soil-structure and soil-structure-soil interaction

This article was downloaded by: [University of Grenoble] On: 06 June 2012, At: 09:28 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Journal of Earthquake Engineering Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/ueqe20

SOIL-STRUCTURE AND SOIL-STRUCTURE-SOIL INTERACTION: EXPERIMENTAL EVIDENCE AT THE VOLVI TEST SITE a

PHILIPPE GUÉGUEN & PIERRE-YVES BARD

a

a

Laboratoire de Géophysique Interne et Tectonophysique, LCPC, CNRS, Universite de Grenoble, Maison des Geosciences, BP53 38041, Grenoble Cedex 9, France Available online: 04 Sep 2008

To cite this article: PHILIPPE GUÉGUEN & PIERRE-YVES BARD (2005): SOIL-STRUCTURE AND SOIL-STRUCTURE-SOIL INTERACTION: EXPERIMENTAL EVIDENCE AT THE VOLVI TEST SITE, Journal of Earthquake Engineering, 9:5, 657-693 To link to this article: http://dx.doi.org/10.1080/13632460509350561

PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.

Journal of Earthquake Engineering, Vo1. 9,. No. 5 (2005) 657-693 @ Imperial College Press

SOILSTRUCTURE AND SOILSTRUCTURESOIL. INTERACTION: EXPERIMENTAL EVIDENCE AT THE VOLVI TEST SITE

Downloaded by [University of Grenoble] at 09:28 06 June 2012

PHILIPPE GUEGUEN and PIERREYVES BARD Labomtoire de Ghphysique Interne et Tecfonophystque LCPC, CNRS, Uniuersite' de Grenoble Maison des Geosciences, BPS3 38041 Gmnoble Cedex 9, h n c e Received 17 August 2004 Reviewed 3 November 2004 Accepted 19 January 2005

This paper shows the results of two passive experiments carried out a t the European Vohi test site where a scaled building has been constructed. The first experiment was performed to study the motion of the structure excited by two small earthquakes. For one month, six strong-motion recorders were installed within the structure, a t the top and a t the basement. The analysis of the deformation of the structure has been assessed by computing the spectral ratio between the top and the bottom, with a special focus on soilstructure interaction. An analytical model was then proposed to reproduce the structure and soil-structure system behaviour. The soil-structure interaction was accounted for by using impedance functions. During the second experiment, we concentrated our efforts on the effect of the building vibration on the surface ground motion. An explosive shot was fired and several strong-motion recorders were installed on the ground dose to the structure that allowed us to clearly identify a monochromatic wave coming from the building, in the time and frequency domains. This experiment allows us to dernbnstrate the non-negligible effect of the soil-structuresoil interaction that may disturb the surrounding ground motion. K e y w o d : Soil-structure; Volvi test site; interaction; seismic instrumentation.

1. Introduction Soil-structure interaction (SSI)effects have been recognised for a very long time. In 1954,Merrit and Housner (19541 found that significant effects could be expected for the rocking mode, in the case of structures resting on exceptional soft soil. Even though Housner [1957] showed using experimental data collected in California that SSI effects in the case of horizontal motion might be insignificant, it has been shown that, in some cases, SSI effects could largely affect the structure behaviour during earthquakes. For example, SSI effects affect the measured shift of fundament a1 frequency fi [Jennings and Bielak, 19'731 or the significant part of rocking motion in the total building motion [Bard, 1988; Bard et ol., 19921, for buildings resting on soft soil. By using experimental data collected at 51 California sites, 657

with a wide range of ground shaking levels, buildings and foundations characteris tics, Stewart et d. [1998,1999bl aiso showed that inertial interaction effects were found to be siD@cant. They consisted mainly of period lengthening and an increase in foundation damping, the inertial interaction eEects being evaluated horn variations between fixed- and flexible-base behaviour. They also confirmed that the eficiency of SSI effects was mainly determined by the soil-to-structure stiffness ratio. This phenomenon has received numerous analytical confirmat ions, e.g. Paolucci . '-. . [1993], Bard et al. [1996],Stewart et al. [1999a],for various types of supporting soils and structures (e.g. shear wave velocity, building mass and footing radius). Moreover, Todorovska and Trifunac [I9921presented an extensive analytical study, using a linear twedimensional model, in which the building model was an equivalent single degree-of-freedom (SDOF) oscillator. This study was devoted to analysing the impact of model features (e.g. building height and mass, and soil flexibility) on the system damping, system frequency and system response. They concluded that the flexibility of the soil, and also the building characteristics, strongly influence the soil-structure system behaviour , and these control the importance of SSI effects. One common characteristic of these numerical and experimental studies is that they limited their ranges of investigation to the structural behaviour, which was considered as being isolated. However, Luco and Contesse [I9731 and Wong and Trifunac [I9751 analytically showed that the building behaviour might be disrupted by the neighbouring structures. Wong and Trifunac 119751 showed that structuresoil-structure interaction is especially prominent when the structure of ink erest is smallerand lighter than its neighbours. Erlingson and Bodare [I9961and Erlingson [I9991confirmed that harmonic energy produced by human activity (e.g. jumping in a stadium during a rock concert) and applied at the surface of soft ground, could induce significant modifications to the motion of nearby structure. They assumed that the transfer of energy was favoured by resonance phenomena between the soil response and the frequency of the input "human" force. These observations tend to prove that the vibration of an oscillator, resting on the surface of the gro-und, can significantly contaminate the close free-field motion. by a radiative process. Jennings [1970], Kanamori et al. [1991],Gu6guen et al. [2000a] and Yun et d.[I9991also showed that wavefields induced by building vibrations and radiated back into the soil through the foundation could be recorded in the vicinity of the building. Jennings [I9701and Gukguen et ul. [2000a]showed that the radiated motion is still significant at two times and ten times the building characteristic length, with about 20% and 5% of the building base motion, respectively. However, all these experimental studies described structures forced into vibration by primary energy provided drectly to the structure (i.e. pull-out test or actuators), and not through the foundation. The radiated wavefield was then easily analysed, because it was not polluted by incident input wavefield. As discussed by Chavez-Garcia and Cardenas-Soto (20021,it becomes more difficult to detect the waves generated by the building vibration when this one is excited by ground input motion. Nevertheless, numerical studies for 2D SH model [Wirgin and Bard, 1996)or more realistic 3D models [Bard et a l , 1996; Gukguen et d., 20021 seem to

Downloaded by [University of Grenoble] at 09:28 06 June 2012

*

.

Downloaded by [University of Grenoble] at 09:28 06 June 2012

Soil-Structure and Soil-Stmctue-Soil Intenaction

659

confirm the significant effects of SSI on the b e e field motion, in the case of seismic input motion. Bard et al. [1996] and Gukguen et al. [ZOO21 analytically showed that, because of resonance, the trapping and scattering of waves radiated from the bases of buildings into the uppermost soft layer was favoured by surface layering, and significantly affects the fiee-field motion, and in particular its duration. Numerous seismic records provided from the Mexico City area, characterised by an uppermost very soft soil layer, also present very long time duration of motion relative to the stiff rock site records and show monochromatic beating. Even if numerous theories have been proposed to explain (with success) the high amplification of motions observed on the soft zone after the 1985 Michoacan earthquake, no satisfactory explanations have been developed to explain these particularities [Chavez-Garcia, I99 1; Chavez-Garcia and Bard, 19941. Nevertheless, Gukguen et al. (20021 recently showed that one explanation could be the presence of the very dense urban environment resting on very soft soil: During an earthquake, the city releases vibrating energy into the soil as seismic waves. Because of the similar frequencies of the soil and the majority of buildings, a global site-city interaction appears to induce the lengthening and the beating observed. Moreover, in some particular cases, structures produce modification of the freefield motion by scattering of incident seismic waves from their foundations, e.g. kinematic interact ion [Trifunac, 1972; Wong and Trifunac, 19751. Trifunac [I9721 assumed that the disturbance generated by waves scattering and diffracting phenomena at the rigid foundation is not a local phenomenon but could be extended to large distances, at least one order of magnitude greater than the size of the footing. The main goal of this paper is therefore to confirm and analyse the contamination of the fiee-field by buildings, subjected to ground shaking. Two field experiments were carried out in 1997 at the test site of Volvi [Euro-Seistest, 19951, on which a 1:3 scale RC-structure is located. The first experiment is devoted.to the analysis of the dynamic behaviour of the structure and of the effects of SSI. The building was temporarily instrumented with accelerometers installed a t the top and base of t h e structure. Two regional seismic events were recorded and the system response (hequencies and damping ratios) are studied. A dynamic model of the soil-foundation-structure system is proposed. For the simplified analysis of the inertial interaction, a widely used linear two-dimensional model is employed. A second experiment was carried out to record the radiation into the soil of the building's vibrational energy. The structure was subjected to two big shots, performed on the Volvi test site. The contarhination of the hee-field ground motion is analysed with the help of a temporary linear array, made up of four velocimeters installed in the vicinity of the structure. 2. Description of the Volvi Eur-Seistest

In order to investigate engineering seismology, earthquake engineering and seismology problems, the EureSeistest project has been carried out since 1993 in the

'

660

P. GueCguen€4 P.-Y. Bard

Sandy day

Gravel and sand 22.0

23.

24.

East

Sand or gravel

Downloaded by [University of Grenoble] at 09:28 06 June 2012

Quartenary undivide fans Mica gneiss

Mica and biotite granite

Fig. 1. ~ocdisationand geological map [after Koekel and Moltat, 19771 of the sedimentary basin of Volvi.

Volvi sedimentary basin, near Thessaloniki (Fig. I), in the scope of a European joint research project of several universities and institutions. The first aim of this project was to improve the knowledge of the regional seismicity as well as to better constrain seismological, geotechnical and structural models. Towards that goal, several field experiments were conducted to define the geological and geotechnical parameters of the basin (around 90krnz). Moreover, detailed geotechnical and geophysicai investigations were carried out on the test-site area on which a 1:3 scaled reinforced concrete structure was erected. Originally, this structure was subjected to various dynamic experiments to define its major dynamic features and the variations of its behaviour relative to its structural evolutions.

2.1. The Volvi basin at the teat site The Volvi basin is situated in the Mygdonian valley, an east-west graben divided by a ridge of the basement into the two basins of Langhadia and Volvi (Fig. 1). It is bounded towards the north and south by two mountain ranges and infilled with quaternary deposits, roughly horizontally plane stratified, but with abrupt discontinuities corresponding to active normal faulting. Geophysical investigations, including seismic and electrical prospecting were performed, as well as a geotechnical survey made up of drilling, sampling, water table measurements, standard penetration tests (SPT)and cone penetration tests (CPT),cross-hole and borehole measurements. These surveys gave a comprehensive and detailed description of the basin, in particular around and beneath the structure of the Test-Site area [Euro-Seistest, 1995; Jongmans et al., 19981.

soil-structure and Soil-Stnrcture'-Soil Interaction 661

Downloaded by [University of Grenoble] at 09:28 06 June 2012

Table 1. Soil profile at the Volvi Eur&eistest Eur~Seistest[I9951 and Jongmans et al. [1998]).

(after

At the test-site, the soil profile is highly stratified, with 13 horizontal layers (Table 1). As usual, the velocity increases from top to bottom, for the S-waves from 130 m/s in the surface 3 m to 1.8km/s at 240 m depth [Euro-Seistest, 19951. Since then, Jongmans et al. (1998) have proposed a new soil profile, which presents variations by comparison with the previous one. They assumed a basement located at around 15qm depth, with S-wave (P) and P-wave (a)velocities estimated in the deeper unweathered rocks as 2.5 km/s and higher than 4.0 km/s, respectively. The impedance contrast between sedimentary infill and bedrock becomes much larger. Nevertheless, the uppermost layers characteristics were unchanged, that will not produce variations in our study. Therefore, in the following study, we use the original soil profile. The quality factors for S- and P-waves (Qsand Q p respectively) were deduced from the analysis of the surface wave attenuation and vary from 10 (at the top) to 50 (at 250m) and from 30 to 100 for S- and P-waves, respectively [Jongmans et al., 19981. Determined by dense geophysical and geotechnical surveys, the very detailed knowledge of the soil dynamic properties provides the basis for a numerical approach. Moreover, on account of the horizontal plane layers, the location of the test-site in the Volvi Basin (Fig. 1)and its small size relative to the basin extension, the ground can be modelled by a 2D semi-infinite horizontally layered half-space. 2.2. The RC-building model

The RC-structure [Manos et aL, 19951 is founded on a (3.5 x 3.5 x 0.40m)surface square foundation (Fig. 2). The overall height h of the fivestorey structure is 5 m, i.e. the structure dimensions are scaled by 3. Each storey is composed of a slab of 80 mm thickness, supported by four columns of 110 x 110mm section and 0.85 m height. In the latest version, the building has masonry infill walls, built with (59 x 87 x 191mm) brick units. In order to simplify the analysis of the measured

662

P. Gw'guen 6 P.-Y. Bard

Cross Section

Downloaded by [University of Grenoble] at 09:28 06 June 2012

Plan

I

Foundation Slab

I

Fig. 2. Schematic view of the Eurc4eistest RGmodel built on the sedimentary basin of Volvi (after Manos et d. [1995]).

response during testing sequences, the structure was built as symmetrically as possible in plan and in elevation, keeping also the material properties as constant as possible. Two kinds of concrete were employed to ensure good concreting of the columns with the beams and to avoid excessive microcracking. During building sequences, three main kinds of experiments were conducted, using ambient vibration noise, weak ground motion from regional seismicity and a Pull-Out Test (POT).The latest consisted of forcing into vibration the structure through a prestressed steel cable, anchored between the building top and the close ground surface, which was suddenly released. The building was instrumented with accelerometers, judiciousiy installed to study its modes of vibration (longitudinal, transverse and torsion). The evolution of the structure behavior with building sequences is provided by Manos et al. [I9951 and Euro-Seistest [1995].The final version of the building is summarised in Table 2. The total weight of the structure ml accounts for added weight corresponding to concrete blocks put on each storey in order to avoid microcracking, which reduces considerably the stiffness of the structure, a n d also to satisfy the similitude law. Details of the reinforcement can be found in Manos et al. [I9951 and Euro-Seistest [1995],as well as the material properties (concrete and reinforcing steel) and the formwork and special features of the structure. Table 2. RC-building characteristics of Votvi model, with 2B and 2L the horizontal dimensions of the footing, ml and rno the mass of the equivalent building and of the footing, respectively and Jo the mass moment of inertia of the footing.

Soil-Structure and Soil-Sttuctue-Soil Intemction

663

3. First Experiment: Soil-Structure Interaction SSI

Downloaded by [University of Grenoble] at 09:28 06 June 2012

The analyses of the SSI effects on the building motion are usually separated into two consecutive steps: 6) Kinematic interaction refers to the diffraction effects on the incident seismic wavefield. Due to the inclination of incident waves, to the shape and the kind of the foundation (embedded or superficial), the incident waves are deviated and scattered from the foundation. Kinematic-interaction anaiysii usually consists in the definition of the foundation input motion (FIM),which represents the excitation transmitted to the structure (i.e. corresponding to the excitation in the inertial interaction). Because of the presence of a rigid body at ground surface, the FIM reflects the modification of the foundation motion relative to the free-field motion, as a result of the diffraction of seismic input motion by the foundation. For surface or embedded foundations subjected to vertical or oblique body and surface waves, the FIM may differ considerably from the free-field values for rotation and for translation. However, ignoring both rotational and translational kinematic components of the FIM usually leads to conservative results [Gazetas and My lonakis, 19981 and the routine practice for non-critical structures is therefore to neglect kinematic interaction effects. Because of the small size and the surface footing of the Volvi RC-structure, the kinematic interaction is neglected here. (ii) Inertial interaction refers to the inertia developed in the structure owing to its own vibrations. Building and soil-foundation systems are described using frequency-dependent impedance functions, made of stiffness and damping coefficients. Inertial effects result in a decrease of base shear force and rocking moment developed at the soil-structure contact [Bard et al., 1996; Stewart et al., 19981.In case of surface foundations, these strengths are mainly located beneath the foundation, where the contact between the soil and the footing is more efficient.

3.1. RC-building behaviour 3.1.1, Seismic instnrrnentation The first instrumentation (Fig. 3j was designed to provide detailed idormation about the seismic response of the Volvi test-site (TS) structure, using earthquake records. Six 3C-accelerometers (Guralp CMG5) were then installed at the corners of the structure, three at the building top (T)and three on the foundation slab (*). The sensors were installed along the longitudinal .L and transverse T edge of the footing, oriented accordingly with the three main directions of the structure-foundation system: The longitudinal L-component is oriented positive outward from the building (i.e. in the east direction), the transverse T-component is 90' counterclockwise position fIom the longitudinal (i.e. in the north direction) and the vertical Z-component is positive in the upward direction. Simultaneously,

'

664

P. Gw'guen €9 P.-Y. Bani Volvi RC-Stnrcture Instrumentation Configuration 1 Accelenzmeters (Top) Asdemmters (Bottom) Vebdmeter (F&eM))

x-axis Downloaded by [University of Grenoble] at 09:28 06 June 2012

-b

Fig. 3. First experiment: Overwiew of the building instrumentation devoted to the analysis of the SSI effects.

one 3C-velocimeter TS5 (Mark Products 2.0 Hz-L22) was installed in the vicinity of the structure (17.40mfrom the base of the building's edge), in the longitudinal and T S ~Accelerometers . were connected to direction, similarly oriented as three Reftek acquisition systems, the same i corner of the structure, at the roof (TS') and at the basement ( T S ~ ) being , connected to the same system, and the free-field velocimeter to one independent acquisition system. They were combined with external GPS for time reference.

TST

3.1.2. Records

Between 15th August and 15th September 1997 two local seismic events were recorded on the test-site, with a high enough signal/noise spectral ratio (i-e. above 3). They occurred on 19th and 25th August 1997. Because of their low magnitude, *they were not detected by the national networks of Greece, so we assume they had magnitudes less than 2. In the following, both events are named by the year and the day of the year, i.e. 97-231 and 97-237.Vertical, transverse and longitudinal accelerations recorded inside and outside the structure are shown Fig. 4. The records provide evidence for the amplification of the acceleration between the basement and the roof of the structure. The maximal acceleration recorded at the basement is around 0.2mm/s2 and 0.4mm/s2 for the 97-231 and 97-237events, whereas the roof acceleration is around 5 and 4 times higher, reaching 1.5mm/s2 and 1.8mm/s2.Vertical acceleration ZTs7 ire small relative to the horizontal acceleration, so we shall disregard its effects in the following. The top records show high frequency signals, which reflect the vibrating bequencies of t he structure. They are also characterised by monochromatic beatings, certainly resulting from coupling effects between vibrating modes. Whatever the. instrument location, accelerations recorded at the same storey (roof or base) and

Soil-Stmctuw and Soil-Structuwdoil Intmctaon 665

Downloaded by [University of Grenoble] at 09:28 06 June 2012

Longitudinal L

Transverse T

Time (s)

Tlme (s)

Longitudinal L I

.

I

Transverse T

I

0

0.5.

-5.4

10 20 Tlme (s)

Vertical Z 97-231

30

rs,

r

Longitudinal L

r

Transverse T

97-237 I

0

10 20 Tbne (s)

30

0

10

20

30

me (s)

Fig. 4. Acceleration time histories of the two events 97-291 and 97-237,in the vertical (left), longitudinal (centre) and transverse (right) directions, at the three corners of the building roof (top) and of the building base (middle). The close free-field acceleration time history is also shown (bottom).

in the same horizontal direction (L or T), show close similarities of wave shape, showing that the structure moves without strong internal horizontal deformations. Note also that for the same event, vertical components differ with respect of their position at the same level. This may result from structural effects. However, the

666

P. Gukguen & P.-Y. Bard

Downloaded by [University of Grenoble] at 09:28 06 June 2012

wave shape of the vertical motion seems to be linked to the horizontal wave shape, which shows the existence of "crosscoupling" horizontal-vertical mot ion. However, significant variations are observed between longitudinal and transverse components, which provide information about the seismic process and the azimuth of source but also may give information about non-symmetrical behaviour of the structure, as detailed in the following section. There are also no doubts that the total foundation acceleration (TS;) shows characteristics (wave shape and amplitude) roughly identical to the free-field acceleration (TS5),especially for horizontal components. This is in contrast to Wong and Trifunac (19751, who computed efficient scattering effects £ram long structures. The modification of the free-field motion owing to the presence of the small footing of the Volvi RC-structure seems to be insignificant, at least at first glance.

3.1.3. Motion of the stmcture in time domain

The contribution of each mode of deformation to the total building acceleration is defined in the time domain for transverse and longitudinal directions, considering both events denoted by their index k (k corresponds to 1and 2 for 97-231 and 97-237 events, respectively). The process for defining all modes is summarised in Fig. 5.

Torsion. Torsion at the top (T;) and at the base (T;) are derived from the difference between horizontal components recorded at each corner, for stations installed on the same edge of the system (i.e. ~ f =i (LTSF- LTSf)/21 and T& = (TTST- T * ~ ? ) /where P ~ , 21 < 2L and 26 < 2 8 are the distances between sensors in the T-and L-directions, respectively). The torsion at the basement (Fig. 6) is very insignificant relative to the horizontal and rocking acceleration (about 10 times less for both modes), a s previously observed for 1:1 scale buildings, hit by ground motion in Mexico City [Meli et ul., 19981. and T ; ~show variations, which result from the flexibility of the Volvi RC-structure. At the base level, and T& show roughly similar wave shapes, except for the 97-231 event. This could result from small inhomogeneities of soil beneath the foundation, producing variations of soil-footing system stiffness in parts of the foundation. This phenomenon was already experimentally observed for foundation systems consisting of friction piles [Trifunac et al., 19991 that produced an eccentricity in the centre of stiffness of the foundation system and then variations of rigidity, also expected in case of large footings. For the Volvi structure, the difference in torsion from one direction to the other can be expressed with respect to the average torsion measured at the base, as follows: 2. - T&)I ( T ' T&). This difference is about 4% and 3% of the average base torsion, for the 97-231 and 97-237events respectively. Therefore we assume no internal deformations. Moreover, due to the small size of the basement, the foundation will be considered ss rigid in the following. Rocking. As mentioned in Bard (19881, rocking acceleration Rjkis computed as being the difference between the vertical components of stations spreading

Tz'

TE

(TL

+

Sod-Structure and Soil-Structuz-Sod Int eractian 667

Fp:

Rocking R,:

Torsion

along L direction: (~,-)hl2b along T direction: (%+)hnl

Top: (Bl-8,)Qb and (D243/21 Base: (b,+@b and (%&)1a

e

and

Downloaded by [University of Grenoble] at 09:28 06 June 2012

T direction

Relative motion of the foundationH~

Structural drift Dli

along L direction: d l d , d 2 d and d,-d along T direction: bl-b, b2-b and b,-b

along L direction: Dldl-(c,+)h/2b and Dpd,-(c, -+)hl2b along T direction: B2-b2-(%%)h/2t and B3-b3-(c+~)h121

Fig. 5. Schematic view of the process used for obtaining the structural drift Dijk, the rocking acceleration (Rjk), the torsion (T; and T; and the horizontal acceleration of the foundation (ffijk)-

.

along the same foundation edge ( j = L or T), normalised by the distance between the two vertical sensors (i.e. 2b or 21) and multiplied by the height of the building h (e.g. rocking for the L-direction during the 97-237 event is computed as RLZ= h(ZTSP- ZTSf ) / 2 b . In reality, this formulation of the rocking represents the horizontal acceleration that the building would experience a t its top in case of rigid assumption. Moreover, this rocking corresponds to the total rocking of the basement, including the rocking of the soil, which is generally considered as negligible. Rocking of the foundation (Fig. 6) reveals variations between L- and T-directions, which may result from source and azimuth effects but also from asymmetrical characteristics of the structure. The maximum of the rocking corresponds t o about 0.2 mm/s2 (e.g. RL1 and RL2)with values reaching about 0.3 mm/s2 recorded during the 97-237event by the T-component ( R T Z )namely , around 25% of the total acceleration of the structure ( TS;).As reported in various papers, e.g. Bard [1988], Paolucci [1993],Meli et ol. [l998],buildings founded on soft soils usually exhibit significant rocking owing to soil-structure interactions, even for structures founded on piles.

668

P. Gue'guen E9 P.-Y. B a d a) 97-231 longitudinal L

Transverse T

Roof

Relative motion of the foundation Structural driff

Downloaded by [University of Grenoble] at 09:28 06 June 2012

Roof torsion

Base torsion

Base racking Fm e-field I

0

20

ftme (I)

10 20 Ttme (s)

tongitudinal L

Transverse T

TO

300

30

b) 97-237

Roof

Relative motion of the foundation

Structural dM

Roof torsion Base torsion Base mcking

Free-field

Fig. 6. Time histories of horizontal displacement at roof level, and comparison with the different components of deformation (see text for comprehension).

Soil-Stnrcture and Soil-Shrctun+Soil Int e m t i o n 669

Downloaded by [University of Grenoble] at 09:28 06 June 2012

(iii) Stmctuml d r i ' . The structural drift Dijk,which corresponds to the fixed-base structure behaviour, is computed by subtracting the total acceleration of the foundat ion (i.e. rocking plus horizont a1 acceleration including the input accel= LTST - LTSf - RLI eration) from the building top acceleration (e-g. D2L1 for the 97-231event). The maximum of the structural drift (Fig. 6) represents about 90% and 80% of the maximum total acceleration recorded at the top of the structure during the 97-231and 97-237events, respectively. Note that, . . . whichever set of stations considered in the same direction, the time history of the structural drift is quite similar. The coherence Crd between rocking and structural drift is plotted in Fig. 7. This one is computed as follows Max [I9851:

where S,, and S,, are the power spectral density of signal x and y, respectively and S,, is the cross spectral density of x and y. For the two events, high coherences are observed at the frequencies of the structure, which will be detailed in the following, between rocking and structural drift (more than 95%) in both directions, which shows strong SSI effects [Bard, 19881. ( i 4 Relative motion oJ the foundation. Horizontal foundation acceleration Hijk relative to the ground is described by subtracting the free-field motion from the total building base acceleration (e.g = LTSf - LTSSfor the 97-231 event). The Hijk's (Fig. 6) are computed relative to the TS5 free-field station ( Land T-direction). The acceleration corresponds to about 0.2 mm/s2, namely eight times less that the building drift. .However, considering no lateral variations of geology, the non-null Hijk 's values confirm the' suspected

vent 97-231

Event 97-237

-Ldirection

-Tdirection 5.-

3 Frequency (hz)

4 .

6.m

5 6 7 Frequency (hz)

a

Fig. 7. Coherence Crdbetween rocking and structural drift, computed in the L (thin line) and the T-directions (thick line) for the two events (continuous and dotted lines are the two sets of stations used to compute the structural drift).

670

P. Gdguen 8 P.-Y.Bard effect of the foundation, which may modify the &+field acceleration through scattering and/or radiating processes.

3.1.4. Motion of the structure in frepuency domain

Downloaded by [University of Grenoble] at 09:28 06 June 2012

The kequency f of the fundamental mode of the soil-structure system is usually obtained from the transfer function between the top of the building and the free-field motion [Paolucci, 1993; Meli et al., 19981.Nevertheless, more relevant t r a d e r ' functions FTijk are obtained by removing the fohdatioo input motion (FIM) effects from the building's top motion, where i refer to the set of stations, j to the direction and k to the event, as follows:

Based on this process, three transfer functions are available for each direction and each event (Fig. 8). For each of them, the fundamental frequency of the soilstructure system f corresponds to the frequency, which presents the maximal amplitude of FTijk.Three quasi-identical values of f are then plotted on Fig. 8. For the sake of simplicity, only their average value J'is displayed in Table 3 (+/- standard deviation).

3

Frequency (Hz)

10

3

Frequency (Hz)

10

Fig. '8. Transfer functions'(njk) computed at each corner of the structure, in the three directions, for the two events. Thmfer functions are displayed for the frequency range, in which the signa17tc+noiseratio is over 3.

Soil-Structure and Soil-Structunz-Soil Intemction 671 Table 3. Average fundamental frequency (Hz)of the soil-structure system j.(+I- standard deviation) computed kith the three sets of stations installed at each corner of the structure.

Downloaded by [University of Grenoble] at 09:28 06 June 2012

Events

.

L-direction

T-direction

Due to the swaying of the structure, the vertical niode observed in the structure results horn both horizontal motions, which introduce a vertical component. The very low peaks observed on the transfer function of vertical motion are probably due to the high radiation damping usually encountered for vertical motion. Moreover, the common assumption of neglecting the effect of the vertical and torsion modes for SSI studies is also confirmed because the vertical response of the structure is negligible compared to the horizontal modes. Therefore, the vertical modes will not be discussed in the following. A decay between the longitudinal and transverse frequency is noted (of about 10%) for the two events, confirming the asymmetry of the structure and/or the inhomogeneities of the SSI system. Note that the longitudinal and transverse j ' are very close (Table 3), which could lead to significant coupling effects between horizontal modes of the superstructure [Meli et aL, 1998; Trifunac et al., 1999; Gu6guen et al., 2000al. Beating observed at the building's top (TS?, in Fig. 4) could then be explained by this coupling. Between 97-231 and 97-237, a slight increase in apparent fundamental fiequencies is observed in the L-directions (about 2%). Nevertheless, this increase is within the standard deviation and, because of very low accelerations produced by these two events, no information about the nonlinear behaviour of the soil-structure system is available. However, the linear assumption was validated by previous studies conducted at the Volvi test-site [Euro-Seistest, 1995; Manos et al., 19951 which gave identical responses of the structure to small earthquakes, ambient noise or pull-out tests. Moreover, in the T-direction, the reverse effect is observed, with a frequency decay of about 0.5%. Damping ratios may be evaluated from spectral responses FTijk by the halfpower (band-width) method [Clough and Penzien, 19751. This method is summarisedinFig. 9. The damping ratio is determined from the frequencies f i and fi at which the response amplitude A is reduced to the level 1 / d times its peak value A,, (corresponding to the maximal amplitude of spectra). The damping ratio is then obtained by the following expression:

5

*


> H 2 k l ) , i.e. when SSI effects are negligible.

In conformity with the experimental results, spectral ratios are computed for the horizontal components (with input motion corresponding to x, and y,) of the two events (Fig. 12). Analysis in the transverse direction is done by substituting x and y indices in the spectral ratios formulation. The parameters set give results that are in good agreement with the observed data (Fig. 12). Following the first event, we observe significant variations of the structural dynamic parameters (e.g. f; from 6.94 to 7-02Hz and C: from 3.00 to 3.72% for the 97-231 and 97-237 events, respectively, in the longitudinal direction) and of the height H' (from 3.86 to 3.66m,which corresponds to 0.8 and 0.7 times the height of the structure). Even if the estimation of the dynamic structura1 parameters in the T-direction during the 97-237 event gives a worse agreement with the amplitude of spectral ratios (1) and (2), rather good fits are observed considering damping and frequency, and also for spectral ratios (3) and (4). Whatever the model, the SSI damping is probably too high (see ratio (2)) but cannot be reduced because it would make the agreement worse for spectral ratio (3). It is relevant to note that the shift of the fundamental frequency from f,' to as well as the decay of the damping ratio from :C to i*,are well modelled (see ratios (1) and (2)), that leads to significant SSI effects. They reduce the fundamental frequency by about 10% in both directions (excepted for the T-direction for which only a 2.5% reduction is observed) and the damping ratio by about 25%. Nevertheless, the rate of damping ratio decay from C; to for experimental as well as for numerical results, may also be dependent on the mass and height of the building Ci, as analytically shown by Todorovska and Trifunac [1992]. A rather good fit is observed for ratio (4) and confirms the predominant effect of the structural drift relative to the rocking motion, at the f hequency. But, at the soil-structure system frequency, spectral ratios (4) have values close to 0.3, which reflect significant SSI effects. However, the soil model used for the estimation of the impedance functions seems to systematically underestimate this ratio. This may be due to the choice of the equivalent soil: The uppermost layers are softer than the equivalent soil and they may increase the SSI effects.

p,

t*,

Sd-Structure and SoiGStructunz-Soil lntemction 681

Thus, a single 3-DOF model gives satisfactory estimation of the building response. Moreover, the impedance functions provided from the impedance b c tions handbook [Sieffert and Cevaer, 19921 reproduce the SSI effects. Note that the structural stiffness and damping are estimated from experimental data, but are slightly dependent on the mass ml which are provided h m Manos et al. (19951. Moreover, up to this point of the study, only inertial effect are considered and the incident seismic motion is considered as unaffected by the presence of the foundation (scattering of incident wavefield), which is usually assumed for surface footings. The motion considered as Yree-field'' motion (x,) is also assumed to be sufficiently far horn the building base as to be considered non-contaminated by the building motion.

'

Downloaded by [University of Grenoble] at 09:28 06 June 2012

4. Second Experiment: Soil-Structure-Soil Interaction SStSI

To evaluate the effects of the SSI on the fiee-field motion, one must consider: (i) The scattering of incident waves from the foundation, which can alter the freefield motion appreciably [Trihnac, 1972; Wong and Trifimac, 19751. Trihnac [I9721 showed that disturbances generated by waves scattered and diffracted in the vicinity of the-foundation are not a local phenomenon. They can extend to large distances (kt least one order of magnitude greater than the characteristic length of the foundation). (ii) The radiation of the wave energy, associated with the building response, via the foundation, into the half-space. Gueguen et al. [2000a]experimentally showed that the radiated wavefield is characterised by the soil-structure system frequency f,with damping proportional to They also showed that geometrical decay of radiated energy was directly dependent on the l / r and 1/fi rate of geometrical decay of body and surface waves.

t.

The main goal of the second experiment is then devoted to the analysis of SStSI effects, the building being forced into vibration by shots. 4.1. Experiment

On 2nd October 1997, two big shots wsre joictly Ered by the Laboratory of Soil Mechanics and Foundation and the Geophysical Laboratory. of the Aristotle University of Thessdoniki. The main goal was to investigate the geometry across the whole valley as well as the position of the buried faults that control the sediment thickness. They consisted of 35 kg and 65 kg of explosives placed at the sedimentaryrock interface, i.e. at about 40 m depth, and in the sedimentary layers, respectively. The first is located around 1km towards the north of the test-site (Fig. 13). Because of the proximity of the shot and of its power, the records saturated. Therefore, it could not be employed'in the following analysis. The second shot is located close to the Stivos site, around 2 km distance to the south-east of the test site area.

682

P. G&pm Ef P.- Y.Bard 368 m

Shot 1

Vohi Test-Site

A

MOO4 1OOm

-....._,......

u

Volvi TestSite t

Downloaded by [University of Grenoble] at 09:28 06 June 2012

0

>: :

StivrJs

&MOO5 ..*.'.......ROO1

A>