Displacement; Stiffness; Ground motion; Soil–structure interaction; Soft soils.
冑
by Avilés and Pérez-Rocha (2003) is used. This equivalent oscil-
Me
lator is defined by an effective ductility together with the effective Te = 2 共1兲
period and damping of the system for the elastic condition. These Ke
writers have demonstrated that modifying both the ductility factor
as well as the relevant natural period and damping ratio of the Ce
e = 共2兲
fixed-base structure is a reliable way of expressing the interaction 2冑KeM e
effects in nonlinear systems. By using this information, the site- where Ce and Ke = viscous damping and initial stiffness of the
dependent reduction rule developed by Ordaz and Pérez-Rocha structure when fixed at the base. The foundation is assumed per-
(1998) for fixed-base systems is suitably adjusted for interacting fectly bonded to the surrounding soil; it is defined by the radius r,
systems. This is a more rational way to assess in practice the yield depth of embedment D, mass M c, and mass moment of inertia Jc
resistance and maximum deformation of flexibly supported struc- about a horizontal centroidal axis at the base. The stratum of
tures. thickness Hs is characterized by the Poisson’s ratio s, mass den-
0
K̃hh共t − 兲 K̃hr共t − 兲
K̃hr共t − 兲 K̃rr共t − 兲
册再 冎
Uc共兲
⌽c共兲
d 共4兲
冥冦 冧
tive frequency of the system for the elastic condition, just as is
冤
Me Me M e共He + D兲 Üe共t兲 done in this investigation, or may be averaged over the frequency
Me Me + Mc M e共He + D兲 + M cE Üc共t兲 range of interest.
The computation of dynamic stiffnesses was achieved by use
M e共He + D兲 M e共He + D兲 + M cE M e共He + D兲2 + Jc ⌽̈c共t兲 of an efficient numerical technique based on the thin layer ele-
ment method (Tassoulas and Kausel 1983). In this technique, the
冦 冧
Pe共t兲
base of the stratum is taken fixed. This is not, however, a serious
+ Ps共t兲 restriction because it is always possible to choose a depth that is
M s共t兲 large enough to simulate the presence of an underlying half-space.
冦 冧 冦 冧
More specifically, the foundation was assumed to be embedded in
Me M e共He + D兲
the actual layer of thickness Hs which, in turn, is underlain by a
= − Üo共t兲 Me + Mc − ⌽̈o共t兲 M e共He + D兲 + M cE fictitious layer with rigid base of thickness 2Hs, representing the
M e共He + D兲 + M cE M e共He + D兲2 + Jc underlying half-space. Modeling the supporting soil in this way is
共3兲 justified because the computation of dynamic stiffnesses belongs
to a radiation problem rather than to a scattering one. In fact, it is
generally considered that the soil region that is affected by the
where an overdot denotes differentiation with respect to time t; force and moment acting on the foundation extends to about 4r
and E = D / 2 = depth to the center of gravity of the foundation below the foundation base for horizontal motion and to about 1.5r
which, for the sake of simplicity, is taken as one-half the depth of for rocking motion. Here, an embedment ratio of the foundation
embedment; this assumption is irrelevant for practical purposes. D / r = 0.5 and a normalized depth of the stratum Hs / r = 3 were
Also, Pe共t兲 = CeU̇e共t兲 + Ve共t兲 = internal force of the structure; and used for illustration. Also, the values of s / o = 0.8, s / o = 0.2,
Ve共t兲 = the restoring force. The interaction force Ps共t兲 and moment s = 0.05, o = 0.03, s = 0.45, and o = 0.33 were assigned to char-
M s共t兲 of the soil acting on the foundation are defined by the acterize the subsoil. The normalized springs and dampers so ob-
convolution integral tained are displayed in Fig. 2. The shear modulus of the stratum,
Gs, is used for normalization of these quantities. Given that the structure until the ratio Um / Uy of its maximum to yield deforma-
springs reflect both the stiffness and inertia of the soil, notice that tions is the same as the specified admissible ductility e. Evi-
they can take negative values. dently, the yield deformation is determined as Uy = Vy / Ke with Vy
Having determined the dynamic stiffnesses, the computation known. The iteration process is stopped when the difference be-
of input motions was accomplished by application of the averag- tween the computed and target ductilities is considered satisfac-
ing method of Iguchi (1984). With this simple but efficient tech- tory for engineering purposes. Due precautions are taken when
nique, the harmonic response of the foundation to vertically the ductility demand does not increase monotonically as the struc-
propagating shear waves is calculated by taking a weighted aver- tural strength decreases. In this case, there is more than one yield
age of the free-field displacements along the soil–foundation in- resistance that produces a ductility demand equal to the target
terface, and adding the displacement and rocking caused by the ductility, but only the largest one of them is of interest for earth-
resultant force and moment associated with the free-field tractions quake resistant design.
along this surface. The normalized, horizontal and rocking input
motions so obtained are exhibited in Fig. 3. Incidentally, they
prove to be independent of the layer depth and the bedrock flex-
ibility. The effects of these parameters are implicit in the free-field Nonlinear Replacement Oscillator
motion used for normalization.
If the transfer functions Qh共兲 = Uo共兲 / Ug共兲 and Qr共兲 Effective Period and Damping of System
= ⌽o共兲 / Ug共兲 are known, the time histories of the foundation
input motion for a particular earthquake are determined from a The common approach used in practice to take the elastic inter-
Fourier analysis as follows: (1) To compute the direct Fourier action effects into account has not changed over the years: A
replacement oscillator represented by the effective period and
transform, Ü*g共兲, of the horizontal free-field acceleration, Üg共t兲;
damping of the system. The most extensive efforts in this direc-
(2) to calculate the Fourier transforms of the horizontal and rock-
tion were made by Veletsos and his co-workers (see, for instance,
ing input accelerations as Ü*o共兲 = Qh共兲Ü*g共兲 and ⌽̈*o共兲 Veletsos and Meek 1974; Veletsos and Nair 1975; and Veletsos
= Qr共兲Ü*g共兲; and (3) to compute the time histories of the foun- 1977). Indeed, their studies form the basis of the interaction pro-
dation input motion, Üo共t兲 and ⌽̈o共t兲, by taking the inverse Fou- visions currently in use in U.S. codes (ATC 1984; FEMA 1997).
rier transforms of Ü*o共兲 and ⌽̈*o共兲. We shall call T̃e and ˜e the effective period and damping of the
To compute the step-by-step nonlinear response of structures system, respectively. They can be determined using an analogy
interacting with the soil, a time-integration scheme based on the between the interacting system excited by the foundation input
Newmark method was implemented. We obtained from Eq. (3) motion and a replacement oscillator excited by the free-field mo-
the structural strength that, for a particular earthquake, is required tion. The mass of this equivalent oscillator is taken to be equal to
to limit the maximum ductility demand to the ultimate ductility that of the actual structure. Under harmonic base excitation, it is
capacity. This is done by iteration on the yield resistance Vy of the imposed that the resonant period and peak response of the inter-
normalized, strength 共Vy / M eg兲 and displacement 共Um / Ug兲 spectra ues of D / r and Hs / r tend to reduce the effects of interaction. The
for constant ductility (e = 1, 2, and 4) and 5% of critical damp- remaining system parameters were fixed constant at typical values
ing, at the exclusion of soil–structure interaction. Here, g is the for building structures: M c / M e = 0.2, Jc / M e共He + D兲2 = 0.05, and
acceleration of gravity and Ug = 21 cm is the peak ground dis- M e / sr2He = 0.2. As Hs = 38 m, the value of Hs / r = 3 is valid for
placement. Also shown in this figure are the corresponding a foundation radius r ⬇ 12.5 m and, thereby, the values of D / r
strength-reduction 共R−⬁兲 and displacement-modification 共C−⬁兲 = 0.5 and He / r = 3 would correspond to a foundation depth D
factors. ⬇ 6.25 m and a structure height He ⬇ 37.5 m, respectively. If we
Acording to the one-dimensional wave propagation theory, the assume an interstory height of 3.5 m and the effective height as
predominant period of the site is given by 0.7 times the total height, the value of He ⬇ 37.5 m would corre-
spond to a building of approximately 15 stories.
4Hs The validity of Eqs. (9) and (11) is verified with the results
Ts = 共15兲
s shown in Fig. 6 for elastic 共e = 1兲 and inelastic (e = 2 and 4)
where Hs = 38 m and s = 76 m / s for the site considered. So, we response. It is clear that the strength and displacement spectra
have that Ts = 2 s, equal to the resonant period observed at the obtained for the interacting system are well predicted by using the
elastic acceleration spectrum. replacement oscillator. The discrepancies observed at short natu-
It can be seen from Fig. 5 that the values of R−⬁ for natural ral periods are not of practical significance, since midrise struc-
periods around the site period are substantially higher than e, the tures such as the one studied normally fall in the medium-period
value predicted by Veletsos and Newmark’s rule for the long- spectral region. For a 15-story building, Te may vary between,
period spectral region. This fact was observed for the first time by say, 1.1 and 1.8 s, depending on the structural system. Out of this
Meli and Avila (1988) analyzing recordings of the great Mexico region, the interest is just to show the trends observed in flexible-
City event, and after documented by Miranda (1993) and Ordaz base response spectra. As happens with fixed-base systems, the
and Pérez-Rocha (1998) analyzing hundreds of accelerograms spectral acceleration for a very short period as well as the spectral
from earthquakes of different origins and site conditions. In con- displacement for a very long period are independent of the value
trast, the values of C−⬁ are notably lower than unity in the period of e. It can be seen from Figs. 5 and 6 that the inelastic displace-
range at which R−⬁ ⬎ e, as inferred directly from Eq. (14). ments near the site period are smaller than the elastic ones, a fact
From these results, it is clear that the equal displacement rule may that is more pronounced for rigid- than for flexible-base struc-
be quite conservative for highly peaked narrow-band response tures.
spectra. Strength-reduction and displacement-modification factors
were computed using the rigorous results given in Fig. 6. The
shapes of R− and C− are compared in Figs. 7 and 8 with those
Results for the Flexible-Base Condition of R−⬁ and C−⬁, respectively. The differences between the re-
To show the influence of foundation flexibility on the R− and sults with and without interaction are noticeable, specially for
C− factors, only one representative interacting system was e = 4. It is apparent that structures on soft soil designed assuming
evaluated, which is characterized by: He / r = 3, for the slenderness rigid base may experience significant changes in their intended
ratio of the structure; D / r = 0.5, for the embedment ratio of the strength and displacement demands if soil–structure interaction
foundation; and Hs / r = 3, for the normalized depth of the stratum. plays an important role. Note that, as required by structural dy-
Naturally, deviations from these system parameters will affect namics, R− = 1 for Te = 0 and R− → e as Te → ⬁, irrespective
results. In general, decreasing values of He / r and increasing val- of the foundation flexibility. For other natural periods, there are
no theoretical indications regarding the values of this factor. Simi- Design Reduction Rule
larly, we have that C− = e when Te = 0 and C− → 1 as Te
→ ⬁, independent of the foundation flexibility. The limits of both As the difference between the shapes of R− and R−⬁ may be
factors for an infinite period have to do with the fact that long- significant, the reduction of elastic strength spectra to assess in-
period structures, whether elastic or plastic, undergo the same elastic strength spectra could not be attained accurately with ap-
maximum absolute deformation, equal to the peak ground dis- proximate rules deduced for fixed-base yielding systems. As seen
placement. before, there is no smooth correlation between the values of R−
It can be seen from Figs. 7 and 8 that the shapes of R− and and Te. In spite of this, it is worthwhile to devise a site-dependent
C− are very sensitive to variations in the structure period. In reduction rule that includes soil–structure interaction. This prob-
some period ranges, the values of these factors are larger for lem is addressed now using the solution for the nonlinear replace-
rigid- than for flexible-base structures, but in others are smaller. ment oscillator.
When R−⬁ ⬎ R−, it is confirmed that C−⬁ ⬍ C−; and vice The shape of R−⬁ has been extensively studied in the last
versa. This is consistent with Eq. (14). Despite this irregular be- years using recorded motions and theoretical considerations. In
havior, one can conclude that site effects, reflected in that R−⬁ particular, Ordaz and Pérez-Rocha (1998) observed that, for a
⬎ e and C−⬁ ⬍ 1 around the site period, are counteracted by wide variety of soft sites, it depends on the ratio between the
soil–structure interaction. This means that, at extreme interaction elastic displacement spectrum and the peak ground displacement,
conditions, we will have that R− = e and C− = 1 for medium in the following way:
冉 冊
and long natural periods. The reason for this is that, if the inter- ␣
Um共Te,e兲
action effects were so large, the structure period would shift to the R−⬁ = 1 + 共e − 1兲 共16兲
long-period spectral region, for which the equal displacement rule Ug
is applied. Results for other soft sites in Mexico City lead essen- where ␣ ⬇ 0.5. It is a simple matter to show that this expression
tially to the same conclusions. has correct limits for very short and long periods of vibration.
It should be noted that the R− and C− factors are to be Contrarily to what happens with available reduction rules, the
used in combination with flexible-base elastic spectra which, in values given by Eq. (16) can be larger than e, which indeed
turn, can be derived from rigid-base elastic spectra using the val- occurs if Um ⬎ Ug. In the conditions of Mexico City, this happens
ues of T̃e and ˜e previously defined. By this way, the yield resis- when Ts ⬎ 1 s. This reduction rule is more general than others
˜ e − 1兲 e
R− = 1 + 共
Te
冉 Ug
˜兲
T̃ Ũm共T̃e, e
冊 ␣
共17兲
It should be pointed out that Eq. (16) will yield the same result
as Eq. (17) if the elastic displacement spectrum without interac-
tion appearing in the former equation is replaced by the corre-
sponding spectrum with interaction. The two spectra Um共Te , e兲
and Ũm共T̃e ,˜e兲 are used to emphasize the fact that the former
corresponds to the actual structure, whereas the latter to the re-
placement oscillator. The steps involved in the application of Eq.
(17) can be summarized as follows.
1. By use of Eqs. (6)–(8), compute the modified period T̃e,
damping ˜e, and ductility ˜ e of the structure whose rigid-
base properties Te, e, and e are known.
2. From the prescribed site-specific response spectrum, deter-
mine the elastic spectral displacement Ũm corresponding to
T̃e and ˜e, just as if the structure were fixed at the base.
Fig. 8. Variations against period of displacement-modification
3. The value of R− is then estimated by application of Eq.
factors with (solid line) and without (dashed line) interaction for (17), provided the peak ground displacement Ug is known.
e = 2 and 4, considering the Secretaría de Comunicaciones y Results presented so far correspond to a structure with He / r
Transportes recording of the 1985 Michoacan earthquake in Mexico = 3. To show the validity of Eq. (17), two different values of this
City. Results correspond to an interacting system with He / r = 3, crucial parameter were chosen: He / r = 1 and 5, corresponding to
D / r = 0.5, and Hs / r = 3. short and tall structures, respectively. These values reasonably
cover the interest in structural engineering. Comparisons are
made in Fig. 9 between real strength-reduction factors and those
reported in literature, because the period and damping depen- obtained with the proposed reduction rule, for e = 2 and 4. It is
dence of R−⬁ is properly controlled by the actual shape of the seen that, although the representation is not perfect, the approxi-
elastic displacement spectrum, and not by a smoothed shape ob- mate rule satisfactorily reproduces the tendencies observed in re-
tained empirically. ality. In view of the many uncertainties involved in the definition
Following the replacement oscillator approach, this reduction of R−, it is judged that such an approximation is appropriate for
rule may be readily implemented for elastically supported struc- design purposes.
Fig. 9. Comparisons of real strength-reduction factors (solid line) with those obtained by the proposed reduction rule (dashed line) for e = 2 and
4, considering the Secretaría de Comunicaciones y Transportes recording of the 1985 Michoacan earthquake in Mexico City. Results correspond
to two interacting systems with He / r = 1 and 5, in both cases with D / r = 0.5 and Hs / r = 3.