M. Mucciarelli 1 , M.R. Gallipoli 1, 2 , A. Gosar 3 , B. Sket-Motnikar 3 , J. Roser 3 , P. Zupancic 3 ,S. Prevolnik 4 , M. Herak 4 , J. Stipcevic 4 , D. Herak 4 , Z. Milutinovic 5 , T. Olumceva 5 1 DiSGG, University oI Basilicata, Potenza, Italy 2 IMAA-CNR, Tito Scalo, Italy 3 EARSO, Ljubljana, Slovenia 4 Dept. oI Geophysics, University oI Zagreb, Croatia 5 IZIIS, University Cyril and Methodius, Skopje, Republic oI Macedonia !"#$%&'$ During a three year-period, the participants to a NATO Science Ior Peace project perIormed ambient noise measurements inside buildings in Iour European nations. This paper reports the results relevant to reinIorced concrete (RC) buildings with height in the range 1-20 Iloors. The total number oI buildings surveyed is 244. The most striking Ieature is the similarity oI the height-period relationships in the Iour countries, that allowed to treat the measurements as a single database. We Iound no signiIicant correlation with other variables, and calculated a regression that is very similar to other empirical height-period relationship and quite diIIerent Irom code provisions and theoretical models. ()$%*+,'$-*) IdentiIying the sites where a local ampliIication oI seismic shaking will occur and identiIying the buildings which will result the weakest under the seismic shaking is the only strategy that allows to eIIectively deIend people Irom earthquake damage at an aIIordable cost, by applying selective reinIorcement only to the constructions that need it. To mitigate seismic risk in urban areas, it is important to estimate dynamic parameters oI buildings and underlying soils in order to identiIy possible resonance phenomena. This resonance approach was used to study the damage pattern oI past earthquakes, see e.g. Mucciarelli and Monachesi (1999) Ior the 1998 Bovec earthquake, Panou et al.(2006) Ior the 1978 Thessaloniki earthquake, Navarro et al.(2007) Ior the 1993-1994 Adra earthquake. The search Ior resonace has been recently introduced in microzonation studies oI cities such as Granada (Spain, Navarro !" $%., 2004) or Senigallia (Italy, Mucciarelli and Tiberi, 2006). The dynamic behaviour oI a building in the elastic domain can also be useIul to calibrate the starting point oI non-linear analyses or to assess its vulnerability. Boutin and Hans (2008) showed that taking Draft of the paper published on Bulletin of Earthquake Engineering04/2012; 8(3):593-607 The original publication is available at www.springerlink.com into account the dynamic parameters that envisage the most adequate beam model, and taking the maximum tensile and compression strains oI concrete and steel as damage criteria, it is possible to quantiIy the levels that onset the structural damages and the plastic hinge, thus providing a useIul tool Ior vulnerability diagnosis. The best method to determine the dynamic parameters is oI course to record earthquakes inside permanently monitored building, but this is a costly option, deserved to a limited number oI case studies (see e.g.,; Celebi, 2000; TriIunac et al., 2001a; 2001b; Clinton et al., 2006, Herak and Herak, 2009). A possible alternative to permanent monitoring is to perIorm noise measurements and estimates the Iundamental period and damping with diIIerent techniques: Horizontal-to-Vertical Spectral Ratio (HVSR, Castro et al., 1998; Gallipoli et al., 2004; Di Giulio et al., 2005), Standard Spectral Ratio (SSR, Parolai et al., 2005), Non Parametric Damping Analysis (NonPaDAn, Mucciarelli and Gallipoli, 2007) and HalI Bandwidth method (HBW, Clough and Penzien, 1975). In this context, the main purposes oI the NATO Science Ior Peace project named 'Assessment of Seismic Site Amplification and Seismic Building Julnerabilitv in the FYR Macedonia, Croatia and Slovenia were: 1) to validate a procedure based on a small number oI ambient noise measurements with the aim to obtain at least the Iundamental Irequencies in horizontal orthogonal directions and the relevant damping with the minimum time and instrumental resources, comparing when possible the results with those derived Irom earthquake recordings; 2) to perIorm these measurements on a large set oI buildings; 3) to take ambient noise measurements on soil in order to look Ior possible cases oI building-soil resonances. The results relevant to the validation phase have been published by Gallipoli et al. (2008). For sake oI brevity we report here just their main conclusions: a) the comparison on a test building between ambient noise and weak-motion earthquake recordings shows a very small decrease in the Iundamental Irequencies, and an increase in damping that is not statistically signiIicant, and b) analysing the perIormances oI the Iour diIIerent techniques mentioned above on 80 typical Italian buildings, all the methods gave very similar results. HVSR and HBW in Iew instances were not able to return the two Iundamental horizontal Irequencies and the damping, respectively. On the other hand, SSR appears to be the more reliable method Ior assessing Irequency, whereas NonPaDAn gives better results Ior damping estimates, both returning physically sound values with a reasonable diIIerence on orthogonal components. The results concerning the presence oI soil-building resonance are described by Gosar et al. (2009) and Herak et al. (2009) while Albarello et al. (2009) report the experience gained on ambient noise measurements on soils Irom the methodological point oI view. This paper Iocuses on the statistical analysis oI the large data set acquired with measurements on buildings. !"#" %&"'()*) During the project the team acquired several hundred measurements on buildings. Here we describe the results relevant only to RC building with height in the range 1-20 Iloors. The measurement were perIormed in the Iour participating countries, so the analysed data base comprises 65 buildings Irom Italy, 47 Irom Slovenia, 62 Irom Croatia and 70 Irom Republic oI Macedonia. For all building we determined the Iundamental period on the two orthogonal components. For some buildings it was possible to extract some more inIormation, according to the number oI instruments installed. The data acquisition on this large data set brought diIIerent experience that were gathered in a manual Ior Iield crews. Appendix 1 reports the highlights oI the experience gained, while Appendix 2 is a table reporting all the data here discussed. Figure 1 reports the Iundamental period vs. the height Ior all the buildings here considered (Ior each building we selected the largest oI the two orthogonal components). 0 10 20 30 40 50 60 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 TA (65) SLO (47) HR (62) MK (70) Height (m) P e r i o d
( s ) Fig. 1 Height-period scatter plot for all the 244 RC buildings surveved Italian and Slovenian teams provided more data in the range 2-5 Iloors, while Croatia and Republic oI Macedonia teams covered more the taller buildings. The distribution is then Iairly sampled all over the height range considered. We then investigated the empirical relationship between the two variables, considering three typical Iormulation provided by other authors in diIIerent parts oI the world. !a! b |1| !ab! |2| !a! |3| The better Iit provided by |1| and|2| is marginal with respect to the need oI an extra parameter. The best statistical result is then provided by eq. |3|, in the Iorm !0.016" |4| We perIormed also a robust Iit with varying weight schemes Ior the statistical outliers, but the Iit proved to be very stable, returning a variation just on the Iourth digit. A boostrap procedure returned the 95 conIidence limits Ior the parameter a , equal to 0.015 and 0.0175 The next step was the investigation oI possible relationships among the residuals and other variables. Fig. 2 shows that the residuals are Iairly distributed according to a Gaussian distribution. - 0 . 5 - 0 . 4 - 0 . 3 - 0 . 2 - 0 . 1 0 0 . 1 0 . 2 0 . 3 0 . 4 0 . 5 0 0 . 1 0 . 2 0 . 3 0 . 4 0 . 5 0 . 6 0 . 7 0 . 8 0 . 9 1 D a t a C u m u l a t i v e
p r o b a b i l i t y
r e s i d u a l s d a t a n o r m a l Fig. 2 Distribution of the residual vs. a normal distribution with 0 mean Fig. 3a shows that there is no secondary correlation between residuals and height, while Iig. 3b reports the scatter oI residuals vs. the second mode, i.e. the period in the orthogonal direction. !"#$ &' ()*+,-./0',,*12)3, 3+ 1*4"56')4 74$ 8*"#8,9 +"# &: (1"#8,-. 40',,*1 3+ 1*4"56')4 74$ ,8* 4*03;5 <35* All these analyses conIirm the robustness oI the obtained regression, and also that the variance in the relationship it is almost explained by the single variable used. Our result is strikingly similar to that obtained by Navarro *, ')$ (2007) Ior Spanish buildings, that proposed the relationship !0.045"# with N equal to the number oI Iloors. We compared our relationship also with the provisions oI EC8 (Fig. 4), reports the comparison, The extension oI the data set conIirms the Iindings oI Gallipoli *, ')$ (2008) that concluded that ambient vibration data Ior RC buildings provide very similar in results in diIIerent countries, and conIirmed the mismatch with theoretical estimates and code provisions. We leave to another paper (Masi and Vona, 2008) the comparison with other relationships, including results obtained Irom theoretical modelling. Finally we considered the damping associated to the Iundamental period. ConIirming the Iindings Irom Gallipoli *, ')$ (2008), the standard assumption oI a 5 damping is Iully justiIied: Iig. 5a shows that the empirical distribution oI the observed damping is Iit by a log-normal distribution whose mean is 4.8 ($1.33, %0.7). The scatter plot oI damping vs. Iundamental period seems to provide an hint oI a inverse relationship, but the presence oI many outliers does not allow to retrieve a statistically signiIicant result. Further investigations are needed to explore iI there is a physical reason Ior the observed outliers. 0 1 0 2 0 3 0 4 0 5 0 6 0 - 0 . 5 - 0 . 4 - 0 . 3 - 0 . 2 - 0 . 1 0 0 . 1 0 . 2 0 . 3 0 . 4 0 . 5 H e i g h t ( m ) R e s i d u a l s
y = - 0 . 0 0 0 6 * x + 0 . 0 1 7 0 0 . 2 0 . 4 0 . 6 0 . 8 1 1 . 2 1 . 4 - 0 . 5 - 0 . 4 - 0 . 3 - 0 . 2 - 0 . 1 0 0 . 1 0 . 2 0 . 3 0 . 4 0 . 5 T 2 ( s ) R e s i d u a ls
y = - 0 . 0 0 9 3 * x - 0 . 0 2 5 Fig 4 Comparison among the Height/Period relationship here proposed and other functions Fig. 5a (left) Cumulative distribution of observed damping and fit of a log-normal distribuition. Fig. 5b (right) Scatter plot of damping vs. fundamental period 0 10 20 30 40 50 60 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 This paper, lower bound This paper,upper bound Navarro et al. (2007) EC8 Height (m) P e r i o d
( s ) 0 2 4 6 8 1 0 1 2 1 4 1 6 1 8 2 0 0 0 . 0 2 0 . 0 4 0 . 0 6 0 . 0 8 0 . 1 0 . 1 2 0 . 1 4 0 . 1 6 0 . 1 8 0 . 2 D a t a D e n s i t y
D a m p i n g ( % ) L o g n o r m a l 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 0.0 5.0 10.0 15.0 20.0 25.0 Period (s) D a m p i n g
( % ) !"#$%&'("#' We collected ambient noise recordings on 244 buildings in Italy, Slovenia, Croazia and Republic oI Macedonia. The height-period relationship we determined is robust and very similar to another relationship proposed Ior Southern European buildings by Navarro et al. (2007). ConIirming the results oI Gallipoli et al. (2008), who analysed a Iirst subset oI our data set, we Iound that our relationship is very diIIerent Irom European code provisions (EC8). This diIIerence cannot be ascribed to soil- structure interaction, because the Irequency determined by microtremors is representative oI the dynamic oI the whole system. Nor the low level oI shaking can be invoked: the ratio between our relationship and EC8 values is much higher than the period increase a building may suIIer due to damage up to incipient collapse. Finally, the distribution oI observed damping is well Iit by a log normal distribution. The average observed damping is 4.8 oI critical, justiIying the standard 5 assumption. However the dispersion oI the damping could make worthwhile in-situ measurements when an existing building have to be modelled or retroIitted. )$#"*%+,-+.+#/' This work was supported by grant NATO Science Ior Peace 980857 'Assessment of Seismic Site Amplification and Seismic Building Julnerabilitv in the FYR Macedonia, Croatia and Slovenia 0+1+2+#$+' Boutin C. and S. Hans (2008) How Far Ambient Noise Measurement May Help to Assess Building Vulnerability? in Increasing Seismic Safetv bv Combining Engineering Technologies and Seismological Data, M. Mucciarelli, M. Herak and J. Cassidy eds., Springer, ISBN 978-1-4020-9196-4, 151-180 Castro R.R., Mucciarelli M., Pacor F., Federici P. and A. Zaninetti (1998); Determination oI the characteristic Irequency oI two dams located in the region oI Calabria, Italy, Bull. Seism. Soc. Am., 88, 2, 503-511. elebi M. (2000) Seismic Instrumentation oI Buildings U.S. Geological Survey Open-File Report 00-157, last download at http://geopubs.wr.usgs.gov/open-Iile/oI00-157/oI00-157.pdI Clinton J.F., BradIord S.C., Heaton T.H. and J. Favela (2006); The Observed Wander oI the Natural Frequencies in a Structure, Bull. Seism. Soc. Am., 96, 237257. Clough R. W., J. Penzien (1975); Dynamics oI structures, II Ed., Mac Graw-Hill, New York. Gallipoli M.R., M. Mucciarelli and M. Vona (2008) Empirical estimate oI Iundamental Irequencies and damping Ior Italian buildings, Earthquake Engng Struct. Dyn. DOI: 10.1002/eqe.878 Herak M. and D., Herak, (2009) Continuous monitoring oI dynamic parameters oI the DGFSM- building (Zagreb, Croatia) , submitted to Bull. Earthq. Eng. Mucciarelli M., G. Monachesi (1999) The Bovec (Slovenia) earthquake, April 1998: a preliminary corrrelation among damage, ground motion ampliIication and building Irequencies, Journ. Earthq. Eng., 3, 317-327. Mucciarelli M., Tiberi PP. (2007) Microzonazione sismica di Senigallia. Tecnoprint Ancona,; 420 pp2 CD-Rom. Mucciarelli M. and M.R. Gallipoli (2007); Damping estimate Ior simple buildings through non- parametric analysis oI a single ambient vibration recording, Annals Geoph, 50, 259-266. Navarro M, Vidal F., Feriche M., Enomoto T., Sanchez F.J., Matsuda I. (2004) Expected groundRC building structures resonance phenomena in Granada city (southern Spain). Proceedings oI the 13th World ConIerence on Earthquake Engineering, Vancouver, BC, Canada, 16 August 2004. Paper No. 3308. Navarro M., F. Vidal, T. Enomoto, F. J. Alcala, A. Garcia-Jerez, F. J. Sanchez, N. Abeki (2007) Analysis oI the weightiness oI site eIIects on reinIorced concrete (RC) building seismic behaviour: The Adra town example (SE Spain); Earthquake Engineering & Structural Dynamics, 36, 1363-1383. Panou AA, Theodulidis N, Hatzidimitriou P, Stylianidis K, Papazachos CB (2005). Ambient noise horizontal-to-vertical spectral ratio in site eIIects estimation and correlation with seismic damage distribution in urban environment: the case oI the city oI Thessaloniki (northern Greece). Soil Dynamics and Earthquake Engineering; 25, 261274. Parolai S., Facke A., Richwalski S.M. and L. Stempniewski (2005); Assessing the Vibrational Frequencies oI the Holweide Hospital in the City oI Cologne (Germany) by Means oI Ambient Seismic Noise Analysis and FE modelling, Natural Hazard, 34, 217-230. TriIunac M.D., Ivanovic S.S., Todorovska M.I., Novikova E.I. and A.A. Gladkov (1999); Experimental evidence Ior Ilexibility oI building Ioundation supported by concrete Iriction piles, Soil Dynamics and Earthquake Engineering, 18, 169187. Appendix 1 : EXPERIENCE GAINED IN 3 YEARS OF MEASUREMENTS ON BUILDINGS 1. PRELIMINARY SURVEY OF THE BUILDING WITH THE HELP OF STANDARD FORM BeIore making any kind oI measure is necessary to retrieve some important characteristics oI the building (age, height, structural system, etc.). During the project, we prepared two standard survey Iorm, diIIering in the amount oI data requested and in the engineering skills required to the surveyors. The two Iorms are available on the project site, http://nato.gIz.hr. 2. HOW MANY MEASUREMENTS AND HOW TO TAKE THEM We did a diIIerent number oI measures according to the degree oI precision we wanted in the estimate oI the translational modes. It is necessary at least a single measure atop the building. II we want to reach a Iair knowledge oI the modes, a pair oI measurements at the lowest Iloor or on the Iree-Iield and on the higher Iloor, are suIIicient. The best solution is to do measurements at each Iloor and on Iree Iield. To estimate the rotational modes it is necessary to make two measurements, at the centre oI the ground Iloor and at the corner/centre oI the top Iloor. To estimate the modal shapes it is necessary at least one measurement at each Iloor in asynchronous way, while the best solution is to do one measurements at each Iloor, synchronous and with a base as a reIerence station. A single measurement atop the building is a Iair solution to estimate the damping oI the main translational modes. In reinIorced concrete buildings, the instrument should be placed near column-beam junction or near the centre oI stiIIness (to estimate the translational modes); do not put the instrument at the centre oI the rooms to avoid membrane modes oI the Iloors (aIIecting mainly the vertical component). II the stairwell is detached Irom the Irame oI the building, avoid to perIorm the measurement in it; be careIul to the non-structural element (Iloor tiling, Iloating Iloors, piping); be careIul to ground Iloor, it could be at a height diIIerent Irom the level oI Ioundations. Flexible Iloors in masonry building could be a problem with respect to reinIorced concrete ones. Be careIul about connection between structural elements, or you might measure local modes (e.g., out-oI-plane movements oI a detached wall). 3. CONDITIONS TO FULFIL DURING MEASUREMENTS BeIore to perIorm measurements it is important to level the instrument (be careIul oI possible tilting oI terrace Iloors), ensuring the proper coupling between the device and the point oI measurement, avoid to touch the semsor or to produce a strong noise near it. The usage oI the building is not a problem, except iI many people are walking nearby the instrument causing point loads on the Iloors. . II all the above conditions are IulIilled, 10 minutes oI ambient noise measurement at 128 Hz are a good compromise between accuracy and logistic needs. II not, longer measurements are advisable in order to possible remove Iaulty piece oI recording during the processing phase. Strong wind or traIIic noise could provide larger input and thus much resolved data processing. For modal shape estimation, variable weather conditions (e.g. strong wind gusts) require the use oI a reIerence station at the base oI the building in synchronous mode. !" $%&'()'&'*+ ,- *%./0'12%3 The Iollowing tables highlight the reliability oI techniques to estimates the dynamic characteristics. (Ior more details see Gallipoli et !"#, 2008). Generally the SSR and HVSR are the more appropriate techniques to estimate the main translational modes. II the building is small and very stiII, avoid HVSR, because only SSR can avoid Iake soil-building resonance due to the little mass. SSR is the best choice to estimate the rotational modes. For modal shape estimation synchronous measurements gave more reliable results than asynchronous ones. To estimate dampings, the NonPaDAn technique seems better because the damping estimates are less dispersed and grouped around values that are more realistic than those estimated with HalI-Band. HalI-band may also suIIer oI close translational and rotational modes.
Table 1: summary oI the reliability oI diIIerent techiques to analysed ambient noise data. !""#$%&' ) * +,# %-.- /#.